ICES Journal of Marine Science: Journal du Conseil

Marshall, Kristin N; Koehn, Laura E; Levin, Phillip S; Essington, Timothy E; Jensen, Olaf P

doi: 10.1093/icesjms/fsy152pmid: N/A

Abstract The appetite for ecosystem-based fisheries management (EBFM) approaches has grown, but the perception persists that implementation is slow. Here, we synthesize progress toward implementing EBFM in the United States through one potential avenue: expanding fish stock assessments to include ecosystem considerations and interactions between species, fleets, and sectors. We reviewed over 200 stock assessments and assessed how the stock assessment reports included information about system influences on the assessed stock. Our goals were to quantify whether and how assessments incorporated broader system-level considerations, and to explore factors that might contribute to the use of system-level information. Interactions among fishing fleets (technical interactions) were more commonly included than biophysical interactions (species, habitat, climate). Interactions within the physical environment (habitat, climate) were included twice as often as interactions among species (predation). Many assessment reports included ecological interactions only as background or qualitative considerations, rather than incorporating them in the assessment model. Our analyses suggested that ecosystem characteristics are more likely to be included when the species was overfished (stock status), the assessment is conducted at a science centre with a longstanding stomach contents analysis program, and/or the species life history characteristics suggest it is likely to be influenced by the physical environment, habitat, or predation mortality (short-lived species, sessile benthic species, or low trophic-level species). Regional differences in stomach contents analysis programs may limit the inclusion of predation mortality in stock assessments, and more guidance is needed on best practices for the prioritization of when and how biophysical information should be considered. However, our results demonstrate that significant progress has been made to use best available science and data to expand single-species stock assessments, particularly when a broad definition of EBFM is applied. Introduction Although ecosystem-based fisheries management (EBFM) is increasingly identified as a way to improve management outcomes, there is little consensus as to the extent to which management decisions are based on, or informed by, EBFM principles. On one hand, management bodies around the world have developed frameworks and policies that broaden considerations in fisheries management decisions to include the human and biophysical systems in which fisheries operate (FAO, 2003; Marine Strategy Framework Directive, 2008; NOAA, 2016). Even so, many have argued that the practice of EBFM has lagged despite the proliferation of EBFM frameworks (Arkema et al., 2006; Pitcher et al., 2009; Berkes, 2012; Cowan et al., 2012; Essington et al., 2016). Here we use the EBFM definition of Essington et al. (2016), “a holistic, place-based framework that seeks to sustain fisheries and other services that humans want and need by maintaining healthy, productive and resilient fishery systems.” Some authors suggest lags between the generation of EBFM thought and action are caused by a need to develop new data sources, analytical tools, and models (Hilborn, 2011; Cowan et al., 2012). However, Patrick and Link (2015a) argue that these challenges to EBFM have largely been resolved in developed countries, and now persist only as “myths”. At least in the US federal fisheries management system, decision-makers and stakeholders in several regions are open to implementing EBFM (Biedron and Knuth, 2016) and the federal fisheries management agency, National Oceanic and Atmospheric Administration, has committed to doing so (NOAA, 2016). Thus, it is possible that EBFM is occurring more often than typically acknowledged, but does not receive attention because it now part of “good practice”. That is, conventional management has incrementally evolved to include ecosystem considerations, but has not labelled such evolution as EBFM. We examine this thesis here. The data and models used for stock assessment, a cornerstone of conventional fisheries management in many areas of the world, have greatly expanded in scope and complexity, and may be one way in which ecosystem considerations inform management. Stock assessment models estimate stock abundance relative to reference points using data such as catch, abundance, life history parameters, and expert knowledge. Output from these models informs decisions about annual catch limits, and as such they are subjected to a great deal of scrutiny from scientists, managers, and stakeholders. While some fisheries scientists have developed approaches for estimating the status of fish stocks that includes environmental relationships or predation mortality (Maunder and Watters, 2003; Kuparinen et al., 2012; Methot and Wetzel, 2013), the degree to which this research has been transferred to assessment models used for management is unknown. We note that expanding single species assessment models to include more environmental context may be considered by some as an ecosystem approach to fisheries management (EAFM), rather than EBFM, which requires consideration of interacting fisheries or fish stocks (Patrick and Link, 2015a). However, in this article we include EAFM under the broader definition of EBFM and do not differentiate between them. A recent global review of stock assessment models found that very few (2%) incorporated data or parameters representing external drivers of productivity (Skern-Mauritzen et al., 2016). However, productivity is only one avenue through which stocks are connected to their environment, and parameters and data in the final assessment model is only one line of evidence in support of considering ecosystem context. Any review of how broader system information is used needs to identify all possible ways such information might be included in management advice in general, and stock assessments in particular. On one end of the continuum is explicit inclusion of external parameters driving key population vital rates into assessment models, such as a predation mortality parameter or environmental driver of recruitment. On the other end, is broader qualitative considerations that inform model development in less obvious ways. Qualitative data could influence management decisions, or quantitative information may be used indirectly in the stock assessment process. For example, Zador et al. (2017) outlined how ecosystem assessments have qualitatively informed decisions by the North Pacific Council. We sought to document how frequently ecosystem information has been incorporated in stock assessment reports in the United States and understand the conditions under which uptake of ecosystem information has occurred. We undertook this exercise aware that it is not reasonable to expect all stock assessment models to include all ecological drivers; the performance of stock assessments and management strategies that include such drivers are highly variable (Myers, 1998; Punt et al., 2014). Nonetheless, exploring cases where ecosystem information has not been included may indicate persistent barriers to implementing EBFM. To that end, we developed three hypotheses describing conditions that we thought could lead to stock assessment reports including ecosystem considerations. First, we hypothesized that assessments for stocks that were in an overfished status would be more likely to include additional ecosystem interactions. Our reasoning is that overfished status could lead to a sense of urgency, which has been suggested to increase the receptiveness to EBFM (Olsson et al., 2008). Additionally, overfished stocks may simply receive higher priority for development of a new assessment. New assessments may create opportunities to update older models, and an overfished status may lead to a desire to understand what caused the stock decline (or lack of recovery) and exploration of causative drivers within the stock assessment model. Furthermore, when stocks collapse, it is often due to combined effects of fishing and environmental variability (Essington et al., 2015; Pinsky and Byler, 2015). Second, we hypothesized that regional differences among National Marine Fisheries Service Fishery Science Centers conducting the assessments may influence how ecosystem information is considered. For example, centres that have longstanding stomach contents analysis programs may be more likely to produce assessments in which predation and diet information are included. Or, centres that have strong oceanography programs may be more likely to produce assessments with climate information. If data limit the development of ecosystem models (Mace, 2001; Hilborn, 2011; Cowan et al., 2012), the availability of such data may spur development of assessments that include novel information. A full assessment of data availability for all stocks considered in this analysis would be outside the scope of this article, however. We focus on regional differences in stomach contents programs because diet data provide information on one of the most common justifications of EBFM, namely that predator–prey interactions change population productivity and reference points (Link, 2002, 2010). Third, we suspected that inclusion of ecosystem considerations will depend on the life history characteristics of stocks. For example, forage species are typically short-lived, highly linked to the physical environment, and may be influenced by predation from higher trophic levels (Pikitch et al., 2012). Therefore, we might expect that stock assessments for forage species would be more likely to include information about environmental drivers or predation than a stock assessment for a high trophic level, longer-lived, generalist piscivorous predator. The goals of our synthesis are to gauge the current status of the use of ecosystem considerations in US assessments, provide examples that can serve as a reference for others seeking to expand the scope of assessments, and consider more broadly how ecosystem information can be used in the institutional context in which assessments occur. We suspect that all of these contextual factors could influence how stock assessment models for fish species evolve as EBFM continues to advance. Methods We reviewed 206 stock assessments conducted by NOAA Fisheries. We obtained a list of the most recent stock assessment for each Council-managed stock in federal waters through a data request to the NOAA Species Information System (SIS) database (a simplified public version of the portal is available at https://www.st.nmfs.noaa.gov/sisPortal/). The SIS database contains metadata on stock assessment models and stock status from 2000 to present. We controlled for variation in model complexity by evaluating reports that had, at a minimum, some sort of production model (assigned level 3 or higher in the database). Simple index-based assessments or per-recruit analyses which lacked an underlying population model were excluded. We examined the extent to which each stock assessment report incorporated information about the interaction of the target stock with its ecosystem and other fisheries. We characterized six types of interactions: interactions with habitat or habitat requirements, environmental or climate interactions, interactions with prey (hereafter referred to as diets), interactions with predators, bycatch of the target species in other fisheries, and bycatch of other species within the target species fishery. These topic areas cover a range of factors that could influence recruitment, growth, movement, or mortality, which are the processes that affect stock biomass and thus most likely to be included in assessments. We scored each category of ecosystem information on an ordinal scale from 1 to 3. A score of 1 was given when the topic was mentioned in the stock assessment report as background information on the species. We scored a report with a 2 for two cases: when quantitative data on the interaction were included in the report, but not used in any analyses, or when the author made an explicit link between the ecosystem category and assessment parameters or output. For example, including numerical data from diet studies on the target species would receive a score of 2, as would discussing a link between sea surface temperature and recruitment predictions. The highest score, 3, was given in cases when the category of information was explicitly included in the assessment model through data inputs or estimated parameters. It is unlikely that any report would score high in every category. Given the step-wise progression of most assessment models, new components are generally only added as needed, or desired, by the assessment working group or the stock assessment author. Moreover, higher scores are not intended to be a judgement of the quality of an assessment. In some cases, an initial screening of the available environmental variables may be sufficient to determine that inclusion of these variables in the stock assessment would not improve model performance. Thus, a model that includes these variables, which would receive a score of 3, is not necessarily more accurate or less biased than a model that does not (Punt et al., 2014). In some cases, ecosystem interactions were included in exploratory model runs, but not the final model used to develop management advice. Our scores reflect the level of consideration given to each category of ecosystem interaction as reflected in the final stock assessment report, not whether the final model used for decision-making included any of these factors. We did this out of a desire to record the consideration of new topics, not track the review process of new components of assessment models. Potential explanatory factors: stock status, availability of stomach contents data, and life history types We explored how characteristics of the target stocks and the context surrounding their management might influence their stock assessments by exploring three aspects. First, we categorized stock status based on its designation by NOAA during the period from 2001 to 2005. We chose this period because NOAA’s Fish Stock Sustainability Index (FSSI) began tracking overfished status in 2001, and the oldest assessment in our database was from 2006. If the stock was given an overfished status designation (defined on a stock by stock basis) during any one of those years, we considered it “overfished” for the purposes of this analysis. Second, we explored how regional differences in stomach contents analysis programs influenced the potential to include information on predation and diets of target stocks in assessment reports. The Northeast Fisheries Science Center (NEFSC, https://www.nefsc.noaa.gov/femad/pbb/fwdp/databases.html) and Alaska Fisheries Science Center (AFSC, https://access.afsc.noaa.gov/REEM/WebDietData/DietDataIntro.php) have long-standing stomach contents analysis programs and sampling as part of their annual surveys, while the other science centres have more opportunistic sampling and support for diet studies, if any. Third, we categorized each target stock as one of four ecological “types” that combine information about taxonomy, habitat, and functional role in the ecosystem: small pelagic fishes, groundfish, benthic invertebrates, or medium/large pelagic fishes. We evaluated statistical support for our hypotheses by comparing the number of stock assessment reports with scores of 2 or higher with the number of reports scoring 1 or lower in each category using non-parametric one-sided Mann-Whitney U-tests. Differences between the four ecological types were described by Wilcoxon Rank Sum tests and post-hoc two-sided tests with Bonferonni corrections for multiple comparisons. Results The quality and quantity of inclusion of the six fishery and ecosystem interactions within the 206 recent stock assessments varied dramatically (Figure 1). Fishery interactions, specifically bycatch of the target species (40% of assessments), were the most common interaction included in quantitative approaches. Quantitative incorporation of other interactions into assessments was less common. Twenty-four percent of assessment reports included at least one of the other ecosystem factors quantitatively: 11% of stock assessments included habitat, 14% included environmental or oceanographic conditions, and 1% included the effects of predation. Bycatch of other species and competition were never incorporated quantitatively. Figure 1. Open in new tabDownload slide Inclusion of fishery and ecosystem interactions across interaction types. Each bar represents the proportion of assessment reports that received each score across topics (n = 206). Shading increases with scores: background information (1), qualitative inclusion of information (2), or quantitative inclusion (3). Figure 1. Open in new tabDownload slide Inclusion of fishery and ecosystem interactions across interaction types. Each bar represents the proportion of assessment reports that received each score across topics (n = 206). Shading increases with scores: background information (1), qualitative inclusion of information (2), or quantitative inclusion (3). Most assessments that scored a 3 in one or more categories included ecosystem information to filter or correct observations of the assessed species in fishery dependent or independent surveys (Supplementary Table S1). Of 23 assessments that included habitat, 19 used habitat factors to filter survey observations or correct catchability. In those assessments, habitat was characterized by bottom depth, bottom type, or the presence of co-occurring species. Three assessments for invertebrate bivalves (Atlantic surfclam, ocean quahog, sea scallop) included total habitat area to inform the biomass estimate. One assessment (Gulf of Alaska demersal shelf rockfish) used the area of rocky habitat as a multiplier for densities observed in the survey. Twenty-nine assessment models quantitatively included climate, and did so in more diverse ways than for habitat. About half used temperature as a covariate for catchability or an index of abundance. Four salmon stock assessments used environmental covariates to forecast returns of adult fish. Five assessments used temperature or other environmental indices to predict recruitment. Growth was modelled as temperature dependent in one assessment (Bering Sea-Aleutian Islands yellowfin sole), and in another growth was time varying with phases of the Pacific Decadal Oscillation (PDO) (Southern Pacific Coast chilipepper rockfish). The Gulf of Mexico Gag assessment used an environmentally driven mortality parameter to account for red-tide events. Catches were assigned to the United States or Mexican fleets based on temperature (which influences spatial distribution of the stock) in the Pacific sardine assessment. Predation was included quantitatively in three assessment reports. Time-varying natural mortality informed by predator abundance was investigated for butterfish and Atlantic herring (but was not retained in the final model for either stock). Predator abundance was used as an indicator of year-class strength for shortbelly rockfish. The Atlantic herring assessment also investigated an index of an egg predator to predict recruitment. Fishery and ecosystem considerations in stock assessment reports were both qualitative and quantitative in nature. Qualitative elements were more common for diet, predation, and bycatch of other species. Quantitative approaches were more common for habitat, climate, and bycatch of the target species (Figure 1). For most of the categories, ecosystem interactions were most commonly incorporated as background. Habitat (68%) and predator (49%) interactions were most frequently included in background information. Bycatch of other species was mentioned in 30% of assessment reports and climate interactions were mentioned in 23% of the reports. Competition was rarely mentioned (5%), and we did not include it in the remaining graphs. Stock status Our hypothesis that overfished status may lead to increased inclusion of ecosystem information was only partly supported (Figure 2). Bycatch of the target species appeared more often in assessments for overfished species (U = 2655, p = 0.002). Inclusion of climate interactions was more common for overfished stocks (U = 2825, p = 0.004). Overfished status did not contribute to the inclusion of bycatch of other species, habitat, predation, or diet in stock assessment reports. Figure 2. Open in new tabDownload slide Stock assessments for species that were in an overfished status for some part of 2001–2005 had relatively higher scores on their assessments for accounting for bycatch of the target species and including climate interactions. Panels show scores by fishery or ecosystem interaction type: bycatch of the target species (a), bycatch of other species (b), habitat (c), climate (d), diet (e), and predation (f). Stacked bars show the proportion of scores within each category, with increased shading for higher scores (none = 0, background = 1, qualitative = 2, quantitative = 3). Figure 2. Open in new tabDownload slide Stock assessments for species that were in an overfished status for some part of 2001–2005 had relatively higher scores on their assessments for accounting for bycatch of the target species and including climate interactions. Panels show scores by fishery or ecosystem interaction type: bycatch of the target species (a), bycatch of other species (b), habitat (c), climate (d), diet (e), and predation (f). Stacked bars show the proportion of scores within each category, with increased shading for higher scores (none = 0, background = 1, qualitative = 2, quantitative = 3). Availability of diet data We found support for our hypothesis that information considered in stock assessments may reflect availability of data (Figure 3). The number of assessment reports scoring 2 or higher for diet and predation was greater in regions that have a long-standing on-site stomach contents lab. Diet was included at a score of 2 or higher in 23 assessments from science centres from these two regions compared with only 2 from elsewhere (U = 3706, p < 0.001). Predation was included at a score of 2 or higher in 22 assessments from Alaska and Northeast compared with 4 from elsewhere (U = 3849.5, p < 0.001). Quantitative incorporation of predation in assessment models was rare (3 out of 206 assessments). Figure 3. Open in new tabDownload slide The incorporation of diets and predation into stock assessments may be explained by data availability. Bar plots show the proportion of assessments that received each score as a function of the co-occurrence of a stomach contents lab at the science centre where the stock assessment was done. Panels show scores by fishery or ecosystem interaction type: diet (a) and predation (b). Stacked bars show the proportion of scores within each category, with increased shading for higher scores (none = 0, background = 1, qualitative = 2, quantitative = 3). Figure 3. Open in new tabDownload slide The incorporation of diets and predation into stock assessments may be explained by data availability. Bar plots show the proportion of assessments that received each score as a function of the co-occurrence of a stomach contents lab at the science centre where the stock assessment was done. Panels show scores by fishery or ecosystem interaction type: diet (a) and predation (b). Stacked bars show the proportion of scores within each category, with increased shading for higher scores (none = 0, background = 1, qualitative = 2, quantitative = 3). Target species life history Of all of the fishery and ecosystem interactions that might be included in stock assessments, only habitat, diet, and predation showed any relationship to stock life history characteristics (Figure 4). Habitat interactions were included more frequently for small pelagic species, demersal species, and invertebrate species compared with large pelagic species (statistics presented in Supplementary Table S2). Assessment reports for demersal species included diet information at a level 2 significantly more often than assessment reports on large pelagic species (Supplementary Table S2). Over 50% of assessments for small pelagic species incorporated predation at least qualitatively (though predation was included quantitatively for only three of these species). The proportion of assessments scoring 2 or higher was significantly higher for small pelagic species compared with demersal or large pelagic species (Supplementary Table S2). Assessment reports for large pelagic fishes had the lowest levels of inclusion of ecosystem and fishery interactions across all types. Figure 4. Open in new tabDownload slide Assessment scores for six categories of fishery and ecosystem interactions, separated by the ecological type of the assessed species. Panels show scores by fishery or ecosystem interaction type: bycatch of the target species (a), bycatch of other species (b), habitat (c), climate (d), diet (e), and predation (f). Stacked bars show the proportion of scores within each category, with increased shading for higher scores (none = 0, background = 1, qualitative = 2, quantitative = 3). Figure 4. Open in new tabDownload slide Assessment scores for six categories of fishery and ecosystem interactions, separated by the ecological type of the assessed species. Panels show scores by fishery or ecosystem interaction type: bycatch of the target species (a), bycatch of other species (b), habitat (c), climate (d), diet (e), and predation (f). Stacked bars show the proportion of scores within each category, with increased shading for higher scores (none = 0, background = 1, qualitative = 2, quantitative = 3). Discussion Our review of over 200 US stock assessments demonstrates more widespread inclusion of interactions among fisheries and with the ecosystem into the stock assessment process than previously reported. One quarter of the assessment models included at least one type of interaction between the assessed species and its ecosystem, especially physical drivers of habitat and climate. Diets and predation were less common, likely because of the paucity of detailed data on fish feeding in many areas of the United States. Together, these findings suggest that ecosystem factors are being considered in the stock assessment process, even if those factors are not specifically called out as EBFM. Moreover, when interactions are included quantitatively, they often involved the influence of habitat or environmental factors on catchability rather than trying to predict recruitment. We found a greater degree of inclusion of ecosystem considerations than the global review by Skern-Mauritzen et al. (2016), using our broader definitions of inclusion and ecosystem information types. The context surrounding ecosystem considerations in European (ICES) assessments they described is similar to what we found in the US context, however. Skern-Mauritzen et al. (2016) noted that inclusion of interactions has been primarily a bottom-up process, driven first by scientific support in the literature, then data availability, and then interest and inclusion in the assessment model. They also found that qualitative inclusion of ecosystem effects on stock productivity was more common than quantitative inclusion, although they did not quantify those differences. Their results and ours suggest that there are likely more opportunities to include and evaluate relationships between harvested species and their ecosystems moving forward. Given the examples we identified in the United States for expanding assessments to include more ecosystem considerations, an important next step will be to develop more formal recommendations for how and when to include ecosystem data and relationships. Identifying where and when specific assessments and management advice could be improved by ecosystem information will be case specific. Including ecosystem information in assessment models does not always improve the accuracy or predictive capacity of models (Punt et al., 2014). However, our findings on current uses of ecosystem information in assessments lead to some preliminary recommendations. Overfished species may be a universally higher needs category of assessment report that could benefit from ecosystem information. Failure of overfished stocks to recover is in some cases tied to overoptimistic assessment results, which can be the result of unaccounted for shifts in productivity. We found support for our hypothesis that overfished status may contribute to greater inclusion of ecosystem interactions, particularly for interactions with other fisheries (bycatch) and oceanographic conditions. Beyond environmental information, identifying other potentially changing sources of mortality could improve the understanding of the causes of population decline. Increasing predator abundance could lead to increasing trends in natural mortality, and documenting patterns of predation could lead to inferences about time-varying natural mortality, even if only qualitatively. Overfished status may also lead to additional scrutiny and a sense of urgency, ultimately supporting innovation of methods or data during the development of subsequent assessment models for that species. Research in product innovation suggests that creating a sense of urgency is a critical component in team dynamics that leads to higher levels of creativity and more competitive new technologies (Im et al. 2013). For species that are not overfished, one way to select assessments for potential inclusion of ecosystem information is a risk analysis and prioritization framework. This approach could triage species most likely to benefit from greater consideration of fishery and ecosystem interactions. For example, NOAA’s recent Stock Assessment Improvement Plan (NOAA 2018) recommends a simple framework for scoring species based on their ecosystem importance (trophic linkages), recruitment variability (likelihood of being linked to environmental driver), and habitat associations. Higher ecosystem importance scores would be given for top predators and dominant forage species, identifying their potential outsized importance in food web dynamics (Heithaus et al., 2008; Smith et al., 2011). High recruitment variability or changes in average recruitment (especially coinciding with high steepness values) may imply that recruitment processes are linked to environmental conditions (Szuwalski et al., 2015). Strong habitat associations might identify the potential for habitat covariates to improve how the observation process is modelled (e.g. catchability). These categories cover similar characteristics as the life history types we characterized, but in a more explicit process-oriented way (e.g. demersal species may have high or low recruitment variability, and some, but not all, have strong habitat associations). Implementing a qualitative scoring approach in each region may be useful to quickly screen species and identify the highest priority candidates for expanding the scope of assessments. A second phase of a risk analysis approach would be to add “ecosystem considerations” sections to assessment reports for those species identified as having high scores in the areas identified above. These sections are already standard for all species in some regions (e.g. AFSC). An ecosystem considerations section could describe the qualities of the stock that make its assessment and management likely to benefit from the inclusion of ecosystem information and identify potential data sources or indicators that could be explored for inclusion in the assessment model or to qualitatively inform the structure of the model and/or discussions on setting annual catch levels. This second phase could also identify data needs where information would be useful but is not currently available. For example, a productive, abundant forage fish with high recruitment variability identifies a potential need for ecosystem information, but environmental data or information on predators and their diets may or may not be available in that region. A third phase of fitting assessment models with new covariates or datasets would then only be employed in the limited set of cases where a potential need was identified and data available to inform the models. Governance and institutional challenges are identified as barriers to implementing EBFM (Hilborn et al., 2005; Bundy et al., 2008; Olsson et al., 2008) and some of these may occur within the stock assessment process itself, limiting further inclusion of ecosystem considerations in assessment models. For example, scepticism about new approaches is inherent to the process of science and particularly to EBFM (Hilborn, 2011). Moreover, fisheries science has been strongly influenced by statistical inference, where the goal is frequently to describe observed data with as simple a model as possible (Burnham and Anderson, 1998; Kuparinen et al., 2012). Any new models, data, or tools are also subject to reviews by Fishery Management Councils’ Science and Statistical Committee and outside reviewers. Together, these factors are designed to protect an important process influencing management decisions and ensure the use of “best available science”. An unintended consequence may be that this high burden of proof presents an obstacle to even positive changes. Developing new stock assessment models and data sources to inform them is a complex and creative scientific process. Research on creativity suggests that negative emotions (such as those created by negative feedback from reviewers, or fears generated by large changes in stock status) can motivate improvement, for which creativity is required (Rasulzada, 2014). But, stress (such as that created by being asked to produce results under very tight deadlines and in a public arena) can also reduce creativity by reducing cognitive resources (Fredrickson, 2004). Bureaucratic climates can threaten employee creativity by fostering a fear of failure and risk avoidance (Ford, 1996). Consideration of the institutional context surrounding stock assessments could create opportunities to improve the process. For example, exploring potential ecosystem expansions to assessment models first using management strategy evaluations (e.g. Punt et al., 2014) can provide some breathing room from the pressure associated with decisions on catch levels. Creating assessment teams that add scientists with expertise in ecological interactions, climate, habitat ecosystem to those with expertise in population dynamics will also encourage broader consideration of ecosystem information. Some regions have developed terms of reference for assessments that recommend consulting with or including ecosystem scientists (e.g. PFMC, 2016) or explicitly require ecosystem factors to be considered. For example, the 2014 butterfish assessment included the following term of reference (TOR): “3. Characterize oceanographic and habitat data as it pertains to butterfish distribution and availability. If possible, integrate the results into the stock assessment (TOR-5).” This TOR has two characteristics that we believe are helpful. First, it specifies a priority for inclusion of ecosystem information—in this case, “oceanographic and habitat data.” Second, it urges quantitative integration of these factors into the stock assessment model, but ultimately leaves the decision to include these factors in the final model up to the assessment scientists and working group. Because inclusion of ecosystem information and the associated additional parameters does not always make for a more a robust and reliable assessment (Myers 1988, Punt et al., 2014), the technical decision of whether it is best to include it should be left to the appropriate experts. One productive approach to expanding the use of ecosystem information in stock assessments is to develop separate “research” and “operational” tracks for stock assessments. Research-track assessments would have greater flexibility to innovate without being constrained by the tight timelines and need for demonstrated robustness associated with operational assessments and their formal review process. A mechanism would be needed to move successful innovations from the research track into the operational assessment. Currently, “benchmark” assessments provide some opportunity to innovate, but they are still constrained by the existing review process and intense assessment schedule. Expanding stock assessments to include more consideration of fishery and ecosystem interactions is only one way ecosystem considerations can influence the management process. Others may be equally or more influential. Stock assessments estimate stock status relative to reference points, which in turn influence the recommended catch. This influence of estimated status on recommended catch is made explicit in harvest control rules. The form of the control rule (how catch should change with biomass), and reference points (targets or limits) are additional targets for including ecosystem information (Punt et al., 2014; Patrick and Link, 2015b; e.g. Holsman et al., 2016). For example, the control rule for Pacific sardine depends on temperature (PFMC, 2018). Moreover, a control rule translates biomass into allowed catch (ABC, allowable biological catch), but actually setting catch limits (TAC, total allowable catch) is a separate decision, which could also be influenced qualitatively or quantitatively by ecosystem status (e.g. Zador et al. 2017) or other fishery or ecosystem considerations (Levin, 2014; Patrick and Link, 2015b). Our analysis provides a summary of the current state of stock assessments in the United States with respect to ecosystem science and highlights numerous examples where broader considerations have been included qualitatively and quantitatively. We identified potential data gaps and also opportunities for further expansion of assessments moving forward. Successful examples of where fishery and ecosystem interactions have been included in stock assessment reports can inform future guidelines for prioritizing which assessments to target for further expansion and funding opportunities to improve data collection for EBFM. Acknowledgements This work emerged from discussions with the Lenfest Fishery Ecosystem Task force supported by the Lenfest Ocean Program. We thank the task force and advisory panel members for their input on earlier versions of our analysis. We thank C. Hudson, B. Shouse, A. Bednarek, and the Lenfest Ocean Program for support throughout the project. Thanks to Rick Methot and Stacey Miller for facilitating our access to the NOAA Species Information System database, and to I. Kaplan, J. Hastie, M. McClure, and two anonymous reviewers whose comments improved our article. This manuscript reflects the views of the authors, not NOAA Fisheries nor the Lenfest Ocean Program. References Arkema K. K. , Abramson S. C., Dewsbury B. M. 2006 . Marine ecosystem-based management: from characterization to implementation . Frontiers in Ecology and the Environment , 4 : 525 – 532 . Google Scholar Crossref Search ADS WorldCat Berkes F. 2012 . Implementing ecosystem-based management: evolution or revolution? Fish and Fisheries , 13 : 465 – 476 . Google Scholar Crossref Search ADS WorldCat Biedron I. S. , Knuth B. A. 2016 . Toward shared understandings of ecosystem-based fisheries management among fishery management councils and stakeholders in the U.S. Mid-Atlantic and New England regions . Marine Policy , 70 : 40 – 48 . Google Scholar Crossref Search ADS WorldCat Bundy A. , Chuenpagdee R., Jentoft S., Mahon R. 2008 . If science is not the answer, what is? An alternative governance model for the world’s fisheries . Frontiers in Ecology and the Environment , 6 : 152 – 155 . Google Scholar Crossref Search ADS WorldCat Burnham K. P. , Anderson D. R. 1998 . Model Selection and Inference: A Practical Information-Theoretic Approach . Springer-Verlag , New York . 353 pp. Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Cowan J. H. , Rice J. C., Walters C. J., Hilborn R., Essington T. E., Day J. W., Boswell K. M. 2012 . Challenges for implementing an ecosystem approach to fisheries management . Marine and Coastal Fisheries , 4 : 496 – 510 . Google Scholar Crossref Search ADS WorldCat Essington T. E. , Moriarty P. E., Froehlich H. E., Hodgson E. E., Koehn L. E., Oken K. L., Siple M. C., et al. 2015 . Fishing amplifies forage fish population collapses . Proceedings of the National Academy of Sciences of the United States of America , 112 : 6648 – 6652 . Google Scholar Crossref Search ADS PubMed WorldCat Essington T. E. , Levin P. S., Marshall K. N., Koehn L. E., Anderson L. G., Bundy A., Carothers C., et al. 2016 . Building effective fishery ecosystem plans: a report from the Lenfest Fishery Ecosystem Task Force. Lenfest Ocean Program, Washington, DC. FAO . 2003 . Fisheries Management. 2. The ecosystem approach to fisheries. FAO Technical Guidelines for Responsible Fisheries, 4 Suppl. 2. Rome, Italy. Ford C. M. 1996 . A theory of individual creative action in multiple social domains . Academy of Management Review , 21 : 1112 – 1142 . Google Scholar Crossref Search ADS WorldCat Fredrickson B. L. 2004 . The broaden-and-build theory of positive emotions . Philosophical Transactions of The Royal Society of London, Series B: Biological Sciences , 359 : 1367. Google Scholar Crossref Search ADS WorldCat Heithaus M. , Frid A., Wirsing A., Worm B. 2008 . Predicting ecological consequences of marine top predator declines . Trends in Ecology & Evolution , 23 : 202 – 210 . Google Scholar Crossref Search ADS PubMed WorldCat Hilborn R. , Orensanz J. M., Parma A. M. 2005 . Institutions, incentives and the future of fisheries . Philosophical Transactions of The Royal Society of London, Series B: Biological Sciences , 360 : 47 – 57 . Google Scholar Crossref Search ADS WorldCat Hilborn R. 2011 . Future directions in ecosystem based fisheries management: a personal perspective . Fisheries Research , 108 : 235 – 239 . Google Scholar Crossref Search ADS WorldCat Holsman K. K. , Ianelli J., Aydin K., Punt A. E., Moffitt E. A. 2016 . A comparison of fisheries biological reference points estimated from temperature-specific multi-species and single-species climate-enhanced stock assessment models . Deep-Sea Research Part II-Topical Studies in Oceanography , 134 : 360 – 378 . Google Scholar Crossref Search ADS WorldCat Im S. , Montoya M. M., Workman J. P. 2013 . Antecedents and consequences of creativity in product innovation teams . Journal of Product Innovation Management , 30 : 170 – 185 . Google Scholar Crossref Search ADS WorldCat Kuparinen A. , Mäntyniemi S., Hutchings J. A., Kuikka S. 2012 . Increasing biological realism of fisheries stock assessment: towards hierarchical Bayesian methods . Environmental Reviews , 20 : 135 – 151 . Google Scholar Crossref Search ADS WorldCat Levin P. S. 2014 . New conservation for the Anthropocene Ocean. Conservation letters. http://onlinelibrary.wiley.com/doi/10.1111/conl.12108/abstract (accessed 10 November 2014). Link J. S. 2002 . What does ecosystem-based management mean? Fisheries , 27 : 18 – 21 . Google Scholar Crossref Search ADS WorldCat Link J. S. 2010 . Ecosystem-Based Fisheries Management: Confronting Tradeoffs . Cambridge University Press , Cambridge . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Mace P. M. 2001 . A new role for MSY in single-species and ecosystem approaches to fisheries stock assessment and management . Fish and Fisheries , 2 : 2 – 32 . Google Scholar Crossref Search ADS WorldCat Marine Strategy Framework Directive . 2008 . Directive 2008/56/EC of the European Parliament and of the Council of 17 June 2008 establishing a framework for community action in the field of marine environmental policy . Official Journal of the European Union L , 164 : 19 – 40 . OpenURL Placeholder Text WorldCat Maunder M. N. , Watters G. M. 2003 . A general framework for integrating environmental time series into stock assessment models: model descriptions, simulation testing and example . Fisheries Bulletin , 101 : 89 – 99 . Google Scholar OpenURL Placeholder Text WorldCat Methot R. D. , Wetzel C. R. 2013 . Stock synthesis: a biological and statistical framework for fish stock assessment and fishery management . Fisheries Research , 142 : 86 – 99 . Google Scholar Crossref Search ADS WorldCat Myers R. A. 1998 . When do environment–recruitment correlations work? Reviews in Fish Biology and Fisheries , 8 : 285 – 305 . Google Scholar Crossref Search ADS WorldCat NOAA . 2016 . NOAA fisheries ecosystem-based fisheries management road map. https://www.st.nmfs.noaa.gov/ecosystems/ebfm/creating-an-ebfm-management-policy (last accessed 17 February 2017). Olsson P. , Folke C., Hughes T. P. 2008 . Navigating the transition to ecosystem-based management of the Great Barrier Reef, Australia . Proceedings of the National Academy of Science of the United States of America , 105 : 9489 – 9494 . Google Scholar Crossref Search ADS WorldCat Patrick W. S. , Link J. S. 2015a . Myths that continue to impede progress in ecosystem-based fisheries management . Fisheries , 40 : 155 – 160 . Google Scholar Crossref Search ADS WorldCat Patrick W. S. , Link J. S. 2015b . Hidden in plain sight: using optimum yield as a policy framework to operationalize ecosystem-based fisheries management . Marine Policy , 62 : 74 – 81 . Google Scholar Crossref Search ADS WorldCat Pacific Fishery Management Council (PFMC) . 2016 . Terms of reference for the groundfish and coastal pelagic stock assessment review process for 2017–2018.http://www.pcouncil.org/wp-content/uploads/2017/01/Stock_Assessment_ToR_2017-18.pdf (last accessed 4 October 2018). OpenURL Placeholder Text WorldCat Pacific Fisheries Management Council (PFMC). 2018 . Coastal Pelagic Species Fishery Management Plan . https://www.pcouncil.org/wp-content/uploads/2018/05/CPS_FMP_as_Amended_thru_A16.pdf (last accessed 5 October 2018). OpenURL Placeholder Text WorldCat Pikitch E. K. , Boersma P. D., Boyd I. L., Conover D. O., Cury P., Essington T. E., Heppell S. S., et al. 2012 . Little fish, big impact: Managing a crucial link in ocean food webs. Lenfest Ocean Program, Washington, DC. http://www.oceanconservationsicence.org/foragefish (last accessed 16 May 2016). Pinsky M. L. , Byler D. 2015 . Fishing, fast growth and climate variability increase the risk of collapse . Proceedings of The Royal Society B: Biological Sciences , 282 : 20151053. Google Scholar OpenURL Placeholder Text WorldCat Pitcher T. J. , Kalikoski D., Short K., Varkey D., Pramod G. 2009 . An evaluation of progress in implementing ecosystem-based management of fisheries in 33 countries . Marine Policy , 33 : 223 – 232 . Google Scholar Crossref Search ADS WorldCat Punt A. E. , A’mar T., Bond N. A., Butterworth D. S., de Moor C. L., De Oliveira J. A. A., Haltuch M. A., et al. 2014 . Fisheries management under climate and environmental uncertainty: control rules and performance simulation . ICES Journal of Marine Science , 71 : 2208 – 2220 . Google Scholar Crossref Search ADS WorldCat Rasulzada F. 2014 . Creativity at work and its relation to well-being. Creativity Research: An Interdisciplinary and Multidisciplinary Research Handbook , pp. 171 – 190 . Ed. by E. Shiu. Routledge, New York. 338 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Skern-Mauritzen M. , Ottersen G., Handegard N. O., Huse G., Dingsør G. E., Stenseth N. C., Kjesbu O. S. 2016 . Ecosystem processes are rarely included in tactical fisheries management . Fish and Fisheries , 17 : 165 – 175 . Google Scholar Crossref Search ADS WorldCat Smith A. D. , Brown C. J., Bulman C. M., Fulton E. A., Johnson P., Kaplan I. C., Lozano-Montes H., et al. 2011 . Impacts of fishing low-trophic level species on marine ecosystems . Science , 333 : 1147 – 1150 . Google Scholar Crossref Search ADS PubMed WorldCat Szuwalski C. S. , Vert-pre K. A., Punt A. E., Branch T. A., Hilborn R. 2015 . Examining common assumptions about recruitment: a meta-analysis of recruitment dynamics for worldwide marine fisheries . Fish and Fisheries , 16 : 633 – 648 . Google Scholar Crossref Search ADS WorldCat Zador S. G. , Holsman K. K., Aydin K. Y., Gaichas S. K. 2017 . Ecosystem considerations in Alaska: the value of qualitative assessments . ICES Journal of Marine Science , 74 : 421 – 430 . Google Scholar Crossref Search ADS WorldCat Published by International Council for the Exploration of the Sea 2018. This work is written by US Government employees and is in the public domain in the US. Published by International Council for the Exploration of the Sea 2018.

Strong, James, Asa;Clements,, Annika;Lillis,, Helen;Galparsoro,, Ibon;Bildstein,, Tim;Pesch,, Roland

doi: 10.1093/icesjms/fsy161pmid: N/A

Abstract The production of marine habitat maps typically relies on the use of habitat classification schemes (HCSs). The choice of which HCS to use for a mapping study is often related to familiarity, established practice, and national desires. Despite a superficial similarity, HCSs differ greatly across six key properties, namely, purpose, environmental and ecological scope, spatial scale, thematic resolution, structure, and compatibility with mapping techniques. These properties impart specific strengths and weaknesses for each HCS, which are subsequently transferred to the habitat maps applying these schemes. This review has examined seven HCSs (that are commonly used and widely adopted for national and international mapping programmes), over the six properties, to understand their influence on marine habitat mapping. In addition, variation in how mappers interpret and apply HCSs introduces additional uncertainties and biases into the final maps. Recommendations are provided for improving HCSs for marine habitat mapping as well as for enhancing the working practices of mappers using habitat classification. It is hoped that implementation of these recommendations will lead to greater certainty and usage within mapping studies and more consistency between studies and adjoining maps. Introduction The pressing need for seabed inventory mapping, marine spatial planning, spatial estimates of anthropogenic impacts [as required by the Marine Strategy Framework Directive (Council Directive 2008/56/EC)], and the designation of seabed conservation features (as required by the Habitats Directive 92/43/EEC) has made the habitat map an indispensable item within marine management and research. The production, and ultimate presentation, of marine habitat maps typically rely on the use of a habitat classification scheme (HCS). Within mapping, HCSs categorize environmental and biological information (e.g. depth, topography, substratum, hydrodynamic energy, community composition) into distinct habitat classes. Each class is assumed to be associated with a distinctive abiotic condition and identifiable biological community, and therefore attempts to produce environmentally or ecologically meaningful units. Habitat classification is an integral part of habitat map production, and as such, the HCS has a significant influence on how mapping information is: (i) interpreted during map production; (ii) displayed within the map; and (iii) interpreted by the end user. This review aims to examine explicitly how HCSs influence the production of marine habitat maps. A wider discussion will follow on what improvements can be made to HCSs, and how mappers should use these HCSs, to provide more consistent, accurate, and useful products for end users. The specific objectives of this review are: Introduce the principles of habitat classification for marine mapping. Describe the properties common to most HCSs. Examine the variation in these common properties for seven established and contemporary HCSs used for benthic habitat mapping internationally. Assess the influence of variations within these common properties on the production and representations of marine habitat maps. Make recommendations for the development of HCSs in habitat mapping. Recommend best practice for marine habitat mappers when using HCSs. Use of habitat classification schemes in marine mapping Although HCSs are developed to support all sorts of environmental work, few activities are as intimately linked to the use of HCSs as habitat mapping. This section introduces HCSs, as well as how and why they are incorporated into marine habitat mapping. The influence that HCSs have on habitat maps is also introduced, before being discussed in more detail at the end of the review. Habitat classification schemes Robinson and Levings (1995) defined a HCS as a set of instructions that identify, delimit, and describe the habitats of distinct biological assemblages (communities or single species). The primary purposes of HCSs, summarized from Galparsoro et al. (2012) and Robinson and Levings (1995), are to: Provide a structured framework for the efficient classification of habitats Provide common and easily understood concepts and language for the description of habitats Hold information in a relational structure that allows for the interrogation of information based on parameters collected by common survey methods. Describe and standardize the physical, chemical, and biological parameters that define habitat classes. Regulate the spatial and thematic scales and thresholds used for habitat classification, and thereby standardize the classification of habitats within and between studies. The use of a HCS benefits marine habitat mapping in several ways. Most importantly, the HCS provides a structured framework for the integration of environmental and biological information (which have different spatial scales, units, and formats) into one, integrated product, via ecologically meaningful decision points along the classification pathway. Ultimately, HCSs facilitate the segmentation of discrete (e.g. categorical data such as substratum) and continuous variables into ecologically relevant spatial units. The influence of habitat classification schemes on the outputs of habitat mapping Although the benefits associated with the consistent classification of habitats during mapping are great, it must also be recognized that the use of a HCS also imposes certain constraints and limitations, which are inherent within the fundamental concepts of habitat classification. For example, many HCSs assume that individual habitats are discrete classes. When used in mapping, these classes usually form mutually exclusive patches when presented spatially, and therefore fail to capture the natural continuities (biocoenoses) and environmental gradients (ecotones) that perhaps better reflect the natural configuration and gradients between different habitat types. The structure of a HCS has a marked effect on the production process for a habitat map, through dictating when different types of information are relevant during the classification pathway. The structure can, therefore, modify the relative importance of physical, chemical, and biological variables in determining the final classification for a unit of habitat. Many HCSs have a hierarchical structure in which the physical information is typically associated with the upper levels of the hierarchy and can sometimes be assigned based on existing, coarse-resolution data such as from hydrodynamic models and digital elevation models. Lower levels of the classification hierarchy may describe biotopes, communities, and single-species distribution, which require biological data and are often applied at a more local scale. Due to insufficient biological data, or because it is not relevant for the specific scheme or level of classification, some HCSs are based purely on physical and environmental features of the seafloor environment, which are used as a proxy for habitats, on the assumption that there may be a correlation between the non-biological features and biological communities (Brown et al., 2011; Huang et al., 2011). Such assumptions are the basis for the use of distribution modelling techniques by employing full spatial coverage data of environmental variables to predict spatial distribution patterns of benthic communities during the map production (Reiss et al., 2014). Although it is a sensible aspiration that a single classification scheme is used for all marine habitat maps, multiple schemes have arisen to cater for the different applications, for example biological conservation, landscape ecology, environmental monitoring, marine spatial planning, fisheries management, geomorphological descriptions, etc. The presence of several HCSs also reflects the fundamental difficulty of dividing natural continuities (biocoenoses) and environmental gradients (ecotones), into discrete and meaningful classes. Furthermore, the number of HCSs is further inflated as individual schemes cater for specific biogeographic areas. Lund and Wibur (2007) and Greene et al. (2008) summarized 14 marine HCSs developed for North America and Europe alone. Interestingly, schemes differ substantially even though (i) the main physico-chemical variables that are known to define habitats are well-established, (ii) the majority of marine mapping studies record the same parameters, and (iii) the predominantly physical nature of the majority of the classifications. The use of different HCSs for mapping can significantly influence the spatial representation of habitats, and consequently ecologically and biologically significant areas, in the final maps, which in turn can hinder the merging of adjoining maps as well as alter management outcomes based on these maps (Gregr et al., 2012). Variation associated with six common properties of marine habitat classification schemes and their influence on habitat maps An examination of the HCSs suggests that they differ according to six properties, namely, (i) purpose of a HCS; (ii) environmental and ecological scope of a HCS; (iii) spatial scale covered by a HCS; (iv) thematic resolution covered by a HCS; (v) structure of a HCS; and (vi) compatibility of a HCS for habitat mapping. Variation in each property can influence the production, and representation, of a marine habitat map. The following section will: (i) introduce each property; (ii) examine seven common and widely adopted HCSs to highlight the variation within each property (these schemes are introduced in Table 1); and (iii) summarize the influence of variation, within each property, on habitat map production. Table 1. Marine (benthic) HCSs used to document the variation in the six scheme properties considered. Habitat classification scheme Description Examples of usage European Nature Information System (EUNIS)—Davies et al. (2004)http://eunis.eea.europa.eu/ (revised in 2007) EUNIS is a pan-European HCS developed between 1996 and 2001 by the European Environment Agency (EEA) (Davies et al., 2004). It considers both marine and terrestrial habitats in Europe. The geographical scope of the EUNIS marine scheme is the marine waters off the European mainland, including offshore islands (British Isles, Cyprus, Iceland, but not Greenland), and the archipelagos of the European Union Member States (Canary Islands, Madeira, and the Azores). EUNIS marine scheme covers the entire seabed from the intertidal zone to the abyss, and also includes a section of pelagic habitats. In the marine sector, it is based on the Joint Nature Conservation Committee (JNCC) Marine Habitat Classification for Britain and Ireland (Connor et al., 2004) and habitat types developed by the Barcelona and Helcom marine conventions. EUNIS supports inventory mapping (EMODnet), ecosystem-based management (Andersen et al., 2018) and policy implementation Marine Strategy Framework Directive (Council Directive 2008/56/EC). HELCOM Underwater Biotope and Habitat classification system (HELCOM HUB) —HELCOM (2013a) http://www.helcom.fi/baltic-sea-trends/biodiversity/helcom-hub HELCOM HUB was developed to be a comprehensive classification system for marine biotopes of the Baltic Sea (HELCOM, 2013a). Its origins go back to the HELCOM EC-NATURE Red List Project (HELCOM, 1998), which was a first Baltic Sea wide classification scheme based on substrate type and bathymetry. Its classification rules mainly relied on expert judgment and biological classification criteria were not included. In 2007, the goal was set to renew the Red List Classification system by a HELCOM Red List Biotope Expert Group. Previous attempts had been made to apply EUNIS to the Baltic Sea region but the system was recognized to poorly represent its biotic and abiotic characteristics (Galparsoro et al., 2012). Nevertheless, HELCOM HUB was to be compatible with EUNIS and account for available biological information on marine biotopes from the Baltic Sea. HELCOM HUB is primarily focused on benthic habitats/biotopes—the pelagic environment is only dealt with in the upper part of the classification system. As one major improvement, HELCOM HUB provides clear quantitative classification rules for both abiotic and biological criteria. It was therefore used as a basis for the development of the national classification system of the German Red List of Threatened Habitat Types for both the North Sea and the Baltic Sea (Finck et al., 2017). Furthermore HELCOM HUB serves as a reference system to declare and assess marine and coastal Red List biotopes and biotope complexes (HELCOM, 2013b). Supports the assessment of marine and coastal Red List biotope and biotope complexes in the Baltic Sea region and the national implementation of the Marine Strategy Framework Directive (Council Directive 2008/56/EC). Potential Habitat Characterization Scheme (PHCS)—Greene et al. (1999, 2005, 2007) This classification was originally developed to encourage the standardization of new mapping techniques being applied to fisheries research in North America. Consequently, the scheme covers deep-water habitats within North America and has been expanded to include shallow water habitats, as well as Arctic, Antarctic, and tropical regions (Vietti et al., 2001) and estuaries (Greene et al., 2007). This scheme has been specifically developed for seafloor mapping and uses common mapping information such as multibeam echosounder data, video, photographs taken with still cameras, and seafloor samples from grabs. The attributions used to classify the seafloor are mainly based on physical parameters and features and therefore, has a “bottom-up” structure. The classification scheme is unusual in that it recognizes four spatial scales. The first three scales (megascale: few kilometres to 100s of kilometres; mesoscale 10s of metres to few kilometres; macroscale: 1 to 10s m) can be defined with acoustic methods whereas the finest scale (microscale: centimetre to metre) can only be delineated with direct observation (via video, photographic still imagery, diver observations, or seafloor sampling) (Greene et al., 2005, 2007). Adopted for fisheries management, for example Alaska Department of Fish and Game, National Park Service, and Washington Department of Fish and Wildlife (Greene et al., 2005, 2007; Yoklavich et al., 2000) and geophysical mapping of the seafloor (Auster et al., 1995) Hierarchical Framework of Marine Habitat Classification for Ecosystem-Based Management (HFMHC)—Guarinello et al. (2010) This classification framework is specifically designed to meet the need for ecosystem-based management of the marine environment (within North America but applicable anywhere). The upper levels of the scheme include the global classification of large marine ecosystems. Subsequent levels include recognizable ecosystem units. The scheme has three parallel (un-nested) “benthic”, “water column”, and “human” hierarchies. The flexibility to add user-defined classes at the lower levels of all three strands means the framework can be applied in any geographic location and is not limited by the methods used to observe any of the three strands. The framework incorporates the central concepts of ecosystem-based management within the structure of the framework. This ensures that the products of this HCS reflect the values and objectives of ecosystem-based management. Ecosystem-based management as required by the Interagency Ocean Policy Task Force (Guarinello et al., 2010) Classification of Sublittoral Habitats (CSH)—Valentine et al. (2005) This classification scheme was designed to describe and classify habitats in terms of geological, biological, and oceanographic attributes. It is unusual in that the scheme also captures information on the effects of natural and anthropogenic processes on habitats. It contains eight, non-hierarchical themes, namely seabed topography, dynamics, texture, grain size, roughness, fauna and flora, habitat association and usage, and habitat recovery from disturbance. The purpose of the classification is to provide a foundation for scientific research and environmental management of seafloor habitats across a relatively large, regional scales. Despite being developed for the Gulf of Maine region (faunal and floral classes are reflective of the northeastern North America region), the non-biological components of the scheme are and can therefore by applied to other continental shelf and shelf basin areas globally (excluding some low-latitude environments). Fisheries and environmental management (Valentine et al., 2005); there is little evidence of this scheme having being adopted widely. Australian National Intertidal/Subtidal Benthic Habitat Classification Scheme (NISB) `http://lwa.gov.au/products/pn21267 The NISB scheme was developed to identify a “uniform definition of communities, habitats and ecosystems” at both regional and national scales, and spatial information that is informative for assessing critical climate change issues and detecting change or loss of habitats or communities. The proposed scheme covers all of Australia’s territorial waters including intertidal habitats and down to the limit of the photic zone (depth of 50–70 m). Inventory mapping of ecoregions (bioregional subregions) Coastal and Marine Ecological Classification Standard (CMECS)—Madden et al. (2009)https://www.cmecscatalog.org/cmecs/ CMECS was developed by the National Oceanic and Atmospheric Administration (NOAA) and NatureServe. The scheme is founded on existing schemes [e.g. Cowardin et al. (1979), Dethier (1992), Greene et al. (1999), Allee et al. (2000), Zacharias and Roff (2000) and Connor et al. (2004)]. CMECS includes all estuarine, coastal, and marine waters under US jurisdiction in North America. This includes wetlands, the intertidal zone, coastal, and deep-water habitats (including the Great Lakes) as well as the pelagic realm. Development of a coastal marine biodiversity inventory for North America; Delineation of regions for marine-protected areas; identification of essential fish habitat Implementation and monitoring of ecosystem-based management strategies for coastal systems (Madden et al., 2009) Habitat classification scheme Description Examples of usage European Nature Information System (EUNIS)—Davies et al. (2004)http://eunis.eea.europa.eu/ (revised in 2007) EUNIS is a pan-European HCS developed between 1996 and 2001 by the European Environment Agency (EEA) (Davies et al., 2004). It considers both marine and terrestrial habitats in Europe. The geographical scope of the EUNIS marine scheme is the marine waters off the European mainland, including offshore islands (British Isles, Cyprus, Iceland, but not Greenland), and the archipelagos of the European Union Member States (Canary Islands, Madeira, and the Azores). EUNIS marine scheme covers the entire seabed from the intertidal zone to the abyss, and also includes a section of pelagic habitats. In the marine sector, it is based on the Joint Nature Conservation Committee (JNCC) Marine Habitat Classification for Britain and Ireland (Connor et al., 2004) and habitat types developed by the Barcelona and Helcom marine conventions. EUNIS supports inventory mapping (EMODnet), ecosystem-based management (Andersen et al., 2018) and policy implementation Marine Strategy Framework Directive (Council Directive 2008/56/EC). HELCOM Underwater Biotope and Habitat classification system (HELCOM HUB) —HELCOM (2013a) http://www.helcom.fi/baltic-sea-trends/biodiversity/helcom-hub HELCOM HUB was developed to be a comprehensive classification system for marine biotopes of the Baltic Sea (HELCOM, 2013a). Its origins go back to the HELCOM EC-NATURE Red List Project (HELCOM, 1998), which was a first Baltic Sea wide classification scheme based on substrate type and bathymetry. Its classification rules mainly relied on expert judgment and biological classification criteria were not included. In 2007, the goal was set to renew the Red List Classification system by a HELCOM Red List Biotope Expert Group. Previous attempts had been made to apply EUNIS to the Baltic Sea region but the system was recognized to poorly represent its biotic and abiotic characteristics (Galparsoro et al., 2012). Nevertheless, HELCOM HUB was to be compatible with EUNIS and account for available biological information on marine biotopes from the Baltic Sea. HELCOM HUB is primarily focused on benthic habitats/biotopes—the pelagic environment is only dealt with in the upper part of the classification system. As one major improvement, HELCOM HUB provides clear quantitative classification rules for both abiotic and biological criteria. It was therefore used as a basis for the development of the national classification system of the German Red List of Threatened Habitat Types for both the North Sea and the Baltic Sea (Finck et al., 2017). Furthermore HELCOM HUB serves as a reference system to declare and assess marine and coastal Red List biotopes and biotope complexes (HELCOM, 2013b). Supports the assessment of marine and coastal Red List biotope and biotope complexes in the Baltic Sea region and the national implementation of the Marine Strategy Framework Directive (Council Directive 2008/56/EC). Potential Habitat Characterization Scheme (PHCS)—Greene et al. (1999, 2005, 2007) This classification was originally developed to encourage the standardization of new mapping techniques being applied to fisheries research in North America. Consequently, the scheme covers deep-water habitats within North America and has been expanded to include shallow water habitats, as well as Arctic, Antarctic, and tropical regions (Vietti et al., 2001) and estuaries (Greene et al., 2007). This scheme has been specifically developed for seafloor mapping and uses common mapping information such as multibeam echosounder data, video, photographs taken with still cameras, and seafloor samples from grabs. The attributions used to classify the seafloor are mainly based on physical parameters and features and therefore, has a “bottom-up” structure. The classification scheme is unusual in that it recognizes four spatial scales. The first three scales (megascale: few kilometres to 100s of kilometres; mesoscale 10s of metres to few kilometres; macroscale: 1 to 10s m) can be defined with acoustic methods whereas the finest scale (microscale: centimetre to metre) can only be delineated with direct observation (via video, photographic still imagery, diver observations, or seafloor sampling) (Greene et al., 2005, 2007). Adopted for fisheries management, for example Alaska Department of Fish and Game, National Park Service, and Washington Department of Fish and Wildlife (Greene et al., 2005, 2007; Yoklavich et al., 2000) and geophysical mapping of the seafloor (Auster et al., 1995) Hierarchical Framework of Marine Habitat Classification for Ecosystem-Based Management (HFMHC)—Guarinello et al. (2010) This classification framework is specifically designed to meet the need for ecosystem-based management of the marine environment (within North America but applicable anywhere). The upper levels of the scheme include the global classification of large marine ecosystems. Subsequent levels include recognizable ecosystem units. The scheme has three parallel (un-nested) “benthic”, “water column”, and “human” hierarchies. The flexibility to add user-defined classes at the lower levels of all three strands means the framework can be applied in any geographic location and is not limited by the methods used to observe any of the three strands. The framework incorporates the central concepts of ecosystem-based management within the structure of the framework. This ensures that the products of this HCS reflect the values and objectives of ecosystem-based management. Ecosystem-based management as required by the Interagency Ocean Policy Task Force (Guarinello et al., 2010) Classification of Sublittoral Habitats (CSH)—Valentine et al. (2005) This classification scheme was designed to describe and classify habitats in terms of geological, biological, and oceanographic attributes. It is unusual in that the scheme also captures information on the effects of natural and anthropogenic processes on habitats. It contains eight, non-hierarchical themes, namely seabed topography, dynamics, texture, grain size, roughness, fauna and flora, habitat association and usage, and habitat recovery from disturbance. The purpose of the classification is to provide a foundation for scientific research and environmental management of seafloor habitats across a relatively large, regional scales. Despite being developed for the Gulf of Maine region (faunal and floral classes are reflective of the northeastern North America region), the non-biological components of the scheme are and can therefore by applied to other continental shelf and shelf basin areas globally (excluding some low-latitude environments). Fisheries and environmental management (Valentine et al., 2005); there is little evidence of this scheme having being adopted widely. Australian National Intertidal/Subtidal Benthic Habitat Classification Scheme (NISB) `http://lwa.gov.au/products/pn21267 The NISB scheme was developed to identify a “uniform definition of communities, habitats and ecosystems” at both regional and national scales, and spatial information that is informative for assessing critical climate change issues and detecting change or loss of habitats or communities. The proposed scheme covers all of Australia’s territorial waters including intertidal habitats and down to the limit of the photic zone (depth of 50–70 m). Inventory mapping of ecoregions (bioregional subregions) Coastal and Marine Ecological Classification Standard (CMECS)—Madden et al. (2009)https://www.cmecscatalog.org/cmecs/ CMECS was developed by the National Oceanic and Atmospheric Administration (NOAA) and NatureServe. The scheme is founded on existing schemes [e.g. Cowardin et al. (1979), Dethier (1992), Greene et al. (1999), Allee et al. (2000), Zacharias and Roff (2000) and Connor et al. (2004)]. CMECS includes all estuarine, coastal, and marine waters under US jurisdiction in North America. This includes wetlands, the intertidal zone, coastal, and deep-water habitats (including the Great Lakes) as well as the pelagic realm. Development of a coastal marine biodiversity inventory for North America; Delineation of regions for marine-protected areas; identification of essential fish habitat Implementation and monitoring of ecosystem-based management strategies for coastal systems (Madden et al., 2009) Table 1. Marine (benthic) HCSs used to document the variation in the six scheme properties considered. Habitat classification scheme Description Examples of usage European Nature Information System (EUNIS)—Davies et al. (2004)http://eunis.eea.europa.eu/ (revised in 2007) EUNIS is a pan-European HCS developed between 1996 and 2001 by the European Environment Agency (EEA) (Davies et al., 2004). It considers both marine and terrestrial habitats in Europe. The geographical scope of the EUNIS marine scheme is the marine waters off the European mainland, including offshore islands (British Isles, Cyprus, Iceland, but not Greenland), and the archipelagos of the European Union Member States (Canary Islands, Madeira, and the Azores). EUNIS marine scheme covers the entire seabed from the intertidal zone to the abyss, and also includes a section of pelagic habitats. In the marine sector, it is based on the Joint Nature Conservation Committee (JNCC) Marine Habitat Classification for Britain and Ireland (Connor et al., 2004) and habitat types developed by the Barcelona and Helcom marine conventions. EUNIS supports inventory mapping (EMODnet), ecosystem-based management (Andersen et al., 2018) and policy implementation Marine Strategy Framework Directive (Council Directive 2008/56/EC). HELCOM Underwater Biotope and Habitat classification system (HELCOM HUB) —HELCOM (2013a) http://www.helcom.fi/baltic-sea-trends/biodiversity/helcom-hub HELCOM HUB was developed to be a comprehensive classification system for marine biotopes of the Baltic Sea (HELCOM, 2013a). Its origins go back to the HELCOM EC-NATURE Red List Project (HELCOM, 1998), which was a first Baltic Sea wide classification scheme based on substrate type and bathymetry. Its classification rules mainly relied on expert judgment and biological classification criteria were not included. In 2007, the goal was set to renew the Red List Classification system by a HELCOM Red List Biotope Expert Group. Previous attempts had been made to apply EUNIS to the Baltic Sea region but the system was recognized to poorly represent its biotic and abiotic characteristics (Galparsoro et al., 2012). Nevertheless, HELCOM HUB was to be compatible with EUNIS and account for available biological information on marine biotopes from the Baltic Sea. HELCOM HUB is primarily focused on benthic habitats/biotopes—the pelagic environment is only dealt with in the upper part of the classification system. As one major improvement, HELCOM HUB provides clear quantitative classification rules for both abiotic and biological criteria. It was therefore used as a basis for the development of the national classification system of the German Red List of Threatened Habitat Types for both the North Sea and the Baltic Sea (Finck et al., 2017). Furthermore HELCOM HUB serves as a reference system to declare and assess marine and coastal Red List biotopes and biotope complexes (HELCOM, 2013b). Supports the assessment of marine and coastal Red List biotope and biotope complexes in the Baltic Sea region and the national implementation of the Marine Strategy Framework Directive (Council Directive 2008/56/EC). Potential Habitat Characterization Scheme (PHCS)—Greene et al. (1999, 2005, 2007) This classification was originally developed to encourage the standardization of new mapping techniques being applied to fisheries research in North America. Consequently, the scheme covers deep-water habitats within North America and has been expanded to include shallow water habitats, as well as Arctic, Antarctic, and tropical regions (Vietti et al., 2001) and estuaries (Greene et al., 2007). This scheme has been specifically developed for seafloor mapping and uses common mapping information such as multibeam echosounder data, video, photographs taken with still cameras, and seafloor samples from grabs. The attributions used to classify the seafloor are mainly based on physical parameters and features and therefore, has a “bottom-up” structure. The classification scheme is unusual in that it recognizes four spatial scales. The first three scales (megascale: few kilometres to 100s of kilometres; mesoscale 10s of metres to few kilometres; macroscale: 1 to 10s m) can be defined with acoustic methods whereas the finest scale (microscale: centimetre to metre) can only be delineated with direct observation (via video, photographic still imagery, diver observations, or seafloor sampling) (Greene et al., 2005, 2007). Adopted for fisheries management, for example Alaska Department of Fish and Game, National Park Service, and Washington Department of Fish and Wildlife (Greene et al., 2005, 2007; Yoklavich et al., 2000) and geophysical mapping of the seafloor (Auster et al., 1995) Hierarchical Framework of Marine Habitat Classification for Ecosystem-Based Management (HFMHC)—Guarinello et al. (2010) This classification framework is specifically designed to meet the need for ecosystem-based management of the marine environment (within North America but applicable anywhere). The upper levels of the scheme include the global classification of large marine ecosystems. Subsequent levels include recognizable ecosystem units. The scheme has three parallel (un-nested) “benthic”, “water column”, and “human” hierarchies. The flexibility to add user-defined classes at the lower levels of all three strands means the framework can be applied in any geographic location and is not limited by the methods used to observe any of the three strands. The framework incorporates the central concepts of ecosystem-based management within the structure of the framework. This ensures that the products of this HCS reflect the values and objectives of ecosystem-based management. Ecosystem-based management as required by the Interagency Ocean Policy Task Force (Guarinello et al., 2010) Classification of Sublittoral Habitats (CSH)—Valentine et al. (2005) This classification scheme was designed to describe and classify habitats in terms of geological, biological, and oceanographic attributes. It is unusual in that the scheme also captures information on the effects of natural and anthropogenic processes on habitats. It contains eight, non-hierarchical themes, namely seabed topography, dynamics, texture, grain size, roughness, fauna and flora, habitat association and usage, and habitat recovery from disturbance. The purpose of the classification is to provide a foundation for scientific research and environmental management of seafloor habitats across a relatively large, regional scales. Despite being developed for the Gulf of Maine region (faunal and floral classes are reflective of the northeastern North America region), the non-biological components of the scheme are and can therefore by applied to other continental shelf and shelf basin areas globally (excluding some low-latitude environments). Fisheries and environmental management (Valentine et al., 2005); there is little evidence of this scheme having being adopted widely. Australian National Intertidal/Subtidal Benthic Habitat Classification Scheme (NISB) `http://lwa.gov.au/products/pn21267 The NISB scheme was developed to identify a “uniform definition of communities, habitats and ecosystems” at both regional and national scales, and spatial information that is informative for assessing critical climate change issues and detecting change or loss of habitats or communities. The proposed scheme covers all of Australia’s territorial waters including intertidal habitats and down to the limit of the photic zone (depth of 50–70 m). Inventory mapping of ecoregions (bioregional subregions) Coastal and Marine Ecological Classification Standard (CMECS)—Madden et al. (2009)https://www.cmecscatalog.org/cmecs/ CMECS was developed by the National Oceanic and Atmospheric Administration (NOAA) and NatureServe. The scheme is founded on existing schemes [e.g. Cowardin et al. (1979), Dethier (1992), Greene et al. (1999), Allee et al. (2000), Zacharias and Roff (2000) and Connor et al. (2004)]. CMECS includes all estuarine, coastal, and marine waters under US jurisdiction in North America. This includes wetlands, the intertidal zone, coastal, and deep-water habitats (including the Great Lakes) as well as the pelagic realm. Development of a coastal marine biodiversity inventory for North America; Delineation of regions for marine-protected areas; identification of essential fish habitat Implementation and monitoring of ecosystem-based management strategies for coastal systems (Madden et al., 2009) Habitat classification scheme Description Examples of usage European Nature Information System (EUNIS)—Davies et al. (2004)http://eunis.eea.europa.eu/ (revised in 2007) EUNIS is a pan-European HCS developed between 1996 and 2001 by the European Environment Agency (EEA) (Davies et al., 2004). It considers both marine and terrestrial habitats in Europe. The geographical scope of the EUNIS marine scheme is the marine waters off the European mainland, including offshore islands (British Isles, Cyprus, Iceland, but not Greenland), and the archipelagos of the European Union Member States (Canary Islands, Madeira, and the Azores). EUNIS marine scheme covers the entire seabed from the intertidal zone to the abyss, and also includes a section of pelagic habitats. In the marine sector, it is based on the Joint Nature Conservation Committee (JNCC) Marine Habitat Classification for Britain and Ireland (Connor et al., 2004) and habitat types developed by the Barcelona and Helcom marine conventions. EUNIS supports inventory mapping (EMODnet), ecosystem-based management (Andersen et al., 2018) and policy implementation Marine Strategy Framework Directive (Council Directive 2008/56/EC). HELCOM Underwater Biotope and Habitat classification system (HELCOM HUB) —HELCOM (2013a) http://www.helcom.fi/baltic-sea-trends/biodiversity/helcom-hub HELCOM HUB was developed to be a comprehensive classification system for marine biotopes of the Baltic Sea (HELCOM, 2013a). Its origins go back to the HELCOM EC-NATURE Red List Project (HELCOM, 1998), which was a first Baltic Sea wide classification scheme based on substrate type and bathymetry. Its classification rules mainly relied on expert judgment and biological classification criteria were not included. In 2007, the goal was set to renew the Red List Classification system by a HELCOM Red List Biotope Expert Group. Previous attempts had been made to apply EUNIS to the Baltic Sea region but the system was recognized to poorly represent its biotic and abiotic characteristics (Galparsoro et al., 2012). Nevertheless, HELCOM HUB was to be compatible with EUNIS and account for available biological information on marine biotopes from the Baltic Sea. HELCOM HUB is primarily focused on benthic habitats/biotopes—the pelagic environment is only dealt with in the upper part of the classification system. As one major improvement, HELCOM HUB provides clear quantitative classification rules for both abiotic and biological criteria. It was therefore used as a basis for the development of the national classification system of the German Red List of Threatened Habitat Types for both the North Sea and the Baltic Sea (Finck et al., 2017). Furthermore HELCOM HUB serves as a reference system to declare and assess marine and coastal Red List biotopes and biotope complexes (HELCOM, 2013b). Supports the assessment of marine and coastal Red List biotope and biotope complexes in the Baltic Sea region and the national implementation of the Marine Strategy Framework Directive (Council Directive 2008/56/EC). Potential Habitat Characterization Scheme (PHCS)—Greene et al. (1999, 2005, 2007) This classification was originally developed to encourage the standardization of new mapping techniques being applied to fisheries research in North America. Consequently, the scheme covers deep-water habitats within North America and has been expanded to include shallow water habitats, as well as Arctic, Antarctic, and tropical regions (Vietti et al., 2001) and estuaries (Greene et al., 2007). This scheme has been specifically developed for seafloor mapping and uses common mapping information such as multibeam echosounder data, video, photographs taken with still cameras, and seafloor samples from grabs. The attributions used to classify the seafloor are mainly based on physical parameters and features and therefore, has a “bottom-up” structure. The classification scheme is unusual in that it recognizes four spatial scales. The first three scales (megascale: few kilometres to 100s of kilometres; mesoscale 10s of metres to few kilometres; macroscale: 1 to 10s m) can be defined with acoustic methods whereas the finest scale (microscale: centimetre to metre) can only be delineated with direct observation (via video, photographic still imagery, diver observations, or seafloor sampling) (Greene et al., 2005, 2007). Adopted for fisheries management, for example Alaska Department of Fish and Game, National Park Service, and Washington Department of Fish and Wildlife (Greene et al., 2005, 2007; Yoklavich et al., 2000) and geophysical mapping of the seafloor (Auster et al., 1995) Hierarchical Framework of Marine Habitat Classification for Ecosystem-Based Management (HFMHC)—Guarinello et al. (2010) This classification framework is specifically designed to meet the need for ecosystem-based management of the marine environment (within North America but applicable anywhere). The upper levels of the scheme include the global classification of large marine ecosystems. Subsequent levels include recognizable ecosystem units. The scheme has three parallel (un-nested) “benthic”, “water column”, and “human” hierarchies. The flexibility to add user-defined classes at the lower levels of all three strands means the framework can be applied in any geographic location and is not limited by the methods used to observe any of the three strands. The framework incorporates the central concepts of ecosystem-based management within the structure of the framework. This ensures that the products of this HCS reflect the values and objectives of ecosystem-based management. Ecosystem-based management as required by the Interagency Ocean Policy Task Force (Guarinello et al., 2010) Classification of Sublittoral Habitats (CSH)—Valentine et al. (2005) This classification scheme was designed to describe and classify habitats in terms of geological, biological, and oceanographic attributes. It is unusual in that the scheme also captures information on the effects of natural and anthropogenic processes on habitats. It contains eight, non-hierarchical themes, namely seabed topography, dynamics, texture, grain size, roughness, fauna and flora, habitat association and usage, and habitat recovery from disturbance. The purpose of the classification is to provide a foundation for scientific research and environmental management of seafloor habitats across a relatively large, regional scales. Despite being developed for the Gulf of Maine region (faunal and floral classes are reflective of the northeastern North America region), the non-biological components of the scheme are and can therefore by applied to other continental shelf and shelf basin areas globally (excluding some low-latitude environments). Fisheries and environmental management (Valentine et al., 2005); there is little evidence of this scheme having being adopted widely. Australian National Intertidal/Subtidal Benthic Habitat Classification Scheme (NISB) `http://lwa.gov.au/products/pn21267 The NISB scheme was developed to identify a “uniform definition of communities, habitats and ecosystems” at both regional and national scales, and spatial information that is informative for assessing critical climate change issues and detecting change or loss of habitats or communities. The proposed scheme covers all of Australia’s territorial waters including intertidal habitats and down to the limit of the photic zone (depth of 50–70 m). Inventory mapping of ecoregions (bioregional subregions) Coastal and Marine Ecological Classification Standard (CMECS)—Madden et al. (2009)https://www.cmecscatalog.org/cmecs/ CMECS was developed by the National Oceanic and Atmospheric Administration (NOAA) and NatureServe. The scheme is founded on existing schemes [e.g. Cowardin et al. (1979), Dethier (1992), Greene et al. (1999), Allee et al. (2000), Zacharias and Roff (2000) and Connor et al. (2004)]. CMECS includes all estuarine, coastal, and marine waters under US jurisdiction in North America. This includes wetlands, the intertidal zone, coastal, and deep-water habitats (including the Great Lakes) as well as the pelagic realm. Development of a coastal marine biodiversity inventory for North America; Delineation of regions for marine-protected areas; identification of essential fish habitat Implementation and monitoring of ecosystem-based management strategies for coastal systems (Madden et al., 2009) Purpose of a habitat classification scheme A number of HCSs have been constructed for differing but specific purposes (Gregr et al., 2012). For example, some schemes, such as the Potential Habitat Characterization Scheme (PHCS: Greene et al. 1999, 2005, 2007), are designed to address the delineation of fisheries habitats, while others specifically include habitats of conservation importance. Most schemes are more generic classifications, which are more suitable for inventory mapping. The purpose of a HCS dictates the emphasis for separation between classes, and therefore the way in which observed variables are partitioned within the scheme. This structuring is reproduced within a habitat map when a specific HCS is used. Variation in the purpose between habitat classification schemes The majority of HCSs are generalist, descriptive schemes that potentially offer the greatest utility to the largest number of users [of those reviewed here, examples include the European Nature Information System (EUNIS: Davies et al., 2004), HELCOM Underwater Biotope (HUB) and Habitat classification system (HELCOM HUB: HELCOM, 2013a), and the Coastal and Marine Ecological Classification Standard (CMECS: Madden et al., 2009)]. Marine habitat maps are of high relevance when supporting the implementation of diverse policies. For instance, in the framework of European policies, (i) maps have been used for the assessment of priority habitats and species for conservation [Habitats Directive (92/43/EEC)], (ii) they have contributed to the improvement of the knowledge and application of several criteria and indicators used to assess environmental status in the European Marine Strategy Framework Directive (MSFD; 2008/56/EC), in relation to the biological diversity and seafloor integrity descriptors (Galparsoro et al., 2015); and (iii) within the Maritime Spatial Planning (Directive 2014/89/EU), maps have been used as the basic cartographic information for marine space ordination as well as in the early identification of impact and opportunities for multiple use of space by maritime activities. Moreover, maps produced using these schemes are most likely to be centrally collated and widely distributed. For that, the directive for the Infrastructure for Spatial Information in the European Community (INSPIRE; 2007/2/EC), lays down a general framework for a spatial data infrastructure (SDI) for the purposes of European Community environmental policies and policies or activities which may affect the environment and specifies three “reference schemes”: EUNIS, MSFD, and Habitats Directive Annex I habitat types. HCSs aimed at marine mapping, assessment, and reporting are increasingly using EUNIS and HUB (within the Baltic Sea) habitat categories and respective codes so as to guarantee a common shared path and technical terminology between Member States (Vasquez et al., 2015). The Australian National Intertidal/Subtidal Benthic (NISB) scheme (Mount et al., 2007) and the classification of sublittoral habitats (CSH) scheme (Valentine et al., 2005) are also broad enough to allow full coverage mapping and use for the environmental management of seafloor habitats (although NISB primarily focused on managing climate change related issues), as well as specifically providing a foundation for scientific research. The primary purpose of CMECS is to be a national standard for the classification of habitats that ensures the consistency of state, national, and international outputs (Madden et al., 2009). Unlike other schemes, CMECS is claimed to be relatively multipurpose in that it also caters for (i) fisheries management; (ii) the identification and administration of marine-protected areas (Madden et al., 2009); and (iii) ecosystem-based management of marine resources. In contrast, the Potential Habitat Characterization Scheme (PHCS: Greene et al. 1999, 2005, 2007) has a clear geological emphasis, which is thought to provide a better basis for fisheries management, that is the identification of Essential Fish Habitat. Consequently, this scheme has been adopted for the contiguous western coast of the United States for rockfish habitat mapping (Greene et al., 2007). Management purposes lie at the heart of the Hierarchical Framework of Marine Habitat Classification for Ecosystem-Based Management (HFMHC: Guarinello et al., 2010), which has been designed specifically for promoting ecosystem-based management (Guarinello et al., 2010). The framework incorporates the central concepts of ecosystem-based management—this ensures that the products of this HCS reflect the values and objectives of this style of management. The HELCOM HUB scheme has also been designed to align with a strategic plan to ensure ecosystem-based management (HELCOM Baltic Sea Action Plan) and to enable a Red List assessment of marine and coastal biotopes and biotope complexes in the entire Baltic Sea region (HELCOM, 2013a, b). The influence of purpose on habitat maps The majority of HCS are generic, inventory schemes that have subsequently been adopted for use in marine management. Several of the European systems were, however, designed initially for the ready identification of habitats of conservation importance. Other schemes are more specific, in either dealing with components of the habitat (e.g. ground fish) or specific management topics (e.g. climate change, fisheries, conservation, ecosystem-based management). The purpose of a HCS will dictate the information that is required within the classification and, ultimately, how this information is partitioned and presented within a map. Most habitat mapping studies adopt just one HCS, and consequently limit the maps to a specific set of purposes. This restricts both the breadth of the maps for other purposes and how exhaustively the mapping data are used. It is likely that the greatest utility, accuracy, and confidence for a purpose can be obtained from a map classified using a scheme dedicated for that particular purpose. Environmental and ecological scope of a habitat classification scheme The scope of a HCS defines which (i) biogeographic region(s), (ii) biological realm(s) (e.g. pelagic/benthic), and (iii) type of habitats included (e.g. coastal area, estuaries, or hard substrata) are covered by the scheme. In some cases, a HCS will have been developed for a specific biological component, study, or geographic location, and the resulting habitat types may not be applicable beyond that subject or area. In other cases, schemes have been developed using broad-scale data or using thresholds in ecologically relevant variables (Vasquez et al., 2015), for example a pre-defined mud fraction in the sediment. Variation in the scope between habitat classification schemes The combined geographical scope of HELCOM HUB and EUNIS covers almost all European marine habitats (Table 1). Both schemes are heavily biased toward parts of Europe that have been well-studied (Galparsoro et al., 2012), as such, some regions are poorly represented (e.g. the Black Sea and the Canary Islands) due to limited information for refining the classes locally. The HELCOM HUB and EUNIS schemes cover the entire seabed from the intertidal zone into deeper, subtidal areas (EUNIS also extends into the abyssal zone), as well as some broad-scale pelagic habitats. Likewise, both the NISB and CMECS schemes are also designed for a broad set of habitats yet within specific geographic regions, that is NISB covers all of Australia’s territorial waters between the high tide mark and out to the limit of the photic zone (Table 1) and CMECS includes all estuarine, coastal, and marine waters under US jurisdiction in North America. Although initially developed for the Gulf of Maine region, the CSH scheme is a generic classification that can be applied to any continental shelf and shelf basin area. Other classifications have an even broader geographical scope. The PHCS covers both deep-water habitats within North America (Greene et al., 1999, 2005, 2007), as well as shallow water habitats (Vietti et al., 2001, Greene et al., 2007). The HFMHC has perhaps the broadest geographic scope through the inclusion of a global classification of large marine ecosystems (Sherman and Alexander, 1986) before moving to smaller and more distinct ecosystem units, for example estuary, and broad, geological formations such as drowned river valley. The influence of scope on habitat maps The sample of HCSs considered within this review span a range of habitats and geographical regions. Some schemes are broad in their scope by design, whereas others have grown to include new areas, such as the PHCS, EUNIS, and the CSH. Classes in locally calibrated classification schemes are more likely to match the observations made in similar habitats or geographical areas. In contrast, classes within broader, generic schemes are likely to have to generalize class descriptions, thereby diminishing the ability of the scheme to reflect localized variation (reduced specificity) in habitats. However, habitat maps generated with broad-scale HCSs are more likely to be compatible with other maps and contribute to national and international mapping efforts. Furthermore, the output format and classes of maps using broad-scale HCSs will be familiar, and hence more applicable, to more end-users that are already acquainted with the coding and purpose of the selected HCS. Spatial scale covered by a habitat classification scheme The seabed can be characterized and classified at different spatial scales ranging from the fine-scale, local environment (∼10s–100s metres), with factors affecting individual organisms, through to landscapes and regions (∼100s–1000s metres) where the substrates and terrain influence biological heterogeneity, and finally to the broadest scales at the national and international level (∼10s–100s kilometres) where oceanographic settings influence communities and populations (see Gregr et al., 2012 for a useful review of HCSs at the biogeographic scale). Variation in the spatial scale between habitat classification schemes Progression through both the EUNIS and HELCOM HUB hierarchies results in finer thematic resolution as well as a finer spatial scale, for example a level 5 habitat is expected to cover a smaller area than its parent habitat at level 4. Helpfully, both schemes also provide an indication of the minimum spatial footprint for the finest units, for example as a working guide, biotope units extends over an area of at least 5 m × 5 m, but can also cover many square kilometres, such as for extensive offshore sediment plains. For fine-scale features, such as rockpools and overhangs on the shore, this “minimum size” can be split into several discrete patches at a site. The NISB scheme may be applied to fine scales, while the upper tiers of the classification hierarchy, which have a reduced number of habitat classes, may be applied to broader, regional scales. The NISB scheme is particularly helpful in that it defines a “reference area” of 9 m2, for the assessment of habitat and biota dominance. Class modifiers applied to fine-scale features must be applied at the scale of the reference area as a minimum. This reference unit was deemed appropriate for a range of techniques and a practical measure that can be easily made in the field with the current observation sensors and methods, such as videography and diver. To allow for the varying scales of map production and use, the PHCS recognizes and defines four, nested spatial scales, that is micro-habitats, macro-habitats, meso-habitats, and mega-habitats (Table 1). The appearance of specific habitat scales can, therefore, be linked to the scale of observation, thereby aiding the production and visual interpretation of the maps, for example using dynamic segmentation methods such as those detailed by Nasby-Lucas et al. (2002). The tiers associated with the HFMHC scheme are also associated with specific spatial scales, but no strict spatial constraints are set for any level, thereby allowing any project to be fitted within the framework. Equally, CMECS is designed to operate at multiple spatial scales and provides the specificity needed for finescale applications. Like the previous two schemes, each level within CMECS is associated with a specific spatial scale, ranging from 10–1000 km2 at the first “regime” level to 1–100 m2 at the final “biotope” level. As such, CMECS allows the aggregation and assessment of classified units across diverse systems at regional, national, or global scales without loss of utility at local levels. These scales are useful in guiding the mapper during the interpretation of both survey observations and the classification scheme. The influence of spatial scale on habitat maps The consideration of scale is relevant for several aspects of habitat classification, map production and usage. First, the scale, and associated spatial resolution of a scheme determines which physical or ecological features can be represented on a map and what level of habitat heterogeneity can be captured. It is recognized by most mappers that many spatial units of classified habitat are mixed classes or mosaics. For simplicity, spatial units are typically labelled according to the dominant class and information regarding secondary habitats either removed or appended as a modifier. HCSs associated with finer spatial scales reduce the need to generalize mosaicked habitats and thereby better reflect heterogeneity at more scales. It should be noted that it is rarely stated within HCSs that units must be mutually exclusive, that is multiple habitat codes can be attributed with either a proportion or probability and then allocated to a single, spatial unit. Second, the scale of the HCS may also determine the type of mapping information, and therefore mapping method, required for the classification. For example, deep-water acoustic surveys may not have the required resolution for the identification of habitat classes with small footprints, whereby requiring the use of autonomous underwater vehicles (AUVs)-mounted sonars for data collection. Furthermore, schemes that stipulate minimum mappable units and area thresholds for habitat classes also benefit the mapper and reduce the number of subjective decisions that might be needed during the production of maps. The final issue is that the scale addressed by the HCS also defines the type of management supported by the maps. For example, localized impact assessments will require maps with a sufficient resolution for the accurate prediction of impact. Thematic resolution covered by a habitat classification scheme The thematic resolution specifies how fine the increments are between classes within a parent habitat. For schemes with a high thematic resolution, one might expect a high number of classes, each separated by relatively small differences in environmental or biological variables. In contrast, low thematic resolution would entail a small number of coarser habitat classes. Variation in the thematic resolution between habitat classification schemes The most detailed levels in the EUNIS and HELCOM HUB classification schemes are predominantly defined by biotopes and therefore separates classes according to small, but significant, biological differences in otherwise similar habitats. In EUNIS, many of the biotopes at levels 5 and 6 originated from statistical clustering analysis of grab sample data (for sediment biotopes) and expert interpretation of data from in situ diver surveys and intertidal surveys (for rocky biotopes) in the EC Life Nature-funded BioMar project (Connor, 1997). Equally, level 5 biotopes in the HELCOM HUB scheme were defined by analysing more than 50 000 data observations (i.e. video data, diving observations, grab samples) using spatial and statistical methods as well as expert judgment. The PHCS, CSH, and the NISB scheme use modifiers to provide greater thematic resolution and flexibility for the finest classes present. The PHCS uses single letter modifiers that describe specific aspects of geology, biology, topography, and seabed texture. These modifiers can be allocated to any of the six-letter habitat codes used by the scheme. There is no limit to the number of modifiers that can be attributed to each habitat code. Similarly, three themes within the CSH classification also provides modifiers that allow the user to describe “biological” “habitat association and usage” as well as short descriptors for “community disturbance and recovery”. Developing the use of modifiers further, the HFMHC scheme permits the use of user-generated classes (typically at the “data analysis” level) and modifiers at most of the levels within the classification, which therefore allows for any type and level of thematic resolution. Units of information at the lowest levels of the framework can include a variety of relevant information such as absolute values of abundance, dietary composition for dominant species, rates for species-specific ecosystem functions, and observed ranges for important physico-chemical characteristics. The influence of thematic resolution on habitat maps For the majority of the schemes, the finest classes are resolved according to biological characteristics of sessile benthic species. In some HCSs, more resolution is provided through the use of class modifiers rather than distinct classes. Such information displayed with classified habitats on the same map is likely to be valuable to a variety of map users. However, modifiers that unduly extend the basic classification of a habitat (i.e. “what it is”) are likely to complicate the habitat representation into maps, their interpretation by end users, and reduce comparability between maps. The greatest level of thematic resolution differs substantially between HCSs. This is due to either a shortage of information for the formation and validation of these most detailed classes or that the overall purpose and scope of the HCS does not concern itself with detailed biological information. Furthermore, it cannot even be assumed that thematic resolution is always consistent within the same level of a hierarchical scheme, as the most logical distinction between one set of communities does not always occur at the same level of detail as another. Regardless of the HCS used, mappers must be aware of the level of the classification that can be safely supported by the survey data, for example what level of community classification can be supported by epibenthic video, and what the intended purpose of their map will be. Equally, to improve the compatibility of maps, attempts should be made not just to standardize the use of HCS (or suite of HCSs) for mapping but also to set the level of classification within a scheme for a specific mapping technique (matched to a specific purpose). Structure of a habitat classification scheme The structure of HCS can be either hierarchical or flat, as well as nested or un-nested (parallel hierarchies). For hierarchical structures, the highest tiers typically separate observations into coarse classes using broad physical and chemical variables (physiographic approaches to classification). Lower tiers proceed to refine the classification based on more localized, physico-chemical variables, as well as biological information on the composition of the communities present (zoogeographic approaches to classification). Flat classification structures do not nest classes under predefined physico-chemical pathways. As such, flat structures allow the user to combine physico-chemical classes with independent biological classes—such classifications may not be possible within hierarchical structures if the required biological class is not nested within the observed physico-chemical pathway. The restrictive nesting of classes within hierarchical structures is only a significant issue when the training data used to develop the HCS was not reflective of habitat conditions apparent throughout the intended area of application. Variation in the structure between habitat classification schemes EUNIS, HELCOM HUB, and CMECS (substrate and biotic components only) are all hierarchical schemes with six levels of marine classification. For example, the first two levels of the CMECS scheme separate observations according to (i) salinity, geomorphology, and depth, and then (ii) by substrate type or water mass characteristics—additional levels sort observations by (iii) physical zones, (iv) macrohabitats (large and physically complex units containing several habitats), (v) habitats defined by physical and energy characteristics and finally, (vi) by characteristic biological composition. This structure is similar to both EUNIS and HELCOM HUB. For both systems, the structure of the hierarchy assumes that classes at the same level are mutually, and hence spatially, exclusive. Equally, specific communities and biotopes in the lower levels of the hierarchy are nested under specific physical conditions (defined by higher levels) and are not transferable between physical habitats. The NISB scheme is also hierarchical but with fewer levels. At the higher levels of the hierarchy, the NISB scheme assumes spatially exclusive habitats. The scheme uses “decision rules” for attributing habitat classes and for allocating geomorphic, biological, and environmental modifiers. These decision rules allow simple, unambiguous interpretation of survey data and facilitating the objective and consistent assignment of habitat classes. The decision rules are framed to be as sensor/method independent as possible. The PHCS is also hierarchical but has an un-nested structure. This scheme has separate attribution pathways for the classification of broad scale (megahabitats and then mesohabitats) and fine scale (macrohabitats and then microhabitats). The broad-scale classification uses various environmental parameters to provide increasingly finer thematic classes. The fine-scale pathway initially attributes the seafloor according to geological and coarse biological classes, and then followed again by textural attributes. Similarly, the lower levels of the HFMHC (Guarinello et al., 2010) scheme has three parallel (un-nested) benthic, water column and human hierarchies (Table 1). The use of separate components within the framework avoids the difficulty of generating a single hierarchy for fundamentally different domains and the flexibility and structure of this framework allow for a broader storage of information. However, the interaction of the three hierarchies generates a large number of unique habitat classes. The CSH scheme is quite different in structure to the other schemes considered, as it is structured round eight “themes” as the major subject elements of the classification. The themes all reside at the top level (i.e. are not hierarchical) and are applied to the classification of each site. Below the themes, a sequence of more hierarchical subclasses, categories, and attributes address habitat characteristics with increasing detail. This scheme was developed to be used exclusively for mapping purposes. As such, it was designed with a flexible structure to account for both data availability while maintaining a framework that is considered the best method of representing the habitats on maps based on the classification. The classification can accommodate new classes, subclasses, categories, and attributes, and it can easily be modified or expanded to address habitats of other regions. The influence of structure on habitat maps Hierarchical schemes allow habitats to be aggregated to a coarser level, thus allowing comparisons to be made between different studies using the same scheme, even when different levels of detailed information are available. These comparisons, however, are only possible if the HSC is interpreted consistently, and rests upon a thorough understanding of the scheme and how best to classify information using the scheme. A nested structure will provide a smaller but more targeted number of possible classifications—this is likely to benefit consistency and compatibility between studies. However, Galparsoro et al. (2012) reported that for EUNIS (a nested hierarchy) some communities occur in different main branches of the hierarchy due to their variations in associated depth or sediment type, whereas in reality, they are very similar. Equally, some communities only occur in a single branch of the hierarchy because they are mainly associated with certain physical conditions; however, if the same community is observed with a different set of physical conditions, then it would not fit precisely in the existing category. This presents problems in interpreting the difference between maps, as one mapper may favour the physical characteristics in classifying the habitat and another may favour the biological characteristics; thereby classifying the same area of seabed as different habitat types. Schemes with an open structure provide the user of the classification more flexibility to generate classes not previously documented during the development of the classification. Open, un-nested structures are perhaps best suited for mapping in areas that may be poorly represented within more trained and structured classifications. Compatibility of a habitat classification scheme for habitat mapping Although several HCSs have been designed specifically for mapping studies, this was not the intended purpose for all of the HCSs used in habitat mapping. As such, some of the decision points or environmental and ecological parameters that structure HCSs may not be routinely collected, or possible to observe, using the methods routinely deployed for marine habitat mapping. As such, the ease with which a HCS can be applied to mapping data can vary. HCSs that are designed specifically for mapping are more likely to be aligned to the commonly collected variables and include quantitative thresholds or decision points appropriate for these types of data and value ranges. Variation in the compatibility for mapping between habitat classification schemes for habitat mapping EUNIS was developed initially by piecing together several national and regional schemes, which were not all developed with mapping in mind. Regardless, EUNIS has been used extensively for mapping and modelling efforts in Europe. Some of this is a result of EU-funded projects designed to encourage a coordinated approach across Europe to marine conservation, assessment of the status of marine waters and spatial planning. For example, the EMODnet Seabed Habitats initiative has produced and maintains a pan-European, broad-scale EUNIS habitat map (known as EUSeaMap: Cameron and Askew, 2011; Vasquez et al., 2015; Populus et al., 2017) as well as collating local habitat maps produced for various purposes and translating them to the common EUNIS scheme. Until now, HELCOM HUB has been applied for the assessment and declaration of marine and coastal Red List biotopes and biotope complexes of the Baltic Sea region (HELCOM, 2013b) and in national case studies (e.g. Schiele et al. 2014, 2015). However, the use of the light penetration depth as a major structural variable in the HELCOM HUB scheme means that additional observations (not typically collected during marine habitat mapping) or external modelling outputs must be combined with the mapped variables to generate a classification. The same holds true for EUNIS regarding light availability and wave exposure at the seabed. The NISB scheme is interesting in that it provides an umbrella scheme that can adopt and amalgamate other classification schemes into its hierarchical system, that is the NISB scheme can be used to translate existing local habitat maps into a single, aligned product (Hilbert et al., 2007). The flexibility of this scheme allows old maps and mapping data to be translated into new and aligned products. The EUNIS scheme has been criticized for incompatibilities between the information used to define classes and that typically collected during a mapping survey. Levels 5 and 6 of the hierarchy are based on data from a wide variety of sampling techniques; as a result, they describe different aspects of seabed habitats. For example, some biotopes describe infaunal communities, while others describe epifaunal communities. It has been argued that some biotopes can only be identified if the method used during survey work is the same as the method used to originally define that biotope. For example, the characteristic species defining the level 5 biotope “Hesionura elongata and Microphthalmus similis with other interstitial polychaetes in infralittoral mobile coarse sand” are tiny polychaetes that would be grossly under-sampled using all but the finer meshes for sieving sediment. The 1 mm sieve used as standard on offshore surveys would not retain meiofauna such as these polychaetes (Parry, 2014). The classes within the PHCS are mostly defined by their geological character. As such, the scheme is well suited for the detection of habitats using acoustic remote sensing and thereby increases the confidence in the resulting classification. However, the biological classes are coarse, exclusively epifaunal and taxonomically distinct, which are perhaps unreflective of the typical composition of many seafloor communities and means that seafloor biota only have a fairly minor influence on the overall classification. The CMECS scheme is designed to be compatible with a range of sampling methods, for example cameras and certain acoustic devices can be used to identify the higher classification levels, while traditional point sampling methods, such as sediment sampling using grabs, can be used for the lower levels of classification. Equally, the sediment classes within CMECS are aligned to the Folk sediment classification, which is an established scheme in marine habitat mapping. This differs from the EUNIS classification which is underpinned by a “modified” (simplified) Folk classification. Another issue that improves the compatibility of HCSs for marine habitat mapping is the use of quantitative thresholds for defining classes. Common schemes, such as EUNIS and CMECS that are both used in national and international mapping programmes, lack quantitative definitions that could define classes. For EUNIS, the absence of these definitions is a result of it being constructed from several classification schemes, making it difficult to achieve consensus on what those definitions should be. The large part of the scheme that originated in Connor et al. (2004) was designed primarily as a biological classification system, with the physical descriptions at the higher levels being convenient groupings that did not necessarily need to adhere strictly to any definitions. HELCOM HUB provides quantitative delineation and classification rules within each of the classification levels. As an example, the system differentiates between soft and hard bottom substrata (level 3), by a spatial coverage percentage of hard substrates within a given area (HELCOM, 2013a). The latter also holds true for the delineation between infaunal and epifaunal dominated biotopes (level 4), and between epifaunal communities (level 5) and dominating species (level 6). Other HCSs also incorporate quantitative thresholds, for example the Australian NISB scheme also uses decision rules (such as quantitative measures, percentage cover thresholds, and particle size bands) at all levels of the hierarchy and for the class modifiers. The PHCS uses objective methods to calculate specific attributes, such as rugosity and slope, to reduce subjective attribution and delineation, and clear thresholds that separate classes, for example depth ranges for megahabitats or particle size for substrata. However, some attribute classes lack quantitative definitions which could lead to subjectivity, and hence variation, during the manual delineation of features. The influence of compatibility for habitat mapping on habitat maps The ease with which habitat mappers can use a HCS is based on the compatibility of the scheme’s classifying variables with survey outputs. For example, in the PHCS several of the classification attributes are generated specifically from common acoustic parameters such as depth (for bathymetric zones, slope, and rugosity) and backscatter (for hardness). Most of the geomorphological classes for other attributes are easily identifiable from full coverage bathymetric surfaces. However, it is clear that the ease and accuracy of classification also varies between habitat types. For example, it may be relatively straightforward to distinguish rock from muddy habitat in multibeam echosounder backscatter data, while there may be no clear boundary between coarse and mixed sediment. At the more detailed levels, many of the differences in the communities cannot be distinguished in acoustic data and therefore they are difficult to map. Difficulties in finding an appropriate class can be further compounded when HCSs are biased toward the habitats used in the initial development of the classification. For example, the marine component of EUNIS is primarily based on the British-Irish BioMar scheme, which was originally developed largely using UK near-shore data, primarily from grab sampling and, to a lesser extent, diver surveys (Connor et al., 2004). This means that EUNIS is less well-developed for offshore habitats, particularly those occurring on hard substrates. Furthermore, EUNIS is arguably less well developed for interpretation of data from remote video techniques which sample different parts of a biological community than divers or grab samples, and at a different scale, therefore posing difficulties in matching the communities from video/photographic techniques to the statistically driven clusters from grab sample and diver surveys. Similarly, certain classifications have been developed to use certain data types, for example schemes developed for the interpretation of satellite imagery (e.g. Mumby and Harborne, 1999), and may therefore not apply to data obtained from other sources. Recommendations for the use of marine habitat classification schemes in marine mapping Following on from the summary of how HCSs vary and how different aspects of HCSs can influence marine habitat maps, recommendations are listed and then discussed in more detail, below: Habitat mappers to label “realized” and “potential” habitat within maps to improve the communication of uncertainty. HCS custodians to include quantitative definitions of classes within HCSs to improve consistency in classification. HCS custodians to consider the applicability to habitat mapping when defining habitat types in a HCS. Provide information on habitat modifiers in habitat maps: a. HCS custodians to provide guidance on when and how to apply modifiers for their HCS. b. Habitat mappers to include information on relevant modifiers in habitat maps. Provide access to a broader array of attribution for each class: a. HCS custodians to make their HCSs available on an online vocabulary server. b. Habitat mappers to include in their digital maps a unique resource identifier that links to the online vocabulary for each habitat class. c. HCS custodians update their habitat-type summaries to include additional attributes such as sensitivity, rarity, and ecosystem service provision. Habitat mappers to use additional attributes described above to produce multi-purpose marine maps. HCS custodians to align and standardize the data inputs required for difference HCSs. Label “realized” and “potential” habitat within maps to improve the communication of uncertainty Many habitat maps present an unspecified mixture of “realized” and “potential” habitats when using HCSs. For example, the upper classification levels of many HCSs divide areas by geomorphology and rely on acoustic survey data to achieve this delineation. Continuous bathymetric surfaces can, therefore, confirm the presence of large, physical features from observations. Observations of biotopes are only provided by point (e.g. grab or photographic still) or line (e.g. video transect) sampling during ground truthing. The continuous distribution of the biotopes is then predicted using geo-spatial modelling or expert judgment, meaning that the resulting distribution is an extrapolated product not fully supported by direct observation (unless one is mapping a biogenic biotope with a detectable structure). The predictor variables typically used to model the distribution of these biotopes also fail to represent influential biological processes such as competition, predation, and dispersal (Brown et al., 2011). As such, one is modelling “potential” habitat for that biotope, which may or may not be occupied by the species constituting that biotope. The distinction between features that are realized versus potential habitat is rarely explicitly expressed when presenting mapped habitat classes. A lack of specificity may contribute to inaccurate assessments of the confidence of habitat maps by end-users, uncertain assessments of extent, and ambiguity about the relevant management action for sites and feature. It is therefore recommended that maps label habitats and biotopes with potential (modelled and potentially not occupied) and realized (delineated by direct observation) habitat labels or modifiers. Include quantitative definitions of classes within HCSs to improve the consistency of habitat classifications The use of habitat classification involves accepting some of the inherent assumptions associated with HCSs. An assumption common to all schemes is that all habitats can be classified into distinct and identifiable classes. It is often the case that observations, collected during habitat mapping surveys, fail to fall neatly into classes within a scheme. The presence of ecotones and mosaics of heterogeneous habitat reduces the clarity of class membership, and hence the ability to accurately reflect conditions on the seabed. The difficulty in classifying a continuous variable into a discrete class is further complicated when HCSs lack a quantitative definition, or clear “decision rules” for each class. Also, as habitat mapping has become more based upon physical measurements in the past 15 years (e.g. Al-Hamdani and Reker, 2007; Cameron and Askew, 2011; Galparsoro et al., 2015; Vasquez et al., 2015; Populus et al., 2017), there has been an increasing demand for quantitative definitions (Galparsoro et al., 2012). Without this information, qualitative classifications are often open to subjective interpretation and inconsistencies between studies or adjoining maps. The use of quantitative attribution will also provide a more robust basis for: (i) initial classification of habitats; (ii) the estimation of how well the observation fits the assigned class; and (iii) greater certainty about the detection of change over time during repeat mapping. Ideally, quantitative thresholds and class definitions should not be biased toward the survey techniques that were used to initially define classes and should include an indication of how the biotope may appear using a variety of survey techniques. Consider the applicability to habitat mapping when defining habitat types in a HCS HCSs designed for habitat mapping, and aligned to the types of information typically collected, are likely to be easier to use, reduce subjectivity during the classification of seabed information and generate more accurate maps. It is recommended that HCS custodians have an effective system in place for updating either their structure or classified units, and to consider strongly the applicability of new or updated habitat-type definitions to mapping. Provide information on habitat modifiers in habitat maps Modifiers are an extremely useful structural component for appending additional information onto a class without necessarily complicating the production or display of habitat maps. For example, modifiers could be used to represent: (i) observations on the condition of habitats; (ii) evidence of anthropogenic pressures (e.g. litter, physical alteration); (iii) labels for habitats that are hard to classify (e.g. fall between classes or units containing a mosaic of classes); or (iv) associations with other biological features not covered by the HCS such as large shoals of fish. Therefore it is recommended that: HCS custodians provide guidance on when and how to apply modifiers for their HCS. Habitat mappers include information on relevant modifiers in habitat maps. Provide access to a broader array of attribution for each class A name or code for a habitat provides a unique and brief title for the classified feature, however, habitat classes are typically supported by a fuller description that may contain, for example the identity and relative abundance of characterizing species as well as the prevailing physico-chemical conditions present. This supporting information is typically detached from the map and just the class names or codes are presented. Therefore, it is recommended that: Custodians of HCSs make their HCSs available on an online vocabulary server. Habitat mappers include in their digital maps a unique resource identifier that links to the online vocabulary for each habitat class. This would make it easier for somebody to interrogate the digital map and find out the full details behind the habitat name or code. The use of unique resource identifiers would have the added benefit of standardizing the recording of habitat information in habitat maps so that they may be easily compared and combined with other maps. Beyond the communication of habitat characteristics, it is likely that additional attribution providing details, for example on class sensitivity, rarity, ecosystem services provided or related habitat types may be of great interest to the end user of a map. While HCSs’ habitat-type summaries (whether online or in report format) typically contain information to help characterize habitats, there is a growing body of work that has assessed some of these additional attributes for some HCSs. For example, Salomidi et al. (2012) assessed the goods and services, vulnerabillity, and conservation status of habitats in the Mediterranean; Tyler-Walters et al. (2018) assessed the sensitivity of EUNIS habitats in UK waters; Galparsoro et al. (2014) mapped ecosystem services provided by benthic habitats in the European North Atlantic Ocean; Evans et al., 2014 summarized the correspondence between EUNIS, MSFD predominant, and Habitats Directive Annex I habitat types; and JNCC (2018) maintains a database of these and other habitat-type correspondences. Therefore, it is also recommended that HCS custodians update their habitat-type summaries to include additional attributes such as sensitivity, rarity, ecosystem service provision, and related habitat types. This will make it easier for maps to display alternative types of information as well as more contextual information for the class name. Use additional attributes described above to produce multi-purpose marine maps Management outcomes are presumed to be more effective when based on specialized HCSs aligned to the topic of interest. Despite this, most mapping studies tend to produce just one map, or set of maps, based on just one adopted HCS. Based on the cost and effort required to gather the data used for habitat mapping, the practice of producing just one map, based on one HCS per study, is potentially inefficient and narrows greatly the breath of the mapping exercise. Each use or purpose should be linked to the most informative and appropriate HCS. It is therefore recommended that habitat mappers use several HCSs and to generate multiple map products, each with a dedicated purpose. For example, a suite of maps that offers the greatest utility might include, among others: (i) a generic, descriptive map for inventory purposes, (ii) a map attributed according to representativity, rarity or conservation value for the protection of species and habitats (design of marine-protected areas networks), (iii) sensitivity maps for supporting marine spatial planning and management, (iv) a map of ecosystem services for regional valuations and assessments, (v) maps of essential fish habitat for fisheries management, and (vi) geomorphological and surficial sediment maps for sediment dynamics, extraction, and mining. The suggested suite of mapped products is perhaps aspirational as it is unlikely that any one mapping programme would have the ability to collection all of the required data to support the production of all six outputs. However, certain data inputs are common to several outputs (e.g. bathymetry for geomorphology, essential fish habitat and inventory mapping), which will facilitate the production of multiple thematic maps. Following the implementation of the previous recommendation would ease this process, but even without this, habitat mappers may make use of existing data and products, such as habitat sensitivity matrices [categorical coding of habitat sensitivity against various benchmarked anthropogenic pressures and activities (e.g. Stelzenmüller et al., 2010, Tyler-Walters et al., 2018)], habitat classes attributed with ecosystem services [e.g. 62 EUNIS classes labelled with 12 ecosystem services provided by Galparsoro et al. (2014)], and habitat correspondence tables (e.g. JNCC, 2018) to enable the translation of inventory maps to those representing sensitivity, ecosystem services, and other HCSs. The production of a suite of map products does not hamper our ability to standardize or merge maps within a thematic area, but would require additional workload for mappers. Align and standardize the data inputs required for different HCSs The production of multiple mapped products and the translation of existing maps into new thematic outputs, using other HCSs, is hampered by inconsistencies in the collection and presentation of raw data. For example, the use of different variations of the Folk sediment classification between HCSs or how light penetration is derived, can generate inconsistencies between adjoining maps using the same HCS or prevent the reuse of data in new HCSs that do not support the raw data format. As such, it is recommended that effective HCSs are shortlisted for specific thematic productions (previous recommendation), and that the raw data requirements for each scheme are identified. It should be then possible to identify incongruous data formats between schemes, and therefore opportunities for data standardization (e.g. through alignment or reporting of multiple formats for raw data) that would ultimately facilitate the reuse of information in different HCSs. Conclusions Marine HCSs differ greatly within six key properties, due in part to their initially intended application and structure (i.e. whether they follow a strictly hierarchical approach to classification and how readily they incorporate modifiers for the incorporation of greater detail). Consequently, each HCS has specific strengths and weaknesses. These strengths and weaknesses, along with the inherent assumptions associated with the classification process, modify the final representation of habitats when mapped. It is important for mappers to be aware of how these properties and assumptions are transferred into marine habitat maps, and whether these constrain their subsequent use for a wider variety of applications. It was not the objective of this review to judge the quality of individual HCSs—other reviews of HCS, such as that by Gregr et al. (2012) for delineating ecologically and biologically significant areas, have addresses value for specific purposes. Decisions on how mappers use HCSs within the mapping process, which is independent of the properties associated with the HCS, also introduces additional artefacts and biases. Having identified all of these issues, recommendations have been provided for improving HCSs for marine mapping as well as enhanced working practices for mappers using these schemes. For example, limiting interpretation of data to fit only one HSC compromises the information we can communicate through our maps and limits their use to a wider range of stakeholders. It is hoped that implementation of these recommendations will lead to: (i) greater certainty and usage within mapping studies; (ii) more consistency between studies and adjoining maps; and (iii) increased use of mapped products by a greater diversity of end users. Acknowledgements Part of this work was supported by VAPEM project (Fisheries and Aquaculture Directorate of the Basque Government). References Al-Hamdani Z. , Reker J. 2007 . Towards marine landscapes in the Baltic Sea ecoregion. BALANCE Interim Report 117. Allee R. J. , Dethier M. , Brown D. , Deegan L. , Ford R. G. , Hourigan T. F. , Maragos J. et al. 2000 . Marine and estuarine ecosystem and habitat classification . Silver Spring, MD, NOAA/National Marine Fisheries Service, 43pp. (NOAA Technical Memorandum NMFS-F/SPO, 43). http://aquaticcommons.org/14565/ (last accessed 25 October 2018). Andersen J. H. , Manca E. , Agnesi S. , Al-Hamdani Z. , Lillis H. , Mo G. , Populus J. et al. 2018 . European broad-scale seabed habitat maps support implementation of ecosystem-based management . Open Journal of Ecology , 8 : 86 – 103 . Google Scholar Crossref Search ADS Auster P. J. , Lewis R. S. , Wahle L. C. , Babb I. G. , Malatesta R. J. 1995 . The use of side-scan sonar for landscape approaches to habitat mapping. In Applications of Side-Scan Sonar and Laser-Line Systems in Fisheries Research: Alaska Department of Fish and Game , pp. 1 – 7 . Ed. by O’Connell V. M. , Wakefield W. . Special Publication , Commercial Fisheries Management and Development Division Juneau, Alaska, pp. 9. Brown C. J. , Smith S. J. , Lawton P. , Anderson J. T. 2011 . Benthic habitat mapping: a review of progress towards improved understanding of the spatial ecology of the seafloor using acoustic techniques . Estuarine, Coastal and Shelf Science , 92 : 502 – 520 . Google Scholar Crossref Search ADS Cameron A. , Askew N. (Eds.). 2011 . EUSeaMap—Preparatory Action for Development and Assessment of a European Broad-Scale Seabed Habitat Map. Final Report. www.emodnet-seabedhabitats.eu/outputs. Connor D. W. (Ed.) 1997 . Classification of benthic marine biotopes of the north-east Atlantic. Proceedings of the Second BioMar-Life Workshop, Dublin, 10 September 1995. Unpublished, Joint Nature Conservation Committee. Connor D. W. , Allen J. H. , Golding N. , Howell K. L. , Lieberknecht L. M. , Northen K. O. , Reker J. B. 2004 . The marine habitat classification for Britain and Ireland version 04.05. Introductory text. In JNCC (2015). Marine Habitat Classification for Britain and Ireland Version, 15.03. ISBN 1 861 07561 8. Joint Nature Conservation Committee, UK. http://jncc.defra.gov.uk/MarineHabitatClassification (last accessed 25 October 2018). Council Directive 92/43/EEC of 21 May 1992 on the conservation of natural habitats and of wild fauna and flora (Habitats Directive), OJ L206, 22.07.92. Cowardin L. M. , Carter V. , Golet F. C. , LaRoe E. T. 1979 . Classification of wetlands and deepwater habitats of the United States. US Department of the Interior, US Fish and Wildlife Service. Davies C. E. , Moss D. , Hill M. O. 2004 . EUNIS habitat classification revised 2004. Report to the European Environment Agency—European Topic Centre on Nature Protection and Biodiversity, October 2004. http://www.emodnet-seabedhabitats.eu/PDF/GMHM4_EUNIS_2004_report.pdf (last accessed 25 October 2018). Dethier M. N. 1992 . Classifying marine and estuarine natural communities: an alternative to the Cowardin system . Natural Areas Journal , 12 : 90 – 98 . Directive 2008 /56/EC of the European Parliament and of the Council of 17 June 2008 establishing a framework for community action in the field of marine environmental policy (Marine Strategy Framework Directive) [2008] OJ L164/19. Evans D. , Condé S. , Royo Gelabert E. 2014 . Crosswalks between European marine habitat typologies–a contribution to the MAES marine pilot. ETC/BD Report for the EEA. Finck P. , Heinze S. , Raths U. , Riecken U. , Ssymank A. 2017 . Rote Liste der gefährdeten Biotoptypen Deutschlands—dritte fortgeschriebene Fassung 2017.—Münster (Landwirtschaftsverlag),—Naturschutz Und Biologische Vielfalt, 156, 637 S. Galparsoro I. , Borja A. , Uyarra M. C. 2014 . Mapping ecosystem services provided by benthic habitats in the European North Atlantic Ocean . Frontiers in Marine Science , 1 : 1 – 23 . Galparsoro I. , Connor D. W. , Borja A. , Aish A. , Amorim P. , Bajjouk T. , Chambers C. et al. 2012 . Using EUNIS habitat classification for benthic mapping in European seas: present concerns and future needs . Marine Pollution Bulletin , 64 : 2630 – 2638 . Google Scholar Crossref Search ADS PubMed Galparsoro I. , Rodríguez J. G. , Menchaca I. , Quincoces I. , Garmendia J. , Borja Á. 2015 . Benthic habitat mapping on the Basque continental shelf (SE Bay of Biscay) and its application to the European Marine Strategy Framework Directive . Journal of Sea Research , 100 : 70 – 76 . Google Scholar Crossref Search ADS Greene H. G. , Bizzarro J. J. , O’Connell V. M. , Brylinsky C. K. 2007 . Construction of digital potential marine benthic habitat maps using a coded classification scheme and its application . Mapping the Seafloor for Habitat Characterization: Geological Association of Canada Special Paper , 47 : 141 – 155 . Greene H. G. , Bizzarro J. J. , Tilden J. E. , Lopez H. L. , Erdey M. D. 2005 . The benefits and pitfalls of geographic information systems in marine benthic habitat mapping. In Place Matters: Geospatial Tools for Marine Science, Conservation, and Management in the Pacific Northwest . Ed. by Wright D. J. , Scholz A. J. . Oregon State University Press , Corvallis, OR . Greene H. G. , O'Connell V. , Brylinsky C. K. , Reynolds J. R. 2008 . Marine benthic habitat classification: what's best for Alaska? Marine Habitat Mapping Technology for Alaska , 2008 : 169 – 184 . Greene H. G. , Yoklavich M. M. , Starr R. M. , O’Connell V. M. , Wakefield W. W. , Sullivan D. E. , McRea J. E. et al. 1999 . A classification scheme for deep seafloor habitats . Oceanologica Acta , 22 : 663 – 678 . Google Scholar Crossref Search ADS Gregr E. J. , Ahrens A. L. , Perry R. I. 2012 . Reconciling classifications of ecologically and biologically significant areas in the world’s oceans . Marine Policy , 36 : 716 – 726 . Google Scholar Crossref Search ADS Guarinello M. L. , Shumchenia E. J. , King J. W. 2010 . Marine habitat classification for ecosystem-based management: a proposed hierarchical framework . Environmental Management , 45 : 793 – 806 . Google Scholar Crossref Search ADS PubMed HELCOM . 1998 . Red List of marine and coastal biotopes and biotopes complexes of the Baltic Sea, Belt Sea and Kattegat. Baltic Sea Environment Proceedings No. 75. HELCOM . 2013a . HELCOM HUB—Technical Report on the HELCOM Underwater Biotope and habitat classification. Baltic Sea Environment Proceedings No. 139. HELCOM . 2013b . Red List of Baltic Sea underwater biotopes, habitats and biotope complexes. Baltic Sea Environmental Proceedings. Baltic Sea Environment Proceedings No. 138. Hilbert D. W. , Hughes L. , Johnson J. , Lough J. M. , Low T. , Pearson R. G. , Sutherst R. W. et al. 2007 . Biodiversity conservation research in a changing climate, Australian Government Department of the Environment and Water Resources. 72. Huang Z. , Brooke B. P. , Harris P. T. 2011 . A new approach to mapping marine benthic habitats using physical environmental data . Continental Shelf Research , 31 : S4 – S16 . Google Scholar Crossref Search ADS JNCC . 2018 . Marine habitat correlation tables version 2018.01. http://jncc.defra.gov.uk/marinehabitatcorrelation (last accessed 25 October 2018). Lund K. , Wilbur A. R. 2007 . Habitat classification feasibility study for coastal and marine environments in Massachusetts. Massachusetts Office of Coastal Zone Management, Boston, MA. Madden C. , Goodin K. , Allee B. , Finkbeiner M. , Bamford D. 2009 . Coastal and marine ecological classification standard . National Oceanic and Atmospheric Administration and NatureServe . NOAA Charleston, SC, 77 pp. Mount R. , Bricher P. , Newton J. 2007 . National Intertidal/Subtidal Benthic (NISB) Habitat Classification Scheme. Australian Greenhouse Office; National Land & Water Resources Audit; School of Geography and Environmental Studies, University of Tasmania, Hobart, Australia. Mumby P. J. , Harborne A. R. 1999 . Development of a systematic classification scheme of marine habitats to facilitate regional management and mapping of Caribbean coral reefs . Biological Conservation , 88 : 155 – 163 . Google Scholar Crossref Search ADS Nasby-Lucas N. M. , Embley B. W. , Hixon M. A. , Merle S. G. , Tissot B. N. , Wright D. J. 2002 . Integration of submersible transect data and high-resolution multibeam sonar imagery for a habitat-based groundfish assessment of Heceta Bank, Oregon . Fishery Bulletin , 100 : 739 – 751 . Reiss H. , Birchenough S. , Borja A. , Buhl-Mortensen L. , Craeymeersch J. , Dannheim J. , Darr A. et al. 2014 . Benthos distribution modelling and its relevance for marine ecosystem management . ICES Journal of Marine Science , 72 : 297 – 315 . Google Scholar Crossref Search ADS Parry M. E. V. 2014 . JNCC marine habitat classification for Britain and Ireland: overview of user issues. JNCC Report No. 529. http://jncc.defra.gov.uk/page-6757 (last accessed 25 October 2018). Populus J. , Vasquez M. , Albrecht J. , Manca E. , Agnesi S. , Al Hamdani Z. , Andersen J. et al. 2017 . EUSeaMap, a European broad-scale seabed habitat map. 174 pp. doi:10.13155/49975. Robinson C. L. K. , Levings C. D. 1995 . An overview of habitat classification systems, ecological models, and geographic information systems applied to shallow foreshore marine habitats. Canadian Manuscript Report of Fisheries and Aquatic Sciences 2322. Department of Fisheries and Oceans Science Branch, British Columbia, Canada. 65 pp. Salomidi M. , Katsanevakis S. , Borja A. , Braeckman U. , Damalas D. , Galparsoro I. , Mifsud R. et al. 2012 . Assessment of goods and services, vulnerability, and conservation status of European seabed biotopes: a stepping stone towards ecosystem-based marine spatial management . Mediterranean Marine Science , 13 : 49 – 88 . Google Scholar Crossref Search ADS Schiele K. , Darr A. , Zettler M. L. , Friedland R. , Tauber F. , von Weber M. , Voss J. 2015 . Biotope map of the German Baltic Sea . Marine Pollution Bulletin , 96 : 127 – 135 . Google Scholar Crossref Search ADS PubMed Schiele K. , Darr A. , Zettler M. L. 2014 . Verifying a biotope classification using benthic communities—an analysis towards the implementation of the European Marine Strategy Framework Directive . Marine Pollution Bulletin , 78 : 181 – 189 . Google Scholar Crossref Search ADS PubMed Sherman K. , Alexander L. M. (Eds). 1986 . Variability and Management of Large Marine Ecosystems. American Association for the Advancement of Science (AAAS) Selected Symposium 99, Westview Press, Boulder, CO. 319 pp. Stelzenmüller V. , Lee J. , South A. , Rogers S. I. 2010 . Quantifying cumulative impacts of human pressures on the marine environment: a geospatial modelling framework . Marine Ecology Progress Series , 398 : 19 – 32 . Google Scholar Crossref Search ADS Tyler-Walters H. , Tillin H. M. , d’Avack E. A. S. , Perry F. , Stamp T. 2018 . Marine Evidence-Based Sensitivity Assessment (MarESA)—a Guide. Marine Life Information Network (MarLIN) . Marine Biological Association of the UK , Plymouth . 91 pp. https://www.marlin.ac.uk/publications (last accessed 25 October 2018). Valentine P. C. , Todd B. J. , Kostylev V. E. 2005 . Classification of marine sublittoral habitats, with application to the northeastern North America . American Fisheries Society Symposium , 41 : 183 – 200 . Vietti R. C. , Harris C. M. , Susanna M. G. , Iampietro P. , Kvitek R. , Greene H. G. , Chiantore M. , Giuliani S. , Giorgi F. 2001 . Terra Nova Bay Antarctica Special Protection Area (ASP Number 161), Bathymetry, Habitat Interpretation and Community Distribution Map. Istituto Idrografico Della Marine, Programma Nazionale di Ricerche in Antartide, Officine Grafiche Novara 1091 S.P.A., Genoa, Italy. Vasquez M. , Mata Chacón D. , Tempera F. , O’Keeffe E. , Galparsoro I. , Sanz Alonso J. L. , Gonçalves J. M. S. et al. 2015 . Broad-scale mapping of seafloor habitats in the north-east Atlantic using existing environmental data . Journal of Sea Research , 100 : 120 – 132 . Google Scholar Crossref Search ADS Yoklavich M. M. , Greene H. G. , Cailliet G. M. , Sullivan D. E. , Lea R. N. , Love M. S. 2000 . Habitat associations of deep-water rockfishes in a submarine canyon: an example of a natural refuge . Fishery Bulletin—National Oceanic and Atmospheric Administration , 98 : 625 – 641 . Zacharias M. A. , Roff J. C. 2000 . A hierarchical ecological approach to conserving marine biodiversity . Conservation Biology , 14 : 1327 – 1334 . Google Scholar Crossref Search ADS © International Council for the Exploration of the Sea 2018. 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)

Ovenden, Jennifer, R

doi: 10.1093/icesjms/fsy132pmid: N/A

Abstract The article is a recollection of the ups and downs of my career as a research scientist. It does not chronicle my career in the standard way; there is no timeline of events and only a few details of the type of research that I have devoted my life to. I feel the take-home messages are relevant no matter what type of science you do. The narrative jumps around between periods of my life, uses anecdotes freely and attempts to be informative, forthright, and entertaining. Although science is perceived as a profession, the career pathway can be uneven and forked. In addition to discipline-specific knowledge and experience, successful scientists need a multitude of skills; for instance, human management, logistics, and social media—to name a few. The field often attracts talented introverts for whom working in teams is challenging. The complexity of science is fascinating and powerful. In the fervour for discoveries, care is needed not to become isolated from friends and family by long working hours. For those of you who have gone part or full way down this path, there will be value in comparing your experiences to mine. Mostly I hope that younger scientists who are beginning on this path, or considering whether to enter, will take away some messages that will help them stand on our shoulders. Introduction The greatest moment of my career was when I picked up The Key. I opened the door to the room which would be my new university office and I stood there, taking it all in. For a minute I slowed my brain, which was trying to leap forward to all the practical details of taking up a new position (hmmm, where was all my equipment and how would I set up the lab?). Instead, I took a moment to reflect on the long, uneven path that had led me to this moment. It was July 2013. I had left a government job as Principal Research Scientist in the field of fisheries genetics and had returned to Australia after a 3-month contract at the Natural History Museum in Paris. I arrived at campus on a cold, rainy winter morning and walked into the building. A door on my left had my name on it; the building manager had the key. There was my new office, and along with it a new position as Associate Professor. I remember feeling a surge of huge relief and happiness. The room was organized for me and was ready, as promised. They were expecting me. They wanted me. Wow, I thought. I’m actually here. Why was this such an enormous moment? Well, for so many reasons, but mostly because for the first time ever, I felt like I had successfully tackled The Myths of what it takes to be a successful scientist. I had carved my own path to where I wanted to be. What are The Myths, you ask? It’s a fair question and not one that gets addressed very often as you’re starting out on your career as a research scientist. But I’m willing to bet you’ll meet at least a few of them, as you step out along your own path. For those of you who have taken this pathway already, reading about my experiences may help you reflect on where you’ve been and where you are going. So, let me give you a “heads up” and run you through The Myths. Myth No. 1: Your career will be a linear unwavering progression to success In my case, I began university when I was young (17 years old). From high school, I’d been given pre-exam matriculation to university, which set me firmly on the tertiary education pathway. I graduated with a BSc Hons at 21 from the Australian National University, and then worked as a laboratory technician in the Research School of Biological Sciences for a year. That was a good time to consider if a PhD and a career as a researcher or academic was really for me. I had no idea if I had the intellectual capacity to succeed at this and was quite surprised to be encouraged by the academic who supervised my honours research project. Having decided to go ahead, I searched for a supervisor who shared my research interests. I was (and remain) fascinated by animal speciation, evolution, and biodiversity, all of which was sparked by my passion in the diversity of the Australian bird fauna. I relocated to the University of New South Wales to work with Professor Ross Crozier studying evolution using genetic markers. Under his supervision, I caught up with all the genetics theory and practise that I had not learnt as an undergraduate, and attempted and achieved an ambitious research agenda in the field and laboratory. I thank him for that, as well as his staunch support for the later development of my career. I graduated with a PhD in 1984 when I was 27 years old. From there, I was awarded the first of two back-to-back Australian Research Council research grants that funded my salary and research for 7 years at the University of Tasmania. With Professor Robert White, I switched to studies on aquatic biodiversity, which was fortuitous in the short (no shortage of fish tissue for DNA extraction, before the days of PCR) and longer term (future career prospects in fisheries research). I felt my career was on the way. Six years after PhD graduation, I’d published five first-authored publications (Ovenden and Mahon, 1984; Ovenden et al., 1987; Ovenden and White, 1988; Ovenden et al., 1988; Ovenden et al., 1989), a prestigious review paper (Ovenden, 1990) and a paper in a highly regarded journal (GENETICS, published by the Genetics Society of America; Ovenden and White, 1990). But then a T-junction appeared in the road ahead. At the time the decision to have children felt just as significant as deciding to proceed with a PhD and a career in research. While it’s possible to step down from being a researcher, it’s impossible to walk away from your children. Everyone experiences parenting differently, but from the beginning it was a magical experience for me. I worked part-time when I had one child but by the time I was the mother of two children, I had willingly resigned from the university to care for them full time at home. I had no thoughts of starting up my research career again. I was proud of the successes that I’d achieved in what was a productive, albeit short, career. Caring for the children at home was a happy time for me. But, our family circumstances changed, and one option was for me to take up paid work again. An academic-friend encouraged me to apply for a post-doctoral fellowship at The University of Queensland. It was an ideal opportunity, as this university reserved some positions for women getting back to their careers like I was. The euphoria of being awarded that fellowship was soon tempered by the challenges of uprooting the family and getting settled in another city. Soon after starting the fellowship I was offered a permanent job as a research scientist in the Queensland state government. The government research station was on the outskirts of the city, and a world away from the familiar university environment. The work underway there was supremely practical and important. With my academic background, I wondered if I could do this type of work. However, there was an urgent need for my skills and I soon had many projects underway and a great team to work with. Onward and upward? Um, no …… After 15 years of constructive and acclaimed research, my team and I were sacked by an over-zealous conservative government that was sceptical of the value of investing in science. You can begin to see why I fell in love with The Key. It represented a fresh start in science following the closure of our government research program. A career in science is going to be stop-start, forwards-backwards, with major wins and potentially devastating lows. Busting the linear-progression model of a career in research is an important first lesson to understand. Myth No. 2: All you have to do to succeed is to be brilliant at science Great scientists make important breakthroughs in pure and applied science, publish articles in high impact journals, lead collaborative teams of esteemed researchers and work at prestigious institutions. But being great at science is not enough. To be “great” a lot of other skills needed: for example; finding mentors, negotiating effectively, being geographically, and mentally flexible enough to work at new places and across disciplines. When my team and I were made redundant by the government in 2013, I had continuing PhD students and ongoing and new research projects. For their sake and mine, I spent several months meeting with mentors and university colleagues. It was a challenging and uncertain time, but eventually led to The Key for me but unhappily not for all of my group members. Through this process I learnt a great deal and happily added negotiation to my skill toolbox. Successes are readily measured by publications, citations, and value of research funds, but none of this is possible without extras skills—negotiator, project manager, and human resources expert. Added to that are the mundane but essential aspects of day-to-day project and laboratory management. As the scientist-in-charge, you are responsible for running your laboratory. If you are lucky, sometimes you may have help. In government I was fortunate to have lots of help; a research assistant as well as various business, building, and station managers who were responsible for infrastructure at a higher level. However, the responsibility (and in some cases liability for safety) is yours. This became apparent soon after I received The Key; my new position required setting up and running a new research lab. I needed all my knowledge of basic genetics, procedures, protocols and safety regulations to set up the new laboratory and get it going. The research students were a huge help in this. I was soon managing the lab from the bottom—where do I empty the clinical waste? How do I order lab consumables?—to the top; advising on experimental design, guiding students into the literature, helping with data analysis and editing student scientific manuscripts and theses. If I didn’t know it already, these experiences made me appreciate the complexity of doing science and wonder at the depth of experience needed to explore its frontiers. Suddenly, skills that seemed to have nothing to do with science became essential. The examples about negotiation and logistics above are trivial given what scientists are expected to be expert in today. You can add outreach to the community, e-skills to engage with social media, ability to understand and contribute to policy development and translation of innovation into intellectual property to support potential commercialization. Basically, it would be two dozen different jobs in most other professions. To address this complexity throughout your career, follow the principles of scientific investigations. Step one is to identifying gaps in your knowledge, and step two is addressing them. Take advantage of formal training if available. Learn by observing the successes and otherwise of others. Have a go yourself. For me, this process is still ongoing. Myth No. 3: Science is a great profession for a talented introvert As a scientist, you will invariably work inside institutions and research groups, which are often nested inside large and complex organizational networks. Aside from the challenges of navigating this hierarchy, there are the challenges of personal day-to-day working relationships. You can see from ever-increasing numbers of co-authors on publications that it is essential to be able to work effectively in groups. Projects are focused less and less around a single type of data collection. Now many data types are brought together to test increasingly expansive and complex scientific ideas. To do this large collaborative teams are essential. It’s becoming more and more important to be able to build and work in teams to maximize outputs from minimal input funding, as well as coping with the increasingly diverse nature of projects. Because of this, in both pure and applied research, the need for talented specialists working in isolation is diminishing. Despite this, try to foster some part of your work that you alone have special expertise in. It’s good for self-esteem and single-authored publications are still possible (Ovenden, 2013). In an ideal world, everyone is friendly, happy, professional, and eager to work together to achieve high quality scientific outputs and team relationships are characterized by respect, resilience, diversity, conflict resolution skills, and willingness to share the load. In many workplaces this is the state-of-play, but be prepared for other situations. Your team mates are the same as the public you catch the bus with; diverse, complicated and at all stages in their personal and professional journeys through life. However, unlike the crowd on the bus, you have to work with them. This means you need to get to know them; if they work nearby learning their names and room numbers is a fine start. If you put on your sociable hat (not always easy for scientists who tend to be introverts) in some cases you’ll make friends for life. In the worse cases, you’ll encounter those who you really can’t understand, or work with, or communicate with. The reasons are as opaque as they are numerous; competition, jealously, and malice—who knows? A limited pool of experts available at any one time for collaborative work in specialist scientific fields may be a contributing factor. The ability to contribute productively to team work across a range of situations is a skill that comes with understanding your own strengths and weaknesses and gaining human relationship experience as you go along. The workgroup at the government research station I mentioned earlier was a challenge for me. I struggled to fit with the local culture. Talk at coffee and lunch centred on sport, fishing, and boating, all generally outside my interests. I remember losing it when plans were being laid for a social event involving running around the bush “shooting” each other. Having arrived from Tasmania in 1997, the Port Arthur mass shooting (April 1996; 35 people died) was too fresh in my mind and “paint-ball” was an activity completely unknown to me at the time. Even as my research portfolio grew (more publications, more projects, more collaborations, more funding), it took several years of growing together before I managed to effectively work side-by-side with them. By then, I had established sound working relationships with fisheries science people in other states (including the Northern Territory, Western Australian, and Tasmania) and other institutions (CSIRO, universities in Australia, and worldwide). It was sobering to learn that it was probably the uneasy daily interactions among workmates that was holding up collaborative work at the research station and not the nature of the science that we were undertaking. A friendly workplace is a happy and productive workplace. Teams definitely work better if everyone gets along. But, the “friendly” approach can backfire for younger scientists. Early in my career, a senior academic appeared to go out of his way to have regular and long conversations with me. It affected my daily productivity to the extent that I needed to catch up after hours. However, we were part of the same research group and it was reasonable that we had regular communication. Looking back, I can see the daily conversations were not strictly necessary, and that I should have developed some strategies to deflect these odd, time-consuming interactions. The #MeToo movement shows that it’s all too easy to cross the line from friendship and mentoring to something more sinister. For career-vulnerable younger scientists, it’s hard to come up with winning strategies in these situations. Happily, the issue has a higher public profile now, and early-career scientists (and workers at all levels) can and should have discussions with a wide range of peers and colleagues around these topics. Back then, navigating “friendly” was a minefield. On this issue we need to move forwards not backwards. Being a scientist can be lonely work, seemingly ideally suited for talented introverts. For me it often involves reading, writing, or doing maths on the computer. But, like I said, research is more and more becoming a group activity. Under these circumstances, I believe that the best work groups are those where folks know and like each other. To achieve this, you have to get to know each other personally, however the more time spent chatting at work means more time spent working outside office hours to maintain outputs. The antidote to this is socializing outside work. Ideally, events should be organized to allow senior and junior scientists to meet as equals. If you are lucky enough to be offered these types of opportunities at your workplace take them up, even if it seems weird to be rubbing shoulders with esteemed scientists on the cricket pitch for example. Out of hours networking helps team building without interfering with nine to five productivity, but events need to be held at family friendly time-slots or involve family friendly activities. Some groups use joint conference attendance to build and strengthen teams. I don’t have the solution here, but I feel strongly that work is best done among folk who can get along with each other as long as the boundaries are clear and accepted. As you become more senior, a “friendly” workplace transforms into another kind of torment. Members of your work-friendship group end up on teams that you manage. Then, you are in a situation where decisions that you have to take to achieve project outcomes will affect those in your work-friendship group. Even if the team members are not pals, these actions often lead to conflict. In my career, I feel there have been times when I should have taken hard decisions but didn’t. I now regret avoiding conflict and not imposing consequences on my team for missed deadlines. I let situations continue when I should have acted. With hindsight I can see that a tougher approach would have had a longer-term benefit for personal development on both sides, at the expense of short-term chaos. Scientists are generally not known for their outgoing nature and excellent verbal communication skills. But, increasingly science is not an individual sport, rather very much a group activity. You will need to recognize your strengths or weaknesses in this situation and take every opportunity to fine tune them. Myth No. 4: You can’t be a successful scientist and “have a life” as well Working one-hundred hours per week is not the only way to scientific success. Again, and again, I have seen colleagues who work seven days per week make significant achievements in their careers. It was (and still is) common for my colleagues to work on the weekend. Undoubtedly, this is the way to get ahead as you can make achievements that are not possible if your hours are constrained from Monday to Friday, and these achievements are important stepping stones for career advancement. Possibly, early in your career placing more emphasis on work rather than “life” may be acceptable, but I recommend pushing back against making it a habit. I could have achieved a great deal more in my career if I’d worked non-stop, and there have been times when I’ve been tempted, however, whether you realize it or not, working non-stop turns into a significant life decision. I didn’t go down this track because my children, my family, and my friends are essential aspects of my life. Although I love my work, I love these people and my non-work activities just as much. I have never underestimated the importance of a strong friend-network. When times get tough and your family is not around, it’s your friends that will help. When you are time-poor and more introvert than extrovert, establishing, and maintaining friend networks takes mental effort and often is hard work. The same applies to being a “good” wife, mother, sister, daughter. But, like I said, I regard all of this as essential. I rationalized that I was certainly not being paid enough as a research scientist to work on the weekends. The range of skills necessary to be an effective scientist would be rewarded with a much higher wage in other professions. “You can always take flex-time”, a colleague said to me when I was upset that I could not attend my son’s swimming carnival during working hours. This was true, and indeed more and more institutions are family-friendly enough to allow you to vary your work hours while maintaining productivity. However, the demands of being a parent and working full time are considerable and no amount of flexible working conditions can truly compensate. When it came to the demands from work compared to demands from my children, I always put my children first. Their need was undeniable, whereas I was never indispensable at work. To be successful under these circumstances is truly challenging and every parent develops their own “hacks” that allows them to cope. It’s essential to be really organized and to never stop thinking about moving things along. It could be planning to buy new school shoes or how to schedule sport and music practices, but equally it could be planning your next experiment, thinking about where your work colleagues are coming from or inventing titles for your next manuscript. I once asked a colleague without children what she thought about while commuting to and from work. I was surprised when she said nothing much, as for me this was (and still is) essential thinking time. Effective and productive scientists need to be happy, and I believe family and friendships are crucial for this. Don’t fall for the cliché that if you’re not working seven days a week, twenty hours a day, you’ll never achieve anything. Likewise, don’t believe that those who do work like that are automatically “better” than you. If you have significant achievements under your belt but keep a lid on your working hours, have confidence that you are not an “impostor” but a talented and valued member of the science community. Finally, not a myth but a fact: you are going to need resilience. Some wise parents give their kids a station-wagon and a road map of Australia when they leave high school. Others encourage their kids to take a more formal “gap-year” before taking up university studies. But, in my case, I began university when I was young and got some life experience working as a technician in a research lab for a year before starting my PhD. This only provided a limited amount of life experience. My PhD studies were challenging, but unlike many of my fellow graduates, I did not feel mature enough at the end of it to work overseas. Life experience is also important to make an informed decision about your career path. You want to be like Roger Federer and Rafael Nadal, who absolutely love tennis even after a lifetime of competition. I got into my corner of science (evolutionary and population genetics) because of my love of birds and their amazing biodiversity. Genetics was one tool to understand evolution and speciation, and so much more as shown by my long publication history (type Jennifer Ovenden into scholar.google.com to find out more about me). For me, my research field is just as fascinating now as it was when I started; perhaps more because now I know so much more. More than that, my love of science is driven by my need to know how things work. If that drives you too, then science is a good career choice. Due to the specialist work we do, scientists can be isolated from the working life of the majority of the community; whether it’s the public or private sector, social services or business, manufacturing, or sales. Working in collaborative, multi-disciplinary teams can mitigate this. Scientists often have a high level of management, human relation, logistic, financial, and social media skills, but when considering a career change soon come to realize that many professionals have these abilities too. It is relatively rare to find a position that values your deep knowledge of particular aspect of science, which often means that scientists largely do not move between jobs by choice. I was extremely fortunate to step back into a research career after several years at home with the children, which I attribute to having a good record of achievement before taking a break. Mental health and a positive state of mind is not negotiable; you can’t work without it. Even within a working day, my mental state can vary from strong (able to face and do anything) to weak (just want to go home). Science can be a cruel profession. Many (but not all) of your peers turn into ego-driven monsters given the opportunity to anonymously comment on your unpublished work or grant applications. Your peers do not do this intentionally. They are busy, sometimes not truly interested in your work but mostly they just forget to be compassionate. They genuinely believe that any comment, even negative ones, are useful. It takes an enormous amount of internal courage sometimes just to read comments received on journal submissions or funding proposals. It’s important to stay calm in these situations, but this often takes a lifetime to learn. Becoming disconnected from friendship-groups is another common cause of mental anguish. I suffered greatly leaving Tasmania and relocating to Queensland. Not only did I suddenly have a full-time job again after being full-time at home with the children, but I had little time to develop a new friendship group to replace the one I left behind. At the time, someone said that it would take 10 years to get happy again, which turned out to be pretty true. Thoughts of self-harm are an essential trigger to get professional help. I found a psychotherapist, whom I saw once a week for 2 years which was hugely beneficial. Research is specialized work and the knowledge and skills needed are considerable. Without them scientific discoveries will not be made, and innovation will come to a halt. Looking back, my resilience largely came from meeting and overcoming challenges in my personal and work life. Be prepared to listen, learn, and absorb from other scientists and people outside of science. Do your gap-analyses again and again; recognize your skills shortage and act on them. Conclusion I’ve covered a lot of ground here but let me go back to the importance of The Key. Receiving that key 29 years into my career was the first time I felt truly secure in science. I had busted through all those myths to get to my goal, which was to be able to continue my research work with freedom, purpose, and self-belief. While separating myths from facts is important, there is no definite recipe for success. The pathway through science is not well trodden, and for me, it’s been a bumpy ride. It’s easy to underestimate the breadth of skills needed to succeed. A higher degree will give you discipline-specific knowledge, but you need to be able to absorb and use other skills such as negotiation, team management, and logistics as you become more senior. Interpersonal communication skills are essential, particularly as science is becoming more integrative across disciplines, but for some of us communication requires great effort. Science is so seductive in its richness and depth, it can rival the importance of family, friends, and community. Guard your mental and physical health as you will need a great deal of this to provide the resilience to move forward. A career in science can be difficult to traverse, but by sharing my triumphs and self-doubts with you I hope I’ve showed that it’s possible to be successful whilst maintaining a healthy work-life balance. Acknowledgements Thanks go to Editor Howard Browman for the invitation to contribute this article. Nick Place gave me the courage to proceed and essential guidance on how to write this piece. Colleagues and graduate students (Amelia Armstrong, Mike Bennett, Danielle Davenport, Christine Dudgeon, Raewyn Street, Samuel Williams, and Jamie Wyatt), my son (Alex Hall) and two anonymous provided important feedback on earlier drafts. Footnotes Food for Thought articles are essays in which the author provides their perspective on a research area, topic, or issue. They are intended to provide contributors with a forum through which to air their own views and experiences, with few of the constraints that govern standard research articles. This Food for Thought article is one in a series solicited from leading figures in the fisheries and aquatic sciences community. The objective is to offer lessons and insights from their careers in an accessible and pedagogical form from which the community, and particularly early career scientists, will benefit. The International Council for the Exploration of the Sea (ICES) and Oxford University Press are pleased to make these Food for Thought articles immediately available as free access documents. References Ovenden J. , Mahon R. 1984 . Venereal transmission of Sindbis virus between individuals of Aedes australis (Diptera, Culicidae) . Journal of Medical Entomology , 21 : 292 – 295 . Google Scholar Crossref Search ADS PubMed Ovenden J. , Mackinlay A. , Crozier R. 1987 . Systematics and mitochondrial genome evolution of Australian rosellas (Aves: platycercidae) . Molecular Biology and Evolution , 4 : 526 – 543 . Ovenden J. , White R. 1988 . Mitochondrial DNA restriction site map for Gadopsis marmoratus . Biochemical Systematics and Ecology , 16 : 355 – 357 . Google Scholar Crossref Search ADS Ovenden J. , White R. , Sanger A. 1988 . Evolutionary relationships of Gadopsis spp. inferred from restriction enzyme analysis of their mitochondrial-DNA . Journal of Fish Biology , 32 : 137 – 148 . Google Scholar Crossref Search ADS Ovenden J. , Smolenski A. , White R. 1989 . Mitochondrial-DNA restriction site variation in Tasmania populations of Orange Roughy (Hoplostethus atlanticus), a deep-water marine teleost . Australian Journal of Marine and Freshwater Research , 40 : 1 – 9 . Google Scholar Crossref Search ADS Ovenden J. R. 1990 . Mitochondrial DNA and marine stock assessment: a review . Australian Journal of Marine and Freshwater Research , 41 : 835 – 853 . Google Scholar Crossref Search ADS Ovenden J. R. , White R. W. G. 1990 . Mitochondrial and allozyme genetics of incipient speciation in a landlock population of Galaxias truttaceus (Pisces, Galaxiidae) . Genetics , 124 : 701 – 716 . Google Scholar PubMed Ovenden J. R. 2013 . Crinkles in connectivity: combining genetics and other types of biological data to estimate movement and interbreeding between populations . Marine and Freshwater Research , 64 : 201 – 207 . Google Scholar Crossref Search ADS © International Council for the Exploration of the Sea 2018. 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)

Peterman, Randall, M

doi: 10.1093/icesjms/fsy141pmid: N/A

Abstract This paper describes my research on fish population dynamics, which has aimed to improve the information available for management and conservation. Through numerous collaborations, my research program addressed three main objectives. (1) Increase the understanding of spatial and temporal variation in productivity of fish populations. (2) Quantify uncertainties and risks in fishery systems and their implications for management and conservation. (3) Develop methods to reduce those uncertainties and risks. To help young scientists, I present 11 general lessons, as well as some specific advice, that emerged from that research. The general lessons include pursuing a path of continuous learning, going beyond your comfort zone to broaden your skills and knowledge, and collaborating with others. More specific advice for fisheries scientists includes evaluating the bias and precision of parameter estimation methods via Monte Carlo simulations, and considering multiple models of whole fishery systems. This paper also illustrates, with examples, how the understanding of some aspects of fish population dynamics has evolved, at least from the limited perspective of my own group's research. Introduction Fisheries scientists face an exciting but challenging future. On the positive side, I have never seen a more promising time for making progress in understanding fisheries systems and improving fisheries management and conservation practices. Scientists can now apply a wide variety of advanced statistical methods to large data bases, and powerful computers can run complex simulation models. There is also a strong desire by most young fisheries scientists to address real-world problems, and many universities recognize the value of such applied research, as well as collaborations with fisheries management agencies. However, despite these positive signs, young fisheries scientists face a wide array of serious challenges in the future. Even with decades of research and improved understanding, considerable uncertainty still remains about the complex structure and dynamics of fisheries systems, and climate change appears to be altering previously well-documented relationships. Scientists are expected to take these uncertainties into account in their analyses. It is also difficult to keep up with advances in knowledge and research methods published in the rapidly growing literature. To help address some of these challenges (not all of them!), I will provide some general lessons that I have learned, and more specific advice for fisheries scientists (in italics in the main text). I will also describe how my career has been shaped by chance events, mentors, and colleagues, including my students, research assistants, and post-doctoral fellows. Formative years Inspiration while an undergraduate My career as an ecologist started in earnest when, as an undergraduate biology major at the University of California at Davis (UC Davis) from 1966 to 1970, I learned in an ecology course about the exciting papers written by C.S. (Buzz) Holling on the components of the “functional response” of predation rate to changes in prey density (Holling, 1959, 1965, 1966). Holling broke the process of predation down into its components, did some critical experiments, and then derived equations for the predation process. As all ecologists now know, functional and numerical responses of consumers to changes in their prey density have subsequently been documented in a wide range of taxa. I was quite fascinated by Holling's new process-oriented “experimental components” approach, and I came back to it later. One day I was browsing in the university bookstore and came across an intriguing, newly published book by Kenneth E. F. Watt called “Ecology and Resource Management” (Watt, 1968). I immediately bought it because, unlike the well-known proverb, I could judge this book by its cover, which displayed a flow chart, computer code, plants, and animals. The book was filled with riveting examples of quantitative analyses of the dynamics of fishes, forests, and insect pests. The book even showed how to translate equations of ecological processes into FORTRAN computer code, a novelty at the time. I knew nothing about computer coding though, having stupidly turned down the chance to take a FORTRAN course in high school in favour of playing on the golf team! Not only did Watt's book pull together exciting advanced work on ecological systems, but his analyses were applied to resource management problems. He even explained how to find the best management options using optimization methods such as dynamic programming, a method that would not become widely used until many years later. Watt was well ahead of his time. Lesson 1: Challenge yourself to go beyond your comfort zone to learn something new, especially if it is a quantitative method. Knowledge of such methods will always be in demand by employers. My accidental discovery of Ken Watt's book is what set my career on a trajectory to be a quantitative ecologist. I could hardly believe my luck that Watt was a faculty member in Zoology at UC Davis, and that he taught a course in “Biomathematics”. Even though it was a graduate course, he allowed me to take it as a fourth-year undergraduate. Although the course was quite challenging, I am forever grateful for that opportunity; I acquired a quantitative perspective on ecology, including simulation modelling. That was not something that I anticipated learning when I started my undergraduate work in biology, so the experience was exciting to say the least! Ken Watt was a former colleague of Buzz Holling at the Canadian Forest Service Laboratory in Ontario, Canada, so when I decided to pursue a graduate degree in quantitative ecology, Watt naturally recommended that I apply to Holling's 2-year-old Institute of Animal Resource Ecology (IARE) at the University of British Columbia (UBC) in Vancouver, Canada. I was accepted there, not by Holling, but instead by a relatively new, young faculty member, Carl J. Walters. Now there was good luck! I was his first Ph.D. student. Branching out at the University of British Columbia Carl Walters is now well-known as one of the world's top quantitative fisheries scientists, but at the time I started at UBC in 1970, he had interests in a broader range of topics. He had come to UBC after finishing his doctorate at Colorado State University in Fort Collins, where he was one of the first people to be trained in the new field of systems ecology, including simulation modelling, as part of the Grasslands Biome Project. Luckily, Carl Walters supported my overly grandiose (in retrospect) initial proposal for a research program, which was to apply Holling's experimental components approach to understanding animal dispersal. I eventually narrowed that goal down to asking about the causes and spread of outbreaks of an insect pest, the mountain pine beetle in British Columbia's lodgepole pine forests. I did field work and simulation modelling on the interaction between mountain pine beetle populations and lodgepole pine trees in eastern British Columbia while collaborating with both Ken Graham in the UBC Forestry School and scientists at the Canadian Forest Service. My four years of Ph.D. work on this topic laid a solid foundation in population dynamics that I would apply later to research on fishes. Carl Walters was a stimulating person to be around. He constantly challenged ideas with his sharp critical thinking. He was also a truly gifted teacher. He taught us graduate students systems thinking and how to program simulation models in FORTRAN. He made it look easy. In February 1971, the first year of my graduate program, Carl offered a lucky few of us students the opportunity to help develop a simulation model of a lake ecosystem in a brief workshop funded by the International Biological Program. Our modelling team was to glean knowledge about state variables, parameters, and processes from visiting experts in limnology, phytoplankton, zooplankton, and fish, and then write the FORTRAN code for a simulation model to explore the system's dynamics. Despite the time pressure, we succeeded in developing a working model, although each 2-year simulation run took 45 min, during which we idly shot rubber bands at the ceiling lights! We later learned that this model helped to identify key areas for future research priorities on lake ecosystems. Lesson 2: Get involved in projects outside of your main thesis topic or research focus. Such experiences will broaden your skills, knowledge, and network of contacts. Although Carl Walters soon became a leading researcher in fisheries science and management, early on he was also well known for conducting week-long simulation modelling workshops, such as the lake biome one, in conjunction with people from various resource management agencies (Walters, 1974; Chapter 4 of Holling et al., 1978; Hilborn et al., 1984). The aim of these workshops varied, depending on the agency. They ranged from merely testing hypotheses to codifying in one place experts' knowledge about a system, identifying top priorities for future research based on sensitivity analyses with a model, and evaluating a range of potential management actions. The 21 simulation modelling workshops in which I participated as a programmer covered topics such as mallard ducks, caribou, the proposed James Bay (Quebec) hydroelectric project, the Mackenzie River basin, the spruce budworm insect pest, Beaufort Sea Lagoon, Pacific salmon, and fishing fleets. Thus, as a Ph.D. student, I was very fortunate to be exposed to issues across such a wide range of ecosystems and taxonomic groups. I learned how to collaborate as we quickly built computer simulation models in late-night “trials by fire”. I also got a first-hand look at the state of knowledge about dynamic components of a broad range of ecosystems (often disappointingly weak at the time). As well, I learned that problems facing resource managers were often compounded by vaguely defined management objectives, a situation that persists to this day. These workshops were thus incredibly valuable windows on the world of resource and environmental management that I otherwise would not have encountered in a normal Ph.D. program. That exposure to resource management issues and a chance to develop my modelling skills built an unplanned foundation for my future career, which included providing input to management policies (see later section, “Influencing policy”). In essence, I lucked out by being at the right place (UBC IARE) at the right time in the early 1970s. That was also a time when, by chance, a large number of outstanding graduate students and post-doctoral fellows, as well as some new faculty members, joined IARE. They were great sources of ideas, skills, and stimulating debates. As a result, Holling's Institute of Animal Resource Ecology soon became one of the world's leading institutions in quantitative ecology. During one of the modelling workshops described above, I learned that most scientists who worked on changes in abundance of Pacific salmon (Oncorhynchus spp.) focused their research on processes in fresh water, and that relatively little work had been done on the marine life stage. I thus decided to spend my post-doc year (1974–1975 at UBC) looking into mechanisms operating on Pacific salmon in the ocean. I convened a meeting in 1975 at which people such as Bill Ricker and Percy Wickett presented data and hypotheses about marine survival of salmon. Bill Ricker's quantitative approach, his incredible knowledge, and his analytical thoroughness inspired me to continue in this field after my post-doc year was over. The next fortuitous event in my career resulted from the actions taken in 1975 by Buzz Holling, who had just returned from a sabbatical year at the International Institute of Applied Systems Analysis (IIASA) in Austria. He managed to convince the Canadian Department of Environment to hire both Ray Hilborn and me as “Policy Analysts” to conduct publishable research at UBC and act as liaisons between UBC and the federal government in the area of applied ecology, including simulation modelling. I am forever grateful to Buzz for his initiative and leadership in setting up those positions. During the next 4 years, I interacted with scientists and managers from several federal management agencies and took part in more modelling workshops. I also continued with research on Pacific salmon. A new initiative at Simon Fraser University In 1979, I was fortunate to be hired for a faculty position at Simon Fraser University (SFU) in Burnaby, British Columbia, which is a suburb of Vancouver. I had a joint appointment, half in the Department of Biological Sciences and half in a brand new graduate program called the Masters of Resource Management (MRM) Program. My experience from the modelling workshops at UBC probably had more to do with my getting this faculty position than my Ph.D. work on mountain pine beetles and lodgepole pine. Both departments wanted me to teach courses in ecological modelling, among other topics. The MRM program literally opened the day that my appointment and that of the Director, Chad Day, began. It was exciting to be part of a brand new graduate school. Two years later I moved full time to MRM. The MRM program (renamed in 1990 as the graduate School of Resource and Environmental Management or REM) was characterized by research and courses that were both interdisciplinary and applied to real-world problems. These two features proved to be attractive to my future graduate students, post-doctoral fellows, and research assistants who were seeking such an environment. Members of my research group came from backgrounds not only in ecology, but also engineering, economics, computer science, mathematics, oceanography, and statistics. I benefited greatly from working with this interdisciplinary team. Lesson 3: If you want to make changes in your institution, create a long-term plan and a strategy to achieve it, and be persistent but patient. I pursued two initiatives to increase the critical mass of fisheries researchers at Simon Fraser University. The first was a plan to bring government-agency researchers to the campus full time, an approach that had been successfully demonstrated by the US Fish and Wildlife Service's Cooperative Research Units on dozens of US university campuses. However, because of the inflexibility of senior Canadian government officials, it took 8 frustrating years to bring the Cooperative Resource Management Institute to fruition, when several researchers from Canada's Department of Fisheries and Oceans finally moved to SFU. My second initiative, to increase the number of fisheries faculty members in my department, took even longer (12 years). I was fortunate to be awarded a senior Canada Research Chair in 2001, and as a result, we hired two more fisheries faculty members, William de la Mare and Sean Cox. The Chair's extra funding, as well as contributions from Canada's Department of Fisheries and Oceans, enabled us to bring in numerous excellent graduate students, research assistants, and post-doctoral fellows, as well as to build a high-end computer simulation lab. These two successful initiatives resulted from having a long-term vision, a strategy to achieve it, and repeatedly presenting the case, including the potential benefits, to all parties. It helped to be patient, too; institutions are typically slow to change. Research on fish population dynamics and management My overarching research goal has been to advance the understanding of fish population dynamics (particularly of Pacific salmon) and thereby improve the information available for their management and conservation. To achieve this broad goal, three main research objectives emerged during my career. Expand the understanding of key processes that cause changes over time and space in reproduction, growth, and survival rates of fish populations. Quantify the main sources of uncertainties in data and fishery systems, the risks created by those uncertainties, and their implications for management and conservation. Develop methods to reduce some of those uncertainties and risks, which in turn should increase the social and economic benefits derived from fish populations. I took two key long-term approaches to meet these three objectives. First I requested data from various management agencies from Alaska, British Columbia, Washington, and Oregon to create what eventually became a quality-controlled data set of 163 populations of four species of Pacific salmon: sockeye, Chinook, pink, and chum. Those data covered annual abundances of spawners, juveniles (where available), and adults produced by those spawners over periods as long as 50 years. We also gathered data on oceanographic variables. My second approach was to actively pursue projects that required learning about advanced statistical and simulation methods. To facilitate these two approaches, I collaborated with scientists outside of the university and also worked closely during my career with 70 graduate students, full-time research assistants, and post-doctoral fellows, plus two undergraduate research students. Although I officially was supervisor for the young scientists in my group, I was open to learning along with them, as well as from them, especially about methods such as hierarchical statistical models, Bayesian analysis, decision analysis, Dynamic Factor Analysis, Markov Chain Monte Carlo methods, and receiver operating curves. Our research was quite productive for these young scientists, and several of our publications were recognized with awards from various organizations. Below I summarize my research and the affiliated lessons under my three main research objectives, although there are inevitable overlaps. Research objective 1: expand the understanding of key processes that cause changes over time and space in reproduction, growth, and survival rates of fish populations Density dependence during the ocean life of salmon Lesson 4: Novel findings that run counter to the prevailing wisdom are often initially ignored by other people, but if your data and methods are solid, persist with that research. High-quality science will eventually prevail, though probably more slowly than you would like. As I mentioned above, most salmon researchers were working on freshwater mechanisms as potential drivers of salmon abundance when I began my research on processes in the marine environment. My first publication on the latter topic showed that density-dependent survival (lower survival rate at high abundance) occurred during the ocean life stage for a few salmon populations (Peterman, 1978). After that paper appeared, one salmon scientist told me that the idea of density-dependent marine survival was preposterous because the ocean was “just one big frog pond”. Fortunately I ignored him, and my research group subsequently documented widespread evidence of important density-dependent interactions in the ocean (1) among fish within a salmon population, (2) among fish between separate salmon populations, and (3) among salmon from different nations (Peterman, 1982a, 1984a, 1987; Pyper and Peterman, 1999). I also found that most between-year variation in density-dependent survival rate and growth rate of sockeye salmon occurs in the first year and a half of the marine life stage (Peterman, 1982a, b, 1984a), which helped focus further research on survival on that portion of marine life. These results were only slowly acknowledged by other salmon scientists, but many other researchers subsequently found extensive evidence that substantial density-dependent growth and survival interactions occur within and among species, and that the density-dependence is more likely due to limited food supply than to processes such as predation (Kaeriyama, 1998; Ruggerone et al., 2003; Ruggerone and Connors, 2015; Batten et al., 2018). Due to the spatial and temporal overlap of salmon from nations that border the North Pacific Ocean, these findings have stimulated discussions about the negative effects of salmon hatchery releases on wild salmon and the need for international coordination to regulate the large and growing number of hatcheries (Peterman, 1984b, 1991; Ruggerone et al., 2003; Holt et al., 2008; Peterman et al., 2012). However, ultimately I have been disappointed that discussions about such international coordination have gone nowhere, probably because of arguments made by business interests. Lag times in innovations Lesson 5: New methods also often take a long time to become widely accepted. Be persistent. The paragraph above refers to an initial reluctance to accept unusual research results, but the same time-lag problem exists for novel research methods. Today's young scientists are often frustrated by how slowly their innovative research methods become widely accepted by other scientists, let alone government decision makers or other potential users. Unfortunately, slow adoption is very common ( Gerlotto, 2017). My own informal review of the history of new quantitative methods in environmental science shows that it is quite normal to have at least a 10- to 20-year lag before seeing wide acceptance of new ideas and methods, even among scientists (Peterman, 2015). Two examples demonstrate this point. First, because of extensive debates between field ecologists and modellers in the 1970s and 1980s, it took about 25 years from the inception of ecological simulation modelling in the late 1960s for it to become widely recognized as a legitimate and useful method of analysis outside the small circle of quantitative ecologists. The second example is the method of closed-loop simulations (Walters, 1986) or Management Strategy Evaluations (MSEs) (Punt, 1992; de la Mare, 1998; Butterworth and Punt, 1999), which use stochastic simulation models of entire fishery systems to find the management procedures that give the best performance in the presence of several sources of uncertainty. It took about 15 years before numerous well-developed cases of this method were implemented by management agencies in South Africa, Australia, North America, and Europe. Young scientists should therefore not expect that new methods that they develop will be readily accepted and used. Instead, acceptance will require patience and repeated demonstration that those methods both work and provide new insights. Multi-population analyses One method that started to become known among ecologists in the late 1990s was hierarchical statistical modelling (Myers and Mertz, 1998; Royle and Dorazio, 2008). My group used it on our large salmon database to conduct several multi-population analyses (so-called meta-analyses) to estimate the importance of physical and biological oceanographic processes in driving salmon population dynamics. Hierarchical statistical modelling provides a solid theoretical foundation for such analyses. It uses multiple populations as “statistical replicates”, which helps to average out the effects of random observation errors. Another advantage of hierarchical models applied to multi-population data sets is that they reduce the chance of being misled by one-off spurious correlations between variables that are found in studies of single populations (Myers and Mertz, 1998). The scope of our salmon database and the emergence of hierarchical statistical modelling in the late 1990s enabled us to investigate a widely cited idea at the time, namely that large-scale oceanographic and/or atmospheric processes tended to drive similar time trends in salmon populations across thousands of kilometers of the Pacific Ocean. However, the evidence for that large-scale effect came from studies of salmon catch data, which are only an indirect and inaccurate indicator of salmon abundance or productivity. We instead used our large salmon data set on biological productivity (adult recruits produced per spawner) to conduct spatial analyses and hierarchical modelling. We found that positive correlations among species-specific time series of productivities for pink, chum, and sockeye salmon were strongest between populations where juveniles entered the ocean at locations less than about 500–800 km apart. This result and others showed the importance of shared environmental conditions at that small, regional scale early in ocean life (Peterman, 1982b, 1987; Peterman et al., 1998; Mueter et al., 2002; Pyper et al., 2005), in contrast to the large scale reported by previous analyses that used only catch data. This result also enabled us to screen out numerous potentially spurious environmental factors as explanatory variables when their spatial scale was inconsistent with the small-scale positive correlations in productivity (Mueter et al., 2003). We found that early-summer sea-surface temperature (SST) was a physical variable that had an appropriate regional spatial scale of positive covariation (as opposed to too large or too small). Inclusion of SST in salmon models substantially improved fits to data, whereas the widely used Pacific Decadal Oscillation (PDO) did not (Mueter et al., 2002). Lesson 6: How can progress be made on particularly difficult problems? Be an early adopter of innovative methods of analysis. Hire or collaborate with people who know more than you do and learn from them. Continually add to your toolbox of methods to gain new insights. The key advance in our multi-population analyses described above came from applying hierarchical statistical models, which were brought to my group by Franz Mueter, a post-doctoral fellow. This innovative technique attributed some of the observed temporal variation in productivity of 120 pink, chum, and sockeye salmon populations to variation that is shared across numerous stocks. Compared to analyses of each stock separately, these multi-stock models produced more precise estimates of stock-specific productivity parameters despite the typical noisy data (Mueter et al., 2002; Su et al., 2004). These productivity parameters are important for estimating maximum sustainable harvest rates as well as recovery rates for low-abundance populations. My research group also estimated time trends in productivity of sockeye salmon populations by applying the Ricker stock-recruitment model cast in the form of a Kalman filter. That method was developed in control engineering to extract signals from noisy electronic data. As far as I know, it was first applied to fisheries problems by Walters (1986), but it has been rarely used in fisheries after that. Brian Pyper and Jeff Grout, two researchers in my group, implemented the Kalman filter algorithm and determined via Monte Carlo simulations that it performed well. It successfully extracted underlying time trends in productivity (the “a” parameter in the Ricker spawner-recruit model) from simulated data that contained high- and low-frequency variation (Peterman et al., 2000). We therefore applied the Kalman filter to productivity time series of 64 sockeye salmon populations and found that productivity has decreased over time in the majority of sockeye salmon populations along the coasts of Washington, British Columbia, and southeast Alaska (Peterman and Dorner, 2012). This result fed directly into the highly publicized $25 million Cohen Commission, which was charged with looking into the causes of the decline in abundance of sockeye salmon populations from the Fraser River, British Columbia—Canada's largest producer of that commercially valuable species (Cohen Commission, 2012). Our finding of widespread declines in sockeye productivity across many west-coast populations indicated that the decrease in abundance of Fraser River sockeye salmon was very unlikely to have resulted from mechanisms unique to that river. Instead, marine processes that were shared across sockeye populations, such as survival during the early ocean period, likely had a major influence on Fraser River sockeye salmon. Another researcher in my group, Brigitte Dorner, introduced us to Dynamic Factor Analysis (DFA), a method developed in econometrics and first applied to fisheries by Zuur et al. (2003), but infrequently used in this field since then. We applied DFA to 24 wild Chinook salmon populations in western North America and found a tendency for decreasing productivity over time (Dorner et al., 2018). The spatial scale of positive correlation in productivity among Chinook populations was much larger than the scale for pink, chum, and sockeye salmon, for unknown reasons. The DFA also showed that productivity time trends were most closely associated with the North Pacific Gyre Oscillation (NPGO) (Dorner et al., 2018), which is highly correlated with variables related to food supply for salmon (Di Lorenzo et al., 2008). Another key finding from our spatial analyses of processes affecting salmon productivity is that in the last 15–20 years, temporal patterns in productivity have become increasingly synchronous across populations of most Pacific salmon species (Peterman and Dorner, 2012; Kilduff et al., 2015; Malick and Cox, 2016; Dorner et al., 2018). These results suggest that the relative importance of different drivers of salmon productivity may have changed recently, with some drivers having a stronger and more widespread influence in recent years. This increased synchronicity may have undesirable consequences: (1) reduced resilience benefits of the portfolio effect of diverse life histories and sub-populations (Griffiths et al., 2014), and (2) increased frequency of fishery closures (and hence create greater year-to-year variability in catches) because there will tend to be fewer high-abundance stocks in years when others are at low abundance. Other multi-population studies helped to explain (1) the inverse correlation between time series of abundances of adult sockeye salmon in British Columbia and those in Alaska (Peterman and Wong, 1984), and (2) the inverse relationship between catches of several salmon species in those two regions (Mantua et al., 1997). Specifically, Mueter et al. (2002) found that higher SST during the early ocean life stage of salmon is associated with increased productivity of pink and sockeye salmon in most of Alaska but with reduced productivity of those species in British Columbia. In addition, Malick et al. (2015) found that early spring blooms of marine phytoplankton are significantly correlated with higher productivity of pink salmon in Alaska and lower productivity of pink salmon in British Columbia. Together, these results not only help to explain why an inverse relationship exists between abundances of some salmon species in British Columbia and Alaska, but they also suggest that changes in SST or spring-bloom timing resulting from climate change could cause latitudinal shifts in salmon productivity and abundance. These examples of my research group's work were conducted relatively efficiently by using our large salmon database in conjunction with advanced statistical methods such as hierarchical statistical models, the Kalman filter, and Dynamic Factor Analysis. Thus, there was a great benefit to being open to testing and applying novel methods. Research objective 2: quantify the main sources of uncertainties in data and fishery systems, the risks created by those uncertainties, and their implications for management and conservation “Uncertainty is an uncomfortable position. But certainty is an absurd one”.—Voltaire (1694–1778) “The domain of our ignorance is greater than the domain of our knowledge”. C. S. Holling (1978) Despite the obvious points made in these two quotes, up until at least the 1980s, most sources of uncertainty were not commonly considered in the theories and quantitative analyses of fish population dynamics and their management. Everyone recognized that uncertainties existed, but quantifying them was difficult, and there were challenges to incorporating such uncertainties into analyses, even if they could be quantified. Considerable progress has been made, but there is still room for improvement. Four key sources of uncertainty pervade fisheries systems: (1) natural variability over time and space in both physical and biological processes, (2) observation error (imperfect knowledge resulting from measurement error and sampling error), (3) structural uncertainty owing to incomplete understanding of the ecosystem's complex structure, and (4) outcome uncertainty (differences between management targets and actual outcomes). These uncertainties are important because together, they create ecological risks to fish populations as well as economic and social risks to the fishing industry and people. Natural variability and observation error At a time when I still used only classical hypothesis testing, I became interested in the management implications of natural variability and observation error in the context of statistical power analysis. Specifically, monitoring programs, experiments, and impact assessments should have high statistical power (i.e. a high probability of correctly detecting the existence of some biologically or economically important effect or relationship). Without high power, studies may miss detrimental effects and appropriate management action may not be taken. Moreover, such mistakes could be common given high levels of natural variability or observation error, both of which reduce statistical power. We carried out several statistical power analyses. The first one used a Monte Carlo simulation to evaluate a proposed 3-year experiment of releasing large numbers of hatchery-reared juvenile coho salmon (smolts) in Oregon to estimate the degree of nonlinearity in the relationship between abundances of smolts and the resulting adults. Nonlinearity would mean that hatchery operators and harvesters would catch fewer adults than expected from a linear assumption. We found that the proposed experiment would have low power to detect any important degree of nonlinearity. To obtain statistical power greater than 0.8, a 10- to 20-year experiment would be needed instead, and with a much larger number of juveniles released than proposed (Peterman and Routledge, 1983). In another analysis (Peterman et al., 1987), we built a stochastic model of English sole and compared under various conditions the probabilities that two simulated stock assessment methods (a trawl survey and VPA or cohort analysis) would successfully estimate the true time trend in abundance. We found that under most realistic conditions, statistical power for those methods was low. In a third study, we compared various designs of a large-scale experiment for testing whether the observed temporal decrease in body size in pink salmon was due to size-selective fishing (McAllister et al., 1992). There we used a decision analysis framework to rank alternative designs based on their statistical power and the expected net economic benefits from improved catches (McAllister and Peterman, 1992a). Lesson 7: Scientists must not forget what they already know. The culmination of my work on statistical power was a comment (Peterman, 1990b) and a review paper (Peterman, 1990a) that surprisingly turned out to be my most widely cited journal paper. That recognition was unexpected because in those two papers, I took the not-so-novel step of merely reminding scientists of what they had once learned, but had since tended to overlook, namely that they must consider two types of error in statistical analyses (Table 1). Type I errors arise when some null hypothesis is incorrectly rejected, as for example, by finding a statistically significant harmful effect of an industrial effluent, even though the null hypothesis is true that the effect does not exist. In contrast, a type II error would be made if the effluent actually is harmful, but the data analysis fails to find a statistically significant effect of the effluent. Unfortunately, scientists have traditionally focused on minimizing the well-known probability (α) of making a type I error without considering the probability (β) of making a type II error. Statistical power is the probability of avoiding a type II error (1-β) by correctly rejecting some null hypothesis (Table 1). The main problem has been that most authors who report a failure to reject some null hypothesis go on to assume that the null hypothesis must be true without asking how the natural variability, observation error, and sample size might have affected statistical power. Instead, I argued that the respective probabilities, as well as short- and long-term costs, of both type I and type II errors should be taken into account when making management decisions, not just the probability of making a type I error, as has been done traditionally. Table 1. Four possible categories of results from conducting a classical statistical analysis to test some null hypothesis (in this example, that there is no relationship between two variables). Actual state of nature Result of classical statistical analysis Do not reject null hypothesis Reject null hypothesis Null hypothesis actually true (i.e. there is no relationship) Correct conclusion (1-α) Type I error (α) Null hypothesis actually false (i.e. there is a relationship) Type II error (β) Correct conclusion (1-β), i.e. statistical power Actual state of nature Result of classical statistical analysis Do not reject null hypothesis Reject null hypothesis Null hypothesis actually true (i.e. there is no relationship) Correct conclusion (1-α) Type I error (α) Null hypothesis actually false (i.e. there is a relationship) Type II error (β) Correct conclusion (1-β), i.e. statistical power Table 1. Four possible categories of results from conducting a classical statistical analysis to test some null hypothesis (in this example, that there is no relationship between two variables). Actual state of nature Result of classical statistical analysis Do not reject null hypothesis Reject null hypothesis Null hypothesis actually true (i.e. there is no relationship) Correct conclusion (1-α) Type I error (α) Null hypothesis actually false (i.e. there is a relationship) Type II error (β) Correct conclusion (1-β), i.e. statistical power Actual state of nature Result of classical statistical analysis Do not reject null hypothesis Reject null hypothesis Null hypothesis actually true (i.e. there is no relationship) Correct conclusion (1-α) Type I error (α) Null hypothesis actually false (i.e. there is a relationship) Type II error (β) Correct conclusion (1-β), i.e. statistical power In both of these 1990 papers on statistical power I also suggested that in some cases, the usual burden of proof should be reversed, given the prevalence of low statistical power that had previously been found across studies (Vaughan and van Winkle, 1982; de la Mare, 1984; Gerrodette, 1987). That meant putting the burden of proof on the fishing industry to show with high statistical power that fishing activities do not have a detrimental effect. This would reverse the traditional onus on management agencies, which is to show that some detrimental effect is occurring, before regulations are changed or accepted by the fishing industry. Other scientists made similar arguments, and some features of this concept of reversing the burden of proof were incorporated into the United Nations FAO document on “The Precautionary Approach to Capture Fisheries and Species Introductions” (FAO, 1995), which I discuss later. Bayesian statistical analysis Ironically, soon after writing my review paper on statistical power analysis (Peterman, 1990a), I began to feel uncomfortable about standard statistical hypothesis testing and its inherent limitations, and instead leaned more toward Bayesian statistics because of its advantages (Berger and Berry, 1988; Ellison, 1996; Wade, 2000). I concluded, as had many before me, that what is important is the probability distribution of estimated parameter values, such as the slope or degree of nonlinearity in relationships between variables, not just whether a statistically significant maximum-likelihood relationship exists with α arbitrarily set at 0.05. One advantage of the Bayesian approach is that a probability can be estimated for each of several alternative hypotheses about possible parameter values. Those probabilities are also more directly interpretable by fishery managers. Therefore, the first example of my specific advice to fisheries scientists is to learn Bayesian statistics, given its advantages compared to classical hypothesis testing. We applied Bayesian statistical analysis to a wide range of problems. Among others, they included topics such as (1) quantifying the effect of lake fertilization on productivity of sockeye salmon (Maxwell et al., 2006), (2) hierarchical Bayesian modelling to improve estimates of pink salmon productivity (Su et al., 2004), (3) Bayesian state-space modelling for stock-recruitment data subject to measurement error (Su and Peterman, 2012), and (4) quantifying uncertainties in target spawner abundances for sockeye salmon (Bodtker et al., 2007). All of these cases improved understanding of how natural variability and observation error can affect scientific advice to managers. Monte Carlo simulations I also highly recommend that fisheries scientists add to their toolkits the ability to develop and run Monte Carlo simulations. This simulation method became a staple of my research team. Among other applications, we used it to evaluate the performance (bias and precision) of parameter estimation methods in the presence of natural variability and observation error. For instance, we compared the effectiveness of various methods for taking temporal autocorrelation into account when testing the statistical significance of correlations in time series data (Pyper and Peterman, 1998). This paper is still frequently cited. In another case, we conducted Monte Carlo simulations of the dynamics of nine populations of marine species ranging from mackerel, herring, and menhaden to cod, halibut and sole. We documented a wide range of conditions in which estimates of stock biomass derived by Virtual Population Analysis (VPA) generated spurious time trends owing to time trends in fishing mortality, F, and errors in assumed values of natural mortality, M (Lapointe et al., 1989). We also showed that spurious correlations between abundance of recruits and environmental factors can result from those conditions (Lapointe and Peterman, 1991). State-space models Another type of natural variation that is common in marine fisheries research is non-stationarity, which is a change over time in the mean and/or variance of some state variables or parameters such as productivity. Such changes complicate effective management of fish populations in a dynamic ocean. To address this issue, we again applied the Kalman filter (an example of a state-space model) to our fisheries data. These data contain considerable natural variability in the underlying productivity (signal), as well as low- and high-frequency natural variation and observation error (noise). Our previous simulations showed that a Kalman filter with a random walk system equation can track temporally autocorrelated changes in productivity of Pacific salmon and even regime shifts (step-functions) more effectively than traditional methods for annually updating parameter estimates (Peterman et al., 2000). Such better tracking can help fisheries managers to set appropriate regulations when productivity changes over time (Peterman et al., 2003), as it is likely to do with climate change. Such state-space models are effective in these situations because they estimate how much of the observed variation is attributable to noise and how much to changes in the true underlying signal. State-space models are now being widely used in several environmental fields as a way to deal with non-stationarity (Clark and Bjornstad, 2004). Therefore, every fisheries scientist should be able to apply state-space models. Structural uncertainty Any fisheries system's true underlying structure is not completely understood. It is now widely accepted in fisheries science that to deal with such structural uncertainty, analyses must include multiple plausible models. Therefore, fisheries scientists should routinely determine how their results are affected by different assumptions about structures of underlying models. Many scientists have appropriately changed their focus from finding the single best model to (1) using delta AICc to identify a set of plausible models consistent with the data, or (2) finding management procedures that are most robust to uncertainties about the real-world model (e.g. harvest control rules that produce acceptably high values of performance indicators across a wide range of models and uncertainties). The latter approach uses Monte Carlo simulation models of entire fishery systems, including simulating the natural system dynamics for each of a plausible set of underlying models and simulating the steps of data collection with observation errors, stock assessment analyses, setting of regulations by fisheries managers, and responses to regulations by harvesters. Such “whole-system models”, as I refer to them here, have been variously labelled as closed-loop simulations (Walters, 1986), Management Strategy Evaluations (MSE) (Sainsbury et al., 2000), or Management Procedure Evaluations (MPE) (Butterworth and Punt, 1999). This approach of whole-system simulation is now considered the gold standard in marine fisheries for conducting stock assessments and providing advice for risk-management decisions. We applied whole-system modelling to Pacific salmon in several cases, only three of which I mention here. One example was the paper described above in which we evaluated the relative performance of a Kalman filter at tracking changes in productivity (Peterman et al., 2000). In another example, we simulated a risk-assessment framework that quantified tradeoffs and found that management targets that changed with time-varying estimates of fish productivity resulted in higher catches of chum salmon and lower conservation risk than standard management practices (Collie et al., 2012). The most comprehensive whole-system model that we developed (our CLIM2 model of 15 sockeye salmon populations) included not only seven spatial and temporal patterns of changes in productivity, but also seven alternative stock assessment models ranging from simple to highly complex (Dorner et al., 2009). The complex stock assessment methods performed better in some cases, but with a caveat explained below under “Outcome uncertainty”. Another useful method for dealing with structural uncertainty is formal, quantitative decision analysis, which came from business and was again pioneered in fisheries by Walters (1977, 1986). In several case studies, we used decision analysis to evaluate management options in the presence of uncertainties. For example, we applied Bayesian decision analysis to evaluate management options to (1) rebuild a depleted sockeye salmon population (Pestes et al., 2008), (2) determine the timing of fishery openings (Robb and Peterman, 1998), and (3) enhance chum salmon populations (MacGregor et al., 2002). We also quantified safety margins that fisheries managers of Atlantic menhaden and Arcto-Norwegian cod should use when setting harvest goals in the presence of large uncertainties in current stock abundance and stock-recruitment parameters (Frederick and Peterman, 1995). Such decision analyses can not only provide information directly to managers, but they can also help rank future research priorities based on the uncertain factors that most affect the rank order of alternative management actions. The whole-system models described in a previous section are also essentially decision analyses in which management options are ranked based on their performance across a wide range of conditions and uncertainties. Outcome uncertainty This final type of uncertainty, “outcome uncertainty”, refers to the inevitable differences that arise between management targets and actual outcomes in the realized annual spawning stock or harvest rate, for example. Although such differences are commonplace (Bocking and Peterman, 1988; Rosenberg and Brault, 1993), surprisingly they have rarely been quantified. They arise from any combination of at least three factors, (1) unexpected changes in fish behaviour that make fish more or less available to fishing gear, (2) errors in estimates of fish abundance prior to setting regulations, or (3) noncompliance with regulations by fishing vessels (the latter two together often called implementation uncertainty). Up until roughly the late 1990s, most fish stock assessments ignored outcome uncertainty, even though it could sometimes be large. My research group analysed harvest rate data on sockeye salmon populations and found that actual realized harvest rates tended to be higher than targets when abundance was low (creating a conservation concern) and lower than targets when abundance was high (leading to foregone catch) (Holt and Peterman, 2006). More importantly, the comprehensive whole-system modelling of sockeye salmon fisheries described above showed that the effects of outcome uncertainty masked or swamped the incremental benefits of the more complex stock assessment models (Dorner et al., 2009). Reduction in outcome uncertainty should therefore be a priority for fisheries managers. My advice to fisheries scientists is that outcome uncertainty should be routinely included in fishery models, not only to obtain better estimates of effects of simulated management regulations, but also to determine whether the effects of outcome uncertainties may overwhelm the expected benefits of some “improvement” in a stock assessment model. Research objective 3: develop methods to reduce some of the uncertainties and risks My third research theme focused on ways to either reduce the magnitude of uncertainties or decrease the potential risks associated with whatever uncertainties existed. I already mentioned how good experimental design can reduce uncertainties about processes of population dynamics (McAllister and Peterman, 1992a, b), and how Kalman filters and hierarchical statistical models can produce more precise parameter estimates than would otherwise be possible. Lesson 8: Widely used methods can only push the frontier of knowledge so far. Methods from quite distant disciplines may have unexpected benefits. One key challenge in reducing the risks that result from inappropriate management decisions is how to best present complex quantitative information to fisheries managers to help them make well-informed decisions in the presence of large uncertainties. I worked with three experts in ergonomics and computer visualization to develop a new software package called Vismon, which visually presents trade-offs, along with their uncertainties, in an interactive and dynamic manner (Booshehrian et al., 2012). Vismon allows managers, scientists, and users of fish to visually compare complex outcomes from various management options, thereby reducing the chance of choosing inappropriate actions. Vismon has been used for evaluating management options in some salmon fisheries in western Alaska. This software and written paper received the “Second-best paper” award (a sachertorte!) in Vienna at the 2012 “Eurographics Conference on Visualization”—Europe's top-ranked conference in computer visualization, demonstrating the value of innovation and collaboration. More recently, we looked at the methods for classifying species into various levels of concern for extinction risk. Those methods have been standardized for many decades in the IUCN Red List categories and criteria (IUCN, 2013). One common way of estimating extinction risk is based on the direction and rate of change in abundance over some period. The widely used IUCN criterion A, for example, only examines abundance estimates over the last 10 years or three generations. However, the typical uncertainties of natural variation over time in productivity, as well as observation error, can create errors in those estimates. Such errors could lead to either failing to correctly identify a conservation concern (false negative) or incorrectly classifying some situation as being a conservation concern (false positive or false alarm). The former would increase the chance of some detrimental outcome occurring, and the latter would divert the usually limited funds for conservation action to a situation that did not need it. Only a few studies have quantitatively evaluated the accuracy of IUCN decline criteria. Notable examples are Punt (2000), Mace et al. (2002), Dulvy et al. (2005), Rice and Legacé (2007), Holt et al. (2009), Wilson et al. (2011), and Regan et al. (2013). We extended those previous analyses by conducting both empirical evaluations and Monte Carlo simulations to estimate the reliability of a wide variety of criteria for classifying population status. In two of those studies (Porszt et al., 2012; d’Eon-Eggertson et al., 2015), we applied receiver operating characteristic (ROC) curves, which are used in medicine to estimate the reliability of diagnostic tests for diseases, as well as in quality-control engineering. We found that the often-used IUCN extinction-risk criterion of a given percentage decline in the last 10 years or three generations often does not perform as well as other criteria, and that the relative ranking of different types of indicators of extinction risk is influenced in quantifiable ways by factors such as the degree of autocorrelation in natural variability, as well as the relative magnitude of that variability and observation error (Connors et al., 2014). Unfortunately, to my knowledge, these results have not yet influenced the long-standing IUCN methods for classifying extinction risks based on time trends in abundance, despite the presentation of our results to IUCN people. Perhaps changes in these risk-assessment practices will require more evidence and new practitioners. I have one final comment about uncertainties. Several years ago I was at a conference, and a fisheries economist pointed out how unclear fish ecologists were in their understanding of fisheries systems, as indicated by our frequent discussion of uncertainties. Yet that same person had the audacity to ignore all uncertainties in the economic system and present numbers for how many jobs and economic value would be lost under certain management regulations! Needless to say, I gave some negative feedback on that lack of logic! Sabbaticals and professional committees Lesson 9: Take advantage of sabbaticals or other forms of outside involvement to diversify your knowledge and keep up with developments in your field. I was also fortunate to be able to learn about novel subjects and interact with new people through sabbaticals from SFU. I will elaborate on two examples. During my first sabbatical in 1985–86, I proposed to compare the dynamics of several populations of clupeids (herring, sardines, and anchovies) in different oceanographic domains while spending a year at the US National Marine Fisheries Service (NMFS) lab in La Jolla, California. The anadromous and semelparous life histories of Pacific salmon species, as well as their associated stock assessment methods, are quite different from those of non-anadromous fish species, so I wanted to expand my experience by conducting research on non-salmonids for that year. I also wanted to learn more about oceanography. The NMFS lab was ideal for the latter purpose because it was located next to the Scripps Institute of Oceanography, where large numbers of physical and biological oceanographers worked. Shortly after I arrived in La Jolla, the division head, Reuben Lasker, asked me whether I would like to change my plans and focus instead on understanding the link between ocean dynamics and recruitment variability in northern anchovy (Engraulis mordax). I would be able to use the long-term data set on eggs, larvae, and adults that had been collected by the unusually intensive sampling program of the California Cooperative Fisheries Investigation (CALCOFI). I jumped at this chance and tested Lasker's (1975, 1981) “stable ocean hypothesis”, which stated that “… the upper mixed layer of the ocean must be in a stable (nonturbulent) state” to generate sufficient concentration of food to ensure good survival of first-feeding young larval anchovy. Our resulting paper in Science reported strong support for one-half of Lasker's hypothesis—a significant inverse relation between mortality rate of larval anchovy and the number of stable-ocean periods during the spawning season (Peterman and Bradford, 1987). However, abundance of the surviving northern anchovy larvae was not significantly correlated with abundance of subsequent 1-year-old recruits (Peterman et al., 1988). Thus, we found that good survival to an early larval stage was a necessary but not sufficient condition for good recruitment to a later age. We therefore did not find support for the dominant hypotheses at the time of Hjort and Lasker that the abundance of recruits in marine fishes is largely determined at an early larval life stage. During my next sabbatical, I set out to learn about Bayesian statistics at the University of Washington in Seattle. I spent considerable time in the Mathematics and Statistics library and also learned quite a bit from Andre Punt, Geof Givens, and Ray Hilborn. I focused on developing a simulation model for Pacific salmon that incorporated annual Bayesian updating of estimates of productivity parameters in response to different types and magnitudes of oceanographic regime shifts. The model assumed that management actions would respond to changing estimates of simulated productivity. That work eventually led to the Peterman et al. (2000) paper, which also documented the merits of a Kalman filter estimation scheme. Another way to learn is to serve on professional committees. I was fortunate to participate in many independent expert panels that looked into fisheries problems and made recommendations to management agencies. Those experiences helped broaden my perspective and not only contributed indirectly to my research, but also enriched my teaching by providing examples of real-world problems for students. I won't discuss more about my teaching career here, but suffice it to say that it was very rewarding. Influencing policy Lesson 10: Accept opportunities to help develop management policies and guidelines. I had an opportunity to incorporate into management policy some of my research experience with uncertainties when I was one of 34 scientists invited to Sweden in 1995 to help write the UN Food and Agriculture Organization's (FAO's) guidelines for the Precautionary Approach to Fisheries. That very interesting week of collaboration with some of the world's best fisheries scientists produced an influential document (FAO, 1995). Not only were some elements of these guidelines incorporated into Annex II of the 1995 United Nations Agreement on Conservation and Management of Straddling Fish Stocks and Highly Migratory Fish Stocks, but they were rapidly adopted by numerous fisheries management agencies around the world (Garcia, 2000). Precautionary approaches to human activities in fisheries, which are more biologically conservative because of uncertainties, are now commonplace. As a result, Andre Punt, a very well respected scientist, subsequently called this meeting “… one of the most important fisheries meetings of the 20th century” (Punt, 2008). Despite the importance of this FAO document, I am frequently disappointed to find that most people are only familiar with a few oft-repeated quotes from its introductory pages, even though there are extensive substantive guidelines in the rest of the 52-page document. I recognize the difficulty of keeping up with the rapidly growing literature, but people who cite documents should read them thoroughly. Conclusion Summary Lesson 11: Pursue a path of continuous learning throughout your career. My overarching theme of continuous learning should now be clear. Actively develop and apply new approaches to your research, and make a habit of being an early adopter of novel ideas and methods of analysis. To stimulate innovation, look outside of your own discipline, as we did when we drew upon methods from engineering, econometrics, medical diagnostics, and business. All of this should be done throughout every stage of your career, even if it requires substantial mental effort and takes you out of your comfort zone. Such an approach will not only help speed progress in understanding, but it will very likely enhance your personal career, regardless of whether you are a young or well-established scientist. Acknowledgements I thank Howard Browman for the invitation to write this essay and for providing very useful suggestions, as did two anonymous reviewers. Brigitte Dorner and Judith Anderson provided valuable comments on a draft manuscript. I am also grateful to the numerous, unnamed people who gathered the long time series of data on Pacific salmon and other fish species that permitted my research group to address questions about dynamic processes. I particularly thank Brian Pyper and Mike Lapointe for their years of diligent quality-control work on our salmon database. I also thank the 70 graduate students, research assistants, and post-doctoral fellows and two undergraduate co-op students for whom I technically was the supervisor, but from whom I learned a great deal as well. I regret that the limited space here did not permit mentioning all of your publications, let alone your names. Thanks also to students in my courses, especially “Risk Assessment and Decision Analysis in Management of Natural Resources”, who forced clarifications of my own thinking. Finally, all of this was made possible by generous funding from the Canada Research Chairs Program, Canada Foundation for Innovation, Natural Sciences and Engineering Research Council of Canada, Gordon and Betty Moore Foundation, Canada's Department of Fisheries and Oceans, and Simon Fraser University. Footnotes Food for Thought articles are essays in which the author provides their perspective on a research area, topic, or issue. They are intended to provide contributors with a forum through which to air their own views and experiences, with few of the constraints that govern standard research articles. This Food for Thought article is one in a series solicited from leading figures in the fisheries and aquatic sciences community. The objective is to offer lessons and insights from their careers in an accessible and pedagogical form from which the community, and particularly early career scientists, will benefit. The International Council for the Exploration of the Sea (ICES) and Oxford University Press are pleased to make these Food for Thought articles immediately available as free access documents. References Batten S. D. , Ruggerone G. T. , Ortiz I. 2018 . Pink salmon induce a trophic cascade in plankton populations in the southern Bering Sea and around the Aleutian Islands . In Fisheries Oceanography , in press. https://doi.org/10.1111/fog.12276 Berger J. O. , Berry D. A. 1988 . Statistical analysis and the illusion of objectivity . American Scientist , 76 : 159 – 165 . Bocking R. C. , Peterman R. M. 1988 . Preseason forecasts of sockeye salmon (Oncorhynchus nerka): comparison of methods and economic considerations . Canadian Journal of Fisheries and Aquatic Sciences , 45 : 1346 – 1354 . Google Scholar Crossref Search ADS Bodtker K. M. , Peterman R. M. , Bradford M. J. 2007 . Accounting for uncertainty in estimates of escapement goals for Fraser River sockeye salmon based on productivity of nursery lakes in British Columbia, Canada . North American Journal of Fisheries Management , 27 : 286 – 302 . Google Scholar Crossref Search ADS Booshehrian M. , Möller T. , Peterman R. M. , Munzner T. 2012 . Vismon: facilitating analysis of trade-offs, uncertainty, and sensitivity in fisheries management decision making . Computer Graphics Forum , 31 : 1235 – 1244 . Google Scholar Crossref Search ADS Butterworth D. S. , Punt A. E. 1999 . Experiences in the evaluation and implementation of management procedures . ICES Journal of Marine of Science , 56 : 985 – 998 . Google Scholar Crossref Search ADS Clark J. S. , Bjornstad O. N. 2004 . Population time series: process variability, observation errors, missing values, lags, and hidden states . Ecology , 85 : 3140 – 3150 . Google Scholar Crossref Search ADS Cohen Commission of Inquiry into the Decline of Sockeye Salmon in the Fraser River—Final Report. 2012 . Vancouver, BC, 1 – 3 . http://publications.gc.ca/site/eng/432516/publication.html (last accessed 3 July 2018). Collie J. S. , Peterman R. M. , Zuehlke B. M. 2012 . A fisheries risk-assessment framework to evaluate trade-offs among management options in the presence of time-varying productivity . Canadian Journal of Fisheries and Aquatic Sciences , 69 : 209 – 223 . Google Scholar Crossref Search ADS Connors B. M. , Cooper A. B. , Peterman R. M. , Dulvy N. K. 2014 . The false classification of extinction risk in noisy environments . Proceedings of the Royal Society, Series B , 281 : 20132935. Google Scholar Crossref Search ADS de la Mare W. K. 1984 . On the power of catch per unit effort series to detect declines in whale stocks . Report of the International Whaling Commission , 34 : 655 – 662 . de la Mare W. K. 1998 . Tidier fisheries management requires a new MOP (management-oriented paradigm) . Reviews in Fish Biology and Fisheries , 8 : 349 – 356 . Google Scholar Crossref Search ADS d’Eon-Eggertson F. , Dulvy N. K. , Peterman R. M. 2015 . Reliable identification of declining populations in an uncertain world . Conservation Letters , 8 : 86 – 96 . Google Scholar Crossref Search ADS Di Lorenzo E. , Schneider N. , Cobb K. M. , Franks P. J. S. , Chhak K. , Miller A. J. , McWilliams J. C. et al. 2008 . North Pacific Gyre Oscillation links ocean climate and eco- system change . Geophysical Research Letters , 35 : 1 – 6 . Google Scholar Crossref Search ADS Dorner B. , Peterman R. M. , Su Z. 2009 . Evaluation of performance of alternative models of Pacific salmon in the presence of climatic change and outcome uncertainty using Monte Carlo simulations . Canadian Journal of Fisheries and Aquatic Sciences , 66 : 2199 – 2221 . Google Scholar Crossref Search ADS Dorner B. , Catalano M. , Peterman R. M. 2018 . Spatial and temporal patterns of covariation in productivity of Chinook salmon populations of the Northeastern Pacific . Canadian Journal of Fisheries and Aquatic Sciences , 75 : 1082 – 1095 . Google Scholar Crossref Search ADS Dulvy N. K. , Jennings S. , Goodwin N. B. , Grant A. , Reynolds J. D. 2005 . Comparison of threat and exploitation status in north-east Atlantic marine populations . Journal of Applied Ecology , 42 : 883 – 891 . Google Scholar Crossref Search ADS Ellison A. M. 1996 . An introduction to Bayesian inference for ecological research and environmental decision-making . Ecological Applications , 6 : 1036 – 1046 . Google Scholar Crossref Search ADS FAO (Food and Agriculture Organization of the United Nations) . 1995 . Precautionary approach to fisheries. Part 1. Guidelines on the precautionary approach to capture fisheries and species introductions. FAO Fisheries Technical Paper, 350/1. 52 pp. [Also reprinted in 1996 with the same title, but in the series FAO Technical Guidelines for Responsible Fisheries, 2. 60 pp.] Frederick S. W. , Peterman R. M. 1995 . Choosing fisheries harvest policies: when does uncertainty matter? Canadian Journal of Fisheries and Aquatic Sciences , 52 : 291 – 306 . Google Scholar Crossref Search ADS Garcia S. M. 2000 . The precautionary approach to fisheries: progress review and main issues. In Current Fisheries Issues and the Food and Agriculture Organization of the United Nations , pp. 479 – 560 . Ed. by Nordquist M. H. , Moore J. N. . Martinus Nijhoff Publishers , The Hague, The Netherlands . Gerlotto F. 2017 . Sixteen lessons from a 40-year quest to understand the mysterious life of the grey triggerfish . ICES Journal of Marine Science , 74 : 2321 – 2332 . Google Scholar Crossref Search ADS Gerrodette T. 1987 . A power analysis for detecting trends . Ecology , 68 : 1364 – 1372 . Google Scholar Crossref Search ADS Griffiths J. R. , Schindler D. E. , Armstrong J. B. , Scheuerell M. D. , Whited D. C. , Clark R. A. , Hilborn R. et al. 2014 . Performance of salmon fishery portfolios across western North America . Journal of Applied Ecology , 51 : 1554 – 1563 . Google Scholar Crossref Search ADS PubMed Hilborn R. , Walters C. J. , Peterman R. M. , Staley M. J. 1984 . Models and fisheries: a case study in implementation . North American Journal of Fisheries Management , 4 : 9 – 14 . Google Scholar Crossref Search ADS Holling C. S. 1959 . The components of predation as revealed by a study of small-mammal predation of the European Pine Sawfly . Canadian Entomologist , 91 : 293 – 320 . ( 5 ): Google Scholar Crossref Search ADS Holling C. S. 1965 . The functional response of predators to prey density and its role in mimicry and population regulation . Memoirs of the Entomological Society of Canada , 97 : 5 – 60 . Google Scholar Crossref Search ADS Holling C. S. 1966 . The functional response of invertebrate predators to prey density . Memoirs of the Entomological Society of Canada , 98 : 5 – 85 . Google Scholar Crossref Search ADS Holling C. S. , (ed.), Bazykin A. , Bunnell P. , Clark W. C. , Gallopin G. C. , Gross J. , Hilborn R. et al. 1978 . Adaptive Environmental Assessment and Management . John Wiley & Sons , Chichester, England . 377 pp. Holt C. A. , Peterman R. M. 2006 . Missing the target: uncertainties in achieving management goals in fisheries on Fraser River, British Columbia, sockeye salmon . Canadian Journal of Fisheries and Aquatic Sciences , 63 : 2722 – 2733 . Google Scholar Crossref Search ADS Holt C. A. , Rutherford M. B. , Peterman R. M. 2008 . International cooperation among nation-states of the North Pacific Ocean on the problem of competition among salmon for a common pool of prey resources . Marine Policy , 32 : 607 – 617 . Google Scholar Crossref Search ADS Holt C. A. , Cass A. , Holtby B. , Riddell B. 2009 . Indicators of Status and Benchmarks for Conservation Units in Canada’s Wild Salmon Policy. Canada Department of Fisheries and Oceans . Canadian Science Advisory Secretariat Research Document , 2009/058: viii + 74 pp. IUCN (International Union for Conservation of Nature) Standards and Petitions Subcommittee . 2013 . Guidelines for using the IUCN Red List Categories and Criteria. Version 10.1 (September 2013). IUCN, Gland, Switzerland. http://www.iucnredlist.org/documents/RedListGuidelines.pdf (last accessed November 2013). Kaeriyama M. 1998 . Dynamics of chum salmon, Oncorhynchus keta, populations released from Hokkaido, Japan . North Pacific Anadromous Fish Commission (NPAFC) Bulletin , 1 : 90 – 102 . Kilduff D. P. , Di Lorenzo E. , Botsford L. W. , Teo S. L. 2015 . Changing central Pacific El Niños reduce stability of North American salmon survival rates . Proceedings of the National Academy of Sciences of the United States of America , 112 : 10962 – 10966 . Google Scholar Crossref Search ADS PubMed Lapointe M. F. , Peterman R. M. , MacCall A. D. 1989 . Trends in fishing mortality rate along with errors in natural mortality rate can cause spurious time trends in fish stock abundances estimated by virtual population analysis (VPA) . Canadian Journal of Fisheries and Aquatic Sciences , 46 : 2129 – 2139 . Google Scholar Crossref Search ADS Lapointe M. F. , Peterman R. M. 1991 . Spurious correlations between fish recruitment and environmental factors due to errors in the natural mortality rate used in virtual population analysis (VPA) . ICES Journal of Marine Science , 48 : 219 – 228 . Google Scholar Crossref Search ADS Lasker R. 1975 . Field criteria for survival of anchovy larvae: the relation between inshore chlorophyll maximum layers and successful first feeding . Fisheries Bulletin (United States) , 73 : 453 – 462 . Lasker R. 1981 . Factors contributing to variable recruitment of the northern anchovy (Engraulis mordax) in the California Current: contrasting years, 1975 through 1978 . Rapports et Proces-Verbaux des Reunion–Conseil International pour l'exploration de la Mer , 178 : 375 – 388 . Mace P. M. , Bruckner A. W. , Daves N. K. , Field J. D. , Hunter J. R. , Kohler N. E. , Kope R. G. et al. 2002 . National Marine Fisheries Service (NMFS)/Interagency Working Group evaluation of CITES criteria and guidelines. Technical Memorandum NMFS-F/SPO-58 . National Oceanic and Atmospheric Administration , Rockville, Maryland . MacGregor B. W. , Peterman R. M. , Pyper B. J. , Bradford M. J. 2002 . A decision analysis framework for comparing experimental designs of projects to enhance Pacific salmon . North American Journal of Fisheries Management , 22 : 509 – 527 . Google Scholar Crossref Search ADS Malick M. J. , Cox S. P. , Mueter F. J. , Peterman R. M. 2015 . Linking phytoplankton phenology to salmon productivity along a north-south gradient in the Northeast Pacific Ocean . Canadian Journal of Fisheries and Aquatic Sciences , 72 : 697 – 708 . Google Scholar Crossref Search ADS Malick M. J. , Cox S. P. 2016 . Regional-scale declines in productivity of pink and chum salmon stocks in western North America . PLoS One , 11 : e0146009. Google Scholar Crossref Search ADS PubMed Mantua N. J. , Hare S. R. , Zhang Y. , Wallace J. M. , Francis R. C. 1997 . A Pacific interdecadal climate oscillation with impacts on salmon production . Bulletin of the American Meteorological Society , 78 : 1069 – 1079 . Google Scholar Crossref Search ADS Maxwell M. R. , Peterman R. M. , Bradford M. J. , MacIsaac E. A. 2006 . A Bayesian analysis of biological uncertainty for a whole-lake fertilization experiment for sockeye salmon in Chilko Lake, British Columbia, and implications for the benefit-cost ratio . North American Journal of Fisheries Management , 26 : 418 – 430 . Google Scholar Crossref Search ADS McAllister M. K. , Peterman R. M. 1992a . Decision analysis of a large-scale fishing experiment designed to test for a genetic effect of size-selective fishing on British Columbia pink salmon (Oncorhynchus gorbuscha) . Canadian Journal of Fisheries and Aquatic Sciences , 49 : 1305 – 1314 . Google Scholar Crossref Search ADS McAllister M. K. , Peterman R. M. 1992b . Experimental design in the management of fisheries: a review . North American Journal of Fisheries Management , 12 : 1 – 18 . Google Scholar Crossref Search ADS McAllister M. K. , Peterman R. M. , Gillis D. M. 1992 . Statistical evaluation of a large-scale fishing experiments designed to test for a genetic effect of size-selective fishing on British Columbia pink salmon (Oncorhynchus gorbuscha) . Canadian Journal of Fisheries and Aquatic Sciences , 49 : 1294 – 1304 . Google Scholar Crossref Search ADS Mueter F. J. , Ware D. M. , Peterman R. M. 2002 . Spatial correlation patterns in coastal environmental variables and survival rates of salmon in the north-east Pacific Ocean . Fisheries Oceanography , 11 : 205 – 218 . Google Scholar Crossref Search ADS Mueter F. J. , Peterman R. M. , Pyper B. J. 2003 . Opposite effects of ocean temperature on survival rates of 120 stocks of Pacific salmon (Oncorhynchus spp.) in northern and southern areas . Canadian Journal of Fisheries and Aquatic Sciences , 59 : 456 – 463 , plus the corrigendum printed in Canadian Journal of Fisheries and Aquatic Sciences, 60: 757. Google Scholar Crossref Search ADS Myers R. A. , Mertz G. 1998 . Reducing uncertainty in the biological basis of fisheries management by meta-analysis of data from many populations: a synthesis . Fisheries Research , 37 : 51 – 60 . Google Scholar Crossref Search ADS Pestes L. R. , Peterman R. M. , Bradford M. J. , Wood C. C. 2008 . Bayesian decision analysis for evaluating management options to promote recovery of a depleted salmon population . Conservation Biology , 22 : 351 – 361 . Google Scholar Crossref Search ADS PubMed Peterman R. M. 1978 . Testing for density dependent marine survival in Pacific salmonids . Journal of the Fisheries Research Board of Canada , 35 : 1434 – 1450 . Google Scholar Crossref Search ADS Peterman R. M. 1982a . Nonlinear relation between smolts and adults in Babine Lake sockeye salmon (Oncorhynchus nerka) and implications for other salmon populations . Canadian Journal of Fisheries and Aquatic Sciences , 39 : 904 – 913 . Google Scholar Crossref Search ADS Peterman R. M. 1982b . Model of salmon age structure and its use in pre-season forecasting and studies of marine survival . Canadian Journal of Fisheries and Aquatic Sciences , 39 : 1444 – 1452 . Google Scholar Crossref Search ADS Peterman R. M. , Routledge R. D. 1983 . Experimental management of Oregon coho salmon (Oncorhynchus kisutch): designing for yield of information . Canadian Journal of Fisheries and Aquatic Sciences , 40 : 1212 – 1223 . Google Scholar Crossref Search ADS Peterman R. M. 1984a . Density-dependent growth in early ocean life of sockeye salmon (Oncorhynchus nerka) . Canadian Journal of Fisheries and Aquatic Sciences , 41 : 1825 – 1829 . Google Scholar Crossref Search ADS Peterman R. M. 1984b . Effects of Gulf of Alaska sockeye salmon (Oncorhynchus nerka) abundance on survival, body size, growth rate, and age at maturity of British Columbia and Bristol Bay, Alaska sockeye populations . Canadian Technical Report of Fisheries and Aquatic Sciences , 1302. iv + 87 pp. Peterman R. M. , Wong F. Y. C. 1984 . Cross correlations between reconstructed ocean abundances of Bristol Bay and British Columbia sockeye salmon (Oncorhynchus nerka) . Canadian Journal of Fisheries and Aquatic Sciences , 41 : 1814 – 1824 . Google Scholar Crossref Search ADS Peterman R. M. 1987 . Review of the components of recruitment of Pacific salmon. In Common Strategies of Anadromous and Catadromous Fishes , 1 , pp. 417 – 429 . Ed. by Dadswell M. , Klauda R. , Moffitt C. , Saunders R. , Rulifson R. , Cooper J. E. , American Fisheries Society Symposium , Bethesda, Maryland, USA. Peterman R. M. 1991 . Density-dependent marine processes in North Pacific salmonids: Lessons for experimental design of large-scale manipulations of fish stocks . ICES Marine Science Symposium , 192 : 69 – 77 . Peterman R. M. , Bradford M. J. 1987 . Wind speed and mortality rate of a marine fish, the northern anchovy (Engraulis mordax) . Science , 235 : 354 – 356 . Google Scholar Crossref Search ADS PubMed Peterman R. M. , Bradford M. J. , Kruse G. H. 1987 . Simulation model of English sole (Parophrys vetulus) population dynamics in Washington and Oregon coastal waters . Canadian Journal of Fisheries and Aquatic Sciences , 44 : 1870 – 1878 . Google Scholar Crossref Search ADS Peterman R. M. , Bradford M. J. , Lo N. C. H. , Methot R. D. 1988 . Contribution of early life stages to interannual variability in recruitment of northern anchovy, Engraulis mordax . Canadian Journal of Fisheries and Aquatic Sciences , 45 : 8 – 16 . Google Scholar Crossref Search ADS Peterman R. M. 1990a . Statistical power analysis can improve fisheries research and management . Canadian Journal of Fisheries and Aquatic Sciences , 47 : 2 – 15 . Google Scholar Crossref Search ADS Peterman R. M. 1990b . The importance of reporting statistical power: the forest decline and acidic deposition example . Ecology , 71 : 2024 – 2027 . Google Scholar Crossref Search ADS Peterman R. M. , Pyper B. J. , Lapointe M. F. , Adkison M. D. , Walters C. J. 1998 . Patterns of covariation in survival rates of British Columbian and Alaskan sockeye salmon (Oncorhynchus nerka) stocks . Canadian Journal of Fisheries and Aquatic Sciences , 55 : 2503 – 2517 . Google Scholar Crossref Search ADS Peterman R. M. , Pyper B. J. , Grout J. A. 2000 . Comparison of parameter estimation methods for detecting climate-induced changes in productivity of Pacific salmon (Oncorhynchus spp.) . Canadian Journal of Fisheries and Aquatic Sciences , 57 : 181 – 191 . Google Scholar Crossref Search ADS Peterman R. M. , Pyper B. J. , MacGregor B. 2003 . Use of the Kalman filter to reconstruct historical trends in productivity of Bristol Bay sockeye salmon (Oncorhynchus nerka) . Canadian Journal of Fisheries and Aquatic Sciences , 60 : 809 – 824 . Google Scholar Crossref Search ADS Peterman R. M. , Dorner B. 2012 . A widespread decrease in productivity of sockeye salmon populations in western North America . Canadian Journal of Fisheries and Aquatic Sciences , 69 : 1255 – 1260 . Google Scholar Crossref Search ADS Peterman R. M. , Holt C. A. , Rutherford M. R. 2012 . The need for international cooperation to reduce competition among salmon for a common Pool of prey resources in the North Pacific Ocean . North Pacific Anadromous Fish Commission Technical Report , 8 . 99 – 101 pp. Peterman R. M. 2015 . Consideration of uncertainties in environmental science and management with examples from Pacific salmon . PeerJ PrePrints , 3 : e1955 . Porszt E. J. , Peterman R. M. , Dulvy N. K. , Cooper A. B. , Irvine J. R. 2012 . Reliability of indicators of decline in abundance . Conservation Biology , 26 : 894 – 904 . Google Scholar Crossref Search ADS PubMed Punt A. E. 1992 . Selecting management methodologies for marine resources, with an illustration for southern African hake . South African Journal of Marine Science , 12 : 943 – 958 . Google Scholar Crossref Search ADS Punt A. E. 2000 . Extinction of marine renewable resources: a demographic analysis . Population Ecology , 42 : 19 – 27 . Google Scholar Crossref Search ADS Punt A. E. 2008 . The FAO precautionary approach after almost 10 years: have we progressed towards implementing simulation-tested feedback-control management systems for fisheries management? Natural Resource Modeling , 19 : 441 – 464 . Google Scholar Crossref Search ADS Pyper B. J. , Peterman R. M. 1998 . Comparison of methods to account for autocorrelation in correlation analyses of fish data . Canadian Journal of Fisheries and Aquatic Sciences , 55 : 2710. plus the erratum printed in Canadian Journal of Fisheries and Aquatic Sciences, 55: 2710. Google Scholar Crossref Search ADS Pyper B. J. , Peterman R. M. 1999 . Relationship among adult body length, abundance, and ocean temperature for British Columbia and Alaska sockeye salmon, 1967-1997 . Canadian Journal of Fisheries and Aquatic Sciences , 56 : 1716 – 1720 . Pyper B. J. , Mueter F. J. , Peterman R. M. 2005 . Across-species comparisons of spatial scales of environmental effects on survival rates of Northeast Pacific salmon . Transactions of the American Fisheries Society , 134 : 86 – 104 . Google Scholar Crossref Search ADS Regan T. J. , Taylor B. L. , Thompson G. G. , Cochrane J. F. , Ralls K. , Runge M. C. , Merrick R. 2013 . Testing decision rules for categorizing species’ extinction risk to help develop quantitative listing criteria for the U.S. Endangered Species Act . Conservation Biology , 27 : 821 – 831 . Google Scholar Crossref Search ADS PubMed Rice J. C. , Legacé E. 2007 . When control rules collide: a comparison of fisheries management reference points and International Union for Conservation of Nature (IUCN) criteria for assessing risk of extinction . ICES Journal of Marine Science , 64 : 718 – 722 . Google Scholar Crossref Search ADS Robb C. A. , Peterman R. M. 1998 . Application of Bayesian decision analysis to the management of a sockeye salmon fishery . Canadian Journal of Fisheries and Aquatic Sciences , 55 : 86 – 98 . Google Scholar Crossref Search ADS Rosenberg A. A. , Brault S. 1993 . Choosing a management strategy for stock rebuilding when control is uncertain . Canadian Special Publication of Fisheries and Aquatic Sciences , 120 : 243 – 249 . Royle J. A. , Dorazio R. M. 2008 . Hierarchical Modeling and Inference in Ecology: The Analysis of Data from Populations, Metapopulations and Communities . Academic Press , San Diego, CA . 444 pp. Ruggerone G. T. , Zimmermann M. , Myers K. W. , Nielsen J. L. , Rogers D. E. 2003 . Competition between Asian pink salmon (Oncorhynchus gorbuscha) and Alaskan sockeye salmon (O. nerka) in the North Pacific Ocean . Fisheries Oceanography , 12 : 209 – 219 . Google Scholar Crossref Search ADS Ruggerone G. T. , Connors B. M. 2015 . Productivity and life history of sockeye salmon in relation to competition with pink and sockeye salmon in the North Pacific Ocean . Canadian Journal of Fisheries and Aquatic Science , 72 : 818 – 833 . Google Scholar Crossref Search ADS Sainsbury K. J. , Punt A. E. , Smith A. D. M. 2000 . Design of operational management strategies for achieving fishery ecosystem objectives . ICES Journal of Marine Science , 57 : 731 – 741 . Google Scholar Crossref Search ADS Su Z. , Peterman R. M. , Haeseker S. L. 2004 . Spatial hierarchical Bayesian models for stock-recruitment analysis of pink salmon (Oncorhynchus gorbuscha) . Canadian Journal of Fisheries and Aquatic Sciences , 61 : 2471 – 2486 . Google Scholar Crossref Search ADS Su Z. , Peterman R. M. 2012 . Performance of a Bayesian state-space model for semelparous species for stock-recruitment data subject to measurement error . Ecological Modelling , 224 : 76 – 89 . Google Scholar Crossref Search ADS Vaughan D. S. , van Winkle W. 1982 . Corrected analysis of the ability to detect reductions on year-class strength of the Hudson River white perch (Morone americana) population . Canadian Journal of Fisheries and Aquatic Sciences , 39 : 782 – 785 . Google Scholar Crossref Search ADS Wade P. R. 2000 . Bayesian methods in conservation biology . Conservation Biology . 14 : 1308 – 1316 . Google Scholar Crossref Search ADS Walters C. J. 1974 . An interdisciplinary approach to development of watershed simulation models . Technological Forecasting and Social Change , 6 : 299 – 323 . Google Scholar Crossref Search ADS Walters C. J. 1977 . Confronting uncertainty. In Pacific Salmon: Management for People , pp. 261 – 297 . Ed. by Ellis D.V. . University of Victoria Press , Victoria, British Columbia, Canada . Walters C. J. 1986 . Adaptive Management of Renewable Resources . MacMillan Press , New York . 374 pp. Watt K. E. F. 1968 . Ecology and Resource Management: A Quantitative Approach . McGraw-Hill Book Co ., New York . 450 pp. Wilson H. B. , Kendall B. E. , Possingham H. P. 2011 . Variability in population abundance and the classification of extinction risk . Conservation Biology , 25 : 747 – 757 . Google Scholar Crossref Search ADS PubMed Zuur A. F. , Tuck I. D. , Bailey N. 2003 . Dynamic factor analysis to estimate common trends in fisheries time series . Canadian Journal of Fisheries and Aquatic Sciences , 60 : 542 – 552 . Google Scholar Crossref Search ADS © International Council for the Exploration of the Sea 2018. 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)

Link, Jason, S;Dickey-Collas,, Mark;Rudd,, Murray;McLaughlin,, Richard;Macdonald, Nicol, M;Thiele,, Torsten;Ferretti,, Johanna;Johannesen,, Ellen;Rae,, Margaret

doi: 10.1093/icesjms/fsy169pmid: N/A

Abstract Mandates to execute ecosystem-based management exist but are not implemented sufficiently enough to reap the benefits of a growing blue economy. The uses, benefits, and value of the ocean from which the blue economy arises are increasingly expanding. So also are the pressures and stressors facing marine ecosystems. Marine transportation, offshore energy production, food production, coastal development, and other activities are placing increasing pressures on large portions of the World’s Ocean (AORA, 2017). There is a strong need to deal with the multiple uses and pressures across multiple sectors with multiple parties that have multiple goals. Doing so also concurrently provides business opportunities to address these challenges. To address these multiple ocean-uses among a wide array of sectors, it is well recognized that an ecosystem-based approach to management (EBM) is warranted (Leslie and McLeod, 2007). Yet despite the clearly recognized value of addressing trade-offs and prioritizing outcomes via execution of EBM, it is not widespread in practice. Several countries and organizations have committed to implementing EBM (Rodriguez, 2017), but it remains an exception in actual ocean management practice. One major impediment has been the perception that there are inadequate mandates to fully authorize EBM (e.g. Sardà et al., 2014; Marshak et al., 2017). The absence of sufficient and appropriate mandates implies that governing authorities do not have the tools necessary to effectively engage in EBM nor the authority to evaluate the current and future successes of policies and management strategies. To address the question of EBM mandate sufficiency, an international group of legal scholars, economists, social, administrative and political scientists, and natural resource practitioners was convened as part of the Atlantic Ocean Research Alliance (AORA, 2018). The spectrum of coverage from over 200 legal mandates from the EU, Canadian, US, and High Seas jurisdictions was considered. From this exercise, two notable outcomes arose. First, there are adequate, existing mandates to authorize EBM in the North Atlantic. In all jurisdictions considered, nearly all of the ocean uses, goods and services, pressures, and stressors have some level of legal or regulatory mandate coverage. In all these jurisdictions, even those ocean uses or pressures without direct mandate coverage have some form of overarching legislation or policy to address facets of cumulative impacts, coordinate planning, and ensure comprehensive, systematic consideration of ocean uses or pressures. It warrants noting that there is a wide range of types of mandates (Figure 1) beyond the typical legislative or regulatory mandates often invoked in mandate discussion. Policy tools and non-regulatory mandates are just as important, and those also exhibit a comparable degree of coverage for ocean use and pressure coverage as the legislative mandates, but are perhaps less widely used. We acknowledge that in some of the AORA jurisdictions the mandate for EBM could be stated more clearly or directly, but in some jurisdictions it is already quite clear and direct. The resulting conclusion is that collectively the existing laws, regulations, treaties, and policies provide sufficient mandate and a clear legal basis to support EBM for ocean management and governance in the North Atlantic. In theory, there is no legal basis not to do EBM, and in many cases a clear need to do so has emerged. Lack of mandate therefore cannot be claimed as a rationale for continued inaction on EBM. Figure 1. View largeDownload slide The conceptual multi-level approach depicting political mandate, legislative structure, and non-regulatory implementing policy. Mandates to implement EBM exist in international, national, and regional levels, and vary in their degree of prescriptiveness—i.e. allowing (enabling) vs. requiring (regulatory)—and in their degree of formality—legislative vs. policy statement. Providers of knowledge into the arena of ecosystem-based management should be aware of that arena and the distinctions between legislative/regulative processes and policy/other tools. Figure 1. View largeDownload slide The conceptual multi-level approach depicting political mandate, legislative structure, and non-regulatory implementing policy. Mandates to implement EBM exist in international, national, and regional levels, and vary in their degree of prescriptiveness—i.e. allowing (enabling) vs. requiring (regulatory)—and in their degree of formality—legislative vs. policy statement. Providers of knowledge into the arena of ecosystem-based management should be aware of that arena and the distinctions between legislative/regulative processes and policy/other tools. Second, the limited adoption of EBM in practice was in most cases not due to lack of mandates, but rather due to implementation failure that is a fragmented and limited use of the EBM approach. Among the key challenges to implementing EBM were the “classics” which have been recognized as critical factors for other venues intending to achieve an integrated perspective of sector-environment uses (Turnpenny et al., 2008), including the lack of specificity of mandates, institutional constraints, failure to provide sufficient and ongoing institutional resources, different ways of producing and integrating knowledge, and the unevenness of capacity to operationalize EBM among administrations and stakeholders. Additionally, mandates can be viewed as providing an authority to allow for the execution of governance activity vs. requiring that such activity be executed. It is this distinction among not only the types (Figure 1), but also the role of mandates that may be hindering the implementation of EBM. Political will is a necessary factor in the successful implementation of EBM given that it usually crosses political and administrative boundaries (AORA, 2018). Political will is often expressed through political mandate and may be reflected in a variety of ways; it can be formally expressed through legislative or programmatic action or more fluidly via informal means such as policy declarations (i.e. statements by Ministers or in the annual budgeting process) as depicted in Figure 1 as non-regulatory tools. Thus, EBM is an approach that varies considerably across jurisdictions in how it is interpreted and implemented, leading to confusion and overlap with other management efforts. We note that issues of linguistic uncertainty regarding what marine EBM means are now much less of an issue (Link and Browman 2014, 2017; Marshak et al., 2017), but how to actually implement EBM in practice for a given set of conditions remains a challenge. EBM can also challenge existing ocean governance paradigms (e.g. Berkes, 2012; Ramírez-Monsalve et al., 2016), thereby potentially exacerbating institutional and sectoral conflicts. It is imperative to address and resolve these potential problems and to develop governance and management solutions if successful implementation of EBM is to be achieved. In many instances these will not be revolutionary (Berkes, 2012) but rather a transformation evolving from the consolidation of existing mandates, ensuring their coherence (Ramírez-Monsalve et al., 2016; Rouillard et al., 2018), and linkage with political initiatives (UN, 2017) and scientific advances (e.g. AORA, 2017; Zador et al., 2017). All these are ongoing in the three AORA jurisdictions and can help to build momentum to nurture alliances for EBM implementation and maturation. The scientific context within which EBM mandates operate have advanced significantly since the origins of the EBM debate 30–40 years ago (WCED, 1987). We now have at our disposal larger and more readily available time series of data, increased computing capacity able to run fully coupled social-ecological system models, increasing appreciation of incorporating social, cultural, and indigenous peoples’ values and knowledge, and recognition that there are carrying capacity constraints with respect to ocean resources and use. From a political and legal perspective, we see a more inter-connected global economy and thus interdependencies and high degree of trade-offs, increased geographic conflict over resource access and utilization, and a clearly acknowledged need for conflict resolution and de-escalation across and within ocean use sectors (e.g. WBGU, 2013). The international policy agenda has also advanced and now reflects increased public awareness of and demand for attention on ocean environmental issues, the recognition of sustainable development principles, and the establishment and adoption of sustainable development goals (UN, 2017). Advances in the social sciences now allow us to better value ecosystem goods and services, conduct trade-off analyses, integrate broader knowledge paradigms, and transparently explore governance strategies and management options within and across jurisdictions. Advances across disciplines should provide the scientific and policy tools to facilitate further implementation of EBM, to empower an integrated and common approach to ocean governance and to take advantage of untapped blue economy potential. The knowledge base we have now compared with even 10–20 years ago gives us the capacity to effectively implement EBM. Ultimately mitigating these challenges of implementation requires recognizing the benefits of EBM. The scientific and management rationales and benefits of doing EBM have been well chronicled (e.g. Leslie and McLeod, 2007; Link and Browman, 2014), but a clearer, stronger business case for EBM, and ultimately for the blue economy, is warranted. The argument that EBM is a wise societal investment involves a number of factors. Any such analysis first needs to include the economic benefits (i.e. profits, resource rent capture, spin-off benefits) of Business as Usual (BAU) scenarios relative to scenarios where EBM is successfully implemented. Second, society must recognize that private benefits are only one part of the overall benefits accruing to society and should also account for a broad range of environment benefits (e.g. human health, community well-being, technological development opportunities) that depend on healthy oceans. Additionally, consideration and comparison of the transactions costs (i.e. the costs of coordination, negotiation, litigation, monitoring, and enforcement) of governance for BAU and EBM scenarios is needed. The business case for EBM is founded on the assertions that economic profitability for the private sector (and spin-offs and tax revenues) will decline if ocean resources are over-exploited over time, that non-market benefits derived from ecosystem services are usually inadequately accounted for in BAU analyses and that non-market and social benefits under BAU will erode as the public recognizes deteriorating ocean conditions. BAU will likely lead to increased levels of contestation and costs over time given competing, multiple objectives across ocean uses and pressures. Implementing EBM implies upfront costs (e.g. negotiations, development or revision of governance venues and structures, revised legislation, etc.) that are higher than BAU but reduce losses in the longer-term (i.e. due to reduced levels of conflict, longer-term profit, continued delivery ecosystem goods and service, etc.). Investments in EBM also serve to deliver more predictable, reliable, ocean governance, with benefits to the private sector that help protect profitability in the face of increasing environmental uncertainty. A more predictable social, economic, and political environment also allows for longer planning horizons, and the ability to invest accordingly for longer-term returns. The deliberative and participatory orientation of EBM (Rudd, 2004) also helps alleviate uncertainty and integrate knowledge from the natural, social, and legal sciences. If the business case for EBM grows stronger, through better articulation of its benefits or increasing awareness of the environmental and political costs of BAU, we should expect to see increased levels of political support for EBM implementation efforts. To improve implementation of EBM and thereby obtain the benefits of doing so, we recommend the following items for AORA jurisdictions, with probable applicability well beyond just the North Atlantic. First would be to facilitate further institutionalization of EBM by realignment of funding from project to base budget to ensure sustained and long-term capacity. Next would be to revisit and consider the effectiveness and impact of existing over-arching, integrative mandate(s), along with more effective use of existing mandates to implement EBM. We recommend mitigation of implementation barriers, partly by identifying and acknowledging implementation challenges and partly by better articulating the benefits of EBM in a political, economic, and social context. Realignment of calls for research need to be increasingly cross-disciplinary to address both institutionalization and implementation barriers of EBM. Ultimately the best business case for EBM is that it offers a dynamic, adaptive, and holistic approach to ocean management, developed over more than two decades, to address the multiple and complex pressures facing our global ocean. Acknowledgements We thank A. Marshak, K. Abrams, and K. Osgood for prior reviews and Robin Anderson, Paul Snelgrove, Dave Fluharty, Ana-Teresa Cetano, Murray Roberts, Alistair Hobday, Sarah Carr, Manuel Barange, and Beth Fulton for prior discussions that informed this topic. This work was supported by the Atlantic Ocean Research Alliance Coordination and Support Action (AORA-CSA) a European Union Horizon 2020 coordination and support action [Grant Agreement number 652677]. References Atlantic Ocean Research Alliance Report (AORA). 2017 . Working Group on the Ecosystem Approach to Ocean Health and Stressors, January 2017, Reykjavik, Iceland. 53 pp. https://atlanticresource.org/aora/sites/default/files/GalleryFiles/Publications/2017–01_EA2OHS_report.pdf AORA. 2018 . Working Group on the Ecosystem Approach to Ocean Health and Stressors. Mandates for Ecosystem-based Ocean Governance across Canada, the EU, and the US March 2018. London, UK. 52 pp. https://atlanticresource.org/aora/sites/default/files/GalleryFiles/Publications/2018–03_EA2OHS_mandates_report.pdf Berkes F. 2012 . Implementing ecosystem-based management: evolution or revolution? Fish and Fisheries , 13 : 465 – 476 . Google Scholar Crossref Search ADS Leslie H. M. , McLeod K. L. 2007 . Confronting the challenges of implementing marine ecosystem‐based management. Frontiers in Ecology and the . Environment , 5 : 540 – 548 . Link J. S. , Browman H. I. 2014 . Integrating what? Levels of marine ecosystem-based assessment and management. Introduction to integrated assessments theme section . ICES Journal of Marine Science , 71 : 1170 – 1173 . Google Scholar Crossref Search ADS Link J. S. , Browman H. I. 2017 . Operationalizing and implementing ecosystem-based management . ICES Journal of Marine Science , 74 : 379 – 381 . Google Scholar Crossref Search ADS Marshak A. R. , Link J. , Shuford R. , Monaco M. , Johannesen E. , Bianchi G. , Anderson M. R. , et al. 2017 . International perceptions of an integrated, multi-sectoral, ecosystem approach to management . ICES Journal of Marine Science , 74 : 414 – 420 . Ramírez-Monsalve P. , Raakjær J. , Nielsen K. N. , Laksá U. , Danielsen R. , Degnbol D. , Ballesteros M. , et al. 2016 . Institutional challenges for policy-making and fisheries advice to move to a full EAFM approach within the current governance structures for marine policies . Marine Policy , 69 : 1 – 12 . Google Scholar Crossref Search ADS Rodriguez N. A. 2017 . A comparative analysis of holistic marine management regimes and ecosystem approach in marine spatial planning in developed countries . Ocean & Coastal Management , 137 : 185 – 197 . Google Scholar Crossref Search ADS Rouillard J. , Lago M. , Abhold K. , Röschel L. , Kafyeke T. , Mattheiß V. , Klimmek H. 2018 . Protecting aquatic biodiversity in Europe: how much do EU environmental policies support ecosystem-based management? Ambio , 47 : 15 – 24 . Google Scholar Crossref Search ADS PubMed Rudd M. A. 2004 . An institutional framework for designing and monitoring ecosystem-based fisheries management policy experiments . Ecological Economics , 48 : 109 – 124 . Google Scholar Crossref Search ADS Sardà R. , O’Higgins T. , Cormier R. , Diedrich A. , Tintoré J. 2014 . A proposed ecosystem-based management system for marine waters: linking the theory of environmental policy to the practice of environmental management . Ecology and Society , 19 : 51 . Google Scholar Crossref Search ADS Turnpenny J. , Nilsson M. , Russel D. , Jordan A. , Hertin J. , Nykvist B. 2008 . Why is integrating policy assessment so hard? A comparative analysis of the institutional capacities and constraints . Journal of Environmental Planning and Management , 51 : 759 – 775 . Google Scholar Crossref Search ADS United Nations. 2017 . UN Conference to Support the Implementation of Sustainable Development Goal 14: Conserve and Sustainably Use the Oceans, Seas and Marine Resources for Sustainable Development. 5–9 June 2017, New York, US (A/RES/71/312, 2017). http://www.un.org/ga/search/view_doc.asp? symbol=A/RES/71/312&Lang=E World Commission on Environment and Development (WCED). 1987 . Our Common Future . Oxford University Press , Oxford . WBGU. 2013 . World in Transition: Governing the Marine Heritage . German Advisory Council on Global Change , Berlin . https://www.wbgu.de/fileadmin/user_upload/wbgu.de/templates/dateien/veroeffentlichungen/hauptgutachten/hg2013/wbgu_hg2013_en.pdf Zador S. , Holsman K. , Aydin K. , Gaichas S. 2017 . Ecosystem considerations in Alaska: the value of qualitative assessments . ICES Journal of Marine Science , 74 : 421 – 430 . Google Scholar Crossref Search ADS Published by International Council for the Exploration of the Sea 2018. This work is written by US Government employees and is in the public domain in the US.

Hilborn,, Ray

doi: 10.1093/icesjms/fsy138pmid: N/A

Abstract Most reporting of stock status accumulated at a national or regional level gives statistics on what proportion of the stocks are below some abundance threshold or above some fishing mortality rate threshold. This approach does not convey useful information on the performance of the fisheries management system in maximizing long-term sustainable yield, which is the primary objective of most national and international fisheries legislation. In this paper, I present a graphical approach for representing how much yield is being lost as a consequence of current suboptimal abundance and fishing pressure. Using the EU stocks assessed by ICES as an example, I show how traditional criteria for overfished and overfishing fail to display realistic information about the performance of the fishery. This approach provides much more useful information for the public and policy makers. Introduction For several decades, national and international agencies have described the overall status of fish stocks and have almost exclusively based this upon the abundance of the fish stock relative to some reference point. For instance, the 2016 report by FAO on the status of fish stocks reports three categories: (i) overfished, (ii) maximally sustainably fished, and (iii) underfished. All of these categories are based on biomass relative to the reference point of stock size that will produce maximum sustainable yield (BMSY) (FAO, 2016). FAO considers a maximally sustainable fished stock to be between 0.8 and 1.2 BMSY, an overfished stock to be <0.8 BMSY, and an underfished stock to be >1.2 BMSY. For US federally managed fisheries, NOAA reports on what fraction of stocks are overfished (https://www.fisheries.noaa.gov/feature-story/status-stocks-2017), usually defined as being <0.5 BMSY, though the exact ratio differs by region and type of stock. NOAA also defines overfishing as a fishing mortality rate being greater than the level that would produce MSY. New Zealand has essentially the same definitions as the US (MFNZ, 2008) except that they define stocks <0.5 BMSY to be “depleted” rather than overfished, as they recognize that stocks may be at low abundance for reasons other than fishing pressure. The European Commission has no formal policy for classifying stock status, but Froese et al. (2018) have recently described the status of stocks in Europe and have stated that any stock with F > FMSY or B < 0.5 BMSY as overfished or beyond safe biological limits. Most classification systems use MSY as a reference point because most national legislation and international agreements have MSY as the default objective for fisheries management. In the United States, the Magnuson–Stevens Act (ACT, 1996) specifies several purposes which include “maintain, on a continuing basis, the optimum yield from each fishery.” The Act further states that “optimum yield” means the amount of fish which “is prescribed as such on the basis of the maximum sustainable yield from the fishery, as reduced by any relevant economic, social or ecological factor.” Hence, the default is maximum sustainable yield in the absence of specific economic, social, or ecological factors. The international Law of the Sea specifies that fisheries management measures “shall also be designed to maintain or restore populations of harvested species at levels which can produce the maximum sustainable yield.” The general intention of these classifications is to inform fishery managers and the public which stocks are in need of rebuilding (overfished stocks) and which have the potential for more intense harvesting (underfished stocks). There are a number of deficiencies in this approach. A key problem with choosing a biomass threshold to define “overfished” is that fish stocks fluctuate naturally and, if managed to produce MSY, will fluctuate around BMSY. Indeed, we would expect stocks managed to achieve MSY to be <BMSY roughly one-half of the time. Using a meta-analysis of the amount of variability in recruitment and the auto-correlation of this variability, Thorson et al. (2015) estimated that stocks managed to produce MSY would be <BMSY 60% of the time and <0.5 BMSY 8% of the time. Thorson’s calculations were conducted assuming perfect control of the exploitation rate and perfect assessment of the stock size, so uncertainties in science and management would undoubtedly make the frequency of dropping <0.5 BMSY greater than 8%. Thus, to label any stock <BMSY (or even <0.8 BMSY) as overfished may be deceptive from either the perspective of supplying management advice or from conveying to the public how fisheries are performing. Because stocks may be at low abundance due to natural variation in recruitment or survival, managers actually look at fishing mortality as the more important metric to determine a need to change management. In the United States, NOAA also reports on “overfishing,” which is defined as the rate of harvest being greater than that which would produce MSY. Since managers can only directly control harvest, when harvest rates are above the level that produces MSY, they need to be reduced, and when stocks are at particularly low abundance, harvest rates should be dramatically reduced to allow the stocks to rebuild more rapidly. Hence, the real use for overfished classifications is reporting to the public and government officials on the overall status of fish resources, not in providing management advice for individual stocks. There is a real need to identify how fisheries are performing, with respect to both yield, but also social and economic objectives. However, social and economic objectives are highly specific to individual fisheries and countries. Hence, in what follows, I will suggest a method of reporting the status of fisheries on fish stocks and the performance of the fisheries management system relative to production of maximum long-term sustainable yield. Material and methods I calculated lost yield at current fishing pressure or biomass by evaluating equilibrium yield at current fishing pressure or biomass compared to MSY yield. I used the Pella and Tomlinson (1969) biomass dynamics model, which has been used previously for this purpose (Branch et al., 2013; Hilborn and Costello, 2018): Bt+1=Bt+φmBtK-φmBtKn-Ct, (1) where m is the maximum sustainable yield, K is carrying capacity, n is a shape parameter that determines the ratio BMSY/K, Bt is the biomass at time t, and Ct is the catch at time t. The value of φ depends on n: φ= nnn-1 n-1. (2) The Schaefer (1954) model has n = 2, the Fox (1970) model has n = 1, and Thorson et al. (2012) estimated using a meta-analysis of 147 stocks that the median is n = 1.736: BMSYk=n11-n . (3) So, the Bt/K in Equation (1) is: B*=BMSYkBBMSY. (4) The fraction of potential yield lost at biomass B* is: 1-φBt*-φBt*n (5) and the shape of this loss function is shown in Figure 1 for different models. Figure 1. View largeDownload slide Fraction of yield lost vs. U/UMSY for three different models. Figure 1. View largeDownload slide Fraction of yield lost vs. U/UMSY for three different models. If we define the exploitation rate Ut as Ct/Bt, the fraction of potential yield lost at any fishing mortality rate U* (defined as Ut/UMSY) is: 1-U*n-U*n-11n-1. (6) The shape of lost yield vs. U* is shown in Figure 2. We can calculate the status of a specific fishery with respect to its capability to produce the best long-term yield on the basis of either its current abundance or its current fishing pressure. We can then summarize, for a group of stocks, how the fishery as a whole is performing relative to the capability of the stocks to produce maximum possible yield. We can do this with respect to either current fishing mortality or current biomass for each stock. Figure 2. View largeDownload slide Fraction of yield lost vs. B/BMSY for three different models. Figure 2. View largeDownload slide Fraction of yield lost vs. B/BMSY for three different models. We can do our lost yield calculations based on either fishing mortality (U*) or current biomass (B*). For a given collection of stocks, we can calculate performance across all stocks simply by weighting the yield for each stock by the MSY for that stock. If Ys is the fraction of the yield lost for stock s (calculated for both U* and B*), the fraction of total potential yield obtained is: T=∑s(1-Ys) MSYs∑sMSYs, (7) where T is the performance measure, and the fraction of yield lost (1 – T) will occur in two ways. If we are overfishing or have overfished stocks, then either U > UMSY or B < BMSY. If we are fishing too little, then U < UMSY or B > BMSY. We can thus separate potential yield into three groups: (i) yield that is being obtained at current U or B, (ii) yield that is lost from stocks fished too hard, and (iii) yield lost from stocks fished too lightly. This has analogies to the Goldilocks fairy tale where some porridge was too hot, some too cold, and some just right. Some stocks are fished too hard (too hot), some are fished too little (too cold), and the yield obtained at current conditions is “just right.” We can undertake these calculations for each year, as shown in Figure 3, which reflects results from such calculations using U/UMSY for Atlantic tuna and billfish stocks from data in the RAM Legacy Stock Assessment Database (http://ramlegacy.org/) assuming the Thorson value for n. Figure 3. View largeDownload slide Goldilocks plot for Atlantic tuna and billfish based on fishing mortality. The solid line shows the fraction of total MSY from stocks assessed that year. Figure 3. View largeDownload slide Goldilocks plot for Atlantic tuna and billfish based on fishing mortality. The solid line shows the fraction of total MSY from stocks assessed that year. This shows that, in the early 1950s, there was no yield lost from fishing too hard (U>UMSY), and about 78% of the potential yield was lost from low fishing pressure. At the fishing pressure in the early 1950s, the long-term yield would have been about 22% of the potential. However, starting in the 1970s, fishing pressure increased, and there was some yield lost from fishing too hard, while the amount lost by fishing too lightly had diminished greatly. In the 1990s, the yield lost by both fishing too lightly and fishing too hard was on the order of 10%. Of course, it is not possible nor necessarily desirable to actually catch MSY from all stocks. Some stocks may be too low in value to warrant full exploitation, and we may want to maintain exploitation rates below MSY for economic or environmental reasons. Similarly, some stocks may be intentionally fished harder than UMSY because they are part of a mixed-stock fishery. Results We can use this approach to examine the history of EU stocks assessed by ICES. This excludes the Mediterranean stocks that ICES does not assess and the non-EU stocks that ICES does assess. The results are shown in Figures 4 and 5 also using the Thorson median value of n. Figure 4. View largeDownload slide Goldilocks plot for European Union ICES stocks based on fishing mortality. The solid line shows the fraction of total MSY from stocks assessed that year. Figure 4. View largeDownload slide Goldilocks plot for European Union ICES stocks based on fishing mortality. The solid line shows the fraction of total MSY from stocks assessed that year. Figure 5. View largeDownload slide Goldilocks plot for European Union ICES stocks based on current abundance. The solid line shows the fraction of total MSY from stocks assessed that year. Figure 5. View largeDownload slide Goldilocks plot for European Union ICES stocks based on current abundance. The solid line shows the fraction of total MSY from stocks assessed that year. These two figures show roughly similar pictures, with the fishing mortality plot showing more loss than the biomass-based plot. Post 2010, there is very little yield lost through excess fishing pressure or by stocks being at low abundance, although there is a growing trend of lost yield from low fishing pressure since the turn of the century. The biomass-based plot (Figure 5) shows more potential yield (blue) and less red than the effort-based plot. This is because there were a number of stocks with relatively high effort, but, at the same time, relatively high biomass. This can be caused by increasing fishing pressure when U is high, but B has not declined to equilibrium levels, or stocks that encounter unusually good environmental conditions. The message for policy-makers and the public is that, overall, there is no need to further reduce fishing pressure and perhaps a need instead to identify how to increase exploitation on lightly exploited stocks. Nevertheless, some stocks may be fished too hard or be at low abundance, and improved management of these stocks requires ways to address the issue. However, combined across all of the stocks assessed, the losses from both excessive fishing pressure and diminished abundance are quite low. For contrast, we can look at the trends of the number of stocks overfished by the US definition (B/BMSY < 0.5), the FAO definition (B/BMSY < 0.8), and the most conservative definition (B/BMSY < 1) (Figure 6). Figure 6. View largeDownload slide Fraction of EU stocks assessed by ICES whose biomass was below the different thresholds shown. Figure 6. View largeDownload slide Fraction of EU stocks assessed by ICES whose biomass was below the different thresholds shown. This graph suggests that overfishing is a serious problem regardless of the metric used and would tend to imply to viewers that there is an urgent need to reduce fishing pressure on an appreciable fraction of stocks and that substantial yield is being lost as a result of excess fishing pressure. This perception would be strongly reinforced by Figure 7, which shows the fraction of stocks with F > FMSY and is a dramatically different picture. This graph suggests that overfishing is a major and continuing problem. Figure 7. View largeDownload slide Fraction of EU stocks assessed by ICES whose fishing mortality rate was >FMSY. Figure 7. View largeDownload slide Fraction of EU stocks assessed by ICES whose fishing mortality rate was >FMSY. Discussion At a time when much fisheries policy seems to be guided by emotion and activist agendas, scientists need to communicate clearly to policy-makers and the public the scientific understanding of the status of our resources and the performance of fisheries management agencies. The often repeated “90% of fish stocks are overexploited or fully exploited” as a condemnation of the fisheries management performance is an example of both (i) the failure to communicate appropriate information and (ii) activist agendas. Traditional methods of describing the impact of overfishing on fisheries production have not appropriately or effectively communicated the cost of overfishing. I suggest that the Goldilocks plots do communicate the heart of the real issue, which relates to the extent of lost potential fisheries production much more effectively and understandably. The calculation of lost yield has been reported before (Branch et al., 2013; Hilborn and Costello, 2018), but systematically plotting it over time as an alternative to plotting the proportion of overfished stocks is new. The Kobi plot has become a common method of displaying stock status; when the size of the points in the Kobi plot is proportional to potential yield (Worm et al., 2009), it conveys considerable information about stock status. However, the extent of potential lost yield is not obvious from a Kobi plot, as the loss depends in a nonlinear way on the distance from the BMSY or FMSY target. The Kobi plot, while very useful within scientific circles, is often too complex for the public and policy-makers to grasp key implications. Overfishing is a concern for more than just lost fish production. Any concerns about the status of species that are not abundant or commercially very important will not be captured by Goldilocks plots. The scientific community needs to consider how best to present information on the status of these smaller stocks as well. The yield lost by underfishing or from stocks at high abundance is a more complex issue than yield lost from overfishing. When we use U/UMSY as a guide to potential increases in yield, some stocks may be at low abundance and rebuilding under low fishing pressure. Managers want to keep fishing pressure low while stocks rebuild, so the blue areas in the Goldilocks plot will show the potential increase in yield, but for stocks at low abundance, the policy prescription would not be an immediate increase in fishing pressure. When we use B/BMSY as the guide for underexploited stocks, the public and policy-makers need to understand that there are several factors that could result in underexploitation, and while the potential yield is certainly there, it may not be economically obtainable or desirable for reasons of ecosystem protection. In summary, in the context of most national fisheries objectives, the primary concern about overfishing is the loss in potential yield. The Goldilocks plots provide a much more informative presentation of this loss than the traditional methods of reporting the status of individual stocks relative to any threshold abundance or fishing mortality. Acknowledgements This analysis was made possible by the RAM Legacy Stock Assessment Database, and I thank all those who have contributed to and financially support the database. Pamela Mace provided details of the New Zealand definitions of overfished and overfishing. Two anonymous reviewers provided very useful suggestions. References ACT A. 1996 . Magnuson–Stevens fishery conservation and management Act . Public Law , 94 : 265 . Branch T. A. , Hively D. J. , Hilborn R. 2013 . Is the ocean food provision index biased? Nature , 495 : E5 – E6 . Google Scholar Crossref Search ADS PubMed FAO . 2016 . The State of World Fisheries and Aquaculture 2016. Contributing to Food Security and Nutrition for All. Food and Agriculture Organization of the United Nations. Rome, 200. pp. Fox W. W. Jr. 1970 . An exponential surplus-yield model for optimizing exploited fish populations . Transactions of the American Fisheries Society , 99 : 80 – 88 . Google Scholar Crossref Search ADS Froese R. , Winker H. , Coro G. , Demirel N. , Tsikliras A. C. , Dimarchopoulou D. , Scarcella G. et al. 2018 . Status and rebuilding of European fisheries . Marine Policy , 93 : 159 – 170 . Google Scholar Crossref Search ADS Hilborn R. , Costello C. 2018 . The potential for blue growth in marine fish yield, profit and abundance of fish in the ocean . Marine Policy , 87 : 350 – 355 . Google Scholar Crossref Search ADS MFNZ . 2008 . Harvest Strategy Standard for New Zealand Fisheries . Ministry for Primary Industries , New Zealand Government . 25 pp. Pella J. J. , Tomlinson P. K. 1969 . A generalized stock production model . Inter-American Tropical Tuna Commission Bulletin , 13 : 419 – 496 . Schaefer M. B. 1954 . Some aspects of the dynamics of populations, important for the management of the commercial marine fisheries . Inter-American Tropical Tuna Commission Bulletin , 1 : 27 – 56 . Thorson J. T. , Cope J. M. , Branch T. A. , Jensen O. P. 2012 . Spawning biomass reference points for exploited marine fishes, incorporating taxonomic and body size information . Canadian Journal of Fisheries and Aquatic Sciences , 69 : 1556 – 1568 . Google Scholar Crossref Search ADS Thorson J. T. , Jensen O. P. , Hilborn R. 2015 . Probability of stochastic depletion: an easily interpreted diagnostic for stock assessment modelling and fisheries management . ICES Journal of Marine Science , 72 : 428 – 435 . Google Scholar Crossref Search ADS Worm B. , Hilborn R. , Baum J. K. , Branch T. A. , Collie J. S. , Costello C. , Fogarty M. J. et al. 2009 . Rebuilding global fisheries . Science , 325 : 578 – 585 . Google Scholar Crossref Search ADS PubMed © International Council for the Exploration of the Sea 2018. 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)

Steele, John, H;Brink, Kenneth, H;Scott, Beth, E

doi: 10.1093/icesjms/fsy149pmid: N/A

Abstract The transition of plants and animals from sea to land required adaptation to a very different physical and chemical environment. In this paper, we focus on the consequences of the differences between the magnitude of the variability of ocean and atmospheric dynamics, with the ocean environment (in particular temperature and currents) being two to three orders of magnitude less variable than that on land. We suggest that greater insights on possible responses of marine vs. terrestrial systems to rapid climate change can be gained by considering that terrestrial vertebrates, invertebrates and plants have evolved from marine organisms that, pre-Cambrian, had early life history developmental stages as planktonic larvae. Marine larvae were/are adapted to the predictable and minimal range of temperature changes and regularities in ocean currents, as most organisms utilize the energy in these currents as an “auxiliary” source for predictable gamete and larvae dispersal. Post-Cambrian, on land, no such simple strategy was available; instead, most terrestrial organisms have evolved reproductive strategies and behaviours to eliminate, or at least minimize, the consequences of much larger atmospheric variability. Adapting our future use of these systems sensibly will require greater understanding of how the two regimes respond to rapid climate change. Introduction About 500 million years ago, give or take a few million years, the oceans were teeming with life and the land was a desert. We assume the invasion of the land began when simple plants—green slime—spread inland from coastal ponds, enhancing the oxygen in the air for animals to follow (Lenton and Watson, 2011); first the invertebrates, crustaceans having common ancestors that evolved into insects; then the more complex transformations from fish to quadrupeds. Within 100 million years there were fully developed terrestrial ecosystems that provided the basic material from which we and all contemporary life on land are derived. Some of the organisms within these ecosystems developed the capability to alter their environment (in particular us humans with our use of land for farming) in ways that marine organisms never achieved. How did terrestrial life—the Johnny-come-lately in the ecosystem game—develop the capabilities to manage or mismanage the world? Why had the earlier frontrunners in the sea not gotten there first? And, what are the drivers that have led to the present-day differences? Apparently it was too difficult to start from scratch on land (Lenton and Watson, 2011), so the tedious and complicated process we call evolution, with all its constraints, was required to transform the marine ecosystem, into a terrestrial one. This adaptation required the transformation of elegant sequences in the sea, where plants, herbivores and carnivores increase in an orderly fashion in body size, life span and ambit, to a messy, non-systematic grouping on land with short and long-lived plants, large warm-blooded herbivores, and social insects with lifespans not related to temporal and spatial scales (Figure 1). Figure 1. View largeDownload slide Schematic presentation of ecological.space and time scales for major features in the dynamics of atmosphere and oceans and food web components (a) marine (Steele, 1978) and (b) terrestrial (Delcourt et al., 1982). Adapted from Steele (1992). Figure 1. View largeDownload slide Schematic presentation of ecological.space and time scales for major features in the dynamics of atmosphere and oceans and food web components (a) marine (Steele, 1978) and (b) terrestrial (Delcourt et al., 1982). Adapted from Steele (1992). Popular writing (Gordon and Olson, 1995; Fortey, 1999) covers the adaptations to the “statics” of the terrestrial environment (Denny, 1990), such as the effects of gravity on plant structures, the problems of desiccation, the elimination of waste products, and other physiological processes that are the subject of paleontology. The tendency is to give primacy to the terrestrial perspective, with titles such as “Concepts of Adaptation in Aquatic Animals: Deviations from the Terrestrial Paradigm” (McFall-Ngai and Manahan, 1990), or “Why life histories evolve differently in the sea” (Strathmann, 1990). This tendency is understandable, but we believe that an appreciation of the intricacies involved in deriving terrestrial solutions from marine origins is critical to an understanding of the evolution of terrestrial species as well as relationships between species at the ecosystem level in both regimes. The main thesis in this paper is that, behind the need to adapt to the very different opportunities presented by the terrestrial environment, one of the main drivers was the adaptation to the much more variable atmospheric component of the terrestrial environment. We suggest that between evolutionary constraints and the need to deal with much higher variability in the environment, that terrestrial organisms evolved life-history traits, especially developmental life-history traits (egg and juvenile), and behaviours that were needed to dampen and eliminate the effects of much higher environmental variability. The basic premise that underlies our desire to ask readers to re-look at the comparisons between present-day terrestrial and marine ecosystems is to provide readers with the information about the contrasts in environmental variability between the regimes and to better appreciate the particular forms and processes we find on land and how they may be affected by rapid climate change, by understanding how these may have evolved from life in the sea. This paper will focus on the consequences of the very different space and time scales of the physical dynamics in the two regimes, while recognizing the constraints imposed by the very different static properties of the two systems. We propose to focus on three principal solutions (illustrated in Figure 2) to the environmental challenges raised by the transition to the use of the land. The point is to view these transitions in the light that they are bounded by evolutionary constraints, especially within developmental life-history traits that require a very predictable stable environment: The three solutions are: (i) that some terrestrial plants evolved very much longer lifespans to smooth out the short-term but large-amplitude environmental perturbations; (ii) many invertebrates and vertebrates built nests to provide an environment that moderated or removed them and especially their eggs from this variability; and (iii) some vertebrates also created their own regular internal environment by becoming warm-blooded and viviparous. We contrast the possible response for adaptation to open-sea predictably low environmental variability with the possible reasons related to high environmental variability and demonstrate these three possible solutions by contrasting differences in evolutionary adaptation by superficially similar marine and terrestrial species. We suggest that it is important to consider the possible environmental variability drivers behind early life history traits by demonstrating the divergent consequences of the evolution of those traits in the two regimes and that looking at the differences between modern marine and terrestrial species in this light will allow more accurate predictions of what will be the consequences of rapid climate change in the contrasting regimes. Figure 2. View largeDownload slide Illustrations of the three major adaptations to terrestrial life for plants, invertebrates, and vertebrates. Figure 2. View largeDownload slide Illustrations of the three major adaptations to terrestrial life for plants, invertebrates, and vertebrates. Our motivation is concern about how we think about affective adaptation to future rapid climate change, especially as the human achievement of combining two of the original strategies for adaptation to a terrestrial environment: “invertebrate” nest making on the largest scale, with efficient viviparous reproduction. This has required major disruption of the first adaptation—long-term ecological stability built upon long-lived perennial plant communities—to achieve high production of selected annual species. Not only has this affected biodiversity; possibly more significant, dependence on annual plants removes the natural ability to dampen out climatic fluctuations at the decadal to century scales, and requires large-scale human intervention. Thus, the original challenge in the transition to land—independence from the effects of environmental variability—has become the ultimate adaptation we still need to make in understanding how to best manage our land and seas under rapid climate change. Future climate looks set to be far more variable than we humans have experienced in the last 10 000 years since our species started to become dependent on annual plant production (Pearce, 2007). The major sections of this paper set out below are firstly “The transition to the land” which describes the range of transitions to the land across plants, invertebrates and vertebrates via the relevant physical processes in the oceans and atmosphere, with examples of ecological adaptations at the species level. The second major section of this paper is “Comparative study: rapid climatic change” which considers the consequences for rapid climatic change of the different adaptations. The arguments in this review about relationships between physical and biological processes and marine and terrestrial communities, and the generalizations about broad groups such as “plants” or “invertebrates”, do not have the authority of physical laws. Rather, they invoke consensus of opinion or preponderance of evidence. We do not try to describe the detailed processes determining the evolution of traits and the estimation of evolutionary trade-offs (Stearns, 1992). Thus, the “conclusions” try to discern patterns that may illuminate certain ecological processes and evolutionary sequences, and in doing so provide insight into particular important sectors of our environment that the world has come to rely on in many contrasting parts of our world. The transition to the land Importance of low environmental variability during early (developmental) life-histories Evolution is a long-term process of adaptation to the environment. The challenges for organisms invading the land were, initially, posed by the utterly different physical conditions between the two regimes. We think of the adaptation of individual species to “static” properties, such as the structural problems posed by gravity and oxygen supply, or of physiological processes required for temperature control and to prevent desiccation. But there were also problems raised by the difference between the levels of predictable variability and rates of change in the different environments. We are aware of the millennial-scale glaciations driven by the Malenkovich Cycles (Kerr, 1987), or, at the other extreme, of the global extinctions caused by meteorites or volcanoes. Both of these have tested the ability of life on land and in the sea to resurrect their ecosystem patterns with new components. But between the millennial-scale trends and the occasional global catastrophes, there is the need to adapt to the daily, monthly, or yearly variations in the environment. And, especially, to evolve modes of reproduction and rearing of young that maximize survival over generations. The early stages in the life cycle have the greatest exposure to the environment and display great diversity in the processes that have been developed by terrestrial plants and animals. It is in this phase that we find the very different solutions evolved in the sea and on the land in relation to the very different characteristics of the two regimes. For most of us land-based inhabitants, the weather is a habitual topic of conversation. Its unpredictability beyond a few weeks is now established scientifically. Even the seasonal cycles can be quite variable. The predictability of weather a year ahead is about the same as for three weeks from now. Uncertainties about the temperature, the humidity, the wind strength and direction can affect our plans for outdoor activities. We find that most terrestrial animals devote a great deal of their acquired energy to mitigating the consequences of this unpredictability; especially for the reproduction and nurture of their young. In the sea, most animals release large numbers of eggs or larvae to the water to be carried along by currents with broadly predictable trajectories and speeds to their next habitat that will provide food. This “auxiliary” source of energy (Mann and Lazier, 1991) is widely used by marine animals for transport of their young and does not have anywhere near the comparable level of predictability in the atmosphere. On land, many animals provide shelter for their offspring within their bodies or in specially constructed nests, and provide parental care by actively foraging for food for their young. Why have these very different strategies for parental care evolved in the two regimes? It is not that the alternatives are infeasible. Nephrops norvegicus (“scampi”) creates extensive burrows, comparable to those of rabbits, in the muddy sea bed and carries its eggs until they are ready to hatch but then releases them to the currents with no more parental care. Within elasmobranch fish, skates are oviparous as they lay pouches containing a few large eggs, known as “maiden’s purses”, and most sharks are viviparous but all parental care ends once young are released (Wourms and Lombardi, 1992). But within the superclass Osteichthyes, with more than 28 000 species, the vast majority of species, such as the commercially important cod (Gadus sp.), will broadcast millions of tiny eggs to the sea with only 2–3% of species described as viviparous and only 15% of marine species showing any signs of parental care (Baylis, 1981). On land, many trees scatter large number of seeds with the wind, and several species of spiders and aphids use the wind for limited juvenile dispersal (Richter, 1970; Loxdale et al., 1993; Bonte and Saastamoinen, 2012), but there is nothing on the scale of the directed transport by marine eggs and young. Why has the vast majority of marine life evolved to make use of its physical surroundings as part of their reproductive plan, whereas terrestrial animals expend their energy in insulating themselves and their young from their environment? Physical differences between regimes Levels of variability The answer to the differences between the marine and terrestrial regimes must, in part, be attributed to the contrasting physical differences between them. Plankton and most fish are close to being neutrally buoyant and, unlike the birds and the bees, do not expend significant energy in staying suspended. The sea, unlike the air, contains the nutrients necessary for plant growth so that a complete food web can exist without a solid base. Thus most marine life depends on the microscopic plants in the water column for the energy and nutrients to support their food webs. But this does not account for the lack of parental care in the life cycles of the great majority of marine animals. For an explanation of the predominance of strategies based on the release of eggs or larvae to the “care” of the physical environment, we need to consider the differences between the marine and terrestrial habitats in terms of their oceanic and atmospheric environments and look more closely at the dynamics as well as the statics. The ocean and the atmosphere are both subject to the same laws of fluid dynamics, but it is because of their very different physical properties (e.g. density, thermal expansion coefficient, and viscosity), that the time scales in the ocean are much longer and the spatial scales for comparable features are much shorter (Figure 1). This is especially true for eddies in the open ocean or cyclonic systems in the atmosphere, where most of the kinetic energy is found and which are generated by essentially random processes. The atmospheric eddy scales are much shorter in time (days vs. weeks) and larger in space (104 vs. 102 km) than ocean eddies (Gill, 1982; Clark, 1985). It is these features that introduce the variability in air currents, temperature, and rainfall that will affect the plants and animals. Over the continental shelf, where direct atmospheric forcing of the ocean is particularly important, oceanic time (but not so much space) scales are influenced by atmospheric processes. It is only on short time scales (10 s of minutes or less) that the differing molecular properties of air and water constrain heat transport between an organism and the environment. From Figure 3, it is clear that the transition to land involved the challenge of variability at quite different scales and would have required adaptation to much more rapid and much larger-scale variance. Figure 3. View largeDownload slide Information about temporal power spectra. In the first three panels, spectral slopes are indicated as, e.g. “f-2” and the logarithmic vertical axis is spectral density of temperature (i.e. temperature variance divided by frequency increment) and the horizontal logarithmic axis is frequency. (a) A spectrum of atmospheric temperature in England, redrawn after (NAS, 1975). This is presented because an unusually long time series allows spectral estimates at unusually low frequencies. (b) A spectrum of ocean sea level estimated using a combination of instrumental and geological data (J. Imbrie, pers. comm.). (c) A spectrum estimated using long-term data from ice cores (an air temperature proxy) merged with higher frequency information from continental records, redrawn after (Pelletier,1998). (d) Vertical axis: the negative coefficients of power spectral slope (i.e. the exponent of frequency) for three regions obtained from near-surface air temperature data, redrawn after (Vasseur and Yodzis, 2004). On the horizontal “axis”, three different environments are labelled. From this we note that the sea surface temperature has the greatest slope, hence the “reddest” spectrum and the longest inherent time scales. Figure 3. View largeDownload slide Information about temporal power spectra. In the first three panels, spectral slopes are indicated as, e.g. “f-2” and the logarithmic vertical axis is spectral density of temperature (i.e. temperature variance divided by frequency increment) and the horizontal logarithmic axis is frequency. (a) A spectrum of atmospheric temperature in England, redrawn after (NAS, 1975). This is presented because an unusually long time series allows spectral estimates at unusually low frequencies. (b) A spectrum of ocean sea level estimated using a combination of instrumental and geological data (J. Imbrie, pers. comm.). (c) A spectrum estimated using long-term data from ice cores (an air temperature proxy) merged with higher frequency information from continental records, redrawn after (Pelletier,1998). (d) Vertical axis: the negative coefficients of power spectral slope (i.e. the exponent of frequency) for three regions obtained from near-surface air temperature data, redrawn after (Vasseur and Yodzis, 2004). On the horizontal “axis”, three different environments are labelled. From this we note that the sea surface temperature has the greatest slope, hence the “reddest” spectrum and the longest inherent time scales. In a classic paper, Hasselmann (1976) developed stochastic climate models to show how the nonlinearities in the combined ocean-atmosphere system converted short-term “weather” forcing into long-term ocean responses with red spectra, where variance decreases with increasing frequency. To get data on spectra for time scales from days to millennia, one has to patch together instrumental and geophysical records (Figure 3). The reconstructed long time series (Figure 3a–c), using temperature, sea level, and geophysical data and eliminating regular solar or lunar cycles, confirm Hasselman’s insights and show the terrestrial environment with a near “white” power spectrum on scales from days to centuries, whereas the ocean is “red” (Figure 3). “White” spectra have relatively equal energy at all frequencies, whereas for “red” spectra, the energy increases with decreasing frequency. An analysis (Figure 3d) by Vasseur and Yodzis (2004) of surface air temperatures over sea, inland, and at the coast at ecological time scales confirms this general pattern in the slope (-β) of the power spectra and shows that intermediate coastal regions have more variable responses that span the land-ocean range. Specifically, the spectra from over the ocean have the steepest slopes, hence longest time scales, and records from over land have the flattest spectra, hence the shortest time scales. The large range of intermediate values in coastal regions is caused in part by the fact that, in water of less than about 200 m depth, the open ocean definition of space-time scales (Figure 1) and of power spectra (Figure 3b) have to be modified. Generally, these data compilations confirm Hasselmann’s (1976) intuition about the role of ocean-atmosphere interactions in transforming the variability in the atmospheric drivers to an ocean-based system with very great millennial variance and, relatively, minimal variability at yearly to decadal scales. It is this transformation in scales that is central to our argument about the ecological and evolutionary consequences. Figure 4. View largeDownload slide Schematic representation of the relationship between atmospheric and marine spectra with major food chain components indicated. Figure 4. View largeDownload slide Schematic representation of the relationship between atmospheric and marine spectra with major food chain components indicated. According to Pelletier (1998), at time scales longer than about 1 000 years, the atmosphere and the oceans are in thermal equilibrium, so that these two sectors will have the same variance. Therefore, at millennial scales, the land and sea form a single system. A schematic presentation of the combined spectra and their relation to ecological time scales (Figure 4) summarizes the effects of the transition to land. What are the implications of these differences in the magnitude and time scales of variability for ecological processes? The critical early life stages of nearly all marine species and of most terrestrial animals have time scales from days to a year. Based on the differences in the oceanic and atmospheric temperature spectra, at these time scales, the variability of the ocean environment can be two to three orders of magnitude less than that on land. At these same scales, the spectral slope of the ocean environment is much steeper than that on land. This spectral slope means that day-to-day variability is more dominant (compared to lower-frequency changes) over land than in the ocean, where the longer time scales dominate. Different variables, such as winds/currents, may have different spectral properties than temperature, but we anticipate that generally the oceanic spectrum will remain “redder”. Given the differences in variability and its time scales, then, it is not surprising that the trade-offs in how animals and plants have evolved to ensure the survival of their offspring in relation to terrestrial and marine environmental variability can be so different. Predictably of currents in oceans vs. atmospheres As well as the differences in temporal variance, in the ocean there are solid boundaries to constrain the currents. In consequence, instead of the atmospheric fronts and high or low pressure systems sweeping erratically across the continents, in the ocean we see predictable current systems at a wide range of scales, from the oceanic gyres (Talley et al., 2001) of the Gulf Stream in the North Atlantic and the Kuroshio in the North Pacific, to the anti-clockwise circulation in the North Sea driven by the Baltic outflow (Hill, 1998) on the east side and Atlantic inflow to the west (Rohde, 1998); and, at smaller scales, the tidal motions that can generate clockwise flow around fishing grounds such as Georges Bank off New England (Simpson, 1998). None of these current systems is “as regular as clockwork”, but it is not merely we humans that recognize the basic patterns, so do many animals that use them as a critical part of their reproductive cycle. North American and European eels (Anguilla rostrata and A. Anguilla, respectively) migrate to the Sargasso Sea to reproduce, and their larvae are then carried back to the continent of origin by physical transports that we still do not fully understand (Tesch, 2003). In the North Sea, herring (Clupea harengus) lay their eggs on gravel beds off the Orkney and Shetland Islands. The larvae released by these eggs are carried southeast by currents across the North Sea basin to their nursery grounds off Germany and Denmark (Corten, 1986). Cod (Gadus morhua) and haddock (Melanogrammus aeglefinus) spawn at the northeast corner of Georges Bank, with each female releasing hundreds of thousands of eggs. The tides around the Bank produce a clockwise residual current trapped against the sides of the Bank; a current that is rich in plankton. The eggs become larvae and feed in this current while being carried to the southwest corner, where they metamorphose to begin life as fish on the seabed of the Bank (Smith and Morse, 1985). It is important to recognize that the physical processes are quite different in these three examples. The common factor is the passive but directed transport of larvae as part of the three life cycles. These examples illustrate the general dependence of the early life strategies on physical processes, for the great preponderance of marine animals. But also, generally and in each case given here, there is significant inter-annual variation in the recruitment of the larvae to the adult population. So nearly all highly fecund fish species have multiple annual spawning, and often prolonged spawning within each season (Hedgecock and Pudovkin, 2011). Red to pink noise: oceans compared to coastal/fresh water But ecology seldom or never has all-or-nothing solutions to environmental challenges. This concern is expressed in the concept of trade-offs between different traits (Stearns, 1992). Evolution can be seen as a search for loopholes (Bakun and Broad, 2003) in any existing ecosystem—in particular, to find the most efficient reproductive strategy as a function of energy intake. Compared to the marine, the very different terrestrial environment involved new factors such as gravity and processes such as desiccation in addition to the changing scales form oceanic to atmospheric dynamics. It is difficult to compare energy balances of marine organisms with completely land-based, air-breathing animals. However, freshwater ponds and rivers are subject to the “white noise” variability of the land (Figure 3d, Vasseur and Yodzis, 2004). Thus, a comparison of the extent of parental care and dependence on physical current systems for “similar” marine and freshwater aquatic animals can partially resolve this issue. Furthermore, many if not most of the transitions to land occurred from brackish or freshwater organisms (Gordon and Olson, 1995; Petit and Hampe, 2006). Oceans At one extreme are cod, which release millions of tiny eggs providing very little stored food in each egg and relying on numbers to propagate the species. At the other extreme are the cartilaginous fish (sharks), which are oviparous or viviparous, producing 2–100 offspring in each reproductive cycle. Then, in between, are viviparous redfish (Sebastes spp.), which release a few tens of thousands of fully formed young. However, the predominant response of marine fish families has been to adopt a “hands-off” approach, with 85% being broadcast spawners (Winemiller and Rose, 1992). In comparison, families in freshwater display a much greater diversity of reproductive strategies, showing that the overriding factor is transition from the regularities of ocean systems, rather than just the adaptation to air. The distinction is most evident in the life cycles of catadromous and anadromous fish. The former, such as eels, feature a pelagic larval stage, whereas salmonids, such as salmon (Salmo salar), lay very large eggs in nests on the upper reaches of rivers. There are similar descriptions for invertebrates (Thorson, 1950; Marshall et al., 2012). The free-swimming crustaceans that dominate the plankton obviously have eggs and nauplii that are at the mercy of the currents, but the offspring of bottom-living crustaceans such as lobsters (e.g. Homarus americanus), which carry their eggs for months, have also a later pelagic phase where the young molt and grow before settling back on the seabed. Again, there are exceptions. Thorson (1950) proposed that high-latitude benthos had fewer planktonic larvae, but later work by Marshall et al. (2012) showed that this applied in the Antarctic but not the Arctic. Marshall and Thorson look for explanations in terms of productivity and temperature. Could the difference be explained in terms of the current patterns? Circumpolar currents in the Antarctic have no return path at the regional scale and so are comparable to atmospheric systems. Coastal Estuaries provide a great diversity of physical environments and current systems, forming interfaces between freshwater and marine environments. Highly fecund oysters and other estuarine shellfish use a strategy of multiple spawns of millions of eggs, which is believed to be a way to play the environmental lottery (North et al., 2008). But the main feature is larval vertical migration over a few tens of meters, to use vertical gradients in currents to counteract the kilometers of net seaward transport and ensure long-term residence (Carriker, 1951; North et al., 2008). This widespread dependence by broadcast eggs and larvae on estuarine currents to retain larvae where there is good habitat forms a further illustration of the use of “auxiliary” energy sources for marine reproductive strategies. Freshwater Mussels provide an interesting contrast between marine and freshwater life cycles. The widespread blue mussel (Mytilus edulis) lives on rocky marine shorelines and is a broadcast spawner. Pearly mussels (Unionacea), which live in rivers and lakes, have what could be considered a bizarre life cycle. They have a specialized larva, the glochidium, which is a parasite of the gills of fish where they can derive some nutrients (Strayer et al., 2004). We suggest that what looks bizarre once considered in the context of this essay is merely an adaption to stop larvae from being flushed out of the one-way river system and an evolved behaviour to serve the role of retention of larvae in good habitat (as described in the “Predictably of currents in oceans vs. atmospheres” section) Evolutionary constraints within “white noise” terrestrial environments Vertebrates There is a predominant pattern for the early life phases of marine organisms that couples ecological behaviour to physical regularities in the chosen environment; patterns that are absent in closely related freshwater species. We must presume that the rewards (larvae survival) outweigh the risks that arise from the variability associated with these physical systems. When we turn to the land, the trend has been in the other direction; a decoupling from the use of the physical environment. As they passed from a relatively slowly changing temperature regime to one that could change rapidly and unpredictably, some vertebrate groups evolved from cold- to warm-blooded, accepting the greater energy costs. But the greatest developments were in the reproductive cycles, with the more recently evolved vertebrate class (mammals) becoming predominantly viviparous. Many invertebrates still produce very large numbers of eggs, but the general trend has been towards smaller numbers of offspring that are given increasing care, feeding, and protection in nests, culminating in vertebrate egg-laying as a strategy to decrease environmental exposure (Royle et al., 2012). Invertebrates Whatever the details of the strategy for any particular terrestrial species, the common factors involve the expenditure of significant energy and resources on the separation of the offspring from the large and unpredictable fluctuations in their environment. In contrast to viviparous reproduction for warm-blooded vertebrates, many of the invertebrate strategies involve the use of “nests”. Furthermore, these structures often require cooperative efforts by a few to hundreds or thousands of other members of the species. Wilson (2012), in “The Social Conquest of the Earth”, has described the wonderful variety of communal structure and behaviour of the ants he has studied and has concluded that nest building is the defining feature of the social ant communities that achieved a level of organization comparable to humans, and some millions of years earlier. In the sea, the crustaceans are a dominant component of the food web. Copepods are a critical link between phytoplankton and fish. Furthermore, they evolved to living on and in the seabed in a way that seems pre-adapted to life on land. But there is no evidence of complex communities, or of communal nests. Nearly all the marine arthropods use a planktonic phase as an essential part of their life cycle, whereas, according to Wilson, on land the evolution of social systems and protected reproductive cycles went hand-in-hand. It is this combination that produced what Wilson calls “eusociality”, where groups contain multiple generations and are prone to performing altruistic acts as part of their division of labour. According to Wilson (2012), “all animal species that have attained eusociality, without exception, at first built nests that they defended against enemies”. We would suggest that these nests were first built as defense against the very variable and unpredictable environment. The more complex structures achieved greater isolation from the external conditions but required more division of labour and more social interaction. Contrast that with the yearly spawning aggregations of cod or haddock. There is quite a lot of “behaviour” in the male-female interactions and inherited “knowledge” of the spawning sites, but no real social patterns and certainly no parental assistance. Between that and the social insects, there is a range of intermediate solutions described by Wilson (2012), from a snapping shrimp species, which has a queen and workers inhabiting sponges, to birds and crocodilians, where the young leave the nest when they mature and disperse to breed and build nests on their own. Wilson (2012) regards dispersal as a barrier to eusociality. However, if, in the sea, egg or larval dispersal provides a more successful life cycle (i.e. higher survival of both larvae and adults—possibly from use of less energy in parenting) for a predominant number of species, then this difference in survival in the “redder” vs. “whiter” environments, is the critical division for responses to the environments on land and in the sea. This difference in fitness is the trade-off between the advantages of pelagic egg and larval transport and of complex parental involvement in long-term evolution. Plants Another challenge concerned the response of plants to the large-scale variability in the terrestrial atmosphere compared to the sea. We still do not appear to know as much as we would like about the early adaptations to air and the acquisition of nutrients (Gensel, 2008). But the later developments indicate two strategies. The first expansion came with the evolution of seed-bearing trees and perennial grasses. One solution to the problems posed by the vagaries of weather was to produce large numbers of wind-borne seeds over an extended life time on a very much longer, almost climatic (>100 s of years), time scale. There is an obvious comparison to the early life stages of fish, particularly gadoids. Both trees and fish have very high mortality of the early stages, followed by a long maturity with low death rates (Figure 5). Thus, some features of the predominant marine adaptation can be transferred to land by greatly increasing the lifespan and so averaging out the annual to decadal weather variability (Petit and Hampe, 2006). But the greatest development occurred just over 100 million years ago. The evolution of flowering plants transferred the problem of dispersal of pollen from the wind to insects as an energetically acceptable solution encouraging great diversity in both components. But the time scales have now become a problem. At climatic scales of centuries, the dispersion rates for deciduous trees of less than a kilometer per year (Davis and Shaw, 2001) could cope with post-glacial poleward spread but would be inadequate to meet the predicted rates of rapid climate change resulting from present human activities. The importance of why we need to appreciate the differences in evolutionary adaption between the regimes in terms of environmental variability and rapid climate changes is taken up in the “Comparative study: rapid climatic change” section. Figure 5. View largeDownload slide Schematic representation of survivorship curves for fish, trees and humans. Adapted from Petit and Hampe (2006) and Kinlan and Gaines (2003). Figure 5. View largeDownload slide Schematic representation of survivorship curves for fish, trees and humans. Adapted from Petit and Hampe (2006) and Kinlan and Gaines (2003). Comparative study: rapid climatic change As we move into an unprecedented period of rapid climate change (Pearce, 2007) it is important to more fully understand how the marine and terrestrial regimes, their ecosystems and species, have evolved differently via the expected environmental variability. There is at least modelling evidence that those regimes with environmental variability in the “red” noise spectrum may face increasing extinction risk with rapid climate change (Mustin et al., 2013). The more accurately we can predict how species and ecosystems will react to climate change, the better we can use our ability to adapt management practices for our land and sea resources. This section provides a few examples of what the temporal and spatial differences may be in terrestrial vs. marine systems. Forests (land/white noise) The most recent and therefore most accessible large shift in climate was the recession from the last ice age that had its maximum about 18 000 years ago. The changes in forests in northeastern North America have been well documented by Margaret Davis and others (Davis, 1981; Davis and Shaw, 2001). The gradual northward move in the ice edge is followed fairly closely by the expansion of the hardwoods. For the future the question is how these forests (or their remnants) would respond to further climatic change induced by human activities. Davis and Zabinski (1994) simulated this by using a climate model projection as the basis for a distribution of one common evergreen, hemlock (Conium maculatum). In more recent studies, Davis and Shaw (2001) have emphasized the differences between the predictions of northward movement of temperature contours and estimates of spreading rate for tree communities. They conclude that current climate projections for the 21st century necessitate range shifts of 300 to 500 km per century, in contrast to commonly observed migration rates in the past of 20 to 40 km per century. Even the exceptional examples from the fossil record, of 100 to 150 km per century, are far below the rates required to track climate changes in the future. Therefore, the evolution of forest systems towards the use of long time scales to integrate higher environmental variability may not be able to respond to rapid climate change as it happens. Pelagic fish (oceans/red noise) These rates of change and their corresponding distributions can be compared to the temporal/spatial patterns observed for dominant pelagic fish, such as anchovy (e.g. Engraulis encrasicolus) and sardine (e.g. Sardina pilchardus). The populations of these species have roughly cyclical changes in abundance with a period of 40–50 years, and correspondingly large amplitude changes in distribution with increases of 20–40 degrees latitude and longitude (2 000–3 000 km), (Chavez et al., 2003). The 30-year expansion in range of the California sardine (Sardinops sagax caerulea) is therefore comparable to the 10 000-year change in hardwoods. Managed fishing effort, such as targeted fishing of top predators can also lead to major increases in prey species abundance that can result in large extensions in their distribution. The very marked increase in abundance of sand lance (Ammodytes americanus) in the northwest Atlantic is an example of a fishery-induced change (Sherman et al., 1981; Steele, 2012). But, as with the sardine and anchovy examples, the change in abundance corresponded to a considerable spatial expansion in distribution close to the latitudinal scale of the millennial change in forests, but occurring in just a few years. Therefore, marine systems seem capable of more rapid, but also more dramatic whole regime shift responses (Steele, 1998). Reef systems (coastal/pink noise) However, it is not a complete land-sea dichotomy as there are the coral reef ecosystems, with the symbiosis between very long-lived animal systems and microscopic plants imposing almost “terrestrial” time scales for coral grazers. Reefs have several terrestrial features: fine-structured solid substrate, long-lived basic food supply to fish, coastal rather than purely marine physical air-sea environment. Coastal environments have spectral exponents intermediate to slopes to ocean vs. land (Figure 3D, Vasseur and Yodzis, 2004). Nevertheless, almost all coral reef fish have a pelagic larval phase and formed the basis for the lottery hypothesis (Sale, 1978). Almany et al. (2007) indicates that dispersion from and return to the reef is more limited than previously supposed, but overall the trade-offs in favour of using “auxiliary energy” appear to hold generally for coral reefs. These constraints may be the basis for reefs’ being diversity “hot spots” comparable to tropical forest systems, and potentially subject to the same concerns about the negative consequences of rapid climatic change (Burrows et al., 2011). Coda We hope we have shown how life cycles on land and sea are related not only to the divergent static properties of the two regimes but depend on the quite distinctive general responses to environmental variability. The very long time and short spatial scales of landscape provide a trade-off against the relatively short time and long space scales encountered on emerging from the sea. These “solid” scales on land provide opportunities for long life spans for many plants, and so determine the terrestrial insulation from shorter-term environmental fluctuation. Together, these scales promote functional diversity (Petchey and Gaston, 2002) on land as the basis for sustainability. In contrast, the relatively short time scales of marine responses allow for regime shifts (Scheffer, 2009) in the sea and encourage adaptability (Steele, 1998). Sustainability and adaptability are the ecological processes at very different time scales imposed in each regime in response to their combination of contrasts in environmental variability and static properties (Levin and Lubchenco, 2008). The story in this essay has followed the divergent development of marine ecosystems that used the auxiliary energy in ocean currents as a critical component of their early life cycle, and the regularities in the physical systems to determine their spatial patterns, as compared to terrestrial systems that followed the alternative path of developing life cycles that minimized the irregularities in their environment by length of life, internal temperature control, or external housing. The persistence of both systems over millions of years is an indication of their relative stability under their very different constraints. The recent utilization of fossil fuels by humans has allowed terrestrial systems to break the rules inherent in our evolutionary past and exceed the limitations that these imposed. It is ironic that the solution to our present “climate change” problems is to develop terrestrial ecosystems that rely on energy derived from unpredictable (wind) and relatively rapid, temperature changing sources (solar) and also from the most regular potential sources of energy (tides and ocean currents), thereby finally combining the options that all of our evolutionary ancestors, marine and terrestrial, have used. Let’s hope it works. Footnotes This article arose from a work in progress with the co-authors when Prof John Steele passed away. It was produced from material in early drafts, copious notes from meetings and discussions such that this work was re-constructed using mostly John's original voice on an issue that he very much wanted people to contemplate and deliberate. Acknowledgements The authors are very grateful to Simon Levin of Princeton University for his generous guidance in the development of this manuscript and his many insightful comments and suggestions. We would also like to thank two anonymous reviewers for helpful comments. Andy Beet and Mary Schumacher of the Marine Policy Center at WHOI helped considerably in preparation of the paper. Funding JHS and KHB acknowledge support from the National Science Foundation, Biological Oceanography section, through grant OCE-1258667. References Almany G. R. , Berumen M. L. , Thorrold S. R. , Planes S. , Jones G. P. 2007 . Local replenishment of fish populations in marine reserve . Science , 316 : 742 – 744 . Google Scholar Crossref Search ADS PubMed Bakun A. , Broad K. 2003 . Environmental ‘loopholes’ and fish population dynamics: comparative pattern recognition with focus on El Niño in the Pacific . Fisheries Oceanog , 12 : 458 – 473 . Google Scholar Crossref Search ADS Baylis J. R. 1981 . The evolution of parental care in fishes with reference to Darwin’s rule of male sexual selection . Environmental Biology of Fishes , 6 : 223 – 251 . Google Scholar Crossref Search ADS Bonte D. , Saastamoinen M. 2012 . Dispersal syndromes in butterflies and spiders. In Dispersal Ecology and Evolution , Chapter 13. Ed. by Colbert et al. Oxford University Press , Oxford . Burrows M. T. , Schoeman D. S. , Buckley L. B. , Moore P. , Poloczanska E. S. , Brander K. M. , Brown C. et al. 2011 . The pace of shifting climate in marine and terrestrial ecosystems . Science , 334 : 652 – 655 . Google Scholar Crossref Search ADS PubMed Carriker M. R. 1951 . Ecological observations on the distribution of oyster larvae in New Jersey estuaries . Ecological Monographs , 21 : 19 – 38 . Google Scholar Crossref Search ADS Chavez F. , Ryan J. , Lluch-Cota S. , Niquen M. 2003 . From anchovies to sardines and back: multidecadal change in the Pacific Ocean . Science , 299 : 217 – 221 . Google Scholar Crossref Search ADS PubMed Clark W. C. 1985 . Scales of climatic impacts . Climatic Change , 7 : 5 – 27 . Google Scholar Crossref Search ADS Corten A. 1986 . On the cause of recruitment failure of herrings in the central and northern North Sea in the years 1972–78 . Journal du Conseil International pour l'Exploration de la Mer , 42 : 281 – 294 . Google Scholar Crossref Search ADS Davis M. B. 1981 . Quaternary history and the stability of forest communities. In Forest Succession: Concepts and Applications , pp. 132 – 153 . Ed. by West Shugart , Botkin . Springer-Verlag , New York . Davis M. B. , Zabinski C. 1994 . Changes in geographical range resulting from greenhouse warming: effects on biodiversity in forests. In Global Warming and Biological Diversity , pp. 297 – 308 . Ed. by Peters , Lovejoy . Yale University Press , New Haven . Davis M. B. , Shaw R. G. 2001 . Range shifts and adaptive responses to quaternary climate change . Science , 292 : 673 – 679 . Google Scholar Crossref Search ADS PubMed Delcourt H. R. , Delcourt P. A. , Webb T. 1982 . Dynamic plant ecology: the spectrum of vegetational change in space and time . Quaternary Science Reviews , 1 : 153 – 175 . Google Scholar Crossref Search ADS Denny M. W. 1990 . Terrestrial versus aquatic biology: the medium and its message . American Zoologist , 30 : 111 – 121 . Google Scholar Crossref Search ADS Fortey R. 1999 Life . Vintage Books , New York . ISBN: 9780375702617 Gensel P. G. 2008 . The earliest land plants . Annual Review of Ecology, Evolution, and Systematics , 39 : 459 – 477 . Google Scholar Crossref Search ADS Gill A. E. 1982 . Atmosphere-Ocean Dynamics. 30, International Geophysics Series . Academic Press , New York . 662 pp. Gordon M. , Olson E. 1995 . Invasions of the Land . Columbia University Press , New York . ISBN: 9780231068765 Hasselmann K. 1976 . Stochastic climate models, part 1: theory . Tellus , 28 : 473 – 485 . Google Scholar Crossref Search ADS Hedgecock D. , Pudovkin A. 2011 . Sweepstakes reproductive success in highly fecund marine fish and shellfish . Bulletin of Marine Science , 87 : 971 – 1002 . Google Scholar Crossref Search ADS Hill A. E. 1998 . Buoyancy effects in coastal and shelf seas. In The Sea, Volume 10: The Global Coastal Ocean Processes and Methods , pp. 21 – 62 . Ed. by Brink K. H. , Robinson A. R. . J. Wiley & Sons, Inc ., New York . ISBN: 0-471-11544-4 Kerr R. A. 1987 . Milankovitch climate cycles through the ages . Science , 235 : 973 – 974 . Google Scholar Crossref Search ADS PubMed Kinlan B. P. , Gaines S. D. 2003 . Propagule dispersal in marine and terrestrial environments . Ecology , 84 : 2007 – 2020 . Google Scholar Crossref Search ADS Lenton T. , Watson A. 2011 . Revolutions that Made the Earth . Oxford University Press . https://doi.org/10.1093/acprof: oso/9780199587049.001.0001 Levin S. A. , Lubchenco J. 2008 . Resilience, robustness, and marine ecosystem-based management . Bioscience , 58 : 27 – 32 . Google Scholar Crossref Search ADS Loxdale H. U. G. H. D. , Hardie J. I. M. , Halbert S. U. S. A. N. , Foottit R. O. B. E. R. T. , Kidd N. E. I. L. A. C. , Carter C. L. I. V. E. I. 1993 . The relative importance of short- and long-range movement of flying aphids . Biological Reviews , 68 : 291 – 311 . Google Scholar Crossref Search ADS Mann K. , Lazier J. 1991 . Dynamics of Marine Ecosystems . Blackwell , Oxford . ISBN: 9780865420823 Marshall D. J. , Krug P. J. , Kupriyanova E. , Byrne M. , Emlet R. 2012 . The biogeography of marine invertebrate life histories . Annual Review of Ecology, Evolution, and Systematics , 43 : 97 – 114 . Google Scholar Crossref Search ADS McFall-Ngai M. J. , Manahan D. T. 1990 . Concepts of adaptation in aquatic animals: deviations from the terrestrial paradigm . American Zoologist , 30 : 109 . Google Scholar Crossref Search ADS Mustin K. , Dytham C. , Benton T. G. , Travis J. M. J. 2013 . Red noise increases extinction risk during rapid climate change . Diversity and Distributions , 19 : 815 – 824 . Google Scholar Crossref Search ADS North E. W. , Schlag Z. , Hood R. R. , Li M. , Zhong L. , Gross T. , Kennedy V. S. 2008 . Vertical swimming behavior influences the dispersal of simulated oyster larvae in a coupled particle-tracking and hydrodynamic model of Chesapeake Bay . Marine Ecology Progress Series , 359 : 99 – 115 . Google Scholar Crossref Search ADS NAS . 1975 . Understanding Climatic Change: A program for action . National Academy Press, Washington DC . Pearce F. 2007 . The Last Generation: How Nature Will Take Her Revenge for Climate Change . Eden Project Books , London . Pelletier J. D. 1998 . The power spectral density of atmospheric temperature from time scales of 10-2 to 106 years . Earth and Planetary Science Letters , 158 : 157 – 164 . Google Scholar Crossref Search ADS Petchey O. , Gaston K. 2002 . Functional diversity, species richness and community composition . Ecology Letters , 5 : 402 – 411 . Google Scholar Crossref Search ADS Petit R. J. , Hampe A. 2006 . Some evolutionary consequences of being a tree . Annual Review of Ecology, Evolution, and Systematics , 37 : 187 – 214 . Google Scholar Crossref Search ADS Richter C. J. 1970 . Aerial dispersal in relation to habitat in eight wolf spider species . Oecologia , 5 : 200 – 214 . Google Scholar Crossref Search ADS PubMed Rohde J. 1998 . The Baltic and North Seas: a process-oriented review of the physical oceanography, In The Sea, Volume 11: The Global Coastal Ocean: Regional Studies and Syntheses , pp. 699 – 732 . Ed. by Robinson A. R. , Brink K. H. . J. Wiley & Sons, Inc ., New York . ISBN 0-471-11545-2 Royle N. J. , Smiseth P. T. , Köllike M. 2012 . The evolution of parental care. In Patterns of Parental Care in Invertebrates , Chapter 5, Oxford Press . doi: 10.1093/acprof: oso/9780199692576.001.0001, ISBN-13: 9780199692576 Sale P. F. 1978 . Coexistence of coral reef fishes—a lottery for living space . Environmental Biology of Fishes , 3 : 85 – 102 . Google Scholar Crossref Search ADS Scheffer M. 2009 . Critical Transitions in Nature and Society . Princeton University Press , Princeton . Sherman K. , Jones C. , Sullivan L. , Smith J. , Berrien P. , Ejsymont I. 1981 . Congruent shifts in sand eel abundance in western and eastern North Atlantic ecosystems . Nature , 291 : 486 – 489 . Google Scholar Crossref Search ADS Simpson J. H. 1998 . Tidal processes in shelf seas. In The Sea, Volume 10: The Global Coastal Ocean Processes and Methods , pp. 113 – 150 . Ed. by Brink K. H. , Robinson A. R. . J. Wiley & Sons, Inc ., New York . ISBN: 0-471-11544-4 Smith W. G. , Morse W. W. 1985 . Retention of larval haddock Melanogrammus aeglefinus in the Georges Bank region, a gyre influenced spawning area . Marine Ecological Progressive Series , 21 : 1 – 13 . Google Scholar Crossref Search ADS Steele J. H. 1978 . Some comments on plankton patches. In Spatial Pattern in Plankton Communities , pp. 1 – 20 . Ed. by Steele J. H. . Plenum Press , New York . Steele J. H. 1998 . Regime shifts in marine ecosystems . Ecological Applications , 8 : S33 – S36 . Google Scholar Crossref Search ADS Steele J. H. 2012 . Prediction, scenarios and insight: the uses of an end-to-end model . Progress in Oceanography , 102 : 67 – 73 . Google Scholar Crossref Search ADS Strathmann R. R. 1990 . Why life histories evolve differently in the sea . American Zoologist , 30 : 197 – 207 . Google Scholar Crossref Search ADS Strayer D. L. , Downing J. A. , Haag R. W. , King T. L. , Layzer J. B. , Newton T. J. , Nichols S. J. 2004 . Changing perspectives on pearly mussels, North America’s most imperiled animals . Bioscience , 55 : 429 – 439 . Google Scholar Crossref Search ADS Talley L. D. , Pickart G. L. , Emery W. J. , Swift J. H. 2001 . Descriptive Physical Oceanography: An Introduction . Academic Press , London , 560 pp. ISBN-13: 978-0750645522, ISBN-10: 0750645520 Tesch F.-W. 2003 . The Eel. Blackwell Science , Oxford, UK . ISBN: 9780632063895 Thorson G. 1950 . Reproductive and larval ecology of marine bottom invertebrates . Biology Reviews , 25 : 1 – 45 . Google Scholar Crossref Search ADS Vasseur D. , Yodzis P. 2004 . The color of environmental noise . Ecology , 85 : 1146 – 1152 . Google Scholar Crossref Search ADS Wilson E. O. 2012 . The Social Conquest of Earth . Norton , New York . ISBN: 9780871403636 Winemiller K. O. , Rose K. A. 1992 . Patterns of life history diversification in North American fishes: implications for population regulation . Canadian Journal of Fisheries and Aquatic Sciences , 49 : 2196 – 2218 . Google Scholar Crossref Search ADS Wourm J. P. , Lombardi J. 1992 . Reflections on the evolution of piscine viviparity . American Zoologist , 32 : 276 – 293 . Google Scholar Crossref Search ADS © International Council for the Exploration of the Sea 2018. 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)

Mat, Audrey, M

doi: 10.1093/icesjms/fsy131pmid: N/A

Abstract Marine habitats are shaped by several geophysical cycles ranging from a few hours (tidal and solar cycles) to a year (seasons). These cycles have favoured the selection of endogenous biological clocks. Such a clock is a molecular time-keeping mechanism that consists of a set of core clock genes whose expression oscillates. The clocks produce biological rhythms and influence virtually all metabolic, physiological, and behavioural functions in organisms. This work highlights the importance to take chronobiology into account in experimental marine biology to avoid faulty results, misinterpretation of results, and/or to strengthen observations and conclusion. A literature survey, based on 150 articles, was conducted and showed that, despite the pervasive imprint of biological rhythms in marine species, environmental cycles such as the 24 h-light/dark cycle and the seasonality are rarely considered in experimental designs. This work emphasizes that better integrating the temporal organization and regulation of marine species within the marine biology community is essential for obtaining representative results. Introduction The Sun, the Moon, and the Earth’s immutable rotation on their orbits deeply influence our living world. The pervasive alternation of nights and days has favoured the selection of an endogenous circadian clock. The genetic basis of the circadian clockwork mechanism was first discovered in fruit flies (Hardin et al., 1990; Sehgal et al., 1995; Allada et al., 1998; Rutila et al., 1998): translation and transcription feedback loops of core clock genes set the tempo for cells, tissues, and ultimately the whole organism (Chaix et al., 2016; Kumar, 2017). Mechanistically, the clock varies between species, but its formal principle is ubiquitous from cyanobacteria, to plants, and animals (Young and Kay, 2001). Although ubiquitous among taxa, biological clocks probably evolved independently at least twice (Rosbash, 2009). The clock is synchronized to the Earth’s 24 h-revolution by external cues called zeitgebers (time-givers), such as the light/dark cycle. Without zeitgeber, the clock free-run at its endogenous period that is circa 24 h, meaning approximately 24 h. Endogenous periods can vary between individuals (Aschoff, 1981; Johnson et al., 2004). The circadian clock drives organisms’ biological rhythms, including the sleep/wake or hormonal cycles (Kumar, 2017). Biological rhythms are considered to be adaptive (Woelfle et al., 2004) in that they allow the anticipation of cyclic environmental changes, and ensure consistency in organisms’ physiology, metabolism, and behaviour (Rosbash, 2009). An example of synchronization and anticipation is the daily rhythm of body temperature in human: the trough occurs at night, it starts rising in anticipation of wakening, reaches a peak in the early evening, and drops in anticipation of sleep (Refinetti and Menaker, 1992). Marine species are influenced not only by environmental cycles associated with the solar day, but also by moon-related environmental cycles, which include the tidal cycle (with period of 12.4 h), the lunar day (24.8 h, the time it takes for the moon to complete an orbit around the earth), and the semi-lunar/lunar cycles (14.8/29.5 days; Tessmar-Raible et al., 2011). The seasons also deeply influence both terrestrial and marine habitats; in Crassostrea gigas, for example, temperature and photoperiod (i.e. the duration of the light phase of a light/dark cycle) regulate oyster annual reproduction (Fabioux et al., 2005). All these cycles make the marine biotope a very complex, yet predictable, cyclic environment. Whereas a deep understanding of the mechanisms underlying the circadian clockwork has been provided in terrestrial species, the timing mechanisms in marine organisms still need to be deciphered (Tessmar-Raible et al., 2011; de la Iglesia and Johnson, 2013). However, endogenous rhythms corresponding to all of these environmental cycles have been described in a variety of marine organisms including annelids (Last et al., 2009), molluscs (Connor and Gracey, 2011), arthropods (Zhang et al., 2013), chordates (Vera et al., 2013), phytoplankton (Bouget et al., 2014), and manifest in many phenotypes including locomotor and feeding activity, as well as metabolism and reproduction. In the laboratory, incubation conditions affect biological rhythms. However, despite the importance of environmental cycles in driving major rhythms of marine organisms, they are often neglected in the design of marine biology experiments. Such a practice may lead to desynchrony within the group under study if no environmental cycle is implemented, add some undesired variance among individuals if irregular sampling is performed, or alter the parameter studied if unnatural incubation conditions are implemented. These can potentially alter all observations. Therefore, the objective of the present study was twofold: (1) to evaluate in which way chronobiological information was mentioned or rather neglected, and (2) to show examples of misinterpretation of results when not considering the biological rhythmicity in marine organisms. Material and methods A literature survey of experimental laboratory work (n = 150 articles; Supplementary Table S1) on marine and brackish species (according to the World Register of Marine Species, http://www.marinespecies.org) was performed, using articles which were published between July and December 2017 in the top 20 journals in the “marine and freshwater biology” category of the 2016 Thomson Reuters Ranking. Articles were selected in the last published issues of the journals at the time of reading, in decreasing ranking. It was reported if and in which way environmental cycles were integrated into the experimental design. As simulating tidal cycles in the laboratory is technically demanding and thus rarely done, they were neglected form the search. In the published articles, to find out how organisms were synchronized to an environmental cycle and at which stage of the cycle they were studied, I focused on the following questions: Were the light conditions mentioned? This include the presence/absence of a day/night cycle, the photoperiod, the light intensity or spectrum, and whether there were abrupt or gradual changes in light intensity at light on and light off. For the light intensity or spectrum, I considered any information including either natural illumination, the type of lamp, the irradiance (photons·m−2·s−1), or the photon flux (lux). Was the temperature mentioned and was it kept constant or cyclic? For animals, was there a regular feeding procedure implemented? Was the time of the year at which the experiment was performed reported, either in terms of month(s) or season(s)? Was the origin of the biological material provided and was the experimental location mentioned? I searched for these elements in the Material and methods section, and scanned the text according to keywords (“light”, “dark”, “cycle”, “photo*” for either photoperiod or photons, “°C”, “temperature”, “fed”, “feed”, “food”, each month and season, “lab*” for lab or laboratory, “center”, “instit*” for either institution or institute, “hous*” for housed). The purpose of this analysis was to assess the integration of biological rhythms within marine biology. Results The results are summarized in Table 1; detailed information is gathered in Supplementary Table S1. Out of the 150 articles reviewed, 32% did not provide information about the presence/absence of a light/dark cycle in the experiment. Another 10% provided only partial information such as for a specific time period during the experiment (i.e. acclimation or incubation), a particular development stage (for adults but not for their progeny) or for a subset of the studied organisms. Information on a diel cycle was rarely missing for plants (13% and 8%) or chromists (2% and 9%) while it was often missing for animals (52% and 9%), revealing a strong difference in treatment between photosynthetic and non-photosynthetic organisms. Of all articles analysed, 52% reported experiments run under a light/dark cycle while 6% of the experiments were explicitly run either under constant light or constant darkness. Only 1 article provided an explicit reason to work under constant conditions. The photoperiod was given in only 47% of the 150 articles: some articles specified the presence of a light/dark cycle but only mentioned a natural photoperiod without detail, for example. The information was not relevant for studies conducted under continuous conditions. The type of light, either in terms of intensity or spectrum and when continuous light or light/dark cycles were implemented, was mentioned in only 43% of the 150 articles considered. Unless implicit when working with outdoor or with natural daylight, the information on whether light on and off were abrupt or gradual was rarely mentioned (7% of the articles). Table 1. Summary of the literature survey. Information Parameter None Partial Full NA Cyclic Constant Light/dark conditions (all phylla, n = 150) 32 10% 58% / 52% 6%  Plants (n = 24) 13% 8% 79% / 75% 4%  Chromists (n = 46) 2% 9% 89% / 78% 11%  Animals (n = 86) 52% 9% 38% / 36% 2% Photoperiod (all phylla, n = 150) 32% 15% 47% 6% / /  Plants (n = 24) 13% 21% 58% 8% / /  Chromists (n = 46) 2% 11% 76% 11% / /  Animals (n = 86) 52% 19% 27% 2% / / Lighting conditions (all phyla, n = 150) 48% 6% 43% 3% / /  Plants (n = 24) 13% 8% 75% 4% / /  Chromists (n = 46) 2% 7% 89% 2% / /  Animals (n = 86) 81% 7% 10% 1% / / Abrupt/gradual change (all phyla, n = 150) 85% 3% 7% 5% / / Temperature cycle (all phyla, n = 150) 20% 13% 67% / 1% 66%  Plants (n = 24) 17% 8% 75% / 4% 71%  Chromists (n = 46) 11% 7% 83% / 0% 83%  Animals (n = 86) 27% 17% 56% / 0% 56% Feeding cycle (all phyla, n = 150) 23% 20% 9% 48% 1% 5%  Plants (n = 24) / / / / / /  Chromists (n = 46) / / / / / /  Animals (n = 86) 40% 34% 14% 13% 8% 1% Time of year (all phyla, n = 150) 70% 11% 19% / / / Experimental facility (all phyla, n = 150) 46% 11% 43% / / / Origin of the biological material (all phyla, n = 150) 4% 5% 91% / / / Information Parameter None Partial Full NA Cyclic Constant Light/dark conditions (all phylla, n = 150) 32 10% 58% / 52% 6%  Plants (n = 24) 13% 8% 79% / 75% 4%  Chromists (n = 46) 2% 9% 89% / 78% 11%  Animals (n = 86) 52% 9% 38% / 36% 2% Photoperiod (all phylla, n = 150) 32% 15% 47% 6% / /  Plants (n = 24) 13% 21% 58% 8% / /  Chromists (n = 46) 2% 11% 76% 11% / /  Animals (n = 86) 52% 19% 27% 2% / / Lighting conditions (all phyla, n = 150) 48% 6% 43% 3% / /  Plants (n = 24) 13% 8% 75% 4% / /  Chromists (n = 46) 2% 7% 89% 2% / /  Animals (n = 86) 81% 7% 10% 1% / / Abrupt/gradual change (all phyla, n = 150) 85% 3% 7% 5% / / Temperature cycle (all phyla, n = 150) 20% 13% 67% / 1% 66%  Plants (n = 24) 17% 8% 75% / 4% 71%  Chromists (n = 46) 11% 7% 83% / 0% 83%  Animals (n = 86) 27% 17% 56% / 0% 56% Feeding cycle (all phyla, n = 150) 23% 20% 9% 48% 1% 5%  Plants (n = 24) / / / / / /  Chromists (n = 46) / / / / / /  Animals (n = 86) 40% 34% 14% 13% 8% 1% Time of year (all phyla, n = 150) 70% 11% 19% / / / Experimental facility (all phyla, n = 150) 46% 11% 43% / / / Origin of the biological material (all phyla, n = 150) 4% 5% 91% / / / Percentages of articles providing information on the presence/absence of a light/dark cycle, the photoperiod, the light conditions (either in terms of intensity or spectrum), the setting of the light/dark change, whether the temperature was constant or cyclic, the feeding procedure and whether it was regular or not, the time of year (either in terms of month(s) or season), the experimental facility used for the experimental work, and the origin of the biological material (collection place or strain). Articles (n = 150) involving the following: plants (n = 24), chromists (n = 46), animals (n = 86), bacteria (n = 4), and fungi (n = 1). Percentages calculated for those n. Some articles may include several experiments and phylla, some run under cyclic conditions, others under constant conditions, and will therefore be counted in both categories, potentially giving a total for a line slightly different than 100%. NA: not applicable. Table 1. Summary of the literature survey. Information Parameter None Partial Full NA Cyclic Constant Light/dark conditions (all phylla, n = 150) 32 10% 58% / 52% 6%  Plants (n = 24) 13% 8% 79% / 75% 4%  Chromists (n = 46) 2% 9% 89% / 78% 11%  Animals (n = 86) 52% 9% 38% / 36% 2% Photoperiod (all phylla, n = 150) 32% 15% 47% 6% / /  Plants (n = 24) 13% 21% 58% 8% / /  Chromists (n = 46) 2% 11% 76% 11% / /  Animals (n = 86) 52% 19% 27% 2% / / Lighting conditions (all phyla, n = 150) 48% 6% 43% 3% / /  Plants (n = 24) 13% 8% 75% 4% / /  Chromists (n = 46) 2% 7% 89% 2% / /  Animals (n = 86) 81% 7% 10% 1% / / Abrupt/gradual change (all phyla, n = 150) 85% 3% 7% 5% / / Temperature cycle (all phyla, n = 150) 20% 13% 67% / 1% 66%  Plants (n = 24) 17% 8% 75% / 4% 71%  Chromists (n = 46) 11% 7% 83% / 0% 83%  Animals (n = 86) 27% 17% 56% / 0% 56% Feeding cycle (all phyla, n = 150) 23% 20% 9% 48% 1% 5%  Plants (n = 24) / / / / / /  Chromists (n = 46) / / / / / /  Animals (n = 86) 40% 34% 14% 13% 8% 1% Time of year (all phyla, n = 150) 70% 11% 19% / / / Experimental facility (all phyla, n = 150) 46% 11% 43% / / / Origin of the biological material (all phyla, n = 150) 4% 5% 91% / / / Information Parameter None Partial Full NA Cyclic Constant Light/dark conditions (all phylla, n = 150) 32 10% 58% / 52% 6%  Plants (n = 24) 13% 8% 79% / 75% 4%  Chromists (n = 46) 2% 9% 89% / 78% 11%  Animals (n = 86) 52% 9% 38% / 36% 2% Photoperiod (all phylla, n = 150) 32% 15% 47% 6% / /  Plants (n = 24) 13% 21% 58% 8% / /  Chromists (n = 46) 2% 11% 76% 11% / /  Animals (n = 86) 52% 19% 27% 2% / / Lighting conditions (all phyla, n = 150) 48% 6% 43% 3% / /  Plants (n = 24) 13% 8% 75% 4% / /  Chromists (n = 46) 2% 7% 89% 2% / /  Animals (n = 86) 81% 7% 10% 1% / / Abrupt/gradual change (all phyla, n = 150) 85% 3% 7% 5% / / Temperature cycle (all phyla, n = 150) 20% 13% 67% / 1% 66%  Plants (n = 24) 17% 8% 75% / 4% 71%  Chromists (n = 46) 11% 7% 83% / 0% 83%  Animals (n = 86) 27% 17% 56% / 0% 56% Feeding cycle (all phyla, n = 150) 23% 20% 9% 48% 1% 5%  Plants (n = 24) / / / / / /  Chromists (n = 46) / / / / / /  Animals (n = 86) 40% 34% 14% 13% 8% 1% Time of year (all phyla, n = 150) 70% 11% 19% / / / Experimental facility (all phyla, n = 150) 46% 11% 43% / / / Origin of the biological material (all phyla, n = 150) 4% 5% 91% / / / Percentages of articles providing information on the presence/absence of a light/dark cycle, the photoperiod, the light conditions (either in terms of intensity or spectrum), the setting of the light/dark change, whether the temperature was constant or cyclic, the feeding procedure and whether it was regular or not, the time of year (either in terms of month(s) or season), the experimental facility used for the experimental work, and the origin of the biological material (collection place or strain). Articles (n = 150) involving the following: plants (n = 24), chromists (n = 46), animals (n = 86), bacteria (n = 4), and fungi (n = 1). Percentages calculated for those n. Some articles may include several experiments and phylla, some run under cyclic conditions, others under constant conditions, and will therefore be counted in both categories, potentially giving a total for a line slightly different than 100%. NA: not applicable. Temperature conditions were mentioned in 67% of the articles. The 20% of articles without information on a temperature cycle are distributed as follow: 6% did not mention any temperature, and 14% did mention either a range of temperature or a mean ± standard deviation/standard error where the latter is equal or superior to 2°C. Without further clarification, it cannot be determined whether there was a planned or unplanned temperature cycle. Again, temperature conditions were less often mentioned in studies conducted on animals (56%) than on plants (75%) or chromists (83%). In the vast majority of cases, the temperature was kept constant, not cyclic. Feeding cycles were considered for animals only. Out of the 86 articles involving animals, 40% did not provide information about the presence/absence of a feeding cycle, another 34% provided only partial information. Most studies that contained partial information on the feeding cycle mentioned that animals were fed either once or twice daily, but did not provide additional information such as: was the feeding implemented at the same time(s) each day? Fourteen percent of the articles provided full information on the feeding procedure, but only those that were on a daily basis were considered as cyclic, not those involving weekly or every other day feeding. The time of year at which the experiment was performed was lacking in 70% of the work reported. Another 11% included only partial information, such as the collection time of animals from the field but not the experimental time or for only part of the experiments. Collection time and experimental time are not necessary the same, as animals might be kept in the lab as broodstock, or for acclimation. Compared to temporal information, spatial information was mostly provided in the literature: the origin of the organisms studied was detailed in the majority of cases (91%). However, the location of where the experiment took place was given in only 43% of the articles analysed (Table 1). Discussion The importance of biological rhythms is valid for potentially all biological functions as biological clocks deeply influence organisms’ physiology and behaviour. In the mouse genome, almost half of all genes show circadian rhythms in transcription in at least one organ (Zhang et al., 2014). Similarly, >40% of Mytilus californianus gill transcriptome is cyclic under light/dark and tidal entrainment (Connor and Gracey, 2011), and ∼23% of the transcripts are rhythmic in the sea anemone Aiptasia diaphana (Sorek et al., 2018). This is valid at all levels of organization, from the molecular to the behavioural level, and can occur even in cells as self-sustained circadian oscillations persisting for over 20 cycles have been reported in isolated tissues of mice (Yoo et al., 2004). The observation has been extended to other organisms and there is potentially a clock in each cell and each tissue (Mohawk et al., 2012). Rhythms are also important for full-length cDNA analysis as differentially spliced mRNA isoforms may show rhythmic oscillations in relative abundance (Preußner et al., 2014). No matter what we study, biological rhythms may thus influence virtually all major biological functions. Environmental cycles and biological rhythms should therefore be taken into consideration in the experimental strategy, as neglecting them may generate several problems. Problem induced when working with desynchronized organisms When an organism possessing a biological clock is kept under free-running conditions, i.e. without zeitgeber, it cycles with its own endogenous period. Due to inter-individual differences in this period, each animal may be in a different phase to its neighbour with obvious implications for variation in the phenotype (Figure 1a). For an experiment that aims at measuring a phenotype, this can artefactually increase the variance and the inter-individual variability. For experiments that aim at studying the effect of a treatment, for example, it might similarly alter reported observations. If the amplitude of the effect analysed is within the range of cyclic variation for the parameter studied, similarities or differences between those organisms could artefactually result from their intrinsic biological rhythm. The sand hopper Talitrus saltator, for example, exhibits a daily locomotor activity rhythm, being active at night. This activity rhythm is under circadian control (Bregazzi and Naylor, 1972). Without zeitgeber, a group of T. saltator won’t be synchronized anymore and at the same time of day, one might have both active and resting animals, depending on their individual endogenous time. This would have an effect on the measure of locomotor activity. Rhythms are not only important for the absolute value of what we measure, but also for the nature of the measure itself. Indeed, they also influence how organisms cope with the same circumstances, either favourable or unfavourable, at different times of the day or seasons. For example, olfactory responses to food-related odours and pheromones in the cockroach Leucophea maderae is under circadian control, with a 5–10-fold change in sensitivity to food-related odours between night and day (Page and Koelling, 2003; Rymer et al., 2007). One might similarly expect clock-regulated responses in the marine environment and these rhythmic changes would again influence our observations. It is thus essential to work with synchronized organisms as working with desynchronized ones might highly disturb the parameter analysed, rendering it potentially unusable. Figure 1. View largeDownload slide (a) Shift of a biological parameter under circadian control for 3 organisms whose internal period are 22h (short dotted line), 24h (continuous line), and 26h (long dotted line), respectively, over a 5-day experiment under free-running conditions. The desynchronization would involve different values for the parameter studied, even if organisms are sampled at the same time. (b) Effect of sampling at different times over the diel cycle in synchronized organisms. The value of the parameter studied would here be 0.9, 0.5, 0, −0.5, or −0.9 AU for sampling performed at 8, 10, 12, 14, or 16h, respectively. AU: arbitrary units. Figure 1. View largeDownload slide (a) Shift of a biological parameter under circadian control for 3 organisms whose internal period are 22h (short dotted line), 24h (continuous line), and 26h (long dotted line), respectively, over a 5-day experiment under free-running conditions. The desynchronization would involve different values for the parameter studied, even if organisms are sampled at the same time. (b) Effect of sampling at different times over the diel cycle in synchronized organisms. The value of the parameter studied would here be 0.9, 0.5, 0, −0.5, or −0.9 AU for sampling performed at 8, 10, 12, 14, or 16h, respectively. AU: arbitrary units. Problems occurring when ignoring the influence of entrainment in experimental design The influence of entrainment should also be carefully considered in the experimental design as omitting it can introduce two types of bias. First, it should be included in the sampling strategy; rhythmic changes present in all treatments could be interpreted as differences between treatments if comparing samples collected at different times (Figure 1b). Second, the technical setup and maintenance of an experiment also needs to be considered within a cyclic frame. Besides light, temperature and food are known potential and powerful zeitgebers for the circadian and circatidal clocks. If the temperature in the experimental setting is intended to be constant and measured once a day at the same time, the parameter may indeed appear constant over the course of the experiment while there might actually be an undesired daily cycle. While the controlled chambers or continuous recording inherently overcome this bias, manual measures should be planned at different phases of the implemented environmental cycle(s). Similarly, any maintenance operation like feeding that is operated daily should occur at the same time(s), both during the week and the weekend. It is also globally important to avoid unplanned entrainment like turning the light on during the dark phase, or an unmonitored change of the temperature of the water because zeitgebers synchronize the clock(s). In both nocturnal and diurnal organisms, light applied during the dark phase can reset the circadian clock and advance or delay the observed rhythm, depending on its administration time (Johnson et al., 2004). In flying squirrels, a 1-s light pulse is sufficient to provide proper photoentrainment (Johnson et al., 2004). Marine species may similarly show great sensitivity to different environmental cycles and proper control of experimental conditions is essential to gather relevant results. Problems occurring when working with unnatural conditions Beyond their constant or cyclic aspect, biological rhythms and clocks are affected by environmental and laboratory conditions. In the seabream Sparus aurata, mealtime determined whether locomotor activity was diurnal or nocturnal and influenced clock gene expression in the liver, while clock gene expression in the brain was determined by the light/dark cycle (Vera et al., 2013). The light intensity and spectrum are also relevant as they can influence organisms’ physiology or behaviour. For example, Nephrops norvegicus lobsters exposed to light/dark cycles showed a nocturnal burrow emergence activity under 10 lux, but a diurnal one under 0.1 lux (Chiesa et al., 2010). Additionally, N. norvegicus’ eyes are very sensitive to light-induced damage (Gaten et al., 2013); in the laboratory, using a light intensity consistent with the animal’s natural environment and working with progressive lights on and off allows avoidance of eye damage (Sbragaglia et al., 2013). Photoperiod and temperature also provide temporal information to organisms on an annual scale, and seasons influence organisms’ morphology, physiology, and behaviour (Helm et al., 2013). They drive the lifecycle of the toxic dinoflagellate Gonyaulax tamarensis (Andersen and Keafer, 1987) and determine the diel valve activity pattern in the oyster Crassostrea gigas that is rather diurnal in spring and summer and nocturnal in autumn and winter (Mat et al., 2012). Seasons also control major life traits like gametogenesis and spawning in several marine species including bivalves (Fabioux et al., 2005), corals (Sorek and Levy, 2014), and worms (Naylor, 2010). They influence the response of daily locomotor activity to temperature changes in the crab Uca pugilator (Mat et al., 2017), and affect the chemical composition in the kelp Eisenia arborea (Landa-Cansigno et al., 2017). The photoperiod, temperature, and time of year therefore have several implications in laboratory experiments. First, an unnatural photoperiod can cause behavioural, physiological, or metabolic changes related to seasonal phenology based on photoperiod measurement. Second, it is also critical to simulate a time that is suitable to investigate the scientific question. Photoperiodic induction of diapause has been reported in the marine copepod Labidocera aestiva (Marcus, 1980). If one wants to study the hatching success of eggs, animals should not be exposed to a photoperiod that triggers diapause. Working with relevant environmental incubation conditions is thus crucial to obtain realistic results. Controlling and informing about the timing of an experiment is also important for comparing experimental results with the existing literature, and to improve reproducibility. Finally, both the origin of the biological material and the experimental facility are important information to understand and compare both the natural and laboratory conditions of the studied species. Organisms’ collection and transfer can strongly affect rhythmicity due to stress, changing conditions, or transport conditions. Guide for future studies The present literature search demonstrates that the consideration of biological rhythms in marine species tend to stay confined to the field of chronobiology and are not yet well integrated into the broader field of marine experimental biology: 32%, 20%, and 23% of the articles analysed did not provide information about the light/dark, temperature, and feeding cycles in the experimental setup, respectively. Similarly, the time of year was not provided in 70% of the manuscripts. The following are suggestions to improve our experimental setups and practices, allowing greater potential for realistic observations and inter-study comparisons: If there is no specific need or relevance for constant darkness or illumination for the experiment (e.g. work on cave species, photoinhibition), organisms should be under a light/dark cycles, either one mimicking the natural condition or a 12:12 light/dark cycle; ideally, using a light intensity and spectrum that are relevant for the studied species. For experiments which run over several weeks or months, mimicking the change in photoperiod would more closely reflect natural conditions. Gradual changes in light intensity would ideally be more realistic than abrupt changes, but are more difficult to implement; one should however be aware that this can influence experimental results. Realistic environmental temperatures should be tightly controlled and monitored for the studied species. This includes frequent monitoring of the both constant or cycling temperatures throughout the various phases of the experimental cycle. Daily feeding should occur at the same time(s), both during the week and the weekend. If the organisms are not fed but kept with running seawater, monitoring chlorophyll a might be one way to control for the absence of an undesired cycle. If several cycles are implemented such as temperature and light/dark cycles for example, they should be consistent with each other. Other environmental cues can also act as zeitgebers for the circadian and circatidal clocks including salinity, pH, and turbulence cycles (Naylor, 2010). The most exhaustive control and monitoring of experimental conditions are therefore globally important to avoid either unwanted cycles or erratic patterns. Under light/dark cycles, sampling per day should be performed at the same time to compare data acquired at the same phase of the cycle. If several cycles are included in the experimental setting, sampling per day should be performed at the most appropriate time to compare data acquired at the same phase of the cycle. Simulating tidal cycles in the laboratory can be challenging, as many parameters can act as a tidal zeitgeber (e.g. salinity, air/water exposure, and temperature) and the organism might be specific in their responsiveness to these parameters. However, tidal cycles could be implemented when possible to reflect more closely the natural conditions marine species encounter in the field. The awareness of biological rhythm should also be extended to field studies, where it can influence sampling time. Sampling at noon on week 1 and 2 correspond to different phases of a tidal cycle as high and low tide drift by ∼48 min every day. Conversely, sampling specifically at low tide might result in sampling occurring both during the day and the night over several weeks. Thoroughly document in the Material and methods section the light conditions (photoperiod, intensity, spectrum, abrupt or gradual light transition), temperature (mean ± standard deviation), feeding protocol if any, the month(s) during which the work was performed, the collection place or strain used, as well as the experimental facility. Consider their potential influence on the obtained results in the Discussion section. In conclusion, integrating the temporal organization and regulation of marine species within the marine biology community is essential for obtaining representative results, strengthening the validity of our observations, and improving reproducibility. Acknowledgements This work was supported by the “Laboratoire d’Excellence” LabexMER (ANR-10-LABX-19) and co-funded by a grant from the French government under the program “Investissements d'Avenir”, and by a grant from the Regional Council of Brittany. A.M. was supported by a LabexMER International Post-doctoral Fellowship. I thank Drs Dominique Cowart, Arnaud Huvet, Delphine Muths, Fabrice Pernet, and Pr. Charalambos P. Kyriacou for their thoughtful comments and advice on this article. I also thank the editor, Dr Browman, and the three anonymous reviewers for their constructive comments that helped strengthen this article. References Allada R. , White N. E. , So W. V. , Hall J. C. , Rosbash M. 1998 . A mutant Drosophila homolog of mammalian clock disrupts circadian rhythms and transcription of period and timeless . Cell , 93 : 791 – 804 . Google Scholar Crossref Search ADS PubMed Andersen D. M. , Keafer B. A. 1987 . An endogenous annual clock in the toxic marine dinoflagellate Gonyaulax tamarensis . Nature , 325 : 616 – 617 . Google Scholar Crossref Search ADS PubMed Aschoff J. 1981 . Freerunning and entrained circadian rhythms. In Biological Rhythms , pp. 81 – 93 . Ed. by Aschoff J. . Springer , Boston, MA . Bouget F.-Y. , Lefranc M. , Thommen Q. , Pfeuty B. , Lozano J.-C. , Schatt P. , Botebol H. et al. 2014 . Transcriptional versus non-transcriptional clocks: a case study in Ostreococcus . Marine Genomics , 14 : 17 – 22 . Google Scholar Crossref Search ADS PubMed Bregazzi P. K. , Naylor E. 1972 . The locomotor activity rhythm of Talitrus Saltator (Montagu) (Crustacea, Amphipoda) . Journal of Experimental Biology , 57 : 375 – 391 . Chaix A. , Zarrinpar A. , Panda S. 2016 . The circadian coordination of cell biology . The Journal of Cell Biology , 215 : 15 – 25 . Google Scholar Crossref Search ADS PubMed Chiesa J. J. , Aguzzi J. , García J. A. , Sardà F. , de la Iglesia H. O. 2010 . Light intensity determines temporal niche switching of behavioral activity in deep-water Nephrops norvegicus (Crustacea: Decapoda) . Journal of Biological Rhythms , 25 : 277 – 287 . Google Scholar Crossref Search ADS PubMed Connor K. M. , Gracey A. Y. 2011 . Circadian cycles are the dominant transcriptional rhythm in the intertidal mussel Mytilus californianus . Proceedings of the National Academy of Sciences of the United States of America , 108 : 16110 – 16115 . Google Scholar Crossref Search ADS PubMed de la Iglesia H. O. , Johnson C. H. 2013 . Biological clocks: riding the tides . Current Biology , 23 : R921 – R923 . Google Scholar Crossref Search ADS PubMed Fabioux C. , Huvet A. , Le Souchu P. , Le Pennec M. , Pouvreau S. 2005 . Temperature and photoperiod drive Crassostrea gigas reproductive internal clock . Aquaculture , 250 : 458 – 470 . Google Scholar Crossref Search ADS Gaten E. , Moss S. , Johnson M. L. 2013 . The Reniform Reflecting Superposition Compound Eyes of Nephrops norvegicus: Optics, Susceptibility to Light-Induced Damage, Electrophysiology and a Ray Tracing Model. In Advances in Marine Biology . Ed. by M. L. Johnson and M. P. Johnson. Elsevier , Amsterdam . 107 – 148 pp. Hardin P. E. , Hall J. C. , Rosbash M. 1990 . Feedback of the Drosophila period gene product on circadian cycling of its messenger RNA levels . Nature , 343 : 536 – 540 . Google Scholar Crossref Search ADS PubMed Helm B. , Ben-Shlomo R. , Sheriff M. J. , Hut R. A. , Foster R. , Barnes B. M. , Dominoni D. 2013 . Annual rhythms that underlie phenology: biological time-keeping meets environmental change . Proceedings of the Royal Society of London B: Biological Sciences , 280 : 20130016. Google Scholar Crossref Search ADS Johnson C. H. , Elliott J. , Foster R. , Honma K.-I. , Kronauer R. 2004 . Fundamental properties of circadian rhythms. In Chronobiology: Biological Timekeeping , p. 406 . Ed. by Dunlap J. C. , Loros J. J. , DeCoursey P. J. . Sinauer Associates , Sunderland, MA . Kumar V. 2017 . Biological Timekeeping . Springer India , New Delhi . Landa-Cansigno C. , Hernández-Carmona G. , Arvizu-Higuera D. L. , Muñoz-Ochoa M. , Hernández-Guerrero C. J. 2017 . Bimonthly variation in the chemical composition and biological activity of the brown seaweed Eisenia arborea (Laminariales: Ochrophyta) from Bahía Magdalena, Baja California Sur, Mexico . Journal of Applied Phycology , 29 : 2605 – 2615 . Google Scholar Crossref Search ADS Last K. S. , Bailhache T. , Kramer C. , Kyriacou C. P. , Rosato E. , Olive P. J. W. 2009 . Tidal, daily, and lunar-day activity cycles in the marine polychaete Nereis virens . Chronobiology International , 26 : 167 – 183 . Google Scholar Crossref Search ADS PubMed Marcus N. H. 1980 . Photoperiodic control of diapause in the marine calanoid copepod Labidocera aestiva . The Biological Bulletin , 159 : 311 – 318 . Google Scholar Crossref Search ADS Mat A. M. , Dunster G. P. , Sbragaglia V. , Aguzzi J. , de la Iglesia H. O. 2017 . Influence of temperature on daily locomotor activity in the crab Uca pugilator . PLoS One , 12 : e0175403. Google Scholar Crossref Search ADS PubMed Mat A. M. , Massabuau J.-C. , Ciret P. , Tran D. 2012 . Evidence for a plastic dual circadian rhythm in the oyster Crassostrea gigas . Chronobiology International , 29 : 857 – 867 . Google Scholar Crossref Search ADS PubMed Mohawk J. A. , Green C. B. , Takahashi J. S. 2012 . Central and peripheral circadian clocks in mammals . Annual Review of Neuroscience , 35 : 445 – 462 . Google Scholar Crossref Search ADS PubMed Naylor E. 2010 . Chronobiology of Marine Organisms . Cambridge University Press , Cambridge . Page T. L. , Koelling E. 2003 . Circadian rhythm in olfactory response in the antennae controlled by the optic lobe in the cockroach . Journal of Insect Physiology , 49 : 697 – 707 . Google Scholar Crossref Search ADS PubMed Preußner M. , Wilhelmi I. , Schultz A.-S. , Finkernagel F. , Michel M. , Möröy T. , Heyd F. 2014 . Rhythmic U2af26 alternative splicing controls PERIOD1 stability and the circadian clock in mice . Molecular Cell , 54 : 651 – 662 . Google Scholar Crossref Search ADS PubMed Refinetti R. , Menaker M. 1992 . The circadian rhythm of body temperature . Physiology & Behavior , 51 : 613 – 637 . Google Scholar Crossref Search ADS PubMed Rosbash M. 2009 . The implications of multiple circadian clock origins . PLoS Biology , 7 : e1000062. Google Scholar Crossref Search ADS Rutila J. E. , Suri V. , Le M. , So W. V. , Rosbash M. , Hall J. C. 1998 . CYCLE is a second bHLH-PAS clock protein essential for circadian rhythmicity and transcription of Drosophila period and timeless . Cell , 93 : 805 – 814 . Google Scholar Crossref Search ADS PubMed Rymer J. , Bauernfeind A. L. , Brown S. , Page T. L. 2007 . Circadian rhythms in the mating behavior of the cockroach, Leucophaea maderae . Journal of Biological Rhythms , 22 : 43 – 57 . Google Scholar Crossref Search ADS PubMed Sbragaglia V. , Aguzzi J. , García J. A. , Chiesa J. J. , Angelini C. , Sardà F. 2013 . Dusk but not dawn burrow emergence rhythms of Nephrops norvegicus (Crustacea: Decapoda) . Scientia Marina , 77 : 641 – 647 . Google Scholar Crossref Search ADS Sehgal A. , Rothenfluh-Hilfiker A. , Hunter-Ensor M. , Chen Y. , Myers M. P. , Young M. W. 1995 . Rhythmic expression of timeless: a basis for promoting circadian cycles in period gene autoregulation . Science , 270 : 808 – 810 . Google Scholar Crossref Search ADS PubMed Sorek M. , Levy O. 2014 . Coral spawning behavior and timing. In Annual, Lunar, and Tidal Clocks . Ed. by Numata H. , Helm B. . Springer Japan, Tokyo . 81 – 97 pp. Sorek M. , Schnytzer Y. , Ben-Asher H. W. , Caspi V. C. , Chen C.-S. , Miller D. J. , Levy O. 2018 . Setting the pace: host rhythmic behaviour and gene expression patterns in the facultatively symbiotic cnidarian Aiptasia are determined largely by Symbiodinium . Microbiome , 6 : Tessmar-Raible K. , Raible F. , Arboleda E. 2011 . Another place, another timer: marine species and the rhythms of life . Bioessays , 33 : 165 – 172 . Google Scholar Crossref Search ADS PubMed Vera L. M. , Negrini P. , Zagatti C. , Frigato E. , Sánchez-Vázquez F. J. , Bertolucci C. 2013 . Light and feeding entrainment of the molecular circadian clock in a marine teleost (Sparus aurata) . Chronobiology International , 30 : 649 – 661 . Google Scholar Crossref Search ADS PubMed Woelfle M. A. , Ouyang Y. , Phanvijhitsiri K. , Johnson C. H. 2004 . The adaptive value of circadian clocks . Current Biology , 14 : 1481 – 1486 . Google Scholar Crossref Search ADS PubMed Yoo S.-H. , Yamazaki S. , Lowrey P. L. , Shimomura K. , Ko C. H. , Buhr E. D. , Siepka S. M. et al. 2004 . PERIOD2: LUCIFERASE real-time reporting of circadian dynamics reveals persistent circadian oscillations in mouse peripheral tissues . Proceedings of the National Academy of Sciences of the United States of America , 101 : 5339 – 5346 . Google Scholar Crossref Search ADS PubMed Young M. W. , Kay S. A. 2001 . Time zones: a comparative genetics of circadian clocks . Nature Reviews Genetics , 2 : 702 – 715 . Google Scholar Crossref Search ADS PubMed Zhang L. , Hastings M. H. , Green E. W. , Tauber E. , Sladek M. , Webster S. G. , Kyriacou C. P. et al. 2013 . Dissociation of circadian and circatidal timekeeping in the marine crustacean Eurydice pulchra . Current Biology , 23 : 1863 – 1873 . Google Scholar Crossref Search ADS PubMed Zhang R. , Lahens N. F. , Ballance H. I. , Hughes M. E. , Hogenesch J. B. 2014 . A circadian gene expression atlas in mammals: implications for biology and medicine . Proceedings of the National Academy of Sciences of the United States of America , 111 : 16219 – 16224 . Google Scholar Crossref Search ADS PubMed © International Council for the Exploration of the Sea 2018. 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)

Diggles, B, K

doi: 10.1093/icesjms/fsy058pmid: N/A

Abstract The scientific literature on the subject of welfare and pain in crustaceans is immature. It is based largely on a few dubious and disputed studies done on a small number of decapod species in instances where nociception was not confirmed, laboratory artefacts occurred, all variables that potentially influence the results were not fully controlled, and interpretations of results were questionable or contradictory. The proposed criteria for pain being applied to crustaceans since 2014 has set the “evidential bar” for pain so low it is impossible to have confidence that the behaviours observed in many experiments are even due to nociception, extinguishing scientific confidence that these behaviours are in any way analogous to how the word pain is defined, used, and understood by humans. Given the critical flaws in design and interpretation of several crustacean “pain” studies, acceptance of claims of pain for these animals, even as a precautionary measure, represents acceptance of a much lower evidential bar than is usually dictated by normal scientific standards. This may lead to circumstances whereby the precautionary principle, underpinned by weak science, is used by decision makers to justify unnecessary constraints on scientific research or other uses of crustaceans, imparting significant costs to scientific programs (and potentially food production industries), which are likely to exceed any benefits from changes in welfare status that may (or may not) accrue to these animals. Introduction An opinion by the European Food Safety Authority’s Scientific Panel on Animal Health and Welfare (EFSA, 2005) concluded “The largest of decapod crustaceans are complex in behaviour and appear to have some degree of awareness. They have a pain system and considerable learning ability”, and that “all decapods should receive protection”. However, decapod crustaceans were subsequently not included for protection under EU animal welfare legislation, due to debate and “fierce resistance” from sections of the scientific community in Europe at the time (Birch, 2017). In the decade since, there have been repeated calls by animal welfare lobby groups in several countries to include crustaceans under welfare legislation as a precautionary measure (Birch, 2017), resulting in the recent inclusion of lobsters and crayfish in welfare legislation in Switzerland on 1 March 2018 (https://www.blv.admin.ch/blv/de/home/tiere/tierschutz/revision-verordnungen-veterinaerbereich.html). The initial scientific resistance back in 2005 was, at least in part, based upon doubt regarding the scientific validity of some of the “pain criteria” being used for crustaceans. Several studies found inconsistencies such as a lack of morphine analgesia to mild electric shocks (Barr and Elwood, 2011), lack of evidence of the presence of nociceptors [the first report of nociceptors in crustaceans was published a decade later by Puri and Faulkes (2015)], and other scientifically critical issues, some of which remain unresolved today (Rose et al., 2014; Puri and Faulkes, 2015; Stevens et al., 2016). The confusion regarding the state of knowledge in this field is exemplified by a recent review by Sneddon (2018), who does not cite Puri and Faulkes (2015) and claims “no studies as yet have identified nociceptors or receptive fields in decapods.” Since 2005 the volume of literature relating to crustacean welfare has increased significantly, as has the number of peer reviewed papers and rebuttal letters highlighting scientific flaws in some of the key “crustacean pain” research papers (e.g. Rose et al., 2014; Puri and Faulkes, 2015; Stevens et al., 2016). In any developing area of scientific research, it is important that points raised by critical reviews and rebuttal papers are also considered by decision makers and legislators whenever the merits of the original literature are assessed. Given that interest in crustacean welfare continues to increase, a critical review of the available scientific literature on the subject of nociception and pain in crustaceans is needed. However, it should be noted that in conducting this review, the author is examining the science and is in no way advocating for careless or indiscriminate use of crustaceans by researchers or industry, out of fundamental respect for life itself (Adamo 2016a). Definitions of pain The scientific problems that have been identified in the field of crustacean pain arise in part from the difficulties with dealing scientifically with subjects that are often subjective. The word “pain” was first developed to describe a human emotional experience often (but not always) associated with trauma or injury (https://www.iasp-pain.org/terminology?navItemNumber=576), so the word “pain” may be accurately used when discussing the relative experiences of humans and closely related primates, other mammals, or even birds. However, as taxa further and further away in evolutionary and morphological terms from humans are considered, it is reasonable to ask how analogous their experiences to noxious stimuli are to the human experience, and therefore how relevant phylogenetically retrospective use of the word “pain” becomes (Derbyshire, 2016; Diggles, 2016). For this reason, some scientists consider it inappropriate (or even mischievous) to use the word “pain” to describe behaviours and experiences of fishes (and crustaceans), as this is essentially a form of anthropomorphism (Rose, 2007; Rose et al., 2014) that invites false equivalence between the experience of those animals and that of human pain (Derbyshire, 2016). For example, the low-voltage electric shocks used by some research groups to generate behavioural changes in crabs have not been shown to specifically induce nociception (Puri and Faulkes, 2015), and therefore could represent an “irritation”, “stimulus”, “unpleasant sensation”, “buzz”, “itch”, or “tingle” if applied to human skin under similar circumstances. In such contexts, the word “pain” is being used instead of other arguably more appropriate terms, possibly because other words do not carry the same legal meaning or headline potential (Stevens et al., 2016; Boutron and Ravaud, 2018). For these reasons, a valid working definition of pain is vital when studying its underlying mechanisms. The key features of the definition of pain by the International Association for the Study of Pain (IASP) are that pain is (i) an unpleasant sensory and emotional experience associated with actual or potential tissue damage, or described in terms of such damage; (ii) pain is always subjective; and (iii) pain is sometimes reported in the absence of tissue damage and the definition of pain should avoid tying pain to an external eliciting stimulus (https://www.iasp-pain.org/terminology?navItemNumber=576, last accessed 8 May 2018). The IASP also define nociception as “the neural process of encoding noxious stimuli, though pain sensation is not necessarily implied”, nociceptors as “a high-threshold sensory receptor of the peripheral somatosensory nervous system that is capable of transducing and encoding noxious stimuli”, and noxious stimuli as “stimulus that is damaging or threatens damage to normal tissues”. Because nociception occurs widely in the animal kingdom (Tobin and Bargmann, 2004; Smith and Lewin, 2009; Sneddon, 2018), it is very important to understand that “activity induced in the nociceptor and nociceptive pathways by a noxious stimulus is not pain, which is always a psychological state” (https://www.iasp-pain.org/terminology?navItemNumber=576, last accessed 8 May 2018). With this in mind, it is clear that nociceptors are not “pain receptors”, as pain (as humans know and understand it) is experienced emotionally in the brain and therefore is likely to only be experienced by animals with brains that are sufficiently developed to generate phenomenal consciousness and sentience (Broom, 2013; Rose et al., 2014; Key, 2015). Hence, conclusions by the EFSA (2005) that decapod crustaceans “have a pain system” were premature, especially considering their statement predated the first scientific confirmation of the presence of nociceptors in a decapod crustacean by a decade (Puri and Faulkes, 2015). Operational definitions of pain have been inconsistent or absent Other problems exist in the area of definitions. Scientists operating in the field of fish and crustacean welfare cannot meet the criteria for pain established for humans, as some of these criteria are untestable in animals. As a response to criticism about their assumptions and conclusions (e.g. Rose, 2002, 2003, 2007; Puri and Faulkes, 2010; Browman and Skiftesvik, 2011; Rose et al., 2014), alternative criteria for defining animal pain were developed (Sneddon et al., 2014). These criteria extended the earlier much criticized “more than a mere reflex” definition and the authors proposed that “if animals fulfill (the) criteria… they should be considered capable, beyond a reasonable doubt, of experiencing pain with implications for their health and welfare” (Sneddon et al., 2014). The criteria included a list of minimal anatomical pre-requisites including presence of nociceptors, central processing in the brain, efficacy of analgesia (but see Barr and Elwood, 2011) as well as behaviours including avoidance responses, behavioural changes, motivational states, and other characteristics, some (but not all) of which needed to be fulfilled in order to “characterize pain beyond a reasonable doubt” (Sneddon et al., 2014). While more detailed than the “more than a mere reflex” definition, the new definitions remain “loose”. For example, Sneddon et al. (2014) do not specify if a minimum level of brain development is required. In the case of crustaceans, this is important as there are over 67 000 described species (Zhang, 2011) exhibiting a huge range in size and anatomical complexity (Regier et al., 2010). Furthermore, as pointed out by Browman and Skiftesvik (2011), the situation leaves room for researchers to pick and choose the criteria that they want to use, and ignore the ones that don’t work for them. The end result has been publication of several more studies since Sneddon et al. (2014) with the same methodological and interpretational problems as the earlier studies. However, these more recent papers not only ignore the earlier criticisms of their methodological and interpretational flaws, but use Sneddon et al. (2014) to support the interpretation that they have found behavioural evidence that is “consistent with criteria for pain” or “consistent with the idea of pain” (e.g. Elwood and Adams, 2015; Rey et al., 2015; Magee and Elwood, 2016; Elwood et al., 2017). In effect, the criteria published by Sneddon et al. (2014) have simply “moved the goalposts”, without addressing the earlier scientific criticisms and have thereby effectively lowered the “evidential bar” needed to claim that the criteria for pain have been fulfilled in fish and invertebrates. The risk of this approach is evidenced by the fact that robots also fulfil many of these criteria (Adamo, 2016a, b). In summary, the criteria for pain proposed by Sneddon et al. (2014) set “the bar” for pain too low to establish any real confidence that the alleged “pain behaviours” in crustaceans (or fish for that matter) are truly related to anything in any way analogous to the definition and use of the word pain that we are familiar with as humans (see Derbyshire, 2016). Problems with experimental administration of putatively “painful” stimuli Application of mild, low-voltage (1–10 V, unknown current) electric shocks to crabs (e.g. Elwood and Adams, 2015), though likely to be sufficient to initiate muscle activity (Stevens et al., 2016), may not represent a truly noxious stimulus to crustaceans as they are unlikely to cause tissue damage (Puri and Faulkes, 2010, 2015). This is an important point that was suggested experimentally by the lack of morphine analgesia to such shocks in shore crabs (Carcinus maenas) (Barr and Elwood, 2011, though see section on Misinterpretation of “morphine analgesia”). The underlying problem is the absence of validated, reliable methods for initiating and measuring nociception in crustaceans using electric shocks. Indeed, electrical stimulation of peripheral nerves has been used to treat pain in humans (Mobbs et al., 2007), demonstrating that all electrical shocks are not necessarily “painful”. Because of the absence of published electrophysiological data on nociceptor activation in crustaceans exposed to mild electric shocks and undisclosed currents, more recent research by Magee and Elwood (2013, 2016) and Elwood and Adams (2015) is critically flawed as activation of nociceptors has been assumed by the authors, but not demonstrated (Table 1). Table 1. Summary of studies using electric shocks on decapod crustaceans. Study Species Voltage (V) Current (mA) Frequency (Hz) Duration (s) Outcome Maldonado and Miralto (1982) Mantis shrimp S. mantis n/a 6.4–74.4 n/a 0.1 Most animals respond with tail flick between 8 and 38.1 mA current. “dose response” to morphine analgesiaa. Lozada et al. (1988) Ghost crab C. granulatus 1–10 n/a 50 1 A total of 22.6% of crabs respond at 6 V, 77.4% respond at 10 V, rest unresponsive. Morphine reduces crabs defensive response to shock in a dose-dependent mannera. Kawai et al. (2004) Crayfish P. clarkii 6.5 0.68 V/cm n/a 0.4 s every 3 s Learning to avoid electric shock after 20 trials/day for 32 days was context dependent. Around half of crayfish facing an escape door learned to avoid shock by walking through the door, vs. 0% for those facing away from the door. Appel and Elwood (2009a) Hermit crab P. bernhardus 0.2–19 n/a 200 1 s every 2 s Crabs respond to electric shock at 9.0–9.1 V regardless of shell type, but more voltage required to evacuate a more preferred type of shellb. Autotomy in 10% of crabs. Appel and Elwood (2009b) Hermit crab P. bernhardus 10 n/a 200 0.2 s every 20 s Shocked crabs more likely to abandon shell and move into new shell faster than unshocked crabsb. Female crabs have lower shock response threshold than males. Elwood and Appel (2009) Hermit crab P. bernhardus 8 n/a 200 1 s every 20 s Shocked crabs more likely to abandon shell and move into new/less preferred shell than unshocked crabsb. Barr and Elwood (2011) European shore crab C. maenas 8 n/a 180 0.2 Morphine inhibits movements with or without shocksb. Magee and Elwood (2013) European shore crab C. maenas 10 n/a 180 0.2 s every 5 s Crabs learned to avoid multiple shocks by avoiding specific sheltersb. Autotomy in 18.9% of crabs, 10 crabs that autotomized 1 leg did not exhibit different behaviour. Elwood and Adams (2015) European shore crab C. maenas 10 n/a 180 0.2 s every 10 s Shocked crabs had higher haemolymph lactate than controlsc. Magee and Elwood (2016) Hermit crab P. bernhardus 1–25 n/a 180 0.2 Initial response 4–6 V (range 1–10 V) in all crabs, but 35.3% of crabs did not evacuate shell up to 25 Vb. Crabs less likely to evacuate shells when exposed to odours from predators or potential food itemsd. Study Species Voltage (V) Current (mA) Frequency (Hz) Duration (s) Outcome Maldonado and Miralto (1982) Mantis shrimp S. mantis n/a 6.4–74.4 n/a 0.1 Most animals respond with tail flick between 8 and 38.1 mA current. “dose response” to morphine analgesiaa. Lozada et al. (1988) Ghost crab C. granulatus 1–10 n/a 50 1 A total of 22.6% of crabs respond at 6 V, 77.4% respond at 10 V, rest unresponsive. Morphine reduces crabs defensive response to shock in a dose-dependent mannera. Kawai et al. (2004) Crayfish P. clarkii 6.5 0.68 V/cm n/a 0.4 s every 3 s Learning to avoid electric shock after 20 trials/day for 32 days was context dependent. Around half of crayfish facing an escape door learned to avoid shock by walking through the door, vs. 0% for those facing away from the door. Appel and Elwood (2009a) Hermit crab P. bernhardus 0.2–19 n/a 200 1 s every 2 s Crabs respond to electric shock at 9.0–9.1 V regardless of shell type, but more voltage required to evacuate a more preferred type of shellb. Autotomy in 10% of crabs. Appel and Elwood (2009b) Hermit crab P. bernhardus 10 n/a 200 0.2 s every 20 s Shocked crabs more likely to abandon shell and move into new shell faster than unshocked crabsb. Female crabs have lower shock response threshold than males. Elwood and Appel (2009) Hermit crab P. bernhardus 8 n/a 200 1 s every 20 s Shocked crabs more likely to abandon shell and move into new/less preferred shell than unshocked crabsb. Barr and Elwood (2011) European shore crab C. maenas 8 n/a 180 0.2 Morphine inhibits movements with or without shocksb. Magee and Elwood (2013) European shore crab C. maenas 10 n/a 180 0.2 s every 5 s Crabs learned to avoid multiple shocks by avoiding specific sheltersb. Autotomy in 18.9% of crabs, 10 crabs that autotomized 1 leg did not exhibit different behaviour. Elwood and Adams (2015) European shore crab C. maenas 10 n/a 180 0.2 s every 10 s Shocked crabs had higher haemolymph lactate than controlsc. Magee and Elwood (2016) Hermit crab P. bernhardus 1–25 n/a 180 0.2 Initial response 4–6 V (range 1–10 V) in all crabs, but 35.3% of crabs did not evacuate shell up to 25 Vb. Crabs less likely to evacuate shells when exposed to odours from predators or potential food itemsd. n/a = data not shown, therefore difficult to determine shock intensity. a Result may be due to “generalized non-responsiveness” due to morphine rather than analgesia of nociception. b Nociception unknown/doubtful at voltages used in absence of current data. c  Result may be due to uncontrolled factors such as muscular contraction or respiratory/cardiac apnoea and/or cardiac bradycardia. d Result may be due to uncontrolled/unrelated factors such as olfaction/gustation. Table 1. Summary of studies using electric shocks on decapod crustaceans. Study Species Voltage (V) Current (mA) Frequency (Hz) Duration (s) Outcome Maldonado and Miralto (1982) Mantis shrimp S. mantis n/a 6.4–74.4 n/a 0.1 Most animals respond with tail flick between 8 and 38.1 mA current. “dose response” to morphine analgesiaa. Lozada et al. (1988) Ghost crab C. granulatus 1–10 n/a 50 1 A total of 22.6% of crabs respond at 6 V, 77.4% respond at 10 V, rest unresponsive. Morphine reduces crabs defensive response to shock in a dose-dependent mannera. Kawai et al. (2004) Crayfish P. clarkii 6.5 0.68 V/cm n/a 0.4 s every 3 s Learning to avoid electric shock after 20 trials/day for 32 days was context dependent. Around half of crayfish facing an escape door learned to avoid shock by walking through the door, vs. 0% for those facing away from the door. Appel and Elwood (2009a) Hermit crab P. bernhardus 0.2–19 n/a 200 1 s every 2 s Crabs respond to electric shock at 9.0–9.1 V regardless of shell type, but more voltage required to evacuate a more preferred type of shellb. Autotomy in 10% of crabs. Appel and Elwood (2009b) Hermit crab P. bernhardus 10 n/a 200 0.2 s every 20 s Shocked crabs more likely to abandon shell and move into new shell faster than unshocked crabsb. Female crabs have lower shock response threshold than males. Elwood and Appel (2009) Hermit crab P. bernhardus 8 n/a 200 1 s every 20 s Shocked crabs more likely to abandon shell and move into new/less preferred shell than unshocked crabsb. Barr and Elwood (2011) European shore crab C. maenas 8 n/a 180 0.2 Morphine inhibits movements with or without shocksb. Magee and Elwood (2013) European shore crab C. maenas 10 n/a 180 0.2 s every 5 s Crabs learned to avoid multiple shocks by avoiding specific sheltersb. Autotomy in 18.9% of crabs, 10 crabs that autotomized 1 leg did not exhibit different behaviour. Elwood and Adams (2015) European shore crab C. maenas 10 n/a 180 0.2 s every 10 s Shocked crabs had higher haemolymph lactate than controlsc. Magee and Elwood (2016) Hermit crab P. bernhardus 1–25 n/a 180 0.2 Initial response 4–6 V (range 1–10 V) in all crabs, but 35.3% of crabs did not evacuate shell up to 25 Vb. Crabs less likely to evacuate shells when exposed to odours from predators or potential food itemsd. Study Species Voltage (V) Current (mA) Frequency (Hz) Duration (s) Outcome Maldonado and Miralto (1982) Mantis shrimp S. mantis n/a 6.4–74.4 n/a 0.1 Most animals respond with tail flick between 8 and 38.1 mA current. “dose response” to morphine analgesiaa. Lozada et al. (1988) Ghost crab C. granulatus 1–10 n/a 50 1 A total of 22.6% of crabs respond at 6 V, 77.4% respond at 10 V, rest unresponsive. Morphine reduces crabs defensive response to shock in a dose-dependent mannera. Kawai et al. (2004) Crayfish P. clarkii 6.5 0.68 V/cm n/a 0.4 s every 3 s Learning to avoid electric shock after 20 trials/day for 32 days was context dependent. Around half of crayfish facing an escape door learned to avoid shock by walking through the door, vs. 0% for those facing away from the door. Appel and Elwood (2009a) Hermit crab P. bernhardus 0.2–19 n/a 200 1 s every 2 s Crabs respond to electric shock at 9.0–9.1 V regardless of shell type, but more voltage required to evacuate a more preferred type of shellb. Autotomy in 10% of crabs. Appel and Elwood (2009b) Hermit crab P. bernhardus 10 n/a 200 0.2 s every 20 s Shocked crabs more likely to abandon shell and move into new shell faster than unshocked crabsb. Female crabs have lower shock response threshold than males. Elwood and Appel (2009) Hermit crab P. bernhardus 8 n/a 200 1 s every 20 s Shocked crabs more likely to abandon shell and move into new/less preferred shell than unshocked crabsb. Barr and Elwood (2011) European shore crab C. maenas 8 n/a 180 0.2 Morphine inhibits movements with or without shocksb. Magee and Elwood (2013) European shore crab C. maenas 10 n/a 180 0.2 s every 5 s Crabs learned to avoid multiple shocks by avoiding specific sheltersb. Autotomy in 18.9% of crabs, 10 crabs that autotomized 1 leg did not exhibit different behaviour. Elwood and Adams (2015) European shore crab C. maenas 10 n/a 180 0.2 s every 10 s Shocked crabs had higher haemolymph lactate than controlsc. Magee and Elwood (2016) Hermit crab P. bernhardus 1–25 n/a 180 0.2 Initial response 4–6 V (range 1–10 V) in all crabs, but 35.3% of crabs did not evacuate shell up to 25 Vb. Crabs less likely to evacuate shells when exposed to odours from predators or potential food itemsd. n/a = data not shown, therefore difficult to determine shock intensity. a Result may be due to “generalized non-responsiveness” due to morphine rather than analgesia of nociception. b Nociception unknown/doubtful at voltages used in absence of current data. c  Result may be due to uncontrolled factors such as muscular contraction or respiratory/cardiac apnoea and/or cardiac bradycardia. d Result may be due to uncontrolled/unrelated factors such as olfaction/gustation. Puri and Faulkes (2015) noted that the behavioural data produced in the various papers by Elwood et al. can be interpreted in ways that do not require nociception to explain the results, and several researchers have pointed out that electric shocks are likely to activate any electrically excitable cell, including non-neural ones (Puri and Faulkes, 2015; Stevens et al., 2016). If nociceptors are not being specifically stimulated by the undisclosed electrical currents used in these experiments (it is electrical current, not voltage, that activates tissues), the behaviours of the crustaceans studied could therefore simply be habituation or associative learning in response to an irritating stimulus, rather than a specific nociceptive response to tissue damage (Puri and Faulkes, 2015; Stevens et al., 2016), and learning in invertebrates is not evidence of awareness or pain (Tobin and Bargmann, 2004; Broom, 2013; Rose et al., 2014). Because of these reasons, studies that utilize electric shock on crustaceans without disclosing electrical current data or demonstrating nociceptor activation should be discouraged. Negative results and alternative interpretations are often ignored It has been previously pointed out that many studies about aquatic animal welfare “ignore negative results” and “inflate the science boundary” (Browman and Skiftesvik, 2011; Rose et al., 2014). When reviewed objectively, scientific claims that fish or crustaceans “may feel pain” have been largely based on a few dubious and disputed studies done on a small number of animals and species in instances where laboratory artefacts are known to have occurred, all variables that potentially influence the results are not fully controlled, and interpretations of results have been questionable and sometimes contradictory (e.g. Newby and Stevens, 2008, 2009; Newby et al., 2009; Puri and Faulkes, 2010; Rose et al., 2014; Stevens et al., 2016; Diggles et al., 2017; Key et al., 2017). For crustacean research, these problems are explored in more detail in the section on Specific problems with the scientific literature on putative pain in crustaceans, but in many cases the flaws in these studies have arisen in part due to failure to control other variables known to affect animal behaviour (e.g. chemosensation, olfaction, gustation), and failure to consider alternative, more parsimonious interpretations (Rose et al., 2014; Stevens et al., 2016; Diggles et al., 2017; Key et al., 2017; Boutron and Ravaud, 2018). Notably, when some of these studies (e.g. Sneddon, 2003; Barr et al., 2008) have been repeated by other research groups (Newby and Stevens, 2008; Puri and Faulkes, 2010), some of the “key findings” were not replicated (Newby and Stevens, 2009; Puri and Faulkes, 2010). These problems are all signs of research in an emerging scientific field that is immature. Specific problems with the scientific literature on putative pain in crustaceans A more detailed review of key crustacean welfare papers highlights several specific scientific problems that are evident in the literature on putative pain in crustaceans. Misinterpretation of “morphine analgesia” Earlier studies (Maldonado and Miralto, 1982), reported apparent dose-dependent “morphine analgesia” after electric shocking (range 8−38.1 mA current) in mantis shrimp (Squilla mantis). A few years later Lozada et al. (1988) reported similar results for a varunid ghost crab (Chasmagnathus granulatus), using low voltages (1–10 V) but unreported amperages (Table 1). However, Bergamo et al. (1992) observed that threat responses in Carcinus mediterraneus tapped on the carapace with a 10 g weight were also halted when the crabs were dosed with morphine, a result consistent with later research showing that morphine has various non-specific effects on crustacean behaviour (Nathaniel et al., 2010, Imeh-Nathaniel et al., 2017). Subsequent studies on shore crabs noted that morphine may result in “general non-responsiveness” (Barr and Elwood, 2011) that could explain the claimed “analgesic effect” of morphine reported in the earlier studies, leaving open the question of whether nociception was ever achieved by Maldonado and Miralto (1982) and Lozada et al. (1988) using electric shocks, especially given nociception was not demonstrated in either study. Experiments employing electric shocks: technical shortcomings and misinterpreted results Other literature relating to application of electric shocks to crustaceans are summarized in Table 1. Kawai et al. (2004) examined if crayfish (Procambarus clarkii) could learn to avoid electric shocks (6.5 V for 0.4 s, every 3 s) by exiting a compartment via an escape door. Electric shocks initiated tail flip escape responses. However, over time around 50% of crayfish could learn to avoid the electric shock by walking through the escape door, but only if they were facing toward the escape door. Crayfish oriented away from the escape door never learned to avoid the shock, showing that avoidance learning was context dependent. Appel and Elwood (2009a) reported motivational trade-offs in hermit crabs (Pagurus bernhardus) exposed to electric shocks as evidence of a “potential pain experience”. They found crabs first reacted to electric shock at the same voltage (9.0–9.1 V) regardless of shell type, but it took more voltage on average to evacuate a more preferred shell type (Littorina, 17.7 V) compared with a less preferred shell type (Gibbula, 14.9 V). However, no measurement of the amperage used was reported in that study (Table 1), and no electrophysiological evidence of presumed nociceptor activity was provided, hence it is difficult to ascertain if sufficient electrical current was used to activate tissues, and what receptors (if any) were activated. Elwood and Appel (2009) exposed the same species to mild electric shocks (8 V) intended, in the authors words “to be below the level that would cause the crab to leave the shell and thus judged not to be severe”. Hence given the absence of current (amperage) data or electrophysiological evidence of nociceptor activity, and given that electric shocks would interact with a wide range of cells and receptors (Derby and Steullet, 2001; Puri and Faulkes, 2010, 2015; Stevens et al., 2016), it is again impossible to determine whether nociception occurred at all, or if so, whether it occurred in the absence of other confounding factors in the Elwood and Appel (2009) study, or for that matter in any of their subsequent studies in which electric shocks were used and electrical current and electrophysiological data were not reported (Appel and Elwood, 2009b; Barr and Elwood, 2011; Magee and Elwood, 2013, 2016; Elwood and Adams, 2015; Elwood et al., 2017; Table 1). The main problem with electric shock, especially for aquatic animals, is that it non-specifically activates any electrically excitable cell, including non-neural ones (e.g. muscle tissue), meaning that assumed “nociceptive behaviours” triggered by such stimuli may represent abnormal responses of the nervous system (or other systems, see Derby and Steullet, 2001), rather than reveal the workings of a nociceptive sensory system tuned to tissue damage by evolution (Puri and Faulkes, 2015). These issues with research using electrical shocks on crabs were highlighted upon publication of a paper by Elwood and Adams (2015) who exposed shore crabs (C. maenas) to electric shocks (10 V, but again unknown current), and measured increased haemolymph lactate in shocked crabs, concluding that their study “fulfils the criteria expected of a pain experience”. A comment on the article (Stevens et al., 2016) highlighted some of the scientific flaws in Elwood and Adams (2015) and advised policy-makers not to make inappropriate decisions about crab welfare based on such studies. First, Stevens et al. (2016) pointed out that Elwood and Adams conflicted the terms “stress” and “pain”, which involve two separate pathways in all animals. This is because Elwood and Adams (2015) used a known stress marker (haemolymph lactate) to suggest that the shocks used induced pain in the experimental crabs without inducing muscle activity, thus correlating stress with pain. Stevens et al. (2016) pointed out that it has been known for many decades that elevated lactate in crabs occurs in association with many other conditions, for example, during any muscular activity or with an increase in water temperature, and that some earlier studies showed the lactate levels reported by Elwood and Adams (2015) were within the normal range for C. maenas. Furthermore, they also pointed out that other studies have reported lactate levels in exercised crabs that were ten times higher than those reported by Elwood and Adams (2015). It is also known that exposure to a “wide range of chemical and environmental stimuli” such as flashes of light or touching the eyes or carapace (Wilkens et al., 1974), and thus also presumably electric shocks, can stop the heartbeat and breathing of crustaceans for short periods (cardiac bradycardia and/or respiratory apnoea), increasing lactate via temporary anaerobic metabolism as part of a normal anti-predator “startle response” that could hide the crustacean from predators that can detect weak electric fields (such as sharks and rays) (Wilkens et al., 1974; Stevens et al., 2016). In summary, Stevens et al. (2016) found that the methods used by Elwood and Adams (2015) could not distinguish between normal stress or startle responses and pain, and that there were reasonable alternative explanations for the elevated lactate reported (i.e. alternative and more parsimonious explanations for their results apparently were not considered and certainly were not tested experimentally). Interactions between electric shocks and learning behaviours in various decapods continue to be investigated, and the observations continue to be interpreted as being “consistent with the idea of pain” (Magee and Elwood, 2016). However, this more recent work has the same problems as the previous studies. For example, in a study by Magee and Elwood (2016), 35.3% of hermit crabs (P. bernhardus) did not evacuate their shells despite having been exposed to voltages of up to 25 V. In the absence of electrical current and electrophysiological data, it is impossible to determine whether non-evacuation of shells was due to insufficient current, as well as the odour interactions that were also being studied. For the latter, hermit crabs were less likely to evacuate their shells when exposed to “predator odours” (water from tanks containing C. maenas), but unexpectedly, hermit crabs were also less likely to evacuate shells when an undiluted “non-predator odour” (=mussel odour) was added. Magee and Elwood (2016) speculated that “This could be due to the concentration of the odour being unusually high and hence novel when compared to a rocky shore, where there would be regular flushing. Alternatively, it could be due to an association of mussel beds with predators of hermit crabs, such as larger shell breaking crabs”. The authors did not mention that mussels are a potent food source for crustaceans, containing large quantities of attractants, which are known olfactory and gustatory stimulants (Kasumyan and Doving, 2003; Derby et al., 2016), which can influence crustacean behaviour (e.g. Stocker and Huber, 2001). Hence, failure of a hermit crab to evacuate its shell in Magee and Elwood (2016) could also be associated with olfactory, gustatory, or other chemosensory stimuli that may have initiated other unforeseen behavioural interactions (e.g. preparation for exploratory food searches). Effects of chemical applications are often not replicable and their interpretation is controversial Several studies have examined the effects of chemicals, temperature, and other potentially noxious non-electrical stimuli on crustacean behaviour (Table 2). Barr et al. (2008) studied the effects of 10% acetic acid (vinegar), bases (sodium hydroxide), and anaesthetic (benzocaine) on glass prawns (Palaemon elegans), concluding that by grooming and rubbing antennae exposed to these compounds (including benzocaine), treated prawns were “attempting to ameliorate a painful effect of the stimulus… … .consistent with the idea of pain”. However, other researchers repeated the study using three other crustacean species [crayfish P. clarkii, white shrimp (Litopenaeus setiferus), and grass shrimp (Palaemonetes sp.)] and failed to replicate these results (Puri and Faulkes, 2010). Instead, they found no responses to extreme pH (hydrochloric acid and sodium hydroxide) in any of these species. They also failed to find any behavioural or physiological evidence (including electrophysiological evidence) that antennae contained nociceptors, suggesting either that the ability to respond to pH was not universal in crustaceans or (more likely), that Barr et al. (2008) had mischaracterized normal grooming or other behaviours as evidence of nociception (Puri and Faulkes, 2010). Table 2. Summary of studies using chemicals and temperature variations on decapod crustaceans. Study Species Chemicals/temperature Concentrations Outcome Barr et al. (2008) Glass prawn P. elegans Acetic acid, sodium hydroxide, benzocaine 10% (acetic acid, sodium hydroxide), or 2% benzocaine Increased grooming and rubbing of treated antennae (including antennae treated with benzocaine)a. Puri and Faulkes (2010) Crayfish P. clarkii white shrimp L. setiferus grass shrimp Palaemonetes sp. Hydrochloric acid, sodium hydroxide, benzocaine 6 mol/l HCL and NaOH for P. clarkii and L. setiferus, 1 mol/l HCL and NaOH for Palaemonetes sp., 2% benzocaine for P. clarkii Responses to extreme pH were not observed in any of the species examined. No behavioural or electrophysiological evidence was found that antennae contained nociceptors for extreme pH or benzocaine/ethanol. Kotsyuba et al. (2010) Shore crab H. sanguineus Formaldehyde 1% formaldehyde in physiological solution, Dose 0.001 ml/g body weight Crabs from a polluted area had higher haemolymph nitric oxide levels than crabs from a clean area. Formalin-injected crabs (into the right cheliped) exhibited hyperactive movement of that cheliped. Autotomy of injected cheliped occurred in 80% of crabs from the polluted area (60% mortality), but only 10% of crabs from the clean area exhibited autotomy (10% mortality). Aggio et al. (2012) Blue crab C. sapidus Aplysia ink, denatonium, quinine, caffeine, cinnamaldehyde Undiluted (ink) or 5 mmol/l (all others) Crabs always searched for deterrent laced food and place it in their mouth parts. The deterrent effect manifests via oesophageal taste receptors as either rejection or extensive manipulation, but in both cases crabs bit the adulterated food. Dyuizen et al. (2012) Shore crab H. sanguineus Formaldehyde 1% formalin in artificial seawater. Dose 0.0005 ml/g body weight Formalin-injected crabs (into the right cheliped) exhibited increased flexion, movement, and rubbing of that cheliped in first 3 min post-injection. Autotomy of injected cheliped occurred in 20% of crabs within 32 s. Reduced use of injected cheliped in remaining crabs over next 10 min post-injection. Puri and Faulkes (2015) Crayfish P. clarkii Low and high temperatures, capsaicin, isothiocyanate Soldering iron at 54°C, dry ice at −78.5°C fed peppers (capsaicin) or wasabi (isothiocyanate) per os ad libitum No aversion or grooming responses to capsaicin or isothiocyanate, no response to low temperatures, however vigorous escape responses when touched with a hot soldering iron but no long-term trauma noted. Nociceptors sensitive to heat confirmed by electrophysiologyb. Elwood et al. (2017) European shore crab C. maenas Acetic acid, capsaicin, mineral oil 10% acetic acid (vinegar), 0.018 g capsaicin per 10 ml mineral oil, undiluted mineral oil Application of acetic acid to the mouth and eyes resulted in vigorous movement of mouth partsc and in the case of application to the mouth rapid movements interpreted as attempts to escape the enclosured. Study Species Chemicals/temperature Concentrations Outcome Barr et al. (2008) Glass prawn P. elegans Acetic acid, sodium hydroxide, benzocaine 10% (acetic acid, sodium hydroxide), or 2% benzocaine Increased grooming and rubbing of treated antennae (including antennae treated with benzocaine)a. Puri and Faulkes (2010) Crayfish P. clarkii white shrimp L. setiferus grass shrimp Palaemonetes sp. Hydrochloric acid, sodium hydroxide, benzocaine 6 mol/l HCL and NaOH for P. clarkii and L. setiferus, 1 mol/l HCL and NaOH for Palaemonetes sp., 2% benzocaine for P. clarkii Responses to extreme pH were not observed in any of the species examined. No behavioural or electrophysiological evidence was found that antennae contained nociceptors for extreme pH or benzocaine/ethanol. Kotsyuba et al. (2010) Shore crab H. sanguineus Formaldehyde 1% formaldehyde in physiological solution, Dose 0.001 ml/g body weight Crabs from a polluted area had higher haemolymph nitric oxide levels than crabs from a clean area. Formalin-injected crabs (into the right cheliped) exhibited hyperactive movement of that cheliped. Autotomy of injected cheliped occurred in 80% of crabs from the polluted area (60% mortality), but only 10% of crabs from the clean area exhibited autotomy (10% mortality). Aggio et al. (2012) Blue crab C. sapidus Aplysia ink, denatonium, quinine, caffeine, cinnamaldehyde Undiluted (ink) or 5 mmol/l (all others) Crabs always searched for deterrent laced food and place it in their mouth parts. The deterrent effect manifests via oesophageal taste receptors as either rejection or extensive manipulation, but in both cases crabs bit the adulterated food. Dyuizen et al. (2012) Shore crab H. sanguineus Formaldehyde 1% formalin in artificial seawater. Dose 0.0005 ml/g body weight Formalin-injected crabs (into the right cheliped) exhibited increased flexion, movement, and rubbing of that cheliped in first 3 min post-injection. Autotomy of injected cheliped occurred in 20% of crabs within 32 s. Reduced use of injected cheliped in remaining crabs over next 10 min post-injection. Puri and Faulkes (2015) Crayfish P. clarkii Low and high temperatures, capsaicin, isothiocyanate Soldering iron at 54°C, dry ice at −78.5°C fed peppers (capsaicin) or wasabi (isothiocyanate) per os ad libitum No aversion or grooming responses to capsaicin or isothiocyanate, no response to low temperatures, however vigorous escape responses when touched with a hot soldering iron but no long-term trauma noted. Nociceptors sensitive to heat confirmed by electrophysiologyb. Elwood et al. (2017) European shore crab C. maenas Acetic acid, capsaicin, mineral oil 10% acetic acid (vinegar), 0.018 g capsaicin per 10 ml mineral oil, undiluted mineral oil Application of acetic acid to the mouth and eyes resulted in vigorous movement of mouth partsc and in the case of application to the mouth rapid movements interpreted as attempts to escape the enclosured. a Result probably due to mistaking normal grooming or chemosensory behaviour as nociception. b First demonstration of nociceptors in crustaceans. c Nociception not demonstrated by electrophysiology. d Movements may be due to uncontrolled/unrelated factors such as preparatory food finding exploration following olfaction/gustation. Table 2. Summary of studies using chemicals and temperature variations on decapod crustaceans. Study Species Chemicals/temperature Concentrations Outcome Barr et al. (2008) Glass prawn P. elegans Acetic acid, sodium hydroxide, benzocaine 10% (acetic acid, sodium hydroxide), or 2% benzocaine Increased grooming and rubbing of treated antennae (including antennae treated with benzocaine)a. Puri and Faulkes (2010) Crayfish P. clarkii white shrimp L. setiferus grass shrimp Palaemonetes sp. Hydrochloric acid, sodium hydroxide, benzocaine 6 mol/l HCL and NaOH for P. clarkii and L. setiferus, 1 mol/l HCL and NaOH for Palaemonetes sp., 2% benzocaine for P. clarkii Responses to extreme pH were not observed in any of the species examined. No behavioural or electrophysiological evidence was found that antennae contained nociceptors for extreme pH or benzocaine/ethanol. Kotsyuba et al. (2010) Shore crab H. sanguineus Formaldehyde 1% formaldehyde in physiological solution, Dose 0.001 ml/g body weight Crabs from a polluted area had higher haemolymph nitric oxide levels than crabs from a clean area. Formalin-injected crabs (into the right cheliped) exhibited hyperactive movement of that cheliped. Autotomy of injected cheliped occurred in 80% of crabs from the polluted area (60% mortality), but only 10% of crabs from the clean area exhibited autotomy (10% mortality). Aggio et al. (2012) Blue crab C. sapidus Aplysia ink, denatonium, quinine, caffeine, cinnamaldehyde Undiluted (ink) or 5 mmol/l (all others) Crabs always searched for deterrent laced food and place it in their mouth parts. The deterrent effect manifests via oesophageal taste receptors as either rejection or extensive manipulation, but in both cases crabs bit the adulterated food. Dyuizen et al. (2012) Shore crab H. sanguineus Formaldehyde 1% formalin in artificial seawater. Dose 0.0005 ml/g body weight Formalin-injected crabs (into the right cheliped) exhibited increased flexion, movement, and rubbing of that cheliped in first 3 min post-injection. Autotomy of injected cheliped occurred in 20% of crabs within 32 s. Reduced use of injected cheliped in remaining crabs over next 10 min post-injection. Puri and Faulkes (2015) Crayfish P. clarkii Low and high temperatures, capsaicin, isothiocyanate Soldering iron at 54°C, dry ice at −78.5°C fed peppers (capsaicin) or wasabi (isothiocyanate) per os ad libitum No aversion or grooming responses to capsaicin or isothiocyanate, no response to low temperatures, however vigorous escape responses when touched with a hot soldering iron but no long-term trauma noted. Nociceptors sensitive to heat confirmed by electrophysiologyb. Elwood et al. (2017) European shore crab C. maenas Acetic acid, capsaicin, mineral oil 10% acetic acid (vinegar), 0.018 g capsaicin per 10 ml mineral oil, undiluted mineral oil Application of acetic acid to the mouth and eyes resulted in vigorous movement of mouth partsc and in the case of application to the mouth rapid movements interpreted as attempts to escape the enclosured. Study Species Chemicals/temperature Concentrations Outcome Barr et al. (2008) Glass prawn P. elegans Acetic acid, sodium hydroxide, benzocaine 10% (acetic acid, sodium hydroxide), or 2% benzocaine Increased grooming and rubbing of treated antennae (including antennae treated with benzocaine)a. Puri and Faulkes (2010) Crayfish P. clarkii white shrimp L. setiferus grass shrimp Palaemonetes sp. Hydrochloric acid, sodium hydroxide, benzocaine 6 mol/l HCL and NaOH for P. clarkii and L. setiferus, 1 mol/l HCL and NaOH for Palaemonetes sp., 2% benzocaine for P. clarkii Responses to extreme pH were not observed in any of the species examined. No behavioural or electrophysiological evidence was found that antennae contained nociceptors for extreme pH or benzocaine/ethanol. Kotsyuba et al. (2010) Shore crab H. sanguineus Formaldehyde 1% formaldehyde in physiological solution, Dose 0.001 ml/g body weight Crabs from a polluted area had higher haemolymph nitric oxide levels than crabs from a clean area. Formalin-injected crabs (into the right cheliped) exhibited hyperactive movement of that cheliped. Autotomy of injected cheliped occurred in 80% of crabs from the polluted area (60% mortality), but only 10% of crabs from the clean area exhibited autotomy (10% mortality). Aggio et al. (2012) Blue crab C. sapidus Aplysia ink, denatonium, quinine, caffeine, cinnamaldehyde Undiluted (ink) or 5 mmol/l (all others) Crabs always searched for deterrent laced food and place it in their mouth parts. The deterrent effect manifests via oesophageal taste receptors as either rejection or extensive manipulation, but in both cases crabs bit the adulterated food. Dyuizen et al. (2012) Shore crab H. sanguineus Formaldehyde 1% formalin in artificial seawater. Dose 0.0005 ml/g body weight Formalin-injected crabs (into the right cheliped) exhibited increased flexion, movement, and rubbing of that cheliped in first 3 min post-injection. Autotomy of injected cheliped occurred in 20% of crabs within 32 s. Reduced use of injected cheliped in remaining crabs over next 10 min post-injection. Puri and Faulkes (2015) Crayfish P. clarkii Low and high temperatures, capsaicin, isothiocyanate Soldering iron at 54°C, dry ice at −78.5°C fed peppers (capsaicin) or wasabi (isothiocyanate) per os ad libitum No aversion or grooming responses to capsaicin or isothiocyanate, no response to low temperatures, however vigorous escape responses when touched with a hot soldering iron but no long-term trauma noted. Nociceptors sensitive to heat confirmed by electrophysiologyb. Elwood et al. (2017) European shore crab C. maenas Acetic acid, capsaicin, mineral oil 10% acetic acid (vinegar), 0.018 g capsaicin per 10 ml mineral oil, undiluted mineral oil Application of acetic acid to the mouth and eyes resulted in vigorous movement of mouth partsc and in the case of application to the mouth rapid movements interpreted as attempts to escape the enclosured. a Result probably due to mistaking normal grooming or chemosensory behaviour as nociception. b First demonstration of nociceptors in crustaceans. c Nociception not demonstrated by electrophysiology. d Movements may be due to uncontrolled/unrelated factors such as preparatory food finding exploration following olfaction/gustation. Kotsyuba et al. (2010) injected the claws of shore crabs (Hemigrapsus sanguineus) with formalin to investigate how stressors affected nitric oxide production. Formalin-injected crabs exhibited hyperactive movement of that cheliped followed by autotomy of the appendage in 80% of crabs from a polluted area, and 60% mortality within 4 h. In contrast, rates of both autotomy and mortality in crabs taken from a relatively unpolluted area were only 10%, showing how environmental stressors could significantly affect results. Dyuizen et al. (2012) also injected formalin into the claws of H. sanguineus, at a dose rate of 50% of that used by Kotsyuba et al. (2010). Formalin-injected crabs exhibited increased flexion, movement and rubbing of the injected claw in first 3 min post injection, with autotomy of the injected claw occurring in 20% of crabs within 32 s. The remaining crabs reduced use of the injected cheliped over the next 10 min compared to saline-injected controls (none of which autotomized any claws) before reverting to normal behaviour. The autotomy behaviour in these two studies is notable as its prevalence was dependent on the formalin dose used. However, given that autotomy of claws in edible crabs (Cancer pagurus) was no more stressful than handling alone, based on haemolymph glucose and lactate measurements (Patterson et al., 2007), the relevance of autotomy to crab welfare in this context remains unclear. Aggio et al. (2012) investigated exposure of blue crabs (Callinectes sapidus) to chemical deterrents and found that they always searched for deterrent-adulterated food and placed it in their mouth parts, with rejection of adulterated feed only occurring after tasting and extensive manipulation. This suggests that chemicals that are deterrents for other animal species are not necessarily deterrents for crabs. In a landmark study Puri and Faulkes (2015) were the first scientists to report the presence of nociceptors in decapod crustaceans when they found behavioural and electrophysiological evidence of activation of nociceptors sensitive to heat by touching crayfish (P. clarkii) with a soldering iron set at 54°C. However, they failed to find any evidence of nociception when crayfish were exposed to dry ice at −78.5°C, or to nocigenic chemicals in foods including capsaicin (in peppers) or isothiocyanate (wasabi) (Puri and Faulkes, 2015). The recent animal welfare legislation in Switzerland states that lobsters and other crayfish cannot be transported on ice or in ice water (https://www.blv.admin.ch/blv/de/home/tiere/tierschutz/revision-verordnungen-veterinaerbereich.html). The evidence used for this decision is unclear, given the observations of Puri and Faulkes (2015) that suggests the absence of nociception in crayfish exposed to cold temperatures. Perhaps this is because, in contrast with crustacean welfare literature published by some other research groups, the study of Puri and Faulkes (2015) did not gain international headlines, but instead they prudently cautioned readers against over interpreting their results. A study by Elwood et al. (2017) tested behavioural responses of the European shore crab (C. maenas) to 10% acetic acid (vinegar) and capsaicin solutions applied to the mouth and eyes. Like Puri and Faulkes (2015), Elwood et al. (2017) did not find evidence of any nocigenic effects of capsaicin. However, application of acetic acid to the mouth and eyes of C. maenas resulted in vigorous movement of mouth parts and, when applied to the mouth rapid movements that were interpreted as “attempts to escape the enclosure” (Elwood et al., 2017). When acetic acid was applied to an eye, the mouth parts moved and the claws scratched at the mouth in a similar manner to when acetic acid was applied to the mouth, but the eyes tended to be held down for longer. The authors suggested that one possible reason for the mouth part movements in response to the eye treatment was that a small groove near to the base of each antenna might allow some of the acetic acid to trickle down to the mouth area. Elwood et al. (2017) claimed (again without electrophysiological evidence) that the increased activity in crabs exposed to acetic acid was evidence of nociception and subsequent behaviours were “consistent with the idea of pain”. They also noted that while Puri and Faulkes (2010) found no evidence of nociception in response to hydrochloric acid, Barr et al. (2008) and Elwood et al. (2017) found that “acetic acid has considerable effect when applied to antennae, eyes and mouth….(and) it seems unlikely that this is a species effect and it might be that different acids act differently on nociceptors”. A more plausible interpretation of the results of Elwood et al. (2017) is that some acidic compounds act on chemoreceptors that are not nociceptors (see Diggles et al., 2017). The fact that different acids affect behaviour of aquatic animals in different ways was demonstrated by Lopez-Luna et al. (2017), who found zebrafish (Danio rerio) larvae exposed via the water to acetic acid exhibited different behaviour compared to fish exposed to citric acid in water of the same pH. Fish exposed to citric acid remained active even at concentrations of 5% when pH dropped as low as 2.6, while fish exposed to the acetic acid showed an initial increase in activity at low concentrations (0.1%), followed by a decrease in activity at higher acetic acid concentrations. The decreased activity was assumed by Lopez-Luna et al. (2017) to be due to “pain”, however earlier researchers had assumed that increased (not decreased) activity was evidence of nociception in zebrafish larvae exposed to acetic acid via the water (Steenbergen and Bardine, 2014), calling the construct validity of the assay into question. Administration of acetic acid via the water (in fish) or via topical application (crabs) has not been shown to trigger nociceptors in any electrophysiological studies to date in either fish (Diggles et al., 2017) or crustaceans, despite the fact that nociception has been assumed to occur by several authors (Steenbergen and Bardine, 2014; Elwood et al., 2017; Lopez-Luna et al., 2017). A critical point that has been overlooked in these studies is that some acids trigger chemoreceptors that initiate non-nociceptive behaviours (Derby and Steullet, 2001; Derby et al., 2016). For example, citric acid is a potent feeding stimulant for some fish species invoking strong exploratory behaviour (Kasumyan and Doving, 2003; Diggles et al., 2017), while dilute (10%) acetic acid (vinegar), has a distinctive taste and smell which is a potent olfactory/gustatory stimulant for humans and arthropods (e.g. Drosophila, see Landolt et al., 2012; Joseph and Carlson, 2015). It is well-known that crustaceans will attempt to eat a wide variety of foods, even foods laced with compounds that are considered by humans to be deterrents (Aggio et al., 2012; Puri and Faulkes, 2015) including unpalatable acids (Derby et al., 1984), although it appears that no research has been published to date in the scientific literature on the chemosensory attractiveness of acetic acid for fish or crustaceans. Hence the “escape” behaviours observed by Elwood et al. (2017) (and for that matter, the “pain”/grooming behaviours reported by Barr et al., 2008) may actually be food seeking exploratory behaviours in animals where chemoreceptors (particularly olfactory and gustatory receptors) have been triggered by vinegar (Table 2). Given the unsubstantiated assumption that exposure to dilute acetic acid/vinegar via the water “only stimulates nociceptors”, research is needed to determine whether acetic acid/vinegar is also a chemosensory (particularly olfactory and gustatory) stimulant for fish and crustaceans, as it is for insects and humans. Only a narrow range of decapod species have been studied to date The literature in Tables 1 and 2 summarizes studies that have been undertaken on ten species of decapod crustaceans. Even within such a small range of species, a wide range of responses to putatively noxious stimuli have been reported, with few consistent scientifically valid outcomes except for the apparent lack of response to the supposedly nocigenic compound capsaicin (Puri and Faulkes, 2015; Elwood et al., 2017). With many different classes and orders of crustaceans (e.g. Martin and Davis, 2001) being utilized by humans, and no exploration of nociception or welfare outside of a miniscule proportion of the Order Decapoda, it is clear that the scientific literature on this subject is limited and immature. How high should the scientific bar be set? In the context of the above, there is a conspicuous lack of scientifically valid evidence of pain in crustaceans at this time, including several examples of nonreplicable results and/or overinterpretation of behavioural responses to assumed aversive stimuli (Boutron and Ravaud, 2018) in situations where it is questionable whether nociception has even occurred (Rose et al., 2014; Puri and Faulkes, 2015; Stevens et al., 2016). Hence, in a strict sense there is currently no scientifically valid reason to change the welfare status of any crustaceans. The scientific problems surrounding methodologies for confirming the presence and activation of nociceptors and initiating nociception, physiological, and behavioural definitions, the problem of inconsistent results that are difficult to interpret, as well as of other perplexing issues such as the significance of autotomy and regrowth of limbs must all be resolved, before this field of research can move forward. The question of “when is the right time?” to protect crustaceans under welfare legislation is a loaded one, and will depend on the height at which the “evidential scientific bar” is set (Birch, 2017). Consequently, there may never be a “right time” based on scientifically valid criteria, if these criteria cannot be met. In the field of aquatic animal welfare this difficult problem usually leads to discussions relating to invoking “the benefit of the doubt”, and the precautionary principle. Birch (2017) discussed the application of the precautionary principle to the problem of animal consciousness/sentience (one of the pre-requisites for pain). He defined the precautionary principle in the context of animal sentience as follows: Where there are threats of serious, negative animal welfare outcomes, lack of full scientific certainty as to the sentience of the animals in question shall not be used as a reason for postponing cost-effective measures to prevent those outcomes. Birch (2017) then outlined how decisions to enact the precautionary principle in this context come down to two main decision criteria, first a burden of scientific proof: When there is a live scientific hypothesis that posits a causal relationship between human action and a seriously bad outcome, we should set an intentionally low evidential bar for the acceptance of that hypothesis in the context of formulating policy. Second, there is a decision rule: Once we have sufficient evidence of a threat of a seriously bad outcome, we should act, in a timely and cost-effective manner, to prevent that outcome. The implication is that the goal of preventing the seriously bad outcome deserves sufficient priority that, once the evidential bar is cleared, it is inappropriate to delay action further. Based on these decision criteria, Birch (2017) pointed out that the key to precautionary reasoning is to delimit carefully what constitutes “seriously bad outcomes”. This is because if there is no imminent threat of seriously bad outcomes, there is no urgency to lower the “evidential scientific bar”. In the context of holding live crustaceans captive at a research facility (for example) the imminent threat of “seriously bad outcomes” must be compared with a valid benchmark—which if a nature-based definition of welfare is used (Diggles et al., 2011), could well be the living conditions of those animals in the wild. Wild caught prawns (Penaeus spp.) held at an aquaculture research facility as broodstock will be used here as a case study to examine their welfare needs (see Specific manipulations used in a prawn research facility for a more detailed case study). Prawns are important components of the natural food chain in the wild and are subject to high predation pressure (Salini et al., 1990). Removing prawns from their natural environment and rearing them in captivity for use as broodstock thus releases them from external predation pressure (though they do cannibalize if given insufficient food), and therefore reduces the majority of the imminent threat of “seriously bad outcomes” for those individuals. Given the high economic value of broodstock and expense of research, much effort is then expended to ensure any experimental manipulations are not compromised by artefacts from captive holding, which is why husbandry procedures aim to maximize survival of captive prawns by optimizing water quality, food supply and biosecurity to control and/or eliminate naturally occurring diseases. In effect, the chances of “seriously bad outcomes” for these experimental animals are extremely low, far less than for prawns in their natural environment, being restricted to only the specific experimental manipulations they are given. For a more detailed assessment of the welfare issues related to such experimental manipulations on prawns, see the section on Specific manipulations used in a prawn research facility. The problem with the “benefit of the doubt” or “precautionary principle” is that by inviting a lowering of scientific standards, it can be misused or lead to undesirable, unintended consequences. It has been suggested that the precautionary principle is basically a socio-political manoeuvre that effectively excludes valid science from policy, and that allowing the “benefit of the doubt” is not benign, nor may it be the best way to protect aquatic animal welfare (Rose et al., 2014). Perhaps this is why Birch (2017) insists that there should not be any lowering of scientific standards, at least initially when considering new taxa. The aim of his appeal to uphold “normal scientific standards” revolves around a premise that “a low evidential bar should not be applied when inferring the presence of credible indicators of sentience. It should instead be applied at a later stage: it should be applied when making a precautionary attribution of sentience on the basis of a single credible indicator, and when extrapolating across a whole order from a single species. There should not be any lowering of standards with regard to the methodology of experiments, or with regard to the analysis of experimental data” (Birch, 2017). Birch (2017) highlighted the problems that surface if scientific standards are lowered with his example of reversal of the burden of proof to assume an animal is sentient unless there is conclusive evidence otherwise. The major problems with this were identified by him as “being unscientific or anti-scientific”, and that such a position would make “the science of animal sentience… more or less irrelevant to the scope of animal protection law: all animals would be assumed sentient unless proven otherwise, and it is hard to see how research could prove otherwise”, leading to “inclusion of nematodes and insects within the scope of animal protection legislation, creating significant practical obstacles to biomedical research” (Birch, 2017). On initial assessment, to the scientist, decision makers or interested layperson the prospects of this reversal of burden of proof actually occurring might appear remote (despite the wishes and activities of animal rights organizations), however in the field of aquatic animal welfare it is already happening given the nascent uncritical acceptance of the “animal pain” criteria summarized by Sneddon et al. (2014) (https://www.hakaimagazine.com/features/fish-feel-pain-now-what/). These criteria essentially lower the “evidential bar” needed to claim pain in fish and invertebrates to a point where insects measured against the criteria are already being considered as “probably sentient” (Barron and Klein, 2016, Klein and Barron, 2016), and robots also fulfil many of the criteria (Adamo, 2016a, b). It is known, however, that insects are descended from crustaceans (Regier et al., 2010). Therefore, even speculation about the possible existence of consciousness and “pain” in insects (Fischer, 2016) does not require that it also occurs in crustaceans (Puri and Faulkes, 2015), as it is possible that nociception evolved in insects after their split from crustaceans, although some authors group all arthropods together in this regard (Mallatt and Feinburg, 2016). Given that crustaceans and other arthropods also commonly and naturally exhibit behaviours such as autotomy [loss and regrowth of entire claws and limbs (Maruzzo et al., 2005; Patterson et al., 2007, Appel and Elwood, 2009a; Kotsyuba et al., 2010; Dyuizen et al., 2012; Magee and Elwood, 2013], this brings the validity of many of the criteria presented by Sneddon et al. (2014) into serious question if the word pain (as used and understood by humans) is to be applied to crustaceans. Certainly, it is difficult to understand what adaptive advantage an ability to “feel pain” would confer to a crustacean that spontaneously sacrifices appendages by autotomy during normal development or when under stress (Patterson et al., 2007; Kotsyuba et al., 2010), only for those appendages to be grown back in later moults (Maruzzo et al., 2005). In fact, Patterson et al. (2007) found that induced autotomy of a claw in edible crabs (C. pagurus) was no more stressful than handling alone, based on measurements of haemolymph glucose and lactate. Indeed, autotomy in crustaceans (which can occur in the absence of mechanical loading), appears fundamentally different to autotomy which follows mechanical damage in lizards or mammals such as African spiny mice (Acomys spp., see Seifert et al., 2012). The same can be said for insects that exhibit behaviours such as eating their own innards and continuing feeding while being eaten (Adamo, 2016a). Accepting pain (in any human understanding of the word and its meaning) for animals that can naturally undergo such behaviours seems bizarre. Another problem with the definitions for pain presented by Sneddon et al. (2014), is that the requirement for “central processing in the brain” does not specify the minimal level of brain complexity needed to achieve the processing required for a “pain” emotional response to nociception. In discussions of insect and crustacean awareness, it has been suggested that while animals with decentralized nervous systems (e.g. jellyfish, annelids) do not meet the criteria for phenomenal consciousness, the number of neurons in those invertebrates with a central ganglia that could be considered as a primitive, crude brain is unimportant (Klein and Barron, 2016). Others, however, consider this position extremely liberal and that it is unscientific to assume that small neuronal numbers are unimportant for phenomenal consciousness, an important pre-requisite for pain (Adamo, 2016b). Indeed, sceptics note if neuron number is not considered important (a honey bee has < 1 million neurons, a mouse 68 million and a human 86 billion, see Klein and Barron, 2016) and “subjective experience” is defined solely as an ability to react to the environment in any purposeful way (e.g. learning and motivated behaviours, see Sneddon et al., 2014), then speculation that nematodes (Tobin and Bargmann, 2004), insects (Adamo, 2016b), and crustaceans may be conscious is likely to be upheld based on existing claims, suggesting that by this definition it would be entirely plausible that they “have feelings” and therefore potentially “feel pain” (Fischer, 2016; Elwood, 2017). It should be made clear that the word “feeling” is consistently misused or unqualified in discussions of animal pain or consciousness. There is abundant evidence that conscious feelings are distinct from unconscious emotions, just as conscious pain is distinct from unconscious nociception (Le Doux, 2012; Rose et al., 2014). In fact, unconscious emotions are better designated as actions of unconscious neural survival circuits, which are doubtless present in all forms of animals (Le Doux, 2012). On the other hand, misuse and loosely defining of the word “feeling” would also mean that the same definitions could be used to argue (despite the protests of some, see Elwood, 2017), that robots can have subjective experiences, which could include pain (see Adamo, 2016b). This possibility puts the argument about the validity of the pain criteria of Sneddon et al. (2014) into perspective and brings a whole range of new problems into welfare science, including such issues as the thousands (or millions) of allegedly “sentient beings” you would kill while driving your car across the countryside (Fischer, 2016). In other words, in embracing such liberal definitions, taking them to their inevitable conclusion, and then suggesting that regulatory changes are required to protect these “sentient beings” from possible harm, there is a risk of making welfare science effectively meaningless (Fischer, 2016). This is exactly why some scientists advocate for more pragmatic, evidence-based approaches to fish and crustacean welfare that are based on objective assessments of measureable, well-established physiological stress and health-related indices, rather than on speculation as to what fish, crustaceans or other invertebrates may (or may not) be feeling (Diggles et al., 2011; Rose et al., 2014; Stevens et al., 2016). Because of the many problems and uncertainties regarding crustaceans (and fish, nematodes and insects) summarized above, and in the absence of imminent threats of “seriously bad outcomes” (see the section on Specific manipulations used in a prawn research facility), there is no urgency to lower the “evidential scientific bar” for these groups of animals at this time, and in the case of decisions affecting regulation of research and food production industries, strong arguments can be made to maintain the highest scientific standards possible when considering such questions, particularly regarding the methodology of experiments and analysis and interpretation of data (Birch, 2017). Otherwise if the precautionary principle is enacted too early, a paradoxical (“Catch 22”) situation may arise and restrictive animal ethics requirements may hinder or even prevent studies that could provide the data needed to solve the unresolved scientific questions (Rose et al., 2014). Anthropomorphic interpretations of stress in crustaceans Unlike the issue with sentience and pain, stress is well established and well defined and can be assessed in fish and crustaceans by measuring hormonal, biochemical, or other changes to the normal physiological state of an animal as it tries to adapt to changing environmental conditions (Stoner, 2012). Stress modulates the immune response and is one of the key factors influencing health and disease states (Adamo, 2012). Therefore, there has been a large amount of research defining useful stress indicators for measuring crustacean welfare (e.g. Paterson et al., 2001, 2007). Stress in crustaceans can be assessed by measuring changes in immune function such as total haemocyte counts and prophenoloxidiase markers (Soderhall and Cerenius, 1992). Crustaceans subject to stressors also release biogenic amines such as epinephrine and serotonin (Adamo, 2012) or hormones such as crustacean hyperglycaemic hormone (CHH), which elevates haemolymph glucose concentrations in response to stressors such as moulting, emersion, and eyestalk ablation (Chang et al., 1999). Measurements of haemolymph lactate have also been used to measure responses to exercise stress (Booth and McMahon, 1985) as well as natural (Wang et al., 2018) or anthropogenic stressors (Patterson et al., 2007; Bakke and Woll, 2014), while measurements of other haemolymph parameters such as metabolites (urea, ammonia, nitric oxide) or even total protein concentration are also useful (Kotsyuba et al., 2010; Dyuizen et al., 2012; Bernardi et al., 2015), as are assessments of impaired reflexes (Stoner, 2012). While the fact that crustaceans become stressed in some situations is not controversial, several authors have attempted to attribute emotional states to stressful situations, including Elwood and Adams (2015) who exposed shore crabs (C. maenas) to electric shocks and attempted to relate a stress marker (increased haemolymph lactate in shocked crabs) to “a pain experience”. Recent studies have also reported that responses of crayfish (P. clarkii) to prolonged (30 min) exposure to stressful stimuli (electric fields) raised serotonin levels in the brain and resulted in avoidance of light for 30–90 min (Fossat et al., 2014). Inhibition of the behaviour by injection of serotonin antagonists, and the lack of involvement of dopamine (Fossat et al., 2015) led to claims that the behaviour was similar to anxiety in humans as defined by “a behavioural response to stress, consisting in lasting apprehension of future events”, and speculation that crayfish have emotions (i.e. conscious feelings) (Fossat et al., 2014, 2015). Similarly, Bacque-Cazenave et al. (2017) studied aggression and fighting and dominance hierarchies in P. clarkii and found that serotonin levels increased in losers, which often continued to be attacked by dominant crayfish leading to them to speculate that the “hostile behaviour resembled psychological harassment in humans”. However, this seems extremely speculative, especially considering that the ongoing “harassment” observed may simply be an experimental artefact of confinement of crayfish in small aquaria (leading to an inability of the loser to leave the home range of the dominant crayfish), while previous studies in other species of crayfish (Astacus astacus) and squat lobsters (Muninda quadrispina) linked increased serotonin to aggression, not anxiety (Antonsen and Paul, 1997; Huber and Delago, 1998). The claims of Bacque-Cazenave et al. (2017) and Fossat et al. (2014, 2015) appear to be classic cases of anthropomorphism, as not only are some of their results inconsistent with previously published work, there is absolutely no empirical justification for casual, inappropriate, and misleading use of language from the field of human psychology to explain crayfish behaviour. The recent increase in use of fish and invertebrates as model organisms by researchers who usually study human health-related topics using more traditional mammalian laboratory animals (such as mice), seems to have led to increased attribution of human-like traits to the subject animals. This is anthropomorphism which, when applied uncritically in the context of animals that are a large distance in evolutionary and morphological terms from humans, reduces the credibility of aquatic animal welfare science (Rose, 2007; Browman and Skiftesvik, 2011; Boutron and Ravaud, 2018). It seems in the field of aquatic animal welfare science, more than most others, there are “many examples of selective preference of scholars and audiences for exciting research results over metaphorical buckets of ice water” (Allen-Hermanson, 2017). The reality is that, using the above instances as an example, dominance hierarchies in crayfish are a behaviour evolved by natural selection to assist in partitioning of sometimes scarce natural resources (e.g. food, shelter) and determine gene progression through mate selection (Stocker and Huber, 2001). Size, past experience, visual cues, olfactory cues, previous social history, moult stage, crayfish density, and a range of other factors influence the outcomes of fighting behaviours in crayfish (Rubenstein and Hazlett, 1974; Schneider et al., 2001; Stocker and Huber, 2001; Cook and Moore, 2009). Even though some of the hormonal mechanisms associated with such behaviour may be conserved between crayfish and humans, crayfish did not evolve from humans so attribution of the human feeling state to the crayfish is highly speculative and must be viewed with extreme caution. The nature of the extent of such speculation could also be illustrated by way of asking whether adult crayfish might experience feelings like sadness, remorse, or guilt while cannibalizing their offspring. Specific manipulations used in a prawn research facility A case study examining welfare of broodstock prawns held in a research facility is of direct relevance to researchers and the prawn (shrimp) farming industry in the Americas and Asia-Pacific regions. This study is performed to clarify whether these captive experimental animals may be exposed to “seriously bad outcomes” requiring intervention from welfare legislation during typical experimental manipulations, including electro ejaculation, pleopod clipping, haemolymph sampling, eye stalk ablation, tagging, stunning, and euthanasia. Electro-ejaculation of male prawns as part of reproductive management For research into methods of improving production of larval and juvenile prawns, manual and electro-ejaculation methods are used to obtain sperm from male broodstock prawns for fertilization and/or sperm assessments (e.g. Sellars et al., 2013). Manual methods involve applying slight pressure to the prawn body to examine whether sperm are released. Electro-ejaculation involves placing two electrodes on each side of the thelycum and applying controlled low (<10 mV) currents from 3 to 9 V to initiate muscle contractions that release spermatophores (Soundarapandian et al., 2013). Single application of the procedure has no notable side effects, but repeated forced ejaculation may be associated with increased frequency of melanization of the reproductive tract (Braga et al., 2018). If results from Appel and Elwood (2009a, b), Elwood and Appel (2009), or Elwood and Adams (2015) were accepted uncritically, the electro-ejaculation process could possibly be considered “painful” due to the use of electric shocks strong enough to trigger muscle contractions. However, as pointed out in the section on Experiments employing electric shocks: technical shortcomings and misinterpreted results, in the absence of evidence of nociceptor activation triggered by such stimuli, there is no scientifically valid evidence that the process of electro-ejaculation in prawns causes nociception, hence it cannot be considered a “potentially painful process”. Instead, the process is likely to result in stress equivalent to exercise (from muscle contraction) and handling, hence suitable precautions to reduce handling and exercise stress should be used to minimize any adverse effects from electro-ejaculation procedures. It is notable that in other areas of veterinary medicine and animal science, it is common to collect semen from domestic ruminants using electro-ejaculation without sedation or anaesthesia. Pleopod clipping for sample acquisition Clipping of between 10 and 75% of the distal part of the pleopods of prawns or lobsters is usually done using sterilized scissors to obtain samples of tissue for genetic analysis, or to test for moult stage, or disease agents as part of routine disease screening (Sahul Hameed et al., 2005) or development of specific pathogen-resistant (SPR) or specific pathogen-free (SPF) stocks (Wyban, 1992; Lightner, 2011). Pleopods are poorly vascularized appendages, and due to the nature of the appendage the majority of the tissue sampled is of ectodermal origin (i.e. cuticle or carapace) that is naturally lost and regenerated each moult, meaning that the clipped pleopod is quickly regenerated. While clipping causes mechanical damage, it is not known whether the pleopods of crustaceans have nociceptors. However, there are many studies of stress markers and immune parameters in crustaceans that show that pleopod clipping is no more stressful than handling alone (e.g. Paterson et al., 2001). Because of the minimal invasiveness and natural redundancy (due to moulting) inherent with this procedure, pleopod clipping should be considered a minor manipulation provided suitable precautions to reduce stress from handling and exercise are employed. Injections and haemolymph sampling Sampling of haemolymph from juvenile and adult prawns is usually done using suitable gauge needles (e.g. 25–27 G hypodermic needles) for a variety of tests (e.g. measurement of haemolymph immune or stress parameters). Sterile needles are inserted through an arthrodial membrane previously swabbed with ethanol, haemolymph drawn into the syringe (which is often pre-primed with anticoagulant), then the needle is removed and the sample is processed. The haemolymph of healthy crustaceans exhibits rapid clotting so the small needle puncture wound closes and heals rapidly. As for pleopods, it is not known whether the arthrodial membranes of crustaceans have nociceptors, however there are many studies of stress markers and immune parameters in crustaceans that show haemolymph sampling is another minor manipulation that is likely to represent little more than the stress usually encountered by handling alone (e.g. Paterson et al., 2001). Again, this example is analogous to human and veterinary medicine; analgesics are not employed when taking a blood sample. Eye stalk ablation of female prawns to stimulate maturation and spawning The eyes of prawns contain a gland called the X organ, which secretes hormones including gonad inhibiting hormone (GIH) that can inhibit maturation of the eggs and sperm of broodstock prawns. Eyestalk ablation is a process involving cauterizing or cutting of one of the eyestalks of a mature (>10-month old) broodstock female prawn using flame-sterilized hot forceps or ligation in order to alter the hormonal balance by reducing production of the inhibition hormones, allowing the prawn to undergo the final stages of maturity in captivity (Uawisetwathana et al., 2011). Use of flame-sterilized forceps ensures a rapid (< 2 s) ablation process as well as cauterization and sterilization of the wound. This process was developed in the 1970s to enable larval production to occur over a narrower time frame, and has allowed the prawn farming industry to develop worldwide as the procedure ensures reliable and timely hatchery production schedules (http://www.fao.org/fishery/culturedspecies/Penaeus_monodon/en). Without eyestalk ablation, spawning is not guaranteed in some prawn species (e.g. Penaeus monodon), making it very difficult to structure research or breeding programs. It is also notable that, like autotomy and regrowth of limbs, some broodstock prawns may grow back ablated eyes over time (<6 months) if permitted to do so (Desai and Achuthankutty, 2000), although this may not be a common occurrence. A study by Taylor et al. (2004) suggesting eyestalk ablation in Penaeus vannamei was stressful (based on observations of erratic or spiral swimming) was not controversial at the time as it is well-known that crustaceans can experience physiological stress. Later, Diarte-Plata et al. (2012) repeated the same study on a different host species but used the word “painful” instead of stressful, based on an unsubstantiated assumption that tail flicking, rubbing, and non-sheltering were evidence of “pain”. However, tail flicking escape responses and rubbing are not specific to nociception and have not been validated as evidence of “pain” in crustaceans (Puri and Faulkes, 2010; Rose et al., 2014). Furthermore, rubbing or tail flicking do not always occur when prawns have their eyestalks ablated (B. K. Diggles, personal observations). While it would be expected that heat from flame-sterilized forceps would activate nociceptors (Puri and Faulkes, 2015), responses to heat occurred in around 2 s in their study (Puri and Faulkes, 2015), hence rapid action (< 2 s for hot forceps) would reduce the duration of the noxious stimulus during ablation. Puri and Faulkes (2015) also stated in their discussion that their results “suggests that crayfish have nociceptors specialized to detect noxious high temperature stimuli. Nevertheless, whether a species has nociceptors or not is not conclusive evidence that the species feels pain”. So while eyestalk ablation procedures lasting >2 s may induce nociception (Puri and Faulkes, 2015) and be stressful, nociception does not necessarily lead to pain, and stress does not equal pain (Stevens et al., 2016). However, if the stress involved with the handling and ablation process is to be minimized, the option to use a topical anaesthetic such as xylocaine (Taylor et al., 2004) may be considered appropriate if its efficacy has been demonstrated and its use does not interfere with other variables to be measured in the experiment. The option to replace ablation with alternative procedures (such as natural spawning) may not be appropriate for those prawn species that do not respond well to environmental manipulations aimed at inducing spawning (e.g. P. monodon). Furthermore, given that the number of eggs obtained from non-ablated P. monodon is significantly lower than that obtained by ablated prawns (Uddin and Rahman, 2015), alternatives to ablation for such species would require use of many more broodstock prawns, which contravenes a fundamental welfare principal of reduction of the numbers of animals impacted by human manipulations. Tagging by injecting elastomer implants, PIT tags, ring tags, or glued tags To facilitate identification of individual prawns, various tagging procedures can be used, including injecting elastomer implants, injecting Passive Integrated Transponder (PIT) tags, gluing tags onto the carapace, or placing circular “bird-band” tags around the eye stalk. The issues relating to the injection of elastomer implants or PIT tags into the muscle of the abdomen are virtually identical to those of injection and haemolymph sampling (see section on Injections and haemolymph sampling) with sterilization of equipment reducing chances of bacterial infection while the small puncture wound from the needle closes and heals rapidly around the tag such that there are no significant health complications with the tagging process when done correctly. The advantage of the injection of tags is that they can persist in growing animals between moults, reducing handling requirements compared to simply gluing tags onto the carapace or use of bird-band tags around the eyestalk. Neither of the latter procedures are invasive, but both require repeated handling of prawns over time and potential loss of data as the tags are shed with the exoskeleton during each moult. Given that repeated handling is likely to be as stressful (or more so) than the process of injection of the tags, there would seem to be no valid reason to require replacement of invasive tagging procedures with non-invasive ones if invasive tagging (elastomer tags, PIT tags) provides advantages for a given experimental design. Moreover, in veterinary medicine, analgesics are not employed when inserting PIT tags in cats or dogs. Stunning and euthanizing prawns (ice slurry, anaesthetics) For purposes of producing healthy prawn stocks free of pathogens, best practice biosecurity arrangements are commonly employed, which involves destructive disease testing of broodstock prawns after spawning to ensure that the progeny stocks will be free from diseases that impact their future health (Wyban, 1992). While this may appear wasteful, euthanizing the relatively small numbers of short-lived broodstock (the lifespan of penaeid prawns seldom exceeds 2 years) can ensure better health outcomes for the millions of progeny animals produced. As such, this approach is considered best practice because control of disease is considered the highest priority for overall welfare of the captive prawn population (Lightner, 2011). Euthanizing prawns is usually done using an ice slurry (a 0°C mixture of >2 parts crushed ice to 1 part salt water) or, very occasionally, anaesthetic (Aqui-S, isoeugenol active). Submerging adult broodstock P. monodon in an ice slurry for >1 min is an effective way to euthanize these large prawns that are reared at tropical water temperatures (29°C) (B. K. Diggles, unpublished data). Ice slurry is also effective for sedating or euthanizing large mud crabs (Scylla serrata), see https://www.youtube.com/watch?v=i-wkdRmdrrE. Broodstock prawns placed in an ice slurry appeared to be initially stunned by the temperature difference, almost ceasing activity and exhibiting no more than two or three tail flips in the first 30 s, before ceasing movement altogether. However, if prawns were removed from the ice slurry before 1 min and placed back in 29°C seawater, it was possible for them to recover to normal activity over the next 30 min or so (B. K. Diggles, unpublished data). In contrast, use of the recommended commercial concentration of Aqui-S anaesthetic (17 ml per 1000 l or 17 mg/l) in 29°C seawater did not induce noticeable anaesthesia after 60 min (B. K. Diggles, unpublished data), resulting in numerous tail flip escape attempts whenever the prawns were touched. This was not surprising, as Coyle et al. (2005) found that MS222 and 2-phenoxyethanol were ineffective while isoeugenol was effective for anaesthetizing freshwater prawns (Macrobrachium rosenbergii), but only at concentrations 5–10 times higher (100–200 mg/l) than those used on finfish. It is notable that exposure to such high concentrations of anaesthetics poses a health and safety risk to researchers, particularly if they are exposed for long periods. Given the fact that low temperatures do not activate nociceptors in some crustaceans (Puri and Faulkes, 2015), ice slurry appears a highly effective, non-chemical method for humanely stunning and euthanizing prawns and also other tropical crustaceans that should be considered superior to the use of anaesthetics, which besides having questionable efficacy for crustaceans (and occupational health and safety risks for users when used at high concentrations), require disposal of the chemicals into the waste water stream (usually down the sink or drain). The latter contributes to unintended downstream effects on the welfare of wild fishes and shellfish as the chemicals enter the waste water stream and, ultimately, the environment as organic contaminants (Diggles et al., 2011, 2017). Conclusions Given the critical flaws in design and interpretation of several crustacean “pain” studies, and the inconsistent results obtained from the extremely narrow range of crustacean species studied to date, acceptance of claims of pain for these animals, even as a precautionary measure, would represent acceptance of a much lower evidential bar than is usually dictated by normal scientific standards. The proposed criteria for animal pain (Sneddon et al., 2014) effectively set “the bar” for pain and sentience so low that it is impossible to have confidence that the behaviours observed in many experiments are even due to nociception, let alone in any way analogous to how the word pain is used and understood by humans. The present situation may be leading to circumstances whereby weak scientific literature on crustacean welfare is being used by decision makers to justify additional, possibly unnecessary constraints on scientific research that uses crustaceans, imparting significant costs (in time and experimental flexibility) to scientific programs (and potentially also food production industries), which would likely exceed any minor beneficial changes in welfare status that may (or may not) accrue to the experimental animals. There are already several examples of non-replicable results in the literature on alleged “pain” in crustaceans (Puri and Faulkes, 2010) and/or overinterpretation of behavioural responses to presumed aversive stimuli (Puri and Faulkes, 2015; Stevens et al., 2016). In the crustacean welfare literature to date, more biologically plausible and parsimonious alternative explanations to conclusions of crustacean “pain” are often being ignored (e.g. Rose et al., 2014; Puri and Faulkes, 2015; Stevens et al., 2016; Allen-Hermanson, 2017; Diggles et al., 2017; Key et al., 2017). This leads to the conclusion that the scientific literature on this subject is immature, and does not include scientifically valid evidence of pain in crustaceans. Appeals for application of the precautionary principal extend the “benefit of the doubt”, which requires that the “scientific evidential bar” for formulating policy be lowered when there is a scientific hypothesis that posits a causal relationship between human action and “seriously bad outcomes” (Birch, 2017). However, in the absence of imminent threats of “seriously bad outcomes” (such as in the case study herein related to prawn aquaculture research), there is no urgency to lower the “evidential scientific bar” for including crustaceans under animal welfare legislation at this time. If the precautionary principle is enacted too early, a paradoxical “Catch 22” situation may arise in that restrictive animal ethics requirements may hinder or even prevent the high quality studies needed to provide the data to solve unresolved scientific questions. Furthermore, animal ethics regulations for use of animals for scientific purposes in some countries appear to have no mechanism for reviewing or withdrawing existing protections once enacted (e.g. NHMRC, 2013). This means that inclusion of new taxa into welfare legislation is often a one way street that should, therefore, require considerable weight of evidence that should require no lowering of scientific standards with regard to the design, methodology, and replication of experiments, or the analysis and interpretation of data (Birch, 2017). It remains the task of the broader scientific community to ensure that the highest scientific standards are upheld in this regard. Alternatively, if the precautionary principle is used to justify inclusion of certain animal groups under welfare regulation, policy-makers should be obliged to regularly review the scientific criteria used to justify such decisions, and/or include provision for “sunset clauses” for the withdrawal of such taxa from protection if more robust scientific data becomes available at a later date which invalidates the preliminary results used to trigger the precautionary decision. High quality science is expected from national research institutions studying, for example, problems in crustacean food production industries that may underpin a nation’s food security. This does not mean that crustaceans should be used (or abused) for such activities carelessly or indiscriminately, but it is important that the quality of any research that may influence regulatory decisions that constrain such national research institutions (or food production industries) should be equally high. In the case of decisions affecting regulation of crustacean research, strong arguments can be made to maintain the highest scientific standards possible when considering such questions. Acknowledgements This work was initiated and partially funded by an independent investigation of the conditions associated with the housing and welfare of prawns and other crustaceans held at a national research facility in Australia. The author thanks D. Stevens, J. Rose, and four anonymous referees for helpful comments on drafts of the manuscript. Declaration of interest The author is an independent scientist who works on topics related to aquatic animal health and welfare for a wide range of government, aquaculture, and fisheries institutions in the Asia-Pacific region. He acts as a reviewer for several aquatic animal health journals and has developed, and administers publically available aquatic animal welfare tools such as the ikijime.com website and ikijime tool phone applications in his own time, at his own expense. The author has previously published articles identifying weaknesses in the science underlying aquatic animal welfare. References Adamo S. A. 2012 . The effects of the stress response on immune function in invertebrates: an evolutionary perspective on an ancient connection . Hormones and Behavior , 62 : 324 – 330 . Google Scholar Crossref Search ADS PubMed Adamo S. A. 2016a . Do insects feel pain? A question at the intersection of animal behaviour, philosophy and robotics . Animal Behaviour , 118 : 75 – 79 . Google Scholar Crossref Search ADS Adamo S. A. 2016b . Subjective experience in insects: definitions and other difficulties . Animal Sentience , 2016 : 127 . Aggio J. F. , Tieu R. , Wei A. , Derby C. D. 2012 . Oesophageal chemoreceptors of blue crabs, Callinectes sapidus, sense chemical deterrents and can block ingestion of food . The Journal of Experimental Biology , 215 : 1700 – 1710 . Google Scholar Crossref Search ADS PubMed Allen-Hermanson S. 2017 . Battlefish contention. Allen-Hermanson on Woodruff on Teleost Consciousness . Animal Sentience , 2017 : 015 . Antonsen B. L. , Paul D. H. 1997 . Serotonin and octopamine elicit stereotypical agonistic behaviors in the squat lobster Munida quadrispina (Anomura, Galatheidae) . Journal of Comparative Physiology A , 181 : 501 – 510 . Google Scholar Crossref Search ADS Appel M. , Elwood R. W. 2009a . Motivational trade-offs and potential pain experience in hermit crabs . Applied Animal Behaviour Science , 119 : 120 – 124 . Google Scholar Crossref Search ADS Appel M. , Elwood R. W. 2009b . Gender differences, responsiveness and memory of a potentially painful event in hermit crabs . Animal Behaviour , 78 : 1373 – 1379 . Google Scholar Crossref Search ADS Bacque-Cazenave J. , Cattaert D. , Delbecque J.-P. , Fossat P. 2017 . Social harassment induces anxiety-like behaviour in crayfish . Scientific Reports , 7 : 39935. Google Scholar Crossref Search ADS PubMed Bakke S. , Woll A. K. 2014 . Evaluation of three handheld lactate meters for measuring haemolymph L-lactate in decapods crustaceans . Journal of Shellfish Research , 33 : 69 – 76 . Google Scholar Crossref Search ADS Barr S. , Elwood R. W. 2011 . No evidence of morphine analgesia to noxious shock in the shore crab, Carcinus maenas . Behavioral Processes , 86 : 340 – 344 . Google Scholar Crossref Search ADS Barr S. , Laming P. R. , Dick J. T. A. , Elwood R. W. 2008 . Nociception or pain in a decapod crustacean? Animal Behaviour , 75 : 745 – 751 . Google Scholar Crossref Search ADS Barron A. B. , Klein C. 2016 . What insects can tell us about the origins of consciousness . Proceedings of the National Academy of Science of the United States of America , 113 : 4900 – 4908 . Google Scholar Crossref Search ADS Bergamo P. , Maldonado H. , Miralto A. 1992 . Opiate effect on the threat display in the crab Carcinus mediterraneus . Pharmacology Biochemistry and Behavior , 42 : 323 – 332 . Google Scholar Crossref Search ADS Bernardi C. , Baggiani L. , Tirloni E. , Stella S. , Colombo F. , Maria Moretti V. et al. 2015 . Hemolymph parameters as physiological biomarkers in American lobster (Homarus americanus) for monitoring the effects of two commercial maintenance methods . Fisheries Research , 161 : 280 – 284 . Google Scholar Crossref Search ADS Birch J. 2017 . Animal sentience and the precautionary principle . Animal Sentience , 2017 : 017 . Booth C. E. , McMahon B. R. 1985 . Short communication: lactate dynamics during locomotor activity in the blue crab, Callinectes sapidus . Journal of Experimental Biology , 5 : 207 – 465 . Boutron I. , Ravaud P. 2018 . Misrepresentation and distortion of research in biomedical literature . PNAS , 115 : 2613 – 2619 . Google Scholar Crossref Search ADS PubMed Braga A. , Lopes D. , Magalhães V. , Klosterhoff M. C. , Romano L. A. , Poersch L. H. , Wasielesky W. 2018 . Hemocytic melanization in shrimp spermatophores . Aquaculture , 486 : 64 – 67 . Google Scholar Crossref Search ADS Broom D. M. 2013 . The welfare of invertebrate animals such as insects, spiders, snails and worms. In Animal Suffering: From Science to Law, International Symposium, pp. 135–152 . Ed. by van der Kemp T. A. , Lachance M. . Éditions Yvon Blais , Paris . Browman H. I. , Skiftesvik A. B. 2011 . Welfare in aquatic organisms – is there some faith- based HARKing going on here? Diseases of Aquatic Organisms , 94 : 255 – 257 . Google Scholar Crossref Search ADS PubMed Chang E. S. , Chang S. A. , Keller R. , Reddy P. S. , Snyder M. J. , Spees J. L. 1999 . Quantification of stress in lobsters: crustacean hyperglycemic hormone, stress proteins, and gene expression . American Zoologist , 39 : 487 – 495 . Google Scholar Crossref Search ADS Cook M. E. , Moore P. A. 2009 . Communication networks and loser effects interact to influence the outcome of aggressive interactions in crayfish Orconectes rusticus . Behaviour , 146 : 263 – 281 . Google Scholar Crossref Search ADS Coyle S. D. , Dasgupta S. , Tidwell J. H. , Beavers T. , Bright L. A. , Yasharian D. K. 2005 . Comparative efficacy of anesthetics for the freshwater prawn Macrobrachium rosenbergii . Journal of the World Aquaculture Society , 36 : 282 – 290 . Google Scholar Crossref Search ADS Derby C. D. , Kozma M. T. , Senatore A. , Schmidt M. 2016 . Molecular mechanisms of reception and perireception in crustacean chemoreception: a comparative review . Chemical Senses , 41 : 381 – 398 . Google Scholar Crossref Search ADS PubMed Derby C. D. , Reilly P. M. , Atema J. 1984 . Chemosensitivity of lobster, Homarus americanus, to secondary plant compounds: unused receptor capabilities . Journal of Chemical Ecology , 10 : 879 – 892 . Google Scholar Crossref Search ADS PubMed Derby C. D. , Steullet P. 2001 . Why do animals have so many receptors? The role of multiple chemosensors in animal perception . Biological Bulletin , 200 : 211 – 215 . Google Scholar Crossref Search ADS PubMed Derbyshire S. W. G. 2016 . Fish lack the brains and psychology for pain. Commentary on Key on Fish Pain . Animal Sentience , 2016 : 025 . Desai U. M. , Achuthankutty C. T. 2000 . Complete regeneration of ablated eyestalk in penaeid prawn Penaeus monodon . Current Science , 79 : 1602 – 1603 . Diarte-Plata G. , Sainz-Hernández J. C. , Aguinaga-Cruz J. A. , Fierro-Coronado J. A. , Polanco-Torres A. , Puente-Palazuelos C. 2012 . Eyestalk ablation procedures to minimize pain in the freshwater prawn Macrobrachium americanum . Applied Animal Behaviour Science , 140 : 172 – 178 . Google Scholar Crossref Search ADS Diggles B. K. 2016 . Fish pain: would it change current best practice in the real world? Commentary on Key on Fish Pain . Animal Sentience , 2016 : 061 . Diggles B. K. , Arlinghaus R. , Browman H. I. , Cooke S. J. , Cowx I. G. , Kasumyan A. O. , Key B. et al. 2017 . Responses of larval zebrafish to low pH immersion assay. Comment on Lopez-Luna et al. (2017) . Journal of Experimental Biology , 220 : 3191 – 3192 . Google Scholar Crossref Search ADS PubMed Diggles B. K. , Cooke S. J. , Rose J. D. , Sawynok W. 2011 . Ecology and welfare of aquatic animals in wild capture fisheries . Reviews in Fish Biology and Fisheries , 21 : 739 – 765 . Google Scholar Crossref Search ADS Dyuizen I. V. , Kotsyuba E. P. , Lamash N. E. 2012 . Changes in the nitric oxide system in the shore crab Hemigrapsus sanguineus (Crustacea, decapoda) CNS induced by a nociceptive stimulus . The Journal of Experimental Biology , 215 : 2668 – 2676 . Google Scholar Crossref Search ADS PubMed EFSA . 2005 . Opinion of the Scientific Panel on Animal Health and Welfare (AHAW) on a request from the Commission related to the aspects of the biology and welfare of animals used for experimental and other scientific purposes. EFSA-Q-2004-105 . EFSA Journal , 292 : 1 – 46 . Elwood R. W. 2016 . Stress was never said to be pain: response to Stevens et al. 2016 . Biology Letters , 12 : 2016.0126. Elwood R. W. 2017 . Might insects experience pain? Elwood on Klein & Barron on Insect Experience . Animal Sentience , 2016 : 133 . Elwood R. W. , Adams L. 2015 . Electric shock causes physiological stress responses in shore crabs, consistent with prediction of pain . Biology Letters , 11 : 2015.0800. Elwood R. W. , Appel M. 2009 . Pain experience in hermit crabs? Animal Behaviour , 77 : 1243 – 1246 . Google Scholar Crossref Search ADS Elwood R. W. , Dalton N. , Riddell G. 2017 . Aversive responses by shore crabs to acetic acid but not to capsaicin . Behavioural Processes , 140 : 1 – 5 . Google Scholar Crossref Search ADS PubMed Fischer B. 2016 . What if Klein & Barron are right about insect sentience? Animal Sentience , 2016 : 115 . Fossat P. , Bacque-Cazenave J. , De Deurwaerdere P. , Cattaert D. , Delbecque J.-P. 2015 . Serotonin, but not dopamine, controls stress response and anxiety-like behavior in crayfish, Procambarus clarkii . Journal of Experimental Biolology , 218 : 2745 – 2752 . Google Scholar Crossref Search ADS Fossat P. , Bacqué-Cazenave J. , De Deurwaerdère P. , Delbecque J.-P. , Cattaert D. 2014 . Anxiety-like behavior in crayfish is controlled by serotonin . Science , 344 : 1293 – 1297 . Google Scholar Crossref Search ADS PubMed Huber R. , Delago A. 1998 . Serotonin alters decisions to withdraw in fighting crayfish, Astacus astacus: the motivational concept revisited . Journal of Comparative Physiology A , 182 : 573 – 583 . Google Scholar Crossref Search ADS Imeh-Nathaniel A. I. , Rincon N. , Orfanakos V. B. , Brechtel L. , Wormack L. , Richardson E. , Huber R. et al. 2017 . Effects of chronic cocaine, morphine and methamphetamine on the mobility, immobility and stereotyped behaviors in crayfish . Behavioural Brain Research , 332 : 120 – 125 . Google Scholar Crossref Search ADS PubMed Joseph R. M. , Carlson J. R. 2015 . Drosophila chemoreceptors: a molecular interface between the chemical world and the brain . Trends in Genetics , 31 : 683 – 695 . Google Scholar Crossref Search ADS PubMed Kasumyan A. O. , Doving K. B. 2003 . Taste preferences in fish . Fish and Fisheries , 4 : 289 – 347 . Google Scholar Crossref Search ADS Kawai N. , Kono R. , Sugimoto S. 2004 . Avoidance learning in the crayfish (Procambarus clarkii) depends on the predatory imminence of the unconditioned stimulus: a behavior systems approach to learning in invertebrates . Behavioural Brain Research , 150 : 229 – 237 . Google Scholar Crossref Search ADS PubMed Key B. 2015 . Fish do not feel pain and its implications for understanding phenomenal consciousness. Biology and Philosophy , 30 : 149 – 165 . Key B. , Arlinghaus R. , Browman H. I. , Cooke C. J. , Cowx I. G. , Diggles B. K. , Rose J. D. et al. 2017 . Problems with equating thermal preference with “emotional fever” and sentience. Comment on Rey et al. (2015) Fish can show emotional fever: stress-induced hyperthermia in zebrafish . Proceedings of the Royal Society of London B , 284 : 20160681. Google Scholar Crossref Search ADS Klein C. , Barron A. B. 2016 . Insects have the capacity for subjective experience . Animal Sentience , 2016 : 100 . Kotsyuba E. P. , Dyuizen I. V. , Lamash N. E. 2010 . Stress induced changes in the nitric oxide system of shore crabs living under different ecological conditions . Russian Journal of Marine Biology , 36 : 201 – 208 . Google Scholar Crossref Search ADS Landolt P. J. , Adams T. , Rogg H. 2011 . Trapping spotted wing drosophila, Drosophila suzukii (Matsumura) (Diptera: Drosophilidae), with combinations of vinegar and wine, and acetic acid and ethanol . Journal of Applied Entamology , 136 : 148 – 154 . Google Scholar Crossref Search ADS Le Doux J. 2012 . Rethinking the emotional brain . Neuron , 73 : 653 – 676 . Google Scholar Crossref Search ADS PubMed Lightner D. V. 2011 . Status of shrimp diseases and advances in shrimp health management. In Diseases in Asian Aquaculture VII, pp. 121–134 . Ed. by Bondad-Reantaso M. G. , Jones J. B. , Corsin F. , Aoki T. . Fish Health Section, Asian Fisheries Society, Selangor , Malaysia . 385 pp. Lopez-Luna J. , Al-Jubouri Q. , Al-Nuaimy W. , Sneddon L. U. 2017 . Reduction in activity by noxious chemical stimulation is ameliorated by immersion in analgesic drugs in zebrafish . Journal of Experimental Biology , 220 : 1451 – 1458 . Google Scholar Crossref Search ADS PubMed Lozada M. , Romano A. , Maldonado H. 1988 . Effect of morphine and naloxone on a defensive response of the crab Chasmagnathus granulatus . Pharmacology Biochemistry and Behavior , 30 : 635 – 640 . Google Scholar Crossref Search ADS Magee B. , Elwood R. W. 2013 . Shock avoidance by discrimination learning in the shore crab (Carcinus maenas) is consistent with a key criterion for pain . Journal of Experimental Biology , 216 : 353 – 358 . Google Scholar Crossref Search ADS PubMed Magee B. , Elwood R. W. 2016 . Trade-offs between predator avoidance and electric shock avoidance in hermit crabs demonstrate a non-reflexive response to noxious stimuli consistent with prediction of pain . Behavioural Processes , 130 : 31 – 35 . Google Scholar Crossref Search ADS PubMed Maldonado H. , Miralto A. 1982 . Effect of morphine and naloxone on a defensive response of the mantis shrimp (Squilla mantis) . Journal of Comparative Physiology , 147 : 455 – 459 . Google Scholar Crossref Search ADS Mallatt J. M. , Feinburg T. E. 2016 . Insect consciousness: fine tuning the hypothesis: Mallatt & Feinberg on Klein & Barron on Insect Experience . Animal Sentience , 2016 : 118 . Martin J. W. , Davis G. E. 2001 . An updated classification of the recent crustacea. No. 39 Science Series, Natural History Museum of Los Angeles County. 124 pp. Maruzzo D. , Bonato L. , Brena C. , Fusco G. , Minelli A. 2005 . Appendage loss and regeneration in arthropods: a comparative view. In Crustacea and Arthropod Relationships, pp. 215–245 . Ed. by Koenemann S. , Jenner R. . Taylor & Francis , Boca Raton, FL . 440 pp. Mobbs R. J. , Nair S. , Blum P. 2007 . Peripheral nerve stimulation for the treatment of chronic pain . Journal of Clinical Neuroscience , 14 : 216 – 221 . Google Scholar Crossref Search ADS PubMed Nathaniel T. I. , Panksepp J. , Huber R. 2010 . Effects of a single and repeated morphine treatment on conditioned and unconditioned behavioral sensitization in Crayfish . Behavioral Brain Research , 207 : 310 – 320 . Google Scholar Crossref Search ADS Newby N. C. , Stevens E. D. 2008 . The effects of the acetic acid “pain” test on feeding, swimming and respiratory responses of rainbow trout (Oncorhynchus mykiss) . Applied Animal Behavior Science , 114 : 260 – 269 . Google Scholar Crossref Search ADS Newby N. C. , Stevens E. D. 2009 . The effects of the acetic acid “pain” test on feeding, swimming, and respiratory responses of rainbow trout (Oncorhynchus mykiss): a critique on Newby and Stevens (2008)—response . Applied Animal Behaviour Science , 116 : 97 – 99 . Google Scholar Crossref Search ADS Newby N. C. , Wilkie M. P. , Stevens E. D. 2009 . Morphine uptake, disposition and analgesic efficacy in the common goldfish (Carassius auratus) . Canadian Journal of Zoology , 87 : 388 – 399 . Google Scholar Crossref Search ADS NHMRC . 2013 . Australian Code for the Care and Use of Animals for Scientific Purposes, 8th ed . National Health and Medical Research Council , Canberra . 92 pp. Paterson B. D. , Davidson G. W. , Spanoghe P. T. 2001 . Physiological studies of stress and morbidity during postharvest handling of western rock lobsters (Panulirus cygnus), I. Physiological stress indicators. Final report of Project no. 96/345 submitted to the Fisheries Research and Development Corporation, November 2001. 141 pp. Patterson L. , Dick J. T. , Elwood R. W. 2007 . Physiological stress responses in the edible crab, Cancer pagurus, to the fishery practice of de-clawing . Marine Biology , 152 : 265 – 272 . Google Scholar Crossref Search ADS Puri S. , Faulkes Z. 2010 . Do decapod crustaceans have nociceptors for extreme pH? PLoS One , 5 : e10244. Google Scholar Crossref Search ADS PubMed Puri S. , Faulkes Z. 2015 . Can crayfish take the heat? Procambarus clarkii show nociceptive behaviour to high temperature stimuli, but not low temperature or chemical stimuli . Biology Open , 4 : 441 – 448 . Google Scholar Crossref Search ADS PubMed Regier J. C. , Shultz J. W. , Zwick A. , Hussey A. , Ball B. , Wetzer R. , Martin J. W. et al. 2010 . Arthropod relationships revealed by phylogenomic analysis of nuclear protein-coding sequences . Nature , 463 : 1079 – 1083 . Google Scholar Crossref Search ADS PubMed Rey S. , Huntingford F. A. , Boltaña S. , Vargas R. , Knowles T. G. , Mackenzie S. 2015 . Fish can show emotional fever: stress-induced hyperthermia in zebrafish . Proceedings of the Royal Society B , 282 : 20152266. Google Scholar Crossref Search ADS PubMed Rose J. D. 2002 . The neurobehavioral nature of fishes and the question of awareness and pain . Reviews in Fisheries Science , 10 : 1 – 38 . Google Scholar Crossref Search ADS Rose J. D. 2003 . A critique of the paper: do fish have nociceptors: evidence for the evolution of a vertebrate sensory system published in Proceedings of the Royal Society: Biological Sciences 270: 1115–1121, 2003 by Sneddon, Braithwaite and Gentle. In Information Resources on Fish Welfare 1970–2003 Animal Welfare Information Resources No. 20, pp. 49–51. Ed. by H. E. Erickson. U.S. Department of Agriculture, Beltsville, 435 pp. Rose J. D. 2007 . Anthropomorphism and ‘mental welfare’ of fishes . Diseases of Aquatic Organisms , 75 : 139 – 154 . Google Scholar Crossref Search ADS PubMed Rose J. D. , Arlinghaus R. , Cooke S. J. , Diggles B. K. , Sawynok W. , Stevens E. D. , Wynne C. D. L. 2014 . Can fish really feel pain? Fish and Fisheries , 15 : 97 – 133 . Google Scholar Crossref Search ADS Rubenstein D. I. , Hazlett B. A. 1974 . Examination of the agonistic behaviour of the crayfish Orconectes virilis by character analysis . Behaviour , 50 : 193 – 215 . Google Scholar Crossref Search ADS Sahul Hameed A. S. , Yoganandhan K. , Sri Widada J. , Bonami J. R. 2004 . Experimental transmission and tissue tropism of Macrobrachium rosenbergii nodavirus (MrNV) and its associated extra small virus (XSV) . Diseases of Aquatic Organisms , 62 : 191 – 196 . Google Scholar Crossref Search ADS PubMed Salini J. P. , Blaber S. J. M. , Brewer D. T. 1990 . Diets of piscivorous fishes in a tropical Australian estuary, with special reference to predation on penaeid prawns . Marine Biology , 105 : 363 – 374 . Google Scholar Crossref Search ADS Schneider R. A. , Huber R. , Moore P. A. 2001 . Individual and status recognition in the crayfish, Orconectes rusticus: the effects of urine release on fight dynamics . Behaviour , 138 : 137 – 153 . Google Scholar Crossref Search ADS Seifert A. W. , Kiama S. , G. , Seifert M. G. , Goheen J. R. , Palmer T. M. , Maden M. 2012 . Skin shedding and tissue regeneration in African spiny mice (Acomys) . Nature , 489 : 561 – 565 . Google Scholar Crossref Search ADS PubMed Sellars M. , Wood A. , Murphy B. , Coman G. , Arnold S. , McCulloch R. , Preston N. 2013 . Reproductive performance and mature gonad morphology of triploid and diploid Black Tiger shrimp (Penaeus monodon) siblings . Aquaculture Research , 44 : 1493 – 1501 . Google Scholar Crossref Search ADS Smith E. , Lewin G. R. 2009 . Nociceptors: a phylogenetic view . Journal of Comparative Physiology A , 195 : 1089 – 1106 . Google Scholar Crossref Search ADS Sneddon L. U. 2003 . The evidence for pain in fish. Use of morphine as an anaesthetic . Applied Animal Behaviour Science , 83 : 153 – 162 . Google Scholar Crossref Search ADS Sneddon L. U. 2018 . Comparative physiology of nociception and pain. Physiology . 33 : 63 – 73 . Sneddon L. U. , Elwood R. W. , Adamo S. A. , Leach M. C. 2014 . Defining and assessing animal pain . Animal Behaviour , 97 : 201 – 212 . Google Scholar Crossref Search ADS Soderhall K. , Cerenius L. 1992 . Crustacean immunity . Annual Review of Fish Diseases , 2 : 3 – 23 . Google Scholar Crossref Search ADS Soundarapandian P. , Varadharajan D. , Anand T. 2013 . Electrode placement and electroejaculation of blue swimmer crab, Portunus pelagicus (Linnaeus) . Journal of Earth Science and Climate Change , 4 : 142 . Google Scholar Crossref Search ADS Steenbergen P. J. , Bardine N. 2014 . Antinociceptive effects of buprenorphine in zebrafish larvae: an alternative for rodent models to study pain and nociception? Applied Animal Behaviour Science , 152 : 92 – 99 . Google Scholar Crossref Search ADS Stevens E. D. , Arlinghaus R. , Browman H. I. , Cooke S. J. , Cowx I. G. , Diggles B. K. , Key B. et al. 2016 . Stress is not pain. Comment on Elwood and Adams (2015) Electric shock causes physiological stress responses in shore crabs, consistent with prediction of pain . Biology Letters , 12 : 20151006. Google Scholar Crossref Search ADS PubMed Stocker A. M. , Huber R. 2001 . Fighting strategies in crayfish Orconectes rusticus (Decapoda, Cambaridae) differ with hunger state and the presence of food cues . Ethology , 107 : 727 – 736 . Google Scholar Crossref Search ADS Stoner A. W. 2012 . Assessing stress and predicting mortality in economically significant crustaceans . Reviews in Fisheries Science , 20 : 111 – 135 . Google Scholar Crossref Search ADS Taylor J. , Vinatea L. , Ozorio R. , Schuweitzer R. , Andreatta E. R. 2004 . Minimising the effects of stress during eyestalk ablation of Litopenaeus vannamei females with topical anesthetic and a coagulating agent . Aquaculture , 233 : 173 – 179 . Google Scholar Crossref Search ADS Tobin D. M. , Bargmann C. I. 2004 . Invertebrate nociception: behaviors, neurons and molecules . Journal of Neurobiology , 61 : 161 – 174 . Google Scholar Crossref Search ADS PubMed Uawisetwathana U. , Leelatanawit R. , Klanchui A. , Prommoon J. , Klinbunga S. , Karoonuthaisiri N. 2011 . Insights into eyestalk ablation mechanism to induce ovarian maturation in the black tiger shrimp . PLOS One , 6 : e24427. Google Scholar Crossref Search ADS PubMed Uddin S. A. , Rahman M. M. 2015 . Gonadal maturation, fecundity and hatching performance of wild caught tiger shrimp Penaeus monodon using unilateral eyestalk ablation in captivity . Journal of Bangladesh Agricultural University , 13 : 315 – 322 . Google Scholar Crossref Search ADS Wang L. , Pan L. , Ding Y. , Ren X. 2018 . Effects of low salinity stress on immune response and evaluating indicators of the swimming crab Portunus trituberculatus . Aquaculture Research , 49 : 659 – 667 . Google Scholar Crossref Search ADS Wilkens J. L. , Wilkens L. A. , McMahon B. R. 1974 . Central control of cardiac and scaphognathite pacemakers in the crab, Cancer magister . Journal of Comparative Physiology , 90 : 89 – 104 . Google Scholar Crossref Search ADS Wyban J. A. 1992 . Development and commercial performance of high health shrimp using specific pathogen free (SPF) broodstock Penaeus vannamei. In Proceedings of the Special Session on Shrimp Farming, Orlando, FL, 22–25 May 1992, pp. 254–260. Ed. by J. Wyban. World Aquaculture Society, Baton Rouge, LA. 301 pp. Zhang Z. 2011 . Animal biodiversity: an outline of higher-level classification and survey of taxonomic richness - Phylum Arthropoda von Siebold, 1848 . Zootaxa , 4138 : 99 – 103 . © International Council for the Exploration of the Sea 2018. 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)

Browman, Howard, I;Cooke, Steven, J;Cowx, Ian, G;Derbyshire, Stuart W, G;Kasumyan,, Alexander;Key,, Brian;Rose, James, D;Schwab,, Alexander;Skiftesvik, Anne, Berit;Stevens, E, Don;Watson, Craig, A;Arlinghaus,, Robert