ICES Journal of Marine Science: Journal du Conseil

Kleiven, Alf, Ring;Moland,, Even;Sumaila, U, Rashid

doi: 10.1093/icesjms/fsz128pmid: N/A

Abstract Recreational fishing, by both local residents and tourists, is a popular activity globally. The behaviour and motivation of recreational fishers is different from those of commercial fishers. Unlike the latter, the former are not dependent on making profits to continue fishing. Rather, the value of recreational fishing to those who engage in it is a combination of catches and experience values. The latter value implies that recreational fishers might continue fishing when they should not, analogous to the effect of subsidy in the commercial fishing sector. Hence, the term “self-subsidizing”: a fishery as one in which fishers subsidize themselves through an economic investment in gear and time from their non-fishery-based earnings. The consequence of which is that recreational fishers can continue fishing long after the commercial fishing industry has stopped fishing because their operations have become economically unviable. There is reason to argue that in many areas, recreational fishing effort, due to the self-subsidizing mechanism, is sustained at a high rate while stocks decrease. In this contribution, we describe the innate self-subsidizing forces in recreational fishing and discuss their implications. Introduction Coastal areas are subject to the highest pressure from human activities, resulting in degraded habitats and decreased biomass and diversity (Jackson et al., 2001). The widespread collapse of fish stocks in coastal ecosystems due to overfishing has been confirmed repeatedly (Pauly et al., 1998; Jackson et al., 2001; Myers and Worm, 2003; Worm et al., 2006). Measures to combat overfishing have been thoroughly discussed (Pauly et al., 2002; Gell and Roberts, 2003; Halpern, 2003; Worm et al., 2006; Francis et al., 2007; Hilborn, 2007). One important driving factor for overfishing has been the weighty subsidy of the commercial fishing industry, where “taxes from outside the fisheries sector are used to maintain fishing at levels that are biologically and economically unsustainable, and which ultimately lead to the depletion and collapse of the underlying resources” (Pauly et al., 2002; Sumaila et al., 2008). Recreational fishing is a popular activity globally, including fishing by local residents and tourists. In recent decades, recreational fishing and its relation to overfishing and biological impacts has gained some needed attention (Schroeder and Love, 2002; Coleman et al., 2004; Cooke and Cowx, 2004; Lewin et al., 2006; Ihde et al., 2011). The activity is mostly conducted in close vicinity to the coast, resulting in concentration of effort in relatively small areas. High fishing intensity from recreational fishers has been observed on coastal populations that are overfished (Coleman et al., 2004; Kleiven et al., 2012,, 2016; Nillos-Kleiven et al., 2019). Moreover, evidence exist that recreational catches for certain species can exceed commercial catches (Schroeder and Love, 2002; Coleman et al., 2004; Kleiven et al., 2012,, 2016). Lewin et al. (2006) argued that there is a growing body of evidence indicating that activities related to recreational fishing can “lead to a decline of fish populations and affect aquatic ecosystems if the degree of fishing mortality is high and the selective exploitation is intensive.” There is now an extensive body of literature pertaining to recreational fishing, both theoretical and empirical. Commercial and recreational fisheries often target the same fish populations, and management must handle both activities to secure sustainability. The innate differences in driving forces between a commercial fishery seeking profit and livelihoods and a non-profit-motivated recreational fishery warrant further discussion. While a commercial fishery will cease when fishing is no longer profitable, recreational fisheries have the potential to continue fishing since the activity is not driven by economical revenue. Fishing without fear of bankruptcy Commercial fishers depend on profit from their fishery. Motivations for recreational fishing are not solely dependent on the value of expected catches. Reasons to fish are a combination of factors related to recreational values, such as enjoying a quality environment and feeling a sense of freedom (Holland and Ditton, 1992), as well as catch and consumption (Cooke et al., 2018). In Richards Bay, South Africa, Beckley et al. (2008) found that recreation was the main motivation to go fishing for 99% of anglers. Furthermore, catch-per-unit-effort (CPUE) for retained fish was as low as 0.064 fish angler−1 h−1. Based on the reasons why people go fishing, it could be expected that a decrease in CPUE will not affect effort by recreational fishers in the same way as for commercial fisheries, as observed by Post et al. (2002) in Canadian lakes. In other words, recreational fishers harbour a potential to continue fishing after the commercial fishery has stopped fishing due to low and economically unsustainable catches. In a social–ecological modelling study, Hunt et al. (2011) examined effects of angling on landscape patterns of overfishing and found responses ranging from self-regulating systems to sequential collapse of walleye fisheries in Canadian lakes, based on highly variable drivers of angler behaviour. They noted that in their model, only very remote fish stocks would be saved from substantial overharvest when regional fishing pressure increased critically to a high level. At some point, one should expect recreational fishers to quit fishing and move on to alternative activities when catch rates hit a critically low level. However, decisions based on such “opportunity costs” of activities will vary in space and time. As shown by Beckley et al. (2008), recreational fishers in Richards Bay were fishing even though catch rates were very low. Tolerable catch rates will vary between different fisheries, but it is hard to predict such behaviour for a recreational fishery. The motivation for—and satisfaction of a fishing trip can change in time, for example due to shifting baselines (sensuPauly, 1995): catching one fish in the present day might generate the same satisfaction—and motivation for continued fishing—as catching three fish on a fishing trip one generation ago, if other non-catch-related values have remained high. Lovell et al. (2016) estimated that recreational saltwater fishers in the US spent $28 billion on fishing equipment and durable goods in 2014. Recently, Hyder et al. (2018) estimated the total expenditure on European marine recreational fisheries to be €5.9 billion annually. Investments in new and better gear (rods, lures, bigger boat, GPS chart-plotter, acoustic fish-finder, etc.) can contribute to technological creep in the recreational fishery with potential to sustain catch rates while stocks are decreasing. Technological creep seems to occur within all types of commercial fisheries (Pauly and Palomares, 2010) and there is reason to expect a similar trend within the recreational sector. Meanwhile, studies addressing technological creep in recreational fisheries are limited, if at all existing. However, not only direct expenditures should be included in estimates of investment in recreational fishing. Time is another expenditure related to the activity. Recreational fishers have no direct economic expenses linked to time spent fishing, in contrast to commercial fishers for whom salaries must be derived and paid to themselves or to crew members. Recreational fishers are using their spare time to fish for the sake of recreation. They can therefore invest a lot of their time in fishing without high economic risk and are in no need to make a profit. Putting it another way: recreational fishers can spend a whole day catching one fish without fear of bankruptcy. Conversely, commercial fishing operations fish for profit—or for a break-even after covering financial costs. Subsidies that reduce the cost of fisheries operations and those that enhance revenues make fishing enterprises more profitable than they would otherwise be. This results directly or indirectly in the build-up of excessive fishing capacity, leading to the overexploitation of fishery resources. Therefore, the elimination of such subsidies, termed overfishing subsidies by Sumaila et al. (2007), is crucial to our ability to manage fisheries sustainably through time. “Self-subsidized” recreational fishing Herein, we argue that recreational fishing might be termed subsidized. However, not in the traditional way as for the commercial fishing sector, where subsidies are transferred from the government to the industry to maintain a break-even in a fishery that in reality is no longer economically viable. Recreational fishers, however, are subsidizing themselves through an economic investment from their earnings into their fishing activity (boats, gear, fuel, etc.) and through an indirect economical investment in time. Thus, we define a self-subsidized fishery as one in which fishers subsidize themselves through an economic investment in gear and time from their non-fishery-based earnings. Even though recreational fishers are a highly heterogenous group with regards to motivations and fishing methods, they all have in common that they transfer money and time into the fishing sector. As the value of recreational fishing is not directly linked to size and economic value of catch, there is a weak relationship between the willingness to invest in recreational fishing and the expected net catch value. If the experience value remains high, a recreational fisher might continue to “hunt the last fish.” The self-subsidizing behaviour of recreational fishing has not been much explored but should be viewed as an important issue in the context of management of coastal fish resources. Studies have examined the motivations for and satisfaction of fishing (Holland and Ditton, 1992; Arlinghaus, 2006; Cooke et al., 2018), the value of recreational fishing (Toivonen et al., 2004; Lawrence, 2005; Lovell et al., 2016; Hyder et al., 2018) as well as the impact and ecological effects of this activity (Schroeder and Love, 2002; Coleman et al., 2004; Cooke and Cowx, 2004; Lewin et al., 2006; Ihde et al., 2011; Fernández-Chacón et al., 2017; Nillos-Kleiven et al., 2019). These are all important factors for understanding the driving forces—and effects of the self-subsidizing behaviour of recreational fishers. Regulating recreational fisheries through input control measures, such as open season duration and gear restrictions, have its limitations. This has been well demonstrated in the Gulf of Mexico red snapper fishery where the recreational fishing sector’s adaptability to new regulations has led to a fishing season lasting for 9 days only in 2014 (Abbott et al., 2018). Therefore, input control and effort regulations alone for the purpose of securing fish populations might be wrought with problems in a highly diverse and adaptable recreational fishery. Arlinghaus et al. (2019) suggested some solutions to the management challenges in recreational fisheries. An important point in their argument is that the quantity of recreational fishing privileges needs to be limited and in line with biological management targets. In our point of view, this needs to include output control, such as total allowable catch (TAC). Due to the innate self-subsidizing forces in recreational fisheries, there is reason to argue that TAC is even more important in recreational- than in un-subsidized commercial fisheries. However, incorporating TAC, including quota allocations, in recreational fisheries is not a straightforward process and might create resistance. Especially in open-access coastal fisheries where resources are shared with the commercial fishing sector (Borch, 2010). To secure sustainability, there is an urgent need to impose management regulations that handle the self-subsidizing forces in the recreational fishery. This might not be possible without strong output controls, such as TACs. In the meantime, there are clear indications that recreational fisheries are increasing their relative impact on many over-harvested populations. Acknowledgements We thank three anonymous reviewers for valuable suggestions that improved the final version of the manuscript. Funding This work was supported by the Research Council of Norway (project numbers: 178376/S40 and 267808/E40). U.R. Sumaila thanks the Pew Charitable Trusts and the Sea Around US project for support. References Abbott J. K. , Lloyd-Smith P. , Willard D. , Adamowicz W. 2018 . Status-quo management of marine recreational fisheries undermines angler welfare . Proceedings of the National Academy of Science , 115 : 8948 – 8952 . Google Scholar Crossref Search ADS WorldCat Arlinghaus R. 2006 . On the apparently striking disconnect between motivation and satisfaction in recreational fishing: the case of catch orientation of German anglers . North American Journal of Fisheries Management , 26 : 592 – 605 . Google Scholar Crossref Search ADS WorldCat Arlinghaus R. , Abbott J. K. , Fenichel E. P. , Carpenter S. R. , Hunt L. M. , Alós J. , Klefoth T. et al. 2019 . Governing the recreational dimension of global fisheries . Proceedings of the National Academy of Science , 16 : 5209 – 5213 . Google Scholar Crossref Search ADS WorldCat Beckley L. E. , Fennessy S. T. , Everett B. I. 2008 . Few fish but many fishers: a case study of shore-based recreational angling in a major South African estuarine port . African Journal of Marine Science , 30 : 11 – 24 . Google Scholar Crossref Search ADS WorldCat Borch T. 2010 . Tangled lines in New Zealand’s quota management system: the process of including recreational fisheries . Marine Policy , 34 : 655 – 662 . Google Scholar Crossref Search ADS WorldCat Coleman F. C. , Figueira W. F. , Ueland J. S. , Crowder L. B. 2004 . The impact of United States recreational fisheries on marine fish populations . Science , 305 : 1958 – 1960 . Google Scholar Crossref Search ADS PubMed WorldCat Cooke S. J. , Cowx I. G. 2004 . The role of recreational fishing in global fish crisis . BioScience , 54 : 857 – 859 . Google Scholar Crossref Search ADS WorldCat Cooke S. J. , Twardek W. M. , Lennox R. J. , Zolderdo A. J. , Bower S. D. , Gutowsky L. F. G. , Danylchuk A. J. et al. 2018 . The nexus of fun and nutrition: recreational fishing is also about food . Fish and Fisheries , 19 : 201 – 224 . Google Scholar Crossref Search ADS WorldCat Fernández-Chacón A. , Moland E. , Espeland S. H. , Kleiven A. R. , Olsen E. M. 2017 . Causes of mortality in depleted populations of Atlantic cod estimated from multi-event modelling of mark-recapture and recovery data . Canadian Journal of Fisheries and Aquatic Science , 74 : 116 – 126 . Google Scholar Crossref Search ADS WorldCat Francis R. C. , Hixon M. A. , Clarke M. E. , Murawski S. A. , Ralston S. 2007 . Fisheries management—ten commandments for ecosystem-based fisheries scientists . Fisheries , 32 : 217 – 233 . Google Scholar Crossref Search ADS WorldCat Gell F. R. , Roberts C. M. 2003 . Benefits beyond boundaries: the fishery effects of marine reserves . Trends in Ecology & Evolution , 18 : 448 – 455 . Google Scholar Crossref Search ADS WorldCat Halpern B. S. 2003 . The impact of marine reserves: do reserves work and does reserve size matter? Ecological Applications , 13 : S117 – 137 . Google Scholar Crossref Search ADS WorldCat Hilborn R. 2007 . Moving to sustainability by learning from successful fisheries . AMBIO , 36 : 296 – 303 . Google Scholar Crossref Search ADS PubMed WorldCat Holland S. M. , Ditton R. B. 1992 . Fishing trip satisfaction: a typology of anglers . North American Journal of Fisheries Management , 12 : 28 – 33 . Google Scholar Crossref Search ADS WorldCat Hunt L. M. , Arlinghaus R. , Lester N. , Kushneriuk R. 2011 . The effects of regional angling effort, angler behaviour, and harvesting efficiency on landscape patterns of overfishing . Ecological Applications , 21 : 2555 – 2575 . Google Scholar Crossref Search ADS PubMed WorldCat Hyder K. , Weltersbach M. S. , Armstrong M. , Ferter K. , Townhill B. , Ahvonen A. , Arlinghaus R. et al. 2018 . Recreational sea fishing in Europe in a global context—participation rates, fishing effort, expenditure, and implications for monitoring and assessment . Fish and Fisheries , 19 : 225 – 243 . Google Scholar Crossref Search ADS WorldCat Ihde T. F. , Wilberg M. J. , Loewensteiner D. A. , Secor D. H. , Miller T. J. 2011 . The increasing importance of marine recreational fishing in the US: challenges for management . Fisheries Research , 108 : 268 – 276 . Google Scholar Crossref Search ADS WorldCat Jackson J. B. C. , Kirby M. X. , Berger W. H. , Bjorndal K. A. , Botsford L. W. , Bourque B. J. , Bradbury R. H. et al. 2001 . Historical overfishing and the recent collapse of coastal ecosystems . Science , 293 : 629 – 638 . Google Scholar Crossref Search ADS PubMed WorldCat Kleiven A. R. , Fernandez-Chacon A. , Nordahl J.-H. , Moland E. , Espeland S. H. , Knutsen H. , Olsen E. M. 2016 . Harvest pressure on coastal cod (Gadus morhua) from recreational fishing assessed from tag-recovery data . PLoS One , 11 : e0149595. Google Scholar Crossref Search ADS PubMed WorldCat Kleiven A. R. , Olsen E. M. , Vølstad J. H. 2012 . Total catch of a red-listed marine species in an order of magnitude higher than official catch . PLoS One , 7 : e31216 . Google Scholar Crossref Search ADS PubMed WorldCat Lawrence K. S. 2005 . Assessing the value of recreational sea angling in South West England . Fisheries Management and Ecology , 12 : 369 – 375 . Google Scholar Crossref Search ADS WorldCat Lewin W. C. , Arlinghaus R. , Mehner T. 2006 . Documented and potential biological impacts of recreational fishing: insights for management and conservation . Reviews in Fisheries Science , 14 : 305 – 367 . Google Scholar Crossref Search ADS WorldCat Lovell S. J. , Hilger J. , Steinback S. , Hutt C. 2016 . The Economic Contribution of Marine Angler Expenditures on Durable Goods in the United States, 2014. U.S. Dep. Commerce, NOAA Tech. Memo. NMFS-F/SPO-165, 72 p. Myers R. A. , Worm B. 2003 . Rapid worldwide depletion of predatory fish communities . Nature , 423 : 280 – 283 . Google Scholar Crossref Search ADS PubMed WorldCat Nillos-Kleiven P. J. , Espeland S. H. , Olsen E. M. , Abesamis R. A. , Moland E. , Kleiven A. R. 2019 . Fishing pressure impacts the abundance gradient of European lobsters across the borders of a newly established marine protected area . Proceedings of the Royal Society B: Biological Sciences , 286 : 20182455. Google Scholar Crossref Search ADS WorldCat Pauly D. 1995 . Anecdotes and the shifting baseline syndrome in fisheries . Trends in Ecology and Evolution , 10 : 430. 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. , Christensen V. , Guénette S. , Pitcher T. J. , Sumaila U. R. , Walters C. J. , Watson R. et al. 2002 . Towards sustainability in world fisheries . Nature , 418 : 689 – 695 . Google Scholar Crossref Search ADS PubMed WorldCat Pauly D. , Palomares M. L. D. 2010 . An Empirical Equation to Predict Annual Increases in Fishing Efficiency. Fisheries Centre Working Paper No. 7, University of British Columbia, Vancouver, 12 p. Post J. R. , Sullivan M. , Cox S. , Lester N. P. , Walters C. J. , Parkinson E. A. , Paul A. J. et al. 2002 . Canada’s recreational fisheries: the invisible collapse . Fisheries , 27 : 6 – 17 . Google Scholar Crossref Search ADS WorldCat Schroeder D. M. , Love M. S. 2002 . Recreational fishing and marine fish populations in California . California Cooperative Oceanic Fisheries Investigations Reports , 43 : 182 – 190 . WorldCat Sumaila U. R. , Khan A. , Watson R. Munro, G., Zeller, D., Baron, N., and Pauly, D. 2007 . The World Trade Organization and global fisheries sustainability . Fisheries Research , 88 : 1 – 4 . Google Scholar Crossref Search ADS WorldCat Sumaila U. R. , Teh L. , Watson R. , Tyedmers P. , Pauly D. 2008 . Fuel price increase, subsidies, overcapacity, and resource sustainability . ICES Journal of Marine Science , 65 : 832 – 840 . Google Scholar Crossref Search ADS WorldCat Toivonen A.-L. , Roth E. , Navrud S. , Gudbergsson G. , Appelblad H. , Bengtsson B. , Tuunainen P. et al. 2004 . The economic value of recreational fisheries in Nordic countries . Fisheries Management and Ecology , 11 : 1 – 14 . Google Scholar Crossref Search ADS WorldCat Worm B. , Barbier E. B. , Beaumont N. , Duffy J. E. , Folke C. , Halpern B. S. , Jackson J. B. C. et al. 2006 . Impacts of biodiversity loss on ocean ecosystem services . Science , 314 : 787 – 790 . Google Scholar Crossref Search ADS PubMed WorldCat © International Council for the Exploration of the Sea 2019. 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)

Endo, Clarissa Akemi, Kajiya;Vikebø, Frode, B;Yaragina, Natalia, A;Hjøllo, Solfrid, Sætre;Stige, Leif, Christian

doi: 10.1093/icesjms/fsaa057pmid: N/A

Abstract The spatial distribution of fish early life stages can impact recruitment at later stages and affect population size and resilience. Northeast Arctic (NEA) cod spawning occurs along the Norwegian coast. Eggs, larvae, and pelagic juveniles drift near-surface towards the Barents Sea nursery area. In this study, a 35-year long time series of NEA cod larvae data was analysed in combination with factors that potentially may affect the distribution of eggs and larvae. These factors included biological aspects of the spawning stock, and environmental variables, such as water temperature, wind, ocean current, and prey abundance. Our aim was to shed light on how these factors influence larval abundance and distribution and how larval abundance and distribution influenced recruitment at age 3. We found that biomass and mean weight of the spawners were positively associated with larval abundance and that a high liver condition index of the spawners was associated with a north-easterly distribution of the larvae. The environmental variables showed generally weak or no correlations with abundance or distribution of larvae. Lastly, we found significant association between larval abundance and year-class abundance at age 3, while the spatial distribution metrics of the larvae, i.e. distribution extent, mean longitude, and mean latitude, showed no significant association with future year-class abundance. Introduction Recruitment variability is regarded as one of the main causes for the observed fluctuations of fish stock abundance over time. According to Hjort (1914), the fluctuations observed in the stocks cannot only be attributed to a constant and regular factor, such as fisheries, but are likely also dependent on variable natural conditions. To explain the variability in year-class success, Hjort (1914) proposed that the dispersal of fish to unfavourable areas during the early life stages would impair the recruitment of fish to fisheries. This means that the spatial distribution of the early life stages has an impact on the survival to later stages, also known as the “aberrant drift” hypothesis (Houde, 2008). Understanding the factors that explain recruitment variability still remains as one of the biggest challenges in fisheries research (Ottersen et al., 2014). The Northeast Arctic (NEA) stock of Atlantic cod (Gadus morhua) is among the largest and most studied cod stocks (Yaragina et al., 2011). Atlantic cod is a long-lived species, with high fecundity and large fluctuation in year-class strength. NEA cod is known to perform long reproductive migrations from its feeding grounds in the Barents Sea towards the spawning banks along the Norwegian coast, between Finnmark (71°N) and Møre (63°N) with highest spawning activity around the Lofoten region (69°N) (Opdal et al., 2008). After the spawning period, the eggs and larvae drift north and east towards the Barents Sea. The pelagic drift follows the Norwegian Coastal current on the shelf and the Norwegian Atlantic current at the shelf-break and more offshore areas (Vikebø et al., 2005). The individual drift routes of early life stages of fish result in highly variable biotic and abiotic exposure, and corresponding survival probabilities (Vikebø et al., 2007; Putman et al., 2016). The potential influence of environmental conditions is especially large for fish that are spawned far away from the nursery areas as they drift for longer distances and experience variable environments over longer periods of time, as is the case for the NEA cod. The early life stages of NEA cod drift between 600 and 1200 km during their pelagic phase from spawning between early March and late April until bottom settlement from September to October in the Barents Sea (Vikebø et al., 2005; Ottersen et al., 2014). During their pelagic phase eggs and larvae of NEA cod drift in the upper mixed layer, being susceptible to temporally and spatially varying climatic conditions that might affect the strength and duration of the flow (Vikebø et al., 2007), and varying temperature, turbulence, prey availability, and predation (Ottersen et al., 2014). The biological condition of the spawners and the demographic structure of the spawning stock also have impacts on the distribution and survival of the early life stages of fish. Demography of the spawning stock, through variable age or size structure, is reported to influence spawning location, time and duration (Kjesbu, 1994; Jørgensen et al., 2008; Opdal and Jørgensen, 2015; Langangen et al., 2019). It has previously been shown that high average age and size of the spawning stock of NEA cod result in high egg abundance and widespread egg distribution (Stige et al., 2017). A wide egg distribution increases the diversity in conditions experienced by early life stages of a fish cohort, increasing the probability that a significant fraction of the cohort reaches the recruitment stage (Brunel, 2010). Moreover, the spawning locations are influenced by other processes, e.g. climate variability (Sundby and Nakken, 2008). Here, we use a combination of time-series and spatiotemporal statistical analysis, to elucidate how demographic factors and environmental conditions influence recruitment dynamics through effects on the spatial distribution and abundance of NEA cod larvae. We have two main questions: (i) what are the effects of the demography of the adult population and of the environment on NEA cod larvae abundance and distribution? (ii) What are the effects of larval abundance and distribution on recruitment at age 3? Material and methods A spatiotemporal dataset of NEA cod larvae for the period from 1959 to 1993 was used to construct seasonal indexes of abundance and distribution of cod larvae for the Barents Sea. The indexes were calculated separately for spring (before day 150, i.e. 30 May) and summer (after day 150) encompassing every year in the dataset (further details on the construction of the indexes can be found in the Plankton data). These indexes were used as response variables in a temporal analysis of the interannual variation in cod larval abundance and distribution. We considered both the abundance and distribution of all cod larvae and the abundance and distribution of only those larvae that were ˃16 mm (hereafter termed “large larvae”). Large larvae analysis was carried out only for the summer, since there are very few NEA cod larvae ˃16 mm in spring. Previous analyses suggest that recruitment at age 3 is more strongly associated with the abundance of large larvae than the abundance of all larvae 3 years previously (Stige et al., 2015). Predictor variables were biological characteristics of the adult spawning stock and biotic and abiotic environmental conditions that the early life stages might have experienced until they settle to the bottom in the Barents Sea (Table 1). Subsequently, we investigated in more detail how the selected variables are associated with cod larval distribution using spatiotemporal statistical analysis. Finally, the annual NEA cod larval indexes were regressed against recruitment at age 3, the age when fish are considered to be recruited to the fisheries stock, to evaluate how larvae distribution and abundance affect recruitment to the adult stock. Table 1. Predictor variables considered for NEA cod larvae abundance and distribution. Variable . Description . SSB Spawning stock biomass [ln(tonnes)]—yearly index MW Mean weight of the spawners (kg)—yearly index LCI Liver condition index (%)—liver wet weight as percentage of total wet weight of cod between 41 and 70 cm in length for January and December on the year before spawning (Yaragina and Marshall, 2000)—yearly index Naup Calanus sp. nauplii abundance [ln(N)] yearly index in spring only Cop Calanus finmarchicus copepodites abundance [ln(N)] yearly index in summer only TSPR, TSUM Mean temperature (°C) averaged over the upper 50 m of the water column for the area sampled in the PINRO stations for spring and summer OCSPR, OCSUM Mean surface ocean current magnitude (m s−1) for the upper 50 m in the water column in between the isobaths of 300–500 m depth for spring and summer NESPR, NESUM North-easterly wind events (fraction of time, scaled from 0 to 1) stronger than 5 m s−1 and with wind events with duration of ˃3 days in the Lofoten region (69°N to 12°E) for spring and summer SWSPR, SWSUM South-westerly wind events (fraction of time, scaled from 0 to 1) stronger than 5 m s−1 and with wind events with duration of ˃3 days in the Lofoten region (69°N to 12°E) for spring and summer Variable . Description . SSB Spawning stock biomass [ln(tonnes)]—yearly index MW Mean weight of the spawners (kg)—yearly index LCI Liver condition index (%)—liver wet weight as percentage of total wet weight of cod between 41 and 70 cm in length for January and December on the year before spawning (Yaragina and Marshall, 2000)—yearly index Naup Calanus sp. nauplii abundance [ln(N)] yearly index in spring only Cop Calanus finmarchicus copepodites abundance [ln(N)] yearly index in summer only TSPR, TSUM Mean temperature (°C) averaged over the upper 50 m of the water column for the area sampled in the PINRO stations for spring and summer OCSPR, OCSUM Mean surface ocean current magnitude (m s−1) for the upper 50 m in the water column in between the isobaths of 300–500 m depth for spring and summer NESPR, NESUM North-easterly wind events (fraction of time, scaled from 0 to 1) stronger than 5 m s−1 and with wind events with duration of ˃3 days in the Lofoten region (69°N to 12°E) for spring and summer SWSPR, SWSUM South-westerly wind events (fraction of time, scaled from 0 to 1) stronger than 5 m s−1 and with wind events with duration of ˃3 days in the Lofoten region (69°N to 12°E) for spring and summer Environmental predictor variables and prey predictor variables were calculated for spring (SPR) and/or summer (SUM). Open in new tab Table 1. Predictor variables considered for NEA cod larvae abundance and distribution. Variable . Description . SSB Spawning stock biomass [ln(tonnes)]—yearly index MW Mean weight of the spawners (kg)—yearly index LCI Liver condition index (%)—liver wet weight as percentage of total wet weight of cod between 41 and 70 cm in length for January and December on the year before spawning (Yaragina and Marshall, 2000)—yearly index Naup Calanus sp. nauplii abundance [ln(N)] yearly index in spring only Cop Calanus finmarchicus copepodites abundance [ln(N)] yearly index in summer only TSPR, TSUM Mean temperature (°C) averaged over the upper 50 m of the water column for the area sampled in the PINRO stations for spring and summer OCSPR, OCSUM Mean surface ocean current magnitude (m s−1) for the upper 50 m in the water column in between the isobaths of 300–500 m depth for spring and summer NESPR, NESUM North-easterly wind events (fraction of time, scaled from 0 to 1) stronger than 5 m s−1 and with wind events with duration of ˃3 days in the Lofoten region (69°N to 12°E) for spring and summer SWSPR, SWSUM South-westerly wind events (fraction of time, scaled from 0 to 1) stronger than 5 m s−1 and with wind events with duration of ˃3 days in the Lofoten region (69°N to 12°E) for spring and summer Variable . Description . SSB Spawning stock biomass [ln(tonnes)]—yearly index MW Mean weight of the spawners (kg)—yearly index LCI Liver condition index (%)—liver wet weight as percentage of total wet weight of cod between 41 and 70 cm in length for January and December on the year before spawning (Yaragina and Marshall, 2000)—yearly index Naup Calanus sp. nauplii abundance [ln(N)] yearly index in spring only Cop Calanus finmarchicus copepodites abundance [ln(N)] yearly index in summer only TSPR, TSUM Mean temperature (°C) averaged over the upper 50 m of the water column for the area sampled in the PINRO stations for spring and summer OCSPR, OCSUM Mean surface ocean current magnitude (m s−1) for the upper 50 m in the water column in between the isobaths of 300–500 m depth for spring and summer NESPR, NESUM North-easterly wind events (fraction of time, scaled from 0 to 1) stronger than 5 m s−1 and with wind events with duration of ˃3 days in the Lofoten region (69°N to 12°E) for spring and summer SWSPR, SWSUM South-westerly wind events (fraction of time, scaled from 0 to 1) stronger than 5 m s−1 and with wind events with duration of ˃3 days in the Lofoten region (69°N to 12°E) for spring and summer Environmental predictor variables and prey predictor variables were calculated for spring (SPR) and/or summer (SUM). Open in new tab Plankton data Ichthyoplankton (NEA cod larvae) and copepod (Calanus sp. nauplii and Calanus finmarchicus copepodites) data were collected by Russian dedicated ichthyoplankton surveys by the Knipovich Polar Research Institute of Marine Fisheries and Oceanography (PINRO, currently the Polar Branch of the Russian Federal Research Institute of Fisheries and Oceanography) from 1959 to 1993 (Nesterova, 1990; Mukhina et al., 2003). The surveys were conducted twice a year: one survey in the spring (April/May)—except for the spring of 1967 due to technical problems—and one survey in the summer (June/July). The surveys covered the main drift areas of NEA cod early life stages, covering an area from ∼7 to 500 km off the coast in the area from 67°30′N to 74°30′N and from 4°E to 33°30′E (Figure 1—orange dots). Note that the more coastal and southern parts of the distribution of the early life stages of NEA cod were not sampled by the surveys (Ottersen et al., 2014). Figure 1. Open in new tabDownload slide Study area. PINRO research cruise stations (orange dots); wind point location (brown star); ocean surface current section (purple line); and surface temperature area (black dashed line). Grid used in the spatiotemporal statistical analysis (dashed grey lines). Main ocean surface circulation pattern in the Barents Sea and Norwegian Sea: Norwegian Coastal Current (green arrows), Norwegian Atlantic Current (red arrows), and Arctic Current (blue arrows). Figure 1. Open in new tabDownload slide Study area. PINRO research cruise stations (orange dots); wind point location (brown star); ocean surface current section (purple line); and surface temperature area (black dashed line). Grid used in the spatiotemporal statistical analysis (dashed grey lines). Main ocean surface circulation pattern in the Barents Sea and Norwegian Sea: Norwegian Coastal Current (green arrows), Norwegian Atlantic Current (red arrows), and Arctic Current (blue arrows). An egg net (IKS-80) with the mesh size of 0.505 mm and a diameter of 80 and 1.5 m height was used for the spring surveys sampling of ichthyoplankton. The sampling at each station consisted of one vertical haul, from the bottom or from a maximum depth of 500 m, and of 10-min horizontal tows from 0 to 25 m depths. In the summer surveys, two types of net were used: an IKS-80 net with the same configuration as for the spring surveys and a ring-trawl net with mesh size of 3 mm, a diameter of 1.5 and 3.0 m height. At each station, vertical hauls with both nets were taken, and 10-min horizontal tows were taken at 0 and 25 m depth with the IKS-80 net and at 50 m depth with the ring-trawl. The zooplankton was sampled using Juday plankton nets (37 cm diameter, 180 μm mesh). For further information on the Russian ichthyoplankton data, we refer to Mukhina et al. (2003) and, for the zooplankton data, we refer to Nesterova (1990) and Kvile et al. (2014). The larvae samples were identified to species, counted, measured to the nearest millimetre, and grouped into five different size intervals: 1–5, 6–10, 11–15, 16–20, and 21 mm or larger. For the purpose of the analysis performed in this study, we have grouped larvae in two groups, one composed of all larvae sizes and the other of larvae ˃16 mm, henceforth referred to as all cod larvae and large cod larvae, respectively. For 17% of the stations with non-zero cod larvae data, only a subsample of the total larvae sampled was length-measured. The size-fractioned abundance data at these stations were rescaled by multiplying with a correction factor defined as the total number of larvae at the station divided by the total number of size-measured larvae. Zooplankton abundance indexes were included among the environmental variables (Table 1), since zooplankton are main prey items for cod larvae (Sundby, 2000). Zooplankton indexes were calculated separately for spring and summer. Zooplankton data was divided into Calanus sp. nauplii (naup) and C. finmarchicus copepodites (cop). For spring, only the nauplii data was considered, while in summer, only the copepodite data were used, based on findings that copepod nauplii are main prey of first-feeding NEA cod larvae while larger cod larvae rely on larger zooplankton prey [reviewed by Ottersen et al. (2014)]. The nauplii and copepodite abundance indexes were calculated following the same procedure as for the cod larvae abundance data, described in the pre-processing of the data. The surveys sampling coverage varied in number, time, and extent among the years, but usually followed regular transects (Mukhina et al., 2003). Spring survey sampling dates varied between day of the year 83 (24 March) and day 150 (30 May), with mean day 128 (8 May). The summer surveys occurred between days 151 (31 May) and 216 (4 August), with mean day 176 (25 June). Adult cod data Spawning stock biomass (SSB, tonnes) and recruitment (number of individuals at age 3) data were obtained from the International Council for the Exploration of the Sea stock assessment report (ICES, 2018). SSB is calculated based on the state-space assessment model using both fisheries and research survey data. It is a sum across ages of age-specific products of stock number, weight, and proportion mature (ICES, 2018), equations 1 and 2. The biomass-weighted mean weight (MW) of the spawning stock was calculated from the same data using equation (3) (Stige et al., 2017): SSBaj=NajWajMaj,(1) SSBj=∑aSSBaj,(2) MWj=∑a=3a=15+WajSSBaj∑a=3a=15+SSBaj,(3) where a is the age, j is the year, N is the number, W is the weight, and M is the proportion mature. By weighting by biomass and not abundance of each age class, the MW index represents the sizes that dominate the spawning stock in terms of potential egg production. The liver condition index (LCI—%) used was calculated for cod measuring 41–70 cm sampled in January–December by PINRO in the year previous to spawning estimated according to equation (4) (Yaragina and Marshall, 2000): LCIj %=∑m=1m=12∑nliverwetweight∑ntotalwetweight*10012,(4) where n is the total number of observations for a given year (j), month (m), and size class. This index has a positive correlation to recruitment in the NEA cod stock (Marshall et al., 1999) and to cod egg distribution and abundance (Stige et al., 2017). Temperature and ocean current data The Regional Ocean Modeling System (ROMS) model is a three-dimensional baroclinic ocean circulation model with terrain following s-coordinates in the vertical (Shchepetkin and McWilliams, 2005). The Nordic Seas 4 km numerical ocean model hindcast (SVIM) archive is available as daily and monthly averaged outputs from an application of ROMS for the Norwegian and the Barents Sea with a resolution of 4 km and 32 sigma levels (Lien et al., 2013). Temperature (TEMP—°C) and ocean current speed (OC—m s−1) were extracted from the monthly averaged SVIM archive for the period 1960–1993. For each variable, annual indexes for spring (SPR) and summer (SUM) were calculated, resulting in the variables TEMPSPR, TEMPSUM, OCSPR and OCSUM. Note that spring was considered as the average from March through May and summer was the average from June through July. The temperature was averaged for the upper 50 m of the water column for the same region as the PINRO survey samples (Figure 1—area outlined by the black dot-dashed line). Average OC speeds were extracted in the Lofoten region (68.2°N–69.2°N latitude and 12.2°E–4.9°E longitude, Figure 1—purple solid line) following the bathymetry of the continental shelf break between the 300- and 500-m isobaths for the upper 50 m in the water column, corresponding to a key area for larval dispersal en route from the spawning grounds towards the nursery areas (Strand et al., 2017). Wind data The wind data were obtained from the MET Norway Reanalysis (NORA 10) downscaled from the European Reanalysis project (ERA-40) to a 10-km grid covering the Norwegian Sea, the North Sea, and the Barents Sea (Reistad et al., 2011). Winds were extracted at a point location (69°N–12°E; Figure 1—brown star) at the shelf edge off the Lofoten archipelago. This is a particular narrow part of the shelf where larval dispersal is highly dependent on wind direction and strength, potentially also resulting in off-shelf transport (Strand et al., 2017). Annual indexes for north-easterly (NE) and south-westerly (SW) winds for the period from 1959 to 1993 were calculated for spring (March–May) and summer (June–July). We only considered winds stronger than 5 m s−1 and with a duration of at least 3 days to cause significant variation in the predominant flow and egg and larval dispersal (Skarðhamar and Svendsen, 2005; Skagseth et al., 2015; Strand et al., 2017). The wind indexes (NESPR, NESUM, SWSPR, SWSUM) were defined as the fractions of time with winds above these thresholds (scaled from 0 to 1). Pre-processing of the data We quantified larval indexes of abundance and distribution (abundance-weighted mean latitude, longitude and distribution extent) for all cod larvae and for large cod larvae. Separate larval indexes for spring and summer of each response and predictor variable were calculated. Note that in the spring analysis of cod larval abundance and distribution, only environmental predictor variables for spring were considered. Whereas in the analysis of cod larval abundance and distribution in summer, environmental predictor variables for both spring and summer were considered, except for spring nauplii abundance, which would reduce the number of analysed years, leaving only 15 years, because of different missing years for spring and summer. As the number of samples (egg-net and ring-trawl hauls) varied between stations, we used station-aggregated data on mean larval abundance per sample in the analyses. The data (cod larvae, Calanus sp. nauplii and C. finmarchicus copepodites separately) were resampled to mean abundance per grid cell of 1° latitude × 3° longitude to account for spatial variability in the sampling. For each year and season (spring or summer), the natural logarithm of the average abundance (N) of NEA cod larvae (or nauplii or copepodites) across all grid cells was then calculated for all cod larvae and for the large larvae. We only used grid cells with consistent sampling across years (Supplementary Figure S1—grid cells marked with an X) in subsequent calculations and only included years with data from all these grid cells, resulting in time series of 21 years in spring and 22 years in summer with good spatial coverage (Supplementary Figure S2). For nauplii and copepodites, the indexes were calculated for the same years as used for the cod larvae indexes. Note that the mean abundance values of 0 (3 years, in spring only) were replaced by the minimum abundance to avoid the logarithm of 0. Mean annual latitude and longitude of occurrence in spring and summer were weighted by NEA cod larvae abundance for each grid cell. A distribution extent index (D, scaled from 0 to 1) was calculated annually for spring and summer, by ranking the grid cells from high to low abundance and counting how many cells represented a cumulative proportion of cod larvae above a threshold of 0.9. We tested how sensitive results were to the choice of threshold by also conducting the analyses with the alternative thresholds of 0.8 and 0.95 (results in Supplementary Table S1). The index was scaled to maximum 1 by dividing on the total number of grid cells. Time series analyses of predictor effects on larval abundance and distribution indexes The first objective of our analyses was to identify which factors explain year-to-year variability in indexes of cod larvae abundance and distribution. We considered 12 cod larval abundance and distribution indexes (i.e. abundance, mean latitude, mean longitude, and distribution extent for all cod larvae in spring and in summer for both all cod larvae and large cod larvae). For each larval index (Supplementary Figure S2), we conducted a multiple linear regression analysis with the adult cod data (Supplementary Figure S3) and the environmental variables as potential predictors (Supplementary Figure S4). Linear models were chosen based on a lack of significant nonlinear associations. The generic formula for the model was: Yt=β0+β1X1,t+β2X2,t+⋯+βnXn,t+εt,(5) where Yt refers to a given larval index for year t; X1, …, Xn are the potential predictor variables as given by Table 2; β0 is the intercept; β1, …, βn are coefficients for the effects of these predictor variables; and ε is the error term. Table 2. Predictor variables (columns) and response variables (rows) considered in analyses of predictor effects (marked with an X) on annual larval abundance and distribution (5). . SSB . MW . LCI . Naup . Cop . TSPR . TSUM . OCSPR . OCSUM . NESPR . NESUM . SWSPR . SWSUM . Larval indexes springa X X X X – X – X – X – X – Larval indexes summerb X X X – X – X X X X X X X . SSB . MW . LCI . Naup . Cop . TSPR . TSUM . OCSPR . OCSUM . NESPR . NESUM . SWSPR . SWSUM . Larval indexes springa X X X X – X – X – X – X – Larval indexes summerb X X X – X – X X X X X X X Each response variable was analysed separately. Acronyms are explained in Table 1. a Abundance [ln(N)], mean latitude, mean longitude, and distribution extent of all cod larvae. b Abundance [ln(N)], mean latitude, mean longitude, and distribution extent of all cod larvae and of large cod larvae. Open in new tab Table 2. Predictor variables (columns) and response variables (rows) considered in analyses of predictor effects (marked with an X) on annual larval abundance and distribution (5). . SSB . MW . LCI . Naup . Cop . TSPR . TSUM . OCSPR . OCSUM . NESPR . NESUM . SWSPR . SWSUM . Larval indexes springa X X X X – X – X – X – X – Larval indexes summerb X X X – X – X X X X X X X . SSB . MW . LCI . Naup . Cop . TSPR . TSUM . OCSPR . OCSUM . NESPR . NESUM . SWSPR . SWSUM . Larval indexes springa X X X X – X – X – X – X – Larval indexes summerb X X X – X – X X X X X X X Each response variable was analysed separately. Acronyms are explained in Table 1. a Abundance [ln(N)], mean latitude, mean longitude, and distribution extent of all cod larvae. b Abundance [ln(N)], mean latitude, mean longitude, and distribution extent of all cod larvae and of large cod larvae. Open in new tab The predictor variables (Tables 1 and 2) were selected by forward stepwise selection based on Akaike’s information criterion corrected (AICc) for small sample size (Hurvich and Tsai, 1989). Terms were added if leading to a reduction in the model’s AICc. If the difference in the AICc between two models was smaller than 2, we considered the two models to have similar statistical support and report the results of both models. Residuals were checked for outliers, deviations from normality and autocorrelation through normal quantile–quantile plots, and plots of the autocorrelation function (not shown). While several potential predictor variables were correlated (Supplementary Table S3), the correlations among the predictor variables in the selected models were not so high (variance inflation factors ≤1.79). Spatiotemporal statistical analysis of larval abundance and distribution In a second step of the analysis, we investigated the spatiotemporal variation in cod larval abundance in more detail. Varying-coefficient generalized additive models (GAMs) (Hastie and Tibshirani, 1993) were used to evaluate the effects of selected predictor variables on the spatial distribution of NEA cod larvae. In this analysis, we used as response variable all the spatiotemporal data for cod larvae abundance for all years (1959–1993, aggregated to average values for each grid cell, season and year). GAMs estimate potentially nonlinear effects as smooth functions; seasonal and spatial patterns can, for example be estimated as smooth functions of day of year, longitude, and latitude. Spatial varying-coefficient GAMs estimate complex interaction effects, in which the spatial pattern depends on one or more continuous variables that represent factors that influence distribution. Specifically, the model assumes the effect of each of these variables to be linear at any given location, but the coefficient for this linear effect to vary as a smooth function of longitude and latitude. The model may also include predictor effects that are spatially homogeneous, i.e. having the same coefficient value at all locations. The predictor variables in our varying-coefficient GAMs were chosen according to the best fitted models in the time series analyses. Specifically, predictor variables that significantly affected cod larval distribution variables were assumed to have spatially varying coefficients, whereas predictor variables that significantly affected larval abundance but not distribution were assumed to have spatially homogenous effects. The generic model formula is described by (6). ln(Nit)=α+f1(x1,t)+f2(x2,t)+⋯+g(dayi)+h0(loni, lati)+h1(loni, lati)·y1,t+h2(loni, lati)·y2,t+⋯+εit.(6) Here, Nit is mean larval abundance in grid cell i and time t; α is an intercept; f1, f2, …, g are smooth functions (i.e. natural cubic splines) of predictor variables with spatially homogeneous effects (here referred to by the generic variables x1, x2, …, and day of year); h0 gives the mean spatial pattern and h1, h2, … are two-dimensional smooth functions (i.e. tensor products of natural cubic spline basis functions) that give spatially varying coefficients for predictor variables y1, y2, …; and εit is a normal distributed error term. Zero data were added the minimum abundance value, and grid cells where there was no sampling were not used in the analysis. Time series analyses of associations between larval abundance and distribution indexes and recruitment In a third step of the analysis we investigated the associations between the larval indexes and recruitment. The relationship between recruitment of 3-year-old cod and the abundance and distribution of the same year-class as larvae 3 years earlier was evaluated by fitting a multi-linear regression model, where the predictor variables were the larval indexes calculated from the PINRO samples. The general formulation for this model was: Rt=γ0+γ1X1,t−3+γ2X2,t−3+γ3X3,t−3+γ4X4,t−3+εt,(7) where Rt is the recruitment at age 3 in year t; X1, X2, X3, X4 are cod larvae abundance, mean latitude, mean longitude, and distribution extent indexes for either all larvae spring, all larvae summer, or large larvae summer; and ε is the error term. Temperature, OC, and wind data were processed in MATLAB version R2019a. All statistical analyses were performed in R version 3.5.1, and the R-package mgcv version 1.8-28 ( Wood, 2017) was used to perform GAM analysis. Results Time series analyses of predictor effects on larval abundance and distribution indexes According to our analysis of the abundance of all cod larvae, the variability in abundance in spring is best explained as a positive function of SSB (Figure 2a). An alternative model with similar statistical support in terms of AICc suggests that the variation in cod larval abundance in spring can also be explained as a function of nauplii abundance (Supplementary Table S2). Note that SSB and nauplii abundance are positively correlated in the investigated period (Supplementary Table S3). In summer, the abundance of all cod larvae is explained as a function of the SSB, the MW of the spawners, and non-significantly (p > 0.05), by the spring OC speed (Figure 3a). Large larvae abundance is best explained as a function of the SSB and the MW of the spawners (Figure 4a). Alternative models for summer abundance of cod larvae include SWSUM and OCSUM instead of OCSPR for all larvae, and OCSPR for large larvae; however, these are non-significant (Supplementary Table S2). Figure 2. Open in new tabDownload slide Linear effects of the AICc-selected predictor variables (acronyms are stated in Table 1) on the abundance and distribution of total NEA cod larvae in spring. Equations for each model (a–c) can be found in italics in Supplementary Table S2—total NEA cod larvae—spring. Each row shows the selected model for one larval index, abundance (a), mean longitude (b), or distribution extent (c). Each panel shows the partial effect of one variable, with partial residuals per year shown by the plotted numbers (= year—1900). Asterisks after the x-axis variable indicate statistical significance: *p < 0.05, **p < 0.01, ***p < 0.001. Figure 2. Open in new tabDownload slide Linear effects of the AICc-selected predictor variables (acronyms are stated in Table 1) on the abundance and distribution of total NEA cod larvae in spring. Equations for each model (a–c) can be found in italics in Supplementary Table S2—total NEA cod larvae—spring. Each row shows the selected model for one larval index, abundance (a), mean longitude (b), or distribution extent (c). Each panel shows the partial effect of one variable, with partial residuals per year shown by the plotted numbers (= year—1900). Asterisks after the x-axis variable indicate statistical significance: *p < 0.05, **p < 0.01, ***p < 0.001. Figure 3. Open in new tabDownload slide Linear effects of the AICc-selected predictor variables (acronyms are stated in Table 1) on the abundance and distribution of total NEA cod larvae in summer. Equations for each model (a–d) can be found in italics in Supplementary Table S2—total NEA cod larvae—summer. Each row shows the selected model for one larval index, abundance (a), mean latitude (b), mean longitude (c), or distribution extent (d). Each panel shows the partial effect of one variable, with partial residuals per year shown by the plotted numbers (= year—1900). Asterisks after the x-axis variable indicate statistical significance: *p < 0.05, **p < 0.01, ***p < 0.001. Figure 3. Open in new tabDownload slide Linear effects of the AICc-selected predictor variables (acronyms are stated in Table 1) on the abundance and distribution of total NEA cod larvae in summer. Equations for each model (a–d) can be found in italics in Supplementary Table S2—total NEA cod larvae—summer. Each row shows the selected model for one larval index, abundance (a), mean latitude (b), mean longitude (c), or distribution extent (d). Each panel shows the partial effect of one variable, with partial residuals per year shown by the plotted numbers (= year—1900). Asterisks after the x-axis variable indicate statistical significance: *p < 0.05, **p < 0.01, ***p < 0.001. Figure 4. Open in new tabDownload slide Linear effects of the AICc-selected predictor variables (acronyms are stated in Table 1) on the abundance and distribution of large NEA cod larvae in summer. Equations for each model (a–d) can be found in italics in Supplementary Table S2—large NEA cod larvae—summer. Each row shows the selected model for one larval index, abundance (a), mean latitude (b), mean longitude (c), or distribution extent (d). Each panel shows the partial effect of one variable, with partial residuals per year shown by the plotted numbers (= year—1900). Asterisks after the x-axis variable indicate statistical significance: *p < 0.05, **p < 0.01, ***p < 0.001. Figure 4. Open in new tabDownload slide Linear effects of the AICc-selected predictor variables (acronyms are stated in Table 1) on the abundance and distribution of large NEA cod larvae in summer. Equations for each model (a–d) can be found in italics in Supplementary Table S2—large NEA cod larvae—summer. Each row shows the selected model for one larval index, abundance (a), mean latitude (b), mean longitude (c), or distribution extent (d). Each panel shows the partial effect of one variable, with partial residuals per year shown by the plotted numbers (= year—1900). Asterisks after the x-axis variable indicate statistical significance: *p < 0.05, **p < 0.01, ***p < 0.001. The spatial distribution of larvae was divided into three indexes indicating their north–south (latitudinal) distribution, east–west (longitudinal) distribution, and an index indicating the general occupation across the sampling area (distribution extent) (Supplementary Figure S2). We found no significant explanatory variables for mean latitudinal variation in the distribution of all cod larvae in spring. In summer, mean latitude of all cod larvae and of large larvae is positively associated with the LCI of the spawners and negatively associated [though non-significantly (p > 0.05) for all cod larvae] with the OC speed in summer (Figures 3b and 4b, respectively). Alternative models suggest that temperature and SSB can replace OC speed as predictor for latitudinal distribution for, respectively, all cod larvae and large larvae in summer (Supplementary Table S2). Note that in summer, temperature in summer and and OC speed in summer are positively correlated during the time period analysed (Supplementary Table S3). The mean longitude of all cod larvae in spring was found to be best explained as a positive function of spring temperature (Figure 2c), with also indication of a non-significant association with SSB (Supplementary Table S2). The mean longitude of all larvae in summer was best explained as a positive function of the LCI of the spawners and a non-significant negative function of the abundance of copepodites (Figure 3c); or, in an alternative model with nearly identical statistical support, as a positive function of SSB and a negative function of abundance of copepodites (Supplementary Table S2). The mean longitude of large larvae was best explained as a positive function of the LCI of the spawners and a negative function of the SW winds in spring (Figure 4c). The NEA cod larvae distribution extent in spring was best explained by the abundance of nauplii (Figure 2c) and, in summer, as functions of the LCI of the spawners, OC speed, and non-significantly, summer temperature (Figure 3d). There was also indication of a non-significant association of summer distribution extent with the MW of the spawners (Supplementary Table S2). Large cod larvae distribution extent was associated positively with the SSB and non-significantly with the MW of the spawners (Figure 4d). LCI and OC speed were selected as predictors of distribution extent of large larvae in summer, if the distribution index represented the distribution of 80% rather than 90% of the larvae; otherwise the choice of threshold for the distribution index had only minor effects on results (Supplementary Table S1). Spatiotemporal statistical analysis of larval abundance and distribution From the previous analysis, we found that SSB and nauplii abundance were the predictor variables that showed strongest association with cod larvae abundance in spring. In the summer for both all larvae and for large larvae, the most consistent predictor variables for abundance and distribution were MW of spawners, SSB, and the LCI of the spawners. The selected predictor variables were used in the varying-coefficient GAM analysis, which showed in more detail how the distribution of cod larvae changed in response to the selected variables. For the spring, our model shows that abundance tends to increase with increasing SSB (Figure 5a). We further observe that in years with low nauplii abundance, the cod larvae in spring tend to be few in all parts of the study area (Figure 5b). When there is higher abundance of nauplii, the cod larvae distribution is expanded to offshore areas, mostly in the Norwegian Sea (Figure 5c), where the Norwegian Atlantic current splits in two branches. Day of the year for survey sampling was not included in the final spring model, because it suggested initially (data not shown) a decrease in cod larvae abundance until mid-spring, which is unlikely to be real, and we believe that this is due to sampling effect. Survey cruises normally started farther south along the Norwegian coast and moved towards the northeast into the Barents Sea (Supplementary Figure S1), which may influence the effect of day of the year in the abundance sampling. Figure 5. Open in new tabDownload slide Spatial patterns of all cod larvae abundance in spring estimated by a varying-coefficient GAM. The broken lines show 95% nominal confidence bands (not accounting for possible residual spatial autocorrelation). (a) Effect of the SSB on all cod larvae abundance in spring. (b, c) Spatial association between Calanus sp. nauplii abundance and all cod larvae abundance in spring. The maps show predicted abundance of cod larvae in periods of low (5th percentile—b) and high (95th—c) abundance of nauplii. Note that red X represents no sampling in the grid cell. Figure 5. Open in new tabDownload slide Spatial patterns of all cod larvae abundance in spring estimated by a varying-coefficient GAM. The broken lines show 95% nominal confidence bands (not accounting for possible residual spatial autocorrelation). (a) Effect of the SSB on all cod larvae abundance in spring. (b, c) Spatial association between Calanus sp. nauplii abundance and all cod larvae abundance in spring. The maps show predicted abundance of cod larvae in periods of low (5th percentile—b) and high (95th—c) abundance of nauplii. Note that red X represents no sampling in the grid cell. The results for the distribution of large larvae in summer resembled the results for all larvae in summer (Figures 6 and 7). Larvae abundance increases to a maximum days around 180–190 (29 June–9 July), and then, it starts to decrease. When there is a low SSB (Figures 6b and 7b), there is also a low abundance of cod larvae and large larvae in the study area. With high SSB (Figures 6c and 7c), larvae tend to spread in most directions, especially closer to the coastal areas. A similar pattern can be observed with low and high MW of the spawning stock (Figures 6d and e and 7d and e). The overall abundance of larvae varies little between years with low and high LCI, but we see that in years with low LCI (Figures 6f and 7f), the larvae are concentrated a little farther south in the coastal regions 68–70°N compared to years with high LCI (Figures 6g and 7g). Figure 6. Open in new tabDownload slide Spatial patterns of all cod larvae abundance in summer estimated by a varying-coefficient GAM. The broken lines show 95% nominal confidence bands (not accounting for possible residual spatial autocorrelation). (a) Effect of the day of the year on all cod larvae abundance in summer. (b, c) Spatial association between the SSB and all cod larvae abundance. (d, e) Spatial association between MW of the spawners and all cod larvae abundance. (f, g) Spatial association between the liver condition index of the spawners and the all cod larvae abundance. The maps show predicted abundance of all cod larvae in the summer for periods of low (5th percentile—b, d, e) and high (95th—c, e, g) SSB or MW or liver condition index. Note that red X represents no sampling in the grid cell. Figure 6. Open in new tabDownload slide Spatial patterns of all cod larvae abundance in summer estimated by a varying-coefficient GAM. The broken lines show 95% nominal confidence bands (not accounting for possible residual spatial autocorrelation). (a) Effect of the day of the year on all cod larvae abundance in summer. (b, c) Spatial association between the SSB and all cod larvae abundance. (d, e) Spatial association between MW of the spawners and all cod larvae abundance. (f, g) Spatial association between the liver condition index of the spawners and the all cod larvae abundance. The maps show predicted abundance of all cod larvae in the summer for periods of low (5th percentile—b, d, e) and high (95th—c, e, g) SSB or MW or liver condition index. Note that red X represents no sampling in the grid cell. Figure 7. Open in new tabDownload slide Spatial patterns of large cod larvae abundance in summer estimated by a varying-coefficient GAM. The broken lines show 95% nominal confidence bands (not accounting for possible residual spatial autocorrelation). (a) Effect of the day of the year on large cod larvae abundance in summer. (b, c) Spatial association between the SSB and large cod larvae abundance. (d, e) Spatial association between MW of the spawners and all cod larvae abundance. (f, g) Spatial association between the liver condition index of the spawners and large cod larvae abundance. The maps show predicted abundance of large cod larvae in the summer for periods of low (5th percentile—b, d, e) and high (95th—c, e, g) SSB, MW, and liver condition index, respectively. Note that red X represents no sampling in the grid cell. Figure 7. Open in new tabDownload slide Spatial patterns of large cod larvae abundance in summer estimated by a varying-coefficient GAM. The broken lines show 95% nominal confidence bands (not accounting for possible residual spatial autocorrelation). (a) Effect of the day of the year on large cod larvae abundance in summer. (b, c) Spatial association between the SSB and large cod larvae abundance. (d, e) Spatial association between MW of the spawners and all cod larvae abundance. (f, g) Spatial association between the liver condition index of the spawners and large cod larvae abundance. The maps show predicted abundance of large cod larvae in the summer for periods of low (5th percentile—b, d, e) and high (95th—c, e, g) SSB, MW, and liver condition index, respectively. Note that red X represents no sampling in the grid cell. Time series analyses of associations between larval abundance and distribution indexes and recruitment The AICc selection for recruitment to the adult stock selected abundance of cod larvae as the main variable (Table 3). For summer, all larvae distribution extent index was also selected, but it was not significant. Table 3. Effect of the NEA cod larvae abundance (N) and distribution (D) on the recruitment at age 3. . Parameter estimates (SE) . ln(recruitment) . Intercept . Parameters . R2 . All larvae spring 13.97 (0.36) +0.28 (0.10) *ln(N) 0.287 All larvae summer 14.68 (0.65) +0.43 (0.12) **ln(N) − 6.07 (3.63) D 0.428 All larvae summer 13.66 (0.23) +0.28 (0.08) **ln(N) 0.377 Large larvae summer 13.75 (0.24) +0.22 (0.06) **ln(N) 0.387 . Parameter estimates (SE) . ln(recruitment) . Intercept . Parameters . R2 . All larvae spring 13.97 (0.36) +0.28 (0.10) *ln(N) 0.287 All larvae summer 14.68 (0.65) +0.43 (0.12) **ln(N) − 6.07 (3.63) D 0.428 All larvae summer 13.66 (0.23) +0.28 (0.08) **ln(N) 0.377 Large larvae summer 13.75 (0.24) +0.22 (0.06) **ln(N) 0.387 Model parameter estimates and standard errors (in parentheses), proportion of variance explained (R2) of the best model for recruitment, and the selected NEA cod larvae abundance and distribution predictors. Note spring and summer were analysed separately due to different year coverage in the time series. Significance levels: * p < 0.05 and ** p < 0.01 Open in new tab Table 3. Effect of the NEA cod larvae abundance (N) and distribution (D) on the recruitment at age 3. . Parameter estimates (SE) . ln(recruitment) . Intercept . Parameters . R2 . All larvae spring 13.97 (0.36) +0.28 (0.10) *ln(N) 0.287 All larvae summer 14.68 (0.65) +0.43 (0.12) **ln(N) − 6.07 (3.63) D 0.428 All larvae summer 13.66 (0.23) +0.28 (0.08) **ln(N) 0.377 Large larvae summer 13.75 (0.24) +0.22 (0.06) **ln(N) 0.387 . Parameter estimates (SE) . ln(recruitment) . Intercept . Parameters . R2 . All larvae spring 13.97 (0.36) +0.28 (0.10) *ln(N) 0.287 All larvae summer 14.68 (0.65) +0.43 (0.12) **ln(N) − 6.07 (3.63) D 0.428 All larvae summer 13.66 (0.23) +0.28 (0.08) **ln(N) 0.377 Large larvae summer 13.75 (0.24) +0.22 (0.06) **ln(N) 0.387 Model parameter estimates and standard errors (in parentheses), proportion of variance explained (R2) of the best model for recruitment, and the selected NEA cod larvae abundance and distribution predictors. Note spring and summer were analysed separately due to different year coverage in the time series. Significance levels: * p < 0.05 and ** p < 0.01 Open in new tab Discussion The results in this study contribute to the knowledge on the effects of demographic structure and environmental variables on the spatial distribution and abundance of cod larvae and on the implications of larval distribution and abundance for recruitment. We show that the demographic characteristics of the spawners were significantly associated with cod larvae abundance and distribution, while the environmental variables considered showed weaker associations. We also show that the abundance of cod larvae, but not spatial distribution metrics of the larvae, correlated significantly with the recruitment to the fisheries at age 3. The role of stock demography and the environment for NEA cod larval abundance and distribution Cod larvae abundance in spring and summer was higher when there was higher SSB, as expected from the higher egg production potential these years. Our results showed that SSB was similarly strongly correlated with the abundance of large larvae as with the abundance of all larvae indirectly suggesting that mean size of the larvae was independent of SSB. Also, other factors showed similar associations with the abundance of all larvae as with the abundance of large larvae, which are the larvae that seem to contribute most to recruitment (Stige et al., 2015). Large body size appears important for the survival of NEA cod juveniles through the first winter of life (Stige et al., 2019), possibly because large individuals have survival advantages compared to smaller individuals through reduced predation risk and increased tolerance of starvation and physical extremes (Miller et al., 1988; Bailey and Houde, 1989). The abundance of both total and large larvae in summer was positively associated with the MW of the spawners. Likewise, egg abundance from the same surveys was found to be higher when there was older and larger individuals in the spawning stock (Stige et al., 2017). Interestingly, the MW of the spawners was estimated to have a much stronger effect on larval than egg abundance. For a change in MW of the spawners from 3.75 to 7.16 kg, which represent 5 and 95% of MW, respectively, we estimated a 12-fold increase in larval abundance, compared to a fourfold increase in egg abundance (Stige et al., 2017). This result supports that the survival of eggs to larvae is higher when the MW of the spawners is high, consistent with large female cod producing larger and more viable eggs (Marteinsdottir and Begg, 2002). Note, however, that spawning of the NEA cod is a complex multilevel process prolonged in space and time. NEA cod are batch spawners with varying egg size per batch. According to laboratory investigations, egg size decreased from first to last batch and the egg dry weight decreased by ∼20–30%; then, the number of eggs liberated in each batch followed a dome‐shaped curve with time (Kjesbu, 1989). NEA cod are also multiple spawners that participate in spawning for several years. There are indications that younger fish spawn later (Jørgensen et al., 2008) and stay at spawning grounds during shorter periods than older ones (Kjesbu et al., 1996). Both mentioned biotic reasons could impact egg size, number of eggs released in different areas, and consequently larvae abundance and mortality. Cod larvae distribution and extent (mean latitude, mean longitude, and distribution extent) in summer, and in particular for large larvae, are more strongly correlated with the LCI of the spawners than with any other variable considered. The higher the LCI the more NE the larvae are distributed. For the Atlantic cod, liver is the primary reserve for lipids, being a good indicator of recent adult feeding opportunity and fecundity (Lambert and Dutil, 1997; Marshall et al., 1999; Yaragina and Marshall, 2000). The liver energy reserve also supports the spawning migration from the Barents Sea. For NEA cod, it has been found that the Kola section temperature was not consistently correlated with the annual mean LCI, but the annual mean liver index was non-linearly related to capelin stock biomass. Also, LCI and the frequency of occurrence of capelin in cod stomachs were positively associated (Yaragina and Marshall, 2000). We hypothesize that years with favourable feeding opportunities and hence good LCI are likely promoted by higher temperatures that has been proposed to shift feeding distribution to the northeast in the Barents Sea and spawning distribution to the northeast along the Norwegian coast (Sundby and Nakken, 2008). Abundance of cod larvae presented few significant associations with the environmental variables. Abundance of cod larvae in spring was, however, associated with the abundance of nauplii. High copepod nauplii abundance and high temperatures in spring have also been found to be positively associated with higher abundance of cod larvae in summer (Stige et al., 2015). Temperature may have an impact on food availability, growth rates, and development of cod larvae (Sundby, 2000), although not detected in our analysis. Temperature was not found to have significant correlation with the abundance of cod larvae, following the same patterns as found for the NEA cod eggs (Stige et al., 2017). Of the environmental variables, OC speed and the wind events were found to be associated with the distribution of NEA cod larvae. Cod eggs and larvae drift near-surface north- and eastwards towards the feeding habitat in the Barents Sea, following the main OCs, i.e. the Norwegian Coastal Current and part of the Norwegian Atlantic Current. Some larvae may be transported off-shelf by episodic events that may vary in frequency and timing from year to year, mostly depending on the occurrence of north-easterly winds and consequent advection of individuals off the shelf (Strand et al., 2017). Eggs and larvae positioning in the water column is also important (not available from the used survey data) since the correlation between winds and the near-surface OC weakens with depth up to 40 m (Strand et al., 2017). OC speed had a negative association with the distribution extent and the mean latitude of the larvae. Though this relation is counter intuitive since both the Norwegian coastal current and the Norwegian Atlantic Current flow northwards, it is possible that strengthened shelf flow is associated with south-westerly winds resulting in downwelling along the coast and Ekman-transport of larvae towards the shore. In fact, the Norwegian Atlantic current is reported to be stronger and confined closer to the Norwegian coast in years with strong south-westerly winds, which occurs during positive phases of the North Atlantic Oscillation (Blindheim, 2004). NEA cod larvae abundance and distribution and their relation to recruitment Similar to previous studies (e.g. Helle et al., 2000; Mukhina et al., 2003; Stige et al., 2015), we found that larval abundance was significantly correlated with recruitment. Recruitment correlated similarly strongly with the abundance of large larvae as with the total abundance of larvae, which is slightly different from previous results showing the highest correlation for large larvae (Stige et al., 2015). Our analysis differs from the previous, Stige et al. (2015), by only including years with good data coverage, which reduces the risk of sampling bias but also reduces the length of the time series and potentially the statistical power. As we found strong associations between MW in the spawning stock and larval abundance and between larval abundance and recruitment at age 3, our results suggest that high MW (larger individuals) in the spawning stock has positive effect on recruitment. Recruitment of NEA cod at age 3 is not generally higher in years with older and larger individuals in the spawning stock (Ottersen, 2008), suggesting that effects of spawning stock structure on egg and larval abundances are often overruled by other factors that influence recruitment. Part of the reason may be changes in which factors drive recruitment variability, as indicated by the finding that correlations between recruitment and environmental indexes and between recruitment and juvenile-abundance indexes change over time (Ottersen et al., 2013). For example, large egg size may hypothetically mainly increase survival under adverse environmental conditions, when a large yolk sac may enable the newly hatched larvae to survive longer without feeding (Nissling et al., 1998). Another hypothetical reason for the lack of association with recruitment is that many of the eggs and larvae in years with older and larger individuals in the spawning stock are located in areas where survival chances are low. Specifically, we found indications that the distribution extent of the larvae was higher in these years, as was also the case for the eggs (Stige et al., 2017); if survival prospects in marginal areas are lower than in central areas, the increased abundance of larvae will be partly offset by reduced survival of the larvae. Although mortality is higher during the first months of life, year-class strength and recruitment can also be affected considerably by processes taking place later, before age 3, e.g. cannibalism and predation by other inhabitants of the Barents Sea (Bogstad et al., 2016). Recruitment to the fisheries at age 3 was not significantly associated with any spatial distribution metrics of the cod larvae, that is distribution extent, mean longitude, and mean latitude. The lack of association with distribution extent is similar to what has been found for NEA cod eggs and questions the biological significance of a wide offspring distribution for increasing offspring survival (Stige et al., 2017). In contrast, for all larvae in summer there was indication of a negative rather than positive association between distribution extent of larvae and recruitment, after controlling for the effect of larval abundance (Table 3). While we should be cautious not to over-interpret this non-significant negative association, it could have a biological explanation by the marginal areas of the distribution being sub-optimal for the survival of larvae to recruitment. The lack of significant association of recruitment at age 3 with mean longitude and mean latitude fails to support a significant role of the “aberrant drift hypothesis” (Hjort, 1914; Houde, 2008) in explaining NEA cod recruitment. The “aberrant drift hypothesis” proposes that eggs and larvae that were transported to unfavourable areas would not be recruited to the fisheries stock, i.e. that recruitment variability of NEA cod can be mostly explained by how large fraction of the larvae reaches the Barents Sea nursery grounds to the north and east of the larval distribution (Ottersen et al., 2014). There may be several explanations to the non-significant associations, including inadequate statistical power, across-shelf transport of fish larvae from the offshore areas back onto the continental shelf (Strand et al., 2017), and trade-offs between favourable locations for larval and juvenile survival up until recruitment (Langangen et al., 2014). Caveats The lack of significant relations between the response variables and the predictors can also be a case of insufficient or inadequate data; similarly, some of the statistical associations may have arisen just by chance. Although care has been taken to correct for temporal and spatial sampling variability in the data, it is possible that the statistical power of the analysis performed is not sufficient to detect signals through noise, which is unavoidable when analysing survey data, mainly due to patchiness in marine population distribution (time and space). Moreover, it is possible that if other indexes had been used for distribution or environmental variables, different associations would be captured. The statistical findings of this study should therefore be considered hypotheses for future research; the possible mechanisms behind the associations can, for example be tested through using a coupled biophysical model. Conclusions Factors regulating the recruitment of fish stocks have been discussed for more than a 100 years, and there are still many unanswered questions (Ottersen et al., 2014). Recruitment and survival of cod larvae are the result of a combination of processes and mechanisms, ranging from the spawning stock characteristics to climate variables. Through the statistical analysis of long-term scientific monitoring data, our results contribute to disentangle the quantitative importance of some of these processes. Our results suggest that spawning stock structure, as measured by the MW of the spawners, has strong effects on larval abundance and distribution and that larval abundance but not distribution correlates significantly with recruitment at age 3. Also, SSB and LCI correlated significantly with larval abundance and/or distribution, while the investigated biotic and abiotic environmental factors showed weaker effect. These results underline the importance of ecological processes prior to spawning for offspring production, such as the availability of suitable prey for the spawners to build up their energy reserves. While the links from spawning stock structure to recruitment remain incompletely understood, the results presented here further motivate fisheries management strategies that support desirable age and size structures and thereby high reproductive potential of harvested fish populations. In sum, our study underlines that sustainable exploitation of the NEA cod stock requires that managers consider not only the biomass of the spawning stock but also the demographic structure and the biomass of prey species. In turn, this increases the chances of sustainability in the exploitation of the fish stocks under a varying and changing climate. Acknowledgements We thank the crew and researchers at the Knipovich Polar Research Institute of Marine Fisheries and Oceanography who have collected and processed the data used in this work. We also thank the reviewers for their comments and suggestions. Funding This work was supported by the Research Council of Norway (project number 280468) under the project “Drivers and effects of spatial shifts in early life stages of marine fish (SpaceShift)”. References Bailey K. M. , Houde E. D. 1989 . Predation on eggs and larvae of marine fishes and the recruitment problem . Advances in Marine Biology , 25 : 1 – 83 . Google Scholar Crossref Search ADS WorldCat Blindheim J. 2004 . Oceanography and climate. In The Norwegian Sea Ecosystem , pp. 65 – 96 . Ed. by Skjoldal H. R.. Tapir academic press , Trondheim . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Bogstad B. , Yaragina N. A., Nash R. D. M. 2016 . The early life-history dynamics of Northeast Arctic cod: levels of natural mortality and abundance during the first 3 years of life . Canadian Journal of Fisheries and Aquatic Sciences , 73 : 246 – 256 . Google Scholar Crossref Search ADS WorldCat Brunel T. 2010 . Age-structure-dependent recruitment: a meta-analysis applied to Northeast Atlantic fish stocks . ICES Journal of Marine Science , 67 : 1921 – 1930 . Google Scholar Crossref Search ADS WorldCat Hastie T. , Tibshirani R. 1993 . Varying-coefficient models . Royal Statistical Society: Series B (Methodological) , 55 : 757 – 779 . Google Scholar Crossref Search ADS WorldCat Helle K. , Bogstad B., Marshall C. T., Michalsen K., Ottersen G., Pennington M. 2000 . An evaluation of recruitment indices for Arcto-Norwegian cod (Gadus morhua L.) . Fisheries Research , 48 : 55 – 67 . Google Scholar Crossref Search ADS WorldCat Hjort J. 1914 . Fluctuation in the great fisheries of Northern Europe viewed in the light of biological research . Rapports et Proces-Verbaux des Reunions Conseil International pour l’Exploration de la Mer , 20 : 1 – 228 . OpenURL Placeholder Text WorldCat Houde E. D. 2008 . Emerging from Hjort’s shadow . Journal of Northwest Atlantic Fishery Science , 41 : 53 – 70 . Google Scholar Crossref Search ADS WorldCat Hurvich C. M. , Tsai C. L. 1989 . Regression and time series model selection in small samples . Biometrika , 76 : 297 – 307 . Google Scholar Crossref Search ADS WorldCat ICES. 2018 . ICES AFWG REPORT 2018 Report of the Arctic Fisheries Working Group (AFWG) . Ispra , Italy . 859 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Jørgensen C. , Dunlop E. S., Opdal A. F., Fiksen Ø. 2008 . The evolution of spawning migrations: state dependence and fishing-induced changes . Ecology , 89 : 3436 – 3448 . Google Scholar Crossref Search ADS PubMed WorldCat Kjesbu O. S. 1989 . The spawning activity of cod, Gadus morhua L . Journal of Fish Biology , 34 : 195 – 206 . Google Scholar Crossref Search ADS WorldCat Kjesbu O. S. 1994 . Time of start spawning in Atlantic cod (Gadus morhua) females in relation to vitellogenic oocyte diameter, temperature, fish length and condition . Journal of Fish Biology , 45 : 719 – 735 . Google Scholar Crossref Search ADS WorldCat Kjesbu O. S. , Solemdal P., Bratland P., Fonn M. 1996 . Variation in annual egg production in individual captive Atlantic cod (Gadus morhua) . Canadian Journal of Fisheries and Aquatic Sciences , 53 : 610 – 620 . Google Scholar Crossref Search ADS WorldCat Kvile K. Ø. , Dalpadado P., Orlova E., Stenseth N. C., Stige L. C. 2014 . Temperature effects on Calanus finmarchicus vary in space, time and between developmental stages . Marine Ecology Progress Series , 517 : 85 – 104 . Google Scholar Crossref Search ADS WorldCat Lambert Y. , Dutil J. D. 1997 . Can simple condition indices be used to monitor and quantify seasonal changes in the energy reserves of Atlantic cod (Gadus morhua)? Canadian Journal of Fisheries and Aquatic Sciences , 54 : 104 – 112 . Google Scholar Crossref Search ADS WorldCat Langangen Ø. , Färber L., Stige L. C., Diekert F. K., Barth J. M. I., Matschiner M., Berg P. R. et al. . 2019 . Ticket to spawn: combining economic and genetic data to evaluate the effect of climate and demographic structure on spawning distribution in Atlantic cod . Global Change Biology , 25 : 134 – 143 . Google Scholar Crossref Search ADS PubMed WorldCat Langangen Ø. , Stige L. C., Yaragina N. A., Ottersen G., Vikebø F. B., Stenseth N. C. 2014 . Spatial variations in mortality in pelagic early life stages of a marine fish (Gadus morhua) . Progress in Oceanography , 127 : 96 – 107 . Google Scholar Crossref Search ADS WorldCat Lien V. S. , Gusdal Y., Albretsen J., Melsom A., Vikebø F. B. 2013 . Evaluation of a Nordic Seas 4 km numerical ocean model archive. Fisken og Havet, 7: 79. Marshall C. T. , Yaragina N. A., Lambert Y., Kjesbu O. S. 1999 . Total lipid energy as a proxy for total egg production by fish stocks . Nature , 402 : 288 – 290 . Google Scholar Crossref Search ADS WorldCat Marteinsdottir G. , Begg G. A. 2002 . Essential relationships incorporating the influence of age, size and condition on variables required for estimation of reproductive potential in Atlantic cod Gadus morhua . Marine Ecology Progress Series , 235 : 235 – 256 . Google Scholar Crossref Search ADS WorldCat Miller T. J. , Crowder L. B., Rice J. A., Marschall E. A. 1988 . Larval size and recruitment mechanisms in fishes: toward a conceptual framework . Canadian Journal of Fisheries and Aquatic Sciences , 45 : 1657 – 1670 . Google Scholar Crossref Search ADS WorldCat Mukhina N. V. , Marshall C. T., Yaragina N. A. 2003 . Tracking the signal in year-class strength of Northeast Arctic cod through multiple survey estimates of egg, larval and juvenile abundance . Journal of Sea Research , 50 : 57 – 75 . Google Scholar Crossref Search ADS WorldCat Nesterova V. N. 1990 . Biomassa planktona na putiakh dreifa lichinok treski (spravochnyi material) [Plankton biomass along the drift route of cod larvae (reference material)]. PINRO, Murmansk, Russia. Nissling A. , Larsson R., Vallin L., Frohlund K. 1998 . Assessment of egg and larval viability in cod, Gadus morhua: methods and results from an experimental study . Fisheries Research , 38 : 169 – 186 . Google Scholar Crossref Search ADS WorldCat Opdal A. F. , Jørgensen C. 2015 . Long-term change in a behavioural trait: truncated spawning distribution and demography in Northeast Arctic cod . Global Change Biology , 21 : 1521 – 1530 . Google Scholar Crossref Search ADS PubMed WorldCat Opdal A. F. , Vikebø F. B., Fiksen Ø. 2008 . Relationships between spawning ground identity, latitude and early life thermal exposure in Northeast Arctic cod . Journal of Northwest Atlantic Fishery Science , 41 : 13 – 22 . Google Scholar Crossref Search ADS WorldCat Ottersen G. 2008 . Pronounced long-term juvenation in the spawning stock of Arcto-Norwegian cod (Gadus morhua) and possible consequences for recruitment . Canadian Journal of Fisheries and Aquatic Sciences , 65 : 523 – 534 . Google Scholar Crossref Search ADS WorldCat Ottersen G. , Bogstad B., Yaragina N. A., Stige L. C., Vikebø F. B., Dalpadado P. 2014 . A review of early life history dynamics of Barents Sea cod (Gadus morhua) . ICES Journal of Marine Science , 71 : 2064 – 2087 . Google Scholar Crossref Search ADS WorldCat Ottersen G. , Stige L. C., Durant J. M., Chan K. S., Rouyer T. A., Drinkwater K. F., Stenseth N. C. 2013 . Temporal shifts in recruitment dynamics of North Atlantic fish stocks: effects of spawning stock and temperature . Marine Ecology Progress Series , 480 : 205 – 225 . Google Scholar Crossref Search ADS WorldCat Putman N. F. , Lumpkin R., Sacco A. E., Mansfield K. L. 2016 . Passive drift or active swimming in marine organisms? Proceedings of the Royal Society B: Biological Sciences , 283 : 20161689 . Google Scholar Crossref Search ADS WorldCat Reistad M. , Breivik Ø., Haakenstad H., Aarnes O. J., Furevik B. R., Bidlot J. R. 2011 . A high-resolution hindcast of wind and waves for the North Sea, the Norwegian Sea, and the Barents Sea . Journal of Geophysical Research Oceans , 116 : 1 – 18 . OpenURL Placeholder Text WorldCat Shchepetkin A. F. , McWilliams J. C. 2005 . The regional oceanic modeling system (ROMS): A split-explicit, free-surface, topography-following-coordinate oceanic model . Ocean Modelling , 9 : 347 – 404 . Google Scholar Crossref Search ADS WorldCat Skagseth Ø ., Slotte A., Stenevik E. K., Nash R. D. M. 2015 . Characteristics of the Norwegian Coastal Current during Years with High Recruitment of Norwegian Spring Spawning Herring (Clupea harengus L.) . PLOS ONE , 10 : e0144117 . Google Scholar Crossref Search ADS PubMed WorldCat Skarðhamar J. , Svendsen H. 2005 . Circulation and shelf–ocean interaction off North Norway . Continental Shelf Research , 25 : 1541 – 1560 . Google Scholar Crossref Search ADS WorldCat Stige L. C. , Langangen Ø., Yaragina N. A., Vikebø F. B., Bogstad B., Ottersen G., Stenseth N. C. et al. . 2015 . Combined statistical and mechanistic modelling suggests food and temperature effects on survival of early life stages of Northeast Arctic cod (Gadus morhua) . Progress in Oceanography , 134 : 138 – 151 . Google Scholar Crossref Search ADS WorldCat Stige L. C. , Rogers L. A., Neuheimer A. B., Hunsicker M. E., Yaragina N. A., Ottersen G., Ciannelli L. et al. . 2019 . Density‐ and size‐dependent mortality in fish early life stages . Fish and Fisheries , 20 : 962 – 976 . Google Scholar Crossref Search ADS WorldCat Stige L. C. , Yaragina N. A., Langangen Ø., Bogstad B., Stenseth N. C., Ottersen G. 2017 . Effect of a fish stock’s demographic structure on offspring survival and sensitivity to climate. Proceedings of the National Academy of Sciences of the United States of America, 114 : 1347 – 1352 . Strand K. O. , Sundby S., Albretsen J., Vikebø F. B. 2017 . The Northeast Greenland Shelf as a Potential Habitat for the Northeast Arctic Cod . Frontiers in Marine Science , 4 : 1 – 14 . Google Scholar Crossref Search ADS WorldCat Sundby S. 2000 . Recruitment of Atlantic cod stocks in relation to temperature and advection of copepod populations . Sarsia , 85 : 277 – 298 . Google Scholar Crossref Search ADS WorldCat Sundby S. , Nakken O. 2008 . Spatial shifts in spawning habitats of Arcto-Norwegian cod related to multidecadal climate oscillations and climate change . ICES Journal of Marine Science , 65 : 953 – 962 . Google Scholar Crossref Search ADS WorldCat Vikebø F. , Jørgensen C., Kristiansen T., Fiksen Ø. 2007 . Drift, growth, and survival of larval Northeast Arctic cod with simple rules of behaviour . Marine Ecology Progress Series , 347 : 207 – 219 . Google Scholar Crossref Search ADS WorldCat Vikebø F. , Sundby S., Ådlandsvik B., Fiksen Ø. 2005 . The combined effect of transport and temperature on distribution and growth of larvae and pelagic juveniles of Arcto-Norwegian cod . ICES Journal of Marine Science , 62 : 1375 – 1386 . Google Scholar Crossref Search ADS WorldCat Wood S. N. 2017. Generalized. additive models: an introduction with R . CRC press . Yaragina N. A. , Aglen A., Sokolov K. M. 2011 . Cod. In The Barents Sea: Ecosystem, Resources, Management , pp. 225 – 270 . Ed. by Jakobsen T. and Ozhigin V. K.. Tapir Academic Press , Trondheim . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Yaragina N. A. , Marshall C. T. 2000 . Trophic influences on interannual and seasonal variation in the liver condition index of Northeast Arctic cod (Gadus morhua) . ICES Journal of Marine Science , 57 : 42 – 55 . Google Scholar Crossref Search ADS WorldCat © International Council for the Exploration of the Sea 2020. 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)