ICES Journal of Marine Science: Journal du Conseil

Mangel,, Marc

doi: 10.1093/icesjms/fsw228pmid: N/A

I review my career in marine science chronologically forward from the time that I decided to become a scientist to the present. Among other themes, I illustrate how much of my career was the result of recognizing good opportunities rather than specific plans, the role that search problems have played in my career, and the power of mathematical methods to allow us to find commonalities in systems appears totally different. I discuss in detail my involvement in the International Court of Justice between Australia and Japan concerning special permit whaling in the Antarctic and conclude with my current activities—showing that surprises can happen at any point in a career. Introduction A retrospective such as this—regardless of one’s stage in life—forces one to understand patterns and processes that shaped a career. For example, I was struck by how much of my career was the result of recognizing good opportunities rather than specific plans. Although Stochastic Dynamic Programming (SDP—see below) teaches us that to understand life, we need to move backwards in time, I have written this essay going forward, from the event in middle school that caused me to become a scientist to my current activities. I have separated sections according to my education and early, middle, and late career. I discuss in detail my involvement in the case in the International Court of Justice between Australia and Japan in which the Court determined that Japanese special permit whaling in the Antarctic contravened the International Convention on the Regulation of Whaling. I close with a description of my current activities, which show that even late in our career we can be surprised by the directions research takes us if we are receptive to opportunities. Search problems have been a theme through my entire career—but emerging in very different ways in different fields of application. Similarly, the methods of applied mathematics—particularly simple mathematics used in mature ways—have been important, since they show us commonalities in questions where one might otherwise only see differences. In a short essay, many details—and people—have to be left out; I apologize in advance to the students, post-docs, and colleagues who helped shape my career for not mentioning them by name. New York, Chicago, Urbana, and Jerusalem, to 1974 One's ideas must be as broad as Nature if they are to interpret Nature—Sherlock Holmes in A Study in Scarlet. I was born in New York, but my family moved to Illinois when I was young and I grew up around Chicago. From an early age, I was interested in physics, chemistry (like many kids growing up in the 1950s and 1960s, I had a "lab" in our basement and did "experiments"), and biology (I loved growing and taking measurements on green beans and spent many hours walking along the shoreline of Lake Michigan). I was never “good at math" but recognized that mathematics was a terrific tool for science. Sunday evening was an important family television time in America then and shaped many careers (e.g. Springsteen, 2016). When I was 11 years old, an episode of the television show Bonanza involved a young Albert Michelson attempting to measure the speeds of sound and light (http://www.imdb.com/title/tt0529603). This struck me as a great life: thinking about how nature works, being outside, and using mathematics (in the case of the TV episode, geometry). Indeed, after the episode, I borrowed one of my father’s books on geometry and went to school the next day asking classmates if they had ever heard of the subject. I graduated from high school in 1968 and went to the University of Illinois as a physics major, but committed to learning chemistry and oceanography, hoping to one day find a career in the marine sciences. I did not know about mathematical biology. The years of university and graduate school are filled with self-discovery; the most important event in 1968–1969 was meeting the girl, Susan Milke, who agreed to become my wife and has been my support, critic, and teacher since then. In the fall semester 1969, I took an upper division course in Oceanography, taught by WW Hay, who is a pre-eminent paleo-oceanographer. His book Experimenting on a Small Planet (Hay, 2016) is a gem. I did not realize it then, but Bill was quite a radical—he taught us plate tectonics as if it were already well accepted, even though Takeuchi’s book had appeared only 2 years before (Takeuchi et al., 1967; Oreskes, 1999). For that course, we had to do a term paper and I focused on complex ions in seawater (complex ions had been one of my favourite topics in high school chemistry). With Bill’s support and tutelage, this became my first paper (Mangel, 1971). Like many first papers (Holmes, 1991), this one had minor significance for science but was essential for my development as a scientist: I saw how one could use simple mathematical methods and existing data to obtain new insights about nature. I kept in touch with Bill over the years and consider him to be one of my most important scientific mentors, in addition to the earliest. Most recently, he, Don Marszalek, and I have been discussing the functional role of the test in foraminifera, something that they worked on together many years ago (Marszalek et al., 1969; Marszalek, 1982). In fall 1969, I discovered the books of Nikolai Rashevsky (Rashevsky, 1969). These books are still useful (Mangel, 2015) and the phrase "mathematical biophysics" captured all the things that I was interested in. We had a department of physiology and biophysics at the University of Illinois, so I wandered over there hoping to do an independent study on membranes, which I had become interested in during the oceanography course studying phyto- and zooplankton. I often wonder how my career path would have differed had I found instead the classic on island biogeography by MacArthur and Wilson (1967), visited the Zoology department instead and thus arrived at mathematical biology much earlier than I did. I found Rashevsky’s books in a university bookstore, but might have easily just as encountered them in a library; there is great value in visiting libraries and bookstores and perusing titles and reading (or at least skimming) the entire issue of a journal—because we never know where the next good idea will come from. I completed MSc in Physiology and Biophysics working on thin lipid membranes (artificial reconstruction of cell membranes to allow better control in experiments) and in 1971, went to my first scientific meeting—that of the Biophysical Society in New Orleans. There, I met an Israeli scientist, Asher Ilani, who also worked on membranes at the Hebrew University in Jerusalem. The 1970s were the first nadir of the job market for PhDs and I saw many of my friends completing PhDs and not finding jobs, so decided that I would use my graduate education as an opportunity to travel and asked Asher about doing a PhD with him. We arrived in Jerusalem in July 1972 and I got right down to the work of learning Hebrew in the mornings and developing experiments in the afternoon. Three important things happened in the first 15-18 months there. First, I had a problem with my experimental system and could not get it to work for about a 6-week period. Second, I had already done some experiments that were qualitatively but not quantitatively reproducible, and wanting to do theory again, I made a visit to the Mathematics Library. I saw a book called Brownian Motion and Diffusion (Friedman, 1971), and thought that this would be perfect for modelling my experiments. The book begins: "(1)Definition. Normalized Brownian motion B is a stochastic process {B(t):0≤t<∞}on a probability triple (Ω, F, P)…". I realized that I was not likely to be able to teach myself the mathematics that I required. Third, in September 1973, I attended a meeting on warm lakes in Tiberias and this rekindled my interest in working in marine/limnological systems. Keeping to my desire of using graduate school for travel, I decided to leave the studies in Israel and pursue a PhD in Applied Mathematics and Statistics at the University of British Columbia. Vancouver 1974–1978 Life is complicated, but not uninteresting—Jerzy Neyman (ca. 1980) When I arrived in Vancouver in August 1974, I was interested fluctuations in systems with oscillatory dynamics. Donald Ludwig arrived at UBC at the same time, recruited to UBC from the Courant Institute of Mathematical Sciences at NYU. I approached Don about doing a PhD with him, working on fluctuations in systems with limit cycles. He said we could work together for a while on a problem that interested him and to see how we got along before he would take me as a student. This problem involved fluctuations in systems with a separatrix (Mangel and Ludwig, 1977; Mangel, 1994), motivated by Thomas Park’s magnificent study of competition between two species of flour beetles (Park, 1954)—something every undergraduate ecologist learns about today—and which had been worked on by every luminary in mathematical biology in the 1950s and 1960s (see the cites in Mangel and Ludwig, 1977; Mangel, 1994), including Jerzy Neyman (Neyman et al., 1956), one of the founders of frequentist statistics. At Don’s insistence, I took the course in Bayesian statistics that Jim Zidek offered; this became very important in my career. Working with Don was a fantastic experience and I continue to try to pay forward the support and goodness that he showed to me. (I am skipping the story of failing my first qualifying examination and a variety of other tribulations!) And, as often happens with things we encounter early in life, I continue to be enthused by flour beetles—and for many years had them in my laboratory. I ultimately wrote my paper on fluctuations in systems with multiple limit cycles (Mangel, 1980) and did work with Dan Gillespie (Gillespie and Mangel, 1981) on such systems before he became famous for his algorithms solving problems in stochastic chemical kinetics small volumes (e.g. Gillespie, 2007). Although Don had research money from Canada’s Natural Science and Engineering Research Council (NSERC) regulations then prohibited such funds from being used to support American students (presumably because of the large influx of Americans during the Viet Nam war). The early 1970s were a period of high inflation and what had seemed like a generous TA salary ($400 a month) when I applied to UBC simply did not go very far, especially since our first daughter was born shortly after I started graduate school and the second one was on the way in 1977. I discussed this problem with Don a few times and one day he told me that Colin Clark, who had written a very insightful paper (Clark, 1974) on schooling, had been approached by NOAA Fisheries to do some work on tuna-porpoise fisheries. I had taken a course in optimal control theory from Colin in my first term at UBC and was sufficiently desperate that I convinced him to hire me as a research assistant on the 1-year grant. I also petitioned the Graduate Dean at UBC to allow me to work both as a Graduate Research Assistant and as a Teaching Assistant, to thus keep my family housed, clothed, and fed. Working with Colin was another fantastic experience, and as with Don, I have tried to pay forward the debt that I owe to him from those days. We ended up with a very nice paper on the relationship between overall abundance of tuna and catch rate of tuna at dolphin-associated schools (Clark and Mangel, 1979); I wrote another paper on a stochastic version of those models (Mangel, 1982a). There were, of, course, difficulties: in the first 6 months of our work, we could not find the right kind of model and the stochastic version of the models was soundly rejected by The American Naturalist. In the course of this work, we learned that Jerzy Neyman—motivated by the collapse of the California sardine fishery—had worked on estimating the number of fish schools (Neyman, 1949) and that tuna purse seine vessels spent the majority of their time searching for schools of dolphin and much less time actually setting tuna. It was my introduction to search theory. I started graduate school in 1974, the year that Ray Hilborn and Carl Walters were at IIASA (see Hilborn, 2016), and when they returned Ray was already on a rapidly accelerating trajectory and I was struggling graduate student. So I knew of him, but really did not know him. In August 1977, Colin organized a special session on Natural Resource Management at a meeting of the American Mathematical Society in Seattle. At that meeting, I spoke about our tuna work and met John Beddington; this becomes important later in my story. The job market had not improved by the late 1970s. For example, a famous applied mathematician—Fritz Oberhettinger (well known to many physical oceanographers)—retired at Oregon State and the department received two jobs back for his one. They received 500 applicants per position. This does not sound unusual today, but it was then. I unsuccessfully sought a permanent faculty position, turned down one post-doc because the faculty mentor had no intention of letting me develop my own ideas (from which I decided that if I ever had post-docs, I would let them spend one day a week working on their own ideas, and have done so), and accepted another with Joel Keizer, a theoretical chemist, at UC Davis (at a salary of $12 000 a year—$11 000 as a post-doc and $1000 for teaching a one quarter course). In one of the bleaker moments of searching for jobs, I had also started applying for non-academic jobs, and one of those applications bore fruit: I was offered a permanent position in the Operations Evaluation Group (OEG) of the Center for Naval Analyses (CNA), at a salary of $24 000. Joel was very gracious in when I declined after having accepted the post-doc. I was anxious to move on and almost surely made mistake of not staying in Vancouver through my formal graduation in May 1978, but instead took the job in OEG, starting in November 1977. Washington, DC and Oak Harbor, Washington 1977–1980 If, instead of sending the observations of seamen to able mathematicians at land, the land would send mathematicians to sea, it would signify much more to the improvement of navigation and safety of men’s lives and estates upon that element—Isaac Newton The Operations Evaluation Group started in 1942 when the US Navy approached Phillip Morse at MIT about providing scientific help to deal with the German submarines and other forces operating off the US east coast (see Little, 2002; Tidman, 1984, and Morse, 1977 for a history of OEG). He recruited Bernard Koopman (an applied mathematician) and George Kimball (a theoretical chemist); together they created the field of operations research in the US while PMS Blackett was doing the same with operations analysis in the UK (Nye, 2004). My boss while I worked there, Phillip DePoy, recently gave an interview that describes both the history of OEG and his role at the time I was there (Sheldon, 2016). When I was hired at CNA, OEG was the main group that sent scientists to work at operational navy commands and had a tradition of staff spending one day a week on their disciplinary research, since operations research is a synthetic field. Most readers of this article indirectly know about OEG and CNA, because my colleague Christine Fox was the role model for Kelly McGillis’s character in Top Gun (http://www.imdb.com/title/tt0092099/). When I met Christine, she was an analyst in DC, before going to the fighter wing in San Diego; she later became Director of OEG and President of CNA. Subsequently, she served as Acting Deputy Secretary of Defense and in that role Acting Secretary of Defense when the Secretary was out of the country, becoming the first woman ever to do so. As of this writing, Christine is Assistant Director for Policy and Analysis of the Johns Hopkins University Applied Physics Laboratory. During WWII, Koopman developed the theory of search (e.g. Koopman, 1956a, b, 1957) and in the late 1970s, as turning his papers into a magnificent book (Koopman, 1980). While I was waiting for my security clearance to come through (which took a while because I had lived abroad), Phil DePoy suggested that I work through a draft of the book. The final chapter left open the question of how to deal with moving targets and I realized that the methods I had used in my PhD thesis could apply to search problems. For example, if the target is moving along a course characterized by mean vector (b1(x,y),b2(x,y)) subject to random fluctuations independent of location that can be characterized as a diffusion process with variance σ2 ⁠, and if ψ(x,y,S)dt is approximately the probability that the target is detected in the next little bit of time dt when it is at (x, y) and the search plan is S, then the probability f(x,y,t,S) that the target is at location (x, y) at time t and still not detected satisfies ∂f∂t=σ22[∂2f∂x2+2∂2f∂x∂y+∂2f∂y2]−∂∂x(b1(x,y)f)−∂∂y(b2(x,y)f)−ψ(x,y,S)f(1) I knew how to obtain approximate solutions of such equations σ2 was small compared with the deterministic component of motion (e.g. Mangel, 1981a) and so once more became involved in search problems. Thus, I continued to work on search problems while in OEG. When I went to the field at Whidbey Island Naval Air Station, Oak Harbor, WA, I developed a method for locating a radio transmitter when there are biases in the angle measured by the receiver and the true angle (Mangel, 1981b). That paper received the Koopman Paper Prize from the Operations Research Society of America in 1982. I describe some of my other work while in OEG in Mangel (1982b). My time in OEG reinforced the power that comes from using simple mathematics in mature and sophisticated ways. Because of this interest in search theory, Phil also suggested that I attend a NATO Advanced Research Workshop (ARW) on search theory, held in Faro, Portugal, in March 1979 (Haley and Stone, 1980). ARWs were wonderful meetings, bringing together colleagues from around the world to spend a week away from distractions discussing common intellectual themes. Ray Hilborn was also there and, as he explains (Hilborn, 2016), we connected again but this time more as equals. Because Ray had a house on Whidbey Island, we began regular contact. Since he has told the story of the The Ecological Detective (Hilborn and Mangel, 1997) and its authorship in his essay in this series, I will not repeat it here, save to mention that had I known how successful the book would be (https://dynamicecology.wordpress.com/2016/02/17/what-is-your-all-time-favorite-ecology-book/), I would have traded authorship order and bought the banjo myself. In an analysis of the Princeton Monographs in Population Biology—of which The Ecological Detective is a part—Elworthy (2007) notes that there are things which cannot be done effectively in a series of articles but can be done in a book. Introducing modern statistical methods to biologists, as Ray and I did, was one of them and making SDP accessible to biologists, as Colin Clark and I did (see below) is another. However, writing a book requires much more time (think of it as 10–12 papers none of which appears until the final one is done) and the full impact may not be seen for many years. Although not required to do so, while in OEG, I kept finding ways to teach—I offered a course on search theory at CNA and taught introductory math for Embry Riddell Aeronautical University while in Oak Harbor. I now understand that I have a personality that requires me to teach and when my students and post-docs are looking for jobs, I ask them "Do you feel that you have to teach formal classes, rather than don’t mind teaching formal classes?" If not, there are plenty of other wonderfully satisfying jobs outside of academia. The Department of Mathematics at UC Davis advertised three positions in applied mathematics in spring 1979. Joel Keizer encouraged me to apply and if offered one to delay coming until 1980 so I could finish my commitment to OEG at Whidbey Island. This is exactly what happened. In September 1980, we moved to Davis, CA, where I planned to do OEG-like work in fisheries and agriculture. By OEG-like work, I mean bringing the scientific approach to problems for which the fundamental laws governing the processes are either unknown or too complex to derive from first principles, or for which we know the fundamental laws but do not know the value of the parameters associated with them—thus merging applied mathematics and statistics. One of the last things that I did for CNA was to write for publication the technical reports of Abraham Wald on aircraft survivability (Mangel and Samaniego, 1984), which was the JASA Applications Invited Paper in 1983. Wald’s thinking still intrigues people (http://www.ams.org/samplings/feature-column/fc-2016-06#mangel) and Jordan Ellenberg begins his book (Ellenberg, 2014) with Wald’s work. UC Davis, 1980–1996 All the business of war, and indeed all the business of life, is to find out what you don't know by what you do; that's what I called “guessing what was on the other side of the hill”—Sir Arthur Wellesley, The Duke of Wellington Since this is a journal of marine science, I will only spend a few sentences on my work in agriculture, making three points. First, the ideas of search had natural application in agricultural pest control, and very nearly from the instant of my arrival at Davis I was applying them (e.g. Mangel et al., 1984; Stefanou et al., 1986). Second, problems involving insects led to two of my longest and most pleasurable collaborations—with Bernie Roitberg at Simon Fraser University and Jay Rosenheim at UCD. Third, mathematical methods allow us to see commonalities in situations in which others only see differences. For example, the inestimable Paul Smith enthused about the use of the Negative Binomial Distribution (Mangel, 2006) for describing the patchiness of fish eggs (Smith, 1973, 1978) and in the early 1980s, we began working together using it (Mangel, 1987; Mangel and Smith, 1990). However, exactly the same methods can be used to describe the aggregation of insect eggs or larva (e.g. Southwood and Henderson, 2009). Indeed, my most recent paper with Bernie Roitberg is “species-independent” and deals with the effects of relaxation of the thermal performance curve on individual growth (Roitberg and Mangel, 2016). Paul Smith also convinced me to study the diversity and longevity of the rockfish Sebastes, which lead to a rich collaboration with Mike Bonsall, first at Imperial College and then at Oxford (e.g. Bonsall and Mangel, 2004; Mangel and Bonsall, 2004, Mangel et al., 2007; Bonsall and Mangel, 2009). Once I started at UCD, my goal was to bring the ideas of search theory—which is fundamentally a Bayesian construct since we are seeking the location of a target conditioned on unsuccessful search—into fisheries (and agriculture) in a modern way. This lead to one of the first applications of Bayesian analysis in stock assessment (Mangel and Beder, 1985). In summer 1981, Colin Clark invited me back to UBC to collaborate with him and we began our work on informational problems in fisheries and natural resource management (Mangel and Clark, 1983; Mangel, 1985; Mangel and Plant 1985; Mangel and Clark, 1986a, b). While working on those informational problems in fisheries, we realized that the behaviour of fishing vessels is a special case of the general problem that almost all foraging organisms face: trading off gaining information about a foraging situation with actual gains in that situation. This work lead us to more general cases in behavioural ecology (Clark and Mangel, 1984, 1986); Clark and Mangel (1986) will shortly be published by Edward Elgar in a collection Biological Economics edited by Andrew Lo and Ruixin Zhang at MIT. According to Elgar: "Our book is designed to bring together the most important and influential material in the subject area as facsimile reprints, to supplement the research resources of newly founded libraries around the world." In the early 1980s, one of the problems in behavioural ecology was the "crisis of the common currency": behavioural ecologists recognized that animals faced the problem of avoiding predation while acquiring food to avoid starvation. The units of predation risk are 1/time and the units of food acquisition are energy/time, so that the risks were non-commensurate. One afternoon in fall 1984, the day before I was planning to visit Colin at UBC, I left my office and walked towards the student union for a coffee. About halfway there, I realized that if we simply thought about overall survival—combining starvation and predation risk—we could use approaches similar to those in search theory, with SDP for optimizing behaviour, to resolve the crisis of the common currency. My recollection is that Colin had a stomach virus during my visit over the next few days, so it was a bit of time until he saw the power of this method—but not long after that we were on the path towards a unified theory of foraging (Mangel and Clark, 1986a, b). As it happened, Alisdair Houston and John McNamara at Oxford had virtually the same idea at the same time (McNamara and Houston, 1986). Over the last 30 years, we have jointly (Houston et al., 1988), pairwise (Mangel and Clark, 1988; Houston and McNamara, 1999; Clark and Mangel, 2000), and separately (Mangel and Ludwig, 1992) worked to make this method a standard tool for life scientists. I provide a bit more history and some simple examples of how state dependent life history theory, implemented by SDP, is developed and applied in Mangel (2015). In spring 1985, Bob Fridley—who ran the aquaculture and fisheries program at UCD—approached me about being involved in a project to develop genetic methods for identifying natal streams of returning chinook salmon. The project was funded by the fishing industry, which had a management zone closed on the basis of the then current method, only to have a very strong run of salmon return to the closed zone. Bob matched Graeme Gall, a top-notch geneticist in the Department of Animal Science, and me as the PIs. I recruited graduate students John Brodziak and Richard Gomukiewicz to work with me and we were very successful in using allozyme methods to determine stream of origin (Gomulkiewicz et al., 1990; Bartley et al., 1992; Brodziak et al., 1992). That project began 30 years of me working on Oncorhynchus species. In spring 1986, I received an unexpected letter from John Beddington on behalf of the Scientific Committee of the Commission for the Conservation of Antarctic Marine Living Resources (SC-CCAMLR). The CCAMLR Convention had been in force for just a few years and the Scientific Committee was wrestling with how to use fishery dependent data to estimate abundance of southern ocean krill, Euphausia superba. The difficulty was that vessels spent much of their time searching for swarms of krill and only a small amount of time harvesting, so that standard proxies for abundance such as Catch Per Unit Effort were not useful. John asked if I would be interested in serving as one of the first two invited experts to SC-CCAMLR; the other would be Doug Butterworth from the University of Capetown who would model the Japanese krill fleet while I would model the Russian fleet (Mangel, 1989, 1990). I attended the meeting of SC-CCAMLR in 1988 as an invited expert, and after my contract with CCAMLR ended continued to work on krill (e.g. Mangel and Switzer, 1998) so attended meetings of the krill working group and the SC-CCAMLR meeting in 1991, where I met George Watters, who now directs the Antarctic Ecosystem Research Division of NOAA Fisheries and is a longtime collaborator. After the SC-CCAMLR meeting in 1988, I was asked to join the Committee of Scientific Advisors of the US Marine Mammal Commission, on which I served from 1990 to 1996. Among other things for the MMC, I lead the development of principles for the conservation of wild living resources (Mangel, 1996). In retrospect, the causal chain was tuna models to search theory to krill fishery to the MMC, a clear set of links that can only be seen looking backwards. I had a sabbatical in 1987–1988, spending November–January at the Hebrew University in Jerusalem and February–August in Oxford with support from the Guggenheim Foundation for the entire period (to work on the development of unified foraging theory) and the Fulbright Foundation for the time in the UK. This was an important year in many ways in addition to the regeneration that a sabbatical provides. Before leaving Davis, I had decided that on my return I would try to move to a biological sciences department—I was ready to be surrounded again by biologists and have coffee with mathematical scientists rather than the reverse. UCD was a big enough campus, and one with a tradition of interdisciplinary work, to allow this to be feasible. As it happened, Robert May left the Department of Biology at Princeton that year to move to Oxford as Royal Society Professor. I applied for the job and was offered it; the retention offer from Davis included moving my appointment to the Department of Zoology (which later became the Department of Evolution and Ecology) and formation of the Center for Population Biology (http://www.cpb.ucdavis.edu) of which I was the founding director. Founding the Center for Population Biology is my greatest administrative accomplishment. Princeton ultimately hired Simon Levin for that position; it is still a perfect match. In Oxford, I was hosted jointly by the Department of Zoology and Centre for Mathematical Biology. Jim Murray, director of the Centre, circulated a list of talks that visitors would be happy to give and I was invited by Felicity Hungtingford to visit Glasgow and Pitlochry. In Glasgow, I learned of the work of Felicity, Neil Metcalfe, and John Thorpe on how size going into the summer and feeding over the summer affected whether a juvenile Atlantic salmon would smolt the next spring or not (Metcalfe et al., 1989). They had developed a statistical predictor based on the size and the growth rate and I realized that this was a perfect situation for applying state dependent life history theory, which we did—although it took us nearly a decade of work to get agreement on all the details (Bull et al., 1996; Mangel, 1996; Thorpe et al., 1998); already a senior Professor, I had the luxury of waiting until the right intellectual structure emerged, rather than rushing to print sooner. In the next section, I discuss application of these ideas to steelhead trout in California. In 1990, I was invited to give a plenary lecture at the meeting of the International Society for Behavioural Ecology (ISBE) in Lund, Sweden. After that, Alasdair Houston and I co-taught a Doctor of Science course organized by Jan Ekman (who unexpected and sadly passed away in August 2016) at the field station of Stockholm University. Such courses were intended for graduate students and post-docs from all Scandinavian countries. There were wild rose bushes on the property, so in addition to lecturing the students on the work that I had done on insect oviposition, we could do field trips! At this course, I meet Jarl Giske, who had just completed his PhD at the University of Bergen. I told Jarl that one day he would have a sabbatical and he should come and spend it with me. And indeed he did—twice (2000–2001; 2010–2011)—as well as sending students and post-docs from Bergen to visit me in California and having me regularly visit Bergen. Since 2010, I have been Adjunct Professor in the Theoretical Ecology Group in Bergen and I continue to collaborate with colleagues there (Eliassen et al., 2009; Giske et al., 2013, 2014; Jorgensen et al., 2016). Jarl also organized two Doctor of Science courses in Bergen that I co-taught with Tony Pitcher (1994) and Colin Clark and Paul Hart (1996), so I have had a 20+ year association with Bergen. Throughout these years, fisheries management had been a theme of my research and I believed that to successfully train students to work in fisheries management, they had to learn social science as well as natural science so that my students took courses such as resource economics as electives. In 1994, a new PhD program Environmental Studies began at UC Santa Cruz and a senior position in that department became available when Michael Soule retired. I moved to UCSC after spending 16 years at Davis, where I could have spent the rest of my career there, but looked forward to new challenges and opportunities. UC Santa Cruz, 1996– The purpose of models is not to fit the data but to sharpen the questions—Sam Karlin (1983) Susan and I moved to Santa Cruz in June 1996, so I have just completed 20 years at UCSC. In August 1996, I travelled to Bergen for a meeting of a SC-CCAMLR working group, a week holiday, and then taught the second Doctor of Science course. I had been working on krill for about a decade and went to the working group intending that this would be my swansong with krill. However, while there I learned that Reuben Lasker—one of my scientific heroes—had worked on the north Pacific congener Euphausia pacifica (Lasker and Theilacker, 1965; Lasker, 1966). I decided to return to Santa Cruz and start to work on individual variation in North Pacific krill (Marinovic and Mangel, 1999) and to continue working on southern ocean krill, with a focus on going from life histories to fisheries. The former project was supported by my start-up funds and the latter by NSF. I also agreed to organize the Second International Symposium on Krill (Mangel and Nicol, 2000). When I could not find a graduate student in the new PhD program in Environmental Studies interested in krill, I used the salary to hire Suzanne Alonzo as a post-doc. We wrote a series of papers (Alonzo and Mangel, 2001; Alonzo et al., 2003a, b) using individual behaviour to predict the effect of krill fisheries on penguin foraging success. When that NSF grant ended, I thought that I was done with working on krill, but this was not to be, as I explain below. Suzanne left Santa Cruz to a faculty position at Yale, but happily for me she returned to UCSC in 2014 as Professor in Ecology and Evolutionary Biology. We continue to interact and I serve on the PhD thesis committee of one of her students who is working on rockfish. Around 2000, Bill Fox, then Senior Scientist at NOAA Fisheries recognized that a very large number of quantitatively trained NOAA scientists would be retiring in the next decade and that NOAA Fisheries needed to be proactive about filling the gap. The Fisheries Ecology Division (FED) had recently moved from Tiburon to Santa Cruz and Churchill Grimes (Director of the lab), Alec MacCall, and I developed a training program between the FED and UCSC. Our idea was to use the funds that NOAA provided as a foundation for training students and post-docs in the quantitative population biology needed to sustain fisheries, thereby increasing the pool of quantitatively trained people who could be hired by NOAA Fisheries. We called our program the Center for Stock Assessment Research (CSTAR) and sought students and post-docs who were committed to learning the relevant quantitative population biology and interacting with NMFS colleagues. For example, CSTAR students and post-docs generally participated in a stock assessment. Our focus was to train them so that they could develop the new quantitative methods that we will require in the 21st century. We attracted UCSC students from programs in Anthropology, Applied Mathematics and Statistics, Ecology and Evolutionary Biology, Environmental Studies, and Ocean Sciences and post-docs from a similar wide range of PhD programs. The discretionary funding for CSTAR was never considerable—alone it could not even support a graduate student and post-doc—but it provided a base that allowed us to send students to meetings or to work with colleagues, to have a robust seminar series, and provide summer support and some research support. This discretionary funding was supplemented by funding that NMFS colleagues or I obtained, or funding that the students themselves obtained. Of the last seven CSTAR students, five had NSF Graduate Research Fellowships and one had a NMFS-Sea Grant Population Dynamics Fellowship; of the last four CSTAR post-docs, two had NSF Postdoctoral Fellowships and one a Marie Curie International Outgoing Fellowship. The first two post-docs appointed to CSTAR were Melissa Snover (whom I met on a visit to the Duke Marine Lab in 2001) and Steve Munch (with whom I had worked when he was a graduate student, Munch et al., 2003) and they set a very high bar for the research by CSTAR students and post-docs for both breadth and depth of investigation and publication outlet (e.g. Munch et al., 2005a, b, Snover et al., 2005, 2006). After CSTAR, Melissa joined NOAA Fisheries Protected Species Division in Hawaii; she is now at the US Fish and Wildlife Service in Corvallis. After CSTAR, Steve joined the faculty at Stonybrook University, where he received tenure and promotion but even so we were able to lure him back to the FED in 2010 and he and I continue to collaborate (e.g. Salinas et al., 2012, Boettiger et al., 2015). CSTAR also had a musical component: Alec MacCall, Steve Ralston, Sarah Newkirk (Steve Munch’s partner), and I played music together from about 2002-2006. One night before music started, Alec asked me if the Fokker–Planck equation might be used to compute a prior for steepness (the fraction of unfished recruitment when the spawning biomass is 20% of its unfished size). I thought about and decided no, but that the Kolmogorov backward equation (e.g. Mangel, 2006) could be. This lead to a series of papers about steepness and its interpretation (He et al., 2006; Enberg et al., 2010; Mangel et al., 2010a, 2013). CSTAR was its biggest from about 2006 to 2011, in part because I was able to bring three very big grants to support the discretionary funding. Susan Sogard and I received a CalFed grant and embarked on a multipronged project concerning steelhead Oncorhynchus mykiss that involved field work, laboratory work, and modelling, with the goal of applying the ideas that I had developed with Atlantic salmon [Mangel and Satterthwaite 2008; Mangel and Satterthwaite 2008, Satterthwaite et al., 2009, 2010, 2012a (selected as best publication for 2009 in The Transactions of the American Fisheries Society); Beakes et al., 2010; Sogard et al., 2012]. Will Satterthwaite continues to work on salmon, and followed up our genetic stock identification work on chinook salmon with a paper (Satterthwaite et al., 2014) that won the Stevan R. Phelps award for best genetics paper in an American Fisheries Society Journal (2015). Margaret Bowman, the program manager for the then newly formed Lenfest Ocean Program, convinced me to work on southern ocean krill once more (I tried to convince her to fund work on longevity and diversity in the Sebastes), to think about krill and their predators, krill fisheries, and climate change. Thus, although I had thought I was done with krill when Suzanne and I finished our work, krill came back into my life and I have given up trying to not work on krill (my current PhD student Ryan Driscoll is working on krill in winter). For this project, I advertised widely and assembled a large and productive team (Wiedenmann et al., 2008; Cresswell et al., 2009; Wiedenmann et al., 2009; Mangel et al., 2010b; Wiedenmann et al., 2011; Cresswell et al., 2012; Shelton et al., 2013; Richerson et al., 2015, 2016), with the goal of doing fundamental work that could feed into the Krill Predator Fisheries Model (KPFM) developed by George Watters and his colleagues at US AMLR (Watters et al., 2013). Finally, I was part of the Bering Sea Integrated Ecosystem Research Program (BSIERP), for which the Secretary of Commerce awarded the Department’s Gold Medal in 2015. The objective of BSIERP was to go from physical forcing through zooplankton and fish to top predators; our role was to predict how a changing environment would influence fur seals and kittiwakes (e.g. Satterthwaite et al., 2010; Satterthwaite and Mangel, 2012; Satterthwaite et al., 2012b; Vincenzi and Mangel, 2013; Vincenzi et al., 2013,; Vincenzi and Mangel, 2014). Each of these grants represented an opportunity; our choice is to seize the opportunity or not. The whaling case in the International Court of Justice In August 2010, I was unexpectedly contacted by the office of the Attorney General of Australia, asking if I would serve as the Independent Scientific Expert witness in a case in the International Court of Justice (ICJ) concerning the Japanese Whale Research Programs Under Special Permit in the Antarctica (JARPA II). This was the second phase of a very controversial program of lethal take that had begun in 1986, just as the commercial moratorium on whaling went into effect. The ICJ is the principal judicial organ of the United Nations and all member states of the UN are automatically members of the Court’s statute. The Court’s role is to settle legal disputes submitted to them by States, so most scientists (including me before 2010) do not know about it. In general, the International Criminal Court (ICC) is in the news; the ICJ has only been so recently because of the dispute in the South China Sea. I was asked by the government of Australia to develop criteria for a program for purposes of scientific research in the context of the conservation and management of whales in the southern ocean, and to assess JARPA II against those criteria. In order to assure my independence as a witness, I was not informed of Australia’s legal strategy and only saw it unfold during oral proceedings. Thus, my task involved considering science as a process, and communicating these ideas along with my assessment of JARPA II to the Judges. My full analysis is included as Appendix 2 to the Australian written submission (http://www.icj-cij.org/docket/files/148/17382.pdf). My determination was that a program for purposes of scientific research in the context of conservation and management of whales required (i) a conceptual framework leading to testable predictions, which is almost a definition of modern science; (ii) a process for setting sample sizes of lethal take based on solid statistical reasoning and analyses of the accuracy required to meet objectives; (iii) regular peer-review of research proposals and results; and (iv) design to avoid adverse effects on the stocks being studied. Assessing JARPA II against these criteria, I concluded that "JARPA II is an activity that collects data in the southern ocean. However, it is not a program for purposes of scientific research". Ultimately, using different reasoning, the Court concluded that JARPA II contravened the International Convention for the Regulation of Whaling and ordered it halted; Japan complied and 2014–2015 was the first time in 100 years that whales were not taken in the Southern ocean. Excellent analyses of the case in general can be found in Fitzmaurice (2016) and Fitzmaurice and Tamada (2016). Analysis of the case from the perspective of scientists can be found in de la Mare et al (2016), Mangel (2016), and Press (2016). It was an amazing and nearly unique experience (the Court had not had an invited expert witnesses for more than 20 years and never really had a scientist). Although I have long had an interest in policy (e.g. service on the Committee of Scientific Advisers of the MMC), and written on policy for scientists (Ludwig et al., 2001), this was my first real foray into international environmental law. The team of lawyers from Australia, lead by William Campbell, and all of the external counsel were stellar colleagues. I worked particularly closely with Philippe Sands, an environmental and human rights lawyer, whose books on environmental law (Sands et al., 2012) and human rights (Sands, 2016) are amazing, and James Crawford, who was elected a Judge of the Court by the UN in fall 2014; his book of Hague lectures (Crawford, 2014) is titled "Chance, Order, Change"—something that every good Darwinian relates to! At one point, I asked the lead lawyer for Australia how they chose me and he said that they asked a variety of people and my name kept coming up; after looking at some of my talks posted on the web they decided I would be an effective witness. Once again, I could draw a causal chain from tuna in graduate school to whales in the ICJ, with each link representing a new opportunity but for which I had to make an active choice. For example, to do the work that the whaling case demanded, I put aside a project that I had planned to do on modelling metabolism and will probably never return to it. Into ‘retirement’ By July 2013, I had served the University of California for more than 33 years and decided that was time to step away from the constraints of formal teaching and professional and university service to concentrate on my research, much of which involved travel. I found it difficult to travel when teaching formal courses, especially to undergraduates, because of my sense of responsibility to the students. I was 62 years old at the time and followed my mentors who had retired at ages 61 (Don Ludwig) and 63 (Colin Clark). I consider that I had then finished the first two-thirds of my career (1971–2013). The University of California reserves the title Research Professor for emeriti who have returned to active service to continue doing research. A one-month break is required between retirement and return to active service. Thus, on 1 August 2013, I returned as Research Professor. As Research Professor, I am allowed to supplement my pension with research grants. When I retired, I expected that in 2016 I would have support to work on linking life history theory to stock assessments and on aspects of Ecosystem Based Fishery Management (Mangel and Dowling, 2016). I could not find funding to investigate those ideas, but I have had continued good fortune of being supported by NSF. I hold an OPUS (Opportunities for Understanding through Synthesis) grant that supports writing a book on ectotherms in changing environments, in which I will show how state dependent life history theory interweaves with other methods for studying how organisms respond to environmental change. I also hold a NSF-NERC grant with Holly Kindsvater (Rutgers) and Jason Matthiopoulos (Glasgow) in which we are developing methods to use life history information to replace missing data for data poor species, motivated by tunas and groupers. This collaboration arose as a result of me telling Holly that she should meet Jason when she went to the International Marine Conservation Congress in Glasgow. As usual, the NSF grants are ones that I had to apply for (and the OPUS more than once), but unexpectedly I was invited to collaborate with Lisa Schwarz, Elizbeth McHuron and Dan Costa at UCSC to use state dependent life history theory to predict the population consequences of disturbance on marine mammals. We have already developed a general framework paper (McHuron et al., 2016) and are in the process of applying the methods to California sea lions, blue whales in the California Current Ecosystem and the western Pacific population of gray whales. Colin once described his retirement as a 10-year long sabbatical, and this is feeling true. I continue an active research program, my appointment at the University of Bergen, and very limited service to the profession (I am a member of the Scientific Review Board of the International Pacific Halibut Commission; http://www.iphc.info/srb). What’s been left out I have not discussed teaching in much detail, but I taught at all levels—from introductory courses to specialty graduate courses; I considered my teaching successful if I could profoundly affect the life of one student each term and almost always did so. Undergraduates in introductory or intermediate courses became undergraduate thesis students, but I generally encouraged them to go elsewhere for graduate studies in order to obtain a different worldview. Some have come back to work with me as post-docs. I still love to teach, and in some sense still have to: in winter 2015, I gave a series of lectures at the FED on quantitative fisheries science. It could not be called a course because I had not been returned to active service to teach. I have not discussed editorial work nor university administration, but did too much of that and should have taught more. As mentioned above, I consider the Center for Population Biology at UCD my greatest administrative achievement. Conclusion Jon Schnute and Laura Richards visited CSTAR to give seminars, and Jon told me about Donald Stokes’s book Pasteur’s Quadrant (Stokes, 1997). Stokes argued that Pasteur was always motivated by an important applied problem (like Edison, but not Bohr) and sought fundamental understanding (like Bohr, but not Edison). I thus had the great fortune of a career in Pasteur’s Quadrant. That career began in a shaky way—not being able to find a permanent academic position but the three years in OEG turned out to be instrumental in the entire rest of my career. Search theory continues to develop (Stone et al., 2016) and search algorithms are a bridge between organisms, evolution, and ecology (Nolting et al., 2015; Barbier and Watson, 2016; Hein et al., 2016). Because I kept publishing when outside of academia, I was able to return to it. The University of California has been a wonderful and almost always supportive place to work. Thus, virtually from the outset, my career was guided by a love of science and a recognition of serendipitous opportunities and seizing them, rather than a careful plan. I constructed the title for this essay from a GLOBEC essay (Mangel, 1993) and the title of Chapter 3 in The Ecological Detective, which together summarize my advice to young scientists embarking on a career. First, truly understand the system in which you are working—make knowledge of your system so deep that it is part of your gut intuition. Second, when applying quantitative methods, be certain that you are choosing the ones that match your system, rather than using something that comes off the electronic or physical shelf. Also make these methods part of your gut intuition. And if a method is lacking for what you need to do, create it because that is the way science advances. Acknowledgements I thank Howard Browman for inviting me to write this essay. Over a long career, I have had support from NOAA Fisheries, NSF, Sea Grant, and USDA and the Lenfest Ocean program; I thank them all. Similarly, I thank the members of my group not mentioned or cited in here for helping create the rich intellectual life that I have enjoyed in UC for almost 40 years. Although I have not published a paper with these colleagues, I thank them for friendship and support over the years: Nancy Reid (since 1974), Simon Levin (since 1977, and an able squash partner for nearly 20 years), John Gillespie and Michael Turelli (at Davis), Joe Travis (since 1996), and John Thompson (at Santa Cruz). I thank Susan Milke Mangel, my partner in this adventure of ideas, an anonymous referee and Howard Browman for comments on a previous version of the manuscript. References Alonzo S. H. , Mangel M. 2001 . Survival strategies and growth of krill: avoiding predators in space and time . Marine Ecology Progress Series , 209 : 203 – 217 . Google Scholar Crossref Search ADS WorldCat Alonzo S. H. , Switzer P. V., Mangel M. 2003a . An ecosystem-based approach to management: using individual behaviour to predict indirect effects of Antarctic krill fisheries on penguin foraging . Journal of Applied Ecology , 40 : 692 – 697 . Google Scholar Crossref Search ADS WorldCat Alonzo S. H. , Switzer P. V., Mangel M. 2003b . Ecological games in space and time: the distribution and abundance of Antarctic krill and penguins . Ecology , 84 : 1598 – 1607 . Google Scholar Crossref Search ADS WorldCat Barbier M. , Watson J. R. 2016 . The spatial dynamics of predators and the benefits and costs of sharing information . PLoS Computational Biology , 12 : e1005147. Google Scholar Crossref Search ADS PubMed WorldCat Bartley D. , Gall G. A. E., Bentley B., Brodziak J. K. T., Gomulkiewicz R., Mangel M. 1992 . Geographical variation in the population genetic structure of Chinook salmon from California and Oregon . Fishery Bulletin (US) , 90 : 77 – 100 . OpenURL Placeholder Text WorldCat Beakes M. P. , Satterthwaite W. H., Collins E., Swank D. R., Merz J. E., Titus R. G., Sogard S. M. et al. 2010 . Smolt transformation in two California steelhead populations: effects of temporal variability in growth . Transactions of the American Fisheries Society , 139 : 1263 – 1275 . Google Scholar Crossref Search ADS WorldCat Boettiger C. , Mangel M., Munch S. 2015 . Avoiding tipping points in fisheries management through Gaussian process dynamic programming . Proceedings of the Royal Society B , 282 : 20141631. Google Scholar Crossref Search ADS PubMed WorldCat Bonsall M. B. , Mangel M. 2004 . Life-history trade-offs and ecological dynamics in the evolution of longevity . Proceedings of the Royal Society of London B , 271 : 1143 – 1150 . Google Scholar Crossref Search ADS WorldCat Bonsall M. B. , Mangel M. 2009 . Density dependence, lifespan and the evolutionary dynamics of longevity . Theoretical Population Biology , 75 : 46 – 55 . Google Scholar Crossref Search ADS PubMed WorldCat Brodziak J. K. T. , Bartley D., Bentley B., Gall G. A. E., Gomulkiewicz R., Mangel M. 1992 . Tests of genetic stock identification using Coded-Wire Tagged fish . Canadian Journal of Fisheries and Aquatic Sciences , 49 : 1507 – 1517 . Google Scholar Crossref Search ADS WorldCat Bull C. D. , Metcalfe N. B., Mangel M. 1996 . Seasonal matching of foraging to anticipated energy requirements in anorexic juvenile salmon . Proceedings of the Royal Society B , 263 : 13 – 18 . Google Scholar Crossref Search ADS WorldCat Clark C. W. 1974 . Possible effects of schooling on the dynamics of exploited fish populations . Journal Du Conseil International Pour L’Exploration De La Mer , 36 : 7 – 14 . Google Scholar Crossref Search ADS WorldCat Clark C. W. , Mangel M. 1979 . Aggregation and fishery dynamics: a theoretical study of schooling and the purse seine tuna fisheries . Fishery Bulletin , 77 : 317 – 337 . OpenURL Placeholder Text WorldCat Clark C. W. , Mangel M. 1984 . Foraging and flocking strategies: information in an uncertain environment . American Naturalist , 123 : 626 – 647 . Google Scholar Crossref Search ADS WorldCat Clark C. W. , Mangel M. 1986 . The evolutionary advantages of group foraging . Theoretical Population Biology , 30 : 45 – 75 . Google Scholar Crossref Search ADS WorldCat Clark C. W. , Mangel M. 2000 . Dynamic State Variable Models in Ecology. Methods and Applications . Cambridge University Press , NY . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Crawford J. 2014 . Chance, Order, Change: The Course of Environmental Law . Hague Academy of International Law , The Hague, The Netherlands . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Cresswell K. A. , Tarling G. A., Thorpe S. E., Burrows M. T., Wiedenmann J. R., Mangel M. 2009 . Diel vertical migration of Antarctic krill (Euphausia superba) is flexible during advection across the Scotia Sea . Journal of Plankton Research , 31 : 1265 . Google Scholar Crossref Search ADS WorldCat Cresswell K. A. , Wiedenmann J. R., Mangel M. 2012 . A model of parental conflict: predicting provisioning behavior of penguin partners in response to local changes in krill . Ecological Modeling , 246 : 68 – 78 . Google Scholar Crossref Search ADS WorldCat de la Mare W. , Gales N., Mangel M. 2016 . Applying scientific principles in international law on whaling . Science , 345 : 1125 – 1126 . Google Scholar Crossref Search ADS WorldCat Eliassen S. , Jorgensen C., Mangel M., Giske J. 2009 . Quantifying the adaptive value of learning in foraging behavior . American Naturalist , 174 : 478 – 489 . Google Scholar Crossref Search ADS PubMed WorldCat Ellenberg J. 2014 . How Not to Be Wrong . Penguin , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Elworthy S. 2007 . The little yellow books that changed the world: a very short history of the Monographs in Population Biology . Bulletin of the British Ecological Society , 38 : 55 – 57 . OpenURL Placeholder Text WorldCat Enberg K. , Jorgensen C., Mangel M. 2010 . Fishing-induced evolution and changing reproductive ecology of fish: the evolution of steepness . Canadian Journal of Fisheries and Aquatic Sciences , 67 : 1708 – 1719 . Google Scholar Crossref Search ADS WorldCat Fitzmaurice M. 2016 . Whaling and International Law . Cambridge University Press , Cambridge, UK . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Fitzmaurice M. , Tamada D. 2016 . Whaling in the Antarctic. Significance and Implications of the ICJ Judgment , Brill, Leiden . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Friedman D. 1971 . Brownian Motion and Diffusion . Holden-Day , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Gillespie D. T. 2007 . Stochastic simulation of chemical kinetics . Annual Review of Physical Chemistry , 58 : 35 – 55 . Google Scholar Crossref Search ADS PubMed WorldCat Gillespie D. T. , Mangel M. 1981 . Conditioned averages in chemical kinetics . Journal of Chemical Physics , 75 : 704 – 709 . Google Scholar Crossref Search ADS WorldCat Giske J. , Eliassen S., Fiksen O., Jakobsen P. J., Aksnes D. L., Jorgensen C., Mangel M. 2013 . Effects of the emotion system on adaptive behavior . The American Naturalist , 182 : 689 – 703 . Google Scholar Crossref Search ADS PubMed WorldCat Giske J. , Eliassen S., Fiksen O., Jakobsen P. J., Aksnes D. L., Mangel M., Jorgensen C. 2014 . The emotion system promotes diversity and evolvability . Proceedings of the Royal Society B , 281 : 20141099. Google Scholar Crossref Search ADS WorldCat Gomulkiewicz R. S. , Brodziak J. K. T., Mangel M. 1990 . Ranking loci for genetic stock identification by curvature methods . Canadian Journal of Fisheries and Aquatic Sciences , 47 : 611 – 619 . Google Scholar Crossref Search ADS WorldCat Haley K. B. , Stone L. D. 1980 . Search Theory and Its Applications . Plenum Press , New York . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Hay W. W. 2016 . Experimenting on a Small Planet . Springer-Verlag , Heidelberg, New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC He X. , Mangel M., MacCall A. 2006 . A prior for steepness in stock-recruitment relationships, based on an evolutionary persistence principle . Fishery Bulletin , 104 : 428 – 438 . OpenURL Placeholder Text WorldCat Hein A. M. , Carrara F., Brumley D. R., Stocker R., Levin S. A. 2016 . Natural search algorithms as a bridge between organisms, evolution, and ecology . Proceedings of the National Academy of Sciences , www.pnas.org/cgi/doi/10.1073/pnas.1606195113 OpenURL Placeholder Text WorldCat Hilborn R. 2016 . Of mice, fishermen, and food . ICES Journal of Marine Science , 73 : 2167 – 2173 . Google Scholar Crossref Search ADS WorldCat Hilborn R. , Mangel M. 1997 . The Ecological Detective. Confronting Models with Data . Princeton University Press , Princeton, NJ . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Holmes F. L. 1991 . Hans Krebs. The Formation of a Scientific Life 1900–1933 . Oxford University Press , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Houston A. , Clark C. W., McNamara J. M., Mangel M. 1988 . Dynamic models in behavioural and evolutionary ecology . Nature , 332 : 29 – 34 . Google Scholar Crossref Search ADS WorldCat Houston A. I. , McNamara J. M. 1999 . Models for Adaptive Behavior. An Approach Based on State . Cambridge University Press , Cambridge . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Jorgensen C. , Engberg K., Mangel M. 2016 . Modelling and interpreting fish bioenergetics: a role for behaviour, life-history traits and survival trade-offs . Journal of Fish Biology , 88 : 389 – 402 . Google Scholar Crossref Search ADS PubMed WorldCat Koopman B. O. 1956a . The theory of search. I. Kinematic . Operations Research , 4 : 324 – 346 . Google Scholar Crossref Search ADS WorldCat Koopman B. O. 1956b . The theory of search, II. Target detection . Operations Research , 4 : 503 – 531 . Google Scholar Crossref Search ADS WorldCat Koopman B. O. 1957 . The theory of search. III. The optimum distribution of searching . Operations Research , 5 : 613 – 626 . Google Scholar Crossref Search ADS WorldCat Koopman B. O. 1980 . Search and Screening. General Principles with Historical Applications . Pergamon Press , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Lasker R. 1966 . Feeding, growth, respiration, and carbon utilization of a Euphausiid Crustacean . Journal of the Fisheries Research Board of Canada , 23 : 1291 – 1317 . Google Scholar Crossref Search ADS WorldCat Lasker R. , Theilacker G. H. 1965 . Maintenance of Euphausiid shrimps in the laboratory . Limnology and Oceanography , 10 : 287 – 288 . Google Scholar Crossref Search ADS WorldCat Little J. D. C. 2002 . Philip M. Morse and the beginnings . Operations Research , 50 : 146 – 148 . Google Scholar Crossref Search ADS WorldCat Ludwig D. , Mangel M., Haddad B. 2001 . Ecology, conservation and public policy . Annual Review of Ecology and Systematics , 32 : 481 – 517 . Google Scholar Crossref Search ADS WorldCat MacArthur R. H. , Wilson E. O. 1967 . The Theory of Island Biogeography . Princeton University Press , Princeton NJ . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Mangel M. 1971 . A treatment of complex ions in sea water . Marine Geology , 11 : M24 – M26 . Google Scholar Crossref Search ADS WorldCat Mangel M. 1980 . Small fluctuations in systems with multiple limit cycles . SIAM Journal of Applied Mathematics , 38 : 120 – 138 . Google Scholar Crossref Search ADS WorldCat Mangel M. 1981a . Search for a randomly moving object . SIAM Journal on Applied Mathematics , 40 : 327 – 338 . Google Scholar Crossref Search ADS WorldCat Mangel M. 1981b . Three bearing method for passive triangulation in systems with unknown deterministic biases . IEEE Transactions on Aerospace Systems , 17 : 814. Google Scholar Crossref Search ADS WorldCat Mangel M. 1982a . Aggregation and fishery dynamics: multiple time scales, times to extinction, and random environments . Ecological Modelling , 15 : 191 – 209 . Google Scholar Crossref Search ADS WorldCat Mangel M. 1982b . Applied mathematicians and naval operators . SIAM Review , 24 : 289 – 300 . Google Scholar Crossref Search ADS WorldCat Mangel M. 1985 . Search models in fisheries and agriculture . Springer Lecture Notes in Biomathematics , 61 : 105 – 137 . Google Scholar Crossref Search ADS WorldCat Mangel M. 1987 . Sampling highly aggregated populations with application to California sardine management . Springer Lecture Notes in Biomathematics , 72 : 247 – 263 . Google Scholar Crossref Search ADS WorldCat Mangel M. 1993 . Models, physics and predictive biological oceanography: KNOW your organism . U.S. Globec NEWS 4 : (August 1993): 1 – 2 . OpenURL Placeholder Text WorldCat Mangel M. 1994 . Barrier transitions driven by fluctuations, with applications to ecology and evolution . Theoretical Population Biology , 45 : 16 – 40 . Google Scholar Crossref Search ADS WorldCat Mangel M. 1989 . Analysis and modelling of the Soviet southern ocean krill fleet, Selected Scientific Papers (1988), Scientific Committee for the Conservation of Antarctic Marine Living Resources, 25 Old Wharf, Hobart, TAS, Australia 7000, SC-CAMLR-SSP/5:127-236. Mangel M. 1990 . Analysis and modeling of the Soviet southern ocean krill fleet, II: Estimating the number of concentrations. Selected Scientific Papers (1989), Scientific Committee for the Conservation of Antarctic Marine Living Resources (25 Old Wharf, Hobart, TAS, Australia 7000), SC-CAMLR-SSP/6 283-322. Mangel M. 1996 . Computing expected reproductive success of female Atlantic salmon as a function of smolt size . Journal of Fish Biology , 49 : 877 – 882 . OpenURL Placeholder Text WorldCat Mangel M. 2006 . The Theoretical Biologist’s Toolbox . Cambridge University Press , Cambridge, UK . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Mangel M. 2015 . Stochastic dynamic programming illuminates the link between environment, physiology, and evolution . Bulletin of Mathematical Biology , 77 : 857 – 873 . Google Scholar Crossref Search ADS PubMed WorldCat Mangel M. 2016 . Whales, Science, and Scientific Whaling in the International Court of Justice . Proceedings of the National Academy of Sciences US, doi/10.1073/pnas.1604988113 . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Mangel M. , Beder J. H. 1985 . Search and stock depletion: theory and applications . Canadian Journal of Fisheries and Aquatic Sciences , 42 : 150 – 163 . Google Scholar Crossref Search ADS WorldCat Mangel M. , Bonsall M. B. 2004 . The shape of things to come: using models with physiological structure to predict mortality trajectories . Theoretical Population Biology , 65 : 353 – 359 . Google Scholar Crossref Search ADS PubMed WorldCat Mangel M. , Ludwig D. 1977 . Probability of extinction in a stochastic competition . Society for Industrial and Applied Mathematics (SIAM) Journal on Applied Mathematics , 33 : 256 – 266 . Google Scholar Crossref Search ADS WorldCat Mangel M. , Ludwig D. 1992 . Definition and evaluation of behavioral and developmental programs . Annual Review of Ecology and Systematics , 23 : 507 – 536 . Google Scholar Crossref Search ADS WorldCat Mangel M. , Plant R. E., Carey J. R. 1984 . Rapid delimiting of pest infestations: a case study of the Mediterranean fruit fly . Journal of Applied Ecology , 21 : 563 – 577 . Google Scholar Crossref Search ADS WorldCat Mangel M. , Clark C. W. 1983 . Uncertainty, search, and information in fisheries . Journal of the International Council for the Exploration of the Seas , 43 : 93 – 103 . OpenURL Placeholder Text WorldCat Mangel M. , Clark C. W. 1986a . Search theory in natural resource modeling . Natural Resource Modeling , 1 : 1 – 54 . Google Scholar Crossref Search ADS WorldCat Mangel M. , Clark C. W. 1986b . Towards a unified foraging theory . Ecology , 67 : 1127 – 1138 . Google Scholar Crossref Search ADS WorldCat Mangel M. , Clark C. W. 1988 . Dynamic Modeling in Behavioral Ecology . Princeton University Press , Princeton, NJ . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Mangel M. , Dowling N. A. 2016 . Reference points for optimal yield: a framework for assessing economic, conservation, and sociocultural tradeoffs in ecosystem-based fishery management . Coastal Management , 44 : DOI: 10.1080/08920753.2016.1208884 OpenURL Placeholder Text WorldCat Mangel M. , Samaniego F. J. 1984 . Abraham Wald's work on aircraft survivability . Journal of the American Statistical Association , 79 : 259 – 271 . Google Scholar Crossref Search ADS WorldCat Mangel M. , Satterthwaite W. H. 2008 . Combining proximate and ultimate approaches to understand life history variation in salmonids with application to fisheries, conservation, and aquaculture . Bulletin of Marine Science , 83 : 107 – 130 . OpenURL Placeholder Text WorldCat Mangel M. , Plant R. E. 1985 . Regulatory mechanisms and information processing in uncertain fisheries . Marine Resource Economics , 1 : 389 – 418 . Google Scholar Crossref Search ADS WorldCat Mangel M. , Nicol S. 2000 . Krill and the unity of biology . Canadian Journal of Fisheries and Aquatic Sciences , 57 : 1 – 5 . Google Scholar Crossref Search ADS WorldCat Mangel M. , Smith P. E. 1990 . Presence–absence sampling for fisheries management . Canadian Journal of Fisheries and Aquatic Sciences , 47 : 1875 – 1888 . Google Scholar Crossref Search ADS WorldCat Mangel M. , Switzer P. V. 1998 . A model at the level of the foraging trip for the indirect effects of krill (Euphausia superba) fisheries on krill predators . Ecological Modelling , 105 : 235 – 256 . Google Scholar Crossref Search ADS WorldCat Mangel M. , Kindsvater H. K., Bonsall M. B. 2007 . Evolutionary analysis of life span, competition, and adaptive radiation, motivated by the Pacific rockfishes (Sebastes) . Evolution , 61 : 1208 – 1224 . Google Scholar Crossref Search ADS PubMed WorldCat Mangel M. , Talbot L. M., Meffe G. K., Tundi Agardy M., Alverson D. L., Barlow J., Botkin D. B. 1996 . Principles for the conservation of wild living resources . Ecological Applications , 6 : 338 – 362 . Google Scholar Crossref Search ADS WorldCat Mangel M. , Brodziak J. K. T., DiNardo G. 2010a . Reproductive ecology and scientific inference of steepness: a fundamental metric of population dynamics and strategic fisheries management . Fish and Fisheries , 11 : 89 – 104 . Google Scholar Crossref Search ADS WorldCat Mangel M. , MacCall A. D., Brodziak J., Dick E. J., Forrest R. E., Pourzand R., Ralston S. 2013 . A perspective on steepness, reference points, and stock assessment . Canadian Journal of Fisheries and Aquatic Sciences , 70 : 930 – 940 . Google Scholar Crossref Search ADS WorldCat Mangel M. , Richerson K., Cresswell K. A., Wiedenmann J. R. 2010b . Modelling the effects of UV radiation on the survival of Antarctic krill (Euphausia superba Dana) in the face of limited data . Ecological Modeling , 221 : 2095 – 2101 . Google Scholar Crossref Search ADS WorldCat Marinovic B. , Mangel M. 1999 . Krill can shrink as an ecological adaptation to temporarily unfavorable environments . Ecology Letters , 2 : 338 – 343 . OpenURL Placeholder Text WorldCat Marszalek D. S. , Wright R. C., Hay W. W. 1969 . Function of the test in foraminifera . Transactions – Gulf Coast Association of Geological Societies , 19 : 341 – 352 . OpenURL Placeholder Text WorldCat Marszalek D. S. 1982 . The role of heavy skeletons in vertical movements of non‐motile zooplankton . Marine Behaviour and Physiology , 8 : 295 – 303 . Google Scholar Crossref Search ADS WorldCat McHuron E. M. , Costa D. P., Schwarz L., Mangel M. 2016 . State-dependent behavioural theory for assessing the fitness consequences of anthropogenic disturbance on capital and income breeders . Methods in Ecology and Evolution , 2041-21/2041-210X.12701. OpenURL Placeholder Text WorldCat McNamara J. M. , Houston A. I. 1986 . The common currency for behavioral decisions . The American Naturalist , 127 : 358 – 378 . Google Scholar Crossref Search ADS WorldCat Metcalfe N. B. , Huntingford F. A., Graham W. D., Thorpe J. E. 1989 . Early social status and the development of life-history strategies in Atlantic salmon . Proceedings of the Royal Society of London B , 236 : 7 – 19 . Google Scholar Crossref Search ADS WorldCat Morse P. 1977 . In at the Beginnings: A Physicist’s Life . MIT Press , Cambridge, MA . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Munch S. B. , Mangel M., Conover D. O. 2003 . Quantifying natural selection on body size from field data: Winter mortality in Menidia menidia . Ecology , 84 : 2169 – 2177 . Google Scholar Crossref Search ADS WorldCat Munch S. B. , Snover M. L., Watters G. M., Mangel M. 2005a . A unified treatment of top-down and bottom-up control of reproduction in populations . Ecology Letters , 8 : 691 – 695 . Google Scholar Crossref Search ADS WorldCat Munch S. B. , Kottas A., Mangel M. 2005b . Bayesian nonparametric analysis of stock-recruitment relationships . Canadian Journal of Fisheries and Aquatic Sciences , 62 : 1808 – 1821 . Google Scholar Crossref Search ADS WorldCat Neyman J. 1949 . On the problem of estimating the number of schools of fish . University of California Publications in STATISTICS , 1 : 21 – 36 . OpenURL Placeholder Text WorldCat Neyman J. , Park T., Scott E. L. 1956 . Struggle for existence . The tribolium model: biological and statistical aspects . In Proceedings of the 3rd Berkeley Symposium on Mathematics, Statistics, and Probability, 4 , pp. 41 – 79 . OpenURL Placeholder Text WorldCat Nolting B. C. , Hinkelman T. M., Brassil C. E., Tenhumberg B. 2015 . Composite random search strategies based on non-directional sensory cues . Ecological Complexity , 22 : 126 – 138 . Google Scholar Crossref Search ADS WorldCat Nye M. 2004 . Blackett: Physics, War, and Politics in the Twentieth Century . Harvard University Press , Cambridge, MA . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Oreskes N. 1999 . The Rejection of Continental Drift. Theory and Method in American Earth Science . Oxford University Press , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Park T. 1954 . Competition: An Experimental and Statistical Study. Statistics and Mathematics in Biology . Iowa State Press , Ames : 175 - 195 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Press A. 2016 . Science in the Court! In Whaling in the Antarctic. Significance and Implications of the ICJ Judgment , pp. 346 – 386 . Ed. by Fitzmaurice M., Tamada D.. Brill, Leiden . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Rashevsky N. 1969 . Mathematical biophysics. Physico-mathematics Foundations of Biology , 1 and 2 . Dover Press , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Richerson K. , Watters G. M., Santora J. A., Schroeder I. D., Mangel M. 2015 . More than passive drifters: a stochastic dynamic model for the movement of Antarctic krill . Marine Ecology Progress Series , 529 : 35 – 48 . Google Scholar Crossref Search ADS WorldCat Richerson K. , Santora J. A., Mangel M. 2016 . Climate variability and multi-scale assessment of the krill preyscape near the north Antarctic Peninsula . Polar Biology , DOI 10.1007/s00300-016-1994-y. OpenURL Placeholder Text WorldCat Roitberg B. , Mangel M. 2016 . Cold snaps, heat waves, and arthropod growth . Ecological Entomology , DOI: 10.1111/een.12324. OpenURL Placeholder Text WorldCat Salinas S. , Brown S. C., Mangel M., Munch S. B. 2012 . Non-genetic inheritance and changing environments . Non-Genetic Inheritance, 1: 38 – 50 . OpenURL Placeholder Text WorldCat Sands P. 2016 . East-West Street: On the Origins of "Genocide" and "Crimes Against Humanity" . AA Knopf , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Sands P. , Peel J., Fabra A., MacKenzie R. 2012 . Principles of International Environmental Law , 3rd edn. Cambridge University Press , Cambridge, UK . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Satterthwaite W. H. , Kitaysky A. S., Hatch S. A., Piatt J. F., Mangel M. 2010 . Unifying quantitative life-history theory and field endocrinology to assess prudent parenthood in a long-lived seabird . Evolutionary Ecology Research , 12 : 779 – 792 . OpenURL Placeholder Text WorldCat Satterthwaite W. H. , Mangel M. 2012 . Behavioral models as a common framework to predict impacts of environmental change on seabirds and fur seals . Deep Sea Research II , 65–70 : 304 – 315 . Google Scholar Crossref Search ADS WorldCat Satterthwaite W. H. , Beakes M. P., Collins E., Swank D. R., Merz J. E., Titus R. G., Sogard S. M. et al. 2009 . Steelhead life history on California's central coast: insights from a state-dependent model . Transactions of the American Fisheries Society , 132 : 532 – 548 . Google Scholar Crossref Search ADS WorldCat Satterthwaite W. H. , Beakes M. P., Collins E., Swank D. R., Merz J. E., Titus R. G., Sogard S. M. et al. 2010 . State-dependent life history models in a changing (and regulated) environment: steelhead in the California Central Valley . Evolutionary Applications , 3 : 221 – 243 . Google Scholar Crossref Search ADS PubMed WorldCat Satterthwaite W. H. , Hayes S. A., Merz J. E., Sogard S. M., Frechette D. M., Mangel M. 2012a . State-dependent migration timing and use of multiple habitat types in anadromous salmonids . Transactions of the American Fisheries Society , 141 : 781. Google Scholar Crossref Search ADS WorldCat Satterthwaite W. H. , Kitaysky A. S., Mangel M. 2012b . Linking climate variability, productivity and stress to demography in a long-lived seabird . Marine Ecology Progress Series , 454 : 221 – 235 . Google Scholar Crossref Search ADS WorldCat Satterthwaite W. H. , Mohr M. S., O’Farrell M. R., Anderson E. C., Banks M. A., Bates S. J., Bellinger M. R. et al. 2014 . Use of genetic stock identification data for comparison of the ocean spatial distribution, size at age, and fishery exposure of an untagged stock and its indicator: California Coastal versus Klamath River Chinook salmon . Transactions of the American Fisheries Society , 143 : 117 – 133 . Google Scholar Crossref Search ADS WorldCat Sheldon R. 2016 . Military Operations Research Society (MORS) Oral History Project Interview of Dr. Phil E. DePoy . Military Operations Research , 22 : 7 – 36 . OpenURL Placeholder Text WorldCat Shelton A. O. , Satterthwaite W. H., Beakes M. P., Munch S. B., Sogard S. M., Mangel M. 2013 . Separating intrinsic and environmental contributions to growth and their population consequences . The American Naturalist , 181 : 0003 – 0147 . /2013/18106. 53990. Google Scholar Crossref Search ADS WorldCat Shelton A. O. , Kinzey D., Reiss C., Munch S., Watters G., Mangel M. 2013 . Among-year variation in growth of Antarctic krill Euphausia superba based on length-frequency data . Marine Ecology Progress Series , 481 : 53 – 67 . Google Scholar Crossref Search ADS WorldCat Smith P. E. 1973 . The mortality and dispersal of sardine eggs and larvae . Rapports et procès-verbaux des réunions / Conseil permanent international pour l'exploration de la mer , 164 : 282 – 292 . OpenURL Placeholder Text WorldCat Smith P. E. 1978 . Biological effects of ocean variability: time and space scales of biological response . Rapp. P-V. Cons. Int. Explor. Mer , 173 : 117 – 127 . OpenURL Placeholder Text WorldCat Snover M. L. , Watters G. M., Mangel M. 2005 . Interacting effects of behavior and oceanography on growth in salmonids with examples for coho salmon (Oncorhynchus kisutch) . Canadian Journal of Fisheries and Aquatic Sciences , 62 : 1219 – 1230 . Google Scholar Crossref Search ADS WorldCat Snover M. L. , Watters G. M., Mangel M. 2006 . Top-down and bottom-up control of life history strategies in coho salmon (Oncorhynchus kisutch) . American Naturalist , 167 : E140 – E157 . Google Scholar Crossref Search ADS PubMed WorldCat Sogard S. M. , Merz J. E., Satterthwaite W. H., Beakes M. P., Swank D. R., Collins E. M., Titus R. G. et al. 2012 . Contrasts in habitat characteristics and life history patterns of Oncorhynchus mykiss in California's Central Coast and Central Valley . Transactions of the American Fisheries Society , 141 : 747 – 760 . Google Scholar Crossref Search ADS WorldCat Southwood R. , Henderson P. A. 2009 . Ecological Methods . Blackwell Publishing , Oxford, UK . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Springsteen B. 2016 . Born to Run . Simon and Schuster , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Stefanou S. E. , Mangel M., Wilen J. E. 1986 . Information in agricultural pest control . Journal of Agricultural Economics , 37 : 77 – 88 . Google Scholar Crossref Search ADS WorldCat Stokes D. 1997 . Pasteur’s Quadrant. Basic Science and Technological Innovation . Brookings Institution Press , Washington, DC . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Stone L. D. , Royset J. O., Washburn A. R. 2016 . Optimal Search for Moving Targets . Springer , Heidelberg . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Takeuchi H. , Uyeda S., Kanamori H. 1967 . Debate About the Earth: Approach to Geophysics through Analysis of Continental Drift . Freeman, Cooper & Co. , San Francisco, CA . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Thorpe J. E. , Mangel M., Metcalfe N. B., Huntingford F. A. 1998 . Modelling the proximate basis of life history variation, with application to Atlantic salmon, Salmo salar L . Evolutionary Ecology , 121 : 581 – 600 . Google Scholar Crossref Search ADS WorldCat Tidman K. 1984 . The Operations Evaluation Group . Naval Institute Press , Annapolis, MD . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Vincenzi S. , Mangel M. 2013 . Linking food availability, body growth and survival in the black-legged kittiwake Rissa tridactyla . Deep-Sea Research IIhttp://dx.doi.org/10.1016/j.dsr2.2013.03.029i. OpenURL Placeholder Text WorldCat Vincenzi S. , Mangel M. 2014 . Food abundance, kittiwake life histories, and colony dynamics in the Northeastern Pacific: implications of climate change and regime shifts . Marine Ecology Progress Series , 515 : 251 – 263 . Google Scholar Crossref Search ADS WorldCat Vincenzi S. , Hatch S., Mangel M., Kitaysky A. 2013 . Food availability affects onset of reproduction in a long-lived seabird . Proceedings of the Royal Society B , 280 : 20130554 . Google Scholar Crossref Search ADS PubMed WorldCat Watters G. M. , Hill S. L., Hinke J. T., Matthews J., Reid K. 2013 . Decision-making for ecosystem-based management: evaluating options for a krill fishery with an ecosystem dynamics model . Ecological Applications , 23 : 710 – 725 . Google Scholar Crossref Search ADS PubMed WorldCat Wiedenmann J. , Creswell K., Mangel M. 2008 . Temperature-dependent growth of Antarctic krill: predictions for a changing climate from a cohort model . Marine Ecology Progress Series , 358 : 191 – 202 . Google Scholar Crossref Search ADS WorldCat Wiedenmann J. , Creswell K. A., Mangel M. 2009 . Connecting recruitment of Antarctic krill and sea ice . Limnology and Oceanography , 54 : 799 – 811 . Google Scholar Crossref Search ADS WorldCat Wiedenmann J. , Cresswell K. A., Goldbogen J., Potvin J., Mangel M. 2011 . Exploring the effects of reductions in krill biomass in the Southern Ocean on blue whales using a state-dependent foraging model . Ecological Modelling , 222 : 3366 – 3379 . Google Scholar Crossref Search ADS WorldCat Author notes " †Food for Thought articles are essays in which the author provides their perspective on a research area, topic, or issue. They are intended to provide contributors with a forum through which to air their own views and experiences, with few of the constraints that govern standard research articles. This Food for Thought article is one in a series solicited from leading figures in the fisheries and aquatic sciences community. The objective is to offer lessons and insights from their careers in an accessible and pedagogical form from which the community, and particularly early career scientists, will benefit. " The International Council for the Exploration of the Sea (ICES) and Oxford University Press are pleased to make these Food for Thought articles immediately available as free access documents. © International Council for the Exploration of the Sea 2017. All rights reserved. For Permissions, please email: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Jobling,, Malcolm

doi: 10.1093/icesjms/fsx012pmid: N/A

This article deals with reflections on a career as a fish biologist that has spanned 40 + years. I provide insights into lessons I have learned over the years, describe some research successes and failures, and end by drawing the conclusion that variety in teaching and research is the spice of academic life. The lessons I have learned might be a guide for those at an early stage of their scientific career: You should try to recognize where your talents lie and what your weaknesses are: play to your strengths, but do not try to hide your weaknesses. Collaboration with perceptive colleagues and receptive associates is the foundation upon which an academic career is built. Research should be designed to address a problem, not apply a technique; use the technology, and do not submit to the temptation of letting it use you. The most rewarding pieces of research are often those that produce unexpected results; you learn more from having your preconceived ideas challenged than by having them confirmed. It is important to know who your readers are, and to select the most appropriate channel for publication of your work. It is not necessary to publish in high impact factor journals to get your work read and cited. You are likely to shift your focus many times during your career; you must continue to grow to remain fresh and enthusiastic. Your students are your scientific legacy; nurture them well because there is nothing more satisfying than to see them succeed. I offer a final piece of advice: Should teaching and research cease to be fascinating and fun, move on; do not risk becoming a square peg in a round hole. Introduction This article is a personal reflection of my 40 + years as a fish biologist with a career in teaching and research. The narrative provides some insights into the lessons I have learned over those years. Although I am sure that each of us has a unique story to tell about how and why we entered academia, we may agree that there are a few universal rules that can guide young scientists. Let me start this retrospection with a couple of confessions. Scientific research has neither been my main interest nor the major driving force of my academic career. That is not to say that I have considered research to be a necessary evil. I reserve the term necessary evil for the administrative duties that occupy an increasing amount of time as academia becomes more regimented and bureaucratic, and as I have climbed the academic ladder. Research has been, and continues to be, fun; but in my psyche research plays second fiddle to teaching. As a result, I think of my book contributions as being my most satisfying and rewarding academic works (e.g. Jobling, 1994, 1995; Houlihan et al., 2001; Le Francois et al., 2010; Huntingford et al., 2012). It is, however, important to realize that teaching and research are not separate entities. I think of them as Siamese twins, sharing much in common and with a need for synergy between the two. My second confession is that very little in my academic career has been planned. In writing this I do not admit to having been a gadabout, it is just that so much in my career has been serendipity. Early inspiration, cold feet, and U-turns When I started secondary school on Tees-side, in the industrial heartland of the north–east of England, I was fortunate to be taught biology and chemistry by two young and dynamic teachers who did not play things according to the book; both were enthusiastic expounders of “learning by doing”. In chemistry classes there was a focus on the running of simple experiments that were written up as short scientific articles in the IMRAD format, and biology was an observational science, both indoors and out. Inspired, my career course seemed set; I looked to follow in the footsteps of these two early mentors and become a secondary school teacher. In senior school, I was given access to the teaching laboratories and could work on small projects: looking at the metabolic rates of insects, measuring enzyme activities and trying out a few histological techniques, for example. All of this stood me in good stead for the transition from school to university. At the University of Hull, I took a B.Sc. degree in Botany and Zoology, and followed this with a year of teacher training. This, to prepare me for what appeared to be my destiny. I was soon to receive a rude awakening; a full-time teaching post in an inner-city secondary school on Tees-side was neither Utopia nor Shangri-La, far from it. In my first, and only, secondary school teaching post I was timetabled to teach more mathematics and English than biology and chemistry, and many of the students were difficult to motivate. In less than a month I had become disillusioned and was looking to escape. In desperation, I contacted my former tutor at the University of Hull for both solace and advice; with hindsight this is one of the best moves I have ever made. He had his Ph.D. in marine biology from the University College of North Wales, Bangor, and had friends and colleagues who were still at the Marine Laboratories in Menai Bridge. By chance he had been told that somebody had withdrawn from the M.Sc. course in marine biology at the last minute, leaving a vacancy; did he know of anybody who might be interested? Although my leanings were mostly botanical, my undergraduate studies had also re-awakened a childhood interest in marine animal life, so I applied and was offered the place on the course. As a result, I resigned from my school-teaching position after less than two months, much to the chagrin of the head of school, and set off for Bangor. Seeing the light and becoming an ichthyophile Little did I realize when I started my M.Sc. studies in Bangor that the experience would lead to both a life-long passion and open the door to a career in academia. When I arrived at the Marine Laboratories in Menai Bridge, I intended to take my M.Sc. thesis project on macroalgae. By Christmas all had changed; I had become fascinated by fish, and had become an ichthyophile. This is a calling I have followed to the present day. How did the transition come about? A case of serendipity. Following a hockey match the team adjourned to the local pub for a post-match pint or two of beer. By chance, I was sitting at the same table as Dave Grove, who was later to supervise my thesis work and become an inspirational scientific sparring partner during the formative years of my career. Thus began my conversion from plant ecologist to fish biologist. Within weeks, I was spending quite a lot of time in the Nuffield Fish Lab, and had been introduced to the fish physiologists Ragnar Fänge, Steffan Nilsson, and Sue Holmgren, who were on a visit from the University of Gothenburg, Sweden. The atmosphere in the Fish Lab was very relaxed, with everybody being allowed to voice an opinion, be heard and be treated as an equal. There was also practical joking and I was the butt of one of Dave’s. My knowledge about fish was extremely rudimentary so I asked Dave to suggest something I could read as a starter. When Dave gave the quick reply “Go and look at Fish Physiology by Hoar and Randall; you’ll find it in the main University library in Bangor” I heard a couple of the doctoral students snigger, but I didn’t know why they found the remark funny. I trotted off to the library to discover that Fish Physiology was a six volume treatise; I felt like the young apprentice sent to the factory stores by the foreman for “a pot of striped paint” or a “long stand”. Having overcome my embarrassment, and suppressed my irritation, I started to browse through the tables of contents, and discovered that the books were a goldmine: I was particularly attracted to the chapters by Phillips (1969), Fry (1971), Gleitman and Rozin (1971) and Hasler (1971) and read these over the ensuing few weeks. One of the first things I did when I was awarded tenure, and was given a budget, was to buy a personal copy of this series of books. By this time, the number of volumes had increased from six to eight, and included the volume on bioenergetics and growth (Hoar et al., 1979) that was my “scientific bible” for many years. To cut a long story short, I did my M.Sc. project, not on macroalgal ecology, but on feeding and gastro-intestinal function in the dab, Limanda limanda. With the benefit of hindsight I can say that the project was too ambitious and the aims unrealistic. We were interested in searching for possible relationships between rates of gastric emptying and appetite return, using X-radiography to study gastric emptying and self-feeders to look at feeding rhythms. At the time both techniques were novel, and there were few studies in which fish had been the experimental animals. We had some background upon which to base the X-ray studies (Molnar et al., 1967; Edwards, 1971; Goddard, 1974) and self-feeders had been used to examine the regulation of food intake and learning abilities of goldfish, Carassius auratus (Rozin and Mayer, 1961, 1964) and rainbow trout, Oncorhynchus mykiss (Adron et al., 1973). The sub-project involving X-radiography went smoothly, but the study of feeding rhythms of a bottom-living marine fish using surface-operated self-feeders was fraught with difficulties. First we had to wean the wild-caught dabs to accept a dry pellet feed, and once weaned we had to train the fish to operate a trigger to obtain food. This took weeks; much exasperation and frustration, but our patience was rewarded in the end. The self-feeders were more like the contraptions drawn by the English cartoonist William Heath Robinson (1872–1944) than scientific equipment; they were made from spring-loaded tea dispensers, reed switches, rubber stop-bars cut from erasers and small permanent magnets attached to thin, swinging metal triggers. Flimsy wiring connected the self-feeders to event recorders wrapped in polythene sheets to protect them from the salt-spray that would have caused short-circuiting and corrosion. Self-feeder designs and reliability have improved substantially since then, and they have been used to study feeding rhythms and behaviour, and diet selection in fish, as well as being adopted as a feeding system in the commercial farming of some fish species (Jobling et al., 1995; Houlihan et al., 2001). Needless to say, our first-generation self-feeders often malfunctioned, and it was not easy to obtain records of feeding activity over prolonged periods of time. Although this part of my thesis project fell flat on its face, it somehow aroused a passion in me for asking questions about what fish do, how they do it and why. The X-ray work from my thesis gave me my first scientific article (Jobling et al., 1977). Dave and Ragnar used two of the figures from the article in their chapter on fish digestion published in the Fish Physiology book series (Fänge and Grove, 1979); as proof of ownership, my finger- and thumb-prints are on the X-radiographs used as Fig. 5 in the chapter. I had been too eager and clumsy when developing the plates and handled them before they were dry, thereby leaving an indelible impression. Such clumsiness when carrying out laboratory work has dogged me throughout my career, and I have never considered hands-on experimental work to be my forte. Learning to know one’s scientific self Fortunately, scientific research is as much a mental activity as a manual one, and a key to success is to recognize where your talents lie and where you have weaknesses. I have always enjoyed writing, and prefer presenting data to collecting them. These are useful attributes to have when editing and reviewing manuscripts; activities that have occupied increasing proportions of my time in recent years (https://publons.com/author/18696/malcolm-jobling#profile). Over the years, I have edited a few books, conference proceedings and special issues of journals (e.g. Houlihan et al., 2001; Le Francois et al., 2010; Huntingford et al., 2012) and have been involved in shaping the development of fisheries and aquaculture journals. It is a privilege to be able to use one’s skills to serve the international scientific community in this capacity. I am grateful to the journal editors, particularly Alwyne Wheeler and John Blaxter, who took the time to help me hone my writing skills as an early career scientist. I hope that I am now paying back some of this debt. I have never considered myself to be a good administrator and organizer, although I have occasionally heard colleagues express the opposite view. This means that I have never actively sought administrative office, and it is only rarely that I volunteer to take the lead role on a scientific panel; nor will I accept the invitation to join a committee unless I feel that I can make an important contribution. As time has gone by, I feel that I have developed into a whole-animal biologist who has a knack for tying together loose threads and weaving tapestries. More specifically, I think of myself as a fish eco-physiologist; I would define fish eco-physiology as being the branch of ecology concerned with the study of how the physiology of fishes is influenced by environmental changes. The principles and techniques of fish eco-physiology can be applied to both wild fish and those held in captivity as farmed species. I feel that I am more a synthesizer, than one who is able to conjure up new and original ideas. This means that mulling over the ideas of others, along with reading their works and discussing science with my peers and students has been a major stimulus for me throughout my career. More than a fair share of good luck has allowed me to carve out the academic career of my choice. Formative years and becoming established I was in Bangor for only one year, and then moved to the University of Glasgow to continue work on fish for my doctorate; looking at feeding and metabolism of the plaice, Pleuronectes platessa. As a doctoral student I was given a lot of freedom to formulate my research and pursue my whims and fancies. It was a relief not to be constrained by a rigid project description and the need to meet the strict reporting deadlines that is the lot of most doctoral students to-day. I enjoyed my academic freedom and independence immensely, and I am sure that I developed a wide range of scientific interests as a consequence. My starting point for building upon my basic background in energetics and metabolism gleaned from reading relevant chapters in Fish Physiology (Phillips, 1969; Fry, 1971) was to turn to the classic works by Krogh (1916), Winberg (1956) and Kleiber (1961). I also continued my interest in gastro-intestinal function, and the medical journals Gut and Gastroenterology became regular reading; providing me with ideas, methods and models that I could apply to my fish studies (Jobling and Davies, 1979; Jobling, 1986). It is my experience that, to use modern parlance, “going outside the box” is stimulating, broadens horizons and opens up new opportunities for research. In parallel with the fish gut work, I started to run respirometry studies. A chance encounter with Ashworth’s (1969) article on malnourished children set me on a track that led me to look at specific dynamic action (SDA) and possible relationships between protein metabolism, metabolic rates, and growth (Jobling and Davies, 1980; Jobling, 1981, 1985). There were also forays into the applied animal nutrition literature to obtain information about principles and analytical techniques (Maynard and Loosli, 1969; Halver, 1972). This stood me in good stead some years later when my research started to shift towards, and eventually become dominated by, studies with application to fish farming (Jobling, 1988, 2004, 2016). Following a stay of just over four years at the University of Glasgow I had a chance to move to Tromsø, in the north of Norway, and have been an incumbent at the university ever since. It was not my intention to carve out a career at the University of Tromsø. I had planned to stay for no more than three-to-four years before seeking employment in Canada; I considered Canadian universities and research institutes to be the centres at which the most exciting research in fish eco-physiology was being carried out. Work during the early years in Tromsø involved developing courses in aquatic ecology, fish biology, physiology and aquaculture for students of fisheries science, writing up the manuscripts from my doctoral thesis and getting them published, and conducting research that underwent a shift in a more applied direction. After three years as a temporary member of staff, I was offered tenure, which I accepted, and three years later I was awarded a professorship; the post I currently hold. Some of the articles resulting from my doctoral work were well-received and resulted in invitations to hold conference talks and contribute chapters to proceedings and books (e.g. Jobling, 1985, 1986). My course teaching also developed over time. Teaching compendia were written, revised and improved to incorporate recent research findings, and this eventually resulted in invitations to write textbooks (Jobling, 1994, 1995). This was hard work, but I found it very rewarding; the necessity to focus on communication and explanation increased my awareness and resulted in me learning a lot. I think it unfortunate that many seem to feel that there is little incentive to write textbooks to-day because the tangible rewards are so few. The monetary rewards are not great, and in the competition for appointments and academic advancement the greatest emphasis is placed on research output. I still teach fish biology, physiology and aquaculture to students of fisheries science; I wonder if I may hold some sort of endurance record here? Course contents have been revised and updated, but the teaching philosophy has remained the same; a throwback to the lessons I learned from my earliest mentors in secondary school—“learning by doing”. My experiences with reviewing and editing have given me insights into the publication process, and I try to communicate these to postgraduate students when giving a course in scientific writing; once again “learning by doing” are the watchwords. Over the years, changes in my personal research interests, along with research funding trends and fashions, have led me into a variety of basic and applied research areas. Those studies have covered areas as diverse as nutritional requirements, biorhythms, fish metabolism and growth, fish as food, endocrine regulatory mechanisms, and gene–environment interactions (https://scholar.google.no/citations?user=NFoaMx4AAAAJ&hl=no&oi=ao; accessed 3 January 2017.). Such a diversity in research during a career spanning 40+ years is probably not unusual. It is essential to blossom and grow during your career, remain fresh and enthusiastic, and make sure that research continues to be both fascinating and fun. No man is an island When I look at my scientific production, it comes as no surprise to me that it is review articles and book chapters that top the list of works with most citations (https://scholar.google.no/citations?user=NFoaMx4AAAAJ&hl=no&oi=ao). My name is the only one on many of these publications, and this may give the false impression that I have been an academic hermit, working alone in an isolation cell of my own construction. All of you know that interactions with knowledgeable colleagues and enthusiastic students are an essential part of academic life; these interactions are your most important stimuli and drivers. I have been fortunate to work with students and colleagues more dexterous than myself in the laboratory, and these collaborations have resulted in the publication of several articles that have been highly-cited. It is particularly pleasing that some of these arose from thesis projects carried out by M.Sc. students. The works that have proved popular cover diverse themes, and it has been a surprise to me that some of these publications have been so widely-cited. I put this down to having a canny knack of being in the right place at the right time; serendipity. One oft-cited publication that came as a surprise arose as a spin-off from the thesis work of my first M.Sc. student, Anne Breiby. Her project involved the investigation of the dietary habits of the squid, Todarodes sagittatus, in which sagittal otoliths present in the stomach contents were used to identify fish prey. The pH of the squid stomach is only very slightly acidic. This meant that recovered fish otoliths were rarely eroded, identification of prey was easy and otolith lengths could be used to estimate the size of the fish prey. As part of the project we had made a compilation of the literature covering the use of fish otoliths in dietary studies. When looking through this literature, I became concerned about possible sources of error when scat analysis was used for assessing predation on fish by marine mammals. Did fish otoliths isolated from scats present a realistic picture of the fish species consumed; the entire fish would need to be consumed for the otolith to appear in the scat? Did erosion of the otoliths in the acidic mammalian stomach introduce serious errors for back-calculation of prey size from otolith size? Could the data be used for a realistic examination of the bioenergetics of predation? We decided to run a very simple in vitro otolith erosion test by exposing fish otoliths to solutions that differed in pH and then measure their lengths at regular intervals. We ran the test over the Easter weekend holiday period, and I presented the data in a short talk given at a meeting of Norwegian marine biologists during the following autumn. Following my talk I was approached by Torleiv Brattegard, who was the editor of the Norwegian journal Sarsia (now Marine Biology Research). He asked me if I would be interested in writing a manuscript for submission to the journal. Sarsia, named after the marine scientist Michael Sars (1805–1869), was produced by the Marine Biological Station, University of Bergen, and was a journal with a limited circulation and readership. That this article (Jobling and Breiby, 1986) has become so widely-cited serves as an indication that even the results of simple experiments published in relatively obscure, parochial journals have the potential to make an impact. During the course of another M.Sc. project, in which Ilona Miglavs was looking at RNA:DNA as a biomarker of recent growth history, we became interested in the growth responses shown by fish following periods of reduced feeding; the compensatory growth phenomenon (Miglavs and Jobling, 1989). The results of the work did not convince us that monitoring RNA:DNA would serve our purposes as a biomarker of growth so we did not pursue this further, but we have returned to studies of compensatory growth on a number of occasions in the intervening years. For example, we looked at how compensatory growth could be exploited in capture-based aquaculture of Atlantic cod, Gadus morhua (Jobling et al., 1994) and also used the compensatory growth response to examine the mechanisms that regulate body growth and composition in Atlantic salmon, Salmo salar (Johansen et al., 2001). Both of these projects involved the participation of, and substantial input from, postgraduate students. I was familiar with the use of X-radiography from my M.Sc. studies on gut physiology of the dab, L.limanda, but the introduction of particulate markers to replace dispersed contrast medium added an extra dimension to the technique (Talbot and Higgins, 1983). When combined with individual recognition of fish by tagging there were new opportunities to use X-radiography in studies of fish feeding and growth (Jobling et al., 1995; Houlihan et al., 2001). We were quick to adopt and apply this technique, and it became a cornerstone of our studies for almost two decades; these studies included our work on the effects of exercise training on salmonids. This work attracted a wide audience and resulted in practical applications to the fish farming industry. It is likely that most work that attracts attention and recognition is the result of researchers being fortunate enough to be in a certain place at the time an opportunity arises, being able to see that opportunity and then to pursue it with as much creativity as they possess. During the course of his thesis research Jørgen Schou Christiansen had seen that Arctic charr, Salvelinus alpinus, exposed to flowing water seemed to grow slightly faster than those held in static water. At the time, the prevailing idea was that slow tangential water flows in fish tanks should be used to facilitate self-cleaning, but that higher flows should be avoided because they were detrimental to growth: “Farmed fish kept in raceways or tanks with a concentric flow must swim to maintain station. In so doing they are using up energy sources which would otherwise have been stored in growing body tissues. Farmed salmon do better in terms of food conversion and growth rate in static water where there is no directional flow… .” (Sedgwick, 1988). I was, understandably, skeptical about Jørgen’s claims and asked him to check his data. If the exercised fish were growing better this needed to be examined in more detail. When we carried out a full-scale study the results revealed that Arctic charr exposed to flowing water reduced haphazard swimming and aggressive behaviour, orientated against the current and formed schools, increased their growth rates and showed improved feed conversion efficiency (Christiansen and Jobling, 1990). Some of the most satisfying research is when your preconceptions are shown to be incorrect. Subsequently, follow-up studies were carried out on both Arctic charr and Atlantic salmon, Salmosalar (Jobling et al., 1993), and we were able to make a contribution to the earliest work on the welfare of farmed fish. We recently returned to some of these data when preparing an overview of environmental requirements for farming of Arctic charr (Sæther et al., 2016). Our work on exercise training of salmonids has been cited regularly over the years, but my feeling is that it has been somewhat neglected by recent recruits to the ranks of those who investigate fish welfare. If I have a criticism of early-career scientists, it is that many of them do not devote sufficient time and effort to exploring the early work that is the foundation upon which recent studies have been built. Perhaps this is the result of a pressure to produce rather than reflect, combined with the disturbing trend of some reviewers and journal editors to shun manuscripts that contain citation of articles that are of venerable age. As a final example of an oft-cited article that owes more to chance than good planning and management, I will recount the tale of a piece of work I did with my Finnish colleague Juha Koskela (Jobling and Koskela, 1996). Juha and I were invited to teach on a practical course “Experimental methods in fish biology” for Finnish postgraduate students, and were asked to demonstrate the use of the X-radiographic technique. We opted to be ambitious; not only to demonstrate the technique but also to get the students to collect data that could show some of its applications. Prior to the course, Juha set up tanks with individually-tagged (PIT-tagged) rainbow trout, O.mykiss, some of which were fed in excess and others that were given a restricted ration. The aim was to use this design to collect data to illustrate the effects of restricted rations on hierarchy development and inter-individual variations in feed intake. We hoped to demonstrate this using X-radiographic measurements of feed intake collected on day 1 of the course. We planned to repeat the exercise on day 2 to demonstrate the negative effects of handling stress on the fish, using feed intake as the metric. The measurements made on day 1 gave results that met our expectations in full, but those from day 2 gave us a surprise; the fish had not reduced their feed intake despite being handled and X-ray photographed the previous day. Fortunately, we had not told the students about our expectations beforehand, so Juha and I did not lose face. We were quick to see that the unexpected result offered us the opportunity to run an interesting small-scale feeding trial, incorporating an investigation of compensatory growth. Juha had the chance to retain the tanks of marked trout for a number of weeks, so the same evening we mapped out an experimental design over a couple of beers in Juha’s summerhouse. This was a piece of research work that was fun to do, but I must confess to being surprised that the resulting article (Jobling and Koskela, 1996) has been cited so often. Admittedly, our approach was novel and we grasped our opportunity when it came along, but the popularity of the resulting article still came as a surprise. Mulling and musing I will start these concluding comments by taking the risk of being accused of stating the obvious: Research should be driven by curiosity, the posing of questions and the desire to solve problems, rather than by the wish to apply a particular technology or technique. Techniques and technologies are there to help answer questions, not to determine the types of experiments performed and shape their designs. Nonetheless, it can be tempting to follow trends, jump aboard the latest technological bandwagon and try to find experiments that can fit a technique. For example, when I started my career molecular biology was a relatively new area of scientific specialization, and very few marine scientists and fish biologists were using molecular techniques in their research. All the molecular biologists I met were technique-orientated, and few seemed to know much about the animals they were studying. This was an anathema to me as a whole-animal biologist. It is probably also the main reason for me being slow to open up to molecular biology, grasp the opportunities offered by the techniques and incorporate them into my research. Although the situation has improved, I can still often detect cases in the published literature where technique seems to have determined problem, rather than the reverse. As researchers, we all like a puzzle, and many of us also have a masochistic streak. We often find that research is more fun when the findings indicate that our working hypothesis is probably incorrect, and a return to the drawing board is called for. I have almost certainly learned more from being wrong than from being right; our work on the effects of exercise on salmonids is a prime example (Jobling et al., 1993). Unexpected findings and outliers in the data should not be ignored, and hidden from sight by being swept under an intellectual carpet; they are most likely to be the inspiration for new experiments and the source of your next research grant application. Throughout my career as a faculty member at a university I have been expected to perform teaching, research and administrative duties, but I have not been either equally motivated to perform, or equally proficient at, each type of duty. Administration has been my Achilles’ heel and in recent years I have become increasingly aware that I am not particularly adept at playing the role of the bureaucrat. These days it seems as though senior faculty are expected to master not only the administrative duties required for the day-to-day running of their academic department, but also have business acumen and accountancy skills. They are expected to be adept at project and personnel management, be industrial entrepreneurs with knowledge about proprietary rights and patenting regulations, and to display their social and political awareness via membership of national and international scientific committees. There are also increasing pressures to popularize science through outreach activities, scientific journalism, blogging and other forms of social media, and open chat-lines with members of the public. More-and-more, I feel as though I am being coerced into becoming a jack-of-all-trades, with a severe risk of ending my career as a master of none. These recent changes in academic life mean that it is an unfortunate fact that the pressures on early career scientists are much greater today than when I started my academic career. These days, young researchers are often constrained within the rigid framework of a project proposal, are expected to publish frequently in top-ranking scientific journals, must meet strict reporting deadlines, are expected to be scientific entertainers and social media gurus, and are all-to-often faced with demands to prove their worth. In my opinion, there are aspects of academic stewardship that are questionable, and should be challenged and debated. Is academia being steered on a safe and secure course, or can we expect to see wreckage on the rocks? Those of us who are nearing the end of productive academic careers must make sure that we provide adequate encouragement to the early career scientists who are to be our successors; we must mentor them well. Even when days are dark and little seems to be going right we must instill them with a sense of self-belief and convince them that their academic activities have far more positives than negatives. We must ensure that they are given sufficient time for intellectual reflection, we must strive to prevent them from succumbing to the temptation of resorting to dubious research and publication practices and we must make sure that they do not burn-out or become disillusioned before they turn 40. As a Parthian shot; Variety in teaching and research is the spice of a life in science. Should you no longer find that teaching and research are fun you should leave the game and search for pastures new; do not risk becoming a square peg in a round hole. Acknowledgements Thanks to Howard Broman for inviting me to write this short memoir. Such invitations are, however, a double-edged sword; both an honour and an indication that your days are probably numbered. Thanks to my close colleagues Per-Arne Amundsen and Jørgen Schou Christiansen for providing feedback on my draft manuscript; as ever, your comments served to hold my ego in check and impose controls on my idiosyncratic and quirky character. Thanks also to my mentors who saw something in a young, obstreperous biologist (= yob) that I was unable to see in myself. Final, a big thank-you to the unsung heroes and heroines, the et alia, whose important efforts and contributions are rarely fully recognized. I hope that this short memoir can serve as a small token of my appreciation. References Adron J. W. , Grant P. T., Cowey C. B. 1973 . A system for the quantitative study of the learning capacity of rainbow trout and its application to the study of food preferences and behavior . Journal of Fish Biology , 5 : 625 – 636 . Google Scholar Crossref Search ADS WorldCat Ashworth A. 1969 . Metabolic rates during recovery from protein-calorie malnutrition: the need for a new concept of Specific Dynamic Action . Nature , 223 : 407 – 409 . Google Scholar Crossref Search ADS PubMed WorldCat Christiansen J. S. , Jobling M. 1990 . The behavior and the relationship between food intake and growth of juvenile Arctic charr, Salvelinus alpinus L., subjected to sustained exercise . Canadian Journal of Zoology , 68 : 2184 – 2191 . Google Scholar Crossref Search ADS WorldCat Edwards D. J. 1971 . Effect of temperature on rate of passage of food through the alimentary canal of the plaice, Pleuronectes platessa L . Journal of Fish Biology , 3 : 433 – 439 . Google Scholar Crossref Search ADS WorldCat Fry F. E. J. 1971 . The effect of environmental factors on the physiology of fish. In Fish Physiology , vol. 6 , pp. 1 – 98 . Ed. by Hoar W. S., Randall D. J.. Academic Press , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Fänge R. , Grove D. 1979 . Digestion. In Fish Physiology , vol. 8 , pp. 161 – 260 . Ed. by Hoar W. S., Randall D. J., Brett J. R.. Academic Press , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Gleitman H. , Rozin P. 1971 . Learning and memory. In Fish Physiology , vol. 6 , pp. 191 – 278 . Ed. by Hoar W. S., Randall D. J.. Academic Press , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Goddard J. S. 1974 . An X-ray investigation of the effects of starvation and drugs on intestinal motility in the plaice, Pleuronectes platessa L . Ichthyologia , 6 : 49 – 58 . OpenURL Placeholder Text WorldCat Halver J. E. (Ed.) 1972 . Fish Nutrition . Academic Press , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Hasler A. D. 1971 . Orientation and fish migration. In Fish Physiology , vol. 6 , pp. 421 – 510 . Ed. by Hoar W. S., Randall D. J.. Academic Press , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Hoar W.S. , Randall D.J., Brett J.R. (Eds.) 1979 . Fish Physiology, Vol. 8 : Bioenergetics and Growth . Academic Press , New York Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Houlihan D. , Boujard T., Jobling M. (Eds.) 2001 . Food Intake in Fish. Blackwell Science , Oxford . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Huntingford F. , Jobling M., Kadri S. (eds). 2012 . Aquaculture and Behavior . Wiley-Blackwell , Oxford . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Jobling M. 1981 . The influences of feeding on the metabolic rate of fishes: a short review . Journal of Fish Biology , 18 : 385 – 400 . Google Scholar Crossref Search ADS WorldCat Jobling M. 1985 . Growth. In Fish Energetics: New Perspectives . pp. 213 – 230 . Ed. by Tytler P., Calow P.. Croom-Helm , London . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Jobling M. 1986 . Mythical models of gastric emptying and implications for food consumption . Environmental Biology of Fishes , 16 : 35 – 50 . Google Scholar Crossref Search ADS WorldCat Jobling M. 1988 . A review of the physiological and nutritional energetics of cod, Gadus morhua L., with particular reference to growth under farmed conditions . Aquaculture , 70 : 1 – 19 . Google Scholar Crossref Search ADS WorldCat Jobling M. 1994 . Fish Bioenergetics . Chapman & Hall , London . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Jobling M. 1995 . Environmental Biology of Fishes . Chapman & Hall , London . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Jobling M. 2004 . On-growing to market size. In Culture of Cold-water Marine Fish . pp. 363 – 432 . Ed. by Moksness E., Kjørsvik E., Olsen Y.. Blackwell Science , Oxford . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Jobling M. 2016 . Fish nutrition research: past, present and future . Aquaculture International , 24 : 767 – 786 . Google Scholar Crossref Search ADS WorldCat Jobling M. , Davies P. S. 1979 . Gastric evacuation in plaice, Pleuronectes platessa L.: effects of temperature and meal size . Journal of Fish Biology , 14 : 539 – 546 . Google Scholar Crossref Search ADS WorldCat Jobling M. , Davies P. S. 1980 . Effects of feeding on metabolic rate, and the Specific Dynamic Action in plaice, Pleuronectes platessa L . Journal of Fish Biology , 16 : 629 – 638 . Google Scholar Crossref Search ADS WorldCat Jobling M. , Breiby A. 1986 . The use and abuse of fish otoliths in studies of feeding habits of marine piscivores . Sarsia , 71 : 265 – 274 . Google Scholar Crossref Search ADS WorldCat Jobling M. , Koskela J. 1996 . Interindividual variations in feeding and growth of rainbow trout during restricted feeding and in a subsequent period of compensatory growth . Journal of Fish Biology , 49 : 658 – 667 . Google Scholar Crossref Search ADS WorldCat Jobling M. , Gwyther D., Grove D. J. 1977 . Some effects of temperature, meal size and body weight on gastric evacuation time in the dab, Limanda limanda (L.) . Journal of Fish Biology , 10 : 291 – 298 . Google Scholar Crossref Search ADS WorldCat Jobling M. , Baardvik B. M., Christiansen J. S., Jørgensen E. H. 1993 . The effects of prolonged exercise training on growth performance and production parameters in fish . Aquaculture International , 1 : 95 – 111 . Google Scholar Crossref Search ADS WorldCat Jobling M. , Meløy O. H., dos Santos J., Christiansen B. 1994 . The compensatory growth response of the Atlantic cod: effects of nutritional history . Aquaculture International , 2 : 75 – 90 . Google Scholar Crossref Search ADS WorldCat Jobling M. , Arnesen A. M., Baardvik B. M., Christiansen J. S., Jørgensen E. H. 1995 . Monitoring feeding behavior and food intake: methods and applications . Aquaculture Nutrition , 1 : 131 – 143 . Google Scholar Crossref Search ADS WorldCat Johansen S. J. S. , Ekli M., Stangnes B., Jobling M. 2001 . Weight gain and lipid deposition in Atlantic salmon, Salmo salar, during compensatory growth: evidence for lipostatic regulation? . Aquaculture Research , 32 : 963 – 974 . Google Scholar Crossref Search ADS WorldCat Kleiber M. 1961 . The Fire of Life: An Introduction to Animal Energetics . John Wiley & Sons , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Krogh A. 1916 . The Respiratory Exchange of Animals and Man . Longmans , London . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Le Francois N. , Jobling M., Carter C., Blier P. (Eds.) 2010 . Finfish Aquaculture Diversification . CABI , Wallingford . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Maynard L. A. , Loosli J. K. 1969 . Animal Nutrition , 6th edn. McGraw-Hill , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Miglavs I. , Jobling M. 1989 . Effects of feeding regime on food consumption, growth rates and tissue nucleic acids in juvenile Arctic charr, Salvelinus alpinus, with particular respect to compensatory growth . Journal of Fish Biology , 34 : 947 – 957 . Google Scholar Crossref Search ADS WorldCat Molnar G. Y. , Tamassy E., Tölg I. 1967 . The gastric digestion of living predatory fish. In The Biological Basis of Freshwater Fish Production . pp. 135 – 149 . Ed. by Gerking S. D.. Blackwell Scientific , Oxford . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Phillips A. M. Jr . 1969 . Nutrition, digestion and energy utilization. In Fish Physiology , vol. 1 , pp. 391 – 432 . Ed. by Hoar W. S., Randall D. J.. Academic Press , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Rozin P. , Mayer J. 1961 . Regulation of food intake in the goldfish . American Journal of Physiology , 201 : 968 – 974 . Google Scholar Crossref Search ADS PubMed WorldCat Rozin P. , Mayer J. 1964 . Some factors influencing short-term food intake of the goldfish . American Journal of Physiology , 206 : 1430 – 1436 . Google Scholar Crossref Search ADS PubMed WorldCat Sæther B. S. , Siikavuopio S. I., Jobling M. 2016 . Environmental conditions required for intensive farming of Arctic charr (Salvelinus alpinus (L.)) . Hydrobiologia , 783 : 347 – 359 . Google Scholar Crossref Search ADS WorldCat Sedgwick S. D. 1988 . Salmon Farming Handbook . Fishing News Books , Farnham . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Talbot C. , Higgins P. J. 1983 . A radiographic method for feeding studies on fish using metallic iron powder as a marker . Journal of Fish Biology , 23 : 211 – 220 . Google Scholar Crossref Search ADS WorldCat Winberg G. G. 1956 . Rate of metabolism and food requirements of fish . Fisheries Research Board of Canada Translation Series , 194 : 253 . OpenURL Placeholder Text WorldCat Author notes " Food for Thought articles are essays in which the author provides their perspective on a research area, topic, or issue. They are intended to provide contributors with a forum through which to air their own views and experiences, with few of the constraints that govern standard research articles. This Food for Thought article is one in a series solicited from leading figures in the fisheries and aquatic sciences community. The objective is to offer lessons and insights from their careers in an accessible and pedagogical form from which the community, and particularly early career scientists, will benefit. © International Council for the Exploration of the Sea 2017. All rights reserved. For Permissions, please email: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Valderrama,, Diego;Fields, KathrynAnn, H.

doi: 10.1093/icesjms/fsw223pmid: N/A

Given its ability to yield predictions for very diverse phenomena based only on two parameters—body size and temperature—the Metabolic Theory of Ecology (MTE) has earned a prominent place among ecology’s efficient theories. In a seminal article, the leading proponents of the MTE claimed that the theory was supported by evidence from Pauly’s (On the interrelationships between natural mortality, growth parameters, and mean environmental temperature in 175 fish stocks. Journal Du Conseil International Pour L’Exploration de la mer 39:175–192) dataset on natural mortality, biomass, and environmental temperature for 175 fish stocks spanning tropical, temperate, and polar locations. We demonstrate that the evidence presented by the proponents of the MTE is flawed because it fails to account for the fact that Pauly re-estimated environmental temperatures for polar fish as ‘physiologically effective temperatures’ to correct for their ‘abnormally’ high natural (mass-corrected) mortalities, which on average turned out to be similar to (rather than lower than) the mortalities recorded for temperate fish. Failing to account for these modifications skews the coefficients from MTE regression models and wrongly validates predictions from the theory. It is important to point out these deficiencies given the broad appeal of the MTE as a theoretical framework for applied ecological research. In a recent application, the MTE was used to estimate biomass production rates of prey fish in a model of invasive Indo-Pacific lionfish (Pterois volitans and P. miles) predation in Bahamian reefs. We show that the MTE coefficients may lead to a drastic overestimation of prey fish mortality and productivity rates, leading to erroneous estimations of target densities for ecological control of lionfish stocks. A set of robust mortality-weight coefficients is proposed as an alternative to the MTE. Introduction Building on the seminal works by Arrhenius (1889) on the effect of temperature on biochemical reactions and by Kleiber (1932) on the scaling of metabolic rate with animal size, Brown et al. (2004) set forth a metabolic theory of ecology (MTE) with the underlying premise that the metabolic rate of organisms is the fundamental biological rate that governs most observed patterns in ecology. The authors compiled numerous examples from the ecological literature to support the concept that metabolic rate controls ecological processes at all levels of organization from individuals to the biosphere, including life history attributes (e.g. population growth rate, mortality rate, age at maturity), population interactions (e.g. carrying capacity, rates of competition and predation), and ecosystem processes (e.g. patterns of trophic dynamics). Given its ability to yield predictions for very diverse phenomena based only on two parameters—body size and temperature—the MTE rapidly emerged as a prominent example of an efficient theory in ecology, i.e. the MTE is a theory that ‘… is grounded in first principles, it is expressed in the language of mathematics, it makes few assumptions and generates a large number of predictions per free parameter, it is approximate, and it entails predictions that provide well-understood standards for comparison with empirical data’ (Marquet et al., 2014). A large body of empirical data has been assembled to support the notion that size and temperature constrain diverse rate processes, including DNA evolution (e.g. Gillooly et al., 2005), population growth (Savage et al., 2004), and ecosystem carbon flux (Enquist et al., 2003), through their effects on metabolic rate. The MTE has also found applicability in fisheries science where it has been used to examine a variety of processes such as nutrient recycling rates in marine food webs (Allgeier et al., 2015); abundance, production, and size-structure of world fish stocks (Jennings and Collingridge, 2015); and the impacts of climate change on marine ecosystems (Barneche et al., 2014), among others. In a recent application, Green et al. (2014) used the MTE to compute the rate of prey fish production in a model of Indo-Pacific lionfish (Pterois volitans and P. miles) predation in invaded Bahamian coral reefs. The authors calculated biomass production P (g year − 1) by each individual lionfish prey as P=ZM,(1) where Z and M are the instantaneous mortality rate and body mass in g, respectively. According to the MTE, the mortality rate Z scales as an allometric function of body mass M with constants j and q ⁠, which approximates the ratio of production rate in grams per hectare per year to standing biomass in grams per hectare, such that Z≈PM=jMqeEkT.(2) The term eE/kT describes the effect of environmental temperature on prey fish production rates, where E is the activation energy (0.65 eV), k is Boltzmann’s constant (8.617 × 10 − 5 eV K − 1), and T is ambient water temperature (in degrees Kelvin). As pointed out by Valderrama and Fields (2015), Green et al. (2014) assumed an incorrect value of 3.08 for constant j ⁠, which was estimated by Lorenzen (1996) in the context of the simpler model Z=jMq that omits the correcting factor eE/kT ⁠. In their response, Green et al. (2015) replaced 3.08 with e26.25 as estimated in the regression model of fish mortality rates presented in Figure 4b of Brown et al. (2004), which was based on data from a study by Pauly (1980) examining the interrelationships between natural mortality, growth parameters, and ambient temperature in 175 fish stocks spanning tropical, temperate, and polar locations. We demonstrate here that the coefficients estimated by Brown et al. (2004) are flawed because the authors failed to account for the fact that Pauly replaced actual environmental temperatures for polar fish (those occurring in waters of ≤3.5 °C mean annual temperature) with ‘physiologically effective temperatures’ to compensate for the relatively high mortality rates observed for polar fish in his dataset. According to Pauly, these mortality rates were more characteristic of temperate fish and therefore suggested a metabolic adaptation to colder waters by polar fish. The most important implication of this finding is that Pauly’s mortality data fail to provide sufficiently strong evidence in favour of the MTE. As such, we recommend that fishery scientists carefully examine the empirical evidence that has been presented in support of the MTE prior to using the theory as a framework for the estimation of life history attributes such as natural mortality rates. Survival and mortality in the MTE Although ecologists have traditionally viewed survival times and their inverse, mortality rates, as being highly variable and consequences of extrinsic environmental conditions rather than intrinsic properties of individual organisms, the fact that most populations are neither continuously increasing nor decreasing is an indication that mortality rates must very nearly equal fecundity rates, and fecundity is fuelled by biomass production (Brown et al., 2004). Therefore, proponents of the MTE argue that the theory can be used to predict much of the variation in field mortality rates Z ⁠. According to Equation (7) in Brown et al. (2004), the following proportionality between mortality rate Z and body size M must hold: Z∝M-1/4eE/kT,(3) where E and k and all other variables are as defined previously. If Equation (3) is fundamentally correct, then a linear regression of the natural log of ZM1/4 (the mass-corrected mortality rate) against 1/kT would yield a slope coefficient equivalent to E (hypothesized to range from 0.60 to 0.70 eV), and an intercept equivalent to the natural log of j in Equation (2). Logically, the intercept corresponds to the proportionality constant implied in Equation (3). Using the fish mortality data from Pauly (1980), Brown et al. (2004) found that the slope of the size-corrected relationship between mortality rate and temperature gives an activation energy of 0.45 eV (95% CI, 0.37–0.54), which is somewhat lower than the predicted range of 0.60–0.70 eV. Pauly’s study is an examination of the interrelationships between natural mortality, growth parameters (from the von Bertalanffy growth model), and mean environmental temperature in 175 different stocks of fish distributed in 84 species, both freshwater and marine, and ranging from tropical to polar waters. The dataset provides a suitable empirical context to test the predictions from the MTE. Pauly included the entire dataset in the original article (Table 1), allowing the regression model of Brown et al. (2004) to be easily replicated. An important caveat about Table 1 in Pauly (1980) is that the temperatures listed for the 22 stocks of polar fish (those occurring in waters of ≤3.5 °C mean annual temperature) are the ‘physiologically effective temperatures’ resulting from modifying the actual mean annual temperature data in order to account for the ‘abnormally high’ [sic] natural mortalities observed for polar fish, a phenomenon suggestive of the process of metabolic cold adaptation (the hypothesis that polar fish show a resting metabolic rate higher than predicted from the overall rate/temperature relationship established for temperate and tropical species) proposed by a number of authors (Scholander et al., 1953; Wohlschlag, 1960) (The notion that higher natural mortality rates are associated with an adaptation to cold temperatures seems counter-intuitive at first but it is a correct one. Rather than referring to the enhanced ability to survive under strenuous environmental conditions, metabolic cold adaptation refers to an increase in the rate of metabolic processes under these strenuous conditions, including respiration, growth, egg production, and mortality). Table 1 Estimation of lionfish target densities in four natural patch reefs (numbers 55, 70, 71, 72) off Eleuthera Island, Bahamas, based on mean prey fish production (g m − 2 year − 1) and average lionfish sizes and densities as reported by Green et al. (2014, 2015). Item . Sites . j=e16.637 . j=e26.25 . 55 . 70 . 71 . 72 . 55 . 70 . 71 . 72 . Prey fish production, P-(g m−2 year−1) 9.69 × 10−4 1.03 × 10−3 9.23 × 10−4 8.56 × 10−4 14.5 15.4 13.8 12.8  Average prey size (g) 1.47  q, E, k, T q = −0.25; E = 0.65 eV; k = 8.617 × 10−5eV·K−1; T = 299.25 K  Z 1.72 × 10−4 2.58  P (g fish−1 year−1) 2.53 × 10−4 3.79  Prey fish density (fish m−2) 3.83 4.07 3.64 3.38 3.83 4.07 3.64 3.38 Prey consumption by lionfish, C-  Average lionfish size (cm) 23 21 16 20 23 21 16 20  Average lionfish size (g) 151 112 46 95 151 112 46 95  Lionfish density (fish reef−1) 18 48 20 33 18 48 20 33  Lionfish density (fish ha−1) 1 440 3 840 1 600 2 640 1 440 3 840 1 600 2 640  p, h p = 0.89; h = −0.29  Consumption (g fish−1 day−1) 17.5 14.2 7.5 12.7 17.5 14.2 7.5 12.7  C- (g m−2 year−1) 923 1 989 439 1 220 923 1 989 439 1 220 Target density (lionfish reef−1) 1.89 × 10−5 2.48 × 10−5 4.20 × 10−5 2.31 × 10−5 0.28 0.37 0.63 0.35 Target density (lionfish ha−1) 1.51 × 10−3 1.99 × 10−3 3.36 × 10−3 1.85 × 10−3 23 30 50 28 Item . Sites . j=e16.637 . j=e26.25 . 55 . 70 . 71 . 72 . 55 . 70 . 71 . 72 . Prey fish production, P-(g m−2 year−1) 9.69 × 10−4 1.03 × 10−3 9.23 × 10−4 8.56 × 10−4 14.5 15.4 13.8 12.8  Average prey size (g) 1.47  q, E, k, T q = −0.25; E = 0.65 eV; k = 8.617 × 10−5eV·K−1; T = 299.25 K  Z 1.72 × 10−4 2.58  P (g fish−1 year−1) 2.53 × 10−4 3.79  Prey fish density (fish m−2) 3.83 4.07 3.64 3.38 3.83 4.07 3.64 3.38 Prey consumption by lionfish, C-  Average lionfish size (cm) 23 21 16 20 23 21 16 20  Average lionfish size (g) 151 112 46 95 151 112 46 95  Lionfish density (fish reef−1) 18 48 20 33 18 48 20 33  Lionfish density (fish ha−1) 1 440 3 840 1 600 2 640 1 440 3 840 1 600 2 640  p, h p = 0.89; h = −0.29  Consumption (g fish−1 day−1) 17.5 14.2 7.5 12.7 17.5 14.2 7.5 12.7  C- (g m−2 year−1) 923 1 989 439 1 220 923 1 989 439 1 220 Target density (lionfish reef−1) 1.89 × 10−5 2.48 × 10−5 4.20 × 10−5 2.31 × 10−5 0.28 0.37 0.63 0.35 Target density (lionfish ha−1) 1.51 × 10−3 1.99 × 10−3 3.36 × 10−3 1.85 × 10−3 23 30 50 28 Item . Sites . j=e27.102 . 55 . 70 . 71 . 72 . Prey fish production, P-(g m−2 year−1) 34.0 36.1 32.4 30.0  Average prey size (g) 1.47  q, E, k, T  Z 6.04  P (g fish−1 year−1) 8.88  Prey fish density (fish m−2) 3.83 4.07 3.64 3.38 Prey consumption by lionfish, C-  Average lionfish size (cm) 23 21 16 20  Average lionfish size (g) 151 112 46 95  Lionfish density (fish reef−1) 18 48 20 33  Lionfish density (fish ha−1) 1 440 3 840 1 600 2 640  p, h p = 0.89; h = −0.29  Consumption (g fish−1 day−1) 17.5 14.2 7.5 12.7  C- (g m−2 year−1) 923 1 989 439 1 220 Target density (lionfish reef−1) 0.66 0.87 1.47 0.81 Target density (lionfish ha−1) 53 70 118 65 Item . Sites . j=e27.102 . 55 . 70 . 71 . 72 . Prey fish production, P-(g m−2 year−1) 34.0 36.1 32.4 30.0  Average prey size (g) 1.47  q, E, k, T  Z 6.04  P (g fish−1 year−1) 8.88  Prey fish density (fish m−2) 3.83 4.07 3.64 3.38 Prey consumption by lionfish, C-  Average lionfish size (cm) 23 21 16 20  Average lionfish size (g) 151 112 46 95  Lionfish density (fish reef−1) 18 48 20 33  Lionfish density (fish ha−1) 1 440 3 840 1 600 2 640  p, h p = 0.89; h = −0.29  Consumption (g fish−1 day−1) 17.5 14.2 7.5 12.7  C- (g m−2 year−1) 923 1 989 439 1 220 Target density (lionfish reef−1) 0.66 0.87 1.47 0.81 Target density (lionfish ha−1) 53 70 118 65 Three different values for constant j from the prey fish production model were assumed: j=e26.25 was obtained from Figure 4b in Brown et al. (2004); j=e16.637 and j=e27.102 are the estimates resulting from the regression models shown in Figures 1a and 2a (this article). The analysis assumes average prey size is 1.47 g across sites and the reef patches are 125 m2 in size. Table 1 Estimation of lionfish target densities in four natural patch reefs (numbers 55, 70, 71, 72) off Eleuthera Island, Bahamas, based on mean prey fish production (g m − 2 year − 1) and average lionfish sizes and densities as reported by Green et al. (2014, 2015). Item . Sites . j=e16.637 . j=e26.25 . 55 . 70 . 71 . 72 . 55 . 70 . 71 . 72 . Prey fish production, P-(g m−2 year−1) 9.69 × 10−4 1.03 × 10−3 9.23 × 10−4 8.56 × 10−4 14.5 15.4 13.8 12.8  Average prey size (g) 1.47  q, E, k, T q = −0.25; E = 0.65 eV; k = 8.617 × 10−5eV·K−1; T = 299.25 K  Z 1.72 × 10−4 2.58  P (g fish−1 year−1) 2.53 × 10−4 3.79  Prey fish density (fish m−2) 3.83 4.07 3.64 3.38 3.83 4.07 3.64 3.38 Prey consumption by lionfish, C-  Average lionfish size (cm) 23 21 16 20 23 21 16 20  Average lionfish size (g) 151 112 46 95 151 112 46 95  Lionfish density (fish reef−1) 18 48 20 33 18 48 20 33  Lionfish density (fish ha−1) 1 440 3 840 1 600 2 640 1 440 3 840 1 600 2 640  p, h p = 0.89; h = −0.29  Consumption (g fish−1 day−1) 17.5 14.2 7.5 12.7 17.5 14.2 7.5 12.7  C- (g m−2 year−1) 923 1 989 439 1 220 923 1 989 439 1 220 Target density (lionfish reef−1) 1.89 × 10−5 2.48 × 10−5 4.20 × 10−5 2.31 × 10−5 0.28 0.37 0.63 0.35 Target density (lionfish ha−1) 1.51 × 10−3 1.99 × 10−3 3.36 × 10−3 1.85 × 10−3 23 30 50 28 Item . Sites . j=e16.637 . j=e26.25 . 55 . 70 . 71 . 72 . 55 . 70 . 71 . 72 . Prey fish production, P-(g m−2 year−1) 9.69 × 10−4 1.03 × 10−3 9.23 × 10−4 8.56 × 10−4 14.5 15.4 13.8 12.8  Average prey size (g) 1.47  q, E, k, T q = −0.25; E = 0.65 eV; k = 8.617 × 10−5eV·K−1; T = 299.25 K  Z 1.72 × 10−4 2.58  P (g fish−1 year−1) 2.53 × 10−4 3.79  Prey fish density (fish m−2) 3.83 4.07 3.64 3.38 3.83 4.07 3.64 3.38 Prey consumption by lionfish, C-  Average lionfish size (cm) 23 21 16 20 23 21 16 20  Average lionfish size (g) 151 112 46 95 151 112 46 95  Lionfish density (fish reef−1) 18 48 20 33 18 48 20 33  Lionfish density (fish ha−1) 1 440 3 840 1 600 2 640 1 440 3 840 1 600 2 640  p, h p = 0.89; h = −0.29  Consumption (g fish−1 day−1) 17.5 14.2 7.5 12.7 17.5 14.2 7.5 12.7  C- (g m−2 year−1) 923 1 989 439 1 220 923 1 989 439 1 220 Target density (lionfish reef−1) 1.89 × 10−5 2.48 × 10−5 4.20 × 10−5 2.31 × 10−5 0.28 0.37 0.63 0.35 Target density (lionfish ha−1) 1.51 × 10−3 1.99 × 10−3 3.36 × 10−3 1.85 × 10−3 23 30 50 28 Item . Sites . j=e27.102 . 55 . 70 . 71 . 72 . Prey fish production, P-(g m−2 year−1) 34.0 36.1 32.4 30.0  Average prey size (g) 1.47  q, E, k, T  Z 6.04  P (g fish−1 year−1) 8.88  Prey fish density (fish m−2) 3.83 4.07 3.64 3.38 Prey consumption by lionfish, C-  Average lionfish size (cm) 23 21 16 20  Average lionfish size (g) 151 112 46 95  Lionfish density (fish reef−1) 18 48 20 33  Lionfish density (fish ha−1) 1 440 3 840 1 600 2 640  p, h p = 0.89; h = −0.29  Consumption (g fish−1 day−1) 17.5 14.2 7.5 12.7  C- (g m−2 year−1) 923 1 989 439 1 220 Target density (lionfish reef−1) 0.66 0.87 1.47 0.81 Target density (lionfish ha−1) 53 70 118 65 Item . Sites . j=e27.102 . 55 . 70 . 71 . 72 . Prey fish production, P-(g m−2 year−1) 34.0 36.1 32.4 30.0  Average prey size (g) 1.47  q, E, k, T  Z 6.04  P (g fish−1 year−1) 8.88  Prey fish density (fish m−2) 3.83 4.07 3.64 3.38 Prey consumption by lionfish, C-  Average lionfish size (cm) 23 21 16 20  Average lionfish size (g) 151 112 46 95  Lionfish density (fish reef−1) 18 48 20 33  Lionfish density (fish ha−1) 1 440 3 840 1 600 2 640  p, h p = 0.89; h = −0.29  Consumption (g fish−1 day−1) 17.5 14.2 7.5 12.7  C- (g m−2 year−1) 923 1 989 439 1 220 Target density (lionfish reef−1) 0.66 0.87 1.47 0.81 Target density (lionfish ha−1) 53 70 118 65 Three different values for constant j from the prey fish production model were assumed: j=e26.25 was obtained from Figure 4b in Brown et al. (2004); j=e16.637 and j=e27.102 are the estimates resulting from the regression models shown in Figures 1a and 2a (this article). The analysis assumes average prey size is 1.47 g across sites and the reef patches are 125 m2 in size. Table XVII in Pauly (1979) can be used to estimate the mean environmental temperatures corresponding to the ‘physiologically effective temperatures’ in Table 1 in Pauly (1980). Thus, ‘physiologically effective temperatures’ ranging from 24.0 to 4.0 °C correspond to environmental temperatures between -2.0 and 4.0 °C (A detailed explanation of the procedure used by Pauly 1979 to derive ‘physiologically effective temperatures’ for polar fish is provided in the online supplementary materials.). The Appendix 1 reproduces the data in Table 1 in Pauly (1980) with stocks arranged in descending order according to mean environmental temperature. For polar fish (≤3.5 °C), both the actual environmental and the physiologically active temperatures (enclosed in brackets) are listed. For these stocks, Table 1 in Pauly (1980) presents only the physiologically effective temperatures enclosed in parentheses. The major goal of the regression models in Pauly (1980) is to develop equations that provide highly reliable estimates of Z for any given fish stock, given the parameters of the von Bertalanffy growth formula and an estimate of the mean water temperature in which the stock in question lives. Because the goal is to obtain valid estimates of Z ⁠, using ‘physiologically effective temperatures’ for polar fish—rather than actual environmental temperatures—is warranted in Pauly’s analysis. However, because the purpose of Brown et al. (2004) is to establish a theoretically consistent relationship between mean environmental temperature and natural mortality Z ⁠, the regression models must be based on the original dataset of mean environmental temperatures rather than the ‘physiologically effective temperatures’. Failing to do so would distort the results of the analysis. Given the above caveat, we proceeded to replicate the linear regression of the natural log of ZM1/4 (the mass-corrected mortality rate) against 1/kT in order to obtain an estimate of the slope coefficient consistent with the theoretical expectation for E in Equation (3). Our results nevertheless produced an activation energy of 0.39 eV (95% CI, 0.31–0.46), which is lower than the 0.45 eV value reported by Brown et al. (2004), and much lower than the predicted range of 0.60–0.70 eV. The relationship between mass-corrected mortality rate and environmental temperature is graphically represented in Figure 1a. Figure 1 Open in new tabDownload slide Relationship between mass-corrected mortality rate, lnZM1/4 ⁠, measured in grams1/4 year − 1, and temperature, 1/kT ⁠, measured in K, based on fish mortality data from Pauly (1980). (a) Using the actual mean environmental temperature records for the subset of polar fish data. (b) Using the ´physiologically effective temperaturé data estimated for polar fish by Pauly (1980). Figure 1 Open in new tabDownload slide Relationship between mass-corrected mortality rate, lnZM1/4 ⁠, measured in grams1/4 year − 1, and temperature, 1/kT ⁠, measured in K, based on fish mortality data from Pauly (1980). (a) Using the actual mean environmental temperature records for the subset of polar fish data. (b) Using the ´physiologically effective temperaturé data estimated for polar fish by Pauly (1980). A comparison of Figure 1a with Figure 4a in Brown et al. (2004) would seem to indicate that polar fish data have been excluded in the latter as the horizontal axis only show observations for which 1/kT < 42; i.e. T > 3.16°. However, a closer examination of Figure 4a in Brown et al. (2004) reveals that, on the contrary, Brown et al. (2004) ran their models using Pauly’s ‘physiologically effective temperatures’ rather than the actual environmental temperatures. Thus, while the 22 temperature observations for polar fish ranged from 5 to 16 °C in Brown et al. (2004), the real environmental temperatures ranged from 3 to -1 °C. To cite just one example of the resulting inconsistencies, American plaice Hippoglossoides platessoides collected in St. Mary Bay, Nova Scotia (⁠ T = -1 °C) is assigned a T value of 15 °C in Brown et al. (2004), which is the same mean environmental temperature estimated for the California anchovy Engraulis mordax (see Appendix 1). Figure 1b recalculates the regression model using ‘physiologically effective’ rather than actual mean environmental temperatures. The estimated activation energy is 0.45 (95% CI, 0.36–0.53), matching the results reported by Brown et al. (2004): 0.45 eV (95% CI, 0.37–0.54). The slight discrepancy in confidence intervals may have resulted from typos in Pauly’s dataset (e.g. 78 g instead of 7.8 g in observation 136). An alternate way to test the validity of the MTE is to run a regression of temperature-corrected mortality rate (⁠ lnZeE/kT ⁠) against body mass (log-transformed data). The slope of the regression should approximate the predicted value of the allometric exponent, -¼. Brown et al. (2004) reported a slope of -0.24 (Figure 4b in their article); we found it to be -0.23 (95% CI, -0.19 to -0.27; Figure 2a). Once again, the -0.24 estimate is replicated when the mean environmental temperatures for polar fish are replaced with the corresponding ‘physiologically effective temperatures’ (Figure 2b). Figure 2 Open in new tabDownload slide Relationship between temperature-corrected mortality rate, lnZeE/kT ⁠, measured as year − 1, and body mass, ln⁡(M) ⁠, measured in grams, based on fish mortality data from Pauly (1980). (a) Using the actual mean environmental temperature records for the subset of polar fish data. (b) Using the ´physiologically effective temperaturé data estimated for polar fish by Pauly (1980). Figure 2 Open in new tabDownload slide Relationship between temperature-corrected mortality rate, lnZeE/kT ⁠, measured as year − 1, and body mass, ln⁡(M) ⁠, measured in grams, based on fish mortality data from Pauly (1980). (a) Using the actual mean environmental temperature records for the subset of polar fish data. (b) Using the ´physiologically effective temperaturé data estimated for polar fish by Pauly (1980). As mentioned previously, the value of constant j in Equation (2) can be computed as the antilog of the intercept in both regression models. If the MTE provides an accurate description of variability in mortality rates, the intercepts in each model should be relatively close to each other. Brown et al. (2004) report two different estimates for j ⁠: e19.24 and e26.25 (i.e. the antilogs of the intercepts in Figure 4a and 4b in Brown et al., 2004), which is to be expected given that the estimated value of the activation energy (-0.45 eV) is significantly lower than the predicted range of 0.60–0.70 eV. Notice that there is a discrepancy between the reported intercept in Figure 4b in Brown et al. (2004) and the graphical intercept shown in the same figure (about 19.2). Green et al. (2015) nevertheless selected e26.25as the empirically validated value of j for their estimation of natural mortality rates of lionfish prey. Because Brown et al. (2004) ran their models with the incorrect sets of temperature data for polar fish, the correct estimates of j resulting from Pauly’s dataset are e16.637 and e27.102 ⁠. They diverge widely because the dataset exhibits a great deal of variability that is not properly accounted for by the theoretical relationships in the MTE. This variability left unexplained by MTE models has been identified previously as a major weakness of the theory (e.g. Tilman et al., 2004; Price et al., 2012). Brown et al. (2004) themselves acknowledged the existence of this residual variation, which suggests that other variables and processes not considered in the general theory are important as well, a view that is shared by many other authors (e.g. Marquet et al., 2004; Enquist et al., 2007; Rüger and Condit, 2012). Although Brown et al. (2004) interpreted residual variations largely as departures from the theory’s predictions, authors such as Dickie et al. (1987) take a more critical view by arguing the existence of two distinct weight relationships: a physiological relationship that reflects the weight dependence of the metabolic rate, and an ecological relationship that reflects ecological factors such as spatial dynamics of predator–prey relationships. In his recent reviews of the evidence countering the predictions of ‘metabolic pacemaker’ theories such as the MTE, Glazier (2014, 2015) convincingly argued that a major weakness of these energy-based growth models is their failure to acknowledge that rates of metabolism and other biological processes are co-regulated by genetic, cellular, and neuroendocrine informational control systems. A truly comprehensive theory must take into consideration the combined influence of energy and information on biological activities, thus implying that growth is not simply determined by metabolic rate but jointly adjusted with it. For example, lifespan in animals is influenced by a number of factors that may be independent or not from body size, even if the general trend if for smaller species to have faster metabolic rates and shorter lifespans than larger species (Austad, 2010). Thus, the long lifespans of small species such as bats and birds can be partially explained by the reduced risk of predation associated with flight as an ecological adaptation. Similarly, although temperature increases are normally associated with increases in the rates of biological activities and underlying metabolic processes, it is also true that the adaptive responses of organisms may directly contradict the predictions of energy-based growth theories (Glazier, 2015). For example, contrary to the postulates of the MTE, the marine snail Echinolittorina malaccana decreases its metabolic rate as ambient temperature increases over the range 30–40 °C in order to conserve energy under warm conditions (Marshall and McQuaid, 2011). On the other hand, marine pteropods can increase their metabolic rate as ambient temperature declines, defying the predictions from the MTE (Seibel et al., 2007). This phenomenon is reminiscent of the metabolic cold adaptation studied by Scholander et al. (1953) and Wohlschlag (1960) in polar fish, which is consistent with the fact that size-corrected mortalities for polar fish (≤ 3.5 °C) in Pauly (1980) are not lower on average than the mortalities observed for temperate fish (3.5 °C ≤  T ≤ 22 °C). In fact, inspection of Figure 1a suggests that the relationship between mass-corrected mortality rate and 1/kT is better described by a curvilinear function (e.g. quadratic) as opposed to the linearity imposed by the MTE. The collective evidence presented by Glazier (2015) and other authors suggests that a major deficiency of the MTE is its assumption of a monotonic relationship between temperature and life history attributes. The studies on metabolic cold adaptation reviewed by Pauly (1979) clearly illustrate this point. Figure 3 depicts the relationship between the log of K—the growth rate in the von Bertalanffy growth function—and environmental temperature for Atlantic cod (Gadus morhua) as reported by May et al. (1965) (cold-adapted cod, ≤3.5 °C) and Taylor (1958) (4.5 °C <  T < 12 °C). While the data for temperate cod confirms the positive relationship suggested by the MTE, the relationship becomes negative for cold-adapted cod. Given the assumption of monotonicity, the MTE fails to capture this adaptation to the extent that an empirical quadratic function such as the curvilinear trend in Figure 3 does. Notice that this function can also be used to approximate ‘physiologically effective temperatures’: a cold-adapted cod growing at 2 °C would have a ‘physiologically effective temperature’ of 9.7 °C, for example. Figure 3 Open in new tabDownload slide Relationship between log of K (growth rate in the von Bertalanffy growth model) and environmental temperature for Atlantic cod (Gadus morhua) stocks. Data for cold-adapted cod (≤3.5 °C) were obtained from May et al. (1965); data for all other cod (4.5 °C <  T < 12 °C) are from Taylor (1958). Figure 3 Open in new tabDownload slide Relationship between log of K (growth rate in the von Bertalanffy growth model) and environmental temperature for Atlantic cod (Gadus morhua) stocks. Data for cold-adapted cod (≤3.5 °C) were obtained from May et al. (1965); data for all other cod (4.5 °C <  T < 12 °C) are from Taylor (1958). Given that the MTE is essentially a broad, monotonic, large-scale attempt to predict variations in the rates of a number of biological processes using only two explaining factors (body size and temperature), it can be used as a general predictive tool as long as it is recognized that it may oversimplify or even misrepresent the processes it is attempting to describe (Glazier, 2015). Therefore, it is important to point out the limitations of applications of the theory such as determination of target densities for invasive lionfish in the Bahamas reefs by Green et al. (2014). The following section provides further details on the study of Green et al. (2014), illustrating how estimates of target densities are strongly influenced by assumptions on the parameter values derived from the MTE. Determining target densities for invasive lionfish stocks in Bahamian coral reefs Green et al. (2014, 2015) estimated target densities for invasive lionfish stocks in 24 natural coral patch reefs off Eleuthera Island, Bahamas. The aggregated rate of biomass production of prey fish (⁠ P- ⁠), computed as the summation of individual Ps in Equation (1) over the assemblage of prey fish (per hectare of habitat), was compared with the rate of prey consumption (⁠ C- in g prey ha − 1 year − 1) by lionfish in each site. Thus, C-=d-W-p-9×10-22e0.16TW-hy,(4) where d- is the density of lionfish per hectare of habitat, W- is the mean body mass (in g) of lionfish, p- is the mean proportion of fish in the total diet of lionfish, and y is the number of days in a year. The function 9×10-22e0.16Tdescribes the scaling relationship between lionfish mass-specific prey consumption rate (g prey g lionfish − 1 d − 1) and body mass (g) derived from field studies of lionfish prey consumption at different water temperatures (Green et al., 2015). The scaling constant h has a value of -0.29 for lionfish (Côté and Green, 2012). The target density d-Target is then the threshold density at which prey consumption by lionfish (⁠ C- ⁠) equals the rate of prey fish biomass production (⁠ P- ⁠). The level of detail provided by Green et al. (2014, 2015) allowed us to re-estimate target lionfish densities (medians of distributions) for four sites (numbers 55, 70, 71, and 72) using different values for constant j in the prey fish production model. Columns 6 through 9 in Table 1 present the target density for each site under the assumption that j=e26.25 ⁠, which Green et al. (2015) selected based on the relationship between body mass and temperature-corrected mortality rate suggested by Brown et al. (2004). Columns 2 through 5 and 10 through 13 re-estimate target densities for the same sites assuming that j=e16.637and j=e27.102 ⁠, which are the two estimates of j that would have emerged from the Brown et al. (2004) regressions (log of mass-corrected mortality rate against 1/kT and log of temperature-corrected mortality rate against log of mass, respectively) if the correct set of environmental temperatures had been assumed for polar fish (Figures 1a and 2a). A number of simplifying assumptions was made for this analysis: average prey size was assumed to be 1.47 g (as reported by Green et al., 2011) whereas each reef patch was assumed to have an area of 125 m2 (Green et al., 2014 indicate that study reefs were 100–150 m2 in size). When j=e26.25 ⁠, average production rate of a prey fish (⁠ P ⁠) is 3.79 g fish − 1 year − 1 (see Equations 1 and 2). Given production rates per unit area reported by Green et al. (2015), prey fish densities for each patch reef can be inferred (between 3.4 and 4.1 fish m − 2; Table 1). On the other hand, consumption rates by lionfish are easily estimated via Equation (6), based on lionfish densities and average lionfish size reported for each site. The resulting target densities are listed in the last two rows of Table 1. In site 55, for example, a 151-g lionfish will consume approximately 6.4 kg of prey fish per year. Given that the reef produces around 1.8 kg of prey biomass per year, the target density is then 0.28 lionfish. Computed target densities varied from 0.28 to 0.63 lionfish per reef (23 to 50 fish ha − 1), representing a minuscule portion of the reported lionfish densities (from 0.8 to 3.1%). Based on the inferred prey fish densities, prey fish production rates (⁠ P- ⁠) and target densities can be re-estimated assuming that j=e16.637and j=e27.102 ⁠. Given the widely diverging estimates of j ⁠, dramatically different results are obtained in each case. When j=e16.637 ⁠, prey mortality Z and productivity P collapse, implying that all lionfish must be removed (target densities are essentially zero). When j=e27.102 ⁠, prey fish mortality Z and productivity P more than double relative to the j=e26.25 scenario, from 2.58 to 6.04 and from 3.79 to 8.88 g fish − 1 year − 1, respectively. This in turn results in much higher target densities, which increase by a factor of 2.3. It is clear then that estimates of prey fish productivity derived from the MTE are very sensitive to parameter assumptions, given the presence of the exponential term for constant j ⁠. Even an apparently minor variation in j from e26.25 to e27.102 more than doubles estimated prey fish mortality and productivity rates as well as lionfish target densities. As indicated by Valderrama and Fields (2015), a more sensible approach would be to replace the production-biomass ratio in Equation (2) with the simpler mortality–weight relationship Z=jMq,(5) which removes the temperature correction factor from Equation (2). Lorenzen (1996) estimated these relationships for a number of natural and artificial (i.e. aquaculture) ecosystems. For tropical fish, Lorenzen found that q = -0.21 (90% CI, -0.356 to -0.111) and j = 3.08 (90% CI, 1.87 to 4.48). Table 2 presents prey fish mortality rates and lionfish target densities re-estimated using Lorenzen’s coefficients. Ranging from 25 to 55 lionfish ha − 1, target densities are much lower than those suggested by the j=e27.102 estimate resulting from the MTE (by a factor of 2.1, approximately). Table 2 Estimation of lionfish target densities in four natural patch reefs (numbers 55, 70, 71, 72) off Eleuthera Island, Bahamas, based on prey fish densities and average lionfish sizes and densities as reported by Green et al. (2014, 2015). Item . Sites . 55 . 70 . 71 . 72 . Prey fish production, P-  Average prey size (g) 1.47  j 3.08  q -0.21  Z 2.84  P (g fish−1 year−1) 4.18  Prey fish density (fish m−2) 3.83 4.07 3.64 3.38  P- (g m−2 year−1) 16.0 17.0 15.2 14.1 Prey consumption by lionfish, C-  Average lionfish size (cm) 23 21 16 20  Average lionfish size (g) 151 112 46 95  Lionfish density (fish reef−1) 18 48 20 33  Lionfish density (fish ha−1) 1 440 3 840 1 600 2 640  p 0.89  h -0.29  Consumption (g fish−1 day−1) 17.5 14.2 7.5 12.7 C- (g m−2 year−1) 923 1 989 439 1 220 Target density (lionfish reef−1) 0.20 0.26 0.44 0.24 Target density (lionfish ha−1) 25 33 55 31 Item . Sites . 55 . 70 . 71 . 72 . Prey fish production, P-  Average prey size (g) 1.47  j 3.08  q -0.21  Z 2.84  P (g fish−1 year−1) 4.18  Prey fish density (fish m−2) 3.83 4.07 3.64 3.38  P- (g m−2 year−1) 16.0 17.0 15.2 14.1 Prey consumption by lionfish, C-  Average lionfish size (cm) 23 21 16 20  Average lionfish size (g) 151 112 46 95  Lionfish density (fish reef−1) 18 48 20 33  Lionfish density (fish ha−1) 1 440 3 840 1 600 2 640  p 0.89  h -0.29  Consumption (g fish−1 day−1) 17.5 14.2 7.5 12.7 C- (g m−2 year−1) 923 1 989 439 1 220 Target density (lionfish reef−1) 0.20 0.26 0.44 0.24 Target density (lionfish ha−1) 25 33 55 31 The constants j=3.08 and q=-0.21 empirically validated by Lorenzen (1996) for fish in tropical marine ecosystems were used to estimate prey fish mortalities and rates of biomass production. The analysis assumes average prey size is 1.47 g across sites and the reef patches are 125 m2 in size. Table 2 Estimation of lionfish target densities in four natural patch reefs (numbers 55, 70, 71, 72) off Eleuthera Island, Bahamas, based on prey fish densities and average lionfish sizes and densities as reported by Green et al. (2014, 2015). Item . Sites . 55 . 70 . 71 . 72 . Prey fish production, P-  Average prey size (g) 1.47  j 3.08  q -0.21  Z 2.84  P (g fish−1 year−1) 4.18  Prey fish density (fish m−2) 3.83 4.07 3.64 3.38  P- (g m−2 year−1) 16.0 17.0 15.2 14.1 Prey consumption by lionfish, C-  Average lionfish size (cm) 23 21 16 20  Average lionfish size (g) 151 112 46 95  Lionfish density (fish reef−1) 18 48 20 33  Lionfish density (fish ha−1) 1 440 3 840 1 600 2 640  p 0.89  h -0.29  Consumption (g fish−1 day−1) 17.5 14.2 7.5 12.7 C- (g m−2 year−1) 923 1 989 439 1 220 Target density (lionfish reef−1) 0.20 0.26 0.44 0.24 Target density (lionfish ha−1) 25 33 55 31 Item . Sites . 55 . 70 . 71 . 72 . Prey fish production, P-  Average prey size (g) 1.47  j 3.08  q -0.21  Z 2.84  P (g fish−1 year−1) 4.18  Prey fish density (fish m−2) 3.83 4.07 3.64 3.38  P- (g m−2 year−1) 16.0 17.0 15.2 14.1 Prey consumption by lionfish, C-  Average lionfish size (cm) 23 21 16 20  Average lionfish size (g) 151 112 46 95  Lionfish density (fish reef−1) 18 48 20 33  Lionfish density (fish ha−1) 1 440 3 840 1 600 2 640  p 0.89  h -0.29  Consumption (g fish−1 day−1) 17.5 14.2 7.5 12.7 C- (g m−2 year−1) 923 1 989 439 1 220 Target density (lionfish reef−1) 0.20 0.26 0.44 0.24 Target density (lionfish ha−1) 25 33 55 31 The constants j=3.08 and q=-0.21 empirically validated by Lorenzen (1996) for fish in tropical marine ecosystems were used to estimate prey fish mortalities and rates of biomass production. The analysis assumes average prey size is 1.47 g across sites and the reef patches are 125 m2 in size. The Mortality–weight model by Lorenzen (1996) generates much more stable predictions than the MTE. At the ecosystem level, he found no significant differences in parameters between lakes, rivers, and the ocean. A joint mortality–weight relationship for all natural ecosystems was estimated with parameters q = -0.288 (90% CI, -0.315 to -0.261) and j = 3.00 (90% CI, 2.70 to 3.30). These joint parameters result in negligible changes to the target densities presented in Table 2, with the re-estimated densities ranging from 24 to 52 lionfish ha − 1. This robustness makes Lorenzen’s model much more appropriate for management purposes than the temperature-corrected MTE. This section ends with a caveat regarding the target densities reported by Green et al. (2014, 2015). These targets are an order of magnitude higher than those presented in Table 1. It is unclear why Green et al. report such high densities given the relatively low productivity of prey fish biomass as compared with the high rates of lionfish predation. For example, prey fish production is around 1600 g reef − 1 year − 1 in site 72 (Green et al., 2015). Given that the average lionfish weighs 95 g (20-cm length), annual consumption per lionfish is 4622 g of prey, according to Equation (4). Hence, the target density must be less than one lionfish per reef (in other words, lionfish must be eradicated given their voracious consumption), not four lionfish as reported by Green et al. Conclusions Marquet et al. (2014) argue for expanding the role of efficient theories in ecology in order to accelerate scientific progress, enhance the ability to address environmental challenges, and foster the development of synthesis and unification in the discipline, among other goals. With its power to abstract and simplify some of the complexity of nature, the MTE is particularly appealing as an efficient theory of ecology. Nevertheless, the MTE has been criticized by its failure to capture the influence of ecological and informational factors—other than body mass and temperature—as determinants of metabolic rates. Brown et al. (2004) attempted to validate the MTE by establishing a relationship between natural mortality rate in fish populations and body weight and ambient temperature using Pauly (1980)’s dataset of 175 fishing stocks. Unfortunately, the data fail to validate the MTE once the regression models are run with the correct set of environmental temperatures (as opposed to physiologically efficient temperatures) for polar fish. Given the failure of the MTE to capture the variability in Pauly’s original dataset and the phenomenon of metabolic cold adaptation in polar fish, the coefficients resulting from the regressions of Brown et al. (2004) may lead to an overestimation of mortality and biomass production rates of prey fish in Green et al. (2014, 2015) model of invasive lionfish predation in Bahamian coral reefs. The study by Lorenzen (1996) on body weight and natural mortality in fish provides a more robust set of parameters by focusing on more homogeneous sets of observations, obviating the need to introduce the temperature correction associated with the MTE. Although Brown et al. (2004) envisioned a wide range of applications for their theory, they also warned about the risks of applying the MTE to practical problems of environmental policy and management (pp. 1787). Given the evidence presented in this study, we urge practitioners to carefully examine the suitability of the MTE prior to adopting the theory as a framework for applied ecological models in fisheries science. Funding Funding for this research was provided by the National Institute of Food and Agriculture, U.S. Department of Agriculture, Multistate Research Project W3004: Marketing, Trade, and Management of Aquaculture and Fishery Resources. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the authors and do not necessarily reflect the view of the U.S. Department of Agriculture. Supplementary data Supplementary material is available at the ICESJMS online version of the manuscript. Acknowledgements We are particularly grateful to two anonymous reviewers for providing insightful comments that significantly improved this article. References Allgeier J. E. , Wenger S. J., Rosemond A. D., Schindler D. E., Layman C. A. 2015 . Metabolic theory and taxonomic identity predict nutrient recycling in a diverse food web . Proceedings of the National Academy of Sciences , 112 : E2640 – E2647 . Google Scholar Crossref Search ADS WorldCat Arrhenius S. 1889 . Über die reaktionsgeschwindigkeit bei der inversion von rohrzucker durch säuren . Zeitschrift für Physikalische Chemie , 4 : 226 – 248 . OpenURL Placeholder Text WorldCat Austad S. N. 2010 . Animal size, metabolic rate, and survival, among and within species. In The Comparative Biology of Aging , pp. 27–41. Ed. by Wolf N. S.. Springer , Dordrecht . 391 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Barneche D. R. , Kulbicki M., Floeter S. R., Friedlander A. M., Maina J., Allen A. P. 2014 . Scaling metabolism from individuals to reef-fish communities at broad spatial scales . Ecology Letters , 17 : 1067 – 1076 . Google Scholar Crossref Search ADS PubMed WorldCat Brown J. H. , Gillooly J. F., Allen A. P., Savage V. M., West G. B. 2004 . Toward a metabolic theory of ecology . Ecology , 85 : 1771 – 1789 . Google Scholar Crossref Search ADS WorldCat Côté I. M. , Green S. J. 2012 . Potential effects of climate change on a marine invasion: The importance of current context . Current Zoology , 58 : 1 – 8 . Google Scholar Crossref Search ADS WorldCat Dickie L. M. , Kerr S. R., Boudreau P. R. 1987 . Size-dependent processes underlying regularities in ecosystem structure . Ecological Monographs , 57 : 233 – 250 . Google Scholar Crossref Search ADS WorldCat Enquist B. J. , Economo E. P., Huxman T. E., Allen A. P., Ignace D. D., Gillooly J. F. 2003 . Scaling metabolism from organisms to ecosystems . Nature , 423 : 639 – 642 . Google Scholar Crossref Search ADS PubMed WorldCat Enquist B. J. , Kerkhoff A. J., Stark S. C., Swenson N. G., McCarthy M. C., Price C. A. 2007 . A general integrative model for scaling plant growth, carbon flux, and functional trait spectra . Nature , 449 : 218 – 222 . Google Scholar Crossref Search ADS PubMed WorldCat Gillooly J. F. , Allen A. P., West G. B., Brown J. H. 2005 . The rate of DNA evolution: Effects of body size and temperature on the molecular clock . Proceedings of the National Academy of Sciences , 102 : 140 – 145 . Google Scholar Crossref Search ADS WorldCat Glazier D. S. 2014 . Metabolic scaling in complex living systems . Systems , 2 : 451 – 540 . Google Scholar Crossref Search ADS WorldCat Glazier D. S. 2015 . Is metabolic rate a universal ‘pacemaker’ for biological processes? Biological Reviews , 90 : 377 – 407 . Google Scholar Crossref Search ADS PubMed WorldCat Green S. J. , Akins J. L., Côté I. M. 2011 . Foraging behaviour and prey consumption in the Indo-Pacific lionfish on Bahamian coral reefs . Marine Ecology Progress Series , 433 : 159 – 167 . Google Scholar Crossref Search ADS WorldCat Green S. J. , Dulvy N. K., Brooks A. M. L., Akins J. L., Cooper A. B., Miller S., Côté I. M. 2014 . Linking removal targets to the ecological effects of invaders: a predictive model and field test . Ecological Applications , 24 : 1311 – 1322 . Google Scholar Crossref Search ADS PubMed WorldCat Green S. J. , Dulvy N. K., Brooks A. M. L., Akins J. L., Cooper A. B., Miller S., Côté I. M. 2015 . Response to Valderrama and fields: effect of temperature on biomass production in models of invasive lionfish control . Ecological Applications , 25 : 2048 – 2050 . Google Scholar Crossref Search ADS PubMed WorldCat Jennings S. , Collingridge K. 2015 . Predicting consumer biomass, size-structure, production, catch potential, responses to fishing and associated uncertainties in the world’s marine ecosystems . PLOS One , 10 : e0133794 . Google Scholar Crossref Search ADS PubMed WorldCat Kleiber M. 1932 . Body size and metabolism . Hilgardia , 6 : 315 – 332 . Google Scholar Crossref Search ADS WorldCat Lorenzen K. 1996 . The relationship between body weight and natural mortality in juvenile and adult fish: a comparison of natural ecosystems and aquaculture . Journal of Fish Biology , 49 : 627 – 647 . Google Scholar Crossref Search ADS WorldCat Marquet P. A. , Labra F. A., Maurer B. A. 2004 . Metabolic ecology: linking individuals to ecosystems . Ecology , 85 : 1794 – 1796 . Google Scholar Crossref Search ADS WorldCat Marquet P. A. , Allen A. P., Brown J. H., Dunne J. A., Enquist B. J., Gillooly J. F., Gowaty P. A., et al. 2014 . On theory in ecology . BioScience , 64 : 701 – 710 . Google Scholar Crossref Search ADS WorldCat Marshall D. J. , McQuaid C. D. 2011 . Warming reduces metabolic rate in marine snails: adaptation to fluctuating high temperatures challenges the metabolic theory of ecology . Proceedings of the Royal Society of London Series B , 278 : 281 – 288 . Google Scholar Crossref Search ADS PubMed WorldCat May A. W. , Pinhorn A. T., Wells R., Fleming A. M. 1965 . Cod growth and temperature in the Newfoundland area . 1965. ICNAF Special Publications , 6 : 545 – 555 . OpenURL Placeholder Text WorldCat Pauly D. 1979 . Gill size and temperature as governing factors in fish growth: a generalization of Von Bertalannfýs growth formula. Berichte aus dem Institut Für Meereskunde Nr. 63. Christian Albrechts Universität, Kiel, Germany. 156 pp. Pauly D. 1980 . On the interrelationships between natural mortality, growth parameters, and mean environmental temperature in 175 fish stocks . Journal du Conseil International pour l’Exploration de la Mer , 39 : 175 – 192 . Google Scholar Crossref Search ADS WorldCat Price C. A. , Weitz J. S., Savage V. M., Stegen J., Clarke A., Coomes D. A., Dodds P. S., et al. 2012 . Testing the metabolic theory of ecology . Ecology Letters , 15 : 1465 – 1474 . Google Scholar Crossref Search ADS PubMed WorldCat Rüger N. , Condit R. 2012 . Testing metabolic theory with models of tree growth that include light competition . Functional Ecology , 26 : 759 – 765 . Google Scholar Crossref Search ADS WorldCat Savage V. M. , Gillooly J. F., Brown J. H., West G. B., Charnov E. L. 2004 . Effects of body size and temperature on population growth . American Naturalist , 163 : 429 – 441 . Google Scholar Crossref Search ADS PubMed WorldCat Scholander P. F. , Flagg W., Walters V., Irving L. 1953 . Climatic adaptation in arctic and tropical poikilotherms . Physiological Zoology , 26 : 67 – 92 . Google Scholar Crossref Search ADS WorldCat Seibel B. A. , Dymowska A., Rosenthal J. 2007 . Metabolic temperature compensation and coevolution of locomotory performance in pteropod molluscs . Integrative and Comparative Biology , 47 : 880 – 891 . Google Scholar Crossref Search ADS PubMed WorldCat Taylor C. C. 1958 . Cod growth and temperature . Journal du Conseil , 23 : 366 – 370 . Google Scholar Crossref Search ADS WorldCat Tilman D. , HilleRisLambers J., Harpole S., Dybzinski R., Fargione J., Clark C., Lehman C. 2004 . Does metabolic theory apply to community ecology? It’s a matter of scale . Ecology , 85 : 1797 – 1799 . Google Scholar Crossref Search ADS WorldCat Valderrama D. , Fields K. H. 2015 . Linking removal targets to the ecological effects of invaders: a predictive model and field test – Letter to the Editor . Ecological Applications , 25 : 2047 – 2048 . Google Scholar Crossref Search ADS PubMed WorldCat Wohlschlag D. E. 1960 . Metabolism of an Antarctic fish and the phenomenon of cold adaptation . Ecology , 41 : 287 – 292 . Google Scholar Crossref Search ADS WorldCat Appendix 1 Table A1 Data on growth, temperature and fish mortality used in Pauly (1980) analysis. Record number . Family . Species . T (°C) . M (g) . Z . 101 Leiognathidae Liognathus splendens 28 64.0 1.80 102 Nemipteridae Nemipterus bleekeri 28 300.0 1.12 103 Nemipteridae Nemipterus delagoae 28 240.0 1.62 104 Nemipteridae Nemipterus hexodon 28 215.0 0.88 105 Nemipteridae Nemipterus japonicas 28 255.0 1.88 106 Nemipteridae Nemipterus marginatus 28 230.0 1.14 107 Nemipteridae Nemipterus marginatus 28 210.0 1.73 108 Nemipteridae Nemipterus mesoprion 28 75.0 1.08 109 Nemipteridae Nemipterus nematophorus 28 200.0 1.63 110 Nemipteridae Nemipterus nemurus 28 245.0 1.03 111 Nemipteridae Nemipterus peronei 28 257.0 1.05 112 Nemipteridae Nemipterus tolu 28 230.0 0.41 113 Nemipteridae Nemipterus sp. 28 406.0 0.53 144 Scombridae Rastrelliger kanagurta 28 160.0 4.44 146 Scombridae Rastrelliger neglectus 28 110.0 7.22 147 Scombridae Rastrelliger neglectus 28 84.0 4.20 148 Scombridae Rastrelliger neglectus 28 205.0 4.56 15 Clupeidae Sardinella longiceps 27 66.0 1.12 16 Clupeidae Sardinella longiceps 27 63.0 0.67 22 Engraulidae Stolothrissa tanganicae 27 6.0 5.20 84 Serranidae Cephalopholis fulva 27 640.0 0.55 86 Serranidae Epinephelus guttatus 27 1 880.0 0.68 87 Serranidae Epinephelus striatus 27 12 900.0 0.24 88 Serranidae Mycteroperca venenosa 27 8 330.0 0.42 92 Carangidae Caranx rubber 27 2 490.0 1.51 94 Lutjanidae Apsilus dentatus ♀ 27 4 000.0 0.83 95 Lutjanidae Apsilus dentatus ♂ 27 4 500.0 1.90 96 Lutjanidae Lutjanus apodus 27 3 800.0 0.54 97 Lutjanidae Lutjaus buccanella ♀ 27 3 200.0 2.24 98 Lutjanidae Lutjanus buccanella ♂ 27 1 890.0 1.83 99 Lutjanidae Lutjanus purpureus 27 11 664.0 0.37 100 Lutjanidae Ocyurus chrysurus 27 3 600.0 0.62 114 Pomadasyidae Haemulon album 27 5 300.0 0.79 115 Pomadasyidae Haemulon plumieri 27 1 360.0 1.77 117 Sciaenidae Cynoscion macdonaldi 27 17 826.0 0.30 119 Sciaenidae Pseudosciaena diacanthus 27 17 400.0 0.80 121 Lethrinidae Lethrinus enigmaticus 27 2 916.0 0.20 122 Mullidae Mulloidichthys martinicus ♀ 27 520.0 1.70 123 Mullidae Mulloidichthys martinicus ♂ 27 360.0 1.73 124 Mullidae Pseudoupeneus macularus 27 360.0 1.89 145 Scombridae Rastrelliger kanagurta 27 23.9 7.80 174 Cynoglossidae Cynoglossus macrolepidus 27 170.0 0.49 175 Balistidae Balistes vetula 27 2 920.0 2.61 120 Sciaenidae Pseudotolithus elongatus 26 715.0 0.34 125 Cichlidae Haplochromis anaphyrmus 26 111.0 1.45 126 Cichlidae Haplochromis mloto 26 74.0 0.92 127 Cichlidae Lethrinops longispinis 26 122.0 1.69 128 Cichlidae Lethrinops parvidens 26 132.0 1.20 150 Scombridae Thunnus alalunga 26 5 150.0 0.20 154 Scombridae Thunnus albacares 26 137 500.0 0.80 155 Scombridae Thunnus atlanticus 26 9 760.0 0.67 157 Scombridae Thunnus maccoyi 26 195 660.0 0.20 129 Cichlidae Tilapia esculenta 25 750.0 1.75 141 Scombridae Katsuwonus pelamis 25 32 300.0 1.68 142 Scombridae Katsuwonus pelamis 25 10 130.0 0.65 149 Scombridae Thunnus alalunga 25 58 760.0 0.23 151 Scombridae Thunnus albacares 25 144 400.0 0.80 152 Scombridae Thunnus albacares 25 199 000.0 0.90 153 Scombridae Thunnus albacares 25 98 790.0 0.77 156 Scombridae Thunnus germo 25 36 900.0 0.22 23 Engraulidae Centengraulis mysticetus 24 56.0 2.40 118 Sciaenidae Cynoscion nobilis 24 27 900.0 0.30 6 Clupeidae Brevoortia tyrannus 23 141.0 1.00 5 Acipenseridae Acipenser transmontanus 22 170 000.0 0.03 69 Merluccidae Merluccius angustimanus 20 255.0 0.84 79 Mugilidae Mugil cephalus ♂ 20 2 450.0 0.31 93 Carangidae Trachurus japonicus 20 1 188.0 0.99 116 Sciaenidae Cynoscion analis 18 1 466.0 0.40 46 Synodontidae Saurida tumbil 17 2 350.0 0.46 132 Blennidae Blennius pholis 17 54.0 0.90 24 Engraulidae Engraulis anchoita 16 212.0 1.42 25 Engraulidae Engraulis anchoita 16 50.0 0.90 29 Engraulidae Engraulis ringens 16 37.0 1.00 30 Engraulidae Engraulis ringens 16 24.0 1.52 42 Engraulidae Osmeridae 16 15.0 2.45 17 Clupeidae Sardinops caeruea 15 337.0 0.40 18 Clupeidae Sardinops caeruea 15 225.0 0.45 19 Clupeidae Sardinops melanosticta 15 209.0 0.50 26 Engraulidae Engraulis encrasicholus 15 24.0 1.80 28 Engraulidae Engraulis mordax 15 21.0 1.70 49 Myctophidae Myctophum punctatum 15 6.6 0.73 71 Merluccidae Merluccius merluccius ♀ 15 1 577.0 0.50 72 Merluccidae Merluccius merluccius ♂ 15 622.0 0.60 77 Cyprinodontidae Aphanius fasciatus ♀ 15 4.0 1.80 78 Cyprinodontidae Aphanius fasciatus ♂ 15 2.5 2.60 143 Scombridae Pneumatophorus japonicus 15 810.0 0.90 70 Merluccidae Merluccius gayi 14 5 830.0 0.37 130 Cheilodactylidae Cheilodactylus macropterus 14 1 390.0 0.08 7 Clupeidae Brevoortia tyrannus 13 863.0 0.50 12 Clupeidae Clupea harengus 12 193.0 0.20 13 Clupeidae Clupea pallassii 12 525.0 0.20 52 Gadidae Gadus minutus ♀ 12 158.0 0.90 53 Gadidae Gadus minutus ♂ 12 59.4 1.10 73 Merluccidae Merluccius productus 12 1 272.0 0.56 74 Scomberesocidae Cololabis saira 12 189.0 1.60 135 Callionymidae Callionymus lyra ♀ 12 27.0 0.86 136 Callionymidae Callionymus lyra ♂ 12 78.0 0.96 8 Clupeidae Clupea harengus 11 200.0 0.25 133 Ammodytidae Ammodytes marinus 11 32.1 1.24 134 Ammodytidae Ammodytes tobianus 11 15.0 1.29 1 Lamnidae Cetorhinus maximus 10 13 820 000.0 0.046 4 Acipenseridae Acipenser transmontanus 10 96 680.0 0.05 14 Clupeidae Clupea pallassii 10 209.0 0.50 75 Gasterosteidae Gasterosteus aculeatus 10 1.8 0.90 76 Gasterosteidae Pungitius pungitius 10 0.4 1.10 85 Moronidae Dicentrarchus labrax 10 6 860.0 0.10 137 Gobiidae Pomatoschistus minutus 10 4.0 3.30 168 Pleuronectidae Limanda ferruginea 10 1 183.0 0.15 11 Clupeidae Clupea harengus 9 160.0 0.36 21 Clupeidae Sprattus sprattus 9 20.4 0.70 40 Salmonidae Salmo trutta 9 300.0 0.94 44 Esocidae Esox lucius ♀ 9 8 810.0 0.26 45 Esocidae Esox lucius ♂ 9 3 288.0 0.24 51 Cyprinidae Phoxinus phoxinus 9 8.2 1.10 67 Gadidae Trisopterus esmarkii 9 48.0 1.60 68 Gadidae Trisopterus esmarkii 9 45.0 1.60 80 Cottidae Cottus gobio ♀ 9 4.3 0.80 81 Cottidae Cottus gobio ♂ 9 4.3 1.10 82 Cottidae Cottus gobio ♀ 9 6.1 0.90 83 Cottidae Cottus gobio ♂ 9 5.9 1.10 158 Pleuronectidae Eopsetta jordani ♀ 9 3 970.0 0.20 159 Pleuronectidae Eopsetta jordani ♂ 9 4 566.0 0.25 3 Acipenseridae Acipenser fulvescens 8 36 000.0 0.01 9 Clupeidae Clupea harengus 8 350.0 0.16 10 Clupeidae Clupea harengus 8 55.0 0.35 20 Clupeidae Sprattus sprattus 8 23.0 0.53 35 Salmonidae Leucichthys artedi 8 70.0 1.10 36 Salmonidae Leucichthys artedi 8 95.0 1.20 37 Salmonidae Leucichthys artedi 8 600.0 0.35 50 Cyprinidae Abramis brama 8 1 593.0 0.165 57 Gadidae Gadus morhua 8 16 350.0 0.17 60 Gadidae Gadus morhua 8 20 000.0 0.20 138 Scorpaenidae Sebastes alutus 8 1 358.0 0.227 169 Pleuronectidae Platichthys flesus 8 1 040.0 0.18 2 Lamnidae Lamna nasus 7 166 800.0 0.18 47 Myctophidae Benthosema glaciale 7 5.6 1.75 48 Myctophidae Benthosema glaciale 7 5.7 0.74 56 Gadidae Gadus morhua 7 11 000.0 0.33 171 Pleuronectidae Pleuronectes platessa ♀ 7 3 430.0 0.12 172 Pleuronectidae Pleuronectes platessa ♂ 7 910.0 0.22 173 Soleidae Solea vulgaris 7 482.0 0.25 55 Gadidae Gadus morhua 6 15 000.0 0.10 62 Gadidae Gadus morhua 6 8 158.0 0.44 65 Gadidae Pollachius virens 6 11 634.0 0.30 89 Percidae Perca fluviatilis 6 435.0 0.29 90 Percidae Perca fluviatilis 6 633.0 0.16 38 Salmonidae Leucichthys sardinella 5 550.0 0.60 39 Salmonidae Salmo trutta 5 500.0 0.31 61 Gadidae Gadus morhua 5 10 834.0 0.31 63 Gadidae Melanogrammus aeglefinus 5 2 150.0 0.20 140 Scorpaenidae Sebastes alutus 5 800.0 0.193 170 Pleuronectidae Pseudopleuronectes americanus 5 1 380.0 0.40 31 Salmonidae Coregonus clupeaformis 3 [5] 2 826.0 0.15 32 Salmonidae Coregonus clupeaformis 3 [5] 6 215.0 0.17 33 Salmonidae Coregonus clupeaformis 3 [5] 4 200.0 1.35 34 Salmonidae Coregonus clupeaformis 3 [5] 27.0 1.30 54 Gadidae Gadus morhua 3 [5] 24 000.0 0.10 64 Gadidae Pollachius virens 3 [5] 20 077.0 0.30 66 Gadidae Pollachius virens 3 [5] 16 514.0 0.30 91 Percidae Stizostedion canadensis 3 [5] 615.0 0.44 58 Gadidae Gadus morhua 2 [6] 11 720.0 0.20 139 Scorpaenidae Sebastes alutus 2 [6] 1 081.0 0.271 41 Salmonidae Salvelinus alpinus 1 [7] 5 000.0 0.17 59 Gadidae Gadus morhua 1 [8] 3 000.0 0.18 166 Pleuronectidae Hippoglossoides platessoides ♀ 1 [7] 3 823.0 0.18 167 Pleuronectidae Hippoglossoides platessoides ♂ 1 [7] 2 035.0 0.26 43 Osmeridae Mallotus villosus 0 [11] 60.0 1.30 160 Pleuronectidae Hippoglossoides platessoides ♀ 0 [11] 2 058.0 0.22 161 Pleuronectidae Hippoglossoides platessoides ♂ 0 [11] 692.0 0.30 162 Pleuronectidae Hippoglossoides platessoides ♀ −0.5 [14] 2 846.0 0.23 163 Pleuronectidae Hippoglossoides platessoides ♂ −0.5 [14] 1 657.0 0.25 164 Pleuronectidae Hippoglossoides platessoides ♀ −1 [15] 2 177.0 0.18 165 Pleuronectidae Hippoglossoides platessoides ♂ −1 [15] 1 241.0 0.27 131 Notothenidae Notothenia neglecta −1 [16] 2 178.0 0.36 Record number . Family . Species . T (°C) . M (g) . Z . 101 Leiognathidae Liognathus splendens 28 64.0 1.80 102 Nemipteridae Nemipterus bleekeri 28 300.0 1.12 103 Nemipteridae Nemipterus delagoae 28 240.0 1.62 104 Nemipteridae Nemipterus hexodon 28 215.0 0.88 105 Nemipteridae Nemipterus japonicas 28 255.0 1.88 106 Nemipteridae Nemipterus marginatus 28 230.0 1.14 107 Nemipteridae Nemipterus marginatus 28 210.0 1.73 108 Nemipteridae Nemipterus mesoprion 28 75.0 1.08 109 Nemipteridae Nemipterus nematophorus 28 200.0 1.63 110 Nemipteridae Nemipterus nemurus 28 245.0 1.03 111 Nemipteridae Nemipterus peronei 28 257.0 1.05 112 Nemipteridae Nemipterus tolu 28 230.0 0.41 113 Nemipteridae Nemipterus sp. 28 406.0 0.53 144 Scombridae Rastrelliger kanagurta 28 160.0 4.44 146 Scombridae Rastrelliger neglectus 28 110.0 7.22 147 Scombridae Rastrelliger neglectus 28 84.0 4.20 148 Scombridae Rastrelliger neglectus 28 205.0 4.56 15 Clupeidae Sardinella longiceps 27 66.0 1.12 16 Clupeidae Sardinella longiceps 27 63.0 0.67 22 Engraulidae Stolothrissa tanganicae 27 6.0 5.20 84 Serranidae Cephalopholis fulva 27 640.0 0.55 86 Serranidae Epinephelus guttatus 27 1 880.0 0.68 87 Serranidae Epinephelus striatus 27 12 900.0 0.24 88 Serranidae Mycteroperca venenosa 27 8 330.0 0.42 92 Carangidae Caranx rubber 27 2 490.0 1.51 94 Lutjanidae Apsilus dentatus ♀ 27 4 000.0 0.83 95 Lutjanidae Apsilus dentatus ♂ 27 4 500.0 1.90 96 Lutjanidae Lutjanus apodus 27 3 800.0 0.54 97 Lutjanidae Lutjaus buccanella ♀ 27 3 200.0 2.24 98 Lutjanidae Lutjanus buccanella ♂ 27 1 890.0 1.83 99 Lutjanidae Lutjanus purpureus 27 11 664.0 0.37 100 Lutjanidae Ocyurus chrysurus 27 3 600.0 0.62 114 Pomadasyidae Haemulon album 27 5 300.0 0.79 115 Pomadasyidae Haemulon plumieri 27 1 360.0 1.77 117 Sciaenidae Cynoscion macdonaldi 27 17 826.0 0.30 119 Sciaenidae Pseudosciaena diacanthus 27 17 400.0 0.80 121 Lethrinidae Lethrinus enigmaticus 27 2 916.0 0.20 122 Mullidae Mulloidichthys martinicus ♀ 27 520.0 1.70 123 Mullidae Mulloidichthys martinicus ♂ 27 360.0 1.73 124 Mullidae Pseudoupeneus macularus 27 360.0 1.89 145 Scombridae Rastrelliger kanagurta 27 23.9 7.80 174 Cynoglossidae Cynoglossus macrolepidus 27 170.0 0.49 175 Balistidae Balistes vetula 27 2 920.0 2.61 120 Sciaenidae Pseudotolithus elongatus 26 715.0 0.34 125 Cichlidae Haplochromis anaphyrmus 26 111.0 1.45 126 Cichlidae Haplochromis mloto 26 74.0 0.92 127 Cichlidae Lethrinops longispinis 26 122.0 1.69 128 Cichlidae Lethrinops parvidens 26 132.0 1.20 150 Scombridae Thunnus alalunga 26 5 150.0 0.20 154 Scombridae Thunnus albacares 26 137 500.0 0.80 155 Scombridae Thunnus atlanticus 26 9 760.0 0.67 157 Scombridae Thunnus maccoyi 26 195 660.0 0.20 129 Cichlidae Tilapia esculenta 25 750.0 1.75 141 Scombridae Katsuwonus pelamis 25 32 300.0 1.68 142 Scombridae Katsuwonus pelamis 25 10 130.0 0.65 149 Scombridae Thunnus alalunga 25 58 760.0 0.23 151 Scombridae Thunnus albacares 25 144 400.0 0.80 152 Scombridae Thunnus albacares 25 199 000.0 0.90 153 Scombridae Thunnus albacares 25 98 790.0 0.77 156 Scombridae Thunnus germo 25 36 900.0 0.22 23 Engraulidae Centengraulis mysticetus 24 56.0 2.40 118 Sciaenidae Cynoscion nobilis 24 27 900.0 0.30 6 Clupeidae Brevoortia tyrannus 23 141.0 1.00 5 Acipenseridae Acipenser transmontanus 22 170 000.0 0.03 69 Merluccidae Merluccius angustimanus 20 255.0 0.84 79 Mugilidae Mugil cephalus ♂ 20 2 450.0 0.31 93 Carangidae Trachurus japonicus 20 1 188.0 0.99 116 Sciaenidae Cynoscion analis 18 1 466.0 0.40 46 Synodontidae Saurida tumbil 17 2 350.0 0.46 132 Blennidae Blennius pholis 17 54.0 0.90 24 Engraulidae Engraulis anchoita 16 212.0 1.42 25 Engraulidae Engraulis anchoita 16 50.0 0.90 29 Engraulidae Engraulis ringens 16 37.0 1.00 30 Engraulidae Engraulis ringens 16 24.0 1.52 42 Engraulidae Osmeridae 16 15.0 2.45 17 Clupeidae Sardinops caeruea 15 337.0 0.40 18 Clupeidae Sardinops caeruea 15 225.0 0.45 19 Clupeidae Sardinops melanosticta 15 209.0 0.50 26 Engraulidae Engraulis encrasicholus 15 24.0 1.80 28 Engraulidae Engraulis mordax 15 21.0 1.70 49 Myctophidae Myctophum punctatum 15 6.6 0.73 71 Merluccidae Merluccius merluccius ♀ 15 1 577.0 0.50 72 Merluccidae Merluccius merluccius ♂ 15 622.0 0.60 77 Cyprinodontidae Aphanius fasciatus ♀ 15 4.0 1.80 78 Cyprinodontidae Aphanius fasciatus ♂ 15 2.5 2.60 143 Scombridae Pneumatophorus japonicus 15 810.0 0.90 70 Merluccidae Merluccius gayi 14 5 830.0 0.37 130 Cheilodactylidae Cheilodactylus macropterus 14 1 390.0 0.08 7 Clupeidae Brevoortia tyrannus 13 863.0 0.50 12 Clupeidae Clupea harengus 12 193.0 0.20 13 Clupeidae Clupea pallassii 12 525.0 0.20 52 Gadidae Gadus minutus ♀ 12 158.0 0.90 53 Gadidae Gadus minutus ♂ 12 59.4 1.10 73 Merluccidae Merluccius productus 12 1 272.0 0.56 74 Scomberesocidae Cololabis saira 12 189.0 1.60 135 Callionymidae Callionymus lyra ♀ 12 27.0 0.86 136 Callionymidae Callionymus lyra ♂ 12 78.0 0.96 8 Clupeidae Clupea harengus 11 200.0 0.25 133 Ammodytidae Ammodytes marinus 11 32.1 1.24 134 Ammodytidae Ammodytes tobianus 11 15.0 1.29 1 Lamnidae Cetorhinus maximus 10 13 820 000.0 0.046 4 Acipenseridae Acipenser transmontanus 10 96 680.0 0.05 14 Clupeidae Clupea pallassii 10 209.0 0.50 75 Gasterosteidae Gasterosteus aculeatus 10 1.8 0.90 76 Gasterosteidae Pungitius pungitius 10 0.4 1.10 85 Moronidae Dicentrarchus labrax 10 6 860.0 0.10 137 Gobiidae Pomatoschistus minutus 10 4.0 3.30 168 Pleuronectidae Limanda ferruginea 10 1 183.0 0.15 11 Clupeidae Clupea harengus 9 160.0 0.36 21 Clupeidae Sprattus sprattus 9 20.4 0.70 40 Salmonidae Salmo trutta 9 300.0 0.94 44 Esocidae Esox lucius ♀ 9 8 810.0 0.26 45 Esocidae Esox lucius ♂ 9 3 288.0 0.24 51 Cyprinidae Phoxinus phoxinus 9 8.2 1.10 67 Gadidae Trisopterus esmarkii 9 48.0 1.60 68 Gadidae Trisopterus esmarkii 9 45.0 1.60 80 Cottidae Cottus gobio ♀ 9 4.3 0.80 81 Cottidae Cottus gobio ♂ 9 4.3 1.10 82 Cottidae Cottus gobio ♀ 9 6.1 0.90 83 Cottidae Cottus gobio ♂ 9 5.9 1.10 158 Pleuronectidae Eopsetta jordani ♀ 9 3 970.0 0.20 159 Pleuronectidae Eopsetta jordani ♂ 9 4 566.0 0.25 3 Acipenseridae Acipenser fulvescens 8 36 000.0 0.01 9 Clupeidae Clupea harengus 8 350.0 0.16 10 Clupeidae Clupea harengus 8 55.0 0.35 20 Clupeidae Sprattus sprattus 8 23.0 0.53 35 Salmonidae Leucichthys artedi 8 70.0 1.10 36 Salmonidae Leucichthys artedi 8 95.0 1.20 37 Salmonidae Leucichthys artedi 8 600.0 0.35 50 Cyprinidae Abramis brama 8 1 593.0 0.165 57 Gadidae Gadus morhua 8 16 350.0 0.17 60 Gadidae Gadus morhua 8 20 000.0 0.20 138 Scorpaenidae Sebastes alutus 8 1 358.0 0.227 169 Pleuronectidae Platichthys flesus 8 1 040.0 0.18 2 Lamnidae Lamna nasus 7 166 800.0 0.18 47 Myctophidae Benthosema glaciale 7 5.6 1.75 48 Myctophidae Benthosema glaciale 7 5.7 0.74 56 Gadidae Gadus morhua 7 11 000.0 0.33 171 Pleuronectidae Pleuronectes platessa ♀ 7 3 430.0 0.12 172 Pleuronectidae Pleuronectes platessa ♂ 7 910.0 0.22 173 Soleidae Solea vulgaris 7 482.0 0.25 55 Gadidae Gadus morhua 6 15 000.0 0.10 62 Gadidae Gadus morhua 6 8 158.0 0.44 65 Gadidae Pollachius virens 6 11 634.0 0.30 89 Percidae Perca fluviatilis 6 435.0 0.29 90 Percidae Perca fluviatilis 6 633.0 0.16 38 Salmonidae Leucichthys sardinella 5 550.0 0.60 39 Salmonidae Salmo trutta 5 500.0 0.31 61 Gadidae Gadus morhua 5 10 834.0 0.31 63 Gadidae Melanogrammus aeglefinus 5 2 150.0 0.20 140 Scorpaenidae Sebastes alutus 5 800.0 0.193 170 Pleuronectidae Pseudopleuronectes americanus 5 1 380.0 0.40 31 Salmonidae Coregonus clupeaformis 3 [5] 2 826.0 0.15 32 Salmonidae Coregonus clupeaformis 3 [5] 6 215.0 0.17 33 Salmonidae Coregonus clupeaformis 3 [5] 4 200.0 1.35 34 Salmonidae Coregonus clupeaformis 3 [5] 27.0 1.30 54 Gadidae Gadus morhua 3 [5] 24 000.0 0.10 64 Gadidae Pollachius virens 3 [5] 20 077.0 0.30 66 Gadidae Pollachius virens 3 [5] 16 514.0 0.30 91 Percidae Stizostedion canadensis 3 [5] 615.0 0.44 58 Gadidae Gadus morhua 2 [6] 11 720.0 0.20 139 Scorpaenidae Sebastes alutus 2 [6] 1 081.0 0.271 41 Salmonidae Salvelinus alpinus 1 [7] 5 000.0 0.17 59 Gadidae Gadus morhua 1 [8] 3 000.0 0.18 166 Pleuronectidae Hippoglossoides platessoides ♀ 1 [7] 3 823.0 0.18 167 Pleuronectidae Hippoglossoides platessoides ♂ 1 [7] 2 035.0 0.26 43 Osmeridae Mallotus villosus 0 [11] 60.0 1.30 160 Pleuronectidae Hippoglossoides platessoides ♀ 0 [11] 2 058.0 0.22 161 Pleuronectidae Hippoglossoides platessoides ♂ 0 [11] 692.0 0.30 162 Pleuronectidae Hippoglossoides platessoides ♀ −0.5 [14] 2 846.0 0.23 163 Pleuronectidae Hippoglossoides platessoides ♂ −0.5 [14] 1 657.0 0.25 164 Pleuronectidae Hippoglossoides platessoides ♀ −1 [15] 2 177.0 0.18 165 Pleuronectidae Hippoglossoides platessoides ♂ −1 [15] 1 241.0 0.27 131 Notothenidae Notothenia neglecta −1 [16] 2 178.0 0.36 Records are organized in descending order according to mean environmental temperature (T). Records for polar fish (≤3.5 °C) indicate both mean environmental temperature and the associated ‘physiologically effective temperature’ (enclosed in brackets). Open in new tab Table A1 Data on growth, temperature and fish mortality used in Pauly (1980) analysis. Record number . Family . Species . T (°C) . M (g) . Z . 101 Leiognathidae Liognathus splendens 28 64.0 1.80 102 Nemipteridae Nemipterus bleekeri 28 300.0 1.12 103 Nemipteridae Nemipterus delagoae 28 240.0 1.62 104 Nemipteridae Nemipterus hexodon 28 215.0 0.88 105 Nemipteridae Nemipterus japonicas 28 255.0 1.88 106 Nemipteridae Nemipterus marginatus 28 230.0 1.14 107 Nemipteridae Nemipterus marginatus 28 210.0 1.73 108 Nemipteridae Nemipterus mesoprion 28 75.0 1.08 109 Nemipteridae Nemipterus nematophorus 28 200.0 1.63 110 Nemipteridae Nemipterus nemurus 28 245.0 1.03 111 Nemipteridae Nemipterus peronei 28 257.0 1.05 112 Nemipteridae Nemipterus tolu 28 230.0 0.41 113 Nemipteridae Nemipterus sp. 28 406.0 0.53 144 Scombridae Rastrelliger kanagurta 28 160.0 4.44 146 Scombridae Rastrelliger neglectus 28 110.0 7.22 147 Scombridae Rastrelliger neglectus 28 84.0 4.20 148 Scombridae Rastrelliger neglectus 28 205.0 4.56 15 Clupeidae Sardinella longiceps 27 66.0 1.12 16 Clupeidae Sardinella longiceps 27 63.0 0.67 22 Engraulidae Stolothrissa tanganicae 27 6.0 5.20 84 Serranidae Cephalopholis fulva 27 640.0 0.55 86 Serranidae Epinephelus guttatus 27 1 880.0 0.68 87 Serranidae Epinephelus striatus 27 12 900.0 0.24 88 Serranidae Mycteroperca venenosa 27 8 330.0 0.42 92 Carangidae Caranx rubber 27 2 490.0 1.51 94 Lutjanidae Apsilus dentatus ♀ 27 4 000.0 0.83 95 Lutjanidae Apsilus dentatus ♂ 27 4 500.0 1.90 96 Lutjanidae Lutjanus apodus 27 3 800.0 0.54 97 Lutjanidae Lutjaus buccanella ♀ 27 3 200.0 2.24 98 Lutjanidae Lutjanus buccanella ♂ 27 1 890.0 1.83 99 Lutjanidae Lutjanus purpureus 27 11 664.0 0.37 100 Lutjanidae Ocyurus chrysurus 27 3 600.0 0.62 114 Pomadasyidae Haemulon album 27 5 300.0 0.79 115 Pomadasyidae Haemulon plumieri 27 1 360.0 1.77 117 Sciaenidae Cynoscion macdonaldi 27 17 826.0 0.30 119 Sciaenidae Pseudosciaena diacanthus 27 17 400.0 0.80 121 Lethrinidae Lethrinus enigmaticus 27 2 916.0 0.20 122 Mullidae Mulloidichthys martinicus ♀ 27 520.0 1.70 123 Mullidae Mulloidichthys martinicus ♂ 27 360.0 1.73 124 Mullidae Pseudoupeneus macularus 27 360.0 1.89 145 Scombridae Rastrelliger kanagurta 27 23.9 7.80 174 Cynoglossidae Cynoglossus macrolepidus 27 170.0 0.49 175 Balistidae Balistes vetula 27 2 920.0 2.61 120 Sciaenidae Pseudotolithus elongatus 26 715.0 0.34 125 Cichlidae Haplochromis anaphyrmus 26 111.0 1.45 126 Cichlidae Haplochromis mloto 26 74.0 0.92 127 Cichlidae Lethrinops longispinis 26 122.0 1.69 128 Cichlidae Lethrinops parvidens 26 132.0 1.20 150 Scombridae Thunnus alalunga 26 5 150.0 0.20 154 Scombridae Thunnus albacares 26 137 500.0 0.80 155 Scombridae Thunnus atlanticus 26 9 760.0 0.67 157 Scombridae Thunnus maccoyi 26 195 660.0 0.20 129 Cichlidae Tilapia esculenta 25 750.0 1.75 141 Scombridae Katsuwonus pelamis 25 32 300.0 1.68 142 Scombridae Katsuwonus pelamis 25 10 130.0 0.65 149 Scombridae Thunnus alalunga 25 58 760.0 0.23 151 Scombridae Thunnus albacares 25 144 400.0 0.80 152 Scombridae Thunnus albacares 25 199 000.0 0.90 153 Scombridae Thunnus albacares 25 98 790.0 0.77 156 Scombridae Thunnus germo 25 36 900.0 0.22 23 Engraulidae Centengraulis mysticetus 24 56.0 2.40 118 Sciaenidae Cynoscion nobilis 24 27 900.0 0.30 6 Clupeidae Brevoortia tyrannus 23 141.0 1.00 5 Acipenseridae Acipenser transmontanus 22 170 000.0 0.03 69 Merluccidae Merluccius angustimanus 20 255.0 0.84 79 Mugilidae Mugil cephalus ♂ 20 2 450.0 0.31 93 Carangidae Trachurus japonicus 20 1 188.0 0.99 116 Sciaenidae Cynoscion analis 18 1 466.0 0.40 46 Synodontidae Saurida tumbil 17 2 350.0 0.46 132 Blennidae Blennius pholis 17 54.0 0.90 24 Engraulidae Engraulis anchoita 16 212.0 1.42 25 Engraulidae Engraulis anchoita 16 50.0 0.90 29 Engraulidae Engraulis ringens 16 37.0 1.00 30 Engraulidae Engraulis ringens 16 24.0 1.52 42 Engraulidae Osmeridae 16 15.0 2.45 17 Clupeidae Sardinops caeruea 15 337.0 0.40 18 Clupeidae Sardinops caeruea 15 225.0 0.45 19 Clupeidae Sardinops melanosticta 15 209.0 0.50 26 Engraulidae Engraulis encrasicholus 15 24.0 1.80 28 Engraulidae Engraulis mordax 15 21.0 1.70 49 Myctophidae Myctophum punctatum 15 6.6 0.73 71 Merluccidae Merluccius merluccius ♀ 15 1 577.0 0.50 72 Merluccidae Merluccius merluccius ♂ 15 622.0 0.60 77 Cyprinodontidae Aphanius fasciatus ♀ 15 4.0 1.80 78 Cyprinodontidae Aphanius fasciatus ♂ 15 2.5 2.60 143 Scombridae Pneumatophorus japonicus 15 810.0 0.90 70 Merluccidae Merluccius gayi 14 5 830.0 0.37 130 Cheilodactylidae Cheilodactylus macropterus 14 1 390.0 0.08 7 Clupeidae Brevoortia tyrannus 13 863.0 0.50 12 Clupeidae Clupea harengus 12 193.0 0.20 13 Clupeidae Clupea pallassii 12 525.0 0.20 52 Gadidae Gadus minutus ♀ 12 158.0 0.90 53 Gadidae Gadus minutus ♂ 12 59.4 1.10 73 Merluccidae Merluccius productus 12 1 272.0 0.56 74 Scomberesocidae Cololabis saira 12 189.0 1.60 135 Callionymidae Callionymus lyra ♀ 12 27.0 0.86 136 Callionymidae Callionymus lyra ♂ 12 78.0 0.96 8 Clupeidae Clupea harengus 11 200.0 0.25 133 Ammodytidae Ammodytes marinus 11 32.1 1.24 134 Ammodytidae Ammodytes tobianus 11 15.0 1.29 1 Lamnidae Cetorhinus maximus 10 13 820 000.0 0.046 4 Acipenseridae Acipenser transmontanus 10 96 680.0 0.05 14 Clupeidae Clupea pallassii 10 209.0 0.50 75 Gasterosteidae Gasterosteus aculeatus 10 1.8 0.90 76 Gasterosteidae Pungitius pungitius 10 0.4 1.10 85 Moronidae Dicentrarchus labrax 10 6 860.0 0.10 137 Gobiidae Pomatoschistus minutus 10 4.0 3.30 168 Pleuronectidae Limanda ferruginea 10 1 183.0 0.15 11 Clupeidae Clupea harengus 9 160.0 0.36 21 Clupeidae Sprattus sprattus 9 20.4 0.70 40 Salmonidae Salmo trutta 9 300.0 0.94 44 Esocidae Esox lucius ♀ 9 8 810.0 0.26 45 Esocidae Esox lucius ♂ 9 3 288.0 0.24 51 Cyprinidae Phoxinus phoxinus 9 8.2 1.10 67 Gadidae Trisopterus esmarkii 9 48.0 1.60 68 Gadidae Trisopterus esmarkii 9 45.0 1.60 80 Cottidae Cottus gobio ♀ 9 4.3 0.80 81 Cottidae Cottus gobio ♂ 9 4.3 1.10 82 Cottidae Cottus gobio ♀ 9 6.1 0.90 83 Cottidae Cottus gobio ♂ 9 5.9 1.10 158 Pleuronectidae Eopsetta jordani ♀ 9 3 970.0 0.20 159 Pleuronectidae Eopsetta jordani ♂ 9 4 566.0 0.25 3 Acipenseridae Acipenser fulvescens 8 36 000.0 0.01 9 Clupeidae Clupea harengus 8 350.0 0.16 10 Clupeidae Clupea harengus 8 55.0 0.35 20 Clupeidae Sprattus sprattus 8 23.0 0.53 35 Salmonidae Leucichthys artedi 8 70.0 1.10 36 Salmonidae Leucichthys artedi 8 95.0 1.20 37 Salmonidae Leucichthys artedi 8 600.0 0.35 50 Cyprinidae Abramis brama 8 1 593.0 0.165 57 Gadidae Gadus morhua 8 16 350.0 0.17 60 Gadidae Gadus morhua 8 20 000.0 0.20 138 Scorpaenidae Sebastes alutus 8 1 358.0 0.227 169 Pleuronectidae Platichthys flesus 8 1 040.0 0.18 2 Lamnidae Lamna nasus 7 166 800.0 0.18 47 Myctophidae Benthosema glaciale 7 5.6 1.75 48 Myctophidae Benthosema glaciale 7 5.7 0.74 56 Gadidae Gadus morhua 7 11 000.0 0.33 171 Pleuronectidae Pleuronectes platessa ♀ 7 3 430.0 0.12 172 Pleuronectidae Pleuronectes platessa ♂ 7 910.0 0.22 173 Soleidae Solea vulgaris 7 482.0 0.25 55 Gadidae Gadus morhua 6 15 000.0 0.10 62 Gadidae Gadus morhua 6 8 158.0 0.44 65 Gadidae Pollachius virens 6 11 634.0 0.30 89 Percidae Perca fluviatilis 6 435.0 0.29 90 Percidae Perca fluviatilis 6 633.0 0.16 38 Salmonidae Leucichthys sardinella 5 550.0 0.60 39 Salmonidae Salmo trutta 5 500.0 0.31 61 Gadidae Gadus morhua 5 10 834.0 0.31 63 Gadidae Melanogrammus aeglefinus 5 2 150.0 0.20 140 Scorpaenidae Sebastes alutus 5 800.0 0.193 170 Pleuronectidae Pseudopleuronectes americanus 5 1 380.0 0.40 31 Salmonidae Coregonus clupeaformis 3 [5] 2 826.0 0.15 32 Salmonidae Coregonus clupeaformis 3 [5] 6 215.0 0.17 33 Salmonidae Coregonus clupeaformis 3 [5] 4 200.0 1.35 34 Salmonidae Coregonus clupeaformis 3 [5] 27.0 1.30 54 Gadidae Gadus morhua 3 [5] 24 000.0 0.10 64 Gadidae Pollachius virens 3 [5] 20 077.0 0.30 66 Gadidae Pollachius virens 3 [5] 16 514.0 0.30 91 Percidae Stizostedion canadensis 3 [5] 615.0 0.44 58 Gadidae Gadus morhua 2 [6] 11 720.0 0.20 139 Scorpaenidae Sebastes alutus 2 [6] 1 081.0 0.271 41 Salmonidae Salvelinus alpinus 1 [7] 5 000.0 0.17 59 Gadidae Gadus morhua 1 [8] 3 000.0 0.18 166 Pleuronectidae Hippoglossoides platessoides ♀ 1 [7] 3 823.0 0.18 167 Pleuronectidae Hippoglossoides platessoides ♂ 1 [7] 2 035.0 0.26 43 Osmeridae Mallotus villosus 0 [11] 60.0 1.30 160 Pleuronectidae Hippoglossoides platessoides ♀ 0 [11] 2 058.0 0.22 161 Pleuronectidae Hippoglossoides platessoides ♂ 0 [11] 692.0 0.30 162 Pleuronectidae Hippoglossoides platessoides ♀ −0.5 [14] 2 846.0 0.23 163 Pleuronectidae Hippoglossoides platessoides ♂ −0.5 [14] 1 657.0 0.25 164 Pleuronectidae Hippoglossoides platessoides ♀ −1 [15] 2 177.0 0.18 165 Pleuronectidae Hippoglossoides platessoides ♂ −1 [15] 1 241.0 0.27 131 Notothenidae Notothenia neglecta −1 [16] 2 178.0 0.36 Record number . Family . Species . T (°C) . M (g) . Z . 101 Leiognathidae Liognathus splendens 28 64.0 1.80 102 Nemipteridae Nemipterus bleekeri 28 300.0 1.12 103 Nemipteridae Nemipterus delagoae 28 240.0 1.62 104 Nemipteridae Nemipterus hexodon 28 215.0 0.88 105 Nemipteridae Nemipterus japonicas 28 255.0 1.88 106 Nemipteridae Nemipterus marginatus 28 230.0 1.14 107 Nemipteridae Nemipterus marginatus 28 210.0 1.73 108 Nemipteridae Nemipterus mesoprion 28 75.0 1.08 109 Nemipteridae Nemipterus nematophorus 28 200.0 1.63 110 Nemipteridae Nemipterus nemurus 28 245.0 1.03 111 Nemipteridae Nemipterus peronei 28 257.0 1.05 112 Nemipteridae Nemipterus tolu 28 230.0 0.41 113 Nemipteridae Nemipterus sp. 28 406.0 0.53 144 Scombridae Rastrelliger kanagurta 28 160.0 4.44 146 Scombridae Rastrelliger neglectus 28 110.0 7.22 147 Scombridae Rastrelliger neglectus 28 84.0 4.20 148 Scombridae Rastrelliger neglectus 28 205.0 4.56 15 Clupeidae Sardinella longiceps 27 66.0 1.12 16 Clupeidae Sardinella longiceps 27 63.0 0.67 22 Engraulidae Stolothrissa tanganicae 27 6.0 5.20 84 Serranidae Cephalopholis fulva 27 640.0 0.55 86 Serranidae Epinephelus guttatus 27 1 880.0 0.68 87 Serranidae Epinephelus striatus 27 12 900.0 0.24 88 Serranidae Mycteroperca venenosa 27 8 330.0 0.42 92 Carangidae Caranx rubber 27 2 490.0 1.51 94 Lutjanidae Apsilus dentatus ♀ 27 4 000.0 0.83 95 Lutjanidae Apsilus dentatus ♂ 27 4 500.0 1.90 96 Lutjanidae Lutjanus apodus 27 3 800.0 0.54 97 Lutjanidae Lutjaus buccanella ♀ 27 3 200.0 2.24 98 Lutjanidae Lutjanus buccanella ♂ 27 1 890.0 1.83 99 Lutjanidae Lutjanus purpureus 27 11 664.0 0.37 100 Lutjanidae Ocyurus chrysurus 27 3 600.0 0.62 114 Pomadasyidae Haemulon album 27 5 300.0 0.79 115 Pomadasyidae Haemulon plumieri 27 1 360.0 1.77 117 Sciaenidae Cynoscion macdonaldi 27 17 826.0 0.30 119 Sciaenidae Pseudosciaena diacanthus 27 17 400.0 0.80 121 Lethrinidae Lethrinus enigmaticus 27 2 916.0 0.20 122 Mullidae Mulloidichthys martinicus ♀ 27 520.0 1.70 123 Mullidae Mulloidichthys martinicus ♂ 27 360.0 1.73 124 Mullidae Pseudoupeneus macularus 27 360.0 1.89 145 Scombridae Rastrelliger kanagurta 27 23.9 7.80 174 Cynoglossidae Cynoglossus macrolepidus 27 170.0 0.49 175 Balistidae Balistes vetula 27 2 920.0 2.61 120 Sciaenidae Pseudotolithus elongatus 26 715.0 0.34 125 Cichlidae Haplochromis anaphyrmus 26 111.0 1.45 126 Cichlidae Haplochromis mloto 26 74.0 0.92 127 Cichlidae Lethrinops longispinis 26 122.0 1.69 128 Cichlidae Lethrinops parvidens 26 132.0 1.20 150 Scombridae Thunnus alalunga 26 5 150.0 0.20 154 Scombridae Thunnus albacares 26 137 500.0 0.80 155 Scombridae Thunnus atlanticus 26 9 760.0 0.67 157 Scombridae Thunnus maccoyi 26 195 660.0 0.20 129 Cichlidae Tilapia esculenta 25 750.0 1.75 141 Scombridae Katsuwonus pelamis 25 32 300.0 1.68 142 Scombridae Katsuwonus pelamis 25 10 130.0 0.65 149 Scombridae Thunnus alalunga 25 58 760.0 0.23 151 Scombridae Thunnus albacares 25 144 400.0 0.80 152 Scombridae Thunnus albacares 25 199 000.0 0.90 153 Scombridae Thunnus albacares 25 98 790.0 0.77 156 Scombridae Thunnus germo 25 36 900.0 0.22 23 Engraulidae Centengraulis mysticetus 24 56.0 2.40 118 Sciaenidae Cynoscion nobilis 24 27 900.0 0.30 6 Clupeidae Brevoortia tyrannus 23 141.0 1.00 5 Acipenseridae Acipenser transmontanus 22 170 000.0 0.03 69 Merluccidae Merluccius angustimanus 20 255.0 0.84 79 Mugilidae Mugil cephalus ♂ 20 2 450.0 0.31 93 Carangidae Trachurus japonicus 20 1 188.0 0.99 116 Sciaenidae Cynoscion analis 18 1 466.0 0.40 46 Synodontidae Saurida tumbil 17 2 350.0 0.46 132 Blennidae Blennius pholis 17 54.0 0.90 24 Engraulidae Engraulis anchoita 16 212.0 1.42 25 Engraulidae Engraulis anchoita 16 50.0 0.90 29 Engraulidae Engraulis ringens 16 37.0 1.00 30 Engraulidae Engraulis ringens 16 24.0 1.52 42 Engraulidae Osmeridae 16 15.0 2.45 17 Clupeidae Sardinops caeruea 15 337.0 0.40 18 Clupeidae Sardinops caeruea 15 225.0 0.45 19 Clupeidae Sardinops melanosticta 15 209.0 0.50 26 Engraulidae Engraulis encrasicholus 15 24.0 1.80 28 Engraulidae Engraulis mordax 15 21.0 1.70 49 Myctophidae Myctophum punctatum 15 6.6 0.73 71 Merluccidae Merluccius merluccius ♀ 15 1 577.0 0.50 72 Merluccidae Merluccius merluccius ♂ 15 622.0 0.60 77 Cyprinodontidae Aphanius fasciatus ♀ 15 4.0 1.80 78 Cyprinodontidae Aphanius fasciatus ♂ 15 2.5 2.60 143 Scombridae Pneumatophorus japonicus 15 810.0 0.90 70 Merluccidae Merluccius gayi 14 5 830.0 0.37 130 Cheilodactylidae Cheilodactylus macropterus 14 1 390.0 0.08 7 Clupeidae Brevoortia tyrannus 13 863.0 0.50 12 Clupeidae Clupea harengus 12 193.0 0.20 13 Clupeidae Clupea pallassii 12 525.0 0.20 52 Gadidae Gadus minutus ♀ 12 158.0 0.90 53 Gadidae Gadus minutus ♂ 12 59.4 1.10 73 Merluccidae Merluccius productus 12 1 272.0 0.56 74 Scomberesocidae Cololabis saira 12 189.0 1.60 135 Callionymidae Callionymus lyra ♀ 12 27.0 0.86 136 Callionymidae Callionymus lyra ♂ 12 78.0 0.96 8 Clupeidae Clupea harengus 11 200.0 0.25 133 Ammodytidae Ammodytes marinus 11 32.1 1.24 134 Ammodytidae Ammodytes tobianus 11 15.0 1.29 1 Lamnidae Cetorhinus maximus 10 13 820 000.0 0.046 4 Acipenseridae Acipenser transmontanus 10 96 680.0 0.05 14 Clupeidae Clupea pallassii 10 209.0 0.50 75 Gasterosteidae Gasterosteus aculeatus 10 1.8 0.90 76 Gasterosteidae Pungitius pungitius 10 0.4 1.10 85 Moronidae Dicentrarchus labrax 10 6 860.0 0.10 137 Gobiidae Pomatoschistus minutus 10 4.0 3.30 168 Pleuronectidae Limanda ferruginea 10 1 183.0 0.15 11 Clupeidae Clupea harengus 9 160.0 0.36 21 Clupeidae Sprattus sprattus 9 20.4 0.70 40 Salmonidae Salmo trutta 9 300.0 0.94 44 Esocidae Esox lucius ♀ 9 8 810.0 0.26 45 Esocidae Esox lucius ♂ 9 3 288.0 0.24 51 Cyprinidae Phoxinus phoxinus 9 8.2 1.10 67 Gadidae Trisopterus esmarkii 9 48.0 1.60 68 Gadidae Trisopterus esmarkii 9 45.0 1.60 80 Cottidae Cottus gobio ♀ 9 4.3 0.80 81 Cottidae Cottus gobio ♂ 9 4.3 1.10 82 Cottidae Cottus gobio ♀ 9 6.1 0.90 83 Cottidae Cottus gobio ♂ 9 5.9 1.10 158 Pleuronectidae Eopsetta jordani ♀ 9 3 970.0 0.20 159 Pleuronectidae Eopsetta jordani ♂ 9 4 566.0 0.25 3 Acipenseridae Acipenser fulvescens 8 36 000.0 0.01 9 Clupeidae Clupea harengus 8 350.0 0.16 10 Clupeidae Clupea harengus 8 55.0 0.35 20 Clupeidae Sprattus sprattus 8 23.0 0.53 35 Salmonidae Leucichthys artedi 8 70.0 1.10 36 Salmonidae Leucichthys artedi 8 95.0 1.20 37 Salmonidae Leucichthys artedi 8 600.0 0.35 50 Cyprinidae Abramis brama 8 1 593.0 0.165 57 Gadidae Gadus morhua 8 16 350.0 0.17 60 Gadidae Gadus morhua 8 20 000.0 0.20 138 Scorpaenidae Sebastes alutus 8 1 358.0 0.227 169 Pleuronectidae Platichthys flesus 8 1 040.0 0.18 2 Lamnidae Lamna nasus 7 166 800.0 0.18 47 Myctophidae Benthosema glaciale 7 5.6 1.75 48 Myctophidae Benthosema glaciale 7 5.7 0.74 56 Gadidae Gadus morhua 7 11 000.0 0.33 171 Pleuronectidae Pleuronectes platessa ♀ 7 3 430.0 0.12 172 Pleuronectidae Pleuronectes platessa ♂ 7 910.0 0.22 173 Soleidae Solea vulgaris 7 482.0 0.25 55 Gadidae Gadus morhua 6 15 000.0 0.10 62 Gadidae Gadus morhua 6 8 158.0 0.44 65 Gadidae Pollachius virens 6 11 634.0 0.30 89 Percidae Perca fluviatilis 6 435.0 0.29 90 Percidae Perca fluviatilis 6 633.0 0.16 38 Salmonidae Leucichthys sardinella 5 550.0 0.60 39 Salmonidae Salmo trutta 5 500.0 0.31 61 Gadidae Gadus morhua 5 10 834.0 0.31 63 Gadidae Melanogrammus aeglefinus 5 2 150.0 0.20 140 Scorpaenidae Sebastes alutus 5 800.0 0.193 170 Pleuronectidae Pseudopleuronectes americanus 5 1 380.0 0.40 31 Salmonidae Coregonus clupeaformis 3 [5] 2 826.0 0.15 32 Salmonidae Coregonus clupeaformis 3 [5] 6 215.0 0.17 33 Salmonidae Coregonus clupeaformis 3 [5] 4 200.0 1.35 34 Salmonidae Coregonus clupeaformis 3 [5] 27.0 1.30 54 Gadidae Gadus morhua 3 [5] 24 000.0 0.10 64 Gadidae Pollachius virens 3 [5] 20 077.0 0.30 66 Gadidae Pollachius virens 3 [5] 16 514.0 0.30 91 Percidae Stizostedion canadensis 3 [5] 615.0 0.44 58 Gadidae Gadus morhua 2 [6] 11 720.0 0.20 139 Scorpaenidae Sebastes alutus 2 [6] 1 081.0 0.271 41 Salmonidae Salvelinus alpinus 1 [7] 5 000.0 0.17 59 Gadidae Gadus morhua 1 [8] 3 000.0 0.18 166 Pleuronectidae Hippoglossoides platessoides ♀ 1 [7] 3 823.0 0.18 167 Pleuronectidae Hippoglossoides platessoides ♂ 1 [7] 2 035.0 0.26 43 Osmeridae Mallotus villosus 0 [11] 60.0 1.30 160 Pleuronectidae Hippoglossoides platessoides ♀ 0 [11] 2 058.0 0.22 161 Pleuronectidae Hippoglossoides platessoides ♂ 0 [11] 692.0 0.30 162 Pleuronectidae Hippoglossoides platessoides ♀ −0.5 [14] 2 846.0 0.23 163 Pleuronectidae Hippoglossoides platessoides ♂ −0.5 [14] 1 657.0 0.25 164 Pleuronectidae Hippoglossoides platessoides ♀ −1 [15] 2 177.0 0.18 165 Pleuronectidae Hippoglossoides platessoides ♂ −1 [15] 1 241.0 0.27 131 Notothenidae Notothenia neglecta −1 [16] 2 178.0 0.36 Records are organized in descending order according to mean environmental temperature (T). Records for polar fish (≤3.5 °C) indicate both mean environmental temperature and the associated ‘physiologically effective temperature’ (enclosed in brackets). Open in new tab © International Council for the Exploration of the Sea 2016. All rights reserved. For Permissions, please email: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Biais,, Gérard;Coupeau,, Yann;Séret,, Bernard;Calmettes,, Beatriz;Lopez,, Rémy;Hetherington,, Stuart;Righton,, David

doi: 10.1093/icesjms/fsw233pmid: N/A

During two surveys in 2011 and 2013, we deployed pop-up satellite archival tags (PSATs) on subadult or adult porbeagles at the Bay of Biscay shelf break. We collected data that enabled the reconstruction of nine migrations (eight females, one male) that uncover the large spatial extent of these sharks in the Northeast Atlantic. The mean duration of each deployment was 292 d, with four reaching 365 d. The reconstructions show that, after migrations that extended up to 2000 km away from the point of release, the tagged porbeagles returned to their location of tagging. All the reconstructed migrations followed the same general pattern of a migration away from the Bay of Biscay in late summer, and a return in spring the following year. The total distance of the migrations was estimated at 5000–13 000 km for PSATs deployed for a full year (n = 4), with examples of migration to the Arctic Circle, southward to Madeira and westward to the mid-Atlantic Ridge. The observed site fidelity to the Bay of Biscay and the common migration pattern of all females provide evidence of complex spatial structure and dynamics that encompasses both the open ocean and heavily fished coastal areas, and highlights the challenge of assessing and managing the porbeagle stock in this area. Introduction The porbeagle (Lamna nasus) is a large pelagic shark found throughout the North Atlantic, between 30° and 75°N (Aasen, 1961; Compagno, 2001), and circumglobally between 25° and 60°S in the southern hemisphere, although it is absent from the North Pacific Ocean (Francis et al., 2008; Semba et al., 2013). Its distribution encompasses the high seas, coastal shelf, and inshore areas; porbeagles have even been reported from rivers (Matheson, 1928). They are born at 60–90 cm total length (TL) [fork length (FL) values are converted to TL in this article using the 1.12 ratio given by Campana et al., 2013], (Aasen, 1963; Jensen et al., 2002), and reach 170 cm TL at 5–6-years old (Natanson et al., 2002). Maximum recorded lengths are over 340 cm TL (Templeman, 1963; Kohler et al., 2002), with males in the Northeast Atlantic reaching maturity at 190 cm TL and females at above 223 cm TL (Hennache and Jung, 2010). Porbeagle sharks have a fusiform body shape, providing a powerful swimming capacity and the potential for long-distance migrations. Like all lamnid sharks, porbeagles have the capacity to raise their body temperature above ambient by conserving metabolic heat (Carey and Teal, 1969), which enables tolerance of cool waters and occupation of the relatively high latitudes where the species is found (Campana and Joyce, 2004). Porbeagles have been exploited in Europe since the 1930s. Market demand led to the development of a Northeast Atlantic porbeagle fishery based initially in Norway (ICES, 2015a). After the Second World War, Danish vessels began to target porbeagle, and landings rose to a peak during the late 1940s. As catches declined through the 1950s, the porbeagle fishery remained predominantly Scandinavian and located in Northern European waters (Norwegian Sea, North Sea, Northwest Scotland and Faroe region) but expanded into Western European waters (Bay of Biscay, Southwest Ireland, Celtic Sea), in the early 1960s before landings declined further when the Norwegian interest in the Northeast Atlantic fishery waned (Rae, 1962). French vessels exploiting porbeagle in western European waters became the dominant source of fishing mortality in the 1970s. However, landings of porbeagle continued to decline through the 1980s and 1990s, eventually leading to concerns in the early 21st century that the North-eastern Atlantic stock was at risk. In consequence, Norway banned directed fishing for porbeagle in 2007 before the European Union prohibited all catches in 2010. However, at present, the true state of the stock still remains unknown (ICES, 2015a) because a full assessment of stock status is not possible, predominantly because abundance indices for the stock are not available at the time the fishery was at its peak (ICCAT, 2009; ICES, 2015b). Developing a greater understanding of porbeagle biology and ecology is therefore desirable. For management and assessment purposes, two porbeagle stocks are considered in the North Atlantic (ICES, 2007; ICCAT, 2009); one to each side of the 42°W meridian. This stock separation is supported by mark-recapture experiments. In the Northwest Atlantic, tag recaptures show that movement occurs between fishing areas all along the North American coast (n = 209 returns from ∼2000 releases; Kohler et al., 2002; Campana et al., 2013). In the northeast Atlantic, the pattern is similar, with recaptures mainly along the western European coast rather than across the ocean (n = 15 returns from 165 releases, one east to west transoceanic trip reported; Stevens, 1990; Kohler and Turner, 2001; Kohler et al., 2002; ICES, 2007). However, the limited number of tag recoveries does not allow an exhaustive analysis of spatial extent or migratory patterns. Furthermore, because observed movements are dependent on the distribution of fishing effort, mark-recapture data can only ever provide a limited insight into movements or distribution. Pop-up satellite archival tags (PSATs) provide a method for collecting direct and detailed evidence of the migrations and distributions of species that are not fished across their geographic range, or which are tagged and recovered only rarely. This technology is now a routine research tool (e.g. Block et al., 2011). PSATs were first deployed on porbeagle in the Northwest Atlantic from 2001 to 2008 (Campana et al., 2010). In the Northeast Atlantic, PSATs have been deployed in 2007 in the Celtic Sea (Pade et al., 2009) and off Northwest Ireland in 2008–2009 (Saunders et al., 2011). In the southern hemisphere, PSATs have also been deployed off New Zealand (Francis et al., 2015). Overall, these deployments have shown that male and female porbeagles may undertake large migrations in the open ocean, whether immature, sub-adult or mature. However, the time at liberty for tagged porbeagles rarely reaches one year (only one tag, deployed on a shark in the Northwest Atlantic, has reported 348 d after release, the other deployments are 10 months long or less) and, consequently, the annual migratory cycle is not well understood for porbeagles in any part of their geographic range. Furthermore, despite the value of the PSAT deployments in the Northeast Atlantic, the number of individuals from which data have been collected is particularly limited (n = 4 and 3 from Pade et al., 2009 and Saunders et al., 2011 respectively), and the duration of the observations, at fewer than 4 months, is relatively short. Tagged individuals have also generally been small, falling in the range 102–207 cm TL (91–185 cm FL), including three mature males but no mature females, leaving an important gap in our knowledge of the seasonal movements or the extent of adult migrations. To address this gap, we undertook a PSAT tagging programme to study the migrations of adult porbeagle, with an emphasis on mature females. Between 2011 and 2013 we tagged 13 large porbeagles (1 male and 12 females 197–265 cm TL) in the region of the Bay of Biscay. Our results show that porbeagles migrate widely across the Northeast Atlantic, but also that they exhibit site fidelity, a behaviour that is shown for the first time for porbeagle sharks. Material and methods Shark tagging Two tagging surveys were carried out in 2011 and 2013 using a chartered commercial long-lining vessel. The fishing area was the Bay of Biscay shelf break between latitudes 46°N and 48°N. The goal of each survey was to deploy PSATs on females preferably larger than 230 cm TL (a length at which they are likely to be reproductively mature; Hennache and Jung, 2010). Pelagic longlines were used to catch porbeagles; line lengths and set durations were limited to ensure hooked sharks were in the best possible condition for tagging. The long-line was set at 200–300 m depth at locations over the continental slope (700–3600 m seabed depths). To achieve the best tag attachment, all sharks were brought on board for tagging. Sharks were supported on a foam block, their eyes were shielded with a damp towel, and their gills were irrigated with running seawater at all times during the procedure. Wildlife Computers Mk10 or MiniPat (both of which record light intensity, depth and temperature) were attached to each shark by inserting a urethane anchor (Wildlife Computer’s “Wilton dart”) about 10 cm into the pterygiophores below the dorsal fin and at one-third of its length from the posterior end. A stainless steel anchor attached to a nylon ring was also inserted behind the dorsal fin to bridle the tag and stream it alongside the body. Tags were programmed to release after 365 d (for large females) and after 190 d (for the male) with an aim to obtain information on potential parturition and mating locations and on the relevant locations and times for future surveys. Depth (±0.5 m), temperature (±0.1 °C) and light intensity were recorded every 10 s. The data were internally binned by 6 h (7 PSAT), 12 h (one PSAT in 2011) or 24 h (one PSAT in 2011 with depth time series generation each 10 min enabled) intervals. PSAT were programmed to release from the shark if a constant depth (±2.5 m) was maintained for 4 d, indicating mortality due to the absence of depth change. Shark tagging was conducted in accordance with the guidelines of the Animal Care Committee of France. Track reconstruction We aimed to estimate one location per day that best explained the daily observed light intensity, depth and temperature data recovered from the PSAT (number of days with observations in Supplementary Table S1). The process relies on a state-space formulation of the tracking problem. In essence, the daily location is the unknown state of a dynamic system governed by a dynamic model (random walk), which means that the sequence of daily state follows a Markov chain. This (hidden) location is directly linked by an observation model to the light, depth, and surface temperature records. Within this framework, the effective resolution resorts to a discretization of the state space in grid cells (0.1 × 0.1 degrees) to infer the probability density of the (hidden) location, given the sequence of observations. This approach was first introduced in the field of fish tracking by Thygesen et al. (2009) and it is known as Grid Filter in the geolocation literature (Neilson et al., 2014). It is a recursive Bayesian estimation technique, in line with the widespread Kalman filter (e.g. Nielsen et al., 2006; Royer and Lutcavage, 2009). In this Bayesian filtering, the daily measurement assimilation is a two-step process. At each sampling time, the method performs a position prediction step by numerically solving the advection-diffusion equation for the 2D probability of animal’s presence. A position update step is then performed to combine the predicted probability density with information recorded by the tag to produce the posterior distribution of the animal. Thus, at each time step the probability of presence of the fish is obtained on each point of the grid. The daily location is computed as the average of the grid locations weighted by their probability. The track of each tag is obtained by connecting the daily position estimates. The standard deviation of a daily location estimate in 2D representations is an ellipse. The orientation angle, the semi-major and semi-minor axis lengths of the ellipse are then deduced from the error covariance matrix of the daily distribution. We implemented a modified version of the geolocation method described in Neilson et al. (2014). First, the positions update step uses the raw locations derived from light intensity as the observed data (obtained with the Wildlife Computers Software “GPE2”). The update step is constrained by the bathymetry as in Thygesen et al. (2009)(i.e. the depth at the updated location must be greater than the daily recorded maximum depth) but also by the satellite-based sea surface temperature SST (by minimizing the deviation of this SST at the updated location from the daily recorded SST). Additionally, the first location of the track is assumed to be the deployment position of the tag and the last location is the pop-up position provided by ARGOS. If the detachment of the tag was premature, the tag will drift 4 d until reporting its position, according to PSAT parameter settings. The tracks were estimated using the same model parameters for all the PSATs: (i) the diffusion coefficient of the random walk model is set empirically to 1000 km2 d−1 because this value minimizes the SST root mean square error (satellite observed vs tag recorded) calculated along the track (a posteriori error), (ii) the standard deviation of the raw light-based location of the update step is set to 1° in longitude and 3.5° in latitude and (iii) The standard deviation of the error in SST is set to 0.5 °C (a priori error). The satellite-based sea surface temperature at the updated location is obtained from Ostia (Operational Sea Surface Temperature and Sea Ice Analysis, see http://ghrsst-pp.metoffice.com/pages/latest_analysis/ – spatial resolution: 0.05 degrees– temporal resolution: daily) and the depth is given by the ETOPO02 bathymetry model (see http://www.ngdc.noaa.gov/mgg/global/etopo2.html – spatial resolution: 2 min). Results Despite setting longlines at locations where large porbeagles had been reported in commercial catches in previous years, catching and tagging females larger than 230 cm TL proved to be difficult. In total, eight large females (197–265 cm TL) and one large male (199 cm TL) were tagged (Table 1), for which tags were retained for over 4 months (mean 292 d; range 128–365 d). In addition, three porbeagles were tagged from which long-term data were not recovered; one porbeagle was captured, while two other tags detached prematurely (shark fate unknown). One tag did not transmit any data. The ratio of long-term tag deployments to premature release shows that our deep and precise insertion of the PSAT anchor and the addition of a ring to bridle the tag alongside the body was extremely successful. Table 1 Tagging summary for the nine tagged porbeagles used for track reconstructions. Shark . Sex . TL (cm) . FL (cm) . Tagging Date . Tagging Lat. (°N) . Tagging Long. (°W) . Pop-up date . Days at liberty . Tagging to pop-up distance (km) . Estimated mean daily move (km) . Estimated trip length (km) . 1 M 199 171 23 Jun 2011 47°47,1′ 8°44,1′ 30 Dec 2011 190 475 34 6512 2 F 265 234 26 Jun 2011 47°30,8′ 7°24,0′ 08 Feb 2012 227 2408 30 6778 3 F 204 180 28 Jun 2011 47°49,1′ 8°05,7′ 27 Jun 2012 365 311 37 13 352 4 F 233 199 03 Jul 2011 46°54,1′ 5°36,1′ 02 Jul 2012 365 12 26 9331 5 F 197 172 09 Jun 2013 47°08,5′ 5°58,6′ 03 May 2014 328 863 24 7905 6 F 235 207 10 Jun 2013 47°09,6′ 5°54,0′ 28 Mar 2014 291 911 27 7857 7 F 233 202 10 Jun 2013 47°09,8′ 5°54,7′ 10 Jun 2014 365 384 15 5594 8 F 250 218 11 Jun 2013 47°08,9′ 5°49,6′ 17 Oct 2013 128 864 30 3871 9 F 228 206 13 Jun 2013 46°42,5′ 5°08,4′ 13 Jun 2014 365 41 26 9410 Mean F 231 202 – – – – 304 724 27 8012 Mean M+F 227 199 – – – – 292 697 28 7846 Shark . Sex . TL (cm) . FL (cm) . Tagging Date . Tagging Lat. (°N) . Tagging Long. (°W) . Pop-up date . Days at liberty . Tagging to pop-up distance (km) . Estimated mean daily move (km) . Estimated trip length (km) . 1 M 199 171 23 Jun 2011 47°47,1′ 8°44,1′ 30 Dec 2011 190 475 34 6512 2 F 265 234 26 Jun 2011 47°30,8′ 7°24,0′ 08 Feb 2012 227 2408 30 6778 3 F 204 180 28 Jun 2011 47°49,1′ 8°05,7′ 27 Jun 2012 365 311 37 13 352 4 F 233 199 03 Jul 2011 46°54,1′ 5°36,1′ 02 Jul 2012 365 12 26 9331 5 F 197 172 09 Jun 2013 47°08,5′ 5°58,6′ 03 May 2014 328 863 24 7905 6 F 235 207 10 Jun 2013 47°09,6′ 5°54,0′ 28 Mar 2014 291 911 27 7857 7 F 233 202 10 Jun 2013 47°09,8′ 5°54,7′ 10 Jun 2014 365 384 15 5594 8 F 250 218 11 Jun 2013 47°08,9′ 5°49,6′ 17 Oct 2013 128 864 30 3871 9 F 228 206 13 Jun 2013 46°42,5′ 5°08,4′ 13 Jun 2014 365 41 26 9410 Mean F 231 202 – – – – 304 724 27 8012 Mean M+F 227 199 – – – – 292 697 28 7846 Table 1 Tagging summary for the nine tagged porbeagles used for track reconstructions. Shark . Sex . TL (cm) . FL (cm) . Tagging Date . Tagging Lat. (°N) . Tagging Long. (°W) . Pop-up date . Days at liberty . Tagging to pop-up distance (km) . Estimated mean daily move (km) . Estimated trip length (km) . 1 M 199 171 23 Jun 2011 47°47,1′ 8°44,1′ 30 Dec 2011 190 475 34 6512 2 F 265 234 26 Jun 2011 47°30,8′ 7°24,0′ 08 Feb 2012 227 2408 30 6778 3 F 204 180 28 Jun 2011 47°49,1′ 8°05,7′ 27 Jun 2012 365 311 37 13 352 4 F 233 199 03 Jul 2011 46°54,1′ 5°36,1′ 02 Jul 2012 365 12 26 9331 5 F 197 172 09 Jun 2013 47°08,5′ 5°58,6′ 03 May 2014 328 863 24 7905 6 F 235 207 10 Jun 2013 47°09,6′ 5°54,0′ 28 Mar 2014 291 911 27 7857 7 F 233 202 10 Jun 2013 47°09,8′ 5°54,7′ 10 Jun 2014 365 384 15 5594 8 F 250 218 11 Jun 2013 47°08,9′ 5°49,6′ 17 Oct 2013 128 864 30 3871 9 F 228 206 13 Jun 2013 46°42,5′ 5°08,4′ 13 Jun 2014 365 41 26 9410 Mean F 231 202 – – – – 304 724 27 8012 Mean M+F 227 199 – – – – 292 697 28 7846 Shark . Sex . TL (cm) . FL (cm) . Tagging Date . Tagging Lat. (°N) . Tagging Long. (°W) . Pop-up date . Days at liberty . Tagging to pop-up distance (km) . Estimated mean daily move (km) . Estimated trip length (km) . 1 M 199 171 23 Jun 2011 47°47,1′ 8°44,1′ 30 Dec 2011 190 475 34 6512 2 F 265 234 26 Jun 2011 47°30,8′ 7°24,0′ 08 Feb 2012 227 2408 30 6778 3 F 204 180 28 Jun 2011 47°49,1′ 8°05,7′ 27 Jun 2012 365 311 37 13 352 4 F 233 199 03 Jul 2011 46°54,1′ 5°36,1′ 02 Jul 2012 365 12 26 9331 5 F 197 172 09 Jun 2013 47°08,5′ 5°58,6′ 03 May 2014 328 863 24 7905 6 F 235 207 10 Jun 2013 47°09,6′ 5°54,0′ 28 Mar 2014 291 911 27 7857 7 F 233 202 10 Jun 2013 47°09,8′ 5°54,7′ 10 Jun 2014 365 384 15 5594 8 F 250 218 11 Jun 2013 47°08,9′ 5°49,6′ 17 Oct 2013 128 864 30 3871 9 F 228 206 13 Jun 2013 46°42,5′ 5°08,4′ 13 Jun 2014 365 41 26 9410 Mean F 231 202 – – – – 304 724 27 8012 Mean M+F 227 199 – – – – 292 697 28 7846 Migrations and daily horizontal movements Reconstructions of migrations showed that porbeagles migrated to a range of distant locations; from the Bay of Biscay, northward to the Arctic Circle, southward to Madeira and westward to the Mid-Atlantic Ridge (Figures 1 and 2 and Supplementary Material S1). The confidence limits of the most probable tracks are relatively large (the mean width of the 50% confidence interval area is ∼250 km). However, despite the low precision, the direction and extent of the migrations were clear. Figure 1 Open in new tabDownload slide North East Atlanctic Ocean. 200, 1000, and 2000 m depth contours and area names cited in the text are shown. Figure 1 Open in new tabDownload slide North East Atlanctic Ocean. 200, 1000, and 2000 m depth contours and area names cited in the text are shown. Figure 2 Open in new tabDownload slide Reconstructed tracks (left) and daily estimated movement distance (right) of the 9 porbeagles tagged in the Bay of Biscay in June-July 2011 (nos. 1–4) and June 2013 (nos. 5–9). 50% CIs are displayed as light grey ellipses and 1000 m depth contours are shown. Downward and upward triangles denote the tagging and pop-up locations, respectively. Figure 2 Open in new tabDownload slide Reconstructed tracks (left) and daily estimated movement distance (right) of the 9 porbeagles tagged in the Bay of Biscay in June-July 2011 (nos. 1–4) and June 2013 (nos. 5–9). 50% CIs are displayed as light grey ellipses and 1000 m depth contours are shown. Downward and upward triangles denote the tagging and pop-up locations, respectively. The mean daily distance travelled was 28 km, ranging from <1 km to nearly 200 km (Figure 2). The total estimated travel distance ranged between 3800 and 13 400 km (Table 1), with a maximum distance of ∼2400 km between release position and tag pop-up. However, when the pop-up time of the tag was 365 d after release (four of the nine sharks), the tags reported a pop-off position within 400 km of release; two of these were within 50 km. In general, sharks at liberty for 9 months or more showed a movement back towards the point of tagging, strongly suggesting a spring (April–June) return to the Bay of Biscay. The pattern of this round-trip was similar between sharks, as follows. They resided at the shelf break of the Bay of Biscay after tagging or when returning the next year (April–September), where daily movement rates were often low. In August–October, all but one (shark no. 7) of the female porbeagles moved in a northwesterly migration along the shelf break to West Ireland (generally on the west side of the Porcupine Bank), before eventually reaching a latitude of 54–55°N. This migration from the Bay of Biscay was generally rapid, but it was sometimes punctuated by short-term residences on the continental slope (sharks nos. 3, 5, 9 on or near Porcupine Bank, no. 8 in the West of Brittany) or an incursion on the continental shelf (no. 6). Sharks nos. 1 and 2 undertook this northward migration slightly earlier than others, and were located to the west of Scotland in August. From this latitude, two routes were identified. Some porbeagles (nos. 1, 2, 3, and 9) travelled northeast to the West of Scotland, the Faroe region, the North Sea or the Norwegian Sea, remaining in residence for up to 5 months. Others (nos. 4–6) adopted a general westerly direction, reaching the mid-Atlantic Ridge, where they stayed for up to 6 weeks. At the end of the autumn period and beginning of winter, all sharks turned to the south. The rate of movement was rapid in some cases (sharks nos. 4 and 9), with sharks eventually reaching latitudes as far south as 33°N (shark no. 3 close to Madeira) or 36°N (shark no. 4 in Azores region) between mid-February to the beginning of April. A return to the Bay of Biscay was observed in March–April each time that the tags remained deployed 12 months. The migration pattern of the adult male (no. 1) was similar to that of the females but the timings of the large-scale movements differed slightly. The shark moved north after release, but turned to the south earlier than the females at the beginning of September and returned to the area of its initial release. It remained in the Bay of Biscay off the northern coast of Spain for 2 months before moving northward again in mid-November, when the track was terminated by the pop-off date. Porbeagles tagged in 2011 migrated further north than those sharks tagged in 2013 and an exception to the large offshore general trip was observed for one shark (no. 7, 230 cm TL). This shark remained in the Bay of Biscay and adjacent waters for its entire period at liberty (365 d). However, it still exhibited the general north–south migration pattern of the other sharks; residing in more northerly waters in autumn (September–December), migrating rapidly to waters off north Portugal in January–February before returning to the Bay of Biscay in spring (April–June). Vertical habitat use Porbeagles ranged between the surface and 1600 m depth during their time at liberty, but rarely ventured deeper than 700 m (Figure 3). All sharks occupied the upper 200 m of the water column predominantly (monthly average time percentage: 59% ±11), but they all spent time in the mesopelagic zone (>200m), with some individuals (nos. 1, 3, 4, and 9) exhibiting an affinity for deeper waters. The timing of deep diving occurred in spring (February to April) for sharks nos. 3, 4, and 9 (respectively in Madeira, Azores, and Galicia Bank areas) and in September for shark no. 1 to the west of Scotland. Figure 3 Open in new tabDownload slide Time at depth distribution by month (left part of each panel) and overall (right) of the nine porbeagles tagged in the Bay of Biscay in June–July 2011 (nos. 1–4) and June 2013 (nos. 5–9). Layers limits from top to down are 50, 100, 150, 200, 300, 400, 500, 600, 700, and >700 m. Time at depth data are shaded to indicate proportion of time at each depth band, while overall depth distribution is shown as a percentage along the x-axis. Figure 3 Open in new tabDownload slide Time at depth distribution by month (left part of each panel) and overall (right) of the nine porbeagles tagged in the Bay of Biscay in June–July 2011 (nos. 1–4) and June 2013 (nos. 5–9). Layers limits from top to down are 50, 100, 150, 200, 300, 400, 500, 600, 700, and >700 m. Time at depth data are shaded to indicate proportion of time at each depth band, while overall depth distribution is shown as a percentage along the x-axis. Patterns of depth occupation were bimodal, with individuals splitting their time between surface waters (0–50 m and sometimes 50–100 m) and depth zones below 200 m, with the exception of shark no. 2 (Figure 3). The mean percentage in the 0–50 m surface waters was 38% ±9 rising to 47% ±9 in summer (July–September) when the sharks were typically in the Bay of Biscay or off southwest Ireland. This bimodal occupation of different ocean layers is partly driven by changes in vertical habitat during the migration. Thus, when sharks resided in the Norwegian Sea (no. 2 in December–February), Madeira (no. 3 in February), the Mid-Atlantic Ridge in the Azores Region (no. 4 in February), they remained in the lower part of the epipelagic layer or in the mesopelagic layer (Figure 3). The method of data transmission from the PSATs (depth records were binned by 6, 12 or 24 h with no time-series information, except shark no. 1) does not permit a detailed examination of diel vertical movements, but the bimodal pattern of depth was also likely a consequence of daily movements between shallow and deep water; time at depth histograms by quarter show this diel pattern clearly (Supplementary Figure S2). The use of the mesopelagic layer increased during the periods that include more daylight hours (6–18 GMT), and generally largely, with few exceptions mostly in quarter 4 (sharks nos. 2, 3, 7). Discussion The reconstructions shown here provide evidence that sub-adult and mature female porbeagles undertake large annual cyclical migrations. The general pattern we observed was a northward movement in summer–autumn (August–October) followed by a northward or westward extension in autumn–winter (September–February), and continued later on by a movement to south of 43°N in winter–spring (January–April). Porbeagles rarely moved north of 62°N but a period of residence north of 52°N in autumn–winter (September–February) was observed in almost all reconstructed migrations. Returns to the Bay of Biscay and southwest Ireland shelf break in spring (March–June) were observed in four of the sharks that were tracked for a full year. The observation of site fidelity is even more remarkable because our data provide evidence that, in some cases, large females moved 1900–2200 km away from the point of release as far south as 33°N (shark no. 3) and as far west as 31°W (shark no. 4). The migratory paths and general pattern of movements exhibited by the porbeagles in this study are consistent with those described elsewhere for porbeagle in the Northeast Atlantic (Pade et al., 2009; Saunders et al., 2011). In previous studies, sharks were tagged close to the coastal shelf in summer or in early autumn and migrated to the shelf break and offshore areas in autumn and winter when deployments were long enough to observe this migration. In Saunders et al., 2011 extensive migrations to the south were observed during winter, similar to sharks tagged in our study. However, unlike those studies, we observed migrations to high latitudes and the mid-Atlantic ridge in the late summer, before the southward migrations occurred. This may indicate that larger sharks have a greater capacity for large migratory movements although it should be noted that one shark of 91 cm FL tagged in Saunders et al., 2011 migrated to the west coast of Morocco over a period of 6 months. Tag retention in Pade et al. (2009) was too low to provide the same information (<90 d; mean 44 d). A recent study on porbeagle in the southern hemisphere also showed also that sharks move predominantly north – south to occupy lower latitudes in winter than in summer (Francis et al., 2015). This study provides 10 tracks (deployment durations 72–300 d; median 221 d) that show that most of the porbeagles remained in offshore waters (depth > 1000 m) adjacent to New Zealand; one immature male (140 cm LF) performed a long circular migration in 300 d. Deployment durations did not provide evidence of site fidelity as clearly as in our study but Francis et al. (2015) as well as Saunders et al. (2011), show that small and immature porbeagles may also undertake north–south migrations with a presumption of site fidelity. Further studies will shed light on this phenomenon. The use of the water column by porbeagles in our study was also similar to that described in other studies. Typically, porbeagles predominantly used the epipelagic zone in summer, before switching to greater use of the mesopelagic zone from autumn to spring, a result reported by Saunders et al., 2011 and Francis et al. (2015). Porbeagles that had migrated further offshore tended to make greater use of the mesopelagic zone, likely related to the greater productivity of these areas in winter, or in areas of ocean in proximity to seabed features known to attract biomass (e.g. seamounts). Further work to establish the drivers of vertical migration is necessary to develop a greater understanding of the links between migration and habitat use. The annual migration cycle is likely to be linked to critical times for feeding and reproductive activity. For example, based on records of historical porbeagle catches, Rae (1962) describes the annual arrival of porbeagle in the North Sea as “an invasion” beginning in May and reaching a peak in August, in synchrony with catches of spawning herring, suggesting that the availability of this food resource is related to the increase in abundance of porbeagle. Herring (Clupea harengus) is the species that is most commonly observed in porbeagle stomachs (Gauld, 1989) in northern European waters. The occurrence of spawning herring aggregations off the northeast Scottish and Shetland coasts during August to September and in the central North Sea during August to October is likely a factor in the increase in porbeagle abundance in the North Sea in summer. In western European waters, the fishery typically began in March–April, peaked between May to August, and ended in September–October; large porbeagles (FL > 200 cm) were caught throughout this period (Lallemand-Lemoine, 1991; Hennache and Jung, 2010). The porbeagle diet in the Bay of Biscay is dominated by horse mackerel (Trachurus trachurus) and blue whiting (Micromesistius poutassou) (Hennache and Jung, 2010). These two preys are abundant during spring and summer on the shelf break (Certain et al., 2011). These timings correspond to the periods when the tagged porbeagles in our study remained close (or returned) to the coastal shelf and predominantly occupied epipelagic depths. Since the historic porbeagle fishery had a broad spatial extent, and none of the tagged sharks exhibited migration from west European waters in spring to north European waters in summer, it seems likely that spatially separate fisheries were exploiting a widely dispersed population. Therefore the evidence for site fidelity in our study suggests that the porbeagle population in the North-East Atlantic may be formed by components which return to spring–summer feeding areas that are widely separated. Similar behaviours are seen in other shark species. For example, the salmon shark Lamna ditropis, a lamnid shark which replaces porbeagle in the North Pacific Ocean (Francis et al., 2008), also makes long distance migration before returning to the productive Alaskan coast (Weng et al., 2008). This behaviour is suggested to improve foraging success of migratory sharks by reducing the cost of research of suitable feeding areas as suggested for the white shark Carcharodon carcharias (Jorgensen et al., 2010), the oceanic whitetip Carcharhinus longimanus (Howey-Jordan et al., 2013) or the tiger shark Galeocerdo cuvier (Lea et al., 2015). Site fidelity may also be linked to reproductive ecology but, for many shark species it is not fully described even if there are evidence that it is common for sharks (Chapman et al., 2015). Within the lamnid sharks, fidelity to mating sites is suggested for the white shark (Jorgensen et al., 2010; Domeier and Nasby-Lucas, 2013) and the salmon shark (Weng et al., 2008). Natal philopatry is also suggested for these two species (Bonfil et al., 2005; Weng et al., 2008), with the possibility of two parturition areas for the salmon shark. One of them is a highly productive region (the California Current) but the other one is the Subtropical Gyre, which is an oligotrophic region. However, this latter fact remains to be demonstrated, particularly because the advantage of locating pupping grounds in an oligotrophic region is questionable. Similarly, from observations of migration in the Western Atlantic, Campana et al. (2010) suggested that porbeagle pupping grounds might be in the southern part of their annual migration to the Sargasso Sea, which is also a low productivity region of the ocean. This inference of a pupping ground in subtropical waters (south of latitude 35°N) was strongly based on the observation that the southward migration is only made by female longer than 218 cm FL at which 50% of porbeagle females are mature (FL50) (Jensen et al., 2002). In our study, six of the tagged females that remained at liberty for more than 9 months (mean TL 222 cm) were at or larger than the size at maturity estimated by Hennache and Jung (2010). The reconstructed tracks do not provide evidence of any incursion into tropical waters; the most southerly location recorded was 33°N, raising the possibility that pupping grounds of the stock of porbeagles in the Northeast Atlantic might also be located in temperate waters. Observations of several large embryos and small free-swimming specimens and the captures of gravid females provide evidence of this possibility. For example, in June 1960, a large female porbeagle was caught off Jersey (Western Channel) containing an 89 cm TL embryo (Caunter, 1961). More recently, a catch of four gravid females with a total of 12 embryos, each about 80 cm long (TL), were reported in May 2008 on the south Celtic Sea shelf break (Hennache and Jung, 2010). Two catches of gravid females containing large embryos (60–63 and 66–76 cm TL) were also reported in East-Scotland and around Shetland in May and June (Gauld, 1989). Further evidence of parturition close to the western European shelf was recently provided by the captures of 9 newborn pups on the Bay of Biscay shelf break in May 2015 and July 2016, during an anchovy sentinel survey (n = 1; 74 cm FL; E. Duhamel, pers. comm.) and a new porbeagle tagging survey carried out in June–July 2016 by some authors of this paper (n = 8; 74–90 cm FL). Based on these observations, the parturition period may begin as early as May, and extend to at least July. This matches the spring–summer residency period of large female porbeagles in the Celtic Sea and the Bay of Biscay shelf and shelf-edge. This area is also a habitat of major importance to juvenile porbeagle (77% of 2008–2009 French catches < 170 cm TL in Celtic Sea– Bay of Biscay; Hennache and Jung, 2010), raising the possibility that mature females and their offspring occupy the same summer-spring feeding area and suggesting natal philopatry (Hueter et al., 2004; Feldheim et al., 2014; Chapman et al., 2015). To what extent the male population might conform to the same migration patterns and consequently might form a discrete demographic unit with the females remains unknown although our single mature male track (shark no. 1) suggests that it might be possible. Genetic and wider tagging studies are required to test this hypothesis, as well that of philopatry behaviour, which remains speculative at the present time. However, our findings provide evidence that the dynamics and life-history processes of porbeagle sharks are spatially structured and complex. The porbeagle stock structure definition in the Northeast Atlantic therefore remains an issue to tackle for assessments and management of this stock. Supplementary data Supplementary material is available at the ICESJMS online version of the article. Acknowledgements We thank the skipper, Adrien Delavaud, and the crew of the fishing vessel ‘Fille du Suet’ for their commitment to the success of the two tagging surveys that we carried out with them. We are thankful to José Jouaneau and to Fanny Brivoal, respectively Chairman and General Secretary of the Pays de la Loire fishermen Committee, and to Marie-Bénédicte Peyrat, Head of Scientific affairs office at the Directorate for Marine Fisheries of the Ecology, sustainable development and energy French Ministry, for their support. Funding This work was funded by the French state (Directorate for Marine Fisheries), the Pays de la Loire Regional Council, the Pays de la Loire fishermen Committee and the UK Department for Environment, Food and Rural Affairs (MB5201). References Aasen O. 1961 . Some observations on the biology of the porbeagle shark (Lamna nasus L.). ICES document C.M. Near Northern Seas Committee: 109. 7 pp. Aasen O. 1963 . Length and growth of the porbeagle (Lamna nasus, Bonnaterre) in the North West Atlantic. Fiskeridirektoratets Skrifter, Serie Havundersøkelser, 13: 20–37. Block B. A. , Jonsen I. D., Jorgensen S. J., Winship A. J., Shaffer S. A., Bograd S. J., Hazen E. L., et al. 2011 . Tracking apex marine predator movements in a dynamic ocean . Nature , 475 : 86 – 90 . Google Scholar Crossref Search ADS PubMed WorldCat Bonfil R. , Meÿer M., Scholl M. C., Johnson R., O’Brien S., Oosthuizen H., Swanson S. et al. 2005 . Transoceanic migration, spatial dynamics, and population linkages of white sharks . Science , 310 : 100 – 103 . Google Scholar Crossref Search ADS PubMed WorldCat Campana S. E. , Gibson A. J. F., Fowler M., Dorey A., Joyce W. 2013 . Population dynamics of Northwest Atlantic porbeagle (Lamna nasus), with an assessment of status and projections for recovery . Research Document 2012/096, Canadian Science Advisory Secretariat, Fisheries and Ocean , Ottawa, Canada . 86 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Campana S. E. , Joyce W. N. 2004 . Temperature and depth associations of porbeagle shark (Lamna nasus) in the northwest Atlantic . Fisheries Oceanography, 13 : 52 – 64 . Google Scholar Crossref Search ADS WorldCat Campana S. E. , Joyce W., Fowler M. 2010 . Subtropical pupping ground for a cold-water shark . Canadian Journal of Fisheries and Aquatic Sciences, 67 : 769 – 773 . Google Scholar Crossref Search ADS WorldCat Carey F. G. , Teal J. M. 1969 . Mako and porbeagle: warm-bodied sharks . Comparative Biochemistry and Physiology, 28 : 199 – 204 . Google Scholar Crossref Search ADS WorldCat Caunter J. A. L. 1961 . Shark angling in Great Britain . George Allen and Unwin Ltd , London, UK . 145 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Certain G. , Masse J., Van Canneyt O., Petitgas P., Doremus G., Santos M. B., Ridoux V. 2011 . Investigating the coupling between small pelagic fish and marine top predators using data collected from ecosystem-based surveys . Marine Ecology Progress Series, 422 : 23 – 39 . Google Scholar Crossref Search ADS WorldCat Chapman D. D. , Feldheim K. A., Papastamatiou Y. P., Hueter R. E. 2015 . There and back again: a review of residency and return migrations in sharks, with implications for population structure and management . Annual Review of Marine Science, 7 : 547 – 570 . Google Scholar Crossref Search ADS WorldCat Compagno L. J. V. 2001 . Sharks of the world. An annotated and illustrated catalogue of shark species known to date. Bullhead, mackerel and carpet sharks (Heterodontiformes, Lamniformes and Orectolobiformes) . Species Catalogue for Fishery Purposes No. 1, Volume 2 . FAO, Rome . 269 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Domeier M. L. , Nasby-Lucas N. 2013 . Two-year migration of adult female white sharks (Carcharodon carcharias) reveals widely separated nursery areas and conservation concerns . Animal Biotelemetry, 1 : 1 – 10 . Google Scholar Crossref Search ADS WorldCat Feldheim K. A. , Gruber S. H., DiBattista J. D., Babcock E. A., Kessel S. A., Hendry A. P., Pikitch E. K., et al. 2014 . Two decades of genetic profiling yields first evidence of natal philopatry and long-term fidelity to parturition sites in sharks . Molecular Ecology, 23 : 110 – 117 . Google Scholar Crossref Search ADS WorldCat Francis M. P. , Holdsworth J. C., Block B. A. 2015 . Life in the open ocean: seasonal migration and diel diving behaviour of Southern Hemisphere porbeagle sharks (Lamna nasus) . Marine Biology, 162 : 2305 – 2323 . Google Scholar Crossref Search ADS WorldCat Francis M. P. , Natanson L. J., Campana S. E. 2008 . The biology and ecology of the porbeagle shark, Lamnanasus. In Sharks of the Open Ocean: Biology, Fisheries and Conservation, pp. 105–113 . Ed. by Camhi M. D., Pikitch E. K., Babcock E. A.. Blackwell Publishing , Oxford, UK . 502 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Gauld J. A. 1989 . Records of porbeagles landed in Scotland, with observations on the biology, distribution and exploitation of the species . Scottish Fisheries Research Report No. 45. Department of Agriculture and Fisheries for Scotland . Aberdeen, UK . 15 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Hennache C. , Jung A. 2010 . Etude de la pêche palangrière de requin-taupe de l’île d’Yeu. Association pour l’étude et la conservation des sélaciens (APECS), Brest, France. 64 pp. http://www.asso-apecs.org/IMG/pdf/APECS_EPPARTIY_Rapport_final_BD.pdf, (downloaded on 14 March 2012). Howey-Jordan L. A. , Brooks E. J., Abercrombie D. L., Jordan L. K., Brooks A., William S., Gospodarczyk E. et al. 2013 . Complex movements, philopatry and expanded depth range of a severely threatened pelagic shark, the oceanic whitetip (Carcharhinus longimanus) in the western North Atlantic . PLoS One, 8 : e56588. Google Scholar Crossref Search ADS WorldCat Hueter R. E. , Heupel M. R., Heist E. J., Keeney D. B. 2005 . Evidence of philopatry in sharks and implications for the management of shark fisheries . Journal of Northwest Atlantic Fishery Science, 35 : 239 – 247 . Google Scholar Crossref Search ADS WorldCat ICCAT . 2009 . Report of the 2009 Porbeagle Stock Assessments Meeting. ICCAT, Collective Volume of Scientific Papers, 65: 1909–2005. ICES . 2007 . Report of the Working Group on Elasmobranch Fishes (WGEF). ICES document C.M. 2007/ACFM: 27. 318 pp. ICES . 2015a . Porbeagle (Lamna nasus) in the North East Atlantic. ICES Advice 2015, 9. 6 pp. ICES . 2015b . Report of the Working Group on Elasmobranch Fishes (WGEF). ICES document C.M. 2015/ACPM: 19. 711 pp. Jensen C. F. , Natanson L. J., Pratt H. L., Kohler N. E., Campana S. E. 2002 . The reproductive biology of the porbeagle shark (Lamna nasus) in the western North Atlantic Ocean . Fishery Bulletin US, 100 : 727 – 738 . OpenURL Placeholder Text WorldCat Jorgensen S. J. , Reeb C. A., Chapple T. K., Anderson S., Perle C., Van Sommeran S. R., Fritz-Cope C., et al. 2010 . Philopatry and migration of Pacific white sharks . Proceeding of the Royal Society B, 277 : 679 – 688 . Google Scholar Crossref Search ADS WorldCat Kohler N. E. , Turner P. A. 2001 . Shark tagging: a review of conventional methods and studies . Environmental Biology of Fishes, 60 : 191 – 223 . Google Scholar Crossref Search ADS WorldCat Kohler N. E. , Turner P. A., Hoey J. J., Natanson L. J., Briggs R. 2002 . Tag and recapture data for three pelagic shark species: blue shark (Prionace glauca), shortfin mako (Isurus oxyrinchus), and porbeagle (Lamna nasus) in the North Atlantic Ocean . iccat, 54 : 1231 – 1260 . Collective Volume of Scientific Papers. OpenURL Placeholder Text WorldCat Lallemand-Lemoine L. 1991 . Analysis of the French fishery for porbeagle Lamma nasus (Bonnaterre, 1788) . ICES Document C.M, 10. 1991/G:71. pp. OpenURL Placeholder Text WorldCat Lea J. S. E. , Wetherbee B. M., Queiroz N., Burnie N., Aming C., Sousa L. L., Mucientes G. R. et al. 2015 . Repeated, long-distance migrations by a philopatric predator targeting highly contrasting ecosystems . Scientific Report, 5 : 11202. Google Scholar Crossref Search ADS WorldCat Matheson C. 1928 . Porbeagle shark in the River Towy . Nature, 3077 : 608 – 609 . Google Scholar Crossref Search ADS WorldCat Natanson L. J. , Mello J. J., Campana S. E. 2002 . Validated age and growth of the porbeagle shark (Lamna nasus) in the western North Atlantic Ocean . Fishery Bulletin US, 100 : 266 – 278 . OpenURL Placeholder Text WorldCat Neilson J. D. , Loefer J., Prince E. D., Royer F., Calmettes B., Gaspar P., Lopez R. et al. 2014 . Seasonal distributions and migrations of Northwest Atlantic swordfish: inferences from integration of pop-up satellite archival tagging studies . PLoS One, 9 : e112736. Google Scholar Crossref Search ADS WorldCat Nielsen A. , Bigelow K. A., Musyl M. K., Sibert J. R. 2006 . Improving light-based geolocation by including sea surface temperature . Fisheries Oceanography, 15 : 314 – 325 . Google Scholar Crossref Search ADS WorldCat Pade N. G. , Queiroz N., Humphries N. E., Witt M. J., Jones C. S., Noble L. R., Sims D. W. 2009 . First results from satellite-linked archival tagging of porbeagle shark, Lamna nasus: area fidelity, wider-scale movements and plasticity in diel depth changes . Journal of Experimental Marine Biology and Ecology, 370 : 64 – 74 . Google Scholar Crossref Search ADS WorldCat Rae B. B. 1962 . Porbeagle sharks . Scottish Fisheries Bulletin, 18 : 17 – 19 . OpenURL Placeholder Text WorldCat Royer F. , Lutcavage M., 2009 . Positioning pelagic fish from sunrise and sunset times: complex observation errors call for constrained, robust modeling. In Tagging and Tracking of Marine Animals with Electronic Devices, Vol. 9 , pp. 323 – 341 . Ed. by Nielsen J. L.. et al. Reviews: Methods and Technologies in Fish Biology and Fisheries , Springer, New York . 452 pp. Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Saunders R. A. , Royer F., Clarke M. W. 2011 . Winter migration and diving behaviour of porbeagle shark, Lamna nasus, in the Northeast Atlantic . ICES Journal of Marine Science, 68 : 166 – 174 . Google Scholar Crossref Search ADS WorldCat Semba Y. , Yokawa K., Matsunaga H., Shono H. 2013 . Distribution and trend in abundance of the porbeagle (Lamna nasus) in the southern hemisphere . Marine and Freshwater Research, 64 : 518 – 529 . Google Scholar Crossref Search ADS WorldCat Stevens J. D. 1990 . Further results from a tagging study of pelagic sharks in the north-east Atlantic . Journal of the Marine Biological Association of the United Kingdom, 70 : 707 – 720 . Google Scholar Crossref Search ADS WorldCat Templeman W. 1963 . Distribution of sharks in the Canadian Atlantic (with special reference to Newfoundland waters) . Bulletin Fisheries Research Board of Canada , Ottawa , 140 , 77 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Thygesen U. H. , Pedersen M. W., Madsen H., 2009 . Geolocating fish using hidden markov models and data storage tags. In Tagging and Tracking of Marine Animals with Electronic Devices, Vol. 9 , pp. 277 – 293 . Ed. by Nielsen J. L.., et al. Reviews: Methods and Technologies in Fish Biology and Fisheries , Springer, New York . 452 pp. Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Weng K. C. , Foley D. G., Ganong J. E., Perle C., Shillinger G. L., Block B. A. 2008 . Migration of an upper trophic level predator, the salmon shark Lamna ditropis, between distant ecoregions . Marine Ecology Progress Series, 372 : 253 – 264 . Google Scholar Crossref Search ADS WorldCat © International Council for the Exploration of the Sea 2017. All rights reserved. For Permissions, please email: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Ichinokawa, Momoko; Okamura, Hiroshi; Kurota, Hiroyuki

doi: 10.1093/icesjms/fsx002pmid: N/A

We present the first quantitative review of the stock status relative to the stock biomass (B) and the exploitation rate (U) that achieved the maximum sustainable yield (MSY) (BMSY and UMSY, respectively) for 37 Japanese stocks contributing 61% of the total marine capture production in Japan. BMSY and UMSY were estimated by assuming three types of stock-recruitment (S-R) relationships and an age-structured population model or by applying a surplus production model. The estimated stock status shows that approximately half of the stocks were overfishing (U/UMSY > 1), and approximately half of the stocks were overfished (B/BMSY < 0.5) during 2011–2013. Over the past 15 years, U decreased and B slightly increased on average. The rate of decrease in the U of the stocks managed by the total allowable catch (TAC) was significantly greater than that of the other stocks, providing evidence of the effectiveness of TAC management in Japan. The above statuses and trends were insensitive to the assumption of the S-R relationship. The characteristics of Japanese stocks composed mainly of resources with relatively high natural mortality, i.e. productivity, suggest that Japanese fisheries have great potential of exhibiting a quick recovery and increasing their yield by adjusting the fishing intensity to an appropriate level. Introduction The question of whether fishery resources are currently globally overfished is among the most concerning and important issues in marine ecosystem assessments. At present, this question is not sufficiently answered because the statuses and trends of fishery resources differ greatly among different regions of the world, and there is a substantial information gap (Worm et al., 2009; Hilborn and Ovando, 2014; Komori et al., 2016). Fishery management effectively reduces the fishing impact in many well-assessed fisheries in developed countries, such as the United States (Melnychuk et al., 2013) and Europe (Fernandes and Cook, 2013), and tuna regional fisheries management organizations (Pons et al., 2017). These well-assessed fisheries, however, represent only part of the world, and the stock statuses in other regions, such as Asia, Africa, and South America, are not sufficiently understood to predict the future global fishery sustainability (Hilborn and Ovando, 2014). Japan is one of the most important fishery countries, with the fifth highest marine production, and currently contains 5% (15% at one time) of the world’s fisheries (FAO, 2016, Figure 1), although its stock status is not fully known. A historical review of Japanese catches in the northwest Pacific revealed peaks during the late 1980s and 1970s with and without Japanese sardine, respectively, and after these peaks the levels exhibited a gradual decline. If this decrease in catches reflects a decrease in the population size (Pauly et al., 2013), the total population size of the northwest (NW) Pacific ecosystems surrounding Japan during the last 10 years would equal half of that during the 1970s and 90% of that during the 1990s. However, catch does not necessarily reflect abundance, particularly when stock management efficiently reduces fishing mortality (Branch et al., 2011). In 1997, the Fishery Agency of Japan introduced fishery management using total allowable catches (TACs) in conjunction with traditional effort management (Makino, 2011). Without an estimation of stock abundance, it is unknown whether the decline in Japanese catches, particularly during the last 15 years after TAC implementation, can be attributed to decreased stock abundances or to TAC management. Figure 1 Open in new tabDownload slide Total amount of landings by Japanese fisheries (grey areas, catches in the northwest Pacific by Japan from FAO 2011–2016) and share of the total world’s catch (solid line). The dashed line represents the total amount of landings of the 37 stocks reviewed in this study. The main difficulty in understanding the status of Japanese fisheries based on an estimation of abundance is the absence of estimated maximum sustainable yield (MSY) reference points. The primary objective of fishery management according to the United Nations Convention on Law of the Sea (UNCLOS) is to maintain the stock level above the level that produces MSY. MSY reference points are commonly used for evaluations of stock statuses throughout the world (Hilborn and Stokes, 2010). However, an estimating MSY reference points is generally difficult, particularly due to the uncertainty in the stock-recruitment (S-R) relationship. The difficulty in estimating the S-R relationship is ubiquitous in fishery sciences and generally annoys most fishery scientists (Walters and Martell, 2004; Szuwalski et al., 2015), including many Japanese fishery scientists dealing with Japanese stock data (cf. Matsuda et al., 1991; Sakuramoto, 2005; Shimoyama et al., 2007). The harvest control rule (HCR) adopted by the Japanese Fisheries Agency also proposes MSY as a primary management reference point (Fisheries Agency and Fisheries Research Agency of Japan, 2015). However, the empirical reference points such as historical stock sizes are commonly used in Japanese stock assessment, and MSY reference points are not estimated in most cases to avoid the large uncertainty associated with their estimation. We addressed the difficulty in estimating MSY reference points by using all of the available biological information in combination with the uncertainty of the assumed S-R relationship. The estimated MSY reference points allowed us to evaluate Japanese stocks from the point of view of MSY and to compare the sustainability of Japanese fisheries with that of other fisheries throughout the world. We also investigated potential factors, such as biological parameters, wholesale prices and the application (or not) of TAC management, using a linear mixed model (LMM, Zurr et al., 2009) to determine whether these could explain the current stock statuses and historical trends. Material and methods Japanese fishery resource data Important Japanese fisheries consisting of 52 species and 84 stocks have been assessed by the Japan Fisheries Research and Education Agency (FRA) every year since 1998. These stocks are assessed because of their relatively high total weight of landings and/or wide distribution across many prefectures. The main objective of this assessment is to estimate the acceptable biological catch (ABC) for the following year and to provide indicators of the stock statuses (high, middle, or low) and trends (increase, constant or decrease). A threshold stock size between “middle” and “low” is generally derived from the reference point “Blimit” as the spawning or total biomass below which strong recruitment cohorts rarely occur due to a low spawning biomass (Fisheries Agency and Fisheries Research Agency of Japan, 2015). The biomass level is determined based on the characteristics of individual stocks mostly through visual inspection and sometimes using objective definitions (e.g. the spawning biomass achieving half of the maximum number of recruits in the assumed S-R relationship). Blimit is used as a threshold of the HCR below which the fishing mortality should present a linear decrease. In this paper, we reviewed Japanese 37 stocks with abundance estimates. The abundance of 26 stocks among these 37 stocks is estimated using classical virtual population analysis (VPA) or tuned VPA (Ichinokawa and Okamura, 2014) (Supplementary Appendix S1). The abundance of the remaining 11 stocks are estimated using species-specific population dynamics models with survey data (nine stocks), sex-specific VPA (one stock), or monthly-based VPA (one stock). The stock assessments are updated every year and are publicly available at http://abchan.fra.go.jp in Japanese. We used the results of the stock assessments published in 2015, which include stock status estimates through 2013 (Fisheries Agency and Fisheries Research Agency of Japan, 2015). Among the 37 stocks that were reviewed in this study, 16 stocks (eight species) have been managed using TACs (TAC stocks). TAC management has been implemented for Japanese sardine, Japanese jack mackerel, chub mackerel, spotted mackerel, Pacific saury, walleye pollock, and snow crab since 1997 and for Japanese flying squid since 1998. The TAC stocks need to have reliable abundance estimates and meet any of the following three criteria (Japan Fisheries Information Service Center, 2002): (1) the stock must have a high amount of landings or high consumption, (2) the abundance must be so depleted that TAC management is urgently needed, and (3) part of the stock must be utilized by foreign countries. Most of the 16 TAC stocks meet the first and third criteria. ABCs provide a basis for determining the TACs for the following fishing year. For some species, the TACs were often markedly higher than the ABCs (Matsuda et al., 2010). However, in 2008, the Fisheries Agency of Japan recommended, as suggested by an advisory panel of well-informed independent personalities, maintaining the TAC below the ABC (Anonymous, 2008), and the discrepancy between the ABC and TAC has recently diminished. The other 21 stocks are not managed by TACs (non-TAC stocks), but some are managed by various management methods often referred to as “self-management” by various management bodies, such as fishery associations. According to the stock assessment reports from 2015 (Fisheries Agency and Fisheries Research Agency of Japan, 2015), the most frequently observed tactics are restricting the size of caught fish and enhancement. Limitation of the number of fishing days and the catch per operation and the marine protected areas for spawning fish are also observed. Because of the nature of self-management, it is difficult to collect comprehensive quantitative information for all of the implemented management strategies. This study focused on the effect of the TAC management and not on other management tools, even though management tools other than TAC are often implemented in conjunction with TAC and are effective (Ichinokawa et al., 2015). MSY reference points We estimated MSY reference points for the 37 stocks with abundance estimates. We defined BMSY as the total or spawning biomass (B) that produces MSY and UMSY as the exploitation rate (U, total catch relative to the total biomass) achieving MSY. The time-series for B and U relative to BMSY (B/BMSY) and UMSY (U/UMSY), respectively, were then calculated to evaluate the statuses of stock abundance and fishing impact. For the 26 stocks assessed by classical and tuned VPA and with an estimate of the annual number of recruits and spawning biomass, we estimated MSY reference points assuming an S-R relationship and age-specific biological parameters (Supplementary Appendix S2). The following functional types of the S-R relationships were investigated: Beverton-Holt (BH) (Beverton and Holt, 1957), Ricker (RI) (Ricker, 1954), and hockey stick (HS) (Clark et al., 1985). We applied these functions to the time-series data for spawning biomass and number of recruits with appropriate offset by the age of recruitment to estimate the parameters of the functions under the assumption of lognormal error distributions. The parameters were estimated using the nls function of the statistical analysis software R (R Development Core Team, 2016). We then conducted 100-year stochastic future projections using an age-structured model assuming an S-R relationship and observed the recruitment variability to determine the UMSY and BMSY values at which the geometric average of the annual catches after 100 years is maximized. We used geometric rather than arithmetic averages because the assumption of a log-normal error distribution for recruitment variability resulted in a long-tailed distribution for the annual catches after 100 years and geometric averages provided more stable estimations of MSY reference points compared with arithmetic averages. The age-structured model and biological parameters were equal to those used in the stock assessment models (VPA) for individual stocks. For the other 11 stocks, we fitted a surplus production model to the catch and biomass time-series data (Supplementary Appendix S3). The method of fitting the surplus production model was generally similar to that used to compile the RAM Legacy database (Ricard et al., 2012). Note that the value of BMSY estimated with the production model is the “total” biomass that produces MSY in this study, whereas the value of BMSY estimated by the age-structured model is the spawning biomass. This inconsistency is inevitable due to a lack of information regarding the spawning biomass. We compared the median trends for the U, B/BMSY and U/UMSY of the 37 stocks with those of stocks around the world. The global data were derived from the RAM Legacy database (v2.5 with production model fits, http://ramlegacy.org/, Ricard et al., 2012). The BMSY and UMSY values provided in the RAM Legacy database are mixtures of estimates from the original stock assessments and those estimated by database compilers using a surplus production model, but we did not distinguish between these estimates in this analysis. We classified the world into five regions where the stock status has been reviewed relatively well: northeast (NE) Atlantic (European Union, Europe non-EU, and Mediterranean-Black Sea), northwest (NW) Atlantic (US East Coast, Canada East Coast, and US Southeast/Gulf), northeast (NE) Pacific (US Alaska, US West Coast, and Canada West Coast), southwest (SW) Pacific (Australia and New Zealand) and open ocean (Indian, Atlantic and Pacific Oceans, mainly tuna and billfishes). We also compared the life history parameter of M and the estimated UMSY of Japanese stocks with those of stocks around the world. The worldwide M and UMSY values were derived from the RAM Legacy database. Due to missing data, we used 194 and 182 data points to obtain the worldwide M and UMSY, respectively. Statistical analyses of the stock status and fishing mortality We explored factors that can affect the statuses and trends of abundances and fishing mortalities of the Japanese stocks using LMMs. The candidates of the predictors are the logarithm of natural mortality coefficient (M), the age at which 50% of the individuals are sexually mature (maturity), the maximum recorded length (max length), and the wholesale price (price). We also included the following categorical variables: habitat (1 for pelagic species and 0 for demersal species) and stock management (1 for TAC stocks and 0 for non-TAC stocks). M and maturity were derived from those used in the stock assessment models or scientific studies. The max lengths of the fish were derived from FishBase (Froese and Pauly, 2016), those of snow crab (maximum recorded carapace width) and Japanese flying squid (maximum dorsal mantle length) were obtained from Ueda et al. (2007) and Sugawara et al. (2013), respectively. The price information was derived from the database provided by the Fisheries Agency at http://www.market.jafic.or.jp/suisan/ (in Japanese). We then used log(B/BMSY) or log(U/UMSY) as the response variable yit and Xit and Zit as the above-described predictors, where i and t represent the stock and year, respectively. To facilitate the interpretation of the estimated coefficients, we defined the “year” t as the actual year (Y) minus 2013 (⁠ t=Y-2013 ⁠). The equation of the model is: (1) where α and β are the vectors of the regression effect of Xit and Zit ⁠, respectively. γi and ηi are random effects for each stock following N(0, α2) and N(0, τ2), respectively, and εit is a residual error term following N(0, υ2). The LMM can be separated into two parts: the first part, (intercept), represents log(B/BMSY) or log(U/UMSY) for the most recent year (2013), and the second part, ) (annual trend), explains the annual trends of the response variables. We performed model selection to determine the most parsimonious model using the bias-corrected Akaike information criterion (AICc; Burnham and Anderson, 2002). Considering the uncertainty of model selection, we present the parameter estimates of the averaged models with ΔAICc < 2 as well as the model that achieved the minimum AICc. Because we were particularly interested in how the TAC and/or biological characteristics affects the statuses and trends of Japanese stocks, we mainly focused on the data from the years after 1997. Results Data overview The total landings of the 37 stocks in 2013 were approximately 2.2 million metric tons (MT), representing 61% of the total marine capture production in Japan (3.6 million MT: Figure 1 and Supplementary Appendix S1). The 16 TAC stocks contributed to 75% of the total landings of the 37 stocks in 2013. The average duration of the stock abundance estimates is 27 years, and the number of the stocks with abundance estimates increased from 19 before 1990 to 31 in 1998. S-R relationship and MSY estimates When applying three different S-R functions, BH, RI and HS, to the spawning and recruitment data, the minimum AICc was achieved assuming BH in eigth, RI in nine, and HS in five stocks (Supplementary Table S2). The three functions yielded the same AICc values for three stocks because the S-R relationships in these stocks became completely linear. Both BH and HS yielded the minimum values for one stock, in which recruitment does not depend on the spawning biomass. However, the ΔAICc values obtained for the three S-R functions were less than two in many cases (20 stocks). According to Burnham and Anderson (2002), models with ΔAIC < 2 cannot be rejected statistically. These results indicate the lack of statistically strong evidence for determining a single S-R function for these stocks. We then classified the 26 stocks into five categories based on the robustness of the estimated BMSY and UMSY to the assumed S-R functions (Supplementary Table S2). The BMSY and UMSY estimates were relatively robust to the assumed S-R functions in eight stocks (four of which were managed by TACs), and these were denoted type I stocks (robust, Figure 2a). The seven type II stocks (three TAC stocks; linear, Figure 2b) show an almost linear relationship between the spawning biomass and recruitment. The BMSY estimates based on the BH and RI in these seven stocks were at least fivefold higher than the maximum historical spawning biomass, whereas HS provided realistic BMSY estimates. This result is due to the characteristics of HS: recruitment is assumed to be constant above a spawning biomass threshold to avoid extreme extrapolation within the range in which the spawning biomass or recruitment was not observed. In the six type III stocks (no TAC-manged stocks; strong density dependence in RI, Figure 2c), the estimates of BMSY based on RI were markedly lower than those based on BH and HS due to the strong density dependence estimated in RI. This strong density dependence resulted in a relatively large UMSY (0.27 for RBRMSETOW and 0.42–0.53 for the others, Supplementary Table S2) compared with their assumed M of 0.2–0.3. The three remaining TAC stocks were categorized as type IV (regime shift), in which the historical maximum spawning biomass was at least 30-fold higher than the historical minimum. These three stocks appeared to be highly subject to environmental factors, i.e. regime shift; thus, we applied the S-R function only to the recent data. A sensitivity analysis using data from different time periods did not reveal a significant effect on the conclusions presented below. The final category, type V, included the remaining two stocks that could not be categorized as any of the other types. Figure 2 Open in new tabDownload slide Examples of the stock-recruitment (S-R) relationships applied to Japanese fishery stock data. Types I (a), II (b), and III (c) are categories based on the uncertainty of the estimated maximum sustainable yield (MSY) reference points. Details of the categories are provided in the main text and Supplementary Appendix S4. Solid: hockey stick (HS); dashed: Beverton-Holt (BH); and dotted: Ricker (RI). All of the figures for the 26 stocks are shown in Supplementary Figure S3. The data points are connected based on year. Denser black points represent more recent data, and the most recent year is indicated with a cross. The downward arrows indicate the stock biomass that achieved the MSY (BMSY) and the number of recruits at BMSY (RMSY) based on each S-R function. The upward arrows indicate that BMSY and/or RMSY is out of range. In summary, almost linear S-R relationships resulted in an implausibly high (outside of the range of the historically observed spawning biomass) BMSY for seven stocks when assuming BH or RI. The strong density dependence in RI resulted in an implausibly high (compared to M) UMSY for six stocks. In contrast, we could derive “realistic” BMSY and UMSY estimates using HS. We therefore used the results derived from HS as the basis for the subsequent stock overview and statistical analysis. For the 11 stocks assessed using a production model, UMSY was positive, and BMSY ranged between the minimum and maximum historical biomass; thus, both BMSY and UMSY can be considered biologically plausible estimates for all of these stocks with the exception of a Pacific stock of snow crab (SNOWCRPAC) and a Pacific stock of cod (CODNP). Because negative intrinsic growth rates were estimated for SNOWCRPAC and CODNP, we fixed the fishing mortality rates that achieved MSY (FMSY) for these two stocks, and only K was estimated. The fixed FMSY was assumed to equal half of M because FMSY/M is generally less than 1 and is approximately 0.5 for long-lived animals (Zhou et al., 2012). Further details of the results and fitted surplus production curves are presented in Supplementary Appendix S3. Stock status The individual trends of B/BMSY and U/UMSY of the 37 assessed stocks (Supplementary Figure S4) and median trends with data coverage (Figure 3) are shown assuming different S-R functions. Roughly similar annual trends of the median B/BMSY and U/UMSY values were obtained for the different S-R assumptions: after 1997, B/BMSY tended to increase, and U/UMSY tended to decrease. Assuming HS, the median absolute values of B/BMSY ranged from 0.5 to 0.7, and those of U/UMSY ranged from 1 to 1.3. This result indicates that half of the stocks had a B/BMSY below 0.5–0.7 and a U/UMSY above 1–1.3 for the last 15 years. The BH assumption provided a more pessimistic view: half of the stocks had a B/BMSY below 0.3–0.5 and a U/UMSY above 1.2–1.7. This pessimistic view was attributed to the seven type II stocks in which an almost linear S-R relationship resulted in substantially high BMSY under BH. However, the RI assumption provided a more optimistic view: half of the stocks had a B/BMSY below 0.6–0.9 and a U/UMSY above 0.8–1.2. This optimistic view was attributed to the six type III stocks in which a strong density dependence was estimated under RI. Figure 3 Open in new tabDownload slide Median of historical biomass (B) relative to the BMSY (B/BMSY, a) and exploitation rates (U, total catch relative to the total biomass) relative to the U achieving MSY (U/UMSY, b) among the 37 Japanese stocks assessed. The assumed S-R functions are HS, BH and RI. The grey areas show the proportion of stocks where abundance estimates are available [# of stocks with abundance estimates/all 37 stocks, %]). The current stock status (geometric mean of the most recent 3 years) indicates that approximately half of the stocks fall into the category of “overfishing” (tentatively defined by U/UMSY > 1) and that half are classified as “overfished” (tentatively defined by B/BMSY < 0.5) regardless of the assumed S-R functions (Figures 4 and Supplementary Figure S5). The actual percentages of overfishing and overfished stocks varied depending on the assumed S-R functions. Under HS, BH and RI, 54, 59, and 44% of the stocks, respectively, suffered from overfishing, and 38, 54, and 30%, respectively, were overfished. The current status determined using the BH and HS S-R functions of the TAC stocks appeared to be better than that of the non-TAC stocks. The RI S-R function yielded similar statuses for the TAC and non-TAC stocks because the stock statuses of the five type III non-TAC stocks improved when using RI instead of BH or HS. The total estimated MSY under the assumption of HS was approximately 3.5 million metric tons (Supplementary Table S2), and the total MSYs of the stocks under the conditions of B2011-2013 < BMSY and B2011-2013 > BMSY were 1.8 and 1.7 metric tons, respectively. Figure 4 Open in new tabDownload slide Most recent status (geometric mean from 2011 to 2013) of the 37 stocks in terms of B/BMSY and U/UMSY under the assumption of the HS S-R function. The positions of the medians of theTAC, non-TAC and all stocks are shown by shaded symbols. The grey figures are the percentages of stocks in the area divided by 0.5 and 1 B/BMSY and by 1 U/UMSY. Comparison with other fisheries throughout the world The median trends of the B/BMSY and U/UMSY of the 37 stocks determined based on HS were compared with those of other five regions of the world (Figure 5a and b). The results revealed that Japanese fisheries were similar to those of the NE Atlantic and presented the most overfished and overfishing conditions among all of the regions investigated. When we compared the stock status of TAC or non-TAC stocks, the median B/BMSY of the TAC stocks in the most recent year of 2013 was comparable to that of the more successful regions, namely SW and NE Pacific. The median U/UMSY of the TAC stocks appeared to decrease at greater rates than those of the non-TAC stocks and gradually approached these of the more successful regions. In contrast, the status of the non-TAC stocks was by far the worst. Figure 5 Open in new tabDownload slide Comparison of the statuses and trends of 37 Japanese stocks based on the HS S-R function with those of five regions of the world: SW Pacific, NW Atlantic, NE Atlantic, NE Pacific and open ocean. Median B/BMSY (a); Median U/UMSY (b); Median U (c). Regardless of the similarity of the median U/UMSY and B/BMSY of Japan to those of the NE Atlantic, the median U (not relative to UMSY) obtained for Japan stood out from those of the other regions (Figure 5c). Although the median U for the Japanese stocks was almost constantly greater than 0.3, no other regions presented such a high median U. This result is attributed to the biological characteristics of the 37 Japanese stocks (Figure 6), which tended to have a relatively high M and consequently a high UMSY compared with those of the stocks of other regions of the world. The median M for the worldwide stocks derived from the RAM Legacy database was 0.17, whereas that of the 37 Japanese stocks was 0.35. Because of the generally positive relationship observed between M and UMSY (Figure 6c, Zhou et al., 2012), the UMSY of the Japanese stocks was also relatively high compared with that of the other regions (Figure 6b). The median UMSY of the worldwide stocks derived from the RAM Legacy database was 0.13, whereas that of the 37 Japanese stocks was 0.27. Figure 6 Open in new tabDownload slide Natural mortality coefficients (M) (a) and UMSY (b) of the Japanese and worldwide fisheries and correlation between M and UMSY (c). The x-axis is the range of M (a) and UMSY (b). In (c), the points by circles refer to Japanese fisheries, and the points by crosses refer to worldwide fisheries. Factors affecting stock status and fishing mortality The results of the LMMs are shown in Tables 1 and 2. To investigate the intercept of the B/BMSY model (Table 1), max length with negative coefficients was included in the minimum AICc model, which indicates that stocks with larger max lengths tended to have a smaller B/BMSY (i.e. were more vulnerable). Regarding the annual trend of the B/BMSY model, no interactions with the year effect were included in the minimum AICc model, but the main effect of “year” was included with a positive coefficient of 0.02. Because the lower limit of the 90% confidence interval (CI) was almost 0 in the minimum AICc model and negative, at –0.02, in the averaged model, this effect was not strongly supported. Nevertheless, the positive coefficient indicates that the B/BMSY has gradually increased since 1998. Table 1 Parameters estimated using liner mixed models (LMMs) with the response variables log (B/BMSY). . . AICc minimum model . Averaged model . . . Estimate . 5% . 95% . Estimate . 5% . 95% . Intercept (Intercept) 3.22 1.96 4.48 2.98 1.32 4.64 TAC 0.38 –0.06 0.82 Pelagic –0.49 –1.01 0.03 M 0.36 –0.11 0.83 Maturity –0.19 –0.51 0.14 Max length –0.96 –1.27 –0.64 –0.88 –1.26 –0.49 Price 0.14 –0.07 0.35 Year: 0.02 0.002 0.03 0.01 –0.02 0.04 TAC Pelagic 0.02 –0.01 0.06 M –0.02 –0.05 0.01 Maturity Max length 0.01 –0.02 0.03 Price . . AICc minimum model . Averaged model . . . Estimate . 5% . 95% . Estimate . 5% . 95% . Intercept (Intercept) 3.22 1.96 4.48 2.98 1.32 4.64 TAC 0.38 –0.06 0.82 Pelagic –0.49 –1.01 0.03 M 0.36 –0.11 0.83 Maturity –0.19 –0.51 0.14 Max length –0.96 –1.27 –0.64 –0.88 –1.26 –0.49 Price 0.14 –0.07 0.35 Year: 0.02 0.002 0.03 0.01 –0.02 0.04 TAC Pelagic 0.02 –0.01 0.06 M –0.02 –0.05 0.01 Maturity Max length 0.01 –0.02 0.03 Price The results of the model achieving the minimum AICc (AICc minimum model) and the average of the models with ΔAICc < 2 (averaged model) are shown. Open in new tab Table 1 Parameters estimated using liner mixed models (LMMs) with the response variables log (B/BMSY). . . AICc minimum model . Averaged model . . . Estimate . 5% . 95% . Estimate . 5% . 95% . Intercept (Intercept) 3.22 1.96 4.48 2.98 1.32 4.64 TAC 0.38 –0.06 0.82 Pelagic –0.49 –1.01 0.03 M 0.36 –0.11 0.83 Maturity –0.19 –0.51 0.14 Max length –0.96 –1.27 –0.64 –0.88 –1.26 –0.49 Price 0.14 –0.07 0.35 Year: 0.02 0.002 0.03 0.01 –0.02 0.04 TAC Pelagic 0.02 –0.01 0.06 M –0.02 –0.05 0.01 Maturity Max length 0.01 –0.02 0.03 Price . . AICc minimum model . Averaged model . . . Estimate . 5% . 95% . Estimate . 5% . 95% . Intercept (Intercept) 3.22 1.96 4.48 2.98 1.32 4.64 TAC 0.38 –0.06 0.82 Pelagic –0.49 –1.01 0.03 M 0.36 –0.11 0.83 Maturity –0.19 –0.51 0.14 Max length –0.96 –1.27 –0.64 –0.88 –1.26 –0.49 Price 0.14 –0.07 0.35 Year: 0.02 0.002 0.03 0.01 –0.02 0.04 TAC Pelagic 0.02 –0.01 0.06 M –0.02 –0.05 0.01 Maturity Max length 0.01 –0.02 0.03 Price The results of the model achieving the minimum AICc (AICc minimum model) and the average of the models with ΔAICc < 2 (averaged model) are shown. Open in new tab Table 2 Parameters estimated using linear mixed models (LMMs) with the response variable log (U/UMSY). . . AICc minimum model . Averaged model . . . Estimate . 5% . 95% . Estimate . 5% . 95% . Intercept (Intercept) –0.92 –2.02 0.18 –1.03 –2.22 0.17 TAC –0.65 –0.90 –0.40 –0.62 –0.88 –0.36 Pelagic M –0.62 –0.98 –0.26 –0.53 –0.96 –0.10 Maturity –0.36 –0.61 –0.11 –0.32 –0.58 –0.06 Max length 0.46 0.21 0.72 0.51 0.22 0.80 Price –0.19 –0.29 –0.09 –0.20 –0.31 –0.09 Year: –0.08 –0.15 –0.01 –0.09 –0.19 0.01 TAC –0.02 –0.04 0.00 –0.02 –0.05 0.00 Pelagic M 0.02 0.00 0.05 Maturity –0.02 –0.03 0.00 –0.02 –0.03 0.00 Max length 0.02 0.00 0.04 0.03 0.00 0.05 Price . . AICc minimum model . Averaged model . . . Estimate . 5% . 95% . Estimate . 5% . 95% . Intercept (Intercept) –0.92 –2.02 0.18 –1.03 –2.22 0.17 TAC –0.65 –0.90 –0.40 –0.62 –0.88 –0.36 Pelagic M –0.62 –0.98 –0.26 –0.53 –0.96 –0.10 Maturity –0.36 –0.61 –0.11 –0.32 –0.58 –0.06 Max length 0.46 0.21 0.72 0.51 0.22 0.80 Price –0.19 –0.29 –0.09 –0.20 –0.31 –0.09 Year: –0.08 –0.15 –0.01 –0.09 –0.19 0.01 TAC –0.02 –0.04 0.00 –0.02 –0.05 0.00 Pelagic M 0.02 0.00 0.05 Maturity –0.02 –0.03 0.00 –0.02 –0.03 0.00 Max length 0.02 0.00 0.04 0.03 0.00 0.05 Price The caption is the same as that of Table 1. Open in new tab Table 2 Parameters estimated using linear mixed models (LMMs) with the response variable log (U/UMSY). . . AICc minimum model . Averaged model . . . Estimate . 5% . 95% . Estimate . 5% . 95% . Intercept (Intercept) –0.92 –2.02 0.18 –1.03 –2.22 0.17 TAC –0.65 –0.90 –0.40 –0.62 –0.88 –0.36 Pelagic M –0.62 –0.98 –0.26 –0.53 –0.96 –0.10 Maturity –0.36 –0.61 –0.11 –0.32 –0.58 –0.06 Max length 0.46 0.21 0.72 0.51 0.22 0.80 Price –0.19 –0.29 –0.09 –0.20 –0.31 –0.09 Year: –0.08 –0.15 –0.01 –0.09 –0.19 0.01 TAC –0.02 –0.04 0.00 –0.02 –0.05 0.00 Pelagic M 0.02 0.00 0.05 Maturity –0.02 –0.03 0.00 –0.02 –0.03 0.00 Max length 0.02 0.00 0.04 0.03 0.00 0.05 Price . . AICc minimum model . Averaged model . . . Estimate . 5% . 95% . Estimate . 5% . 95% . Intercept (Intercept) –0.92 –2.02 0.18 –1.03 –2.22 0.17 TAC –0.65 –0.90 –0.40 –0.62 –0.88 –0.36 Pelagic M –0.62 –0.98 –0.26 –0.53 –0.96 –0.10 Maturity –0.36 –0.61 –0.11 –0.32 –0.58 –0.06 Max length 0.46 0.21 0.72 0.51 0.22 0.80 Price –0.19 –0.29 –0.09 –0.20 –0.31 –0.09 Year: –0.08 –0.15 –0.01 –0.09 –0.19 0.01 TAC –0.02 –0.04 0.00 –0.02 –0.05 0.00 Pelagic M 0.02 0.00 0.05 Maturity –0.02 –0.03 0.00 –0.02 –0.03 0.00 Max length 0.02 0.00 0.04 0.03 0.00 0.05 Price The caption is the same as that of Table 1. Open in new tab Regarding the intercept of the U/UMSY model (Table 2), the AICc minimum model included the effects of TAC, M, maturity, max length, and price but not pelagic. This result indicates that TAC stocks tended to have a smaller U/UMSY when considering other biological factors. Similar results were observed for the annual trend of U/UMSY: the AICc minimum model included the biological factors of maturity and max length and the effect of TAC with a negative coefficient of –0.02. The interaction effect of year and TAC of –0.02 demonstrates that the reduction rate of the TAC stocks was higher than that of the other stocks. However, because the 90% CI of the interaction effect of year and TAC was –0.04 to 0.00 in the AICc minimum model and –0.05 to 0.00 in the averaged model, the uncertainty of the interaction effect was relatively high. The estimated annual trends of U and B for each stock in the LMMs, corresponding to the term in Eq. 1, show a negative relationship (Figure 7). Thirty-one stocks show the negative trend of U, and the median values were –0.025 for TAC stocks (2.5% reduction in U per year) or –0.012 for non-TAC stocks (1.2% reduction in U per year). Twenty-two stocks show the positive trend of B and the median value was 0.011 (1% increase in B pear year). The slope and intercept of the simple regression of B against U trends with excluding the outlier of SNOWCRPAC were –1.03 (p = 0.011) and –0.002 (p = 0.863), respectively, which indicates that a reduction in U can potentially increase biomass although there is the large variance in the data. Figure 7 Open in new tabDownload slide Annual trends of U (U trend) and B (B trend) estimated using the LMMs for each stock. The solid diagonal line is the predicted regression line for the B trend against the U trend, and broken lines show the 90% prediction interval. The regression line was fitted to all of the points with the exception of an outlier at –0.20 along the x-axis and 0 along the y-axis (for SNOWCRPAC), which U was substantially reduced by the Great East Japan Earthquake while B did not increased probably due to the predation of cod increasing after the earthquake (Hattori et al., 2016). A sensitivity analysis of the LMMs was conducted with BMSY and UMSY assuming BH and RI instead of HS (Supplementary Appendix S7). Although the parameter estimates on annual trends were similar to the results obtained using HS, the factors that explain the intercept of B/BMSY in the minimum AICc model differred under the BH and RI assumptions. Assuming BH and RI, the main factors that affected B/BMSY were M and maturity, whereas max length was the main factor under HS, probably due to the potential relationship among M, maturity and max length. However, the qualitative results did not change substantially, even assuming different S-R functions. In other words, B/BMSY has slightly but significantly increased over the past 15 years, and U/UMSY has significantly decreased, particularly for the TAC stocks. Discussion Current stock status in Japan The current stock status of 37 Japanese stocks is not very good because approximately half of the stocks are in the “overfishing” (U/UMSY > 1) range and half are “overfished” (B/BMSY < 0.5) (Figure 4 and Supplementary Figure S5). However, LMMs indicated that the fishing mortality significantly decreased and that the total biomass slightly increased over the past 15 years (Tables 1 and 2). In particular, the rate of decrease in the fishing mortality in the stocks that were managed by TACs was greater than that of the other stocks (Table 2). These results provide complementary information to proceeding studies that review statuses and trends of fisheries in various regions that have undergone stock assessment (Worm et al., 2009; Costello et al., 2012; Hilborn and Ovando, 2014) and show the effectiveness of TAC management (Melnychuk et al., 2013; Pons et al., 2017). Although the main analysis was conducted assuming an HS S-R relationship, the sensitivity analyses using MSY reference points that were estimated with BH and RI (Figure 3 and Supplementary Appendixes S6 and S7) also yielded conclusions similar to those described above. The higher rates of decrease of U obtained for the TAC stocks compared with those found for the non-TAC stocks are particularly important, as this study details the world’s first quantitative evaluation of Japanese fishery management using TAC. In addition, U decreased not only in the TAC stocks but also in the non-TAC stocks. The combination of various management tools, such as input control, is common in Japanese self-management even for the non-TAC stocks (Makino, 2011; Ichinokawa et al., 2015). Therefore, the decrease in U obtained for the non-TAC stocks can be considered the outcome of such self-management using management tools other than TACs. We could not incorporate the effect of self-management into the statistical analysis of explanatory variables due to insufficient information, but a collection of all of the management efforts and an evaluation of the quantitative effects are needed in future studies. The abundances of the TAC stocks did not significantly increase at greater rates than those of the non-TAC stocks, whereas the overall abundance slightly increased. An insufficient reduction in U is the first possible reason for this result. The significant negative relationship between the trends of U and B (Figure 7) indicates that we can expect an increase in stock abundance as an outcome of the reduction in fishing mortality. However, a large variation was observed in this relationship. According to the estimated prediction interval shown in Figure 7, the achievement of stock recovery, i.e. obtaining a positive B trend, with a 90% probability requires an annual reduction in U of –0.066 for 15 years, i.e. at least 6% per year. The large variation in this relationship might obscure the effect of a greater reduction in U obtained for the TAC stocks on abundance recovery. This large variation could be attributed to various factors, such as time-lag in the response of abundance to a reduction in fishing intensity, environmental effects on recruitment deviation, regime shift, and uncertainty of the catch statistics from foreign countries. Our statistical analysis indicates that B/BMSY tended to be smaller in the stocks with a larger max length (Table 1). The negative effect of max length on B/BMSY is supported by previous studies using the RAM Legacy database (Costello et al., 2012; Thorson et al., 2012; Komori et al., 2016). These previous studies reported significant effects of species category, max length, and trophic level on B/BMSY or the probability of B/BMSY being less than 0.2. We did not incorporate the trophic level because it appeared to correlate with max length and other biological parameters, and such information for crabs and squids is not available. Pinsky et al. (2011) and Pinsky and Byler (2015) reported possible interaction effects between growth rates and U, but our dataset was not large enough to detect such complex interaction effects. According to our finding that approximately half of the 37 stocks are overfished, we can expect a substantial increase in the future yield if appropriate management measures are applied to these overfished stocks. Currently, the total catch in 2013 for the overfished stocks (B2011-2013/BMSY < 0.5) among the 37 stocks was approximately 1.1 million MT, whereas the total MSY of these stocks was 1.8 million MT. We can theoretically expect an additional yield of 0.7 million MT if the overfished conditions of these stocks are improved. Japan had the fifth largest marine production of 3.6 million MT in 2014 and one of the greatest consumers of marine products (> 60 kg per capita per year, corresponding to 7.2 million MT) throughout the world (FAO, 2016). Japan relies on imports for the shortfall, but the potential addition of 0.7 metric tons can keep up with approximately 1/4 of the shortfall, which would largely affect the trade flows of marine products and food provision at the global scale. Japan in the world fisheries This study clearly identified some characteristics of Japanese fishery resources through a comparison with those from fisheries around the world. A higher M and consequently a higher UMSY (Figure 6) were the first noteworthy characteristics. These characteristics can be attributed to the large contribution of small pelagic fish and squid to Japanese fishery resources. Sixteen of the 37 assessed stocks consisted of small pelagic fish or squid and contributed 79% of the total landings in 2013. Note that having a majority of small pelagic fish does not mean that “fishing down” (Pauly et al., 1998) has occurred in recent years because the mean trophic level of Japanese catches did not show a long-term decline (Matsuda et al., 2010). The large contribution of small pelagic resources to Japanese stocks with relatively high natural mortality, i.e. productivity, suggests that Japanese fisheries have great potential to exhibit quick recovery and therefore increase their yield through an adjustment of fishing intensity to an appropriate level. We could also characterize the coefficient of variation (CV) of recruitment deviation in Japanese fishery resources (Supplementary Table S2). Thorson et al. (2014) reported that the average CV of recruitment deviation in the world is 0.74 (SD = 0.34), whereas that in Japan is 0.39 when assuming HS or BH (SD = 0.2 or 0.21, respectively), and 0.38 when assuming RI (SD = 0.2). The average first-order autocorrelation coefficients of residuals were 0.36 (SD = 0.26), which are lower than those of the world (0.43, SD = 0.28) (Thorson et al., 2014). In general, the recruitment variation is considered higher in pelagic fish than in demersal fish, and interestingly, the CVs of recruitment deviation in Japanese stocks dominated by pelagic species are lower than those of other fisheries around the world on average. Thorson et al. (2014) also found no significant differences in the CVs of recruitment deviation among taxonomic groups. We expect the fishery management of stocks with a smaller CV in recruitment deviation to be more tractable. A HCR considering these advantages, lower CVs in recruitment deviation and high productivity, is required to utilize the great potential of Japanese fisheries stocks. The third characteristic is large uncertainties in the S-R relationships and MSY reference points, although similar uncertainty is generally observed in any fishery stock. In half of the 26 Japanese stocks examined, it was difficult to apply the commonly used S-R functions of BH and RI due to an implausibly high BMSY in seven stocks (type-II) and an implausibly high UMSY in six stocks (type-III) (Supplementary Table S2). The difficulty in estimating MSY reference points when using BH or RI, particularly with stocks in which an almost linear S-R relationship is estimated, has been a challenge for many Japanese fishery scientists (cf. Matsuda et al., 1991; Sakuramoto, 2005; Shimoyama et al., 2007). Faced with the large uncertainty of MSY reference points in BH and RI, we utilized the HS function and evaluated the Japanese stock statuses based on HS S-R relationship. HS could provide a biologically and empirically plausible estimation (within the range of historical spawning biomass) of MSY reference points, even for stocks for which a naive application of BH or RI is difficult, because we can avoid extreme extrapolation. This characteristic of HS is preferred and HS has thus been recommended as an alternative SR function in many studies (Clark et al., 1985; Barrowman and Myers, 2000; Mesnil and Rochet, 2010). HS can also overcome the problem of predicting an increase in the number of recruits per spawner at low population sizes in BH and RI (Barrowman and Myers, 2000). However, HS has some drawbacks. For example, the nondifferentiability of the HS function frequently results in multiple local minima on the likelihood surface, and the grid search method is recommended (Barrowman and Myers, 2000). In addition, when recruitment appears to be independent on spawning biomass or increase almost linearly against spawning biomass, we cannot determine the single break point of HS based on likelihood. In these cases, we cannot help but arbitrarily set the minimum or maximum historical spawning biomass as the break point of HS. In particular, the almost linear S-R relationship might suggest that these stocks have been so depleted that we cannot observe recruitment compensation within the observed range. Therefore, BMSY estimates under the HS assumption would serve as a lower limit of the possible range of BMSY. The large uncertainty in the estimation of MSY reference points frequently forces us to use many adhoc assumptions in stock assessments to derive a tactical management target or use an implausible management target. Although the use of MSY reference points as the primary management reference is now widely accepted, the concept to MSY itself has a history of near-death and reincarnation (Larkin, 1977; Mace, 2001). The perception of FMSY as a limit to be avoided rather than a target is a solution to the use of the concept in fisheries management when considering a precautionary approach under the potential uncertainty and the ecosystem and economy points of view (Mace, 2001). Although we used MSY reference points according to their use in other studies evaluating global fisheries trends (Worm et al., 2009; Costello et al., 2012), MSY is not a single management target. Individual management authorities would take into account a wide range of management targets considering inter-species relationships and ecosystem aspects. The statuses and trends of fisheries differ among the different regions of the world. Forecasting the worldwide future of fisheries is an interesting issue for many fishery scientists and also for the public (Worm et al., 2006, 2009). Forecasting should be based on the accumulation of knowledge on individual regional fisheries around the world. However, Asian fisheries have never been sufficiently reviewed using stock abundance estimates and have not been included as part of the “world” in some high-impact studies evaluating the world’s fishery trends, such as studies by Worm et al. (2009), Costello et al. (2012) and Neubauer et al. (2013). However, the most recent study by Costello et al. (2016) suggests the great potential of Asian fisheries. Japanese fishery resources are on their way to reaching the target level with respect to MSY reference points, but we expect that this quantitative evaluation will aid future improvements in the fishery and management systems in Japan and help us understand the global stock status of fisheries in more detail. Acknowledgements We are very grateful to all of the scientists who were involved in the stock assessments of the 37 fishery stocks in the FRA and the local fishery institutes. These assessments are funded by the Fisheries Agency of Japan. Our sincere thanks are extended to Dr. Y. Oozeki, three anonymous reviewers and the associate editor Dr. Shijie Zhou for the useful and constructive comments provided, which greatly improved the manuscript. This work was supported by CREST, JST and JSPS KAKENHI Grant Number JP26520313. Supplementary data Supplementary material is available at the ICESJMS online version of the manuscript. References Anonymous . 2008 . Summary report of advisory panel for the current problem and future improvement on total allowable catch system and individual quota. Fisheries Agency of Japan. http://www.jfa.maff.go.jp/j/suisin/s_yuusiki/pdf/matome_01.pdf (last accessed 8 August 2016) (in Japanese). Barrowman N. J. , Myers R. A. 2000 . Still more spawner-recruitment curves: the hockey stick and its generalizations . Canadian Journal of Fisheries and Aquatic Sciences , 57 : 665 – 676 . Google Scholar Crossref Search ADS WorldCat Beverton R. J. H. , Holt S. J. 1957 . On the dynamics of exploited fish populations. Her Majesty’s Stationary Office , London . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Branch T. A. , Jensen O. P., Ricard D., Ye Y., Hilborn R. 2011 . Contrasting global trends in marine fishery status obtained from catches and from stock assessments . Conservation Biology , 25 : 777 – 786 . Google Scholar Crossref Search ADS PubMed WorldCat Burnham K. P. , Anderson D. R. 2002 . Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach . Springer Verlag , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Clark C. W. , Charles A. T., Beddington J. R., Mangel M. 1985 . Optimal capacity decisions in a developing fishery . Marine Resource Economics , 2 : 25 – 53 . Google Scholar Crossref Search ADS WorldCat Costello C. , Ovando D., Clavelle T., Strauss C. K., Hilborn R., Melnychuk M. C., Branch T. A. et al. 2016 . Global fishery futures under contrasting management regimes . Proceedings of the National Academy of Sciences of the United States of America , 113 : 5125 – 5129 . Google Scholar Crossref Search ADS PubMed WorldCat Costello C. , Ovando D., Hilborn R., Gaines S. D., Deschenes O., Lester S. E. 2012 . Status and solutions for the world’s unassessed fisheries . Science , 338 : 517 – 520 . Google Scholar Crossref Search ADS PubMed WorldCat FAO . 2016 . The state of world. Fisheries and Aquaculture. FAO, Rome. http://www.fao.org/3/a-i5555e.pdf (last accessed 8 August 2016). Fernandes P. G. , Cook R. M. 2013 . Reversal of fish stock decline in the northeast Atlantic . Current Biology , 23 : 1432 – 1437 . Google Scholar Crossref Search ADS PubMed WorldCat Fisheries Agency and Fisheries Research Agency of Japan . 2015 . Marine fisheries stock assessment and evaluation for Japanese waters (fiscal year 2014/2015). http://abchan.fra.go.jp/digests26/index.html (last accessed 8 August 2016) (in Japanese). Froese R. , Pauly D. 2016 . Fishbase. www.fishbase.org version (last accessed June 2016). Hattori T. , Shibata Y., Narimatsu Y., Nagao J. 2016 . Stock assessment of the Pacific stock of snow crab in fiscal 2015 year. In Marine Fisheries Stock Assessment and Evaluation for Japanese Waters . Fisheries Agency of Japan and Fisheries Research Agency , Japan , pp 504 – 544 . http://abchan.fra.go.jp/digests26/index.html (last accessed 8 August 2016) (in Japanese). Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Hilborn R. , Ovando D. 2014 . Reflections on the success of traditional fisheries management . ICES Journal of Marine Science , 71 : 1040 – 1046 . Google Scholar Crossref Search ADS WorldCat Hilborn R. , Stokes K. 2010 . Defining overfished stocks: have we lost the plot? Fisheries , 35 : 113 – 120 . Google Scholar Crossref Search ADS WorldCat Ichinokawa M. , Okamura H. 2014 . Review of stock evaluation methods using VPA for fishery stocks in Japan: implementation with R . Bulletin of the Japanese Society of Fisheries Oceanography , 78 : 104 – 113 . in Japanese with English abstract). Google Scholar OpenURL Placeholder Text WorldCat Ichinokawa M. , Okamura H., Watanabe C., Kawabata A., Oozeki Y. 2015 . Effective time closures: quantifying the conservation benefits of input control for the Pacific chub mackerel fishery . Ecological Applications , 25 : 1566 – 1584 . Google Scholar Crossref Search ADS PubMed WorldCat Japan Fisheries Information Service Center . 2002 . What Is Total Allowable Catch (TAC)? Fisheries Agency of Japan. http://www.jfa.maff.go.jp/j/suisin/s_tac/pdf/tacpanfu201501.pdf (last accessed 8 August 2016) (in Japanese). Komori O. , Eguchi S., Ikeda S., Okamura H., Ichinokawa M., Nakayama S. 2016 . An asymmetric logistic regression model for ecological data . Methods in Ecology & Evolution , 7 : 249 – 260 . Google Scholar Crossref Search ADS WorldCat Mace P. M. 2001 . A new role for MSY in single-species and ecosystem approaches to fisheries stock assessment and management . Fish and Fisheries , 2 : 2 – 32 . Google Scholar Crossref Search ADS WorldCat Makino M. 2011 . Fisheries Management in Japan: Its Institutional Features and Case Studies . Springer Verlag , Dordrecht, Netherlands . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Matsuda H. , Makino M., Tomiyama M., Gelcich S., Castilla J. C. 2010 . Fishery management in Japan . Ecological Research , 25 : 899 – 907 . Google Scholar Crossref Search ADS WorldCat Matsuda H. , Wada T., Takeuchi Y., Matsumiya Y. 1991 . Alternative models for species replacement of pelagic fishes . Researches on Population Ecology , 33 : 41 – 56 . Google Scholar Crossref Search ADS WorldCat Melnychuk M. C. , Banobi J. A., Hilborn R. 2013 . Effects of management tactics on meeting conservation objectives for Western North American groundfish fisheries . PLOS ONE , 8 : e56684. Google Scholar Crossref Search ADS PubMed WorldCat Mesnil B. , Rochet M. 2010 . A continuous hockey stick stock-recruit model for estimating MSY reference points . ICES Journal of Marine Science , 67 : 1780 – 1784 . Google Scholar Crossref Search ADS WorldCat Larkin P. A. 1977 . An epitaph for the concept of maximum sustained yield . Transaction of the American Fisheries Society , 106 : 1 – 11 . Google Scholar Crossref Search ADS WorldCat Neubauer P. , Jensen O. P., Hutchings J. A., Baum J. K. 2013 . Resilience and recovery of overexploited marine populations . Science , 340 : 347 – 349 . Google Scholar Crossref Search ADS PubMed WorldCat Pauly D. , Christensen V., Dalsgaard J., Froese R., Torres F. 1998 . Fishing down marine food webs . Science , 279 : 860 – 863 . Google Scholar Crossref Search ADS PubMed WorldCat Pauly D. , Hilborn R., Branch T. A. 2013 . Fisheries: does catch reflect abundance? Nature , 494 : 303 – 306 . Google Scholar Crossref Search ADS PubMed WorldCat Pinsky M. L. , Byler D. 2015 . Fishing, fast growth and climate variability increase the risk of collapse . Proceedings of the Royal Society Biological Sciences , 282 : 20151053. Google Scholar Crossref Search ADS PubMed WorldCat Pinsky M. L. , Jensen O. P., Ricard D., Palumbi S. R. 2011 . Unexpected patterns of fisheries collapse in the world’s oceans . Proceedings of the National Academy of Sciences of the United States of America , 108 : 8317 – 8322 . Google Scholar Crossref Search ADS PubMed WorldCat Pons M. , Branch T. A., Melnychuk M. C., Jensen O. P., Brodziak J., Fromentin J. M., Harley S. J. et al. 2017 . Effects of biological, economic and management factors on tuna and billfish stock status . Fish and Fisheries , 18: 1–21. Google Scholar OpenURL Placeholder Text WorldCat R Development Core Team . 2016 . R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing , Vienna . http://www.r-project.org (last accessed 8 August 2016). Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Ricard D. , Minto C., Jensen O. P., Baum J. K. 2012 . Examining the knowledge base and status of commercially exploited marine species with the RAM Legacy Stock Assessment Database . Fish and Fisheries , 13 : 380 – 398 . Google Scholar Crossref Search ADS WorldCat Ricker W. E. 1954 . Stock and recruitment . Journal of the Fisheries Research Board of Canada , 11 : 559 – 623 . Google Scholar Crossref Search ADS WorldCat Sakuramoto K. 2005 . Does the Ricker or Beverton and holt type of stock-recruitment relationship truly exist? Fisheries Science , 71 : 577 – 592 . Google Scholar Crossref Search ADS WorldCat Shimoyama S. , Sakuramoto K., Suzuki N. 2007 . Proposal for stock-recruitment relationship for Japanese sardine Sardinops melanostictus in North-Western Pacific . Fisheries Science , 73 : 1035 – 1041 . Google Scholar Crossref Search ADS WorldCat Sugawara M. , Yamashita N., Sakaguchi K., Sato T., Sawamura M., Yasue N., Mori K., Fukuwaka M. 2013 . Effects of hatch timing and sex on growth of the winter-spawning stock of Japanese common squid Todarodes pacificus migrating in the Pacific Ocean . Nippon Suisan Gakkaishi , 79 : 823 – 831 . in Japanese with English abstract). Google Scholar Crossref Search ADS WorldCat Szuwalski C. S. , Vert-pre K. A., Punt A. E., Branch T. A., Hilborn R. 2015 . Examining common assumptions about recruitment: a meta-analysis of recruitment dynamics for worldwide marine fisheries . Fish and Fisheries , 16 : 633 – 648 . Google Scholar Crossref Search ADS WorldCat Thorson J. , Branch T., Jensen O., Quinn T. 2012 . Using model-based inference to evaluate global fisheries status from landings, location, and life history data . Canadian Journal of Fisheries and Aquatic Sciences , 69 : 645 – 655 . Google Scholar Crossref Search ADS WorldCat Thorson J. T. , Jensen O. P., Zipkin E. F., Rose K. 2014 . How variable is recruitment for exploited marine fishes? A hierarchical model for testing life history theory . Canadian Journal of Fisheries and Aquatic Sciences , 71 : 973 – 983 . Google Scholar Crossref Search ADS WorldCat Ueda Y. , Ito M., Hattori T., Narimatsu Y., Fujiwara K., Yoshida T., Kitagawa D. 2007 . Growth of the snow crab Chionoecetes opilio estimated by carapace width frequency analysis in the waters off the Pacific coast of northern Honshu, Japan . Nippon Suisan Gakkaishi , 73 : 487 – 494 . in Japanese with English abstract). Google Scholar Crossref Search ADS WorldCat Walters C. J. , Martell S. J. D. 2004 . Fisheries Ecology and Management . Princeton University Press , Princeton . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Worm B. , Barbier E. B., Beaumont N., Duffy J. E., Folke C., Halpern B. S., Jackson J. B. et al. 2006 . Impacts of biodiversity loss on ocean ecosystem services . Science , 314 : 787 – 790 . Google Scholar Crossref Search ADS PubMed WorldCat Worm B. , Hilborn R., Baum J. K., Branch T. A., Collie J. S., Costello C., Fogarty M. J. et al. 2009 . Rebuilding global fisheries . Science , 325 : 578 – 585 . Google Scholar Crossref Search ADS PubMed WorldCat Zhou S. , Yin S., Thorson J., Smith A. M., Fuller M., Walters C. J. 2012 . Linking fishing mortality reference points to life history traits: an empirical study . Canadian Journal of Fisheries and Aquatic Sciences , 69 : 1292 – 1301 . Google Scholar Crossref Search ADS WorldCat Zurr A. , Ieno E. N., Walker N., Saveliev A. A., Smith G. M. 2009 . Mixed Effects Models and Extensions in Ecology with R . Springer Verlag , New York . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC © International Council for the Exploration of the Sea 2017. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. © International Council for the Exploration of the Sea 2017.

Raybaud,, Virginie;Bacha,, Mahmoud;Amara,, Rachid;Beaugrand,, Grégory

doi: 10.1093/icesjms/fsx003pmid: N/A

Anthropogenic climate change is already affecting marine ecosystems and the responses of living-resources to warming waters are various, ranging from the modifications in the abundance of key species to phenologic and biogeographic shifts. Here, we used a recently developed Ecological Niche Model (ENM) to evaluate the potential effects of global climate change on the future geographical distribution of the European anchovy. We first modelled the ecological niche (sensu Hutchinson) of the fish and projected its future spatial range using new IPCC representative concentration pathways (RCPs) scenarios and five of the latest generation of ocean-atmosphere global circulation models. We chose this multi-model and multi-scenario approach to evaluate the range of possible trajectories until the end of the century. Our projections indicate that substantial poleward shifts in the probability of anchovy occurrence are very likely and highlight areas where European anchovy fisheries are forecasted to change most. Whatever the warming scenario, our results project a reduction in the probability of occurrence in all the regions located under 48°N and an increase in more northern areas. However, increases or decreases in the probability of occurrence are greater under the “business-as-usual” scenario RCP8.5 than under the low-emission scenario RCP2.6. Introduction Climate warming is causing major changes in both terrestrial and marine ecosystems and assessing the responses of living organisms to warmer environmental conditions has become a growing concern for scientists. Species geographical distribution is highly correlated with environmental conditions and many studies showed climate-induced range shifts in numerous species belonging to a large variety of taxonomic groups (Hickling et al., 2006). Range shifts and range contractions can lead to local extinctions and/or alterations in community interactions; two phenomena that threaten marine biodiversity (Beaugrand et al., 2015). When a range shift affects a harvested species, socio-economic impact can be substantial. Therefore, in the framework of marine fisheries, the evaluation of the effects of climate warming on fish range is a major challenge in sustainable resources management (Cheung et al., 2013). European anchovy (Engraulis encrasicolus) is a small pelagic fish, known to play a crucial ecological and economic role in coastal ecosystems. From an ecological point of view, the anchovy has a low position in the marine foodweb and therefore plays a key role in the transfer of matter and energy from plankton to upper trophic levels (Cury et al., 2000). From an economical perspective, the European anchovy is one of the world’s most traded fish species (FAO, 2010). Its life history traits (short lifespan, r reproductive strategy, high early stage mortality) make this species particularly sensitive to environmental changes and many studies have focused on the sensitivity of its recruitment to environmental variability at a regional scale (Borja et al., 1998; Santojanni et al., 2006; Irigoien et al., 2009). However, the extent to which climate warming may reshape global European anchovy distributional range remains poorly understood. Generalized additive models (GAM) have been extensively applied on European anchovy to characterize the potential spatial distribution of its different life stages in different regions: North Sea (Kanstinger and Peck, 2009), Bay of Biscay (Planque et al., 2007), North-east Atlantic (Bernal et al., 2007), Mediterranean Sea (Martín et al., 2008), and Aegean Sea (Giannoulaki et al., 2008; Schismenou et al., 2008). These studies were very useful to reveal the relationship between the European anchovy and environmental factors at a local scale but none of them deals with the global future spatial range of the species. European anchovy is a warm-temperate fish, largely spread from the southern part of the North Sea to South Africa, also occurring in the Mediterranean, Azov and Black Seas. Like other marine pelagic fish, anchovies exhibit migratory behaviour and great dispersal capabilities both at the larval and adult stages. Several recent observations showed an expansion of European anchovy populations in the North Sea since the 1990s (Beare et al., 2004; Alheit et al., 2012; Petitgas et al., 2012; Raab et al., 2013). On the other hand, anchovy stocks have declined since the 1990s above the continental shelves of warmer regions such as the Mediterranean Sea (Palomera et al., 2007; Tugores et al., 2010). Here, we investigate the potential influence of climate change on the spatial distribution of the European anchovy at a global scale. We first model the ecological niche (sensu Hutchinson) of the fish and project its future spatial range using the latest generation of IPCC representative concentration pathways (RCP) scenarios (Moss et al., 2010) and five of the latest generation of Atmosphere-Ocean General Circulation Models (AOGCMs) in the aim of evaluating the uncertainties on the projections related to AOGCMs. We subsequently discuss the advantages and potential shortcomings of this type of approach on the European anchovy and conclude on the potential ecological and economic consequences of the changes in anchovy spatial distribution. Material and methods Biological and environmental data Occurrence data of the European anchovy The observed distribution of E. encrasicolus was obtained from the Food and Agriculture Organization of the United Nations (FAO) Aquatic species distribution Map Viewer (http://www.fao.org/figis/geoserver/factsheets/species.html). We completed these data with occurrence records from two databases: OBIS (http://www.iobis.org/) and GBIF (http://data.gbif.org/) and the following published literature: (Huggett et al., 2003; Mullon et al., 2003; Fairweather et al., 2006; Roy et al., 2007). We considered here only adult fish occurrence data, i.e. fish that have reached their first sexual maturity (spawners). The population structure was not taken into account in our modelling approach because it is idiosyncratic and cannot be realistically modelled at global scale. Origin of the large-scale environmental data We used bathymetric data from the "Smith and Sandwell Global Seafloor topography" (Smith and Sandwell, 1997). Sea surface salinity (SSS) data were retrieved from the Levitus’ climatology (Levitus, 1982) and completed with ICES database (http://www.ices.dk). Sea surface temperature (SST) data for the period 1982–2009 were extracted from NOAA 4-km Advanced Very High Resolution Radiometer (AVHRR) Pathfinder (Casey et al., 2010). The climatology 1998–2010 of surface Chlorophyll a concentration was obtained from Seawifs-9 km data (Acker and Leptoukh, 2007). Satellites only measure the environmental conditions of a layer corresponding to a quarter of the depth of the euphotic layer. These data were used as a proxy of the environmental conditions encountered by anchovy populations located deeper. As the spatial resolution of the different environmental databases was not equivalent, all environmental data were interpolated using the inverse-squared distance method (Beaugrand et al., 2000) on a 0.1°longitude × 0.1°latitude spatial grid, in a geographical domain ranging from 30°W to 60°E and from 43°S to 70°N. Georeferenced occurrence data of E. encrasicolus were reorganized on a similar spatial grid and the corresponding environmental data were assigned to each occurrence point. Ecological niche model description NPPEN model We modelled the ecological niche and the geographical distribution of the European anchovy using the Non-Parametric Probabilistic Ecological Niche (NPPEN) model (Beaugrand et al., 2011). This method is based on the concept of the ecological niche sensu Hutchinson (Hutchinson, 1957) and presents several advantages: - NPPEN requires only presence data, a considerable advantage for a study that focuses on marine environment where absences data cannot be inferred with certainty. - NPPEN is based on a non-parametric procedure using a simplified version of the Multiple Response Permutation Procedure (Mielke et al., 1981). - the use of the Mahalanobis generalized distance instead of the Euclidean distance enables the correlation between abiotic variables to be taken into account (Farber and Kadmon, 2003). We will briefly recall here the main calculation steps of the NPPEN model because a full description is available in the reference publication (Beaugrand et al., 2011) and in the articles that used NPPEN for different applications (Lenoir et al., 2011; Chaalali et al., 2013; Raybaud et al., 2013; Goberville et al., 2015; Raybaud et al., 2015). First, we constructed a reference matrix (⁠ Zm,p ⁠) of the environmental data encountered by the species for each occurrence record data. m is the number of species occurrences and p is the number of corresponding abiotic variables. Second, we calculated the Mahalanobis generalized distance between the observations and the homogenized reference matrix using the following formula: Dx,Z2=(x−Z¯)′R−1(x−Z¯)(1) x is a vector of length p and represents the values of the abiotic variable to be tested; Rp,p is the correlation matrix of Zm,p ⁠, and Z- the average environmental conditions assessed from Zm,p ⁠. Third, we used the Multiple Response Permutation Procedure (Mielke et al., 1981) to calculate the probability of each geographical cell to belong to the reference matrix. This probability (⁠ v ⁠) is the number of times the simulated distance was found greater or equal than the observed average distance: ν=qεs≥ε0n(2) ε0 is the average observed distance, εs is the recalculated distance after permutation, and n is the maximum number of permutations. Finally, the probabilities of occurrence were projected into a geographical space to map the current spatial distribution of the European anchovy. To evaluate the possible range shifts of the fish by the end of the 21st century, we used the latest generation of climate models (see paragraph “Future climate projections”). Selection of the best combination of abiotic variables The first step to model the geographical range of a species is to identify which abiotic variables are the most influential. The NPPEN model was ran with eleven combinations of abiotic variables (Supplementary Table S1) and we evaluated the ability of each run to reproduce the observed spatial distribution. The choice of environmental data to test in the ecological niche model (SST, bathymetry, SSS, and chlorophyll a) was based on expert knowledge and bibliography on habitat preference of small pelagic fish a global scale. We are aware that, at regional scale, other covariates could be more important (see discussion). SST is known to be an important factor that drives the geographical distribution of species at a global scale (Beaugrand et al., 2015). Bathymetry was chosen because European anchovy occurs mainly over continental shelves (Checkley et al., 2009; Giannoulaki et al., 2013). Capacity dispersal can also be modulated by SSS through the ability of fish to osmoregulate (Schmidt-Nielsen, 1997). Sea surface chlorophyll a concentration was used here as an indicator of food availability, given that anchovy is a plankton feeder species (Basilone et al., 2004b; Bacha and Amara, 2009). Phytoplankton will be consumed by zooplankton, which is the main food items for many fish larvae of many species, and for small pelagic fish such as anchovy. We compared the performance of each run using two different model evaluation metrics: the Continuous Boyce’s Index (CBI; Hirzel et al., 2006) and the Area Under Curve (AUC; Swets, 1988; Fielding and Bell, 1997). CBI is a modification of Boyce’s index (Boyce et al., 2002) especially designed for presence-only models (Braunisch and Suchant, 2010). This index is based on a moving window analysis on the predicted-to-expected (P/E) frequency curve and uses the Spearman rank correlation coefficient to measure the monotonic increase of the curve. Values of CBI vary from −1 for an inverse model to 1 for a perfect prediction. Values close to zero indicate a random model. AUC plots the false positive rate (1-specificity) against the true positive rate (sensitivity) and calculates the area under the curve. AUC ranges from 0.5 for a random model to 1 for a perfect one. This threshold independent method was first developed for presence–absence models, although absences can be replaced by pseudo-absences, also termed background locations (Wiley et al., 2003; Philips et al., 2006; Tittensor et al., 2009). To model the current geographic distribution of E. encrasicolus, we selected the combination of variables with the highest performance using the two metrics. We used the cross-validation procedure recommended by Merow et al. (2013) that randomly selects 70% of the occurrence data to run the model and 30% of the data for validation. Future climate projections To establish projections for the 21st century, we used outputs of SST from the latest generation of climate models called RCPs (Moss et al., 2010), which are a part of the fifth Intergovernmental Panel on Climate Change (IPCC) assessment report. RCPs have replaced Special Report on Emissions Scenarios (SRES) and provide a range of possible futures for the evolution of atmospheric composition. Before using RCP-SST data to project the future distribution of European anchovy, we first statistically compared the probabilities based on AVHRR data with those calculated from RCPs for 2006–2009 using a Taylor diagram (Taylor, 2001; Raybaud et al., 2011). This step was performed to detect and prevent potential bias in ENM results related to the use of two different SST databases (see Supplementary Figure S1). We used the SST outputs for the decades 2020–2029 (near future), 2050–2059 (mid-century), and 2090–2099 (end of the century), derived from five climate models, also termed Atmosphere-Ocean General Circulation Models (AOGCMs), listed in Supplementary Table S2. For each, we used the optimistic scenario RCP2.6 (global temperature warming below 2 °C), the intermediate RCP4.5, and the most pessimistic RCP8.5 (the “business-as-usual” scenario, in which global mean temperature increase could reach 4–5 °C). We did not use the scenario RCP6.0 because it was not available in all AOGCMs. We used SST data provided by these climate models as input in our ENM to calculate and map the probability of occurrence of E. encrasicolus by the end of the century (decade 2090–2099). For each RCP scenario, we calculated the mean and the coefficient of variation (CV) between the probabilities of occurrence based on all AOGCMs. The map of the CV showed the areas where projections most vary from one AOGCM to another. Results Modelling of the ecological niche of the European anchovy Description of the 11 simulations performed to choose the best set of abiotic variables to model European anchovy spatial distribution is presented in Supplementary Table S1. The run which gives the best performance (CBI = 0.8861 ± 0.0115, AUC = 0.9425 ± 0.0017) is “run1”, performed with only 2 abiotic variables: SST and bathymetry. This is therefore the set of parameters that we chose to model the ecological niche and the spatial distribution of European anchovy. As explained in a recent study on ecological niche modeling (Raybaud et al., 2013), this does not necessarily mean that other factors do not affect species spatial distribution at smaller temporal and/or spatial scale. Our results show that, at a global scale, the spatial distribution of European anchovy is primary constrained by SST and bathymetry. The modelled two-dimensional ecological niche of European anchovy based on SST and bathymetry is presented in Figure 1a. The highest probabilities of occurrence are found at bathymetry ranging from 20 to 300 m and annual SST from 16 to 19 °C. These results are in agreement with the knowledge of thermal and bathymetric preferences of the European anchovy (Basilone et al., 2004a; Planque et al., 2007; Schismenou et al., 2008; Giannoulaki et al., 2013). The interaction between the environment and the species vary according to the position of the environmental regime along the niche (Beaugrand and Kirby, 2010; Beaugrand et al., 2013). For example, at the centre of the thermal niche of a species, thermal fluctuations have no effect. At the left side of the niche (cold regime), an increase in temperature augments strongly the probability of occurrence and at the right side (warm regime), an increase in temperature diminishes the probability of occurrence. Interactions are therefore non-linear. Figure 1 Open in new tabDownload slide (a) Modelled two-dimensional ecological niche of the European anchovy E. encrasicolus based on sea surface temperature (SST) and bathymetry. The colourbar indicates the probability of species occurrence. (b) Observed spatial distribution of E. encrasicolus. Each point represents an occurrence record. (c) Probability of occurrence of E. encrasicolus modelled by NPPEN model for the reference period 1982–2009. White mesh-grid cells denote a nil probability. Figure 1 Open in new tabDownload slide (a) Modelled two-dimensional ecological niche of the European anchovy E. encrasicolus based on sea surface temperature (SST) and bathymetry. The colourbar indicates the probability of species occurrence. (b) Observed spatial distribution of E. encrasicolus. Each point represents an occurrence record. (c) Probability of occurrence of E. encrasicolus modelled by NPPEN model for the reference period 1982–2009. White mesh-grid cells denote a nil probability. Current geographical distribution of European anchovy The observed current geographical distribution of E. encrasicolus (Figure 1b) and the modelled current probability of occurrence (Figure 1c) show similar patterns (CBI = 0.8861 ± 0.0115). The map of the observed presence records show that the European anchovy is a coastal species, found on the continental shelf and the species ranges from the south of the Norwegian coasts to South Africa, in the Mediterranean, the Azov and the Black Seas (Figure 1b). In the southern Benguela, E. encrasicolus was formerly called Engraulis capensis. Probability of occurrence, achieved by projecting the two-dimensional-modelled ecological niche of the species in a geographical space (Figure 1c), shows highest values (>0.7) from the Bay of Biscay to the Mauritanian coasts, in the Mediterranean Sea, the Azov Sea and in the Benguela system. These areas correspond to regions where the habitat is the most suitable for European anchovy. The probabilities of occurrence are medium to low in the English Channel and the North Sea for the reference period 1982–2009, indicating that these regions are currently in the colder part of the ecological niche of the species. Despite an overall good accuracy of the projections at a global scale, a discrepancy between model and observations occurs in some regions. The modelled distribution extents slightly more northwards than the observed distribution along Norwegian coasts. In the northern Persian Gulf and along eastern coasts of Africa (Indian Ocean), the model also predicts a non-zero probability of occurrence while the species does not occur. However, it should be noted that, in all cases, the modelled probability of occurrence is very low (<0.1) in these areas, indicating a position near the edge of the thermal niche. The model that we retained in this study (i.e. the one that gave the best representation of the current distribution of European anchovy at a global scale) took into account only mean SST and bathymetry. We are aware that, at a regional scale, other factors such as dispersal capacities, biotic interactions or hydrology can shape the species distribution patterns (see discussion for details on mechanisms not taken into account in ecological niche models). On the contrary, in the Gulf of Guinea, the model failed to predict the presence of European anchovy. This area is characterized by a narrow continental shelf and a relatively high mean SST (near the maximum threshold of species’ tolerance). In a recent genetic study (Silva et al., 2014b), it has been shown that European anchovy displayed a two-clade mitochondrial structure that shift smoothly along its latitudinal distribution. Clade A dominates at lower latitudes (i.e. in warm-water areas) whereas clade B is more frequent at higher latitudes (i.e. in cold-water regions). Clade B was even absent off Senegal, Guinea, and Ghana, highlighting a local genetic adaptation of anchovy populations to warm temperatures. Our model, that has been calibrated to represent the geographical distribution of the anchovy at a global scale, has trouble representing regional thermal adaptation. Projections of geographical range of European anchovy for the end of the 21st century Maps of the projected probability of occurrence of the European anchovy for the end of the 21st century We calculated probabilities of European anchovy occurrence for the decades 2020–2029 (near future), 2050–2059 (mid-century), and 2090–2099 (end of the century) using SST data from five AOGCMs. Here, we will focus only on the end of the century but the other periods (2020s and 2050s) are presented in Supplementary Figures S2 and S3. For each of the three warming scenarios (RCP scenarios), results are presented as mean probability of occurrence and CV between probabilities based on each climate model. The use of a multi-AOGCM ensemble and three RCP scenarios allow a multitude of potential trajectories to be covered for the end of the century (Laepple et al., 2008). With the optimistic scenario RCP2.6, the geographical distribution of the mean probability of occurrence calculated for 2090–2099 was close to the reference distribution (Figure 2a). With the intermediate RCP4.5 and the pessimistic RCP8.5, the model forecasted sometimes substantial regional changes (Figure 2b and 2c). For example, an increase in the probability of occurrence is forecasted in the English Channel and the North Sea with scenarios RCP4.5 and 8.5. On the contrary, the probability of occurrence is expected to diminish in South Africa (Supplementary Figure S4). In the northern hemisphere, the distributional range of European anchovy may shift northwards by the end of the century in the case of moderate (RCP4.5) to strong (RCP8.5) increase in SST. In the southern hemisphere, only a decrease in the probability of occurrence is expected to occur during the 21st century because there is no continental shelf south of South Africa. Figure 2 Open in new tabDownload slide Top panels: probability of occurrence of E. encrasicolus for the decade 2090–2099, based on five climate models. (a) Mean probability of occurrence with the scenario RCP2.6; (b) with the scenario RCP4.5; (c) with the scenario RCP8.5. White mesh-grid cells denote a nil probability. Bottom panels: coefficient of variation of the probability of occurrence based on the five climate models (d) for the scenario RCP2.6; (e) for the scenario RCP4.5; (f) for the scenario RCP8.5. Figure 2 Open in new tabDownload slide Top panels: probability of occurrence of E. encrasicolus for the decade 2090–2099, based on five climate models. (a) Mean probability of occurrence with the scenario RCP2.6; (b) with the scenario RCP4.5; (c) with the scenario RCP8.5. White mesh-grid cells denote a nil probability. Bottom panels: coefficient of variation of the probability of occurrence based on the five climate models (d) for the scenario RCP2.6; (e) for the scenario RCP4.5; (f) for the scenario RCP8.5. The examination of the spatial patterns in the CVs show for all RCP scenarios regions where forecasted probabilities vary most among AOGCM outputs (Figure 2d–f). In regions, where the European anchovy currently presents the highest probability of occurrence (20–50°N and 30–35°S), CVs were low, indicating that all climate models forecasted similar changes for 2090–2100. In contrast, probabilities were more variable towards the edge of the range of European anchovy (Norwegian, Baltic, and Red Seas). Differences between current probability of occurrence and projections of E. encrasicolus for the end of the 21st century To highlight areas where European anchovy fisheries are forecasted to change most by the end of the century, we mapped the differences between projected and current probabilities (Figure 3, top panels). With Scenario RCP2.6, differences are low (< ±0.2) in most anchovy’s regions (Figure 3a). The only areas where a decrease in the probability of occurrence is notable are along Namibia and South-African coasts. Differences maps based on scenarios RCP4.5 and RCP8.5 (Figure 3b and c) exhibit a low to moderate reduction in the probability of European anchovy occurrence in several regions such as the Mediterranean coasts and the north western African coasts while a substantial decrease is patent along Namibian and South-African coasts. On the opposite, our results also forecast an increase in the probability of occurrence in many regions: the north-western part of the Black Sea, Bay of Biscay, Celtic Sea, English Channel, North Sea, and the southern part of the Norwegian Sea. These increases are larger with the warmer scenario RCP8.5. Figure 3 Open in new tabDownload slide Top panels: differences between the current probabilities of occurrence of E. encrasicolus calculated from the Ecological Niche Model NPPEN and projections for the end of the century (decade 2090–2099) (a) with scenario RCP2.6; (b) with scenario RCP4.5; (c) with scenario RCP8.5. Bottom panels: latitudinal gradients of the differences (d) with scenario RCP2.6; (e) with scenario RCP4.5; (f) with scenario RCP8.5. Figure 3 Open in new tabDownload slide Top panels: differences between the current probabilities of occurrence of E. encrasicolus calculated from the Ecological Niche Model NPPEN and projections for the end of the century (decade 2090–2099) (a) with scenario RCP2.6; (b) with scenario RCP4.5; (c) with scenario RCP8.5. Bottom panels: latitudinal gradients of the differences (d) with scenario RCP2.6; (e) with scenario RCP4.5; (f) with scenario RCP8.5. We also examined projected latitudinal changes in the geographical range of European anchovy (Figure 3, bottom panels). This representation was obtained by averaging differences in occurrence probabilities for each latitude. Whatever the warming scenario, our results forecasted a reduction in the probability of occurrence in all the regions located under 48°N and an increase in all more northern areas. Increasing warming from RCP2.6 to RCP8.5 produced larger responses but latitudinal patterns of increase or decrease seemed independent on the level of warming. Discussion ENMs and European anchovy ENMs have commonly been applied to evaluate the current and future fish distribution. Some studies assessed range shifts of a large number of marine fish (Cheung et al., 2008; Ben Rais Lasram et al., 2010; Albouy et al., 2013). Such studies are very useful to highlight global patterns in potential changes of marine fish biodiversity but the results are often presented as species assemblages. The online AquaMaps approach of Kaschner et al. (2008) is another relatively recent tool for generating model-based, large-scale predictions of probability of occurrence of many marine species, but projections are only based on one SRES scenario (IPCC A2 emissions scenario) for 2100. Other ENMs or species distribution models (SDMs) applied to E. encrasicolus have focused on a restricted part of its spatial range. Palialexis et al. (2011) applied an ensemble of ENMs/SDMs (GAM, Boosted Regression Trees, Multivariate Analysis and Regression Splines, MAXENT, Bioclim Envelope Model) on sardine and anchovy. In this study, the authors focused on the Thermaikos Gulf, an area too small to be compared with our results. Despite some differences in methodology, our study can more easily be compared with those of Lenoir et al. (2011). They modelled the future potential range shift of eight fish (including European anchovy) under climate change scenarios. In the study of Lenoir et al. (2011), the authors arbitrary used three environmental variables (SST, SSS, and bathymetry) and did not test different combination of abiotic variables. Moreover, they focused on the North Atlantic (from 80°W to 70.5°E and 25°N to 85°S) and to do not consider the full range of European anchovy. The assessment of the niche from a limited part of the species global distribution might alter niche shape and bias our projections (Beaugrand et al., 2013). In addition, their projections for the 21st century were based on SRES A2 and B2 climate scenarios and the use of only one AOGCM (ECHAM 4 model), which make impossible to evaluate the uncertainties related to climate models. As no evaluation of the model accuracy has been performed in the study of Lenoir el al. (2011), we cannot statistically compare the relative performance of the two models. However, a comparison of the maps of expected changes between the 2090s and the current distribution can be performed. Lenoir et al. (2011) forecasted only a gain in the North Sea and no changes in the rest of the distribution. Although our study also shows a likely increase in the probability of occurrence in the North Sea during the 21st century, our model predicts an increase in the Bay of Biscay with RCP4.5 and 8.5, which was not captured by Lenoir et al. (2011). Moreover, we also forecast a low to moderate reduction in the probability of occurrence in several regions such as the Mediterranean Sea and the north western African coasts whereas Lenoir et al. (2011) found no potential changes in these areas. The substantial decrease in the probability of occurrence that our study forecasts along Namibian and South-African coasts cannot be compared with the results of Lenoir because they did not consider the southern hemisphere in their study. Here we investigated the potential impact of climate change on European anchovy distribution by the end of the 21st century using all scenarios of warming from the most optimistic scenario (RCP2.6) where expected changes by the end the century are low to the most pessimistic scenario (RCP8.5) where projections forecast significant biogeographical shifts by 2100. In addition, the use of five AOGCMs enabled to highlight the geographical areas where uncertainties are likely to be the highest and where projections vary most from one AOGCM to another (Figure 3d–f). However, as any modelling framework, ENMs/SDMs do not provide a holistic view of processes affecting species distribution and results must be interpreted cautiously (Albouy et al., 2013). These models rely on two core assumptions that need to be discussed in the case of European anchovy: (i) the species is currently in equilibrium with its environment and (ii) E. encrasicolus will conserve its ecological niche until the end of the century (i.e. niche conservatism). Niche conservatism is the tendency of species to retain ancestral ecological characteristics (Crisp et al., 2009) and it is the process that makes rapid environmental changes a threat for species. Although this hypothesis remains debated (Warren et al., 2008), it is unlikely that a species can strongly alter its ecological niche on time-scales ranging from decades to centuries and some authors have shown for zooplankton that the ecological niche remains stable over several decades (Helaouët and Beaugrand, 2007). In addition, high mobility of European anchovy promotes range-shift rather than the genetic adaptation of the species. A recent study from Silva et al. (2014a) based on genetic analysis showed that a range shift of anchovies already occurred 20 000 years ago, during the Last Glacial Maximum. Climate change impacts on the spatial distribution of anchovy may not only be related to temperature changes. For example, hydrology, ocean stratification or coastal upwelling intensity changes are likely to affect biological production and hence anchovy’s distribution and stratification. The mechanisms related to the responses of anchovies to climate change seem to be also highly associated with expansion of suitable spawning and feeding habitats (Petitgas et al., 2013). Biotic interactions were not taken into account in our approach and changes in the concentration or in the geographical distribution of prey, predators, or competitors of the European anchovy could reshape our projections (Engelhard et al., 2014). For example, in the areas where the ENM forecasts an increase of the populations by the end of the century (i.e. the Bay of Biscay, the Celtic Sea, the English Channel, the North Sea, and the Norwegian Sea), the increase in anchovy could be counterbalanced by a lack of adequate preys or an excessive concentration of predators or competitors. However, these regions are known to be highly productive and plankton concentration is unlikely to become a limiting factor for the expansion of anchovy population. On the contrary, changes in the populations of top predators’ such as large fish, marine mammals and birds, may affect European anchovy populations. The expansion of alien invasive species such as the highly opportunistic ctenophore Mnemiopsis leidyi could also alter our projections at local scale. Native to America, M. leidyi was introduced in Europe in the 1980s and the ctenophore is a potential competitor of anchovy as well as a predator during certain periods of the year (Oguz et al., 2008). In the Black Sea, the presence of M. leidyi was one of the most important reasons for the sharp decrease in anchovy and other pelagic fish stocks in late 1980s (Kideys et al., 2000). Currently, M. leidyi has already been observed in many regions of the European anchovy range: the Black Sea and adjacent waters (Shiganova, 1998), the Mediterranean Sea (Boero et al., 2009), the Baltic Sea (Lehtiniemi et al., 2007), the North Sea, and the English Channel (Antajan et al., 2014). Finally, another important phenomenon, which could alter our projections, is fishing. Our projections are exclusively based on the environmental conditions of European anchovy and do not take into account the influence of fisheries. European anchovy is one of the most important commercial species in Europe and Africa and some stocks are fully exploited or over-exploited, especially in the Mediterranean and the Black Sea (FAO, 2010). In the areas where our results forecast a significant decrease in the probability of occurrence of European anchovy (i.e. the Mediterranean Sea and the African coasts), catches should be particularly monitored. A non-sustainable management of anchovy fisheries in these regions could lead to a local collapse of the species. Synergistic effects of climate and fishing may precipitate the decline of European anchovy in these areas. Potential ecological and economic implications of the European anchovy range shift Our study suggests that ocean warming may induce large changes in the geographical distribution of European anchovy, which may have major ecological and economic implications (Figure 4). The first obvious ecological consequence of anchovy range shift is the alteration in anchovy abundance in many regions. Some regions are forecasted to see an expansion of anchovy populations whereas others are expected to experience a severe decline or local depletions. European anchovy constitutes a considerable biomass at intermediate levels of the marine foodweb and the fish has been identified as an important bottom-up flow control group (Cury et al., 2000). Hence, the collapse of anchovy populations in some regions could induce ecosystem-wide responses, by reducing food supply to larger fish, seabirds, and marine mammals. Both the structure and functioning of pelagic ecosystems may in turn be altered (Alekseenko et al., 2014; Chaalali et al., 2016). Figure 4 Open in new tabDownload slide Conceptual scheme of the possible ecological and economic consequences of the forecasted range shift of the European anchovy by the end of the century. Figure 4 Open in new tabDownload slide Conceptual scheme of the possible ecological and economic consequences of the forecasted range shift of the European anchovy by the end of the century. In addition to ecological implications, anchovy range shifts may have considerable social and economic consequences (Pinsky and Fogarty, 2012). Large populations of European anchovy support important fisheries in many regions of the world upon which the economies of many countries depend (FAO, 2010). The first direct implication for fisheries is the redistribution of European anchovy stocks and catches. This forecasted reshaping in the anchovy fisheries could have important economic consequences, especially for the poorest countries. In the regions, where European anchovy populations will decline during the 21st century, fishers will have two options. At short time-scale, local fishers may increase their fishing effort to maintain their profits but it may generate additional costs for fisheries and pressure on the stocks. Anchovy populations may move away from the area where fishing fleets currently operate and distribution changes may have significant consequences for the distance that fishing vessels must accomplish to reach anchovies, with implications for fuel usage and time at sea (Sumaila et al., 2011). Over the longer term, fishers may try to catch other forage fish to compensate for the anchovy decline (Pinsky and Fogarty, 2012). For example, during the decline of the sardine fishery between the 1950s and 1960s, California fishers switched from sardines and other small pelagic fish to higher trophic level, predatory species such as albacore (Checkley et al., 2009). However, a change in the fishing strategy and the target species will induce a considerable investment (e.g. changes in vessel or gear types). In addition, an alteration in anchovy catches could have antagonist consequences on anchovy predators and competitors fisheries (Salomon et al., 2010). For example, a decline in anchovy population may induce a depletion of some large predatory fish such as tuna, which have also a great economic importance (FAO, 2010). In contrast, the anchovy collapse may also promote landings of other harvested forage species which are currently competing with anchovy. Finally, anchovies may migrate across state boundaries and political frontiers. Anchovy stocks may therefore be harvested by fishers from other countries, exacerbating inequalities between fishers from different regions and conflicts between countries (Coulthard, 2009). The “Mackerel War” in 2010 between Icelandic and British fishers illustrates this problem. Previously almost absent from the Icelandic economic zone, the Atlantic mackerel (Scomber scombrus) is now found in large quantities in many areas around Iceland since 2007, probably because of the rise in sea temperature (Astthorsson et al., 2012). Icelandic fishers began fishing intensively this poleward-shifting mackerel population but the government of Iceland was not satisfied with the quota offered by the European Union. Thereby, Iceland unilaterally set a large quota for herself (Hannesson, 2013). Soon after, finding their quota allocation unacceptably low, the Faroe Islands did the same. This Mackerel war led to tensions between the Icelandic government and other countries from the European Union. Above, we outlined a preliminary list of the most possible straightforward implications of European anchovy range shift by the end of the century but we are aware that this list is far from exhaustive. Complex non-linear relationships take place in marine ecosystems and many interactive processes are likely, making it difficult the assessment of all possible consequences. In the present study, we exclusively investigated the potential influences of changing sea temperatures on the spatial distribution of European anchovy. Our longer-term projections are rather pessimistic for Mediterranean and African anchovy fisheries. Other anthropogenic pressures are also likely to affect anchovy populations (e.g. pollution, ocean acidification, or exotic species introduction) and could lead to unexpected outcomes in the future. Despite the magnitude of the task, future studies should attempt to consider all those interactive effects to evaluate the future of anchovy populations and stocks. Supplementary data Supplementary material is available at the ICESJMS online version of the manuscript. Acknowledgements We acknowledge the World Climate Research Programme's Working Group on Coupled Modeling, which is responsible for CMIP, and we thank the climate modeling groups for producing and making available their model output. Funding This work has been supported by the ANR programme in the frame of the GlobCoast ANR-11-BLANC-BS56-018_01 project. This work is also a contribution to the CPER research project CLIMIBIO. The authors thank the French Ministère de l'Enseignement Supérieur et de la Recherche, the Hauts de France Region and the European Funds for Regional Economical Development for their financial support to this project. References Acker J. G. , Leptoukh G. 2007 . Online analysis enhances use of NASA earth science data . Eos, Transactions American Geophysical Union 88 : 14 . Google Scholar Crossref Search ADS WorldCat Albouy C. , Guilhaumon F., Leprieur F., Lasram F. B. R., Somot S., Aznar R., Velez L., et al. 2013 . Projected climate change and the changing biogeography of coastal Mediterranean fishes . Journal of Biogeography 40 : 534 – 547 . Google Scholar Crossref Search ADS WorldCat Alekseenko E. , Raybaud V., Espinasse B., Carlotti F., Queguiner B., Thouvenin B., Garreau P., et al. 2014 . Seasonal dynamics and stoichiometry of the planktonic community in the NW Mediterranean Sea: a 3D modeling approach . Ocean Dynamics 64 : 179 – 207 . Google Scholar Crossref Search ADS WorldCat Alheit J. , Pohlmann T., Casini M., Greve W., Hinrichs R., Mathis M., O’Driscoll K., et al. 2012 . Climate variability drives anchovies and sardines into the North and Baltic Seas . Progress in Oceanography 96 : 128 – 139 . Google Scholar Crossref Search ADS WorldCat Antajan E. , Bastian T., Raud T., Brylinski J. M., Hoffman S., Breton G., Cornille V., et al. 2014 . The invasive ctenophore Mnemiopsis leidyi A. Agassiz, 1865 along the English Channel and the North Sea French coasts: another introduction pathway in northern European waters? Aquatic Invasions, 9: 167–173. Astthorsson O. S. , Valdimarsson H., Gudmundsdottir A., Óskarsson G. J. 2012 . Climate-related variations in the occurrence and distribution of mackerel (Scomber scombrus) in Icelandic waters . ICES Journal of Marine Science 69 : 1289 – 1297 . Google Scholar Crossref Search ADS WorldCat Bacha M. , Amara R. 2009 . Spatial, temporal and ontogenetic variation in diet of anchovy (Engraulis encrasicolus) on the Algerian coast (SW Mediterranean) . Estuarine, Coastal and Shelf Science 85 : 257 – 264 . Google Scholar Crossref Search ADS WorldCat Basilone G. , Guisande C., Patti B., Mazzola S., Cuttitta A., Bonanno A., Kallianiotis A. 2004a . Linking habitat conditions and growth in the European anchovy (Engraulis encrasicolus) . Fisheries Research 68 : 9 – 19 . Google Scholar Crossref Search ADS WorldCat Basilone G. , Guisande C., Patti B., Mazzola S., Cuttitta A., Bonanno A., Kallianiotis A. 2004b . Linking habitat conditions and growth in the European anchovy (Engraulis encrasicolus) . Fisheries Research 68 : 9 – 19 . Google Scholar Crossref Search ADS WorldCat Beare D. , Burns F., Greig A., Jones E., Peach K., Kienzle M., McKenzie E., et al. 2004 . Long-term increases in prevalence of North Sea fishes having southern biogeographic affinities . Marine Ecology. Progress Series 284 : 269 – 278 . Google Scholar Crossref Search ADS WorldCat Beaugrand G. , Edwards M., Raybaud V., Goberville E., Kirby R. R. 2015 . Future vulnerability of marine biodiversity compared with contemporary and past changes . Nature Clim. Change 5 : 695 – 701 . Google Scholar Crossref Search ADS WorldCat Beaugrand G. , Kirby R. R. 2010 . Spatial changes in the sensitivity of Atlantic cod to climate-driven effects in the plankton . Climate Research 41 : 15 – 19 . Google Scholar Crossref Search ADS WorldCat Beaugrand G. , Lenoir S., Ibanez F., Manté C. 2011 . A new model to assess the probability of occurrence of a species based on presence-only data . Marine Ecology Progress Series 424 : 175 – 190 . Google Scholar Crossref Search ADS WorldCat Beaugrand G. , Mackas D., Goberville E. 2013 . Applying the concept of the ecological niche and a macroecological approach to understand how climate influences zooplankton: advantages, assumptions, limitations and requirements . Progress in Oceanography 111 : 75 – 90 . Google Scholar Crossref Search ADS WorldCat Beaugrand G. , Reid P. C., Ibañez F., Planque B. 2000 . Biodiversity of North Atlantic and North Sea calanoid copepods . Marine Ecology Progress Series 204 : 299 – 303 . Google Scholar Crossref Search ADS WorldCat Ben Rais Lasram F. , Guilhaumon F., Albouy C., Somot S., Thuiller W., Mouillot D. 2010 . The Mediterranean Sea as a ‘cul‐de‐sac’for endemic fishes facing climate change . Global Change Biology 16 : 3233 – 3245 . Google Scholar Crossref Search ADS WorldCat Bernal M. , Stratoudakis Y., Coombs S., Angelico M., de Lanzós A. L., Porteiro C., Sagarminaga Y., et al. 2007 . Sardine spawning off the European Atlantic coast: characterization of and spatio-temporal variability in spawning habitat . Progress in Oceanography 74 : 210 – 227 . Google Scholar Crossref Search ADS WorldCat Braunisch, V., and Suchant, R. 2010. Predicting species distributions based on incomplete survey data: the trade-off between precision and scale. Ecography, 33: 826–840. Boero F. , Putti M., Trainito E., Prontera E., Piraino S., Shiganova T. A. 2009 . First records of Mnemiopsis leidyi (Ctenophora) from the Ligurian, Thyrrhenian and Ionian Seas (Western Mediterranean) and first record of Phyllorhiza punctata (Cnidaria) from the Western Mediterranean . Aquatic Invasions 4 : 675 – 680 . Google Scholar Crossref Search ADS WorldCat Borja A. , Uriarte A., Egana J., Motos L., Valencia V. 1998 . Relationships between anchovy (Engraulis encrasicolus) recruitment and environment in the Bay of Biscay (1967–1996) . Fisheries Oceanography 7 : 375 – 380 . Google Scholar Crossref Search ADS WorldCat Boyce, M. S., Vernier, P. R., Nielsen, S. E., and Schmiegelow, F. K. 2002. Evaluating resource selection functions. Ecological Modelling, 157: 281–300. Casey K. S. , Brandon T. B., Cornillon P., Evans R. 2010 . The past, present, and future of the AVHRR Pathfinder SST program. In Oceanography from space. Ed. by V. Barale, J. F. R. Gower and L. Alberotanza. Springer, Netherlands. pp. 273–287. Chaalali A. , Beaugrand G., Raybaud V., Goberville E., David V., Boët P., Sautour B. 2013 . Climatic facilitation of the colonization of an estuary by Acartia tonsa . PLoS ONE 8 : e74531. Google Scholar Crossref Search ADS PubMed WorldCat Chaalali A. , Beaugrand G., Raybaud V., Lassalle G., Saint-Béat B., Le Loc’h F., Bopp L., et al. 2016 . From species distributions to ecosystem structure and function: a methodological perspective . Ecological Modelling 334 : 78 – 90 . Google Scholar Crossref Search ADS WorldCat Checkley D. , Alheit J., Oozeki Y., Roy C. 2009 . Climate change and small pelagic fish , Cambridge University Press , Cambridge . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Cheung W. W. , Close C., Lam V., Watson R., Pauly D. 2008 . Application of macroecological theory to predict effects of climate change on global fisheries potential . Marine Ecology Progress Series 365 : 187 – 197 . Google Scholar Crossref Search ADS WorldCat Cheung W. W. , Watson R., Pauly D. 2013 . Signature of ocean warming in global fisheries catch . Nature 497 : 365 – 368 . Google Scholar Crossref Search ADS PubMed WorldCat Coulthard S. 2009 . Adaptation and conflict within fisheries: insights for living with climate change. Adapting to Climate Change: Thresholds, Values, Governance, pp. 255. Crisp M. D. , Arroyo M. T. K., Cook L. G., Gandolfo M. A., Jordan G. J., McGlone M. S., Weston P. H., et al. 2009 . Phylogenetic biome conservatism on a global scale . Nature 458 : 754 – 756 . Google Scholar Crossref Search ADS PubMed WorldCat Cury P. , Bakun A., Crawford R. J., Jarre A., Quiñones R. A., Shannon L. J., Verheye H. M. 2000 . Small pelagics in upwelling systems: patterns of interaction and structural changes in "wasp-waist" ecosystems . ICES Journal of Marine Science: Journal Du Conseil 57 : 603 – 618 . Google Scholar Crossref Search ADS WorldCat Engelhard G. H. , Peck M. A., Rindorf A., Smout S. C., van Deurs M., Raab K., Andersen K. H., et al. 2014 . Forage fish, their fisheries, and their predators: who drives whom? ICES Journal of Marine Science: Journal Du Conseil 71 : 90 – 104 . Google Scholar Crossref Search ADS WorldCat Fairweather T. , Van Der Lingen C., Booth A., Drapeau L., Van Der Westhuizen J. 2006 . Indicators of sustainable fishing for South African sardine Sardinops sagax and anchovy Engraulis encrasicolus . African Journal of Marine Science 28 : 661 – 680 . Google Scholar Crossref Search ADS WorldCat FAO . 2010 . Aquaculture Department: The State of World Fisheries and Aquaculture 2010. A report of the Food and Agriculture Organization of the United Nations, Rome. Farber O. , Kadmon R. 2003 . Assessment of alternative approaches for bioclimatic modeling with special emphasis on the Mahalanobis distance . Ecological Modelling 160 : 115 – 130 . Google Scholar Crossref Search ADS WorldCat Fielding, A. H., and Bell, J. F. 1997. A review of methods for the assessment of prediction errors in conservation presence/absence models. Environmental Conservation, 24: 38–49. Giannoulaki M. , Iglesias M., Tugores M. P., Bonanno A., Patti B., De Felice A., Leonori I., et al. 2013 . Characterizing the potential habitat of European anchovy Engraulis encrasicolus in the Mediterranean Sea, at different life stages . Fisheries Oceanography 22 : 69 – 89 . Google Scholar Crossref Search ADS WorldCat Giannoulaki M. , Valavanis V. D., Palialexis A., Tsagarakis K., Machias A., Somarakis S., Papaconstantinou C. 2008 . Modelling the presence of anchovy Engraulis encrasicolus in the Aegean Sea during early summer, based on satellite environmental data . Hydrobiologia 612 : 225 – 240 . Google Scholar Crossref Search ADS WorldCat Goberville E. , Beaugrand G., Hautekèete N. C., Piquot Y., Luczak C. 2015 . Uncertainties in the projection of species distributions related to general circulation models . Ecology and Evolution OpenURL Placeholder Text WorldCat Hannesson R. 2013 . Sharing the northeast Atlantic mackerel . ICES Journal of Marine Science: Journal Du Conseil 70 : 259 – 269 . Google Scholar Crossref Search ADS WorldCat Helaouët P. , Beaugrand G. 2007 . Macroecology of Calanus finmarchicus and C. helgolandicus in the North Atlantic Ocean and adjacent seas . Marine Ecology Progress Series 345 : 147 – 165 . Google Scholar Crossref Search ADS WorldCat Hickling R. , Roy D. B., Hill J. K., Fox R., Thomas C. D. 2006 . The distributions of a wide range of taxonomic groups are expanding polewards . Global Change Biology 12 : 450 – 455 . Google Scholar Crossref Search ADS WorldCat Hirzel, A. H., Le Lay, G., Helfer, V., Randin, C., and Guisan, A. 2006. Evaluating the ability of habitat suitability models to predict species presences. Ecological Modelling, 199: 142–152. Huggett J. , Fréon P., Mullon C., Penven P. 2003 . Modelling the transport success of anchovy Engraulis encrasicolus eggs and larvae in the southern Benguela: the effect of spatio-temporal spawning patterns . Marine Ecology Progress Series 250 : 247 – 262 . Google Scholar Crossref Search ADS WorldCat Hutchinson G. E. 1957 . Concluding remarks . Cold Spring Harbor Symposium Quantitative Biology 22 : 415 – 427 . Google Scholar Crossref Search ADS WorldCat Irigoien X. , Fernandes J. A., Grosjean P., Denis K., Albaina A., Santos M. 2009 . Spring zooplankton distribution in the Bay of Biscay from 1998 to 2006 in relation with anchovy recruitment . Journal of Plankton Research 31 : 1 – 17 . Google Scholar Crossref Search ADS WorldCat Kanstinger P. , Peck M. A. 2009 . Co-occurrence of European sardine (Sardina pilchardus), anchovy (Engraulis encrasicolus) and sprat (Sprattus sprattus) larvae in southern North Sea habitats: abundance, distribution and biochemical-based condition . Scientia Marina 73 : 141 – 152 . Google Scholar Crossref Search ADS WorldCat Kaschner K. , Ready J. S., Agbayani E., Rius J., Kesner-Reyes K., Eastwood P. D., South A. B., et al. 2008 . AquaMaps: Predicted range maps for aquatic species. Version 10/2008. www.aquamaps.org. World wide web electronic publication. Kideys A. E. , Kovalev A. V., Shulman G., Gordina A., Bingel F. 2000 . A review of zooplankton investigations of the Black Sea over the last decade . Journal of Marine Systems 24 : 355 – 371 . Google Scholar Crossref Search ADS WorldCat Laepple T. , Jewson S., Coughlin K. 2008 . Interannual temperature predictions using the CMIP3 multi-model ensemble mean . Geophysical Research Letters 35 : L10701. Google Scholar Crossref Search ADS WorldCat Lehtiniemi M. , Pääkkönen J. P., Flinkman J., Katajisto T., Gorokhova E., Karjalainen M., Viitasalo S., et al. 2007 . Distribution and abundance of the American comb jelly (Mnemiopsis leidyi)—a rapid invasion to the northern Baltic Sea during 2007 . Aquatic Invasions 2 : 445 – 449 . Google Scholar Crossref Search ADS WorldCat Lenoir S. , Beaugrand G., Lecuyer E. 2011 . Modelled spatial distribution of marine fish and projected modifications in the North Atlantic Ocean . Global Change Biology 17 : 115 – 129 . Google Scholar Crossref Search ADS WorldCat Levitus S. E. 1982 . Climatological atlas of the world ocean. NOAA Professional Paper 13. US Government Printing Office, Washington, DC. Martín P. , Bahamon N., Sabatés A., Maynou F., Sánchez P., Demestre M. 2008 . European anchovy (Engraulisencrasicolus) landings and environmental conditions on the Catalan Coast (NW Mediterranean) during 2000–2005. In Essential Fish Habitat Mapping in the Mediterranean , pp. 185 – 199 . Springer , Berlin . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Merow C. , Smith M. J., Silander J. A. 2013 . A practical guide to MaxEnt for modeling species’ distributions: what it does, and why inputs and settings matter . Ecography , 36: 1058–1069. OpenURL Placeholder Text WorldCat Mielke P. W. , Berry K. J., Brier G. W. 1981 . Application of multiresponse permutation procedures for examining seasonal changes in monthly mean sea-level pressure patterns . Monthly Weather Review 109 : 120 – 126 . Google Scholar Crossref Search ADS WorldCat Moss R. H. , Edmonds J. A., Hibbard K. A., Manning M. R., Rose S. K., van Vuuren D. P., Carter T. R., et al. 2010 . The next generation of scenarios for climate change research and assessment . Nature 463 : 747 – 756 . Google Scholar Crossref Search ADS PubMed WorldCat Mullon C. , Fréon P., Parada C., Van Der Lingen C., Huggett J. 2003 . From particles to individuals: modelling the early stages of anchovy (Engraulis capensis/encrasicolus) in the southern Benguela . Fisheries Oceanography 12 : 396 – 406 . Google Scholar Crossref Search ADS WorldCat Oguz T. , Fach B., Salihoglu B. 2008 . Invasion dynamics of the alien ctenophore Mnemiopsis leidyi and its impact on anchovy collapse in the Black Sea . Journal of Plankton Research 30 : 1385 – 1397 . Google Scholar Crossref Search ADS WorldCat Palialexis A. , Georgakarakos S., Karakassis I., Lika K., Valavanis V. 2011 . Prediction of marine species distribution from presence–absence acoustic data: comparing the fitting efficiency and the predictive capacity of conventional and novel distribution models . Hydrobiologia 670 : 241 – 266 . Google Scholar Crossref Search ADS WorldCat Palomera I. , Olivar M. P., Salat J., Sabatés A., Coll M., García A., Morales-Nin B. 2007 . Small pelagic fish in the NW Mediterranean Sea: an ecological review . Progress in Oceanography 74 : 377 – 396 . Google Scholar Crossref Search ADS WorldCat Petitgas P. , Alheit J., Peck M. A., Raab K., Irigoien X., Huret M., Van Der Kooij J., et al. 2012 . Anchovy population expansion in the North Sea . Marine Ecology-Progress Series 444 : 1 – 13 . Google Scholar Crossref Search ADS WorldCat Petitgas P. , Rijnsdorp A. D., Dickey‐Collas M., Engelhard G. H., Peck M. A., Pinnegar J. K., Drinkwater K., et al. 2013 . Impacts of climate change on the complex life cycles of fish . Fisheries Oceanography 22 : 121 – 139 . Google Scholar Crossref Search ADS WorldCat Philips, S. J., Anderson, R. P., and Shapire, R. E. 2006. Maximum entropy modeling of species geographic distributions. Ecological Modelling, 190: 231–259. Pinsky M. L. , Fogarty M. 2012 . Lagged social-ecological responses to climate and range shifts in fisheries . Climatic Change 115 : 883 – 891 . Google Scholar Crossref Search ADS WorldCat Planque B. , Bellier E., Lazure P. 2007 . Modelling potential spawning habitat of sardine (Sardina pilchardus) and anchovy (Engraulis encrasicolus) in the Bay of Biscay . Fisheries Oceanography 16 : 16 – 30 . Google Scholar Crossref Search ADS WorldCat Raab K. , Llope M., Nagelkerke L. A., Rijnsdorp A. D., Teal L. R., Licandro P., Ruardij P., et al. 2013 . Influence of temperature and food availability on juvenile European anchovy Engraulis encrasicolus at its northern boundary . Marine Ecology. Progress Series 488 : 233 – 245 . Google Scholar Crossref Search ADS WorldCat Raybaud V. , Beaugrand G., Dewarumez J. M., Luczak C. 2015 . Climate-induced range shifts of the American jackknife clam Ensis directus in Europe . Biological Invasions 17 : 725 – 741 . Google Scholar Crossref Search ADS WorldCat Raybaud V. , Beaugrand G., Goberville E., Delebecq G., Destombe C., Valero M., Davoult D., et al. 2013 . Decline in kelp in west Europe and climate . PLos One 8 : e66044. Google Scholar Crossref Search ADS PubMed WorldCat Raybaud V. , Nival P., Prieur L. 2011 . Short time-scale analysis of the NW Mediterranean ecosystem during summer-autumn transition: A 1D modelling approach . Journal of Marine Systems 84 : 1 – 17 . Google Scholar Crossref Search ADS WorldCat Roy C. , Van der Lingen C., Coetzee J., Lutjeharms J. 2007 . Abrupt environmental shift associated with changes in the distribution of Cape anchovy Engraulis encrasicolus spawners in the southern Benguela . African Journal of Marine Science 29 : 309 – 319 . Google Scholar Crossref Search ADS WorldCat Salomon A. K. , Gaichas S. K., Shears N. T., Smith J. E., Madin E. M., Gaines S. D. 2010 . Key features and context‐dependence of fishery‐induced trophic cascades . Conservation Biology 24 : 382 – 394 . Google Scholar Crossref Search ADS PubMed WorldCat Santojanni A. , Arneri E., Bernardini V., Cingolani N., Di Marco M., Russo A. 2006 . Effects of environmental variables on recruitment of anchovy in the Adriatic Sea . Climate Research 31 : 181 – 193 . Google Scholar Crossref Search ADS WorldCat Schismenou E. , Giannoulaki M., Valavanis V. D., Somarakis S. 2008 . Modeling and predicting potential spawning habitat of anchovy (Engraulis encrasicolus) and round sardinella (Sardinella aurita) based on satellite environmental information . Hydrobiologia 612 : 201 – 214 . Google Scholar Crossref Search ADS WorldCat Schmidt-Nielsen K. 1997 . Animal Physiology: Adaptation and Environment . Cambridge University Press , Cambridge . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Shiganova T. 1998 . Invasion of the Black Sea by the ctenophore Mnemiopsis leidyi and recent changes in pelagic community structure . Fisheries Oceanography 7 : 305 – 310 . Google Scholar Crossref Search ADS WorldCat Silva G. , Horne J. B., Castilho R. 2014a . Anchovies go north and west without losing diversity: post‐glacial range expansions in a small pelagic fish . Journal of Biogeography , 41: 1171–1182. OpenURL Placeholder Text WorldCat Silva G. , Lima F. P., Martel P., Castilho R. 2014b . Thermal adaptation and clinal mitochondrial DNA variation of European anchovy . Proceedings of the Royal Society of London, Series B: Biological Sciences 281 : 20141093 . Google Scholar Crossref Search ADS WorldCat Smith W. H. F. , Sandwell D. T. 1997 . Global sea floor topography from satellite altimetry and ship depth soundings . Science 277 : 1956 – 1962 . Google Scholar Crossref Search ADS WorldCat Sumaila U. R. , Cheung W. W., Lam V. W., Pauly D., Herrick S. 2011 . Climate change impacts on the biophysics and economics of world fisheries . Nature Climate Change 1 : 449 – 456 . Google Scholar Crossref Search ADS WorldCat Swets, J. A. 1988. Measuring the accuracy of diagnostic systems. Science, 240: 1285–1293. Taylor K. E. 2001 . Summarizing multiple aspects of model performance in a single diagram . Journal of Geophysical Research 106 : 7183 – 7192 . Google Scholar Crossref Search ADS WorldCat Tittensor, D. P., Baco, A. R., Brewin, P. E., Clark, M. R., Consalvey, M., Hall-Spencer, J., Rowden, A. A., et al. 2009. Predicting global habitat suitability for stony corals on seamounts. Journal of Biogeography, 36: 1111–1128. Tugores M. P. , Iglesias M., Díaz N., Oñate D., Miquel J., Giráldez A. 2010 . Latitudinal and interannual distribution of the European anchovy (Engraulis encrasicolus) and sardine (Sardina pilchardus) in the western Mediterranean, and sampling uncertainty in abundance estimates . ICES Journal of Marine Science: Journal Du Conseil 67 : 1574 – 1586 . Google Scholar Crossref Search ADS WorldCat Warren D. L. , Glor R. E., Turelli M. 2008 . Environmental niche equivalency versus conservatism: quantitative approaches to niche evolution . Evolution 62 : 2868 – 2883 . Google Scholar Crossref Search ADS PubMed WorldCat Wiley, E., Mcnyset, K. M., Peterson, A. T., Robins, C. R., and Stewart, A. M. 2003. Niche modeling and geographic range predictions in the marine environment using a machine-learning algorithm. Oceanography, 16: 120–127. © International Council for the Exploration of the Sea 2017. All rights reserved. For Permissions, please email: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Westgaard,, Jon-Ivar;Staby,, Arved;Aanestad Godiksen,, Jane;Geffen, Audrey, J.;Svensson,, Anders;Charrier,, Gregory;Svedäng,, Henrik;André,, Carl

doi: 10.1093/icesjms/fsw249pmid: N/A

Recently, there have been reports of increased abundance and landings of European hake in the northern part of the species range. Biological studies are however scarce and information about finer scale population structure important for stock assessments and fishery management is largely lacking. Here, we report on a population genetic study using neutral and outlier SNP loci assessing population structure in hake in the north-eastern parts of its range in the Atlantic. Hake samples from localities along the west coast of Norway, the Kattegat, the northern North Sea, and one locality in the Bay of Biscay were analysed using 53 SNPs, six of which were outliers potentially influenced by natural selection. We detected small-scale structure among northern samples, all of which were also distinct from Bay of Biscay hake, with the exception of a few individuals from the North Sea and the coast of Norway who clustered genetically together with Bay of Biscay hake. Our findings suggest that the present management unit of a single northern stock of hake is not biologically correct, and that there is more detail in the fine-scale population structure indicating that independent population dynamics could be expected in response to fishing patterns or changing environmental conditions. Introduction Identification of population subdivision in harvested marine fish is of vital importance since undetected population structure may lead to overfishing of local populations, their subsequent decline in biomass, and in the worst-case scenario, extirpation of subpopulations (Kell et al., 2009), or a collapse of the resource (Hutchings, 2000; Svedäng, 2003; Svedäng and Bardon, 2003). The European hake is a commercially important fish that is common in the Northeast Atlantic and the Mediterranean (Murua, 2010). Its distribution in the Northeast Atlantic stretches from the coast off Northwest Africa into Portuguese and Spanish waters, to the west of France, Ireland and Scotland, into the North Sea and Kattegat, and along the north-western Norwegian coast (Hickling, 1927; Bergstad, 1991; Casey and Pereiro, 1995). European hake has been heavily fished for centuries in the Mediterranean and around the Iberian Peninsula and the Bay of Biscay (e.g. Fortibuoni et al., 2010). The biology, ecology, distribution, and population structure are relatively well studied in this southern part of the range. Hake is also fished commercially in the Celtic Sea, west of Scotland, and in the North Sea, but it is not as highly valued in the northern part of the range (Statistics Norway; FAO). Consequently, there is less information available on the biology and status of this species in areas above 56oN. Despite the lack of supporting evidence, hake in the Northeast Atlantic is arbitrarily divided into two stocks for management purposes–northern and southern stocks (Roldan et al., 1998). The division between the northern and southern stocks lies along the Cap Breton canyon, which extends west from the Bay of Biscay, and is assumed to hinder movement of fish between the Iberian Peninsula to the south and the northern Bay of Biscay and Celtic Sea to the north (Roldan et al., 1998). Several studies have shown that this division into two management stocks ignores the biological evidence for population structure. Hake caught in the Celtic Sea and the Bay of Biscay are similar genetically and distinct from fish farther south along the Iberian coast (Lundy et al., 1999; Castillo et al., 2004; Milano et al., 2011; Pita et al., 2011). Fisheries landings and scientific survey data from north of Ireland have indicated strong increases in abundance, especially in the northern North sea with larger commercial catches since 2005 (Cormon et al., 2014; Baudron and Fernandes, 2015). Records describing European hake in Norwegian and adjacent waters date to the turn of the 19th century (Brunchorst, 1898; Schmidt, 1909). Official catch records for the Norwegian fishery show that hake have been caught in the North Sea and along the entire Norwegian coast, from the eastern Skagerrak in the south-east and as far north as the Lofoten Islands, since 1935 (Norges_Fiskerier, 1935). Historically landings were low (less than 1000 tonnes), but the last decade has seen a steep increase in spawning stock biomass in the North Sea (Baudron and Fernandes, 2015), and concurrently Norwegian landings increased sixfold during the same period (Bakketeig et al., 2015). One of the major questions arising from such a change is whether the increase represents an increase in population abundance or a change in population distribution patterns, or both. In the Skagerrak and Kattegat, European hake is a frequent demersal fish species (DATRAS database at ICES, www.ices.dk) where it has been targeted for centuries (Smitt, 1892). It spawns during summer in shallow areas; ripe hake are found, for instance, near Kummelbank (“Hake bank” in Swedish) at depths between 30–70 m. Information drawn from commercial fisheries shows that hake are mostly caught between June and September on the western side of the Norwegian Trench off the Jutland coast, i.e. close to the alleged spawning grounds (Smitt, 1892). Hake are often taken as by-catch and the species is of minor importance for the Swedish fishery, amounting to 50–100 tonnes annually. The processes behind the apparent increase in hake abundance at more northern latitudes are difficult to evaluate because the population structure of hake is poorly known in this part of its distribution range. Indeed, few studies have investigated more than one sample of the “northern stock” outside the Celtic Sea. Lundy et al. (1999) analysed a large sample of fish from the west coast of Norway (Trondheimsfjord) and found that fish from Norway were genetically distinct from those sampled in the Celtic Sea and the Bay of Biscay. Significant differences in otolith composition were also found between hake from the west coast of Norway (Romsdalsfjord), Reykjanes Ridge, and the Portuguese coast (Swan et al., 2006). Tanner et al. (2014), on the other hand, found some evidence of exchange between hake sampled off Galicia and the Celtic sea to the north and Portugal to the south, based on the otolith elemental composition. In a SNP marker-based study, Milano et al. (2014) included three sampling locations in the northern range: Celtic Sea, west coast of Scotland, and northern North Sea. They found no genetic differentiation among these three samples with neutral loci, but a clear divergence using “outlier” loci, potentially affected by natural selection. These results underline the utility of considering not only neutral genetic markers but also gene-associated markers for stock identification (Nielsen et al., 2012; Berg et al., 2015; Gagnaire et al., 2015). As there is a limited amount of data regarding the genetic population structure in the northern part of the distribution range of European hake, we aimed at resolving this question. By including two fjord samples from the Norwegian coast, one offshore sample from the North Sea and two samples from Kattegat, we were able to cover major parts of the species' northern range. One sample from the Bay of Biscay was included as an outgroup. An additional objective was to correlate genetic structure to environmental factors, such as temperature and salinity, as such correlations have previously been indicated (e. g. Milano et al., 2014). Material and methods Samples Fish were collected between 2004 and 2012 at six locations (Table 1, Figure 1). Hake from the Norwegian Sea off central Norway (TRØ) were sampled at three locations (two inside fjord systems and one outside) toward the end of the spawning season in October and consisted mainly of mature but spent fish. Hake from Fanafjord (FAN) on the Norwegian west coast were collected from August to October and consisted mainly of immature fish and only a few non-spawning adults (> 35 cm). The North Sea fish (NS) were caught at two locations in July and consisted of both immature (< 40 cm) and maturing/spawning fish (Figure 2). The Kattegat hake sample consisted of fish from two locations: KAN representing large spawning hake at Kummelbank (Hake bank in Swedish) during summer, and KAS consisting of juveniles and smaller sized adults during the non-spawning season from a large part of the Kattegat (Figure 2). The Bay of Biscay (BB) sample was collected outside the spawning season and was composed of both immature (<∼30 cm) and adult (>30 cm) hake (Figure 2). Figure 1 Open in new tabDownload slide Map showing sampling locations (crosses) and surface currents (solid lines indicate Atlantic water, and dotted lines coastal water). The division between “southern” and “northern” European hake stocks is shown as a dashed line. Stations sampled, respectively, in the North Sea (NS) and Trøndelag (TRØ) are shown in circles. The highlighted area of the Kattegat shows locations of samples from the northern Kattegat (KAN) and southern Kattegat (KAS). Abbreviations of other sampling locations: Fanafjorden (FAN) an Bay of Biscay (BB). Figure 1 Open in new tabDownload slide Map showing sampling locations (crosses) and surface currents (solid lines indicate Atlantic water, and dotted lines coastal water). The division between “southern” and “northern” European hake stocks is shown as a dashed line. Stations sampled, respectively, in the North Sea (NS) and Trøndelag (TRØ) are shown in circles. The highlighted area of the Kattegat shows locations of samples from the northern Kattegat (KAN) and southern Kattegat (KAS). Abbreviations of other sampling locations: Fanafjorden (FAN) an Bay of Biscay (BB). Figure 2 Open in new tabDownload slide Length-frequency distribution of fish in samples included in this study. Locations indicated in the top bar. Figure 2 Open in new tabDownload slide Length-frequency distribution of fish in samples included in this study. Locations indicated in the top bar. Table 1 Overview of sample origin, abbreviations used in the text (material used for DNA extraction: M muscle tissue, O otolith), year samples were collected and sample size. Location . Abbreviation . Year . N . HE . HO . FIS . Trøndelag TRØ (M) Oct/Nov 2012 56 0.283 0.316 –0.117 Fanafjord FAN (O) Sept/Oct 2004 47 0.244 0.272 0.103 North sea NS (O) July 2012 74 0.302 0.303 –0.003 Kummelbank KAN (M) July 2010 53 0.318 0.282 –0.128 Kattegat KAS (M) Nov/Dec 2010 38 0.285 0.299 0.049 Bay of Biscay BB (M) Nov/Dec 2012 94 0.271 0.288 0.058 Location . Abbreviation . Year . N . HE . HO . FIS . Trøndelag TRØ (M) Oct/Nov 2012 56 0.283 0.316 –0.117 Fanafjord FAN (O) Sept/Oct 2004 47 0.244 0.272 0.103 North sea NS (O) July 2012 74 0.302 0.303 –0.003 Kummelbank KAN (M) July 2010 53 0.318 0.282 –0.128 Kattegat KAS (M) Nov/Dec 2010 38 0.285 0.299 0.049 Bay of Biscay BB (M) Nov/Dec 2012 94 0.271 0.288 0.058 Also included is observed and expected heterozygosity, HO and HE, respectively, and the inbreeding coefficient (FIS). Table 1 Overview of sample origin, abbreviations used in the text (material used for DNA extraction: M muscle tissue, O otolith), year samples were collected and sample size. Location . Abbreviation . Year . N . HE . HO . FIS . Trøndelag TRØ (M) Oct/Nov 2012 56 0.283 0.316 –0.117 Fanafjord FAN (O) Sept/Oct 2004 47 0.244 0.272 0.103 North sea NS (O) July 2012 74 0.302 0.303 –0.003 Kummelbank KAN (M) July 2010 53 0.318 0.282 –0.128 Kattegat KAS (M) Nov/Dec 2010 38 0.285 0.299 0.049 Bay of Biscay BB (M) Nov/Dec 2012 94 0.271 0.288 0.058 Location . Abbreviation . Year . N . HE . HO . FIS . Trøndelag TRØ (M) Oct/Nov 2012 56 0.283 0.316 –0.117 Fanafjord FAN (O) Sept/Oct 2004 47 0.244 0.272 0.103 North sea NS (O) July 2012 74 0.302 0.303 –0.003 Kummelbank KAN (M) July 2010 53 0.318 0.282 –0.128 Kattegat KAS (M) Nov/Dec 2010 38 0.285 0.299 0.049 Bay of Biscay BB (M) Nov/Dec 2012 94 0.271 0.288 0.058 Also included is observed and expected heterozygosity, HO and HE, respectively, and the inbreeding coefficient (FIS). Muscle tissue was the source of DNA for TRØ, KAN, KAS, and BB hake, and otoliths for FAN and NS hake were used to extract DNA due to the lack of tissue samples. Here after, the term “sample” is defined as a collection of fish from a specific location. Genetic analysis DNA from otoliths was isolated using the QIAamp DNA Micro Kit (Qiagen N. V.) with two modifications: (i) the volume of the lysate was increased by 100% to cover as much of the otolith as possible, and (ii) samples were incubated in a thermo mixer at 56ºC and 750 rpm for 24 h. The elution volume was set to 30 µl. DNA from muscle tissue was isolated using the E.Z.N.A 96 kit (Omega Bio-Tek) according to the manufacturer’s protocol. SNP loci previously characterized as neutral and outliers, possibly affected by selection within the Atlantic, were selected based on the work of Milano et al. (2011). We selected a subset of the 381 SNP markers used by Milano et al. (2014) based on the following criteria: (i) a set of SNPs detected as outliers within the Atlantic (22) and (ii) a random selection of “neutral” SNPs (61). This resulted in a total of 83 SNP markers arranged in three multiplexes that were genotyped using matrix-assisted laser desorption/ionization time-of-flight mass spectroscopy (MALDI-TOF MS) assays. Primers used for genotyping are listed in Supplementary data Table S1. Genotyping was performed using the IPLEX protocol following the manufacturer's instructions (Agena Bioscience Inc., Hamburg, Germany). The MassARRAY Typer software was used for automated genotype calling. Loci with more than 10% missing data per sample and a minor allele frequency below 5% were discarded. This resulted in the removal of 24 loci. Thus, the final dataset consisted of 59 SNP loci. Genetic variation Genetic variability at SNP markers was evaluated through basic population genetic parameters such as observed and expected heterozygosities, gametic disequilibrium and conformity to gametic and Hardy-Weinberg expectations, calculated using the software Arlequin 3.5 (Excoffier and Lischer, 2010). In cases of multiple testing, the Type I error rates were corrected using the False Discovery Rate (FDR) approach (Benjamini and Hochberg, 1995). Sampling stations were pooled if they were in close proximity and after having been tested for population differentiation among stations from the same location (see Figure 1). This increased the sample size and thereby statistical power, especially for TRØ and NS. Outlier detection Genetic loci were tested for diversifying or balancing selection using outlier detection tests. These tests assess the probability that the loci reflect neutral population genetic processes such as drift and gene flow among populations rather than local adaptation or selection (e. g. Narum and Hess, 2011; Gagnaire et al., 2015). We used two approaches to identify such loci. First, a coalescent-based simulation approach was used to identify outlier loci potentially influenced by natural selection. This was done by comparing observed FST at each locus with expected values under neutrality. Then, loci displaying unusually high and low FST values were characterized as outliers (Beaumont and Nichols, 1996). This procedure is implemented in the Lositan Selection Workbench (Antao et al., 2008). An initial run was performed with 50 000 simulations over all loci, using the mean neutral FST as a preliminary value. A more accurate estimate of the mean neutral FST was obtained by excluding all loci lying outside the 99% confidence interval following the first run, as their distribution could be a result of selection rather than neutral evolution. The refined estimate was used for a final set of 50 000 simulations over all loci, and this approach was used both on the global data and in pairwise tests between locations. We also employed a Bayesian simulation-based test implemented in BayeScan 2.1 (Foll and Gaggiotti, 2008). The locus-population FST coefficients were separated into a population specific component across all loci (β) and a locus-specific component across all populations (α). An α significantly different from 0 (negative or positive) indicates diversifying or balancing selection, respectively. The software uses a reversible-jump MCMC algorithm to estimate the posterior probability that a locus is showing signs of diversifying or balancing selection in relation to the global FST. We based our analysis on ten pilot runs each consisting of 5000 iteration, followed by 100 000 iterations with a burn-in of 50 000 iterations. Population structure An initial estimation of genetic structure was performed using pairwise FST values assessed according to Weir and Cockerham (1984) as implemented in Genepop 4.3 (Rousset, 2008). A Pearson’s traditional χ2 test in CHIFISH (Ryman, 2006) was used to assess whether the samples were significantly different from each other. The chi-squared test was chosen because it has been shown that it performs best when summing p–values across bi-allelic loci (Ryman and Jorde, 2001). The robustness of the observed structure and the direction of gene flow was assessed using the assignment test implemented in Arlequin 3.5 (Excoffier and Lischer, 2010). Here, the log likelihood that the genotype of each individual belongs to each population sample is computed as if it was drawn from a population having equal allele frequencies to those estimated for each sample. Since our data comprise both samples from different years (temporal samples) and a mix of juveniles and adults, we performed an AMOVA analysis (Excoffier et al., 1992) implemented in the Arlequin 3.5 (Excoffier and Lischer, 2010) to assess the influence of these factors on the genetic structure. To account for temporal samples, we grouped the data according to the year sampled. If temporal sampling contributed more than spatial sampling, then the variation among groups (year; FCT) is expected to be larger than the variation within groups (spatial location; FSC). Assessing if the genetic structure depended on whether juveniles or adults were analysed, we used a similar approach as described above. Here each sample was split into two size groups, above and below 30 cm total length (adults and juveniles, respectively). Total length was used as it was available for all individuals and was considered as a proxy for age. Differences in the genetic structure between adults and juveniles should then reflect variation among age groups (adults/juveniles; FCT) and may be larger than variation within groups (spatial location; FSC). This approach was, however, hampered by small sample sizes in some cases, especially for the TRØ and KAT samples (n = 9 and 14, respectively), and for the KAN sample, which included no individuals with a total length below 30 cm. An individual-based analysis of population structure was conducted by estimating the number of clusters (K) in the data using the Bayesian cluster analysis implemented in BAPS 6.0 (Corander et al., 2006). BAPS treats K as an unknown parameter and uses a stochastic optimization algorithm to estimate the posterior mode of K by modelling the underlying population allele frequencies assuming Hardy-Weinberg equilibrium and non-linkage of markers (Corander et al., 2003). The analysis in BAPS was made with two models incorporated in the software. First we used the “clustering of individuals” model, where only the genetic data were considered. The second model, “spatial clustering of individuals”, incorporates spatial data as a prior as described in Corander et al. (2008). A predefined number of clusters (K), here 1–10, were explored over 10 different runs to ensure proper replication for the given number of clusters. This approach was pursued for both steps. Genotype-environment association The correlation between the genetic data and the environmental variables temperature and salinity was explored with multiple regression on distance matrices (MRDM) using the “ecodist” package in R (Goslee and Urban, 2007). This analysis was based on a single dependant matrix, here the FST matrix based on all 53 loci, expressed as a function of several independent matrices, which were temperature and salinity at the surface and 25 meters depth. The analysis was conducted both with and without the Bay of Biscay sample. Annual mean values of seawater temperature and salinity at the different depths were retrieved from the National Oceanographic Data Center (NODC) database (Boyer et al., 2013), using the closest geographic coordinates to the actual sampling locations. In addition, the relationship between outlier SNP loci and environmental variables was tested with a simple linear model, where allele frequencies at each of the six loci were fitted against the environmental variables temperature and salinity at different depths. Results Genetic variation The observed and expected heterozygosity within loci (HO) ranged from 0 to 0.723 and 0.018 to 0.527, respectively (Supplementary data Table S1) and was similar for samples from which the DNA originated from otoliths or muscle (Supplementary data Table S1). Conformity to Hardy-Weinberg expectations was violated in 6 of 318 tests. Of these six significant tests, three were at the locus X2592_fpt, and the remaining five were scattered across the remaining loci (Supplementary data Table S1). Evidence of gametic disequilibrium was discovered in 10 locus-pair combinations in a majority of the samples. The most informative SNPs in these combinations were used in the subsequent analysis. A total of six loci were discarded. Details are found in Supplementary data Table S3. Outlier detection The global test for outlier loci returned 6 of 53 loci likely to be influenced by positive directional selection as suggested by the results from Lositan. The BayeScan analysis suggested four outlier loci, which were also found by Lositan. Thus, combining the results from the two software we found four outlier loci in common (Figure 3, Supplementary data Table S2). In the study of Milano et al. (2014), three of these loci were also reported as outliers. The one outlier unique to the present study was x890_fpt. The analysis for the northern samples, without the Bay of Biscay sample, returned two outlier loci in common for the two software (results not shown), where one (X778ms) was in common with the global dataset including all six samples. The second outlier locus (X891_fpt) was detected only in the dataset without the Bay of Biscay sample. Only the four outlier loci detected by both software were considered true outliers and treated accordingly in the subsequent analysis. This applies both to the global dataset and the dataset with only the five northern samples. Figure 3 Open in new tabDownload slide Outlier detection among 53 SNP loci using (a) FST vs. heterozygosity for the 53 SNP loci from the software Lositan (Antao et al., 2008), based on the Fdist approach (Beaumont and Nichols, 1996). The dark grey area indicates the upper 99% confidence interval, while the light grey area indicates the lower 99% confidence interval. The black line in the middle of the plot shows mean FST. (b) Results from the software BayeScan (Foll and Gaggiotti, 2008), showing the logarithmic Bayes factor on the X-axis and FST values on the Y-axis. The vertical line indicates the threshold for significant outliers, which was set to “decisive”, corresponding to a posterior probability of 0.99. Figure 3 Open in new tabDownload slide Outlier detection among 53 SNP loci using (a) FST vs. heterozygosity for the 53 SNP loci from the software Lositan (Antao et al., 2008), based on the Fdist approach (Beaumont and Nichols, 1996). The dark grey area indicates the upper 99% confidence interval, while the light grey area indicates the lower 99% confidence interval. The black line in the middle of the plot shows mean FST. (b) Results from the software BayeScan (Foll and Gaggiotti, 2008), showing the logarithmic Bayes factor on the X-axis and FST values on the Y-axis. The vertical line indicates the threshold for significant outliers, which was set to “decisive”, corresponding to a posterior probability of 0.99. Population structure Pairwise estimates of genetic differentiation, FST, showed a high degree of structure (Table 2), reflected by the number of significant tests, both with (15/15) and without (14/15) outlier loci (Table 2). The pattern of genetic differentiation among population samples for all 53 SNPs was congruent with their geographical distribution. The Bay of Biscay sample was most differentiated, with FST values ranging from 0.027 to 0.056, while the genetic structure was much lower among northern samples, with FST between 0.002 and 0.017 (Table 2). Excluding the four outlier loci resulted in markedly reduced, but still significant, differentiation between Bay of Biscay and the northern samples. Among the northern samples, only minor changes in the levels of differentiation were observed with this approach. That is, the outliers’ influence of the genetic structure was most profound in differentiating the Bay of Biscay and Scandinavia. Table 2 Population-pairwise genetic differentiation (FST). . TRØ . FAN . NS . KAN . KAS . BB . TRØ * 0.010 0.008 0.009 0.015 0.014 FAN 0.009 * 0.002 0.015 0.007 0.006 NS 0.008 0.005 * 0.004 0.001 0.011 KAN 0.006 0.017 0.009 * 0.012 0.016 KAS 0.013 0.010 0.008 0.012 * 0.011 BB 0.054 0.046 0.032 0.069 0.062 * . TRØ . FAN . NS . KAN . KAS . BB . TRØ * 0.010 0.008 0.009 0.015 0.014 FAN 0.009 * 0.002 0.015 0.007 0.006 NS 0.008 0.005 * 0.004 0.001 0.011 KAN 0.006 0.017 0.009 * 0.012 0.016 KAS 0.013 0.010 0.008 0.012 * 0.011 BB 0.054 0.046 0.032 0.069 0.062 * All 53 SNPs, including outlier loci below diagonal and 49 neutral SNPs above diagonal. Significant values based on Pearson’s traditional χ2 test are indicated in bold. Table 2 Population-pairwise genetic differentiation (FST). . TRØ . FAN . NS . KAN . KAS . BB . TRØ * 0.010 0.008 0.009 0.015 0.014 FAN 0.009 * 0.002 0.015 0.007 0.006 NS 0.008 0.005 * 0.004 0.001 0.011 KAN 0.006 0.017 0.009 * 0.012 0.016 KAS 0.013 0.010 0.008 0.012 * 0.011 BB 0.054 0.046 0.032 0.069 0.062 * . TRØ . FAN . NS . KAN . KAS . BB . TRØ * 0.010 0.008 0.009 0.015 0.014 FAN 0.009 * 0.002 0.015 0.007 0.006 NS 0.008 0.005 * 0.004 0.001 0.011 KAN 0.006 0.017 0.009 * 0.012 0.016 KAS 0.013 0.010 0.008 0.012 * 0.011 BB 0.054 0.046 0.032 0.069 0.062 * All 53 SNPs, including outlier loci below diagonal and 49 neutral SNPs above diagonal. Significant values based on Pearson’s traditional χ2 test are indicated in bold. Despite smaller FST values, significant genetic structure was present among the northern population samples, for both sets of SNPs. In general, Trøndelag was the most differentiated sample among the northern samples (without Bay of Biscay) (Table 2). Interestingly, adults (Kummelbank) and juveniles (Kattegat) were significantly differentiated from each other within the Kattegat (FST = 0.012). However, the North Sea sample was weakly differentiated from the Kattegat (Kummelbank and Kattegat) and southern Norwegian (Fanafjord) samples (FST < 0.01). North Sea and Kattegat were not significantly differentiated when considering only the 49 “neutral” SNPs (FST = 0.001, Table 2). The robustness of the population structure was shown in the assignment test results, which confirmed the pattern of the FST estimates. In general, the majority of the individuals from Kattegat, Fanafjord and Bay of Biscay were reassigned to the sample of origin (Table 3a). Fish from the spawning aggregation at Kummelbank were less often reassigned to the sample of origin, as were hake collected at Trøndelag and North Sea. The assignment without outlier loci showed a similar pattern, but the self-assignment rate was lower for the Bay of Biscay population (Table 3b). The AMOVA analysis showed that sampling year did not contribute to the genetic variation (Table 4). Also, a comparison of juveniles vs. adults did not show any difference in structure between life stages (Table 4). Table 3 Genetic assignment showing the number of individuals from each sample (rows) assigned to each of the samples (column). (a) . . . . . . . . TRØ12 . FAN04 . NS12 . KAN10 . KAS10 . BB12 . TRØ12 31 0 3 2 19 1 FAN04 2 38 1 1 3 2 NS12 4 6 26 2 24 12 KAN10 11 4 3 24 11 0 KAS10 1 0 1 0 36 0 BB12 1 4 4 0 6 79 (b) TRØ12 FAN04 NS12 KAN10 KAS10 BB12 TRØ12 28 0 5 5 15 3 FAN04 1 38 2 0 2 4 NS12 2 4 25 5 27 11 KAN10 8 4 5 22 10 4 KAS10 1 0 2 1 32 2 BB12 3 4 3 2 29 54 (a) . . . . . . . . TRØ12 . FAN04 . NS12 . KAN10 . KAS10 . BB12 . TRØ12 31 0 3 2 19 1 FAN04 2 38 1 1 3 2 NS12 4 6 26 2 24 12 KAN10 11 4 3 24 11 0 KAS10 1 0 1 0 36 0 BB12 1 4 4 0 6 79 (b) TRØ12 FAN04 NS12 KAN10 KAS10 BB12 TRØ12 28 0 5 5 15 3 FAN04 1 38 2 0 2 4 NS12 2 4 25 5 27 11 KAN10 8 4 5 22 10 4 KAS10 1 0 2 1 32 2 BB12 3 4 3 2 29 54 Highlighted in bold is the number of individuals assigned to the sample of origin (diagonal). The individual assigned was not included in the reference population (the leave one out procedure). (a) With all 53 loci, and (b) 49 neutral loci only. The numbers behind sample abbreviations indicate sampling year. Table 3 Genetic assignment showing the number of individuals from each sample (rows) assigned to each of the samples (column). (a) . . . . . . . . TRØ12 . FAN04 . NS12 . KAN10 . KAS10 . BB12 . TRØ12 31 0 3 2 19 1 FAN04 2 38 1 1 3 2 NS12 4 6 26 2 24 12 KAN10 11 4 3 24 11 0 KAS10 1 0 1 0 36 0 BB12 1 4 4 0 6 79 (b) TRØ12 FAN04 NS12 KAN10 KAS10 BB12 TRØ12 28 0 5 5 15 3 FAN04 1 38 2 0 2 4 NS12 2 4 25 5 27 11 KAN10 8 4 5 22 10 4 KAS10 1 0 2 1 32 2 BB12 3 4 3 2 29 54 (a) . . . . . . . . TRØ12 . FAN04 . NS12 . KAN10 . KAS10 . BB12 . TRØ12 31 0 3 2 19 1 FAN04 2 38 1 1 3 2 NS12 4 6 26 2 24 12 KAN10 11 4 3 24 11 0 KAS10 1 0 1 0 36 0 BB12 1 4 4 0 6 79 (b) TRØ12 FAN04 NS12 KAN10 KAS10 BB12 TRØ12 28 0 5 5 15 3 FAN04 1 38 2 0 2 4 NS12 2 4 25 5 27 11 KAN10 8 4 5 22 10 4 KAS10 1 0 2 1 32 2 BB12 3 4 3 2 29 54 Highlighted in bold is the number of individuals assigned to the sample of origin (diagonal). The individual assigned was not included in the reference population (the leave one out procedure). (a) With all 53 loci, and (b) 49 neutral loci only. The numbers behind sample abbreviations indicate sampling year. Table 4 Results from the AMOVA test (Excoffier et al., 1992), assessing (i) the impact of sampling year on the genetic structure and (ii) differences in genetic structure between juveniles and adults. Source of variation . Percentage variation . p . The effect of sampling year  Among sampling years –0.12 0.55  Among population within years 3.49 0.00  Within populations 96.63 Juveniles vs. Adults  Among life stages –0.14 0.43  Among populations within life stages 3.21 0.00  Within populations 96.94 Source of variation . Percentage variation . p . The effect of sampling year  Among sampling years –0.12 0.55  Among population within years 3.49 0.00  Within populations 96.63 Juveniles vs. Adults  Among life stages –0.14 0.43  Among populations within life stages 3.21 0.00  Within populations 96.94 Table 4 Results from the AMOVA test (Excoffier et al., 1992), assessing (i) the impact of sampling year on the genetic structure and (ii) differences in genetic structure between juveniles and adults. Source of variation . Percentage variation . p . The effect of sampling year  Among sampling years –0.12 0.55  Among population within years 3.49 0.00  Within populations 96.63 Juveniles vs. Adults  Among life stages –0.14 0.43  Among populations within life stages 3.21 0.00  Within populations 96.94 Source of variation . Percentage variation . p . The effect of sampling year  Among sampling years –0.12 0.55  Among population within years 3.49 0.00  Within populations 96.63 Juveniles vs. Adults  Among life stages –0.14 0.43  Among populations within life stages 3.21 0.00  Within populations 96.94 The most likely number of genetic clusters was K = 4, when using all 53 loci and K = 3, using the 49 “neutral” SNPs. The individual-based BAPS analysis corroborated much of the findings from the pairwise FST estimates in that most of the individuals from the Bay of Biscay were placed in two clusters considering all 53 loci (shown in red and yellow; Figure 4a) and a shallower structure among the northern samples, with a majority of the individuals placed within the same cluster (blue; Figure 4a). However, a higher proportion of the individuals from the samples Trøndelag, North Sea and Fanafjord were placed in the Bay of Biscay group, which was most pronounced for the North Sea sample (Figure 4a). This feature was also found in the FST estimates (Table 2). Considering the 49 “neutral” loci (K = 3), we found a more erratic pattern with the Bay of Biscay sample no longer as differentiated from the northern samples. Thus the influence of the four outlier loci on the genetic structure is clearly demonstrated (Figure 4b). Figure 4 Open in new tabDownload slide Individual genetic clustering using BAPS (Corander et al., 2006). Bar plot with K = 4 clusters, using all 53 loci including four outlier loci (a), where the Bay of Biscay cluster is dominated by red and yellow, and K = 3 using 49 neutral loci (b), demonstrating the effect of the outlier loci in which the Bay of Biscay cluster is no longer well defined. Each bar represents one individual and the different colours represent the cluster to which the individual belongs. Figure 4 Open in new tabDownload slide Individual genetic clustering using BAPS (Corander et al., 2006). Bar plot with K = 4 clusters, using all 53 loci including four outlier loci (a), where the Bay of Biscay cluster is dominated by red and yellow, and K = 3 using 49 neutral loci (b), demonstrating the effect of the outlier loci in which the Bay of Biscay cluster is no longer well defined. Each bar represents one individual and the different colours represent the cluster to which the individual belongs. Genotype-environment association Correlations of the FST-based differences from the four outlier loci only (all samples) with the environmental variables showed a significant association between genetic distance and temperature at both depths (Table 5). A weaker, but significant correlation was also found between FST and salinity at 25 m depth. Removing the Bay of Biscay sample led to no significant association between the genetic and environmental data (Table 5), suggesting that the environmental heterogeneity among the northern localities was not great enough to act as a selective agent. At the individual locus level, all four outliers were correlated to at least one of the environmental variables, temperature and salinity (Table 6). This was most pronounced regarding temperature. Table 5 p-Values for the correlation between genetic differentiation (FST) and environmental variables. . T0 . T25 . S0 . S25 . All six samples 0.0002 0.0010 0.4634 0.0410 Northern samples 0.8993 0.1445 0.2414 0.3907 . T0 . T25 . S0 . S25 . All six samples 0.0002 0.0010 0.4634 0.0410 Northern samples 0.8993 0.1445 0.2414 0.3907 FST was calculated for all six samples based on four outlier loci, or the five northern samples only based on two outlier loci. T0 and T25 refer to temperature at the surface (0 meters) and 25 meters, respectively. S refers to salinity at the same depths. Table 5 p-Values for the correlation between genetic differentiation (FST) and environmental variables. . T0 . T25 . S0 . S25 . All six samples 0.0002 0.0010 0.4634 0.0410 Northern samples 0.8993 0.1445 0.2414 0.3907 . T0 . T25 . S0 . S25 . All six samples 0.0002 0.0010 0.4634 0.0410 Northern samples 0.8993 0.1445 0.2414 0.3907 FST was calculated for all six samples based on four outlier loci, or the five northern samples only based on two outlier loci. T0 and T25 refer to temperature at the surface (0 meters) and 25 meters, respectively. S refers to salinity at the same depths. Table 6 Estimation of the fit between individual outlier loci and different environmental variables using all six samples. . . X1522ms . X2186_fpt . X778ms . X890_fpt . T0 p 0.008** 0.001*** 0.005** 0.011* Adjusted R2 0.822 0.931 0.855 0.793 T25 p 0.009** 0.004** 0.008** 0.013* Adjusted R2 0.812 0.869 0.823 0.779 S0 p 0.153 0.356 0.194 0.109 Adjusted R2 0.296 0.213 0.222 0.393 S25 p 0.031* 0.123 0.048* 0.017* Adjusted R2 0.660 0.487 0.580 0.743 . . X1522ms . X2186_fpt . X778ms . X890_fpt . T0 p 0.008** 0.001*** 0.005** 0.011* Adjusted R2 0.822 0.931 0.855 0.793 T25 p 0.009** 0.004** 0.008** 0.013* Adjusted R2 0.812 0.869 0.823 0.779 S0 p 0.153 0.356 0.194 0.109 Adjusted R2 0.296 0.213 0.222 0.393 S25 p 0.031* 0.123 0.048* 0.017* Adjusted R2 0.660 0.487 0.580 0.743 T0 and T25 refer to temperatures at the surface (0 meters) and 25 meters, respectively. S refers to salinity at the same depths. * p < 0.05, **p < 0.01, ***p < 0.001. Table 6 Estimation of the fit between individual outlier loci and different environmental variables using all six samples. . . X1522ms . X2186_fpt . X778ms . X890_fpt . T0 p 0.008** 0.001*** 0.005** 0.011* Adjusted R2 0.822 0.931 0.855 0.793 T25 p 0.009** 0.004** 0.008** 0.013* Adjusted R2 0.812 0.869 0.823 0.779 S0 p 0.153 0.356 0.194 0.109 Adjusted R2 0.296 0.213 0.222 0.393 S25 p 0.031* 0.123 0.048* 0.017* Adjusted R2 0.660 0.487 0.580 0.743 . . X1522ms . X2186_fpt . X778ms . X890_fpt . T0 p 0.008** 0.001*** 0.005** 0.011* Adjusted R2 0.822 0.931 0.855 0.793 T25 p 0.009** 0.004** 0.008** 0.013* Adjusted R2 0.812 0.869 0.823 0.779 S0 p 0.153 0.356 0.194 0.109 Adjusted R2 0.296 0.213 0.222 0.393 S25 p 0.031* 0.123 0.048* 0.017* Adjusted R2 0.660 0.487 0.580 0.743 T0 and T25 refer to temperatures at the surface (0 meters) and 25 meters, respectively. S refers to salinity at the same depths. * p < 0.05, **p < 0.01, ***p < 0.001. Discussion This study detected an overall population structure in European hake in the Northeast Atlantic, with Bay of Biscay hake clearly distinct from the northern samples. Within the northern samples we found a significant population structure, but using only neutral loci, a connectivity between the North Sea and Kattegat was discovered. One mechanism contributing to this pattern may be the transport of eggs and larvae from the spawning grounds in the North Sea into Kattegat, as documented for e. g. Atlantic cod (André et al., 2016). While some of the observed genetic heterogeneity was due to loci likely influenced by local selection pressures, a neutral population structure was also evident, indicating a restricted migration/gene flow between the populations. Fjord and coastal waters The low genetic differentiation between Fanafjord and North Sea samples (Table 2) implies a large degree of genetic exchange either via direct migration of juvenile or adult fish, or the dispersal of eggs and larvae via the interaction of the Norwegian coastal current and Atlantic water flowing southwards. Mixing of seasonal migrating hake and local populations was suggested by both Hickling (1930) and Hart (1948). The Fanafjord sample, a mixture of juveniles and adults, was also a mixture of a local fjord population and North Sea hake, as supported by both the pairwise FST estimates (Table 2) and the BAPS analysis (Figure 4). Hake at Trøndelag were also affected by North Sea fish, but to a lesser extent. The absence of physical barriers and short geographic distances do not necessarily reduce the potential for the establishment of genetically different populations (Hauser and Carvalho, 2008). Low, but statistically significant, genetic sub-structuring between inner-fjord and outer-fjord environments, as well as between coastal skerries has been shown for cod in the Skagerrak (Knutsen et al., 2003; Jorde et al., 2007; Knutsen et al., 2011). The topographic characteristics of a fjord, including its sill depth, climatic variables and current regimes influence the volume and layer of water exchanged with the outer water masses (Aksnes et al., 1989). These factors impact the amount of eggs and larvae transported into or out of the fjord, knowing that it is not the only prerequisite for genetic differentiation. Provided that the home range of hake is largely within fjord bounds and spawning products are retained, it is likely that genetic divergence of fjord populations can develop, as would be the case for the Trøndelag sample and perhaps to some extent for the Fanafjord sample. Both published and anecdotal information indicates that hake in the northern North Sea and in some Norwegian fjords and coastal waters, as well as off the coast of Møre and Romsdal (Norwegian Sea), spawn primarily in the period July–October (Hickling, 1927; Kjesbu et al., 2006; Groison et al., 2010; Werner et al., 2016). According to local fishermen adult hake are less available in late autumn (October–November) in coastal waters of southern Norway, implying a possible emigration of spawning fish. However, the certainty that adult fish may migrate between fjords and open ocean is obscured by observations of adult fish in some fjords in January–February (Staby unpublished data). The fjord sample included in this study (Fanafjord) shows that 81% of the individuals here are assigned back to the population of origin. A more comprehensive survey of fjords with wider temporal/seasonal coverage would be required to determine the extent of population connectivity for hake in different parts of the coastline. An emerging pattern is that both Fanafjord and Trøndelag are affected by the North Sea hake, which again is affected by the Bay of Biscay hake (Table 2 and Figure 4a). Two scenarios could cause this observed pattern. One is that present-day hake populations share a common ancestry. If the populations diverged recently and have not yet reached genetic drift-migration equilibrium, this will result in no, or shallow, divergence in the neutral regions of the genome (Roderick and Navajas, 2003). Evolution in the parts of the genome subjected to selection is a faster process (Roesti et al., 2012). Therefore, we find much larger divergences between the Bay of Biscay and Scandinavian samples when we include the outlier loci in our analysis. The second scenario includes connectivity/gene flow between the North Sea and the Bay of Biscay, and that connectivity is more limited farther north toward TRØ. The loci detected as outliers, possibly influenced by selection, affect the genetic structure first and foremost by clearly differentiated the Bay of Biscay from Scandinavian samples (Table 3a), indicating that these outliers are temperature and salinity driven (see below). Cod and herring have complex population structures, with smaller local units and a larger migratory unit that covers long distances in north-south annual migrations for spawning or feeding (Neuenfeldt et al., 2013). Hake are suspected of undertaking long migrations northward in summer from the Celtic Sea northward to the west of Scotland, entering the northern North Sea over the north of Scotland and the Shetland Islands. They continue their way southwards along the western slope of the Norwegian trench, possibly as far south as the Skagerrak (cf. Baudron and Fernandes, 2015). In addition, fish spawning in the North Sea, representing the “native” North Sea stock, may undertake shorter migrations along the Norwegian coasts. The presence of additional, local populations along the coast and in the fjords would complete the picture, with various amounts of contact between groups. What role the increased population abundance will play in replenishing, or replacing, these local populations is not known, but could be modelled (Kerr et al., 2010). Kattegat The juvenile hake sampled in the Kattegat (Kattegat; KAS) were genetically differentiated from adult fish spawning in the northern Kattegat (Kummelbank; KAN). The genetic similarity between KAS and samples from the North Sea suggests high connectivity between the two areas. Hake progeny (eggs or larvae) might drift from spawning grounds in the North Sea into the Skagerrak and Kattegat, or hake with a “North Sea” genetic signal might have spawned in the Skagerrak/Kattegat, giving rise to progeny identified genetically as “KAS”. This scenario may explain the asymmetry in gene flow as shown in the assignment test (Table 3), where a large part of the individuals from Trøndelag and the North Sea was assigned to Kattegat (KAN), whereas this is the case for only two individuals the other way around. It is possible that these individuals return to the North Sea as adults. Locally spawning hake in the Kattegat aggregate at the shallow grounds in the Kattegat, such as the Kummelbank, during the spawning period in summer from June to August (Fiskeriverket, 2011). This indicates that migrations may take place during the spawning season when hake presumably move in from the deeper parts of the Skagerrak or the Norwegian trench (Hickling, 1927). This enables a mixed spawning stock to occur, comprising local hake and migrating hake from the North Sea. The latter is “represented” as individuals from Trøndelag in the current data (Table 3). Such a mixture of local spawning stocks and incoming juveniles from offshore spawning grounds has been found for Atlantic cod (Gadus morhua) in the Skagerrak-Kattegat area (Svedäng et al., 2007; André et al., 2016). What is missing from our picture is the fate of the progeny from the Kummelbank spawning grounds. Adaptive vs. neutral population structure Determining the selective agents for adaptive divergence is generally difficult, but the effect of environmental factors provides valuable insights. Here, the regression analyses showed a significant association between the global outlier data (four loci) and average annual water temperature, both at the surface and at 25 m. Also, a weaker association was found related to salinity at 25 meters depth. These findings are in concordance with the previous study by Milano et al. (2014) who also demonstrated a correlation with salinity and temperature for outlier loci. The relationship was no longer significant after removing the Bay of Biscay sample from the analysis, indicating that temperature, and to a degree salinity, may drive the divergence between Bay of Biscay and the more northern populations. However, Milano et al. (2014) detected three clusters of hake based on outlier loci in Atlantic samples: two distinct groups represented by the northern North Sea and northern Portugal, and a third cluster composed of fish from the west of Scotland, the Celtic Sea and the coast of Galicia. Therefore, the pattern of divergence observed here may actually represent a longer north-south continuum along the Atlantic shelf. The effect of environmental variables on the genetic structure of teleost fishes has previously been reported in several studies (e.g. Nanninga et al., 2014; Sexton et al., 2014; Henriques et al., 2016). Genetic variation and climate change The North Sea has become warmer in recent decades (Perry et al., 2005) as a result of global climate change. Changes in distribution are predicted as marine fish species shift or expand into new areas with optimal conditions (Beare et al., 2004; Rijnsdorp et al., 2009; Portner and Peck, 2010). Hake is often considered a southern species, but is rarely observed in waters warmer than 11–13° C. Alternatively, if the southern populations of European hake are adapted to higher temperatures, an increase in temperature could open the northern areas for colonization as suggested by an larger proportion of Bay of Biscay individuals in the most recent sample collected in 2012 in the North Sea (Table 3). Few studies have dealt with the effect of climate change on the genetic variation of commercially exploited fish species specifically (Pauls et al., 2013; Crozier and Hutchings, 2014). But changes in the distribution of genetic variants as well as evolutionary responses in life history traits such as maturation schedules and migration timing have been documented (Crozier and Hutchings, 2014). Implications for fisheries management Hakes in the North Sea (ICES subarea IV) and Kattegat (ICES Division IIIa) are currently assessed as part of the “northern stock”, even though the management documentation states that there is no biological basis for this designation (ICES, 2015). Our results clearly show that the “northern stock” is actually subdivided into several genetically differentiated units, of which Kattegat, Norwegian Coast, North Sea could be identified in this study. However, to fully understand the interconnectivities that seem to exist among the northern populations, further studies are needed to specifically resolve this issue. For now, we suggest that this structure should be included in any predictive modelling, especially population responses to climate change. In recent years, the spawning stock biomass of the northern stock has reached record high levels (ICES, 2015), and annual TACs allocated to the North Sea and Kattegat have increased accordingly. The annually allocated TAC for these areas is based on a fixed percentage of the total “northern stock” TAC, i.e. 3% for the Kattegat and 3.5% for the North Sea. Neither the North Sea nor the Kattegat hake are assessed independently of the annual “northern stock” assessment, and the common management of the stock may obscure changes occurring on a smaller local scale, resulting in inappropriate exploitation patterns for the different components. Expanded reporting of catch data from coastal areas may allow for assessment in the North Sea and adjacent areas. Conclusions The present study provides evidence for both large scale and fine scale population structure in European hake in the Northeast Atlantic. The division between the Bay of Biscay and hake populations located in the North Sea and beyond is evident both from presumed neutral and outlier loci; the structure in the outlier loci correlates with water temperature most likely indicating adaptive differentiation. Moreover, significant population structuring was found at a regional scale in the North Sea and adjacent coastal areas. Our data have highlighted likely movement of hake between the Kattegat and North Sea, while indicating differentiation of hake found to the north of the North Sea (above 62oN). Supplementary data Supplementary material is available at the ICESJMS online version of the manuscript. References Aksnes D. L. , Aure J., Kaartvedt S., Magnesen T., Richard J. 1989 . Significance of advection for the carrying capacities of fjord populations . Marine Ecology Progress Series , 50 : 263 – 274 . Google Scholar Crossref Search ADS WorldCat André C. , Svedäng H., Knutsen H., Dahle G., Jonsson P., Ring A. K., Sköld M. et al. 2016 . Population structure in Atlantic cod in the eastern North Sea-Skagerrak-Kattegat: early life stage dispersal and adult migration . BMC Research Notes , 9 : 63. Google Scholar Crossref Search ADS PubMed WorldCat Antao T. , Lopes A., Lopes R. J., Beja-Pereira A., Luikart G. 2008 . LOSITAN: a workbench to detect molecular adaptation based on a F(st)-outlier method . BMC Bioinformatics , 9 , doi: 10.1186/1471-2105-9-323. OpenURL Placeholder Text WorldCat Bakketeig I. E. , Gjøsæther H., Hauge M., Sunnset B. H., Toft K. 2015 . Fisken og havet. særnr. 1-2015 pp. Baudron A. R. , Fernandes P. G. 2015 . Adverse consequences of stock recovery: European hake, a new “choke” species under a discard ban? Fish and Fisheries , 16 : 563 – 575 . Google Scholar Crossref Search ADS WorldCat Beare D. J. , Burns F., Greig A., Jones E. G., Peach K., Kienzle M., McKenzie E. et al. 2004 . Long-term increases in prevalence of North Sea fishes having southern biogeographic affinities . Marine Ecology Progress Series , 284 : 269 – 278 . Google Scholar Crossref Search ADS WorldCat Beaumont M. A. , Nichols R. A. 1996 . Evaluating loci for use in the genetic analysis of population structure . Proceedings of the Royal Society B-Biological Sciences , 263 : 1619 – 1626 . Google Scholar Crossref Search ADS WorldCat Benjamini Y. , Hochberg Y. 1995 . Controlling the False Discovery Rate–a practical and powerful approach to multiple testing . Journal of the Royal Statistical Society Series B-Methodological , 57 : 289 – 300 . OpenURL Placeholder Text WorldCat Berg P. , Jentoft S., Staar B., Ring K. H., Knutsen H., Lien S., Jakobsen K. et al. 2015 . Adaptation to low salinity promotes genomic divergence in Atlantic cod (Gadus morhua L.) . Genome Biology and Evolution , 7 : 1644 – 1663 . Google Scholar Crossref Search ADS PubMed WorldCat Bergstad O. A. 1991 . Distribution and trophic ecology of some gadoid fish of the Norwegian Deep. 1. Accounts of IndividualSpecies . Sarsia 75 : 269 – 313 . Google Scholar Crossref Search ADS WorldCat Boyer T. P. , Antonov J. I., Baranova O. K., Coleman C., Garcia H. E., Grodsky A., Johnson D. R. et al. 2013 . World Ocean Database 2013, NOAA Atlas NESDIS 72, S. Levitus, Ed., A. Mishonov, Technical Ed.; Silver Spring, MD, 209 pp. Brunchorst J. 1898 . Norges Fiske, deres udbredelse og levevis, John Griegs Boktrykkeri, Bergen. Casey J. , Pereiro J. 1995 . European hake (Merluccius merluccius) in the Northeast Atlantic. In Hake Fisheries, Ecology and Markets , Ed. by Alheit J., Pitcher T. J.. Chapman and Hall , London . 125 – 147 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Castillo A. G. F. , Martinez J. L., Garcia-Vazquez E. 2004 . Fine spatial structure of Atlantic hake (Merluccius merluccius) stocks revealed by variation at microsatellite loci . Marine Biotechnology , 6 : 299 – 306 . Google Scholar Crossref Search ADS PubMed WorldCat Corander J. , Marttinen P., Mantyniemi S. 2006 . A Bayesian method for identification of stock mixtures from molecular marker data . Fishery Bulletin , 104 : 550 – 558 . OpenURL Placeholder Text WorldCat Corander J. , Siren J., Arjas E. 2008 . Bayesian spatial modeling of genetic population structure . Computational Statistics , 23 : 111 – 129 . Google Scholar Crossref Search ADS WorldCat Corander J. , Waldmann P., Sillanpaa M. J. 2003 . Bayesian analysis of genetic differentiation between populations . Genetics 163 : 367 – 374 . Google Scholar PubMed OpenURL Placeholder Text WorldCat Cormon X. , Loots C., Vaz S., Vermard Y., Marchal P. 2014 . Spatial interactions between saithe (Pollachius virens) and hake (Merluccius merluccius) in the North Sea . Ices Journal of Marine Science , 71 : 1342 – 1355 . Google Scholar Crossref Search ADS WorldCat Crozier L. G. , Hutchings J. A. 2014 . Plastic and evolutionary responses to climate change in fish . Evolutionary Applications , 7 : 68 – 87 . Google Scholar Crossref Search ADS PubMed WorldCat Excoffier L. , Lischer H. E. L. 2010 . Arlequin suite ver 3.5: a new series of programs to perform population genetics analyses under Linux and Windows . Molecular Ecology Resources , 10 : 564 – 567 . Google Scholar Crossref Search ADS PubMed WorldCat Excoffier L. , Smouse P. E., Quattro J. M. 1992 . Analysis of Molecular Variance inferred from metric distances among DNA haplotypes–Application to Human Mitochondrial-DNA Restriction Data . Genetics 131 : 479 – 491 . Google Scholar PubMed OpenURL Placeholder Text WorldCat Fiskeriverket . 2011 . Fiskbestånd och miljö i hav och sötvatten–Resurs- och miljööversikt 2011. Foll M. , Gaggiotti O. 2008 . A genome-scan method to identify selected loci appropriate for both dominant and codominant markers: a Bayesian perspective . Genetics 180 : 977 – 993 . Google Scholar Crossref Search ADS PubMed WorldCat Fortibuoni T. , Libralato S., Raicevich S., Giovanardi O., Solidoro C. 2010 . Coding early naturalists' accounts into long-term fish community changes in the Adriatic Sea (1800-2000) . PLOS ONE , 5 : e15502 . Google Scholar Crossref Search ADS PubMed WorldCat Gagnaire P. A. , Broquet T., Aurelle D., Viard F., Souissi A., Bonhomme F., Arnaud-Haond S. et al. 2015 . Using neutral, selected, and hitchhiker loci to assess connectivity of marine populations in the genomic era . Evolutionary Applications , 8 : 769 – 786 . Google Scholar Crossref Search ADS PubMed WorldCat Goslee S. C. , Urban D. L. 2007 . The ecodist package for dissimilarity-based analysis of ecological data . Journal of Statistical Software , 22 : 1 – 19 . Google Scholar Crossref Search ADS WorldCat Groison A. L. , Suquet M., Cosson J., Mayer I., Severe A., Bouquet J. M., Geffen A. J. et al. 2010 . Sperm motility in European hake, Merluccius merluccius, and characterization of its spermatozoa concentration and volume, spermatocrit, osmolality and pH . Aquaculture 301 : 31 – 36 . Google Scholar Crossref Search ADS WorldCat Hart T. J. 1948 . The distribution and biology of Hake . Biological Reviews of the Cambridge Philosophical Society , 23 : 62 – 80 . Google Scholar Crossref Search ADS PubMed WorldCat Hauser L. , Carvalho G. R. 2008 . Paradigm shifts in marine fisheries genetics: ugly hypotheses slain by beautiful facts . Fish and Fisheries , 9 : 333 – 362 . Google Scholar Crossref Search ADS WorldCat Henriques R. , Von der Heyden S., Lipinski M. R., Du Toit N., Kainge P., Bloomer P., Matthee C. A. 2016 . Spatio-temporal genetic structure and the effects of longterm fishing in two partially sympatric offshore demersal fishes . Molecular Ecology , 25 : 5843 – 5861 . Google Scholar Crossref Search ADS PubMed WorldCat Hickling C. F. 1927 . The natural history of the hake. Parts I and II . Fishery Investigations Series II , 10 : 7 – 100 . OpenURL Placeholder Text WorldCat Hickling C. F. 1930 . The natural history of hake. Part III . Fishery Investigations Series II , 12 : 1 – 60 . OpenURL Placeholder Text WorldCat Hutchings J. A. 2000 . Collapse and recovery of marine fishes . Nature 406 : 882 – 885 . Google Scholar Crossref Search ADS PubMed WorldCat ICES . 2015 . Report of the Arctic Fisheries Working Group (AFWG). ICES Document ICES CM 2015/ACOM:05. 590 pp. Jorde P. E. , Knutsen H., Espeland S. H., Stenseth N. C. 2007 . Spatial scale of genetic structuring in coastal cod Gadus morhua and geographic extent of local populations . Marine Ecology Progress Series , 343 : 229 – 237 . Google Scholar Crossref Search ADS WorldCat Kell L. T. , Dickey-Collas M., Hintzen N. T., Nash R. D. M., Pilling G. M., Roel B. A. 2009 . Lumpers or splitters? Evaluating recovery and management plans for metapopulations of herring . ICES Journal of Marine Science , 66 : 1776 – 1783 . Google Scholar Crossref Search ADS WorldCat Kerr L. A. , Cadrin S. X., Secor D. H. 2010 . Simulation modelling as a tool for examining the consequences of spatial structure and connectivity on local and regional population dynamics . ICES Journal of Marine Science , 67 : 1631 – 1639 . Google Scholar Crossref Search ADS WorldCat Kjesbu O. S. , Taranger G. L., Trippel E. A. 2006 . Gadoid mariculture: Development and future challenges–Introduction . Ices Journal of Marine Science , 63 : 187 – 191 . Google Scholar Crossref Search ADS WorldCat Knutsen H. , Jorde P. E., Andre C., Stenseth N. C. 2003 . Fine-scaled geographical population structuring in a highly mobile marine species: the Atlantic cod . Molecular Ecology , 12 : 385 – 394 . Google Scholar Crossref Search ADS PubMed WorldCat Knutsen H. , Olsen E. M., Jorde P. E., Espeland S. H., Andre C., Stenseth N. C. 2011 . Are low but statistically significant levels of genetic differentiation in marine fishes ‘biologically meaningful’? A case study of coastal Atlantic cod . Molecular Ecology , 20 : 768 – 783 . Google Scholar Crossref Search ADS PubMed WorldCat Lundy C. J. , Moran P., Rico C., Milner R. S., Hewitt G. M. 1999 . Macrogeographical population differentiation in oceanic environments: a case study of European hake (Merluccius merluccius), a commercially important fish . Molecular Ecology , 8 : 1889 – 1898 . Google Scholar Crossref Search ADS PubMed WorldCat Milano I. , Babbucci M., Cariani A., Atanassova M., Bekkevold D., Carvalho G. R., Espineira M. et al. 2014 . Outlier SNP markers reveal fine-scale genetic structuring across European hake populations (Merluccius merluccius) . Molecular Ecology , 23 : 118 – 135 . Google Scholar Crossref Search ADS PubMed WorldCat Milano I. , Babbucci M., Panitz F., Ogden R., Nielsen R. O., Taylor M. I., Helyar S. J. et al. 2011 . Novel tools for conservation genomics: comparing two high-throughput approaches for SNP discovery in the transcriptome of the European hake . PLOS ONE , 6 : e28008 . Google Scholar Crossref Search ADS PubMed WorldCat Murua H. 2010 . The biology and fisheries of European hake, Merluccius Merluccius, in the North-East Atlantic . Advances in Marine Biology , 58 : 97 – 154 . Google Scholar Crossref Search ADS PubMed WorldCat Nanninga G. B. , Saenz-Agudelo P., Manica A., Berumen M. L. 2014 . Environmental gradients predict the genetic population structure of a coral reef fish in the Red Sea . Molecular Ecology , 23 : 591 – 602 . Google Scholar Crossref Search ADS PubMed WorldCat Narum S. R. , Hess J. E. 2011 . Comparison of F-ST outlier tests for SNP loci under selection . Molecular Ecology Resources , 11 : 184 – 194 . Google Scholar Crossref Search ADS PubMed WorldCat Neuenfeldt S. , Righton D., Neat F., Wright P. J., Svedäng H., Michalsen K., Subbey S. et al. 2013 . Analysing migrations of Atlantic cod Gadus morhua in the north-east Atlantic Ocean: then, now and the future . Journal of Fish Biology , 82 : 741 – 763 . Google Scholar Crossref Search ADS PubMed WorldCat Nielsen E. E. , Cariani A., Mac Aoidh E., Maes G. E., Milano I., Ogden R., Taylor M. et al. 2012 . Gene-associated markers provide tools for tackling illegal fishing and false eco-certification . Nature Communications , 3 : 851 . Google Scholar Crossref Search ADS PubMed WorldCat Norges_Fiskerier . 1935 . Norges offisielle statistikk, IX 108. 116 pp. Pauls S. U. , Nowak C., Balint M., Pfenninger M. 2013 . The impact of global climate change on genetic diversity within populations and species . Molecular Ecology , 22 : 925 – 946 . Google Scholar Crossref Search ADS PubMed WorldCat Perry A. L. , Low P. J., Ellis J. R., Reynolds J. D. 2005 . Climate change and distribution shifts in marine fishes . Science 308 : 1912 – 1915 . Google Scholar Crossref Search ADS PubMed WorldCat Pita A. , Perez M., Cervino S., Presa P. 2011 . What can gene flow and recruitment dynamics tell us about connectivity between European hake stocks in the Eastern North Atlantic? Continental Shelf Research , 31 : 376 – 387 . Google Scholar Crossref Search ADS WorldCat Portner H. O. , Peck M. A. 2010 . Climate change effects on fishes and fisheries: towards a cause-and-effect understanding . Journal of Fish Biology , 77 : 1745 – 1779 . Google Scholar Crossref Search ADS PubMed WorldCat Rijnsdorp A. D. , Peck M. A., Engelhard G. H., Mollmann C., Pinnegar J. K. 2009 . Resolving the effect of climate change on fish populations . Ices Journal of Marine Science , 66 : 1570 – 1583 . Google Scholar Crossref Search ADS WorldCat Roderick G. K. , Navajas M. 2003 . Genes in new environments: genetics and evolution in biological control . Nature Reviews Genetics , 4 : 889 – 899 . Google Scholar Crossref Search ADS PubMed WorldCat Roesti M. , Hendry A. P., Salzburger W., Berner D. 2012 . Genome divergence during evolutionary diversification as revealed in replicate lake-stream stickleback population pairs . Molecular Ecology , 21 : 2852 – 2862 . Google Scholar Crossref Search ADS PubMed WorldCat Roldan M. I. , Garcia-Marin J. L., Utter F. M., Pla C. 1998 . Population genetic structure of European hake, Merluccius merluccius . Heredity 81 : 327 – 334 . Google Scholar Crossref Search ADS WorldCat Rousset F. 2008 . GENEPOP ' 007: a complete re-implementation of the GENEPOP software for Windows and Linux . Molecular Ecology Resources , 8 : 103 – 106 . Google Scholar Crossref Search ADS PubMed WorldCat Ryman N. 2006 . CHIFISH: a computer program testing for genetic heterogeneity at multiple loci using chi-square and Fisher's exact test . Molecular Ecology Notes , 6 : 285 – 287 . Google Scholar Crossref Search ADS WorldCat Ryman N. , Jorde P. E. 2001 . Statistical power when testing for genetic differentiation . Molecular Ecology , 10 : 2361 – 2373 . Google Scholar Crossref Search ADS PubMed WorldCat Schmidt J. 1909 . The distribution of the pelagic fry and the spawning regions of the gadoids in the North Atlantic from Iceland to Spain. ICES Document 229: X. 10 pp. Sexton J. P. , Hangartner S. B., Hoffmann A. A. 2014 . Genetic isolation by environment or distance: WHich pattern of gene flow is most common? . Evolution 68 : 1 – 15 . Google Scholar Crossref Search ADS PubMed WorldCat Smitt F. A. 1892 . Skandinaviens Fiskar, P.A Norstedt & söners förlag, Stockholm. Svedäng H. 2003 . The inshore demersal fish community on the Swedish Skagerrak coast: regulation by recruitment from offshore sources . Ices Journal of Marine Science , 60 : 23 – 31 . Google Scholar Crossref Search ADS WorldCat Svedäng H. , Bardon G. 2003 . Spatial and temporal aspects of the decline in cod (Gadus morhua L.) abundance in the Kattegat and eastern Skagerrak . Ices Journal of Marine Science , 60 : 32 – 37 . Google Scholar Crossref Search ADS WorldCat Svedäng H. , Righton D., Jonsson P. 2007 . Migratory behaviour of Atlantic cod Gadus morhua: natal homing is the prime stock-separating mechanism . Marine Ecology Progress Series , 345 : 1 – 12 . Google Scholar Crossref Search ADS WorldCat Swan S. C. , Geffen A. J., Morales-Nin B., Gordon J. D. M., Shimmield T., Sawyer T., Massuti E. 2006 . Otolith chemistry: an aid to stock separation of Helicolenus dactylopterus (bluemouth) and Merluccius merluccius (European hake) in the Northeast Atlantic and Mediterranean . Ices Journal of Marine Science , 63 : 504 – 513 . Google Scholar Crossref Search ADS WorldCat Tanner S. E. , Perez M., Presa P., Thorrold S. R., Cabral H. N. 2014 . Integrating microsatellite DNA markers and otolith geochemistry to assess population structure of European hake (Merluccius merluccius) . Estuarine Coastal and Shelf Science , 142 : 68 – 75 . Google Scholar Crossref Search ADS WorldCat Weir B. S. , Cockerham C. C. 1984 . Estimating F-Statistics for the Analysis of Population-Structure . Evolution 38 : 1358 – 1370 . Google Scholar PubMed OpenURL Placeholder Text WorldCat Werner K. M. , Staby A., Geffen A. J. 2016 . Temporal and spatial patterns of reproductive indices of European hake (Merluccius merluccius) in the northern North Sea and Norwegian coastal areas . Fisheries Research , 183 : 200 – 209 . Google Scholar Crossref Search ADS WorldCat © International Council for the Exploration of the Sea 2017. All rights reserved. For Permissions, please email: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Thorson, James T.; Barnett, Lewis A. K.

doi: 10.1093/icesjms/fsw193pmid: N/A

Several approaches have been developed over the last decade to simultaneously estimate distribution or density for multiple species (e.g. “joint species distribution” or “multispecies occupancy” models). However, there has been little research comparing estimates of abundance trends or distribution shifts from these multispecies models with similar single-species estimates. We seek to determine whether a model including correlations among species (and particularly species that may affect habitat quality, termed “biogenic habitat”) improves predictive performance or decreases standard errors for estimates of total biomass and distribution shift relative to similar single-species models. To accomplish this objective, we apply a vector-autoregressive spatio-temporal (VAST) model that simultaneously estimates spatio-temporal variation in density for multiple species, and present an application of this model using data for eight US Pacific Coast rockfishes (Sebastes spp.), thornyheads (Sebastolobus spp.), and structure-forming invertebrates (SFIs). We identified three fish groups having similar spatial distribution (northern Sebastes, coastwide Sebastes, and Sebastolobus species), and estimated differences among groups in their association with SFI. The multispecies model was more parsimonious and had better predictive performance than fitting a single-species model to each taxon individually, and estimated fine-scale variation in density even for species with relatively few encounters (which the single-species model was unable to do). However, the single-species models showed similar abundance trends and distribution shifts to those of the multispecies model, with slightly smaller standard errors. Therefore, we conclude that spatial variation in density (and annual variation in these patterns) is correlated among fishes and SFI, with congeneric fishes more correlated than species from different genera. However, explicitly modelling correlations among fishes and biogenic habitat does not seem to improve precision for estimates of abundance trends or distribution shifts for these fishes. Introduction There are several benefits to simultaneously analysing the distribution and density of multiple species within a natural community. Multispecies models of spatial distribution can estimate associations among species (Latimer et al., 2009; Ovaskainen et al., 2016; Thorson et al., 2015a, 2016a), such that the presence or absence of a given species can be used as an indicator of habitat for other species when reliable habitat variables are otherwise lacking (Ovaskainen et al., 2010). Multispecies models fitted to presence/absence data (termed “multispecies occupancy models”) can also be used in some cases to identify the impact of management actions more efficiently than using single-species occupancy models (Zipkin et al., 2010). Furthermore, research shows that estimating the distribution for each species individually and then summarizing community-level properties by stacking results from single-species analyses can result in improper inference about ecological communities (Clark et al., 2014). The predictive performance of species distribution models is often improved when available covariates are included that are informative about habitat quality. Unfortunately, environmental variables associated with habitat quality are difficult to measure for many species, including demersal marine fishes. To overcome this difficulty, new species distribution modelling (SDM) techniques may allow differences in habitat to be inferred from spatial variation in the density of species with similar habitat requirements (Latimer et al., 2009; Ovaskainen et al., 2010). For example, joint species distribution models have previously been used to show strong covariation in population density among US Pacific Coast rockfishes and thornyheads (Sebastes and Sebastolobus spp.), and these correlations imply that the population density of one species is informative about the density of correlated species (Thorson et al., 2015a). Similarly, joint dynamic species distribution models (JDSDMs) can estimate abundance trends for infrequently encountered species and have revealed similarities in spatio-temporal dynamics among related butterfly species (Thorson et al., 2016a,,b). However, JDSDMs have not previously been used to explore associations between fishes and species that are associated with specific habitat features (e.g. structure-forming invertebrates, SFI). Marine fishes are intensively managed in many parts of the developed world, and the management of marine fisheries is strongly linked to estimates of population status and productivity from population models (termed “stock assessment models”) throughout North America and Europe (Methot, 2009; Maunder and Punt, 2013). Although these stock assessment models often integrate many different types of information, time-series that are proportional to population abundance (“abundance indices”) are often among the most critical (Francis, 2011). For this reason, there is considerable research regarding best practices for minimizing error when estimating abundance indices for fishes from survey data (Walters, 2003; Maunder and Punt, 2004; Shelton et al., 2014). Similarly, survey data are increasingly used to estimate shifts in fish distribution over time (e.g. due to climate change), and distribution shifts are often measured by estimating the centroid of the population's distribution and shifts in this centroid over time (Perry et al., 2005; Pinsky et al., 2013). Research suggests that spatio-temporal models are statistically efficient and can improve precision for estimates of abundance indices or distribution shifts relative to nonspatial models, given limited available data (Thorson et al., 2015b, 2016b). Recently, novel methods have been proposed for estimating abundance indices by simultaneously fitting a JDSDM to data for multiple species (Thorson et al., 2016). However, there is little research comparing the single- and multispecies approaches to estimating abundance indices for marine fishes. For three reasons, Pacific rockfishes and their close relatives provide an interesting example when studying associations between fishes and species that affect habitat suitability (“biogenic habitat”) or the potential benefit of these associations when estimating abundance indices or distribution shifts. Most importantly, Pacific rockfishes manifest an astounding diversity of species, with more than 65 species co-occurring in the Northeast Pacific (Hyde and Vetter, 2007) and exhibit a wide range of life history strategies (Love et al., 2002; Mangel et al., 2007). Given this life history diversity, rockfishes likely include species whose spatial distributions are both strongly correlated and relatively uncorrelated with SFI. Second, Pacific rockfishes differ in functional traits related to the feeding type and efficiency [eye and gill raker size, Ingram and Shurin, 2009], so species with similar spatial distribution and feeding types might exhibit correlated changes in productivity over time in response to variable food supply. Therefore, bottom-up drivers of abundance or distribution changes would result in correlated abundance or distribution changes over time for species with similar feeding types. Third, many Pacific rockfishes have low and extremely variable population densities (Thorson et al., 2011), such that single-species estimates of trends in population abundance or population distribution are frequently imprecise (Thorson et al., 2015b, 2016b). Given these characteristics of the rockfish assemblage, the inclusion of information about species associations and biogenic habitat when estimating population abundance may increase precision and thereby improve stock assessments. Given the potential benefit of estimating habitat quality from the density of co-occurring marine species when estimating abundance indices, we seek to simultaneously estimate the density of Pacific rockfishes and structure-forming invertebrates at a coastwide scale. Specifically, we seek to answer three questions: (i) do Pacific rockfishes have an association with structure-forming invertebrates on the US West Coast? (ii) Is this association similar or variable among rockfish species? and (iii) Does the inclusion of information regarding co-occurrence (either among rockfishes or between rockfish and SFI) improve predictions of local rockfish density or increase precision when estimating rockfish abundance trends or population distribution? To address these questions, we develop a vector-autoregressive spatio-temporal (VAST) model for jointly analysing catch-rate data for fish and structure-forming invertebrates and apply the model to data for eight rockfish species and SFI during 2003–2014. Methods Pacific rockfishes Pacific rockfishes (genus Sebastes) and thornyheads (genus Sebastolobus), hereafter collectively called “rockfishes”, are one of the dominant species groups within the assemblage of bottom-associated fishes off the US West Coast. Pacific rockfishes in this region are monitored by the West Coast groundfish bottom trawl survey (WCGBTS) conducted annually by the Northwest Fisheries Science Center since 2003 (Bradburn et al., 2011). The WCGBTS covers areas between the Canada and Mexico borders in 55–1280 m depth, and survey stations for each year are chosen at random within strata defined by depth and latitude (two regions divided at Point Conception, CA). Four commercial vessels (20–28 m length) are chartered each year to sample from mid-May to late October, conducting ca. 15-min tows at a speed of 2 knots using a standard Aberdeen-type trawl with a 3.8-cm mesh codend liner, 25.9-m headrope, and 31.7-m footrope. All fishes and invertebrates are sorted at sea to the lowest possible taxon, and their wet weight is measured. For the purposes of our analysis, we take the midpoint of each haul to represent the location of each biological sample. We analyse these survey data between the years 2003 and 2014, focusing on structure-forming invertebrates and eight species of Pacific rockfish (Table 1) that are frequently captured within the survey and for which there was previous documentation of association with structure-forming invertebrates at fine spatial scales (Love et al., 2002). We aggregate the structure-forming invertebrate taxa into a single grouping to obtain adequate encounter rates for estimating the distribution for structure-forming invertebrates. This SFI group primarily consists of sponges (phylum Porifera), anemones (order Actiniaria), and sea pens (order Pennatulacea), along with fewer observations of true corals (subclass Hexacorallia) and other soft corals (subclass Octocorallia). Although the survey is primarily designed to capture demersal fishes and is not as effective as visual methods for assessing structure-forming invertebrates, it is the primary source of data for estimating spatio-temporal associations between demersal fishes and biogenic habitat at large spatial and temporal scales off the US West Coast. Bottom-trawl samples have been shown to be a good predictor of biogenic habitat distribution in areas such as the eastern Bering Sea based on validation using camera surveys (Rooper et al., 2016). Table 1 List of taxa (common and scientific name), the abbreviations used to indicate taxa in plots, and the total number of encounters during 2003–2014. Common name . Scientific name . Plotting code . Encounters . Structure-forming invertebrates (SFIs) — SFI 6383 Longspine thornyhead Sebastolobus altivelis L. spine 2758 Shortspine thornyhead Sebastolobus alascanus S. spine 3891 Darkblotched rockfish Sebastes crameri Dark 1338 Pacific Ocean perch Sebastes alutus POP 547 Sharpchin rockfish Sebastes zacentrus Sharp 490 Splitnose rockfish Sebastes diploproa Split 1619 Stripetail rockfish Sebastes saxicola Stripe 1630 Greenspotted rockfish Sebastes chlorostictus Green 434 Common name . Scientific name . Plotting code . Encounters . Structure-forming invertebrates (SFIs) — SFI 6383 Longspine thornyhead Sebastolobus altivelis L. spine 2758 Shortspine thornyhead Sebastolobus alascanus S. spine 3891 Darkblotched rockfish Sebastes crameri Dark 1338 Pacific Ocean perch Sebastes alutus POP 547 Sharpchin rockfish Sebastes zacentrus Sharp 490 Splitnose rockfish Sebastes diploproa Split 1619 Stripetail rockfish Sebastes saxicola Stripe 1630 Greenspotted rockfish Sebastes chlorostictus Green 434 Open in new tab Table 1 List of taxa (common and scientific name), the abbreviations used to indicate taxa in plots, and the total number of encounters during 2003–2014. Common name . Scientific name . Plotting code . Encounters . Structure-forming invertebrates (SFIs) — SFI 6383 Longspine thornyhead Sebastolobus altivelis L. spine 2758 Shortspine thornyhead Sebastolobus alascanus S. spine 3891 Darkblotched rockfish Sebastes crameri Dark 1338 Pacific Ocean perch Sebastes alutus POP 547 Sharpchin rockfish Sebastes zacentrus Sharp 490 Splitnose rockfish Sebastes diploproa Split 1619 Stripetail rockfish Sebastes saxicola Stripe 1630 Greenspotted rockfish Sebastes chlorostictus Green 434 Common name . Scientific name . Plotting code . Encounters . Structure-forming invertebrates (SFIs) — SFI 6383 Longspine thornyhead Sebastolobus altivelis L. spine 2758 Shortspine thornyhead Sebastolobus alascanus S. spine 3891 Darkblotched rockfish Sebastes crameri Dark 1338 Pacific Ocean perch Sebastes alutus POP 547 Sharpchin rockfish Sebastes zacentrus Sharp 490 Splitnose rockfish Sebastes diploproa Split 1619 Stripetail rockfish Sebastes saxicola Stripe 1630 Greenspotted rockfish Sebastes chlorostictus Green 434 Open in new tab Vector-autoregressive spatio-temporal (VAST) model We seek to estimate the association among fishes and structure-forming invertebrates and, therefore, model correlations among density d(s,c,t) for each taxon c (indicating fish species or the SFI group) at location s and time t (all symbols are summarized in Table 2). To do so, we build upon recent research regarding JDSDMs. In particular, we propose a VAST model, where the probability distribution for catch data bi is decomposed into two components representing (i) the probability of encounter psi,ci,ti for the location si ⁠, taxon ci ⁠, and year ti of the ith sample, and (ii) the expected catch rate rsi,ci,ti ⁠, given that taxon ci is encountered. Decomposing catch rates into encounter-probability p and positive catch rates r is commonly conducted using delta models (Maunder and Punt, 2004; Martin et al., 2005), although delta models have not previously been used within JDSDMs. Using a delta model allows us to separately identify species with similar distribution (similarities in occupied habitat) vs. similar density (similarities in hotspots within their distribution). Therefore, we specify: Pr⁡(bi=B)={1−p(si,ci,ti) if B=0p(si,ci,ti)×Lognormal{B|log[wi×r(si,ci,ti)],σc2} if B>0(1) where Lognormal(x|μ,σ2) is a lognormal probability distribution function for value x ⁠, given a log-mean of μ and a variance of σ2 ⁠, and wi is the area swept for the ith sample. Table 2 List of symbols representing indices, data, fixed effects, random effects, and derived quantities defined in the main text. Name . Symbol . Dimension . Type . Sample i — Index Taxon c — Index Location s — Index Year t — Index Taxon (species or species-group) c — Index Vessel v — Index Probability distribution for catches B — Random variable Catch data bi ni Data Area-swept for each sample wi ni Data Area associated with each location a(s) ns Data Statistic associated with each location x(s) ns Data Variance in positive catch rates σ2 — Fixed effect Intercept for p γpc,t nc×nt Fixed effect Intercept for r γrc,t nc×nt Fixed effect Decorrelation distance in ɛps,c,t κp — Fixed effect Decorrelation distance in ɛrs,c,t κr — Fixed effect Geometric anisotropy H 2×2 Fixed effect Factor approximation to Vɛp Lɛp nc×nf Fixed effect Factor approximation to Vɛr Lɛr nc×nf Fixed effect Factor approximation to Vδp Lδp nc×nf Fixed effect Factor approximation to Vδr Lδr nc×nf Fixed effect Spatio-temporal variation in p ɛps,c,t ns×nc×nt Random effect Spatio-temporal variation in r ɛrs,c,t ns×nc×nt Random effect Vessel effect for p δp(c,v) nc×nv Random effect Vessel effect for r δr(c,v) nc×nv Random effect Encounter probability p(s,c,t) ns×nc×nt Derived quantity Positive catch rates r(s,c,t) ns×nc×nt Derived quantity Local density d(s,c,t) ns×nc×nt Derived quantity Spatio-temporal variation in p in year t Ep(t) ns×nc Derived quantity Spatio-temporal variation in r in year t Er(t) ns×nc Derived quantity Spatial correlation in ɛps,c,t Rp ns×ns Derived quantity Spatial correlation in ɛrs,c,t Rr ns×ns Derived quantity Correlation among species in ɛps,c,t Vɛp nc×nc Derived quantity Correlation among species in ɛrs,c,t Vɛr nc×nc Derived quantity Correlation among species in δp(c,v) Vδp nc×nc Derived quantity Correlation among species in δr(c,v) Vδr nc×nc Derived quantity Index of abundance I(c,t) nc×nt Derived quantity Center of spatial distribution X(c,t) nc×nt Derived quantity Name . Symbol . Dimension . Type . Sample i — Index Taxon c — Index Location s — Index Year t — Index Taxon (species or species-group) c — Index Vessel v — Index Probability distribution for catches B — Random variable Catch data bi ni Data Area-swept for each sample wi ni Data Area associated with each location a(s) ns Data Statistic associated with each location x(s) ns Data Variance in positive catch rates σ2 — Fixed effect Intercept for p γpc,t nc×nt Fixed effect Intercept for r γrc,t nc×nt Fixed effect Decorrelation distance in ɛps,c,t κp — Fixed effect Decorrelation distance in ɛrs,c,t κr — Fixed effect Geometric anisotropy H 2×2 Fixed effect Factor approximation to Vɛp Lɛp nc×nf Fixed effect Factor approximation to Vɛr Lɛr nc×nf Fixed effect Factor approximation to Vδp Lδp nc×nf Fixed effect Factor approximation to Vδr Lδr nc×nf Fixed effect Spatio-temporal variation in p ɛps,c,t ns×nc×nt Random effect Spatio-temporal variation in r ɛrs,c,t ns×nc×nt Random effect Vessel effect for p δp(c,v) nc×nv Random effect Vessel effect for r δr(c,v) nc×nv Random effect Encounter probability p(s,c,t) ns×nc×nt Derived quantity Positive catch rates r(s,c,t) ns×nc×nt Derived quantity Local density d(s,c,t) ns×nc×nt Derived quantity Spatio-temporal variation in p in year t Ep(t) ns×nc Derived quantity Spatio-temporal variation in r in year t Er(t) ns×nc Derived quantity Spatial correlation in ɛps,c,t Rp ns×ns Derived quantity Spatial correlation in ɛrs,c,t Rr ns×ns Derived quantity Correlation among species in ɛps,c,t Vɛp nc×nc Derived quantity Correlation among species in ɛrs,c,t Vɛr nc×nc Derived quantity Correlation among species in δp(c,v) Vδp nc×nc Derived quantity Correlation among species in δr(c,v) Vδr nc×nc Derived quantity Index of abundance I(c,t) nc×nt Derived quantity Center of spatial distribution X(c,t) nc×nt Derived quantity Throughout, we use subscripts to indicate either properties of the ith sample (e.g. area-swept wi ⁠, location si ⁠), or parameters associated with encounter probability or positive catch rates (e.g. κp vs. κr ⁠). Indices or scalars have dimension of zero, indicated using a “—” symbol. Open in new tab Table 2 List of symbols representing indices, data, fixed effects, random effects, and derived quantities defined in the main text. Name . Symbol . Dimension . Type . Sample i — Index Taxon c — Index Location s — Index Year t — Index Taxon (species or species-group) c — Index Vessel v — Index Probability distribution for catches B — Random variable Catch data bi ni Data Area-swept for each sample wi ni Data Area associated with each location a(s) ns Data Statistic associated with each location x(s) ns Data Variance in positive catch rates σ2 — Fixed effect Intercept for p γpc,t nc×nt Fixed effect Intercept for r γrc,t nc×nt Fixed effect Decorrelation distance in ɛps,c,t κp — Fixed effect Decorrelation distance in ɛrs,c,t κr — Fixed effect Geometric anisotropy H 2×2 Fixed effect Factor approximation to Vɛp Lɛp nc×nf Fixed effect Factor approximation to Vɛr Lɛr nc×nf Fixed effect Factor approximation to Vδp Lδp nc×nf Fixed effect Factor approximation to Vδr Lδr nc×nf Fixed effect Spatio-temporal variation in p ɛps,c,t ns×nc×nt Random effect Spatio-temporal variation in r ɛrs,c,t ns×nc×nt Random effect Vessel effect for p δp(c,v) nc×nv Random effect Vessel effect for r δr(c,v) nc×nv Random effect Encounter probability p(s,c,t) ns×nc×nt Derived quantity Positive catch rates r(s,c,t) ns×nc×nt Derived quantity Local density d(s,c,t) ns×nc×nt Derived quantity Spatio-temporal variation in p in year t Ep(t) ns×nc Derived quantity Spatio-temporal variation in r in year t Er(t) ns×nc Derived quantity Spatial correlation in ɛps,c,t Rp ns×ns Derived quantity Spatial correlation in ɛrs,c,t Rr ns×ns Derived quantity Correlation among species in ɛps,c,t Vɛp nc×nc Derived quantity Correlation among species in ɛrs,c,t Vɛr nc×nc Derived quantity Correlation among species in δp(c,v) Vδp nc×nc Derived quantity Correlation among species in δr(c,v) Vδr nc×nc Derived quantity Index of abundance I(c,t) nc×nt Derived quantity Center of spatial distribution X(c,t) nc×nt Derived quantity Name . Symbol . Dimension . Type . Sample i — Index Taxon c — Index Location s — Index Year t — Index Taxon (species or species-group) c — Index Vessel v — Index Probability distribution for catches B — Random variable Catch data bi ni Data Area-swept for each sample wi ni Data Area associated with each location a(s) ns Data Statistic associated with each location x(s) ns Data Variance in positive catch rates σ2 — Fixed effect Intercept for p γpc,t nc×nt Fixed effect Intercept for r γrc,t nc×nt Fixed effect Decorrelation distance in ɛps,c,t κp — Fixed effect Decorrelation distance in ɛrs,c,t κr — Fixed effect Geometric anisotropy H 2×2 Fixed effect Factor approximation to Vɛp Lɛp nc×nf Fixed effect Factor approximation to Vɛr Lɛr nc×nf Fixed effect Factor approximation to Vδp Lδp nc×nf Fixed effect Factor approximation to Vδr Lδr nc×nf Fixed effect Spatio-temporal variation in p ɛps,c,t ns×nc×nt Random effect Spatio-temporal variation in r ɛrs,c,t ns×nc×nt Random effect Vessel effect for p δp(c,v) nc×nv Random effect Vessel effect for r δr(c,v) nc×nv Random effect Encounter probability p(s,c,t) ns×nc×nt Derived quantity Positive catch rates r(s,c,t) ns×nc×nt Derived quantity Local density d(s,c,t) ns×nc×nt Derived quantity Spatio-temporal variation in p in year t Ep(t) ns×nc Derived quantity Spatio-temporal variation in r in year t Er(t) ns×nc Derived quantity Spatial correlation in ɛps,c,t Rp ns×ns Derived quantity Spatial correlation in ɛrs,c,t Rr ns×ns Derived quantity Correlation among species in ɛps,c,t Vɛp nc×nc Derived quantity Correlation among species in ɛrs,c,t Vɛr nc×nc Derived quantity Correlation among species in δp(c,v) Vδp nc×nc Derived quantity Correlation among species in δr(c,v) Vδr nc×nc Derived quantity Index of abundance I(c,t) nc×nt Derived quantity Center of spatial distribution X(c,t) nc×nt Derived quantity Throughout, we use subscripts to indicate either properties of the ith sample (e.g. area-swept wi ⁠, location si ⁠), or parameters associated with encounter probability or positive catch rates (e.g. κp vs. κr ⁠). Indices or scalars have dimension of zero, indicated using a “—” symbol. Open in new tab Using this delta model, we separately develop a spatio-temporal model for encounter probabilities p and positive catch rates r ⁠. We approximate spatio-temporal variation in encounter probability psi,ci,ti using a logit-linked linear predictor : logit⁡[psi,ci,ti]=γpci,ti+ɛpsi,ci,ti+δp(ci,vi)(2) where γpci,ti is an intercept for encounter probability for each taxon c and time t ⁠, ɛpsi,ci,ti approximates spatio-temporal variation in encounter probability (in logit-space), and δp(ci,vi) is a “vessel effect” for the vessel vi conducting the ith sample when catching taxon ci ⁠. Vessel effects are included because the WCGBTS is obtained using 3–4 different vessels per year, and previous research indicates that vessels in each year have small, but important, variation in fishing behaviour and resulting catch rates (Helser et al., 2004; Thorson and Ward, 2014). Expected catch rates when a species is encountered rsi,ci,ti are similarly approximated using a log-linked linear predictor: log⁡[rsi,ci,ti]=γrci,ti+ɛrsi,ci,ti+δr(ci,vi)(3) where γrti,ci is an intercept, ɛrsi,ci,ti is the spatio-temporal variation, and δr(ci,vi) is a vessel effect for expected catch rates. ɛrsi,ci,ti ⁠, ɛpsi,ci,ti ⁠, δp(ci,vi) ⁠, and δr(ci,vi) approximate processes that affect density and catchability, respectively, but are not otherwise modelled explicitly (Thorson et al., 2016). Including these effects in the VAST model allows catch data bi for nearby locations to be correlated (via correlations in encounter probability p or positive catch rate r ⁠) and also improves density predictions at locations that otherwise have little data (Shelton et al., 2014). The VAST model involves specifying a probability distribution for spatio-temporal variation (⁠ ɛps,c,t and ɛr(s,c,t) ⁠) and vessel effects (⁠ δp(c,v) and δr(c,v) ⁠). For each modelled year, we therefore specify a three-dimensional Gaussian process for spatio-temporal variation: vec[Ep(t)]∼MVN(0,Rp⊗Vɛp)(4) where Ep(t) is a matrix composed of ɛps,c,t at every modelled location s and taxon c in a given year t ⁠, vec[Ept] is a vector composed of stacking every column of Ep(t) ⁠, Rp is the correlation matrix approximating similar encounter probability among locations, Vɛp is the covariance matrix approximating similar encounter probability among species, and ⊗ is the Kronecker product such that Rp⊗Vɛp is a covariance matrix between any taxon c and location s in year t [ Er(t) follows an identical distribution, but with Rr and Vɛr in place of Rp and Vɛp]. Spatial correlation Rp between location s and location s+h generally declines with an increased distance |h| between the two locations [sometimes termed Tobler's law of geography; Tobler, 1970]. We specify a Matérn function for this correlation, which includes a parameter κp governing the distance at which locations are essentially uncorrelated (increased κp leads to a decreased decorrelation distance) as well as a transformation matrix H representing geometric anisotropy (the tendency for correlations to decline faster in one direction than another): Rp(s,s+h)=12ν−1Γ(n)×(κp|hH|)n×Kν(κn|hH|)(5) where n is a smoothness parameter [fixed at 1.0; Simpson et al., 2012] and Kn is the Bessel function (⁠ Rr is defined identically but with κr in place of κp ⁠). Including geometric anisotropy is generally important for fishes along a narrow continental shelf like the US West Coast, where correlations decline faster moving onshore–offshore rather than moving alongshore (Thorson et al., 2015b). We do not know a priori which taxa are likely to be more or less correlated. Therefore, we model covariance Vɛp among species using a factor-analysis decomposition: Vɛp=LɛpLɛpT(6) where Lɛp is a nc by nf matrix defining the first nf columns of the Cholesky decomposition of covariance matrix Vɛp ⁠, LɛpT is the matrix-transpose of Lɛp ⁠, and Vɛr is defined identically but with Lɛr in place of Lɛp [see Thorson et al., 2016a,b and Warton et al., 2015 for details regarding this factor-analysis decomposition]. Similarly, we specify a factor-analysis decomposition for the covariance Vδp=LδpLδpT among vessel effects: δp(v)∼MVN(0,Vδp)(7) where δp(v) is the vector of vessel effects affecting encounter probability δpc,v for all taxa c and a given vessel v (and where Vδr is defined identically, but with Lδr in place of Lδp ⁠). This factor-analysis decomposition allows the analyst to select an appropriate number of factors nf for approximating spatio-temporal covariation or covariation among vessels, where 0<nf≤nc ⁠. Specifying a reduced number of factors (⁠ nf<nc ⁠) decreases the number of estimated parameters and, therefore, may result in smaller standard errors for other parameters or more precise predictions of local density (Thorson et al., 2015a). However, reducing the number of factors could also result in biased estimates of abundance trends (e.g. by shrinking dynamics for all species onto a small number of dimensions). We leave exploration of this bias-variance trade-off as a topic for future research and instead specify full rank for each covariance (⁠ nf=nc ⁠) to eliminate this source of potential bias. Parameters are estimated for the VAST model by maximizing the marginal likelihood of fixed effects given available data. We treat the intercept parameters for each species (⁠ γp(c,t) and γrc,t ⁠), the spatial scale of spatio-temporal variation (⁠ κλ and κr ⁠), the shape of geometric anisotropy (two parameters in H ⁠), the covariation among species (⁠ Lɛp and Lɛr ⁠), the covariation among vessels (⁠ Lδp and Lδr ⁠), and the magnitude of residual variation in positive catch rates for each species (⁠ σc2 ⁠) as fixed effects. We treat spatio-temporal variation [ ɛps,c,t and ɛrs,c,t] and catchability variation [ δp(c,v) and δr(c,v)]as random effects (Thorson and Minto, 2015). We define the joint likelihood as the product of the probability of random effects (given fixed effects) and the probability of the data (given random and fixed effects). We then calculate the marginal likelihood of fixed effects while integrating the joint likelihood with respect to random effects. We specifically use the Laplace approximation to approximate the multidimensional integral required to calculate the marginal likelihood (Skaug and Fournier, 2006). The Laplace approximation is implemented using Template Model Builder (Kristensen et al., 2016), and Template Model Builder also provides the gradient of the approximated marginal likelihood with respect to all fixed effects. We use a gradient-based nonlinear minimizer within the R statistical environment (R Core Team, 2015) to identify maximum-likelihood estimate (MLE) of fixed effects. To improve computational efficiency, we use Revolution Open R for low-level parallelization of matrix computations (http://www.revolutionanalytics.com/revolution-r-open), a stochastic partial differential equation (SPDE) approximation for all spatial processes (Lindgren et al., 2011), and the R-INLA software (Lindgren, 2012) to compute the triangulated mesh used in the SPDE approximation. We distribute code for applying the VAST model to other datasets as an R package on the author's website (www.github.com/james-thorson/VAST) and have confirmed that the VAST model provides identical parameter estimates to a previous spatio-temporal index standardization model (SpatialDeltaGLMM, Thorson et al., 2015b) when applied to data for a single species. However, the VAST model also incorporates most capabilities of spatial dynamic factor analysis for monitoring trends in community abundance or conducting species ordination (Thorson et al., 2016a,b). After parameters are estimated, we predict the value of random effects by identifying their values that maximize the joint likelihood, given the data and maximum likelihood estimates of fixed effects. We then use the predicted values for random effects to estimate total biomass I(c,t) for each taxon in each year [an “index of abundance”; Thorson et al., 2015b]: I(c,t)=∑s=1nsa(s)× logit ⁡−1[γp(c,t)+ɛp(s,c,t)]× exp ⁡[γr(c,t)+ɛr(s,c,t)](8) where a(s) is the area associated with location s [ I(c,t) does not include vessel effects δp or δr ⁠, because these are interpreted as representing variation in catchability]. We also estimate the centroid of the distribution X(c,t) for each species in each year [termed “center of gravity”; Thorson et al., 2016b]: X(c,t)=∑s=1nsx(s){a(s) logit ⁡−1[γp(c,t)+ɛp(s,c,t)]× exp ⁡[γr(c,t)+ɛr(s,c,t)]I(c,t)}(9) where x(s) can be any statistic used to summarize distribution. We are particularly interested in estimating shifts in fish distribution northward or southward along the US West Coast, so we define x(s) as the distance north of the equator (in kilometres) for location s ⁠. Standard errors for I(c,t) and X(c,t) are then calculated by Template Model Builder using a generalization of the delta method. We compare model performance when fitting all species simultaneously (the “multi-species analysis”) to a conventional “single-species analysis”, where each species is fitted individually using the VAST model. To compare performance between single- and multispecies models, we compute the Akaike information criterion (AIC) (Akaike, 1974). The AIC is a measure of model “parsimony” (see Figure 1.3 from Burnham and Anderson, 2002), which we use to identify the level of complexity that likely minimizes the combination of bias (from an overly simple model) and imprecision (from an overly complex model). We compute the “single-species” AIC as the sum of the AIC for the VAST model fitted to each individual species; this comparison is justified because the aggregate of single-species models is fitted to the same dataset as the multispecies model. We also conduct a tenfold cross-validation analysis to determine whether multi- or single-species analyses have greater predictive ability. To do so, we divide the data into 10 similarly sized partitions. For the first cross-validation, we estimate model parameters only using data in partitions 2–10, and calculate the probability of data in partition 1 using the predictive distribution, given estimated parameters. This process is repeated for all 10 partitions for the multispecies model. For the single-species model, we conduct this tenfold cross-validation for each species individually, and then sum the resulting log-predictive probabilities for each species. Results Inspection of density estimates for eight fishes and structure-forming invertebrates using the multispecies VAST model shows that species are unevenly distributed throughout the California Current (Figure 1, bottom row). By distribution, the fishes can be broadly classified into three groups: coastwide Sebastes spp. (splitnose, stripetail, and greenspotted), northern Sebastes spp. (POP, sharpchin, and darkblotched), and Sebastolobus spp. (longspine and shortspine thornyheads). The thornyheads are distinguished by having increased densities in the deepest waters farthest from the US coast. Structure-forming invertebrates are found at highest densities offshore near northern Oregon, close to the Oregon–California border and offshore from the south of Monterey Bay through the Southern California Bight. Figure 1 Open in new tabDownload slide Comparison of estimated density functions (averaged across all years for each taxon) for structure-forming invertebrates, six Sebastes, and two Sebastolobus species using the VAST multispecies model (bottom row) compared with a single-species estimate for each taxon (top row; see Table 1 for plotting code for taxa; inset colour bar shows average log-density in kg km-2 ⁠, and colours are defined identically for single- and multispecies models for each taxon). Figure 1 Open in new tabDownload slide Comparison of estimated density functions (averaged across all years for each taxon) for structure-forming invertebrates, six Sebastes, and two Sebastolobus species using the VAST multispecies model (bottom row) compared with a single-species estimate for each taxon (top row; see Table 1 for plotting code for taxa; inset colour bar shows average log-density in kg km-2 ⁠, and colours are defined identically for single- and multispecies models for each taxon). Our species classifications are supported by the estimated covariance matrices (Figure 2), where longspine and shortspine thornyheads have high pairwise correlations in both encounter probability and positive catch rates (0.5–0.7). Encounter probability of longspine is negatively associated with encounter probability of Sebastes spp., while encounter probability of shortspine has both positive and negative associations with different Sebastes spp. This difference between thornyhead species is also apparent in distribution maps (Figure 1), where longspine has the deepest distribution of any fish in our analysis, whereas shortspine occupies a more shoreward distribution that overlaps with the spatial distribution for several Sebastes (e.g. darkblotched, POP, and splitnose). The Sebastes spp. all generally have high correlations (0.5–0.9) with one another for encounter probability (⁠ Ep ⁠; Figure 2 top panel), but the northern vs. coastwide groups are strongly distinguished by correlations in positive catch rates (⁠ Er ⁠; Figure 2 bottom panel), where darkblotched, POP, and sharpchin have higher correlations with one another (0.9) than with splitnose, stripetail, and greenspotted (0.2–0.7). At this coastwide spatial scale, Sebastolobus spp. generally have increased encounter probability when structure-forming invertebrates are found, whereas coastwide and northern rockfish groups have somewhat decreased encounter probability in these cases. When fishes and SFI are encountered, however, an increased catch of SFI is associated with increased catch for all fishes except stripetail and sharpchin rockfish. Figure 2 Open in new tabDownload slide Analytic estimate of correlation among fishes and SFI using the VAST model (see Table 1 for taxa abbreviations) for encounter probabilities (⁠ Rεr ⁠, top panel) or positive catch rates (⁠ Rεp ⁠, bottom panel). Numbered columns correspond to the species groups indicated by the row labels, ordered from top to bottom. Figure 2 Open in new tabDownload slide Analytic estimate of correlation among fishes and SFI using the VAST model (see Table 1 for taxa abbreviations) for encounter probabilities (⁠ Rεr ⁠, top panel) or positive catch rates (⁠ Rεp ⁠, bottom panel). Numbered columns correspond to the species groups indicated by the row labels, ordered from top to bottom. We next compare estimates of biomass trends using multispecies and single-species estimates (Figure 3). Biomass trends are broadly similar between models, and particularly for SFI, which shows a trend of increased biomass since 2008. Biomass trend estimates are most different between multi- and single-species models for the group of northern Sebastes spp. (darkblotched, sharpchin, and POP; Figure 3 middle row). For example, the multispecies model estimates lower abundance for POP in 2008 than the single-species model. This lower estimate for POP in 2008 using the multispecies model reflects a similar decrease in abundance for darkblotched rockfish in 2008 using either model – the estimate for POP in this year for the multispecies model is “shrunk” towards the estimate for darkblotched rockfish. Figure 3 Open in new tabDownload slide Relative log-biomass I(c,t) in units of log-metric tons (Equation 8) for each species using single-species (grey) or multispecies (red) modelling (note different y-axis ranges for each species and see Table 1 for taxa abbreviations; top row: SFI and Sebastolobus group; middle row: northern Sebastes group; bottom row: coastwide Sebastes group). Figure 3 Open in new tabDownload slide Relative log-biomass I(c,t) in units of log-metric tons (Equation 8) for each species using single-species (grey) or multispecies (red) modelling (note different y-axis ranges for each species and see Table 1 for taxa abbreviations; top row: SFI and Sebastolobus group; middle row: northern Sebastes group; bottom row: coastwide Sebastes group). Estimates of variation and trends in center-of-gravity (COG) are also generally similar between multi- and single-species model outputs (Figure 4). The notable exceptions are again the northern Sebastes spp., specifically POP and sharpchin rockfish, which both have relatively few encounters relative to other species (ca. 500 each, see Table 1). For POP and sharpchin, the single-species estimates of COG are nearly 100 km farther south than COG estimates from the multispecies model (Figure 4 middle row). By sharing information about positive catch rates (as shown in the lower panel of Figure 2), the multispecies model estimates greater variation in density for these species between different locations off Oregon and Washington and, therefore, estimates a more northward distribution than the single-species model for POP and sharpchin (Figure 1 top row). We again interpret this as a consequence of statistical “shrinkage” for these species, where the multispecies model is sharing information among northern Sebastes spp. to infer density hotspots. Figure 4 Open in new tabDownload slide Northward center-of-gravity (COG), X(c,t) ⁠, in units of kilometres north of the equator (Equation 9) for each species using single-species (grey) or multispecies (red) modelling (see Table 1 for taxa abbreviations, and Figure 3 caption for details). Figure 4 Open in new tabDownload slide Northward center-of-gravity (COG), X(c,t) ⁠, in units of kilometres north of the equator (Equation 9) for each species using single-species (grey) or multispecies (red) modelling (see Table 1 for taxa abbreviations, and Figure 3 caption for details). Finally, a comparison of standard errors (Figure 5) shows that the multispecies model generally has slightly larger standard error for estimating log-biomass (median 0.01 increase relative to single-species model) and center-of-gravity (median 2.4 km increase). This increased standard error presumably occurs because the multispecies model estimates greater spatial variation in density (Figure 5). For POP, for example, the single-species model estimates little spatial pattern except an increase in density moving northward along the coast, while the multispecies model estimates density hotspots in the same mid-depth areas off the Washington coast as it estimates as good habitat for splitnose and darkblotched rockfishes (Figure 1). Figure 5 Open in new tabDownload slide Comparison of standard error estimates for log-index of abundance, ln⁡[Ic,t] in units of log-metric tonnes (Equation 8, top row) and centre-of-gravity (COG) X(c,t) in units of kilometres north of the equator (Equation 9, bottom row) from single-species and multispecies VAST models, where the scatterplots compare standard error estimates for each year t and taxon c (first column; x-axis shows single-species and y-axis shows multispecies standard errors) and the histograms show the difference between these standard error estimates (second column; where a positive value indicates that the multispecies model had a wider confidence interval than the single-species model for that taxon and year). The dotted line in each histogram indicates the median increase in log-standard error (top row) or standard error (bottom row) for the multispecies relative to the single-species model, and the number in the top-right corner indicates the median increase. Figure 5 Open in new tabDownload slide Comparison of standard error estimates for log-index of abundance, ln⁡[Ic,t] in units of log-metric tonnes (Equation 8, top row) and centre-of-gravity (COG) X(c,t) in units of kilometres north of the equator (Equation 9, bottom row) from single-species and multispecies VAST models, where the scatterplots compare standard error estimates for each year t and taxon c (first column; x-axis shows single-species and y-axis shows multispecies standard errors) and the histograms show the difference between these standard error estimates (second column; where a positive value indicates that the multispecies model had a wider confidence interval than the single-species model for that taxon and year). The dotted line in each histogram indicates the median increase in log-standard error (top row) or standard error (bottom row) for the multispecies relative to the single-species model, and the number in the top-right corner indicates the median increase. Despite estimating wider standard errors for abundance indices and distribution shifts, the multispecies model provides a more fit to available data. The multispecies model has an AIC score that is 5692.0 better than the combined AIC for single-species models, despite the multispecies model estimating an additional 112 parameters (409 fixed effects for the multispecies vs. 297 total among all single-species models). The improvement in fit for the multispecies model is also supported by the tenfold cross-validation analysis, where the multispecies model has a 4–5% greater predictive probability than when analysing each species individually (Table 3). This improvement in predictive score presumably arises because the multispecies model identifies fine-scale differences in species density for all taxa (e.g. comparing Fig. 1 top and bottom rows for splitnose), and these fine-scale density estimates are on average a useful prediction of variation in catch rates. Table 3 Predictive negative log-likelihood for left-out samples (where a low number indicates better fit for the multispecies model) from a tenfold cross-validation experiment comparing single-species models to a multispecies VAST model that was estimated for all species simultaneously, as well as the ratio of predictive probability for the multispecies model relative to the single-species model (a value >1.0 indicates better predictive performance for the multispecies model than the single-species model). Partition . Number of cross-validation samples . Predictive negative log-likelihood . Ratio of predictive probability . Single-species VAST model . Multispecies VAST model . 1 6889 8459.71 8078.84 1.057 2 6832 8560.26 8209.90 1.053 3 6835 8768.53 8377.45 1.059 4 6890 8203.66 7815.19 1.058 5 6799 8335.34 8009.96 1.049 6 6828 8548.48 8154.91 1.059 7 6800 8426.61 8133.02 1.044 8 6997 8688.89 8335.62 1.052 9 6743 8439.27 8077.01 1.055 10 6859 8594.30 8271.42 1.048 Partition . Number of cross-validation samples . Predictive negative log-likelihood . Ratio of predictive probability . Single-species VAST model . Multispecies VAST model . 1 6889 8459.71 8078.84 1.057 2 6832 8560.26 8209.90 1.053 3 6835 8768.53 8377.45 1.059 4 6890 8203.66 7815.19 1.058 5 6799 8335.34 8009.96 1.049 6 6828 8548.48 8154.91 1.059 7 6800 8426.61 8133.02 1.044 8 6997 8688.89 8335.62 1.052 9 6743 8439.27 8077.01 1.055 10 6859 8594.30 8271.42 1.048 Open in new tab Table 3 Predictive negative log-likelihood for left-out samples (where a low number indicates better fit for the multispecies model) from a tenfold cross-validation experiment comparing single-species models to a multispecies VAST model that was estimated for all species simultaneously, as well as the ratio of predictive probability for the multispecies model relative to the single-species model (a value >1.0 indicates better predictive performance for the multispecies model than the single-species model). Partition . Number of cross-validation samples . Predictive negative log-likelihood . Ratio of predictive probability . Single-species VAST model . Multispecies VAST model . 1 6889 8459.71 8078.84 1.057 2 6832 8560.26 8209.90 1.053 3 6835 8768.53 8377.45 1.059 4 6890 8203.66 7815.19 1.058 5 6799 8335.34 8009.96 1.049 6 6828 8548.48 8154.91 1.059 7 6800 8426.61 8133.02 1.044 8 6997 8688.89 8335.62 1.052 9 6743 8439.27 8077.01 1.055 10 6859 8594.30 8271.42 1.048 Partition . Number of cross-validation samples . Predictive negative log-likelihood . Ratio of predictive probability . Single-species VAST model . Multispecies VAST model . 1 6889 8459.71 8078.84 1.057 2 6832 8560.26 8209.90 1.053 3 6835 8768.53 8377.45 1.059 4 6890 8203.66 7815.19 1.058 5 6799 8335.34 8009.96 1.049 6 6828 8548.48 8154.91 1.059 7 6800 8426.61 8133.02 1.044 8 6997 8688.89 8335.62 1.052 9 6743 8439.27 8077.01 1.055 10 6859 8594.30 8271.42 1.048 Open in new tab Discussion We have used a JDSDM to illustrate strong associations (both positive and negative) between deep-water demersal fishes and structure-forming invertebrates at broad spatial scales along the US portion of the California Current. These associations vary substantially between two genera Sebastolobus (thornyheads) and Sebastes, where Sebastes can be further divided into northern and coastwide species. Previous work has shown phylogenetic signals in covariation among fishes (Thorson et al., 2015a, 2016a,b) or other species (Ovaskainen et al., 2010), but ours is the first study to (i) use a spatio-temporal statistical model to estimate covariance between fishes and structure-forming invertebrates, and (ii) decompose this covariation into components representing encounter probabilities vs. positive catch rates [i.e. using the delta models that are conventional in fisheries science; Maunder and Punt, 2004]. Although the JDSDM was more parsimonious and had better predictive performance than single-species models (as shown by AIC and cross-validation analysis), the multispecies analysis resulted in slightly wider confidence interval estimates than analysing data for each species individually. At a coastwide spatial scale, we estimate an increased encounter probability for Sebastolobus and a decreased encounter probability for Sebastes species where SFIs are present. In contrast, alternative visual sampling at fine spatial scales often shows a large increase in Sebastes density, given the presence of SFIs, and Sebastolobus densities are less often reported to be associated with biogenic habitat (Brodeur, 2001; Tissot et al., 2008; Yoklavich and O’Connell, 2008; du Preez and Tunnicliffe, 2011). Recent research suggests that correlations in distribution among species will often differ when looking at small and large scales (Ovaskainen et al., 2016), and this may explain why our results differ from those from fine-scale visual sampling. Alternatively, differences in results may arise because visual sampling often occurs in rocky habitats, whereas our analysis relies on bottom trawl data that are primarily available in soft-sediment habitats. We recommend future research combining data from small and coastwide scales (and both hard- and soft-bottom habitat) within a single spatio-temporal statistical model, where density-variation at fine scales could be obtained by either fishery-dependent catch-rate data or direct observations (Jagielo et al., 2003; Shelton et al., 2014; Rooper et al., 2016; Thorson et al., 2016). We also recommend future research to include habitat variables and associations within size-structured spatio-temporal models (e.g. Kristensen et al., 2014; Nielsen et al., 2014). These models could then estimate separate habitat associations for juvenile and adult fishes and be used to target spatial management towards the more vulnerable or sensitive life stage for protected species. Based on our results, we find that simultaneously modelling fishes and SFI yields more parsimonious predictions of density and also facilitates estimating variation in density at finer spatial scales than single-species models, even for species with few encounters (e.g. POP and sharpchin rockfishes). However, incorporating these associations when estimating trends in abundance or distribution does not shrink confidence intervals. For an ecologist conducting a stock assessment, incorporating multispecies data may complicate their description of estimated abundance indices, thereby decreasing stakeholder trust in the stock assessment process. Therefore, we imagine that our results will encourage many assessment scientists to continue using single-species models for estimating abundance indices. From a broader perspective, however, the increased parsimony and out-of-sample predictive ability of the multispecies model indicate that estimates of local density are generally improved by jointly modelling multiple species (including both fishes and biogenic habitat). Precise predictions of local density for rare species might be particularly useful for ecosystem modellers, who often initialize spatial ecosystem models using sparse sampling data for rare species or ecosystem components. These estimates of local density could also be used to prioritize areas for spatial management that have a high density of structure-forming invertebrates and fishes. Therefore, we suggest further research regarding the association of fished species and biogenic habitat, including the likely impact of spatial management on fishery productivity in the West Coast groundfish fishery. Acknowledgements We thank the NWFSC FRAM Fisheries Research Survey Team and the crew on the US West Coast Groundfish Bottom Trawl Survey for collecting the data, and we thank Michelle McClure, Trevor Branch, and Tim Essington for helpful comments and discussion. We also thank Mary Yoklavich, Joe Bizzarro, Chris Rooper, and two anonymous reviewers for comments on an earlier draft. Funding L.A.K.B. gratefully acknowledges funding from the Joint Institute for the Study of the Atmosphere and Ocean (JISAO) under NOAA Cooperative Agreement No. NA15OAR4320063, Contribution No. 2016-01-38. References Akaike H. 1974 . New look at statistical-model identification . IEEE Transactions on Automatic Control , AC19 : 716 – 723 . Google Scholar Crossref Search ADS WorldCat Bradburn M. J. , Keller A. A., Horness B. H. 2011 . The 2003 to 2008 US West Coast bottom trawl surveys of groundfish resources off Washington, Oregon, and California: Estimates of distribution, abundance, length, and age composition . NOAA Technical Memorandum, NMFS-NWFSC-114 . 323 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Brodeur R. D. 2001 . Habitat-specific distribution of Pacific ocean perch (Sebastes alutus) in Pribilof Canyon, Bering Sea . Continental Shelf Research , 21 : 207 – 224 . Google Scholar Crossref Search ADS WorldCat Burnham K. P. , Anderson D. 2002 . Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach , 2nd edn. Springer , New York . 488 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Clark J. S. , Gelfand A. E., Woodall C. W., Zhu K. 2014 . More than the sum of the parts: forest climate response from joint species distribution models . Ecological Applications , 24 : pp. 990 – 999 . Google Scholar Crossref Search ADS PubMed WorldCat Francis R. I. C. C. 2011 . Data weighting in statistical fisheries stock assessment models . Canadian Journal of Fisheries and Aquatic Sciences , 68 : 1124 – 1138 . Google Scholar Crossref Search ADS WorldCat Helser T. E. , Punt A. E., Methot R. D. 2004 . A generalized linear mixed model analysis of a multi-vessel fishery resource survey . Fisheries Research , 70 : 251 – 264 . Google Scholar Crossref Search ADS WorldCat Hyde J. R. , Vetter R. D. 2007 . The origin, evolution, and diversification of rockfishes of the genus Sebastes (Cuvier) . Molecular Phylogenetics and Evolution , 44 : 790 – 811 . Google Scholar Crossref Search ADS PubMed WorldCat Ingram T. , Shurin J. B. 2009 . Trait-based assembly and phylogenetic structure in northeast Pacific rockfish assemblages . Ecology , 90 : 2444 – 2453 . Google Scholar Crossref Search ADS PubMed WorldCat Jagielo T. , Hoffmann A., Tagart J., Zimmermann M. 2003 . Demersal groundfish densities in trawlable and untrawlable habitats off Washington: Implications for the estimation of habitat bias in trawl surveys . Fishery Bulletin , 101 : 545 – 565 . Google Scholar OpenURL Placeholder Text WorldCat Kristensen K. , Nielsen A., Berg C. W., Skaug H., Bell B. M. 2016 . TMB: Automatic differentiation and Laplace approximation . Journal of Statistical Software , 70 : 1 – 21 . Google Scholar Crossref Search ADS WorldCat Kristensen K. , Thygesen U. H., Andersen K. H., Beyer J. E. 2014 . Estimating spatio-temporal dynamics of size-structured populations . Canadian Journal of Fisheries and Aquatic Sciences , 71 : 326 – 336 . Google Scholar Crossref Search ADS WorldCat Latimer A. M. , Banerjee S., Sang H. Jr, Mosher E. S., Silander J. A. Jr, 2009 . Hierarchical models facilitate spatial analysis of large data sets: a case study on invasive plant species in the northeastern United States . Ecology Letters , 12 : 144 – 154 . Google Scholar Crossref Search ADS PubMed WorldCat Lindgren F. 2012 . Continuous domain spatial models in R-INLA . ISBA Bulletin , 19 : 14 – 20 . Google Scholar OpenURL Placeholder Text WorldCat Lindgren F. , Rue H., Lindström J. 2011 . An explicit link between Gaussian fields and Gaussian Markov random fields: the stochastic partial differential equation approach . Journal of the Royal Statistical Society: Series B (Statistical Methodology) , 73 : 423 – 498 . Google Scholar Crossref Search ADS WorldCat Love M. S. , Yoklavich M. M., Thorsteinson L. K. 2002 . The Rockfishes of the Northeast Pacific . University of California Press , Berkeley . 416 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Mangel M. , Kindsvater H. K., Bonsall M. B. 2007 . Evolutionary analysis of life span, competition, and adaptive radiation, motivated by the Pacific rockfishes (Sebastes) . Evolution , 61 : 1208 – 1224 . Google Scholar Crossref Search ADS PubMed WorldCat Martin T. G. , Wintle B. A., Rhodes J. R., Kuhnert P. M., Field S. A., Low-Choy S. J., Tyre A. J. et al. 2005 . Zero tolerance ecology: Improving ecological inference by modelling the source of zero observations . Ecology Letters , 8 : 1235 – 1246 . Google Scholar Crossref Search ADS PubMed WorldCat Maunder M. N. , Punt A. E. 2004 . Standardizing catch and effort data: a review of recent approaches . Fisheries Research , 70 : 141 – 159 . Google Scholar Crossref Search ADS WorldCat Maunder M. N. , Punt A. E. 2013 . A review of integrated analysis in fisheries stock assessment . Fisheries Research , 142 : 61 – 74 . Google Scholar Crossref Search ADS WorldCat Methot R. D. 2009 . Stock assessment: Operational models in support of fisheries management. In The Future of Fisheries Science in North America , pp. 137 – 165 . Ed. by Beamish R. J., Rothschild B. J.. Springer , Netherlands, Dordrecht . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Nielsen J. R. , Kristensen K., Lewy P., Bastardie F. 2014 . A statistical model for estimation of fish density including correlation in size, space, time and between species from research survey data . PLoS One , 9 : e99151. Google Scholar Crossref Search ADS PubMed WorldCat Ovaskainen O. , Abrego N., Halme P., Dunson D. 2016 . Using latent variable models to identify large networks of species-to-species associations at different spatial scales . Methods in Ecology and Evolution , 7 : 549 – 555 . Google Scholar Crossref Search ADS WorldCat Ovaskainen O. , Hottola J., Siitonen J. 2010 . Modeling species co-occurrence by multivariate logistic regression generates new hypotheses on fungal interactions . Ecology , 91 : 2514 – 2521 . Google Scholar Crossref Search ADS PubMed WorldCat Perry A. L. , Low P. J., Ellis J. R., Reynolds J. D. 2005 . Climate change and distribution shifts in marine fishes . Science , 308 : 1912 – 1915 . Google Scholar Crossref Search ADS PubMed WorldCat Pinsky M. L. , Worm B., Fogarty M. J., Sarmiento J. L., Levin S. A. 2013 . Marine taxa track local climate velocities . Science , 341 : 1239 – 1242 . Google Scholar Crossref Search ADS PubMed WorldCat du Preez C. , Tunnicliffe V. 2011 . Shortspine thornyhead and rockfish (Scorpaenidae) distribution in response to substratum, biogenic structures and trawling . Marine Ecology Progress Series , 425 : 217 – 231 . Google Scholar Crossref Search ADS WorldCat R Core Team . 2015 . R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Available from https://www.R-project.org/. Rooper C. N. , Sigler M. F., Goddard P., Malecha P., Towler R., Williams K., Wilborn R. et al. 2016 . Validation and improvement of species distribution models for structure-forming invertebrates in the eastern Bering Sea with an independent survey . Marine Ecology Progress Series , 551 : 117 – 130 . Google Scholar Crossref Search ADS WorldCat Shelton A. O. , Thorson J. T., Ward E. J., Feist B. E. 2014 . Spatial semiparametric models improve estimates of species abundance and distribution . Canadian Journal of Fisheries and Aquatic Sciences , 71 : 1655 – 1666 . Google Scholar Crossref Search ADS WorldCat Simpson D. , Lindgren F., Rue H. 2012 . In order to make spatial statistics computationally feasible, we need to forget about the covariance function . Environmetrics , 23 : 65 – 74 . Google Scholar Crossref Search ADS WorldCat Skaug H. , Fournier D. 2006 . Automatic approximation of the marginal likelihood in non-Gaussian hierarchical models . Computational Statistics & Data Analysis , 51 : 699 – 709 . Google Scholar Crossref Search ADS WorldCat Thorson J. T. , Fonner R., Haltuch M., Ono K., Winker H. 2016 . Accounting for spatiotemporal variation and fisher targeting when estimating abundance from multispecies fishery data . Canadian Journal of Fisheries and Aquatic Sciences , doi: 10.1139/cjfas-2015-0598. Google Scholar OpenURL Placeholder Text WorldCat Thorson J. T. , Ianelli J. N., Larsen E. A., Ries L., Scheuerell M. D., Szuwalski C., Zipkin E. F. 2016a . Joint dynamic species distribution models: a tool for community ordination and spatio-temporal monitoring . Global Ecology and Biogeography , 25 : 1144 – 1158 . Google Scholar Crossref Search ADS WorldCat Thorson J. T. , Minto C. 2015 . Mixed effects: a unifying framework for statistical modelling in fisheries biology . ICES Journal of Marine Science , 72 : 1245 – 1256 . Google Scholar Crossref Search ADS WorldCat Thorson J. T. , Pinsky M. L., Ward E. J. 2016b . Model-based inference for estimating shifts in species distribution, area occupied and centre of gravity . Methods in Ecology and Evolution , 7 : 990 – 1002 . Google Scholar Crossref Search ADS WorldCat Thorson J. T. , Scheuerell M. D., Shelton A. O., See K. E., Skaug H. J., Kristensen K. 2015a . Spatial factor analysis: a new tool for estimating joint species distributions and correlations in species range . Methods in Ecology and Evolution , 6 : 627 – 637 . Google Scholar Crossref Search ADS WorldCat Thorson J. T. , Shelton A. O., Ward E. J., Skaug H. J. 2015b . Geostatistical delta-generalized linear mixed models improve precision for estimated abundance indices for West Coast groundfishes . ICES Journal of Marine Science , 72 : 1297 – 1310 . Google Scholar Crossref Search ADS WorldCat Thorson J. T. , Stewart I., Punt A. 2011 . Accounting for fish shoals in single- and multi-species survey data using mixture distribution models . Canadian Journal of Fisheries and Aquatic Sciences , 68 : 1681 – 1693 . Google Scholar Crossref Search ADS WorldCat Thorson J. T. , Ward E. J. 2014 . Accounting for vessel effects when standardizing catch rates from cooperative surveys . Fisheries Research , 155 : 168 – 176 . Google Scholar Crossref Search ADS WorldCat Tissot B. N. , Wakefield W. W., Hixon M. A., Clemons J. E. 2008 . Twenty years of fish-habitat studies on Heceta Bank, Oregon. In Marine Habitat Mapping Technology for Alaska , pp. 203 – 217 . Ed. by Reynolds J. R., Greene H. G.. Alaska Sea Grant College Program, University of Alaska Fairbanks . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Tobler W. R. 1970 . A computer movie simulating urban growth in the Detroit region . Economic Geography , 46 : 234 – 240 . Google Scholar Crossref Search ADS WorldCat Walters C. 2003 . Folly and fantasy in the analysis of spatial catch rate data . Canadian Journal of Fisheries and Aquatic Sciences , 60 : 1433 – 1436 . Google Scholar Crossref Search ADS WorldCat Warton D. I. , Blanchet F. G., O'Hara R. B., Ovaskainen O., Taskinen S., Walker S. C., Hui F. K. 2015 . So many variables: Joint modeling in community ecology . Trends in Ecology & Evolution , 30 : 766 – 779 . Google Scholar Crossref Search ADS PubMed WorldCat Yoklavich M. M. , O’Connell V. 2008 . Twenty years of research on demersal communities using the Delta submersible in the Northeast Pacific. In Marine Habitat Mapping Technology for Alaska , pp. 143 – 155 . Ed. by Reynolds J. R., Greene H. G.. Alaska Sea Grant College Program, University of Alaska Fairbanks . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Zipkin E. F. , Andrew Royle J., Dawson D. K., Bates S. 2010 . Multi-species occurrence models to evaluate the effects of conservation and management actions . Biological Conservation , 143 : 479 – 484 . Google Scholar Crossref Search ADS WorldCat Published by International Council for the Exploration of the Sea 2017. This work is written by US Government employees and is in the public domain in the US. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) Published by International Council for the Exploration of the Sea 2017. This work is written by US Government employees and is in the public domain in the US.

Orio,, Alessandro;Florin,, Ann-Britt;Bergström,, Ulf;Šics,, Ivo;Baranova,, Tatjana;Casini,, Michele

doi: 10.1093/icesjms/fsx005pmid: N/A

Standardized indices of abundance and size-based indicators are of extreme importance for monitoring fish population status. The main objectives of the current study were to (i) combine and standardize recently performed trawl survey with historical ones, (ii) explore and discuss the trends in abundance, and (iii) the trends in maximum length (Lmax) for cod (Gadus morhua) and flounder (Platichthys flesus) stocks in the Baltic Sea. Standardization of catch per unit of effort (CPUE) from trawl surveys from 1978 to 2014 to swept area per unit of time was conducted using information on trawling speed and horizontal opening of the trawls. CPUE data for cod and flounder stocks were modelled using generalized additive models (GAMs) in a delta modelling approach framework, while the Lmax data were modelled using ordinary GAMs. The CPUE time series of the Eastern Baltic cod stock closely resembles the spawning stock biomass trend from analytical stock assessment. The results obtained furnish evidence of the cod spill-over from Subdivisions (SD) 25–28 to SD 24. The decline of Lmax in recent years was evident for both species in all the stocks analysed indicating that the demersal fish community is becoming progressively dominated by small individuals. It is concluded that the standardization of long time series of fisheries-independent data constitutes a powerful tool that could help improve our knowledge on the dynamics of fished populations, thus promoting a long-term sustainable use of these marine resources. Introduction Indices of abundance based on fisheries-independent survey data are one of the most crucial inputs in analytical fish stock assessments (Maunder and Punt, 2004; Francis, 2011). Moreover, when analytical assessments cannot be performed, fisheries-independent data can be used to follow temporal trends in stock abundance to evaluate the state of the stock in relation to historical baselines. The lack of historical baselines, however, could lead to overly optimistic or misleading assessments of the status of fished populations that may affect management actions (Shifting baseline syndrome; Pauly, 1995; Pinnegar and Engelhard, 2008; Cardinale et al., 2009). Ideally, these indices of abundance should be derived from data collected during standardized scientific surveys that have used the same gear and sampling scheme throughout the entire time series in order to avoid changes in catchability (Maunder and Punt, 2004; Cosgrove et al., 2014; Thorson et al., 2015). In reality, changes in gear types and sampling schemes almost always occur especially when surveys have been conducted for a long period. Standardization is, therefore, obviously a key element in order to be able to make use of the enormous effort allocated in data collections throughout the years and thus to increase the temporal and spatial extent of the analyses. Survey data, besides being important for stock assessment, also give information on the status of fish populations through indicators that can be derived from size frequency distribution, such as maximum length (Lmax) or length at first maturity (Blanchard et al., 2005; Shin et al., 2005). Changes in size structure of a population can be caused by direct or indirect effects of fishing, changes in the environmental conditions, genetic variability as well as inter- and intraspecific interactions (Nicholson and Jennings, 2004; Shin et al., 2005; Walsh et al., 2006). If they remain undetected, these changes could result in the use of erroneous reference levels in stock assessments with possibly severe effects on management efficiency (Heino et al., 2013). In the Marine Strategy Framework Directive (EU-COM, 2008), indicators of age and size structure are pivotal for assessing the status of exploited fish stocks, especially for stocks for which analytical assessments are not currently performed (Probst et al., 2013). In particular, Lmax is considered a good indicator of the status of a fish population because bigger individuals have higher fecundity, better egg quality, and higher reproductive success (Hixon et al., 2014). Further, Lmax is sensitive to fishing pressure since most fisheries are selectively removing the largest individuals of a population, but it can also respond to environmental factors such as, for example, changes in temperature (Piet and Jennings, 2004; Maschner et al., 2008; ICES, 2012). Cod (Gadus morhua, Gadidae) and flounder (Platichthys flesus, Pleuronectidae) are two key species of the Baltic Sea, both ecologically and commercially (Casini et al., 2008; Florin and Höglund, 2008; Lindegren et al., 2009). Abundance trends of cod in the Baltic Sea are known from analytical stock assessments, but long-term trends from fisheries-independent data are lacking (Eero et al., 2011; ICES, 2015a). For flounder, on the other hand, information on populations’ development is scarce because of the lack of analytical assessments and because the abundance trends derived from fisheries-independent data cover only the last 15 years (ICES, 2015a). The use of fisheries-independent information collected through scientific trawl surveys is limited in the Baltic Sea by the lack of long-term standardized time series. This is due to the fact that from 2001 the Baltic International Trawl Survey (BITS; ICES, 2014a) has been carried out with a new sampling scheme and standard gear (ICES, 2015a) and that BITS data before and after 2001 have been standardized only for cod. Moreover, no data from historical surveys performed prior to BITS (i.e. before 1991) have been available to date. Given the central importance of the indices of abundance and of size-based indicators for monitoring fish population status, the main objectives of the current study were (i) to combine and standardize recently performed trawl survey with historical ones, (ii) to explore and discuss the trends in abundance, and (iii) the trends in Lmax for cod and flounder stocks in the Baltic Sea. Material and methods Available data Catch and individual data for cod and flounder collected during the BITS (ICES, 2014a) in ICES Subdivisions (SDs) 22–29 (Figure 1) between 1991 and 2014 were downloaded from the ICES DATRAS database (datras.ices.dk; accessed on the 28 April 2015). Additionally, we compiled historical catch and individual data collected during bottom trawl surveys in the Baltic Sea carried out in the years 1978–1990 by the former Swedish Board of Fisheries (currently the Swedish University of Agricultural Sciences, Department of Aquatic Resources) and the former Baltic Fisheries Research institute (BaltNIIRH; currently the Latvian Institute of Food Safety, Animal Health and Environment). These historical data have been recently digitized. The catch data were constituted by catch in numbers per 1-cm length-class and total catch in weight for each trawl haul and were accompanied with information on haul duration, towing speed, fishing date, quarter (Q1  =  January–March, Q2  =  April–June, etc.), as well as setting and hauling position in latitude and longitude. Length frequency distribution (LFD) data were not available for all the trawl hauls. The individual data consisted of information on total length, total weight, age, sex, and maturity stage of individual cod and flounder caught during the trawl surveys. Figure 1 Open in new tabDownload slide Map of the study area divided in ICES Subdivisions. Figure 1 Open in new tabDownload slide Map of the study area divided in ICES Subdivisions. For all surveys, trawl hauls classified as “Valid”, “Additional” and “No Oxygen” were included in the analyses (ICES, 2014a). “Additional” hauls are valid hauls not used to calculate indices of abundance for stock assessment but mainly to collect biological parameters while “No Oxygen” hauls are hauls not performed because the bottom oxygen level is < 1.5 ml*l − 1 and the catch is assumed to be zero. Catch per unit of efforts (CPUEs) for each length-class of cod and flounder were calculated as number of fish caught in 1 hour of trawling (no*h − 1), for the hauls in which the LFDs were available. CPUEs in weight (kg*h − 1) per length-class were estimated using the year-specific length–weight relationships (W  =  a*Lb) of the individual data. For the two Baltic cod stocks (SDs 22–24 and SDs 25–32), an analysis on the individual data showed different temporal trends in length–weight relationship (data not shown). Therefore, we used different year-specific length–weight relationships for the two stocks. For cod in SDs 22–24 individual weights were not available in 1988–1991 and an average of the parameters (a and b) estimated from 1992 to 1994 was used. For the four flounder stocks (SDs 22–23, SDs 24–25, SDs 26 & 28, SDs 27 & SDs 29–32), on the other hand, a common year-specific length-weight relationship was used since the temporal trends in the parameters a and b did not show any difference between stocks. Individual weights were not available for flounder in 1978, and thus an average of the parameters estimated from 1979 to 1981 was used. For the hauls in which the LFDs were not available, CPUEs in weight (kg*h − 1) were estimated from the total catch in weight. Standardization of CPUE Standardization of CPUEs from BITS and historical Swedish and Latvian surveys to swept area per unit of time was conducted using information on horizontal opening of the trawls and trawling speed, following the approach proposed by Cardinale et al. (2009). The horizontal opening of each trawl is usually estimated as two-thirds of the length of the trawl fishing line (Rijnsdorp et al., 1996). However, with the introduction of the sweeps (extensions of the ground rope between the wings of the net and the trawl doors) in the beginning of the 1920s, the area of seabed swept by the gear has increased considerably and substantially improved trawl efficiency at very little cost in terms of additional towing power (Galbraith and Rice, 2005). Therefore, the horizontal opening of the sweeps was summed to that of the fishing line to estimate the total horizontal opening between the trawl doors (total horizontal opening). The distance between the doors is dependent on the sweeps’ length but also on the angle between the sweep and the direction of the tow. We standardized all the trawl hauls to the total horizontal opening of a TVL (standard trawl currently used in the BITS; ICES, 2014a) assuming an average angle of 15° between the sweep and the direction of the tow. We set the total horizontal opening of TVL with 75 m sweeps to 1 and the relative trawl size (RTS) of the other gears was expressed in relation to that. We gathered information about fishing line length and sweeps’ length of all the gears except for seven gears that were then removed from the analysis, managing to standardize almost 90% of the hauls. The gears that have been standardized are: Grand Overture Verticale (GOV; according to the BITS gear code), Föto bottom trawl (FOT), Latvian bottom trawl (LBT), Sonderborg trawl (SON), Herring ground trawl (H20), Herring bottom trawl (P20), Cod hopper (CHP), TV3 930 meshes (TVL), and TV3 520 meshes (TVS). The gears that were removed are Russian bottom trawl (DT), Hake-4M trawl (HAK), Danish winged bottom trawl (EXP), Estonian small bottom trawl (ESB), Granton trawl (GRT), and two unspecified trawls (CAM and EGY). We standardized all the trawl hauls to a trawling speed of three knots. We set the trawling speed of three knots to 1 and estimated the relative speed (RS) of the trawl hauls swept with a different speed. When the trawling speed was not available, the average speed of the same vessel using the same gear in the same year was used. When the vessel information was not available, the overall average speed in the same year or adjacent years were used. All CPUEs were then multiplied by the reciprocal of the RS and of the RTS in order to make the catches of all the trawl hauls comparable. The CPUE value obtained is usually defined as an area-swept abundance estimate (Harley and Myers, 2001) and it corresponds here to the abundance of fish caught by trawling for 1 h a standard bottom swept area of 0.45 km2 using a TVL trawl with 75 m sweeps at the standard speed of three knots. No standardization was performed for the different mesh sizes used by the different gears during the study period because the biggest registered stretched mesh size in our data was 30 mm, which we considered to be small enough not to introduce any bias in our analyses. Data included in the modelling of CPUE in weights and maximum length (Lmax) We decided to include only the flounder stock in SDs 24–25 and the stock in SDs 26 & 28 in the analyses. The flounder stock in SDs 22–23 was excluded as it was impossible to standardize the gears that were used to fish in those areas; the flounder stock in SDs 27 & 29–32 was excluded because the BITS survey does not cover SDs 29–32. The merging of SD 26 and 28 in the current stock definition by ICES has been questioned since tagging show very little exchange between these SDs (ICES, 2010). We therefore also investigated the trends in these SDs separately. For cod, we performed the analysis on the Eastern Baltic cod stock (SDs 25–32) only in SDs 25–28 because the BITS survey does not cover the SDs 29–32, and because we assume that the stock temporal dynamics in SDs 25–28 (main area of distribution) are consistent with the overall stock dynamics. For the Western Baltic cod stock (SDs 22–24), we decided to perform the analysis only for SD 24 because it was not possible to standardize the gears that were used in SDs 22–23. It is known that SD 24 is an important mixing area between the Eastern and the Western Baltic cod stocks (Hüssy et al., 2016) and therefore the temporal trends of CPUE and Lmax in SD 24 were compared with those in SDs 25–28 and also with that in SD 25 separately. Hereafter, for simplicity, we will refer to cod in SDs 25–28 and cod in SD 24 as the Eastern and the Western Baltic cod stock, respectively. For both cod and flounder we excluded from the analyses the Gulf of Riga (SD 28-1) because the BITS survey does not cover this area. For the CPUE analyses, we aimed at following the spatiotemporal changes in the spawning part of the stocks, which correspond to fish  ≥ 30 cm for cod (ICES, 2015a) and  ≥ 20 cm for flounder (ICES, 2014b). However, for flounder, the CPUEs per length class (i.e. the LFDs) were not available for around one third of the hauls performed before the BITS (i.e. before 1991). For the hauls in which LFDs were available, the proportion of flounder  < 20 cm in the catches was relatively constant (both temporally and spatially, i.e. among SDs) and below 10% of the total catches, except for some years in SD 26. The low proportion of small flounder in the catches is explained by the fact that the surveys do not cover the shallowest areas were juvenile flounders are found (ICES, 2015a). Therefore, we assumed that the spatiotemporal changes in the total CPUEs would reliably represent the trends of the spawning part of the flounder population. For the Lmax analyses, the maximum length (Lmax, [cm]) was defined as the maximum observed fish length in each haul. Statistical analysis The distribution of marine species is the result of the connections between the intrinsic characteristics of the populations, trophic interactions, hydrological constraints and anthropogenic factors. Because of all these interdependencies, we expect the abundance of cod and flounder to be better described by non-linear functions of space and time. In order to capture this non-linearity, we decided to use generalized additive models (GAMs) to model the trends in CPUE and Lmax. Nonlinear approaches, like GAMs, have been found to perform better than linear models for the standardization of CPUEs (Maunder, 2001). Because of the large amount of zero catches in our dataset (between 4.5 and 23.1% of all the hauls depending on the stock), the CPUE data for cod in SD 24, cod in SDs 25–28, flounder in SDs 24–25, and flounder in SDs 26 & 28 was modelled at first using ordinary GAMs with different error distributions that could deal with zero-inflated data (e.g. the quasi-Poisson distribution) but none of the models had acceptable residuals. We then decided to adopt GAMs in a delta modelling approach framework. This modelling approach has been found to be appropriate for the analysis of zero-inflated data (Stefánsson, 1996; Barry and Welsh, 2002; Maunder and Punt, 2004) and has been used to estimate the spatial distribution of marine organisms at large spatial scales (Loots et al., 2010; Lauria et al., 2011; Grüss et al., 2014; Parra et al., 2016), as well as to standardize CPUE data and indices of abundance (Berg et al., 2014; Cosgrove et al., 2014; Thorson and Ward, 2013, 2014). The delta models have become widely adopted especially in the case of survey indices standardization because they allow the separation of the model into two ecologically meaningful components (Thorson and Ward, 2013): the first estimates the probability of encountering the target species and the second estimates the population density within its range of distribution. The total abundance is then the product of the probability of encounter and the population density. The two components are essential because both the distribution range and the densities are likely changing over time. Delta GAM for the CPUE The delta GAM approach used in these analyses consists of two steps: the first involves modelling the presence/absence of the species using a binomial error distribution with a logit link function, and the second is modelling the abundance of only positive CPUE records, log-transformed, using a Gaussian error distribution with an identity link function (Lauria et al., 2011; Parra et al., 2016). The predicted probability of presence, resulting from the binomial model, was then multiplied by the log CPUE prediction, resulting from the Gaussian model, to obtain the final CPUE predictions. The full binomial model for presence/absence and the full Gaussian model for the positive CPUE values were formulated as follows: presence/absence=β(quarter)+s(long,lat)+te1(depth,year)+f1(year)+f2(depth)+f3(lat)+f4(long)+ε(1) log(CPUE)=β(quarter)+s(long,lat)+te1(long,year)+te2(lat,year)+te3(depth,year)+f1(year)+f2(depth)+f3(lat)+f4(long)+ε(2) where β is an overall intercept different for each quarter, s is an isotropic smoothing function (thin-plate regression spline; Wood, 2003), tei are tensor product smoothing functions used for representing interaction terms, fi are natural cubic splines, and ε are error terms. The interactions were introduced to take into account the changes in the spatiotemporal distribution of the species in the time period analysed. Model selection for both models was done through a backward stepwise selection approach based on statistical significance (Wood, 2006). From the full model, the non-significant predictor with the lowest significance level was excluded at each step and the model run again. This procedure was repeated until all the predictors were significant (final model). To make the interpretation of the model results easier, we set a limit to the maximum degrees of freedom (number of knots, k) allowed to the smoothing functions of the variables latitude, longitude and depth (k  =  4) and of the interaction between latitude and longitude (k  =  20). GAM for Lmax The Lmax data for cod in SD 24, cod in SDs 25–28, flounder in SDs 24–25, and flounder in SDs 26 & 28 were modelled with a GAM using a Gaussian distribution since the Lmax values were normally distributed (Hastie and Tibshirani, 1990). The full model was formulated as follows: Lmax=β(quarter)+s(long,lat)+te1(long,year)+te2(lat,year)+te3(depth,year)+f1(year)+f2(depth)+f3(lat)+f4(long)+ε(3) where β is an overall intercept different for each quarter, s is an isotropic smoothing function (thin-plate regression spline; Wood, 2003), tei are tensor product smoothing functions, fi are natural cubic splines, and ε is an error term. The interactions were introduced to take into account the changes in the spatio-temporal distribution of the species in the time period analysed. No correlation was found between trawl duration and Lmax of cod (r  =  0.18) and flounder (r  =  0.04), therefore we did not include it in the full model formulation. As for the previous models, model selection was done through a backward stepwise selection approach based on statistical significance (Wood, 2006). To make the interpretation of the model results easier, also for Lmax we set a limit to the maximum number of knots (k) allowed to the smoothing functions of the variables latitude, longitude and depth (k  =  4) and of the interaction between latitude and longitude (k  =  20). Reconstructing the trends in CPUE and Lmax of different stocks The final models for each stock were used to predict the annual CPUE and Lmax over a regular grid of 0.02° × 0.01°. The area in SD 27 north of 58° was removed from the predictions due to incomplete spatial coverage. Because of poor survey coverage in shallow and deep parts of the different areas, depths shallower than 10 m were excluded from the predictions of cod in SD 24, depths shallower than 8 m and deeper than 150 m were excluded from the predictions of cod in SDs 25–28, depths shallower than 10 m and deeper than 100 m were excluded from the predictions of flounder in SDs 24–25 and depths shallower than 8 m and deeper than 150 m were excluded from the predictions of flounder in SDs 26 & 28. To produce the trends in mean CPUE and mean Lmax, all the predicted estimates were averaged over the whole grid in the respective stock area for each year for quarter 1 (Maunder and Punt, 2004; Beare et al., 2005; Cardinale et al., 2011). Reconstructing the trends in exploitation rate of different stocks The CPUE trends produced from the final models were used to obtain trends in exploitation rate, estimated as the ratio of the commercial catches (ICES, 2015a) to the CPUEs. The commercial catches (landings and discards) of cod ≥ 30 cm (size range considered in the CPUEs analyses) have constituted more than ∼ 98% of the total catches in weights between 2000 and 2014 (ICES, 2015a) (fish  ≥ 30 cm includes discards since the minimum landing size has varied between 35 and 38 cm in this period). In earlier years (before 1994), the minimum landing size was 33 cm, therefore still higher than the lower length boundary used in the CPUE estimations. We are therefore confident that our CPUEs for fish  ≥ 30 cm include the component of the cod populations exploited by the fishery. For flounder, the temporal trends in exploitation rate were estimated using the ratio of the commercial landings (official ICES landings; ICES, 2013; ICES, 2015a) to the CPUEs obtained from our final models. Flounder ≥ 20 cm constitute by far the largest part of the commercial catches (more than 90%), while fish < 20 cm are caught (and then landed or discarded) very seldom (ICES, 2014b). We are therefore confident also for flounder that the CPUEs estimated in our study include the component of the populations exploited by the fishery. On the other hand, discards of flounder could be quite substantial in the Baltic demersal fishery, but their survival is also high, over 50% in cold seasons (ICES, 2014b). Therefore, although uncertainties exist with respect to survival rate of discarded flounder, we believe that exploitation rate as estimated in our study is a reliable proxy for fishing mortality also for this species. All the analyses were performed using R software and the mgcv library of R (Wood, 2011; R Core Team, 2015). Results Trends in CPUE Standardized CPUEs from a total of 10 198 hauls for cod and 9548 hauls for flounder, were included in the analyses (Supplementary Table S1). The equations of the final models varied between modelling approaches (Binomial and Gaussian) and stocks, but the interactions between latitude and longitude and the effect of quarters were always retained after the backward stepwise selection procedure (Table 1). Final binomial models explained between 23.3% of the deviance for cod in SDs 25–28 and 40.9% for cod in SD 24, while Gaussian models explained between 32% for flounder in SDs 24–25 and 42.9% for cod in SDs 25–28. In general, the adjusted R2 indicated a better model fit for the Gaussian models compared to the binomial models. Analysis of the residuals in some cases revealed slight departures from the model assumptions, but we considered the overall quality of the residuals to be satisfactory (Supplementary Figures S1–S8). Table 1 Summary statistics of the Delta GAMs used to estimate the CPUE trends for each stock analysed. Stocks . Years . Quarters . Models . n . df . Variables retained . Dev% . Adj-R2 . Cod 24 1988–2014 1,4 Binomial 2036 19.3 Lat:Long, Depth (as linear effect), Year, Quarter 40.9 0.277 Gaussian 1944 48.6 Lat:Long, Depth:Year, Lat, Long, Year, Quarter 40.9 0.394 Cod 25–28 1978–2014 1,2,3,4 Binomial 8162 47.4 Lat:Long, Depth:Year, Lat, Long, Quarter 23.3 0.250 Gaussian 6784 78.0 Lat:Long, Lat:Year, Long:Year, Depth:Year, Lat, Year, Quarter 42.9 0.423 Flounder 24–25 1988–2014 1,3,4 Binomial 5230 33.5 Lat:Long, Depth:Year, Lat, Year, Quarter 31.1 0.310 Gaussian 4709 73.2 Lat:Long, Lat:Year, Long:Year, Depth:Year, Lat, Year, Quarter 32.0 0.309 Flounder 26 & 28 1978–2014 1,2,3,4 Binomial 4318 41.0 Lat:Long, Depth:Year, Year, Quarter 26.8 0.276 Gaussian 3320 70.7 Lat:Long, Lat:Year, Long:Year, Depth:Year, Depth, Quarter 34.9 0.335 Stocks . Years . Quarters . Models . n . df . Variables retained . Dev% . Adj-R2 . Cod 24 1988–2014 1,4 Binomial 2036 19.3 Lat:Long, Depth (as linear effect), Year, Quarter 40.9 0.277 Gaussian 1944 48.6 Lat:Long, Depth:Year, Lat, Long, Year, Quarter 40.9 0.394 Cod 25–28 1978–2014 1,2,3,4 Binomial 8162 47.4 Lat:Long, Depth:Year, Lat, Long, Quarter 23.3 0.250 Gaussian 6784 78.0 Lat:Long, Lat:Year, Long:Year, Depth:Year, Lat, Year, Quarter 42.9 0.423 Flounder 24–25 1988–2014 1,3,4 Binomial 5230 33.5 Lat:Long, Depth:Year, Lat, Year, Quarter 31.1 0.310 Gaussian 4709 73.2 Lat:Long, Lat:Year, Long:Year, Depth:Year, Lat, Year, Quarter 32.0 0.309 Flounder 26 & 28 1978–2014 1,2,3,4 Binomial 4318 41.0 Lat:Long, Depth:Year, Year, Quarter 26.8 0.276 Gaussian 3320 70.7 Lat:Long, Lat:Year, Long:Year, Depth:Year, Depth, Quarter 34.9 0.335 The variables retained in the final models are indicated; n  =  numbers of hauls used in the models; df  =  degrees of freedom; Dev%  =  explained deviance; Adj-R2  =  adjusted R2. Table 1 Summary statistics of the Delta GAMs used to estimate the CPUE trends for each stock analysed. Stocks . Years . Quarters . Models . n . df . Variables retained . Dev% . Adj-R2 . Cod 24 1988–2014 1,4 Binomial 2036 19.3 Lat:Long, Depth (as linear effect), Year, Quarter 40.9 0.277 Gaussian 1944 48.6 Lat:Long, Depth:Year, Lat, Long, Year, Quarter 40.9 0.394 Cod 25–28 1978–2014 1,2,3,4 Binomial 8162 47.4 Lat:Long, Depth:Year, Lat, Long, Quarter 23.3 0.250 Gaussian 6784 78.0 Lat:Long, Lat:Year, Long:Year, Depth:Year, Lat, Year, Quarter 42.9 0.423 Flounder 24–25 1988–2014 1,3,4 Binomial 5230 33.5 Lat:Long, Depth:Year, Lat, Year, Quarter 31.1 0.310 Gaussian 4709 73.2 Lat:Long, Lat:Year, Long:Year, Depth:Year, Lat, Year, Quarter 32.0 0.309 Flounder 26 & 28 1978–2014 1,2,3,4 Binomial 4318 41.0 Lat:Long, Depth:Year, Year, Quarter 26.8 0.276 Gaussian 3320 70.7 Lat:Long, Lat:Year, Long:Year, Depth:Year, Depth, Quarter 34.9 0.335 Stocks . Years . Quarters . Models . n . df . Variables retained . Dev% . Adj-R2 . Cod 24 1988–2014 1,4 Binomial 2036 19.3 Lat:Long, Depth (as linear effect), Year, Quarter 40.9 0.277 Gaussian 1944 48.6 Lat:Long, Depth:Year, Lat, Long, Year, Quarter 40.9 0.394 Cod 25–28 1978–2014 1,2,3,4 Binomial 8162 47.4 Lat:Long, Depth:Year, Lat, Long, Quarter 23.3 0.250 Gaussian 6784 78.0 Lat:Long, Lat:Year, Long:Year, Depth:Year, Lat, Year, Quarter 42.9 0.423 Flounder 24–25 1988–2014 1,3,4 Binomial 5230 33.5 Lat:Long, Depth:Year, Lat, Year, Quarter 31.1 0.310 Gaussian 4709 73.2 Lat:Long, Lat:Year, Long:Year, Depth:Year, Lat, Year, Quarter 32.0 0.309 Flounder 26 & 28 1978–2014 1,2,3,4 Binomial 4318 41.0 Lat:Long, Depth:Year, Year, Quarter 26.8 0.276 Gaussian 3320 70.7 Lat:Long, Lat:Year, Long:Year, Depth:Year, Depth, Quarter 34.9 0.335 The variables retained in the final models are indicated; n  =  numbers of hauls used in the models; df  =  degrees of freedom; Dev%  =  explained deviance; Adj-R2  =  adjusted R2. Time series of the estimated CPUEs for the cod and flounder stocks are presented in Figure 2. For cod in SD 24 (Figure 2a), the highest CPUE of around 140 kg*h − 1 was observed at the beginning of the time series (1988) and two other smaller peaks occurred around 1995 and 2007. For cod in SDs 25–28 (Figure 2b), the CPUE was at the maximum around 1981–1982 (220 kg*h − 1) and a smaller peak of CPUE was revealed around 2008–2010. However, the CPUE maximum in SD 25 occurred a little later (1982–1984) than in SDs 26–28. Moreover, since the early 1990s, the CPUEs in SD 25 were always higher than in SDs 26–28, while the temporal variations were coincident. The CPUE has decreased from 2009 to 2014 by around 60% in the entire area. Flounder in SDs 24–25 (Figure 2c) shows a decline in CPUEs from a maximum of 44 kg*h − 1 at the beginning of the time series (1988) to the end of 1990s (approximately 8 kg*h − 1), whereas thereafter an increase of around 70% occurred up to 2014. Flounder in SDs 26 & 28 (Figure 2d) shows the highest CPUEs of 110 kg*h − 1 at the beginning of the time series (1978–1979) and a sharp decline of around 90% up to mid-1980s. Thereafter, a general increase occurred up to early 2000s followed by a decrease of approximately 85%. However, the CPUEs in SD 28 were always higher than in SD 26, especially during the periods of highest CPUEs. In general, flounder CPUE decreased in both SDs 26 and 28 in the last 5 years but in SD 28 flounder started to decline already in the early 2000s, whereas in SD 26 the decline was evident only since 2010. Figure 2 Open in new tabDownload slide Estimated average yearly CPUE (kg*h − 1) predicted by the models for (a) cod in SD 24, (b) cod in SDs 25–28, (c) flounder in SDs 24–25 and (d) flounder in SDs 26 & 28. Figure 2 Open in new tabDownload slide Estimated average yearly CPUE (kg*h − 1) predicted by the models for (a) cod in SD 24, (b) cod in SDs 25–28, (c) flounder in SDs 24–25 and (d) flounder in SDs 26 & 28. Trends in Lmax Lmax values from a total of 9005 hauls for cod and 7627 hauls for flounder were included in the analyses. The formulas of the final models varied between stocks, but the interactions between latitude and longitude and between depth and year, the smoother on the year and the effect of quarters were always retained (Table 2). Final models explained between 19.6% of the deviance for cod in SD 24 and 34.6% for cod in SDs 25–28. Analysis of the residuals in some cases revealed slight departures from the model assumptions, but we considered the overall quality of the residuals to be satisfactory (Supplementary Figures S9–S12). Table 2 Summary statistics of GAMs used to estimate the Lmax trends for each stock analysed. Stocks . Years . Quarters . n . df . Variables retained . Dev% . Adj-R2 . Cod 24 1988–2014 1,4 1979 40.4 Lat:Long, Depth:Year, Year, Quarter 19.6 0.180 Cod 25–28 1978–2014 1,2,3,4 7026 79.2 Lat:Long, Lat:Year, Long:Year, Depth:Year, Depth, Year, Quarter 34.6 0.339 Flounder 24–25 1988–2014 1,3,4 4706 57.4 Lat:Long, Lat:Year, Long:Year, Depth:Year, Depth, Year, Quarter 22.8 0.218 Flounder 26&28 1978–2014 1,2,3,4 2921 65.7 Lat:Long, Lat:Year, Long:Year, Depth:Year, Year, Quarter 31.7 0.301 Stocks . Years . Quarters . n . df . Variables retained . Dev% . Adj-R2 . Cod 24 1988–2014 1,4 1979 40.4 Lat:Long, Depth:Year, Year, Quarter 19.6 0.180 Cod 25–28 1978–2014 1,2,3,4 7026 79.2 Lat:Long, Lat:Year, Long:Year, Depth:Year, Depth, Year, Quarter 34.6 0.339 Flounder 24–25 1988–2014 1,3,4 4706 57.4 Lat:Long, Lat:Year, Long:Year, Depth:Year, Depth, Year, Quarter 22.8 0.218 Flounder 26&28 1978–2014 1,2,3,4 2921 65.7 Lat:Long, Lat:Year, Long:Year, Depth:Year, Year, Quarter 31.7 0.301 The variables retained in the final models are indicated; n  =  numbers of hauls used in the models; df  =  degrees of freedom; Dev%  =  explained deviance; Adj-R2  =  adjusted R2. Table 2 Summary statistics of GAMs used to estimate the Lmax trends for each stock analysed. Stocks . Years . Quarters . n . df . Variables retained . Dev% . Adj-R2 . Cod 24 1988–2014 1,4 1979 40.4 Lat:Long, Depth:Year, Year, Quarter 19.6 0.180 Cod 25–28 1978–2014 1,2,3,4 7026 79.2 Lat:Long, Lat:Year, Long:Year, Depth:Year, Depth, Year, Quarter 34.6 0.339 Flounder 24–25 1988–2014 1,3,4 4706 57.4 Lat:Long, Lat:Year, Long:Year, Depth:Year, Depth, Year, Quarter 22.8 0.218 Flounder 26&28 1978–2014 1,2,3,4 2921 65.7 Lat:Long, Lat:Year, Long:Year, Depth:Year, Year, Quarter 31.7 0.301 Stocks . Years . Quarters . n . df . Variables retained . Dev% . Adj-R2 . Cod 24 1988–2014 1,4 1979 40.4 Lat:Long, Depth:Year, Year, Quarter 19.6 0.180 Cod 25–28 1978–2014 1,2,3,4 7026 79.2 Lat:Long, Lat:Year, Long:Year, Depth:Year, Depth, Year, Quarter 34.6 0.339 Flounder 24–25 1988–2014 1,3,4 4706 57.4 Lat:Long, Lat:Year, Long:Year, Depth:Year, Depth, Year, Quarter 22.8 0.218 Flounder 26&28 1978–2014 1,2,3,4 2921 65.7 Lat:Long, Lat:Year, Long:Year, Depth:Year, Year, Quarter 31.7 0.301 The variables retained in the final models are indicated; n  =  numbers of hauls used in the models; df  =  degrees of freedom; Dev%  =  explained deviance; Adj-R2  =  adjusted R2. Time series of the estimated Lmax for the cod and flounder stocks are presented in Figure 3. Cod in SD 24 (Figure 3a) shows a decrease in Lmax from around 64 cm in 1988 to around 49 cm in 2014. Cod in SDs 25–28 (Figure 3b) shows the highest Lmax of approximately 77 cm in 1983–1985, then the Lmax decreased steadily down to around 40 cm in 2014. However, the Lmax in SD 25 remained relatively stable between the mid-1990s and late 2000s before dropping afterwards down to 47 cm. In SDs 26–28, Lmax declined continuously throughout the time period analysed down to 38 cm in 2014. For flounder in SDs 24–25 (Figure 3c) Lmax fluctuated between 33.5 and 36 cm through the entire time series. In SDs 26 & 28 (Figure 3d) flounder Lmax increased 10% from the beginning of the time series (1978) until 1994 where it reached the maximum value of approximately 37.5 cm, and then has decreased steadily down to around 33 cm in 2014. However, Lmax was lower in SD 26 than in SD 28 before the mid-1990s, whereas afterwards the spatial difference was reversed with Lmax lower in SD 28 than in SD 26. In particular, Lmax of flounder in SD 26 was lower than 30 cm at the beginning of the time series, then increased until reaching a maximum of around 37.5 cm in 1994 and in the last part of the time series decreased down to around 33 cm. In SD 28, on the other hand, Lmax of flounder increased from around 37.5 cm in 1978 to around 39.5 cm in 1985 and then decreased to around 28 cm in 2014. Figure 3 Open in new tabDownload slide Estimated average yearly Lmax (cm) predicted by the model for (a) cod in SD 24, (b) cod in SDs 25–28, (c) flounder in SDs 24–25 and (d) flounder in SDs 26 & 28. Figure 3 Open in new tabDownload slide Estimated average yearly Lmax (cm) predicted by the model for (a) cod in SD 24, (b) cod in SDs 25–28, (c) flounder in SDs 24–25 and (d) flounder in SDs 26 & 28. Trends in exploitation rate Cod in SD 24 was subject to a fairly constant exploitation rate throughout the time series (Figure 4a). For cod in SDs 25–28, the exploitation rate continuously increased from the beginning of the time series until reaching a maximum in the year 2000 and then decreased reaching a minimum value in 2008. In the last 6 years, the exploitation rate slightly increased (Figure 4b). For flounder in SDs 24–25, the exploitation rate strongly increased from the late 1980s up to late 1990s, then fluctuated around the same level and then decreased from 2005 onwards (Figure 4c). In SDs 26 & 28 (Figure 4d), the exploitation rate was less dynamic than in SDs 24–25 with the exception of one peak in the late 1980s and one in the last 2 years of the time series. However, the exploitation rate in SD 28 was almost constant compared to SD 26 that presented an important peak in the late 80s. Figure 4 Open in new tabDownload slide Estimated average yearly exploitation rate (catches/CPUE) for (a) cod in SD 24 and (b) cod in SDs 25–28, (c) (landings/CPUE) flounder in SDs 24–25 and (d) flounder in SDs 26 & 28 (for the whole stock and separately for SD 26 and SD 28). Figure 4 Open in new tabDownload slide Estimated average yearly exploitation rate (catches/CPUE) for (a) cod in SD 24 and (b) cod in SDs 25–28, (c) (landings/CPUE) flounder in SDs 24–25 and (d) flounder in SDs 26 & 28 (for the whole stock and separately for SD 26 and SD 28). Discussion In this study, we reconstructed the trends in CPUE and Lmax of two cod and two flounder stocks in the Baltic Sea by analysing an extensive and unique dataset from scientific trawl surveys. The primary and most relevant steps that allowed us to perform these analyses were the collection of modern and historical trawl survey data complete with gear geometries and the subsequent standardization of the data. The standardized time series of CPUE and Lmax provide unprecedented opportunities for utilizing an impressive amount of data collected during the past 40 years in the Baltic Sea. Cod The CPUE time series of the Eastern Baltic cod stock we have produced, closely resembles the spawning stock biomass (trend of the latest accepted analytical stock assessment (ICES, 2013). The biomass of this cod stock had a major increase in the late 1970s and beginning of the 1980s and the spatial distribution of the stock was the widest ever recorded, with spill-over in areas where cod normally do not occur, such as the Gulf of Riga (SD 28-1) and the Bothnian Sea (SD 30; Casini et al., 2012; Casini, 2013). When the cod stock crashed during the mid-1980s, it started to contract to the southern areas and especially to SD 25 (Eero et al., 2012). The results of our model also revealed this spatiotemporal change, showing that in the early 1980s, the CPUEs in SD 25 and SDs 26–28 were similar, whereas after the cod crash in the early 1990s, the CPUE in SD 25 has been twice than in SDs 26-28. Since the mid-2000s, the CPUE has generally increased but mainly in SD 25 (∼60% of the CPUE maximum in 1981), while the persisting low CPUE in SDs 26–28 indicates that cod has not yet succeeded in re-expanding its distribution into more northern areas. After the late 2000s, a drop in CPUE has occurred picturing a current situation with very low spawning population size. The temporal dynamics in cod abundance have been historically attributed to the concomitant effects of changes in fishing pressure, seal predation and hydrological conditions acting on recruitment (Eero et al., 2011). Notably, the lack of cod recovery and re-expansion of its distribution since the early 1990s can be attributable to persisting high fishing pressure, low body condition, decrease in suitable spawning and feeding areas due to oxygen deficiency in the northern areas and loss of subpopulations (Möllmann et al., 2011; Eero et al., 2012, 2015; ICES, 2015b). The Lmax of Eastern Baltic cod showed a constant decline from the mid-1980s onwards, which is in line with the findings by Svedäng and Hornborg (2014) who evidenced a decrease in the asymptotic length for this stock between 1991 and 2014. Our analysis however, extending back to the late 1970s, was able to reveal that the decrease in Lmax started already in the mid-1980s, concomitant with the stock collapse. The drop in Lmax during the past 30 years was probably caused by a mix of excessive fishing pressure and changes in growth. Fishing mortality has been high, far above safe reference points, since the late 1980s (ICES, 2013). This could have caused the drop in Lmax due to the selective nature of the fishery, targeting and therefore selectively removing the largest and most valuable fish (Vainikka et al., 2009). However, the decrease in Lmax could also be due to a reduction in individual growth rates that could be linked to food shortage, physiological responses to increased hypoxic areas and/or density-dependence (Eero et al., 2012; Svedäng and Hornborg, 2014; ICES, 2015c; Casini et al., 2016). The trends in Lmax of SD 25 and SDs 26–28 were almost identical until the late 90s, when they diverged: SDs 26–28 showed a continuous decline while in SD 25 Lmax showed a relatively stable pattern until 2008 followed by a steep decline. Beside spatial heterogeneity in the fishery (ICES, 2015a), food availability (Gårdmark et al., 2015) and hydrological conditions (Casini et al., 2012), we speculate that the spatial differences in Lmax during the past 15 years were also related to the absence of suitable spawning areas in SDs 26–28 causing a higher concentration of mature and larger fish in SD 25. Considering its ecological and economic relevance, explaining the continuous decrease in cod Lmax during the past 30 years should be a priority for future investigations. The dynamics of CPUE and Lmax in SD 24 (eastern part of the Western Baltic stock) resemble those of the Eastern Baltic stock. In fact, the peaks of CPUEs in SD 24 occurred in the same years of high CPUE values in SD 25 with the exception of the last peak in mid 2000s that seems to occur a couple of years earlier in SD 24. The slight asynchrony of the last high CPUE value could be explained by the use of different models to predict the trends in the two different areas that could have slightly different smoothing parameters. Also Lmax showed a striking synchrony between SD 24 and SD 25. In SD 24 and SD 25, the hydrological conditions, management, and biology of the two stocks are different (Hüssy, 2011; ICES, 2015a) and we hypothesize that the synchrony in the dynamics of the two stocks is not caused by common drivers. We therefore conclude that these results furnish evidence of the cod spill-over from SD 25 to SD 24 and of the occurrence of mixing between the Eastern and Western Baltic cod stocks in SD 24, especially in periods of high abundances in SD 25 (ICES, 2015a; Hüssy et al., 2016). Flounder The reconstructed time series of flounder CPUE in SDs 24–25 shows that the population in this area was more abundant in the late 1980s compared to the current situation, while the stock index of abundance that have been used so far in stock assessment and advice shows only an increase of abundance due to the shorter time series used (ICES, 2015a). The limited amount of years in the assessment time series could therefore lead to an overoptimistic view of the stock status. The flounder in this area suffered a drastic decrease in abundance between 1920s and 1940s presumably caused by intense fishery (Molander, 1955). During the same time period, the mean size at age and size at maturity increased and Molander (1938) suggested that this was due to the relaxation of the earlier density dependent limitation in growth. Lmax in our study, however, showed no apparent correlation with the fluctuations in abundance suggesting that other mechanisms were also involved in the variations of Lmax in SDs 24–25. The results of the model of the CPUE of flounder in SDs 26 & 28 show that the stock abundance was high at the end of the 1970s and then crashed, reaching its minimum around the mid-1980s. This high abundance and consequent collapse of this flounder stock has never been shown before in the literature but it is known that in the 1980s the flounder in SD 28 reached an extremely low level of abundance and a fishing ban on the specialized flounder fishery was enforced by the Soviet Union (D. Ustups, pers. comm.). If we take into account the two areas separately, the difference between the dynamics in SD 26 and SD 28 is striking; while the part of the stock in SD 26 shows a less variable CPUE time series, the flounder in SD 28 reveals strong fluctuations driving the trend of the whole stock. One reason for the fluctuation in stock size in SD 28 might be the variations in available reproductive volume. Ustups et al. (2013) found a strong relationship between reproductive volume (determined by salinity and oxygen) in the Gotland deep and the subsequent larval production in the area. The decline of flounder in the northernmost Baltic (SDs 29 and 32) is speculated to depend on environmental change such as pollution and eutrophication but also changes in salinity might be important (Jokinen et al., 2015). Salinity is a limiting factor for reproduction of marine teleosts like flounder and cod, and salinity has earlier been shown to be related to flounder abundance in the central Baltic (SD 27; Olsson et al., 2012) and Gulf of Finland (SD 32; Ojaveer et al., 1985; Ojaveer and Kalejs, 2005). SD 28 might be more sensitive to saltwater inflow as it is on the margin for successful reproduction of flounder (10 PSU for offshore pelagic eggs and 7 PSU for coastal demersal eggs; Nissling et al., 2002) potentially explaining the more dynamic CPUE time series in this SD. Moreover, it is possible that the dissimilar trends between the areas are driven by different causes; SD 28 could potentially be mostly driven by environmental factors such as salinity, while SD 26 by higher fishing pressure compared to SD 28 as indicated by higher landings in SD 26 compared to SD 28 in the analysed timeframe (ICES, 2013; ICES, 2015a). Finally, the different CPUE trends of flounder in SDs 24–25 and in SDs 26 & 28 suggest that the degree of mixing between the two stocks is low. The Lmax for the flounder stock in SDs 26 & 28 overall decreased in the last 20 years. The results obtained for SD 26 and SD 28 separately are interesting, showing a steep decline in Lmax in SD 28 throughout the whole time series and a relatively dome-shaped trend in SD 26. The interpretation of these results is quite complicated especially in SD 28 since this area is occupied also by another stock of flounder, the coastal spawning flounder with demersal eggs (Nissling et al., 2002; Florin and Höglund, 2008). This spawning type mainly resides in SDs 27 & 29–32, but it is known to occur also in SD 28, although at low densities (ICES, 2014b). The demersal spawning flounder is known to have smaller body size (Nissling and Dahlman, 2010), and thus we cannot exclude that the decrease in Lmax in SD 28 is partly due to a change of the proportion of the two flounder ecotypes in this area. Potential interactions between cod and flounder The flounder stock started to decline during the rapid increase of the cod stock in the late 1970s–early 1980s and after the cod stock collapsed the flounder in SDs 26 & 28 began to recover. This potential negative link between cod and flounder dynamics has not been studied before, even though large cod can feed on flounder (Almqvist et al., 2010; ICES, 2016) and the two species potentially compete for benthic prey (Arntz and Finger, 1981; Gjøsæter, 1988). Only Persson (1981) speculated on the fact that the low abundances of cod in the southern part of the Baltic at the beginning of the 20th century could have been caused by the effects of high competition for benthic preys between young cod and flatfishes, especially the dab (Limanda limanda, Pleuronectidae). Studies performed in some areas of the North Atlantic, in the Bering Sea and on the Scottish coast have shown that gadoid predation on juvenile flatfishes are quite widespread (Bailey, 1994 and references therein; Ellis and Gibson, 1995). In the Bering Sea predation of arrowtooth flounder (Atheresthes stomias, Pleuronectidae) on juvenile pollock (Theragra chalcogramma, Gadidae) has been proposed to affect recruitment success of the gadoid population (Hunsicker et al., 2013). At Georges Bank studies have shown competition between haddock (Melanogrammus aeglefinus, Gadidae) and yellowtail flounder (Limanda ferruginea, Pleuronectidae) (Link et al., 2005 and reference therein). Similar ecological links between cod and flounder may play a role in their population dynamics in the Baltic Sea. Changes in the demersal fish community The decline of Lmax evidenced in all the stocks considered in this study raise concern about the ecosystem state of the Baltic Sea. On a population level, the presence of large and old individuals plays a role in population resilience: e.g. large and old fishes usually are characterized by high fecundity and fitness, large eggs, a prolonged spawning period and are considered as reservoirs of desirable genes (Vallin and Nissling, 2000; Froese, 2004; Nissling and Dahlman, 2010; Hixon et al., 2014). On a community level, our results show that the fish demersal component is becoming progressively dominated by small individuals. Similar results have been shown for the Baltic pelagic community (Oesterwind et al., 2013). These structural changes in the Baltic fish communities may reflect changes in the trophic interactions within the community and could be caused by disproportionate high fishing mortality on larger individuals and/or by changes in the environmental conditions affecting growth. Concluding remarks The delta modelling framework used to produce the CPUE time series for cod and flounder captured the known dynamics of cod in SD 25–28. Still, our models suffer from a number of limitations and could be improved. The most important issue is related to the assumed independence between the presence/absence and the abundance model (Thorson and Ward, 2013). The inclusion of environmental variables such as salinity, oxygen and temperature in the presence/absence models could potentially increase the model fit and the predictive power of the binomial model. The collection and standardization of historical survey data are important in the Baltic, with its changes in salinity, temperature, oxygen conditions as well as eutrophication, and fishing effort (Niiranen et al., 2013). An extraordinary amount of data has been collected through time by the states bordering the Baltic Sea during nationally and internationally trawl surveys. Our standardization of these survey data and the subsequent modelling of the time series constitute a powerful tool that improves our knowledge on fished populations in the Baltic Sea, thus promoting long-term sustainable use of these marine resources. The long time series of CPUE and Lmax presented here are a step forward in the knowledge of the dynamics of the four stocks considered. In future analyses, the standardization of CPUE might be considered to be integrated in stock assessments as suggested by Maunder (2001). For the cod stocks, these stock assessments already exists, although the assessment for Eastern Baltic cod stock has been rejected in the last years and the fisheries-independent time series used before have been shorter. For flounder, which is equally important for fisheries and has a central role in the ecosystem (Florin et al., 2013; Östman et al., 2013; ICES, 2016) analytical stock assessments are not available. The CPUE time series and size-based indicators we developed hopefully help the conservation and management of these stocks in the Baltic Sea. Supplementary data Supplementary material is available at the ICESJMS online version of the manuscript. Acknowledgements We thank Valerio Bartolino for his valuable comments on the modelling. We are also grateful to all the colleagues and experts that helped us gathering the gear geometries for the standardization. Two anonymous reviewers provided constructive comments on a previous version of the manuscript. This work was financed by the BONUS INSPIRE project supported by the Joint Baltic Sea Research and Development Programme BONUS (Art 185), funded jointly by the EU and the Swedish Research Council Formas (Sweden) and the State Education Development Agency (Latvia). References Almqvist G. , Strandmark A. K., Appelberg M. 2010 . Has the invasive round goby caused new links in Baltic food webs? Environmental Biology of Fishes , 89 : 79 – 93 . Google Scholar Crossref Search ADS WorldCat Arntz W. E. , Finger I. 1981 . Demersal fish in the western Baltic: their feeding relations, food coincidence and food selection . ICES CM/J , 6 : 1 – 28 . OpenURL Placeholder Text WorldCat Bailey K. M. 1994 . Predation on juvenile flatfish and recruitment variability . Netherlands Journal of Sea Research , 32 : 175 – 189 . Google Scholar Crossref Search ADS WorldCat Barry S. C. , Welsh A. H. 2002 . Generalized additive modelling and zero inflated count data . Ecological Modelling , 157 : 179 – 188 . Google Scholar Crossref Search ADS WorldCat Beare D. J. , Needle C. L., Burns F., Reid D. G. 2005 . Using survey data independently from commercial data in stock assessment: an example using haddock in ICES division VIa . ICES Journal of Marine Science , 62 : 996 – 1005 . Google Scholar Crossref Search ADS WorldCat Berg C. W. , Nielsen A., Kristensen K. 2014 . Evaluation of alternative age-based method for estimating relative abundance from survey data in relation to assessment models . Fisheries Research , 151 : 91 – 99 . Google Scholar Crossref Search ADS WorldCat Blanchard J. L. , Dulvy N. K., Jennings S., Ellis J. R., Pinnegar J. K., Tidd A., Kell L. T. 2005 . Do climate and fishing influence size-based indicators of Celtic Sea fish community structure? ICES Journal of Marine Science , 62 : 405 – 411 . Google Scholar Crossref Search ADS WorldCat Cardinale M. , Bartolino V., Llope M., Maiorano L., Sköld M., Hagberg J. 2011 . Historical spatial baselines in conservation and management of marine resources . Fish and Fisheries , 12 : 289 – 298 . Google Scholar Crossref Search ADS WorldCat Cardinale M. , Linder L., Bartolino V., Maiorano L., Casini M. 2009 . Conservation value of historical data: reconstructing stock dynamics of turbot during the last century in the Kattegat-Skagerrak . Marine Ecology Progress Series , 386 : 197 – 206 . Google Scholar Crossref Search ADS WorldCat Casini M. 2013 . Spatio-temporal ecosystem shifts in the Baltic Sea: top-down control and reversibility potential . Advances in Environmental Research , 28 : 149 – 167 . OpenURL Placeholder Text WorldCat Casini M. , Blenckner T., Möllmann C., Gårdmark A., Lindegren M., Llope M., Kornilovs G., et al. 2012 . Predator transitory spillover induces trophic cascades in ecological sinks . Proceedings of the National Academy of Sciences , 109 : 8185 – 8189 . Google Scholar Crossref Search ADS WorldCat Casini M. , Käll F., Hansson M., Plikshs M., Baranova T., Karlsson O., Lundstrom K., et al. 2016 . Hypoxic areas, density dependence and food limitation drive the body condition of a heavily exploited marine fish predator . Royal Society Open Science , 3 : 160416. Google Scholar Crossref Search ADS PubMed WorldCat Casini M. , Lövgren J., Hjelm J., Cardinale M., Molinero J. C., Kornilovs G. 2008 . Multi-level trophic cascades in a heavily exploited open marine ecosystem . Proceedings of the Royal Society B , 275 : 1793 – 1801 . Google Scholar Crossref Search ADS PubMed WorldCat Cosgrove R. , Sheridan M., Minto C., Officer R. 2014 . Application of finite mixture models to catch rate standardization better represents data distribution and fleet behavior . Fisheries Research , 153 : 83 – 88 . Google Scholar Crossref Search ADS WorldCat Eero M. , Hjelm J., Behrens J., Buchmann K., Cardinale M., Casini M., Gasyukov P., et al. 2015 . Eastern Baltic cod in distress: biological changes and challenges for stock assessment . ICES Journal of Marine Science , 72 : 2180 – 2186 . Google Scholar Crossref Search ADS WorldCat Eero M. , MacKenzie B. R., Köster F. W., Gislason H. 2011 . Multi-decadal responses of a cod (Gadus morhua) population to human-induced trophic changes, fishing, and climate . Ecological Applications , 21 : 214 – 226 . Google Scholar Crossref Search ADS PubMed WorldCat Eero M. , Vinther M., Haslob H., Huwer B., Casini M., Storr-Poulsen M., Köster F. W. 2012 . Spatial management of marine resources can enhance the recovery of predators and avoid local depletion of forage fish . Conservation Letters , 5 : 486 – 492 . Google Scholar Crossref Search ADS WorldCat Ellis T. , Gibson R. N. 1995 . Size-selective predation of 0-group flatfishes in a Scottish coastal nursery ground . Marine Ecology Progress Series , 127 : 27 – 37 . Google Scholar Crossref Search ADS WorldCat EU-COM . 2008 . Directive 2008/56/EC of the European parliament and of the council of 17 June 2008 establishing a framework for community action in the field of marine environmental policy (Marine Strategy Framework Directive). http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2008:16 4:0019:0 0 4 0:EN: PDF. Francis R. I. C. C. 2011 . Data weighting in statistical fisheries stock assessment models . Canadian Journal of Fisheries and Aquatic Sciences , 68 : 1124 – 1138 . Google Scholar Crossref Search ADS WorldCat Florin A. B. , Bergström U., Ustups D., Lundström K., Jonsson P. R. 2013 . Effects of a large northern European no-take zone on flatfish populations . Journal of Fish Biology , 83 : 939 – 962 . Google Scholar PubMed OpenURL Placeholder Text WorldCat Florin A. B. , Höglund J. 2008 . Population structure of flounder (Platichthys flesus) in the Baltic Sea: differences among demersal and pelagic spawners . Heredity , 101 : 27 – 38 . Google Scholar Crossref Search ADS PubMed WorldCat Froese R. 2004 . Keep it simple: three indicators to deal with overfishing . Fish and Fisheries , 5 : 86 – 91 . Google Scholar Crossref Search ADS WorldCat Galbraith R. D. , Rice A. 2005 . An introduction to commercial fishing gear and methods used in Scotland . Scottish Fisheries Information Pamphlet , 25 : 44 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Gårdmark A. , Casini M., Huss M., van Leeuwen A., Hjelm J., Persson L., de Roos A. M. 2015 . Regime shifts in exploited marine food webs: detecting mechanisms underlying alternative stable states using size-structured community dynamics theory . Philosophical Transactions of the Royal Society B , 370 : 20130262. Google Scholar Crossref Search ADS WorldCat Gjøsæter J. 1988 . Competition for food and predator-prey relationships among young cod (Gadus Morhua) and some other fish from shallow waters . Flødevigen Rapportserie , 1 : 1 – 15 . OpenURL Placeholder Text WorldCat Grüss A. , Drexler M., Ainsworth C. H. 2014 . Using delta generalized additive models to produce distribution maps for spatially explicit ecosystem models . Fisheries Research , 159 : 11 – 24 . Google Scholar Crossref Search ADS WorldCat Harley S. J. , Myers R. A. 2001 . Hierarchical Bayesian models of length-specific catchability of research trawl survey . Canadian Journal of Fisheries and Aquatic Sciences , 58 : 1569 – 1584 . Google Scholar Crossref Search ADS WorldCat Hastie T. , Tibshirani R. 1990 . Generalized Additive Models . Chapman and Hall , London . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Heino M. , Baulier L., Boukal D. S., Ernande B., Johnston F. D., Mollet F. M., Pardoe H., et al. 2013 . Can fisheries-induced evolution shift reference points for fisheries management? ICES Journal of Marine Science , 70 : 707 – 721 . Google Scholar Crossref Search ADS WorldCat Hixon M. A. , Johnson D. W., Sogard S. M. 2014 . BOFFFFs: on the importance of conserving old-growth age structure in fishery populations . ICES Journal of Marine Science , 71 : 2171 – 2185 . Google Scholar Crossref Search ADS WorldCat Hunsicker M. E. , Ciannelli L., Bailey K. M., Zador S., Stige L. C. 2013 . Climate and Demography Dictate the Strength of Predator-Prey Overlap in a Subarctic Marine Ecosystem . PLoS ONE , 8 : e66025 . Google Scholar Crossref Search ADS PubMed WorldCat Hüssy K. 2011 . Review of western Baltic cod (Gadus morhua) recruitment dynamics . ICES Journal of Marine Science , 68 : 1459 – 1471 . Google Scholar Crossref Search ADS WorldCat Hüssy K. , Hinrichsen H. H., Eero M., Mosegaard H., Hemmer-Hansen J., Lehmann A., Lundgaard L. S. 2016 . Spatio-temporal trends in stock mixing of eastern and western Baltic cod in the Arkona Basin and the implications for recruitment . ICES Journal of Marine Science , 73 : 293 – 303 . Google Scholar Crossref Search ADS WorldCat ICES . 2010 . Report of the ICES/HELCOM Workshop on Flatfish in the Baltic Sea (WKFLABA), 8 - 11 November 2010, Öregrund, Sweden. ICES CM 2010/ACOM:68. 85 pp. ICES . 2012 . Marine Strategy Framework Directive - Descriptor 3+, ICES CM 2012/ACOM:62. 173 pp. ICES . 2013 . Report of the Baltic Fisheries Assessment Working Group (WGBFAS), 10–17 April 2013, ICES Headquarters, Copenhagen. ICES CM 2013/ACOM:10. 747 pp. ICES . 2014a . Report of the Baltic International Fish Survey Working Group (WGBIFS), 24–28 March 2014, Gdynia, Poland. ICES CM 2014/SSGESST:13. 527 pp. ICES . 2014b . Report of the Benchmark Workshop on Baltic Flatfish Stocks (WKBALFLAT), 27–31 January 2014, Copenhagen, Denmark. ICES CM 2014/ACOM:39. 320 pp. ICES . 2015a . Report of the Baltic Fisheries Assessment Working Group (WGBFAS), 14–21 April 2015, ICES HQ, Copenhagen, Denmark. ICES CM 2015/ACOM:10. 806 pp. ICES . 2015b . Report of the Benchmark Workshop on Baltic Cod Stocks (WKBALTCOD), 2–6 March 2015, Rostock, Germany. ICES CM 2015/ACOM:35. 172 pp. ICES . 2015c . Cod (Gadus morhua) eastern Baltic stock in Subdivisions 25–32 (Eastern Baltic Sea) and Subdivision 24. In: Report of the ICES Advisory Committee 2015. ICES Advice, 2015. Book 8. ICES. 2016 . Report of the Workshop on Spatial Analyses for the Baltic Sea (WKSPATIAL), 3-6 November 2015, Rome, Italy. ICES CM 2015/SSGIEA:13. 37 pp. Jokinen H. , Wennhage H., Lappalainen A., Ådjers K., Rask M., Norkko A. 2015 . Decline of flounder (Platichthys flesus (L.)) at the margin of the species' distribution range . Journal of Sea Research , 105 : 1 – 9 . Google Scholar Crossref Search ADS WorldCat Lauria V. , Vaz S., Martin C. S., Mackinson S., Carpentier A. 2011 . What influences European plaice (Pleuronectes platessa) distribution in the eastern English Channel? Using habitat modelling and GIS to predict habitat utilization . ICES Journal of Marine Science , 68 : 1500 – 1510 . Google Scholar Crossref Search ADS WorldCat Lindegren M. , Möllmann C., Nielsen A., Stenseth N. C. 2009 . Preventing the collapse of the Baltic cod stock through an ecosystem-based management approach . Proceedings of the National Academy of Sciences , 106 : 14722 – 14727 . Google Scholar Crossref Search ADS WorldCat Link J. S. , Fogarty M. J., Langton R. W. 2005 . The trophic ecology of flatfishes. In Flatfishes: Biology and Exploitation . Ed. by R. N. Gibson. Blackwell Publishing , Oxford , pp. 185 – 204 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Loots C. , Vaz S., Planque B., Koubbi P. 2010 . What controls the spatial distribution of the North Sea plaice spawning population? Confronting ecological hypotheses through a model selection framework . ICES Journal of Marine Science , 67 : 244 – 257 . Google Scholar Crossref Search ADS WorldCat Maschner H. D. G. , Betts M. W., Reedy-Maschner K., Trites A. W. 2008 . A 4500 year time series of Pacific Cod (Gadus macrocephalus): archaeology, regime shifts, and sustainable fisheries . Fishery Bulletin , 106 : 386 – 394 . OpenURL Placeholder Text WorldCat Maunder M. N. 2001 . A general framework for integrating the standardization of catch per unit of effort into stock assessment models . Canadian Journal of Fisheries and Aquatic Sciences , 58 : 795 – 803 . Google Scholar Crossref Search ADS WorldCat Maunder M. N. , Punt A. E. 2004 . Standardizing catch and effort data: a review of recent approaches . Fisheries Research , 70 : 141 – 159 . Google Scholar Crossref Search ADS WorldCat Molander A. R. 1938 . Investigations into the growth-rates of the common dab and the flounder in the southern Baltic . Rapports Et Procès-Verbaux Des Réunions, Conseil International Pour L’Exploration De La Mer , 108 : 90 – 101 . OpenURL Placeholder Text WorldCat Molander A. R. 1955 . Swedish investigations on plaice and flounder in the southern Baltic . Journal Du Conseil, Conseil International Pour L’Exploration De La Mer , 21 : 25 – 43 . Google Scholar Crossref Search ADS WorldCat Möllmann C. , Blenckner T., Casini M., Gårdmark A., Lindegren M. 2011 . Beauty is in the eye of the beholder: management of Baltic cod stock requires an ecosystem approach . Marine Ecology Progress Series , 431 : 293 – 297 . Google Scholar Crossref Search ADS WorldCat Nicholson M. D. , Jennings S. 2004 . Testing candidate indicators to support ecosystem-based management: the power of monitoring surveys to detect temporal trends in fish community metrics . ICES Journal of Marine Science , 61 : 35 – 42 . Google Scholar Crossref Search ADS WorldCat Niiranen S. , Yletyinen J., Tomczak M. T., Blenckner T., Hjerne O., MacKenzie B. R., Müller-Karulis B., et al. 2013 . Combined effects of global climate change and regional ecosystem drivers on an exploited marine food web . Global Change Biology , 19 : 3327 – 3342 . Google Scholar PubMed OpenURL Placeholder Text WorldCat Nissling A. , Dahlman G. 2010 . Fecundity of flounder, Pleuronectes flesus, in the Baltic Sea – reproductive strategies in two sympatric populations . Journal of Sea Research , 64 : 190 – 198 . Google Scholar Crossref Search ADS WorldCat Nissling A. , Westin L., Hjerne O. 2002 . Reproductive success in relation to salinity for three flatfish species, dab (Limanda limanda), plaice (Pleuronectes platessa), and flounder (Pleuronectes flesus), in the brackish water Baltic Sea . ICES Journal of Marine Science , 59 : 93 – 108 . Google Scholar Crossref Search ADS WorldCat Ojaveer E. , Kalejs M. 2005 . The impact of climate change on the adaptation of marine fish in the Baltic Sea . ICES Journal of Marine Science , 62 : 1492 – 1500 . Google Scholar Crossref Search ADS WorldCat Ojaveer E. , Kalejs M., Aps R., Labaika I., Vitins M. 1985 . The impact of recent environmental change on the main commercial fish stocks in the Gulf of Finland . Finnish Fisheries Research , 6 : 1 – 14 . OpenURL Placeholder Text WorldCat Oesterwind D. , Psuty I., Pachur M., Von Dorrien C., Lejk A., Casini M., Larson N. 2013 . Proportion of large fish in the community. HELCOM Core Indicator Report. Olsson J. , Bergström L., Gårdmark A. 2012 . Abiotic drivers of coastal fish community change during four decades in the Baltic Sea . ICES Journal of Marine Science , 69 : 961 – 970 . Google Scholar Crossref Search ADS WorldCat Parra H. E. , Pham C. K., Menezes G. M., Rosa A., Tempera F., Morato T. 2016 . Predictive modelling of deep-sea fish distribution in the Azores. Deep-Sea Research II. http://dx.doi.org/10.1016/j.dsr2.2016.01.004i. Pauly D. 1995 . Anecdotes and the shifting baseline syndrome of fisheries . Trends in Ecology & Evolution , 10 : 430. Google Scholar Crossref Search ADS PubMed WorldCat Persson L. E. 1981 . Were macrobenthic changes induced by thinning out of flatfish stocks in the Baltic proper? Ophelia , 20 : 137 – 152 . Google Scholar Crossref Search ADS WorldCat Piet G. J. , Jennings S. 2004 . Response of potential fish community indicators to fishing . ICES Journal of Marine Science , 62 : 214 – 225 . Google Scholar Crossref Search ADS WorldCat Pinnegar J. K. , Engelhard G. H. 2008 . The “shifting baseline” phenomenon: a global perspective . Reviews in Fish Biology and Fisheries , 18 : 1 – 16 . Google Scholar Crossref Search ADS WorldCat Probst W. N. , Kloppmann M., Kraus G. 2013 . Indicator-based status assessment of commercial fish species in the North Sea according to the EU Marine Strategy Framework Directive (MSFD) . ICES Journal of Marine Science , 70 : 694 – 706 . Google Scholar Crossref Search ADS WorldCat R Core Team . 2015 . R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/. Rijnsdorp A. D. , Van Leeuwen P. I., Daan N., Heessen H. J. L. 1996 . Changes in abundance of demersal fish in the North Sea between 1906–1909 and 1990–1995 . ICES Journal of Marine Science , 53 : 1054 – 1062 . Google Scholar Crossref Search ADS WorldCat Shin Y. J. , Rochet M. J., Jennings S., Field J. G., Gislason H. 2005 . Using size-based indicators to evaluate the ecosystem effects of fishing . ICES Journal of Marine Science , 62 : 384 – 396 . Google Scholar Crossref Search ADS WorldCat Stefánsson G. 1996 . Analysis of groundfish survey abundance data: combining the GLM and delta approaches . ICES Journal of Marine Science , 53 : 577 – 588 . Google Scholar Crossref Search ADS WorldCat Svedäng H. , Hornborg S. 2014 . Selective fishing induces density-dependent growth . Nature Communications , 5 : 4152. Google Scholar Crossref Search ADS PubMed WorldCat Thorson J. T. , Shelton A. O., Ward E. J., Skaug H. J. 2015 . Geostatistical delta-generalized linear mixed models improve precision for estimated abundance indices for West Coast groundfishes . ICES Journal of Marine Science , 72 : 1297 – 1310 . Google Scholar Crossref Search ADS WorldCat Thorson J. T. , Ward E. J. 2013 . Accounting for space–time interactions in index standardization models . Fisheries Research , 147 : 426 – 433 . Google Scholar Crossref Search ADS WorldCat Thorson J. T. , Ward E. J. 2014 . Accounting for vessel effects when standardizing catch rates from cooperative surveys . Fisheries Research , 155 : 168 – 176 . Google Scholar Crossref Search ADS WorldCat Ustups D. , Müller-Karulis B., Bergström U., Makarchouk A., Sics I. 2013 . The influence of environmental conditions on early life stages of flounder (Platichthys flesus) in the central Baltic Sea . Journal of Sea Research , 75 : 77 – 84 . Google Scholar Crossref Search ADS WorldCat Vainikka A. , Gårdmark A., Bland B., Hjelm J. 2009 . Two- and three-dimensional maturation reaction norms for the eastern Baltic cod, Gadus morhua . ICES Journal of Marine Science , 66 : 248 – 257 . Google Scholar Crossref Search ADS WorldCat Vallin L. , Nissling A. 2000 . Maternal effects on egg size and egg buoyancy of Baltic cod, Gadus morhua: implications for stock structure effects on recruitment . Fisheries Research , 49 : 21 – 37 . Google Scholar Crossref Search ADS WorldCat Walsh M. R. , Munch S. B., Chiba S., Conover D. O. 2006 . Maladaptive changes in multiple traits caused by fishing: impediments to population recovery . Ecology Letters , 9 : 142. Google Scholar Crossref Search ADS PubMed WorldCat Wood S. N. 2003 . Thin-plate regression splines . Journal of the Royal Statistical Society (B) , 65 : 95 – 114 . Google Scholar Crossref Search ADS WorldCat Wood S. N. 2006 . Generalized Additive Models: An introduction with R . Chapman and Hall/CRC , Boca Raton, Florida . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Wood S. N. 2011 . Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models . Journal of the Royal Statistical Society (B) , 73 : 3 – 36 . Google Scholar Crossref Search ADS WorldCat Östman Ö. , Boström M. K., Bergström U., Andersson J., Lunneryd S. G. 2013 . Estimating competition between wildlife and humans – a case of cormorants and coastal fisheries in the Baltic Sea . PLoS One , 8 : e83763. doi:10.1371/journal.pone.0083763 Google Scholar Crossref Search ADS PubMed WorldCat © International Council for the Exploration of the Sea 2017. All rights reserved. For Permissions, please email: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Trifonova, Neda; Maxwell, David; Pinnegar, John; Kenny, Andrew; Tucker, Allan

doi: 10.1093/icesjms/fsw231pmid: N/A

Abstract The recent adoption of Bayesian networks (BNs) in ecology provides an opportunity to make advances because complex interactions can be recovered from field data and then used to predict the environmental response to changes in climate and biodiversity. In this study, we use a dynamic BN model with a hidden variable and spatial autocorrelation to explore the future of different fish and zooplankton species, given alternate scenarios, and across spatial scales within the North Sea. For most fish species, we were able to predict a trend of increase or decline in response to change in fisheries catch; however, this varied across the different areas, outlining the importance of trophic interactions and the spatial relationship between neighbouring areas. We were able to predict trends in zooplankton biomass in response to temperature change, with the spatial patterns of these effects varying by species. In contrast, there was high variability in terms of response to productivity changes and consequently knock-on effects on higher level trophic species. Finally, we were able to provide a new data-driven modelling approach that accounts for multispecies associations and interactions and their changes over space and time, which might be beneficial to give strategic advice on potential response of the system to pressure. Introduction The North Sea is a dynamic system, heavily modified by humans and climate. Thus, there is an increasing demand for tools with which to explore alternative hypotheses about ecosystem response to change in pressures (Mackinson and Daskalov, 2007). In this study, we present an approach to explore how species and trophic groups respond to change in human and climate pressures and understand potential trade-offs between such ecosystem components, given a set of alternate scenarios. The North Sea has been exploited for centuries by the surrounding countries and the state of its environment has been altered greatly by human activities (Jennings and Kaiser, 1998). Fishing pressure can change the structure of marine populations and consequently influence the nature of their responses to climate (Planque et al., 2010). However, in late 1990s the EU began a fleet reduction scheme and most recently, the EU Common Fisheries Policy introduced significant changes to how fisheries are to be managed, including a landings obligation and management plans that take account of biological and technical interactions (EC, 2013). The ecosystem-based approach to fisheries management acknowledges that fisheries are part of the environment and cannot be managed in isolation (Cury et al., 2005) and requires recognition of the ecosystem dynamics and structure. One way to understand ecosystem dynamics is to incorporate multispecies information and interactions with both physical and biological components that would reduce uncertainty in predicting the species response to change in fisheries and climate. The biological characteristics of any species stock are dependent upon and shaped over time by its interactions with other species and the rest of the ecosystem (Mackinson and Daskalov, 2007). As such, by using multispecies ecosystem models, the species effects can be quantified across space and over time, under different fisheries exploitation and climate scenarios. Many studies using different techniques have been undertaken to utilize environmental information and provide advice to meet management needs and understand future environmental states (Lewy and Vinther, 2004; Mackinson and Daskalov, 2007; Ulrich et al., 2011; Lynam and Mackinson, 2015). Although, such models incorporate a large percentage of the higher trophic groups, they lack important extrinsic drivers, such as climate variation (e.g. Ecopath with Ecosim in Mackinson and Daskalov, 2007), which is fundamental for interpreting community dynamics. In addition, for such models to be valuable, they would also need to reflect the link between an input that can be managed (fisheries catch) and the response (e.g. change in species biomass), and therefore require an anthropogenic involvement (García-Carreras et al., 2015). Our modelling approach of utilizing multiple associations between species and their environment presents a more comprehensive route to projecting future ecosystem change allowing empirical data to be combined with some existing knowledge to build scenarios that describe possible alternative futures. Predicting species response to ecosystem changes is challenging because of the variability in observations and uncertainty in potential associations. However, machine learning techniques have been proposed to be an appropriate approach with desired properties to address uncertainty in prediction (Uusitalo, 2007). In particular, probabilistic methods such as Bayesian Networks (BNs) provide estimates of the uncertainty associated with predictions, as demonstrated by Fernandes et al. (2010). With the recent adoption of BNs in predictive ecology, few assumptions can be made about the data and complex, spatially varying interactions can be recovered from collected field data, as demonstrated by Trifonova et al. (2015). Such probabilistic models allow predictions to be made across very different platforms and organisms (Smith et al., 2006) through the use of a network structure and inference that allow us to ask “what if” type questions of the data. For example, one could ask, what is the probability of seeing a change in the biomass of cod, given that we have observed a change in the probability distribution of catch and/or herring biomass? Originally, BNs were introduced in the context of bioinformatics research but there has been significant progress in their application to environmental problems (Chen and Pollino, 2012; Uusitalo et al., 2012; Hamilton et al., 2015), to manage fisheries resources (Lee and Rieman, 1997) and for other uses (Olson et al., 1990). As applied in ecology, BNs represent probabilistic dependencies among species and ecosystem factors that influence the variables’ likelihood in an intuitive, graphic form (Jensen, 2001), therefore different expertise can have a quantitative indication of the range of possible scenarios consistent with the data to give strategic advice on potential ecosystem response. The visual nature of BNs can help to communicate modelling results and they allow a variety of perspectives of natural and anthropogenic effects to be represented (Levontin et al., 2011). In this study, we are interested in the characteristics of BNs to demonstrate the effects of change in human and environmental pressures on the forward projections of variables of interest. A dynamic BN model was applied to investigate the consequences of fisheries catch, temperature and primary productivity scenarios on different fish and zooplankton species. Through the developed scenarios, we explore the specific trends of species in response to change in pressures and examine potential trade-offs between the species of interest but also with other trophic groups of species. The approach we are using is a modified version of the model in Trifonova et al. (2015), which uses the functional network approach to predict the dynamics of species groups, accounting for trophic associations and interactions with external stressors and unmeasured hidden effects at spatial and temporal scale. Now, we extend this approach to model individual fish and zooplankton species data further into the future by developing a set of scenarios, accounting for their spatially differentiated biotic and abiotic associations, which are important because species interactions can increase or reduce future changes at different scales, influencing the emergence of winners and losers (Barange et al., 2014). Hence, we aim at predicting species year-to-year variations and understanding their dynamics, which is essential to give strategic advice on potential response of the system to pressure. Methods We used a modelling approach that integrates the functional network approach (combination of known topological features of food webs with quantitative variation in species interactions with their environment and surrounding stressors) with a dynamic BN model. We first modify the model to make future projections of species (and trophic groups). Then, we use the model in combination with alternate scenarios of fisheries catch, temperature and productivity to explore species (and trophic groups) trends in response to change in pressures. Data The analyses are based on the database of the International Bottom Trawl Survey (IBTS, https://datras.ices.dk) for Quarter 1 (January to March), maintained by the International Council for the Exploration of the Sea (ICES) and conducted within ICES areas between 51 and 62° latitude (Figure 1, only areas 1–7 were considered in the study here due to limited quality and consistency of the data on the remaining spatial areas). In the study, catch per unit effort data were extracted for the years: 1983–2015 and converted to biomass (kg/h), using length–weight relationships and summing by species and year (www.fishbase.org). Next, individual fish species were aggregated by summing up the data into the relevant trophic group: pelagics (P), small piscivorous (SP), and large piscivorous and top predators (LP). (FishBase was used as a guidance point). The following fish species were separated as specific variables of interest: cod (Gadus morhua), haddock (Melanogrammus aeglefinus), herring (Clupea harengus), European plaice (Pleuronectes platessa), sole (Solea solea), saithe (Pollachius virens), and whiting (Merlangius merlangus). The species were chosen due to their high commercial importance and contribution to total landings (http://www.ices.dk/marine-data/dataset-collections/Pages/Fish-catch-and-stock-assessment.aspx). We also used biomass data for zooplankton species and data for sea surface temperature (temperature), net primary production (Net PP) and fisheries catch. See the Supplementary Materials, section 1.1 for a detailed description of these variables and their sources. The data were standardized (sample mean removed from each observation, which is then divided by the standard deviation) prior to conducting the modelling experiments but when visualizing the results, we reversed the standardization of the modelled values. Figure 1 Open in new tabDownload slide ICES statistical rectangles within the North Sea (areas 1–7 were used in this study). Source: ICES, Manual for the International Bottom Trawl Surveys. Figure 1 Open in new tabDownload slide ICES statistical rectangles within the North Sea (areas 1–7 were used in this study). Source: ICES, Manual for the International Bottom Trawl Surveys. Bayesian networks Formally, a BN describes the joint distribution (a way of assigning probabilities to every possible outcome over a set of variables, X1…XN) by exploiting conditional independence relationships, represented by a directed acyclic graph (DAG) (Friedman et al., 1999). The conditional probability distribution associated with each variable X encodes the probability of observing its values given the values of its parents, and can be described by a continuous or a discrete distribution. The DAG consists of nodes (or variables) and edges (or links) between the variables. “Parent” nodes are those from which arrows originate and “child” nodes are those to which arrows are pointing. Edges between nodes represent dependence relationships. Here, the observed variable nodes in the network are Gaussian nodes, so we assume continuous distribution with mean mu and covariance Sigma. Each node in the DAG is characterized by a state which can change depending on the state of other nodes and information about those states propagated through the DAG. By using this kind of inference, one can change the state or introduce new data or evidence (change a state or confront the DAG with new data) into the network, apply inference and inspect the posterior distribution (which represents the distributions of the variables given in the observed evidence). Given a graphical structure, BNs naturally perform prediction using inference. Modelling time series is achieved by using an extension of the BN known as the Dynamic Bayesian Network (DBN), where nodes represent variables at particular time slices, Figure 2a (Friedman et al., 1999). The semantics of a DBN can be defined by “unrolling” the two-slice DBN into T time slices (Figure 2b). The parameters for slices t = t, t + 1, … do not change over time, i.e. the model is time invariant which allows unbounded amount of data to be modelled with a finite number of parameters (Murphy, 2002). In the study, DBNs allow us to integrate heterogeneous data at different scales and make robust predictions of the temporal species dynamics under modelled scenario interactions with external stressors. DBNs can model the dynamics of a dataset through the use of a latent or hidden variable (HV). This latent variable is used to model unobserved variables and missing data and can infer some underlying state of the series when applied through an autoregressive link that can capture relationships of a higher order (Murphy, 2001). Specifically, the HV was chosen to most easily reflect complex interdependencies between and among species and their environment that might represent something external to the community, which is not purely constrained within the model structure. Figure 2 Open in new tabDownload slide (a) A two slice DBN (b) the same model unrolled for T = 4 slices. Figure 2 Open in new tabDownload slide (a) A two slice DBN (b) the same model unrolled for T = 4 slices. Model description Here, the modelling approach is a modified version of the hidden spatial dynamic Bayesian network model developed in Trifonova et al. (2015) (we will refer to the model as HSDBN). The model structure represents a potential “end-to-end” ecosystem model of each area’s trophic dynamics by incorporating data driven interactions with some expertise knowledge (known topological features of food webs) on the zooplankton dynamics. This model is an extension of the published model in terms of predicting species data further into the future and modelling individual fish species dynamics under different effects from biotic and abiotic scenarios. In addition to modelling individual fish species, we also model the aggregated species groups: P, SP, and LP to account for the trophic effect in predicting future changes. We incorporate only one HV and instead of a second HV, as originally in Trifonova et al. (2015), we incorporate the observed zooplankton biomass for the North Sea. In addition to the three spatial nodes: P sp., SP sp., and LP sp., we add an additional spatial node (the average biomass of the relevant fish species from the spatial neighbourhood (the three or four nearest neighbours) of the current area) as a parent node to the fish species variable, to account for the effect of spatial autocorrelation. In this way, we build the notion that one area’s dynamics is likely to affect another into the model and analysis. The observed variables in the model include total catch, a single fish species catch, temperature, Net PP, total zooplankton biomass, a single fish species and three aggregated trophic species groups: P, SP, and LP and the equivalent spatial nodes from above. This totals 13 observed variables per area. The HSDBN structure varies but the general form is presented in Figure 3a, with example for one of the areas in Figure 3b. Hence, we can explore multiple species associations and model their future dynamics with interactions from external stressors and under specific scenario conditions. Using a recognized model structure, we can compare the modelled scenario outputs across spatial and temporal scales, accounting for the spatial heterogeneity and ecological complexity. Figure 3 Open in new tabDownload slide General structural form of the HSDBN model (a). Solid line represents fixed edges across areas. The spatial nodes (P sp., SP sp., LP sp., Fish species sp.), HV, catch and temperature are individually linked to either P, SP, or LP (represented by the dotted surrounding), depending on the spatial area (grey line). Connectivity between P, SP, and LP and with the fish species also differs spatially. Network structure for area 4 (b) that models the dynamics of cod. The edges shown by a dotted line are defined by expert knowledge. Figure 3 Open in new tabDownload slide General structural form of the HSDBN model (a). Solid line represents fixed edges across areas. The spatial nodes (P sp., SP sp., LP sp., Fish species sp.), HV, catch and temperature are individually linked to either P, SP, or LP (represented by the dotted surrounding), depending on the spatial area (grey line). Connectivity between P, SP, and LP and with the fish species also differs spatially. Network structure for area 4 (b) that models the dynamics of cod. The edges shown by a dotted line are defined by expert knowledge. Experiments The experiments involved prediction of survey data under scenarios of fisheries catch, temperature and Net PP. The network architecture varied with the areas but the method of prediction was universal. Given the probability distribution over X[t] where X = X1…Xn are the n variables observed along time t, to predict the biomass of each species and/or trophic group, we inferred the biomass at time t + 1…t + 5 by using the observed evidence (or available data) from t−1 and t. The choice of 2020 as the horizon for this study was chosen to limit uncertainty and, more importantly, to reflect the need for short-term predictions in fish stock management. We used an exact inference method: the junction tree algorithm (Murphy, 1998). The HV is specified as a discrete node which is parameterized using the Expectation Maximization algorithm in a maximum likelihood sense and assumes a discrete distribution. Non-parametric bootstrap [re-sampling with replacement from the training set, (Friedman et al., 1999)] was applied 250 times for each modelling scenario to obtain statistical validation in the predictions for each area (number of iterations was found to be optimum through experimentation). First, we predict the survey data for each area using historical observations, we refer to this model output as Historical. Then, we use different fixed year levels from each individual fish species catch data to design our fisheries catch scenarios. We use scenarios at varying levels of fisheries catch: low, medium, and high (these to be referred from now on as scenarios of L.FC., M.FC., and H.FC., respectively). We choose from the fisheries catch data 3 years equivalent to these levels and keep each level fixed from the chosen “scenario” year until the year 2015. We keep the other measured variables unchanged. For example, in order to model the dynamics of cod in area 4 in response to change in fisheries catch, we chose from the cod catch data the year 1995 to represent the year from which the scenario of M.FC. starts. Figure 4a illustrates the data input assuming this scenario and the generated output. Note, that the data input for testing the M.FC. model, prior to the chosen scenario year, includes all of the observed variables (and one unmeasured HV) up to 1995 and after 1995–2015, the input is only the fixed values of the total fisheries catch and cod catch (5 × 104 tonnes live weight) (Figure 4b). In this way, we rule out the simple idea that observed values after the “scenario” year are causing the results to stabilize. Figure 4 Open in new tabDownload slide An example matrix from a Medium Fisheries Catch scenario model with initial input used in model definition, the input during model testing and the generated output (a). The time window for each variable is shown in brackets. Note, the time window for the output starts from 1989. “[]” represents variables for which no evidence is introduced and which are predicted. Z stands for zooplankton. The observed cod catch prior to the scenario year of 1995 (solid line) and fixed catch level for the Medium Fisheries Catch scenario (dashed line) is shown in (b). Figure 4 Open in new tabDownload slide An example matrix from a Medium Fisheries Catch scenario model with initial input used in model definition, the input during model testing and the generated output (a). The time window for each variable is shown in brackets. Note, the time window for the output starts from 1989. “[]” represents variables for which no evidence is introduced and which are predicted. Z stands for zooplankton. The observed cod catch prior to the scenario year of 1995 (solid line) and fixed catch level for the Medium Fisheries Catch scenario (dashed line) is shown in (b). We perform this for each individual fish species and across each area, according to the originally published model structure. For example, in area 4 catch is a direct parent to LP, so in this area, we would investigate fisheries catch scenarios for individual LP fish species such as cod (Figure 3b). At the same time, we predict other fish species which are represented by the trophic species groups (P, SP, and LP). Essentially, each area is characterized by a sub-model, driven by the spatial dynamics of the species of interest (there could be more than one sub-model for an area) that accounts for any specific biotic and abiotic interactions between that species and other variables. In this way, we can keep the historically driven interactions between variables and examine their modelled trends under potential changes in stressors such as fisheries catch. Hence, we can examine how different ecosystem components respond to varying levels of fisheries catch, accounting for the heterogeneous nature of the modelled variables and driving factors within each area and their changes over time. Data input and output for Medium Fisheries Catch scenario for cod, area 4 Fisheries Catch Level for Medium Fisheries Catch scenario for cod, area 4 We generate a 10% increase temperature scenario (T.I.) and Net PP scenarios: 30% increase and 30% decline (referred from now on as: Net.I. and Net.D) to understand the effects of temperature on primary production and its potential knock-on effects on different zooplankton species and trophic species higher up the food chain. We did consider a scenario of temperature decline but we only present the results following a potential increase in temperature. We used 1990 as the “divergent year”, which is the year to start the scenario changes from by manipulating the temperature or Net PP data to either increase or decline but keeping the rest of the observed data unchanged, e.g. if the average sea surface temperature for 1990 is 9°C, then for 1991 it would be 9.9°C. For these two types of scenarios, the number of observed variables in the experimental set-up is 12 (total catch, temperature, Net PP, Calanus finmarchicus, Calanus helgolandicus, small copepods, P sp., SP sp., LP sp., P, SP, and LP). Results In the following, we describe the outputs from the modelled fisheries catch, temperature and Net PP scenarios by examining future trends of individual fish and zooplankton species at spatial and temporal scales. We explain the results from the scenarios by examining if the predictions of the ecosystem components were to increase or decline. Our results demonstrate some variability in the future trends of different species, which we explain through the use of “what if” type descriptions of the model structures in response to predicted changes in the other variables. Fisheries catch scenarios Cod First, looking at the Historical output, the model managed to capture the cod variations throughout time and predicted some increase in near future years which were then followed by some decline (Figure 5c). Figure 5 Open in new tabDownload slide Recorded spatial cod data is shown in (a). The observed cod catch (live weight in tonnes) with the three fixed year levels of fisheries catch scenarios for the time window 1983–2015 is shown in (b). Recorded survey cod data (solid line) with the generated output by the Historical model (dotted line) for the time window 1989–2020 for area 4 is shown in (c). Recorded survey cod (solid line) with the modelled cod is shown in (d) under fisheries catch scenarios of high (black dashed line), medium (grey dashed line) and low (black dotted line) levels for the time window 1989–2020. Figure 5 Open in new tabDownload slide Recorded spatial cod data is shown in (a). The observed cod catch (live weight in tonnes) with the three fixed year levels of fisheries catch scenarios for the time window 1983–2015 is shown in (b). Recorded survey cod data (solid line) with the generated output by the Historical model (dotted line) for the time window 1989–2020 for area 4 is shown in (c). Recorded survey cod (solid line) with the modelled cod is shown in (d) under fisheries catch scenarios of high (black dashed line), medium (grey dashed line) and low (black dotted line) levels for the time window 1989–2020. Second, looking at the scenario outputs, as we would expect, the scenario of High Fisheries Catch (H.FC.) resulted in the lowest modelled cod survey data in areas 4 (Figure 5d) and 6 (thus, addressing in detail only area 4 but look at Figure 3a and b in the Supplementary Materials for area 6). We notice a sudden decline in early 1990s (as a result from the high scenario catch level), but then the modelled values were characterized by some fluctuating trend, that was higher than the observed data. This does not mean that if cod could continue to be fished at the highest recorded level the stock would still be ok but more likely when the cod survey values and spatial biomass in neighbouring areas (Figure 5a) are low and catch is high (Figure 5b), another species might increase and a year later that would cause the cod to increase. For example, in this area, cod is influenced by the dynamics of species group P (Figure 3b), which were predicted to be relatively stable with an increasing trend in the near future, partly explaining the modelled cod results here. Under the scenario of Medium Fisheries Catch (M.FC.), the modelled survey data seemed to be genuinely stable throughout time that was higher than the scenario of H.FC. However, we notice that these two scenarios seem to converge in the near future, highlighting the similarity in species response to contrasting levels of fisheries catch, thus still having the need to identify a potential “optimum” level of fisheries catch. The scenario of Low Fisheries Catch (L.FC.) resulted in the highest modelled cod survey data, highlighting the importance of fisheries catch on this species dynamics and identifying a potential “optimum” level of fisheries exploitation comparing to the medium and high levels from above. Cod spatial data, area 4 Cod catch, area 4 Cod survey data and Historical output, area 4 Cod survey data and modelled scenario cod, area 4 Whiting The Historical model managed to reflect on the declining trend of whiting throughout time and predicted some rising trends in the near future which were then followed by some decline (Figure 6c). Figure 6 Open in new tabDownload slide The model structure for area 3 is shown (a). The dotted edges are defined by the expert. The observed whiting catch (live weight in tonnes) with the three fixed year levels of fisheries catch scenarios for the time window 1983–2015 is shown in (b). Recorded whiting survey (solid line) data with the generated output by the Historical model (dotted line) for the time window 1989–2020 for area 3 is shown in (c). Recorded survey (solid line) with the modelled whiting data under fisheries catch scenarios of high (black dashed line), medium (grey dashed line), and low (black dotted line) levels for the time window 1989–2020 is shown in (d). Figure 6 Open in new tabDownload slide The model structure for area 3 is shown (a). The dotted edges are defined by the expert. The observed whiting catch (live weight in tonnes) with the three fixed year levels of fisheries catch scenarios for the time window 1983–2015 is shown in (b). Recorded whiting survey (solid line) data with the generated output by the Historical model (dotted line) for the time window 1989–2020 for area 3 is shown in (c). Recorded survey (solid line) with the modelled whiting data under fisheries catch scenarios of high (black dashed line), medium (grey dashed line), and low (black dotted line) levels for the time window 1989–2020 is shown in (d). We found the opposite of what we were expecting from the fisheries catch scenarios for whiting in area 3: a scenario of L.FC. produced whiting predictions that were characterized with the lowest trend throughout time (Figure 6d). The surrounding predictions of the whiting spatial node were also characterized by a declining trend, which in combination with the medium to high catch from M.FC. and H.FC. and relatively low values of the P species group (network shown in Figure 6a) might allow for another species to increase (e.g. larger predator), which in turn would cause the projected whiting values here. We also note that the predicted trends from the M.FC. and H.FC. scenarios were relatively similar. Interestingly, the hidden variable (HV) captured some of the expected “correct” characteristics: the scenario of L.FC. projected a strongly increasing trend of the HV, that was much higher than the HV from the Historical model. The HV is linked to the LP species group (which includes cod), so it is capturing changes in the variance of their survey data, due to species associations and interactions (LP is influenced by SP and P sp.) and consequent trade-offs between species, that were not easily detected by the model predictions alone. Thus, still having the need to identify a potential “optimum” level of fisheries catch to account for the effect of trade-offs between species. Area 3 Whiting catch, area 3 Whiting survey data and Historical output, area 3 Whiting survey data and modelled whiting, area 3 To summarize, for most species we were able to predict trends that were modelled to either increase or decrease in response to change in fisheries catch but this varied across areas, thus highlighting the spatial heterogeneity in terms of species-specific response to ecosystem change, the spatial relationship between neighbouring areas and trophic interactions. Finally, we need to mention that the aggregated species group biomass might include species not directly targeted by fisheries, which could potentially influence the scenario interpretations, however the fact that we accounted for “what if” type descriptions of all ecosystem components in the network model, should help us in the interpretation of our results. Temperature and Net PP scenarios We are now looking at the potential influence of temperature on the future projections of productivity and consequently how the productivity will influence the future trends of different zooplankton species. We have chosen to present results only for areas 1, 3, and 6 due to the contrasting nature of the physical and bio-chemical characteristics of these areas. For area 1 (and area 3), the scenario of T.I. resulted in an increasing trend of Net PP throughout time that was also higher than the Historical model. However, the T.I. scenario projected some Net PP decline in the near future that was characterized by a converging trend with the projections of the Historical model, possibly indicating a drop in productivity. Conversely, for area 6, the scenario of T.I. projected a trend of lower Net PP values than the Historical model, potentially due to larger temperature changes in southern areas. Similarly to areas 1 and 3, there was a drop in productivity projected from 2017 onwards. Following a scenario of temperature increase, a lower trend (compared with the Historical model) throughout time was projected for C. finmarchicus, whilst the opposite was found for C. helgolandicus. In some areas, it was also the scenario of Net PP decline that led to higher values of both zooplankton species, as a consequence of temperature influence on productivity. However, at the same time, a distinct drop was found in the projected values of the C. finmarchicus species in the near future, highlighting that potential trade-offs will also emerge between lower trophic level species. Look at the Supplementary Material (2.2 Temperature and Net PP Scenarios) for a more detailed description of the zooplankton results in terms of the different spatial areas and the modelled predictions (Supplementary Figure S7). We were able to detect a knock-on effect on the future dynamics of the P species group survey data, following changes in temperature and productivity. For some of the areas, it was the scenario of Net PP decline that led to an increase in the trends of the herring and P species group survey data. Look at the Supplementary Materials, section 2.2 for a more detailed description of these results in terms of the different spatial areas and modelled predictions (Supplementary Figure S8). To summarize, we found the modelled future zooplankton trends to be species-specific but there seems to be consistency in terms of their response to temperature change across the different areas, whilst more variability was found relating to productivity changes. In addition, we were able to confirm the potential influence from productivity and to some extent temperature (depending on the area) changes to species, higher up the food chain. Discussion In this study, we explored the trends of ecosystem change in response to anthropogenic and environmental scenarios by modifying a dynamic data-driven functional network model, accounting for spatial heterogeneity and unmeasured spatial effects. It is important to note that we did not attempt to indicate levels of plausibility between these scenarios but rather explore the predictive results of species response to fisheries and environmental change. Our results highlighted that reducing fisheries catch will not necessarily lead to recovery of all commercially important fish species because fish consume one another, thus the total catch of one species will consequently affect that of others through knock-on effects in the food web. Overall, we found some spatial variability in terms of species response to different fisheries catch and productivity scenarios, highlighting the influence from factors such as trophic associations, spatial connectivity between areas and species interactions with their environment, that could potentially contribute towards the better understanding of ecological stability and resilience in a changing environment. However, at the same time, we found some universal species trends to changes in catch and temperature that could provide some strategic advice on potential response of the system to such pressures. Controlling for the level of fisheries exploitation but also considering trophic interactions and spatial values are of high significance in terms of short-term management. Our results allow dynamic assessment of choices, which should be able to provide strategic advice on potential system response to pressure. In terms of management objectives and expectations, we support the idea that for a given area, reorganization of the management strategies will be required to ensure that the right species are targeted and harvested sustainably (Simpson et al., 2011). Management strategies must also take into account the local population dynamics and processes in a wider sense in order to maximize biodiversity and survival. Fisheries management measures will contribute to improvements in the biodiversity of the fish community, but food web interactions will mediate changes. In the scenarios modelled here, some trade-offs between species emerged in terms of how they would respond to different levels of fisheries catch. Specifically, the potential recovery that we found for cod in the near future (and variance explained by the HV) could explain the modelled results for whiting because cod feeds on juvenile whiting (Mackinson et al., 2009). Similar results were found by Lewy and Vinther (2004) and Lynam and Mackinson (2015), suggesting a more dominant role of the cod in the food web after recovering from exploitation. The potential recovery trend that we found for cod could be due to strict management regulations placed since the Millennium (Horwood et al., 2006), which if continued, will hopefully give the stock a chance to rebuild completely in some areas where the cod was formerly abundant (Engelhard et al., 2014). One of the differences between our model and others is the incorporation of a HV, adopted to capture unmeasured spatial effects and changes in species variance that are not purely constrained within the model structure. For some of the areas, the HV was characterized by a decline and showed high sensitivity in terms of catch variation, outlining that such areas seem to exhibit a range of discontinuous disturbance exacerbated by spatial differences in recruitment and survival. Conversely, for some of the other areas, the learned HVs were projected to increase, following some of the scenarios, which are reflective of the underlying biomass changes, relating to potential knock-on effects, as it was found for area 3. Specifically, our results of modelling whiting in response to different fisheries levels and consequent rising trophic interactions and sensitivities that were captured by the HV, suggest that for effective management, reorganization of the fishing strategies in the mixed-fisheries context will be required to ensure that the right species are targeted and harvested sustainably (Simpson et al., 2011). These results highlight that the use of a HV when modelling species response to change is potentially useful in providing insights on the spatially specific dynamics and patterns in terms of ecological stability and resilience that can contribute towards the general advice on potential response of the system to pressure. Overall, our results showed there were spatial differences in terms of “optimum” level of fisheries catch, suggesting spatial variability regarding community stability and the potential higher influence of trophic interactions in some areas or spatial connectivity in others, compared with fisheries exploitation. For example, we found some similarity in the modelled whiting predictions from the medium and high fisheries scenarios for area 3, which might be due to similarity in the level of fisheries catch but also due to the fact that trophic interactions are potentially more important for controlling the whiting dynamics compared with fisheries, as discussed in Trifonova et al. (2015) for this area. This suggests that stocks cannot be managed in isolation from each other (Cury et al., 2005). Thus, highlighting the need to use multi-species models accounting for spatial connectivity. Multispecies models have been proved useful in terms of providing long-term information on stock recovery and most importantly, have been used to evaluate precautionary reference points for fishery management (Pinnegar et al., 2008). In doing so, multispecies and ecosystem models are anticipated as being helpful to guide strategic management decisions (Mackinson and Daskalov, 2007). As such, multispecies stock assessments and simulation models (e.g. SMS, 4M, Gadget, multispecies IBMs) are becoming more refined (Plagányi, 2007). Although, we analyse the different scenarios in respect to the species of interest in the relevant area, we do acknowledge that one area's dynamics likely affect another by introducing the spatial nodes into the model structure. In this way, we also increase the confidence in the robustness of the approach and contribute to increased knowledge of model behaviour. One main issue encountered is the uncertainty in future trends, which is obviously inherent to any model linking external factors to species interactions. These linkages are of major importance for mixed-fisheries management (Ulrich et al., 2011). However, the fact that we were able to recover genuine trends of species dynamics throughout space and time in Trifonova et al., (2015) and that we were able to identify similarity in our results here with other modelled species predictions (Lewy and Vinther, 2004; Vinther et al., 2004; Lynam and Mackinson, 2015) contributes to strengthening the confidence that our approach can provide some strategic advice on modelling species response to change. Here, the modelling framework was built to handle complex systems such as the North Sea, so consequently we assume there is a degree of complexity when modelling fisheries. The assumptions are based on key processes within the environment accounting for influence from external factors such as fisheries catch. One aspect of the underlying processes that could be further investigated includes fishermen behaviour or effort information to estimate catch potentials for distinct fleets. An example of one model that incorporates this is the Fcube (Fleet and Fisheries Forecast) model. However, for the Fcube to be established at a regional scale requires substantial analysis and due to its short-term applicability, is often used as a routine advice model at the same level as a single-stock assessment model (Ulrich et al., 2011). Another example of a model that uses information on technical interactions alongside biological information from stock assessments is the MTAC developed by Vinther et al. (2004). However, the MTAC did not prove to be robust and flexible enough for mixed- fisheries and there were also problems with data availability (Ulrich et al., 2011). The HSDBN model represents a flexible framework of medium complexity between single-stock assessments and multi-species models such as Ecopath (Mackinson and Daskalov, 2007). By extending our model to use scenarios rather than optimization and adding additional parameters compared with more traditional approaches, we extract simple proxies that are indicative at the regional scale but also work at the level of the broad picture. A similar dynamic framework for the North Sea, accounting for multiple-species interactions, was developed by Lynam et al. (unpublished) (presented at PICES Symposium on “Effects of climate change on the world’s oceans”, March 2015), using a threshold-Generalized Additive Model. The approach is data-demanding and it includes external factors but does not include a spatial component. In our model, we account for the complexity of the spatio-temporal distribution by allowing a framework that accounts for the heterogeneous nature of the driving factors within each area (unique model structure for each area) and their changes over time. Explicit spatiality is a key parameter in our model which does add some complexity to the model structure and it is data-demanding but accounting for additional sources of variation seems to remove spurious interactions and reveal the genuine complexity of such diverse and exploited ecosystems such as the North Sea. Although, we allow for some variability, the model has proven its high flexibility enabling latent effects and testing alternative hypotheses about species and their dynamics to reduce scientific uncertainty. Finally, in the modelled scenarios here, we found that some species appear more robust to changes in fisheries exploitation, compared with others; however, changes in temperature and productivity might be more important in terms of the species long-term sustainability. It was interesting to see that our results of modelling a drop in future productivity coincides with other work that could be related to the overall future productivity conditions expected in the North Sea (Blanchard et al., 2012). Increase in temperature leads to an increase in lower trophic level species and consequently their predators, which we found true for some areas, whilst in others, the effect of temperature on fish was less evident due to interactions with productivity, which could be acting more strongly than the effect of fishing (Blanchard et al., 2010). For example in area 3, the influence of productivity is likely to mask the effects from fisheries, or cause a mixture of responses due to multiple causal mechanisms and stressors on the ecosystem (Halpern et al., 2008). Such results confirm that species response to any future changes in temperature will be determined by their spatial habitat because temperature variations consequently lead to spatial variability in productivity, potentially causing further forcing on higher level trophic species and mixture of responses at spatial scales. Supplementary data Supplementary material is available at the ICESJMS online version of the article. Acknowledgements We would like to thank Daniel Duplisea from DFO, Canada and Simon Jennings from CEFAS for providing comments and feedback. Johan Van Der Molen from CEFAS for providing the ERSEM model outputs, the ICES DATRAS database for the North Sea IBTS data and Historical Catch Statistics, ICES North Sea Integrated Assessment Working Group (WGINOSE) and the organizations which provide data for the ICES assessment process, in particular SAHFOS who have provided the North Sea plankton data. We gratefully acknowledge the Natural Environment Research Council UK that has funded this research, along with support from the European Commission (OCEAN-CERTAIN, FP7-ENV-2013-6.1-1; no: 603773) for David Maxwell and from CEFAS for Andrew Kenny and David Maxwell. References Barange M. , Merino G., Blanchard J., Scholtens J., Harle J., Allison E., Allen J. et al. 2014 . Impacts of climate change on marine ecosystem production in societies dependent on fisheries . Nature Climate Change , 4 : 211 – 216 . Google Scholar Crossref Search ADS WorldCat Blanchard J. , Coll M., Trenkel V., Vergnon R., Yemane D., Jouffre D., Link J. et al. 2010 . Trend analysis of indicators: a comparison of recent changes in the status of marine ecosystems around the world . ICES Journal of Marine Science , 67 : 732 – 744 . Google Scholar Crossref Search ADS WorldCat Blanchard J. , Jennings S., Holmes R., Harle J., Merino G., Allen J., Holt J. et al. 2012 . Potential consequences of climate change for primary production and fish production in large marine ecosystems . Philosophical Transactions of the Royal Society of London B: Biological Sciences , 367 : 2979 – 2989 . Google Scholar Crossref Search ADS WorldCat Chen S. H. , Pollino C. A. 2012 . Good practice in Bayesian network modelling . Environmental Modelling and Software , 37 : 134 – 145 . Google Scholar Crossref Search ADS WorldCat Cury P. M. , Mullon C., Garcia S. M., Shannon L. J. 2005 . Viability theory for an ecosystem approach to fisheries . ICES Journal of Marine Science , 62 : 577 – 584 . Google Scholar Crossref Search ADS WorldCat EC ,. 2013 . Regulation (EU) No 1380/2013 of the European Parliament and of the Council of 11 December 2013 on the Common Fisheries Policy, amending Council Regulations (EC) No 1954/2003 and (EC) No 1224/2009 and repealing Council Regulations (EC) No 2371/2002 and (EC) No 639/2004 and Council Decision 2004/585/EC . Official Journal of the European Union. L , 354 : 22 – 61 . OpenURL Placeholder Text WorldCat Engelhard G. H. , Righton D. A., Pinnegar J. K. 2014 . Climate change and fishing: a century of shifting distribution in North Sea cod . Global Change Biology , 20 : 2473 – 2483 . Google Scholar Crossref Search ADS PubMed WorldCat Fernandes J. A. , Irigoien X., Goikoetxea N., Lozano J. A., Inza I., Perez A., Bode A. 2010 . Fish recruitment prediction, using robust supervised classification methods . Ecological Modelling , 221 : 338 – 352 . Google Scholar Crossref Search ADS WorldCat Friedman N. , Goldszmidt M., Wyner A. 1999 . Data analysis with Bayesian networks: A bootstrap approach. In: Proceedings of the Fifteenth Conference on Uncertainty in Artificial Intelligence, pp. 196–205. Morgan Kaufmann Publishers Inc. García-Carreras B. , Dolder P., Engelhard G. H., Lynam C. P., Bayliss-Brown G. A., Mackinson S. 2015 . Recent experience with effort management in Europe: implications for mixed fisheries . Fisheries Research , 169 : 52 – 59 . Google Scholar Crossref Search ADS WorldCat Halpern B. S. , Walbridge S., Selkoe K. A., Kappel C. V., Micheli F., D’Agrosa C., Bruno J. F. et al. 2008 . A global map of human impact on marine ecosystems . Science , 319 : 948 – 952 . Google Scholar Crossref Search ADS PubMed WorldCat Hamilton S. H. , Pollino C. A., Jakeman A. J. 2015 . Habitat suitability modelling of rare species using Bayesian networks: Model evaluation under limited data . Ecological Modelling 299 : 64 – 78 . Google Scholar Crossref Search ADS WorldCat Horwood J. , O’Brien C., Darby C. 2006 . North Sea cod recovery? ICES Journal of Marine Science , 63 : 961 – 968 . Google Scholar Crossref Search ADS WorldCat Jennings S. , Kaiser M. J. 1998 . The effects of fishing on marine ecosystems . Advances in Marine Biology , 34 : 201 – 352 . Google Scholar Crossref Search ADS WorldCat Jensen F. V. 2001 . Bayesian networks and decision graphs. In Statistics for Engineering and Information Science . Springer , New York . 268 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Lee D. C. , Rieman B. E. 1997 . Population viability assessment of salmonids by using probabilistic networks . North American Journal of Fisheries Management , 17 : 1144 – 1157 . Google Scholar Crossref Search ADS WorldCat Levontin P. , Kulmala S., Haapasaari P., Kuikka S. 2011 . Integration of biological, economic, and sociological knowledge by Bayesian belief networks: the interdisciplinary evaluation of potential management plans for Baltic salmon . ICES Journal of Marine Science , 68 : 632 – 638 . Google Scholar Crossref Search ADS WorldCat Lewy P. , Vinther M. 2004 . A stochastic age-length-structured multispecies model applied to North Sea stocks . ICES Document CM , 2004/FF : 20 . Google Scholar OpenURL Placeholder Text WorldCat Lynam C. P. , Mackinson S. 2015 . How will fisheries management measures contribute towards the attainment of good environmental status for the North Sea ecosystem? Global Ecology and Conservation , 4 : 160 – 175 . Google Scholar Crossref Search ADS WorldCat Mackinson S. , Daskalov G. 2007 . An ecosystem model of the North Sea to support an ecosystem approach to fisheries management: description and parameterisation. CEFAS Science Series Technical Report, pp. 142–195. Mackinson S. , Deas B., Beveridge D., Casey J. 2009 . Mixed-fishery or ecosystem conundrum? Multispecies considerations inform thinking on long-term management of North Sea demersal stocks . Canadian Journal of Fisheries and Aquatic Sciences , 66 : 1107 – 1129 . Google Scholar Crossref Search ADS WorldCat Murphy K. 1998 . A brief introduction to graphical models and Bayesian networks. http://www.cs.ubc.ca/∼murphyk/Bayes/bnintro.html Murphy K. P. 2002 . Dynamic Bayesian networks: representation, inference and learning. PhD thesis, University of California, Berkeley. Murphy K. 2001 . The Bayes net toolbox for Matlab . Computing Science and Statistics , 33 : 1024 – 1034 . Google Scholar OpenURL Placeholder Text WorldCat Olson R. , Willers J., Wagner T. 1990 . A framework for modelling uncertain reasoning in ecosystem management. II. Bayesian belief networks . AI Applications in Natural Resource Management 4 : 11 – 24 . Google Scholar OpenURL Placeholder Text WorldCat Pinnegar J. , Trenkel V., Blanchard J. 2008 . 80 years of multi-species fisheries modelling- significant advances and continuing challenges. In Advances in Fisheries Science: 50 years on from Beverton and Holt , pp. 325 – 357 . Ed. by Payne A., Cotter J., Potter T.. Blackwell Publishing . Oxford, UK . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Plagányi ÉE. 2007 . Models for an ecosystem approach to fisheries. FAO Fisheries Technical Paper, No. 477, Rome, FAO. Planque B. , Fromentin J., Cury P., Drinkwater K. F., Jennings S., Perry R. I., Kifani S. 2010 . How does fishing alter marine populations and ecosystems sensitivity to climate? . Journal of Marine Systems , 79 : 403 – 417 . Google Scholar Crossref Search ADS WorldCat Simpson S. D. , Jennings S., Johnson M. P., Blanchard J. L., Schon P., Sims D. W., Genner M. J. 2011 . Continental shelf-wide response of a fish assemblage to rapid warming of the sea . Current Biology , 21 : 1565 – 1570 . Google Scholar Crossref Search ADS PubMed WorldCat Smith V. A. , Yu J., Smulders T. V., Hartemink A. J., Jarvis E. D. 2006 . Computational inference of neural information flow networks . PLoS Computational Biology , 2 : e161. Google Scholar Crossref Search ADS PubMed WorldCat Trifonova N. , Kenny A., Maxwell D., Duplisea D., Fernandes J., Tucker A. 2015 . Spatio-temporal Bayesian network models with latent variables for revealing trophic dynamics and functional networks in fisheries ecology . Ecological Informatics , 30 : 142 – 158 . Google Scholar Crossref Search ADS WorldCat Ulrich C. , Reeves S. A., Vermard Y., Holmes S. J., Vanhee W. 2011 . Reconciling single-species TACs in the North Sea demersal fisheries using the Fcube mixed-fisheries advice framework . ICES Journal of Marine Science , 68 : 1535 – 1547 . Google Scholar Crossref Search ADS WorldCat Uusitalo L. 2007 . Advantages and challenges of Bayesian networks in environmental modelling . Ecological Modelling , 203 : 312 – 318 . Google Scholar Crossref Search ADS WorldCat Uusitalo L. , Kuikka S., Kauppila P., Söderkultalahti P., Bäck S. 2012 . Assessing the roles of environmental factors in coastal fish production in the northern Baltic Sea: A Bayesian network application . Integrated Environmental Assessment and Management , 8 : 445 – 455 . Google Scholar Crossref Search ADS PubMed WorldCat Vinther M. , Reeves S. A., Patterson K. R. 2004 . From single-species advice to mixed-species management: taking the next step . ICES Journal of Marine Science , 61 : 1398 – 1409 . Google Scholar Crossref Search ADS WorldCat © International Council for the Exploration of the Sea 2017. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. © International Council for the Exploration of the Sea 2017.