Continental Shelf-Wide Response of a Fish Assemblage to Rapid Warming of The

Home , Bandfish, Caproidae, Cepola macrophthalma, Greater weever, John Dory, Lotidae

Please cite this article in press as: Simpson et al., Continental Shelf-Wide Response of a Fish Assemblage to Rapid Warming of the Sea, Current Biology (2011), doi:10.1016/j.cub.2011.08.016 Current Biology 21, 1–6, September 27, 2011 ª2011 Elsevier Ltd All rights reserved DOI 10.1016/j.cub.2011.08.016 Report Continental Shelf-Wide Response of a Fish Assemblage to Rapid Warming of the Sea

1, 2,3 Stephen D. Simpson, * Simon Jennings, a further 1.5C to 2.6C by 2100 [9]. Marine ecosystems in Mark P. Johnson,4 Julia L. Blanchard,5,6 Pieter-Jan Scho¨ n,7 the northeast Atlantic have warmed particularly rapidly, with David W. Sims,8,9 and Martin J. Genner1,8 mean sea temperatures in the North Sea and Celtic-Biscay 1 School of Biological Sciences, University of Bristol, Shelf regions increasing between 1982 and 2006 by 1.31C Bristol BS8 1UG, UK and 0.72C, respectively [5], four times faster than the global 2The Centre for Environment, Fisheries and Aquaculture average [4]. Climate warming is affecting marine biological Science, Lowestoft Laboratory, Lowestoft, processes from the genetic to ecosystem level of organization, Suffolk NR33 0HT, UK with implications for commercial fisheries and food security 3School of Environmental Sciences, University of East Anglia, [1–3]. Significant progress has been made with identifying Norwich NR4 7TJ, UK mechanisms by which climate change can affect fish popula- 4Martin Ryan Marine Science Institute, tion dynamics [10, 11] in describing shifts in the distributions National University of Ireland, Galway, Galway, Ireland of some fish species along latitudinal and depth gradients 5Division of Biology, Imperial College London, Silwood Park, associated with climate change [6, 7, 12] and in developing Ascot SL5 7PY, UK climate envelope models to predict effects of climate change 6Department of Animal and Plant Sciences, on future species distributions [8]. To date, however, macro- University of Sheffield, Sheffield S10 2TN, UK ecological analyses of the effects of climate change on marine 7Agri-Food and Biosciences Institute, Belfast BT9 5PX, fish assemblages have not accounted for constraints on distri- Northern Ireland, UK butional shifts due to population dependence on essential 8Marine Biological Association of the United Kingdom, habitat, for example, favored substrates, appropriate predator The Laboratory, Plymouth PL1 2PB, UK and prey fields, and close proximity to nursery grounds, all of 9School of Marine Sciences and Engineering, Marine Institute, which are often unknown and difficult to quantify. Accommo- University of Plymouth, Plymouth PL4 8AA, UK dation of spatial habitat heterogeneity when measuring climate impacts requires an alternative Eulerian (grid-based) approach of analyzing multiple local associations of species Summary abundance and community diversity with climatic variables, controlling for depth and latitude and allowing for complex Climate change affects marine biological processes from oceanography. genetic to ecosystem levels [1–3]. Recent warming in the To investigate the effects of temperature variability on abun- northeast Atlantic [4, 5] has caused distributional shifts in dance of demersal species within the European continental some fish species along latitudinal and depth gradients shelf fish assemblage, we compiled and analyzed three [6, 7], but such changes, as predicted by climate envelope decades of high-resolution fisheries-independent data. Our 2 models [8], may often be prevented because population analysis spanned w1.2 million km of seabed (15 latitude 3 movement requires availability of suitable habitat. We as- 25 longitude) and depths ranging from 5 to 592 m (mean = sessed the full impacts of warming on the commercially 75 m) and used data from 25,612 bottom-trawl sample hauls important European continental shelf fish assemblage using between 1980 and 2008 comprising >100 million fish from a data-driven Eulerian (grid-based) approach that accommo- 177 species (or species groups; see Tables S1 and S2 available dates spatial heterogeneity in ecological and environmental online). Temperature in this region has increased by 0.06C 21 21 conditions. We analyzed local associations of species abun- years on average at the surface and by 0.04C years on dance and community diversity with climatic variables, the seabed during the past 30 years (Figure 1A). Significant assessing trends in 172 cells from records of >100 million warming trends were evident in all 172 1 3 1 cells within individuals sampled over 1.2 million km2 from 1980–2008. the region and were most intense in the southern and eastern We demonstrate responses to warming in 72% of common North Sea and the Irish Sea (Figure 1B). Trends in SST and species, with three times more species increasing in abun- sea-bottom temperature (SBT) correlated closely (Pearson’s dance than declining, and find these trends reflected in inter- r = 0.94, p < 0.001), whereas there was no significant correla- national commercial landings. Profound reorganization of tion between regional rates of fishing mortality (another key the relative abundance of species in local communities driver of trends in fish abundance) and temperature (SST: r = occurred despite decadal stability in the presence-absence 20.29, p = 0.16; SBT: r = 20.13, p = 0.56), which increased of species. Our analysis highlights the importance of during the 1980s but declined thereafter [13](Figure 1A). focusing on changes in species abundance in established local communities to assess the full consequences of Stability in Species Composition in Communities Despite climate change for commercial fisheries and food security. Warming An unsupervised Bayesian clustering analysis using species Results and Discussion presence-absence data identified 12 biogeographic clusters across the studied shelf region (Figure 2A). The distribution Average global sea surface temperature (SST) increased by of these clusters was generally stable between consecutive 0.7C over the last 100 years [4] and is predicted to rise by 5 year time periods, although during the period of the most rapid warming in the late 1980s (Figure 1A), more than twice *Correspondence: [email protected] as many cells changed in identity between 5-year periods Please cite this article in press as: Simpson et al., Continental Shelf-Wide Response of a Fish Assemblage to Rapid Warming of the Sea, Current Biology (2011), doi:10.1016/j.cub.2011.08.016 Current Biology Vol 21 No 18 2

Figure 1. Environmental Conditions on the European Continental Shelf (A) Temperature and fishing trends. Annual temperature trends are calculated for the 172 1 3 1 cells in this study and are shown with a linear trend (sea surface tempera- 21 2 ture [SST]: +0.06C years ,R = 0.67; sea-bottom 21 2 temperature [SBT]: +0.04C years ,R = 0.50). Standardized fishing mortality is the mean estimate of multispecies fishing mortality for six subregions (North Sea, Eastern Channel, Western Channel, Celtic Sea, Irish Sea, Northwest Scotland) weighted by the spawning-stock biomass of each species that is assessed, as reported in International Council for the Exploration of the Sea working group reports and is shown with two linear trends (1982–1990: R2 = 0.80; 1990–2007: R2 = 0.81). (B) Spatial variation in temperature trends during the study period (1980–2008), expressed as the slope of a fitted linear relationship between SST and year for each cell.

compared with the 1990–2000s (Figure S1). In total, 53% of timescales than implied by reports of changing distributions cells changed identity at some time during the 30 years, but due to recent warming [6]. only four of the 12 clusters changed significantly in latitudinal distribution. The western central North Sea and Channel Seventy-Two Percent of Common Species Are Responding assemblages moved southwards and the southeastern North to Warming Sea and Irish Sea clusters moved northwards. The spatial Observations of latitudinal shifts of some North Sea species patterns of clusters were not driven by rare species (Figure S2) over decadal timescales [6] raise questions about how and were not associated with boundaries between different changes in abundance relate to changes to species distribu- surveys. Instead, similar patterns were seen in regional anal- tions. Across the region, we identified significant positive yses of data from single surveys (Figure S3). Thus, the general associations between fish annual abundance and temperature pattern on the European continental shelf is one of relative for 27 of the 50 most abundant species, whereas nine species stability of overall species presence-absence, despite ubiqui- showed significant declines in abundance relative to warm tous recent warming, a pattern that concurs with an earlier years (Figure 3; Table S2). The mean central latitude (derived study of North Sea fish that found no overall consistent trend from FishBase [15]) of the 27 species that increased in relative in latitudinal shifts among the 36 species studied [6]. Together, abundance with warm conditions across the region (44.5N; these analyses suggest that management zones based on Table S2) was significantly lower than the nine species biogeographical criteria [14] are acceptable on longer that declined (55.2N; t test [equal variances not assumed],

Figure 2. Assemblage Composition in Relation to Warming (A) Twelve biogeographic clusters, identified by Bayesian analysis of presence-absence data, that maintained similar distributions over six 5-year periods. White cells in (A) and (B) indicate locations with insufficient fish-abundance data. (See Supplemental Information for further analysis of cell cluster changes [Figure S1], patterns based solely on common species [Figure S2], and patterns within single surveys [Figure S3].) (B) Eighty-two percent of cells with a positive assemblage response to warming (percentage variation above random in beta-diversity trend explained by temperature). (See Figure S4 for comparison of trends with different temperature measures.) (C) Mean assemblage response to temperature within 1 3 1 cells classified by intensity of warming. Letters above 95% confidence interval bars indicate significantly different categories

(analysis of variance, F6,162 = 9.865, p < 0.001; subgroups identiﬁed using Tukey’s honestly signiﬁcant difference test, a = 0.05). Please cite this article in press as: Simpson et al., Continental Shelf-Wide Response of a Fish Assemblage to Rapid Warming of the Sea, Current Biology (2011), doi:10.1016/j.cub.2011.08.016 Response of Fish Assemblage to Warming 3

Northern Figure 3. Species Abundance Responses to Warming affinity Norway pout Mean species-level relationships between abundance Norway redfish pogge and temperature for the 50 most common species in shorthorn sculpin long-rough dab 172 1 3 1 cells. Red indicates an increase in abundance Atlantic cod in warm years, blue indicates a decline, and gray indi- dab lemon sole cates no significant response. (See Table S2 for scientific haddock ling names, species life history traits, and regional-scale saithe responses.) witch pollack whiting thorny skate anglerfish blue whiting flounder thornback ray determining and predicting abundance re- silvery cod stickleback sponses of species to warming seas. hagfish megrim poor cod hake International Commercial Landings Reflect spurdog greater weever Abundance Changes fourbeard rockling plaice grey gurnard Although commercial landings data may give pouting Dover sole biased estimations of natural abundance [17], dragonet lesser weever they give a valuable indication of the impor- scaldfish spotted ray tance of species to the fishing industry. We spotted dragonet red mullet investigated whether commercial landings cuckoo ray tub gurnard [18] of species from northeast Atlantic fisheries thickback sole lesser spotted dogfish reflected the abundance response of species red gurnard boarfish to warming over the last 30 years derived red bandfish solenette from fisheries-independent data. There was John dory blackbelly rosefish a significant positive relationship between the Southern pearlsides affinity splendid alfonsino trend in commercial landings (between 1980 and 2007) and the abundance response to -0.2 -0.1 0 0.1 0.2 0.3 temperature from survey data for the 33 Abundance response to temperature species for which fisheries data were available (Pearson’s r = 0.41, p = 0.018; Figure 4C). There was also a positive relationship between preferred temperature of species in t24.3 = 4.33, p < 0.001). This difference held for the central surveys and landings trend (r = 0.51, p = 0.002) and a negative latitudinal tendency of these species when calculated from relationship between central latitudinal range and landings the main data set (53.5N versus 55.3N, t14.5 = 2.58, p = trend, with higher latitude species declining in landings (r = 0.021). Generally, central latitudinal range was significantly 20.57, p = 0.001). These results indicate that observed correlated with the abundance response of species to changes in survey data are mirrored in landings of commercial warming (Pearson’s r = 20.52, p < 0.001; Figure 4A), and this species. relationship was also apparent using the central latitudinal tendency of each species within the data set (r = 20.44, p = Local-Scale Reorganization of Communities across 0.001). Although preferred temperatures of species (mean the Region temperature at capture of all individuals) that increased in Finally, we investigated whether changes in local assemblage abundance with warm conditions (mean = 11.3C) were not composition have occurred despite stability in species pres- significantly different from those of species that declined ence-absence. We used principal component analysis (PCA) (mean = 10.8C, t21.6 = 1.66, p = 0.112), there was a significant to reduce matrices of species abundance data for each year overall correlation between preferred temperature and in each 1 3 1 cell and used principal component 1 (PC1) abundance response of species to warming (r = 0.31, p = as a single metric summarizing the major temporal trend in 0.027; Figure 4B). Maximum body size provides a proxy beta-diversity in subsequent analyses [19]. On average, PC1 for ecological performance, as well as other life history captured 34% of variation within cells (range 20%–74%), and parameters, demography, production, consumption, and mean annual SST with a 1 year time lag explained more varia- metabolism [7, 16], and the 27 species that were more tion in PC1 than expected by chance in 143 of the 172 cells abundant during warm conditions (mean = 493 mm) were (Figure 2B). Annual SBT with a 1 year lag explained more significantly smaller than the nine species that declined variation than expected by chance in 131 cells. This stronger (956 mm, t9.709 = 23.20, p = 0.01). However, maximum body association of surface temperature with fish assemblage size was not directly correlated with the strength of response change may indicate that temperature exerts its effect via of species to warming (r = 20.147, p = 0.309). There was also pelagic larval stages. At a 1 3 1 resolution, neighboring cells no significant difference in depth preference of species often had contrasting assemblage responses to warming, (mean depth at capture of all individuals) that increased in implying differential effects of local habitat and/or interspecific warm conditions (mean = 99 m) from those that declined ecological interactions [19]. Temperature-driven changes in (mean = 108 m, t31.70 = 20.92, p = 0.364) and no overall relation- assemblage composition were most pronounced in the ship between preferred depth and index of response (r = 0.02, southern and northern North Sea and the Irish Sea (Figure 2B), p = 0.892). In combination, the results from this analysis matching warming hotspots (Figure 1B). Assemblage-level indicate the importance of overall latitudinal occurrence change was significantly greater in cells that have experienced and temperature preference, but not depth preference, in more intense warming (Figure 2C), and the reduced response Please cite this article in press as: Simpson et al., Continental Shelf-Wide Response of a Fish Assemblage to Rapid Warming of the Sea, Current Biology (2011), doi:10.1016/j.cub.2011.08.016 Current Biology Vol 21 No 18 4

Conclusions The results here are unique in exploring observed spatial heterogeneity in the response of the assemblage at a resolution that matches the one used in climate envelope models [20]. Our finding of stability in presence-absence of species over decadal periods, but significant temperature-driven responses in local species abundance and assemblage composition, suggests that climate envelope models based on species presence-absence alone will not predict the most ecologically and economically significant effects of climate change. Moreover, studies exploring shifts in central latitudes of species ranges using abundance data may miss important spatial variability and local responses. For example, poleward shifts in abundance, as would be expected under climate envelope scenarios, have been identified in only 13 of 36 studied species (36%) in the North Sea assemblage [6] despite significant and ubiquitous changes in temperature. In contrast, by analyzing local changes in abundance without assuming a range shift during warming and allowing for uneven abundance distributions linked to suitable habitat, we identified temperature-associated species-level changes in abundance in the North Sea in 39 of the 50 (78%) most common species (Figure 3; Table S2) and identified change in assemblage composition associated with warming in 93% of the North Sea cells (Figure 2B). This is consistent with many species responding in abundance without necessarily changing their spatial distributions of occurrence within the study region. The impact of these abundance changes to commercial fisheries was apparent, where landings of the nine species identified as declining in warm conditions (blue species in Figure 3, including haddock and cod) fell by a half during the period of this study, whereas landings of the 27 species identified as increasing in warm conditions (red species in Figure 3, including hake and dab) increased 2.5 times. For a given region, reorganization of the fishing fleet and management strategies will be required to ensure that the right species are targeted and harvested sustainably. Our analysis highlights the value of data from high resolution large-scale surveys of species abundance in regions with a known history of climate change. The focus of many studies on the ecological effects of climate change in marine fish or fisheries has been on changing abundance distributions along latitudinal or depth gradients. Our analyses demonstrate that such distributional changes have been rela- tively benign on the European continental shelf during rapid warming of the sea over the last 30 years. Instead, there has been profound climate-driven reorganization of species abundance in established local communities over much of the shelf region, without spatial reorganization of species presence-absence. This result is important because local abundance changes in established fish communities have the greatest implications for both ecosystem function and societies dependent on marine natural resources. By contrast, over the timescale, studied changes in species ranges Figure 4. Species Responses to Warming and Impact on Fisheries are arguably less important, because they result only from (A and B) Abundance response of species to temperature for the 50 most common species in relation to species characteristics: mean latitude of colonizations or extirpations that occur when species are occurrence (A) and preferred temperature (B). necessarily scarce and below the abundances required for (C) Increase in commercial landings of species with positive abundance commercially viable exploitation. Thus, studies of distribu- responses to warming and decline in landings of species with negative tional shifts can overlook ecologically and economically sig- responses. nificant climate effects, except for rare examples of species range expansion coupled with large increases in abundance for cells with lower levels of warming suggests resilience to (e.g., [21]). The next challenge is to use this knowledge to warming to a threshold, beyond which change becomes develop effective indicators and predictive models to assess inevitable and pronounced. consequences of climate change for marine ecosystems and Please cite this article in press as: Simpson et al., Continental Shelf-Wide Response of a Fish Assemblage to Rapid Warming of the Sea, Current Biology (2011), doi:10.1016/j.cub.2011.08.016 Response of Fish Assemblage to Warming 5

fisheries. Such models will need to consider growth, reproduc- References tion, habitat use, essential habitat requirements, and location 1. Brierley, A.S., and Kingsford, M.J. (2009). Impacts of climate change on of nursery and spawning grounds if they are to make predic- marine organisms and ecosystems. Curr. Biol. 19, R602–R614. tions about likely future impacts of warming on food web 2. Brander, K. (2010). Impacts of climate change on fisheries. J. Mar. Syst. dynamics and fish-driven ecological processes. Moreover, 79, 389–402. because the northeast Atlantic assemblage is heavily exploited 3. Hare, J.A., Alexander, M.A., Fogarty, M.J., Williams, E.H., and Scott, J.D. and species responses to fishing are influenced by body size (2010). Forecasting the dynamics of a coastal fishery species using [22], models are needed that integrate effects of warming and a coupled climate—population model. Ecol. Appl. 20, 452–464. fishing. This gained understanding will be valuable for predict- 4. Smith, T.M., Reynolds, R.W., Peterson, T.C., and Lawrimore, J. (2008). Improvements to NOAA’s historical merged land-ocean surface ing future exploitable resources and crucial for determining temperature analysis (1880-2006). J. Clim. 21, 2283–2296. appropriate harvesting and conservation strategies at relevant 5. Sherman, K., Belkin, I., Friedland, K., O’Reilly, J., and Hyde, K. (2009). spatial scales. Accelerated warming and emergent trends in fisheries biomass yields of the world’s large marine ecosystems. In The UNEP Large Marine Ecosystem Report: A Perspective on Changing Conditions in LMEs of Experimental Procedures the World’s Regional Seas, K. Sherman and G. Hempel, eds. (Nairobi, Kenya: United Nations Environment Programme), pp. 41–80. Data were collated from 11 demersal trawl surveys conducted over 30 years 6. Perry, A.L., Low, P.J., Ellis, J.R., and Reynolds, J.D. (2005). Climate by European fisheries agencies in 218 1 3 1 cells (Table S1). The change and distribution shifts in marine fishes. Science 308, 1912–1915. combined data comprised 25,612 hauls and an estimated 105,304,825 7. Dulvy, N.K., Rogers, S.I., Jennings, S., Stelzenmuller, V., Dye, S.R., and fish from 177 species or taxa in 172 cells. Bathymetry data were obtained Skjoldal, H.R. (2008). Climate change and deepening of the North Sea from General Bathymetric Chart of the Oceans [23]. Modeled SST [24] fish assemblage: A biotic indicator of warming seas. J. Appl. Ecol. 45, and SBT [25] data used to describe warming trends correlated closely 1029–1039. with in situ measurements (Pearson’s correlation; SST: n = 5,666, r = 8. Cheung, W.W.L., Lam, V.W.Y., Sarmiento, J.L., Kearney, K., Watson, R., 0.94, p < 0.001; SBT: n = 2,789, r = 0.78, p < 0.001) from the 22% and Zeller, D., and Pauly, D. (2010). Large-scale redistribution of maximum 11% of hauls for which they were available, respectively. International fisheries catch potential in the global ocean under climate change. Council for the Exploration of the Sea stock assessment data [26] were Glob. Change Biol. 16, 24–35. used to calculate weighted multispecies estimates of fishing mortality [27] for each International Council for the Exploration of the Seas region. 9. Meehl, G.A., Stocker, T.F., Collins, W.D., Friedlingstein, P., Gaye, A.T., Unsupervised Bayesian clustering [28] of presence-absence data for six 5 Gregory, J.M., Kitoh, A., Knutti, R., Murphy, J.M., Noda, A., et al. year time periods identified biogeographically distinct subassemblages. (2007). Global climate projections. In Climate Change 2007: The Abundance responses of species to temperature were calculated for the Physical Science Basis. Contribution of Working Group I to the Fourth 50 most abundant species (comprising 99.9% of individuals sampled) as Assessment Report of the Intergovernmental Panel on Climate the mean correlation coefficient for all cells between catch and temperature Change, S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. data. For each species, we determined biogeographic affinities from within Averyt, M. Tignor, and H.L. Miller, eds. (Cambridge: Cambridge our data set and using the central latitudinal range from FishBase [15] and University Press), pp. 747–846. obtained maximum body size from all records in the trawl survey database 10. O’Brien, C.M., Fox, C.J., Planque, B., and Casey, J. (2000). Climate (Table S2). Preferred temperature and depth for each species were derived variability and North Sea cod. Nature 404, 142. using frequency of occurrence across the data set. We defined an assem- 11. Beaugrand, G., Brander, K.M., Alistair Lindley, J., Souissi, S., and Reid, blage-level response to temperature for each 1 3 1 cell as the association P.C. (2003). Plankton effect on cod recruitment in the North Sea. Nature above that expected by chance between PC1 from a PCA of a matrix of log 426, 661–664. abundance (per species per year) that associated with mean annual SST 12. European Commission. (2008). Facts and Figures on the CFP: Basic from the previous year. This was identified as the best predictor from 18 Data on the Common Fisheries Policy (Luxembourg: Office for Official potential drivers for each cell (SST and SBT; annual, summer, and winter; Publications of the European Communities). 1–3 year time lag; Figure S4). 13. Spalding, M.D., Fox, H.E., Allen, G.R., Davidson, N., Ferdana, Z.A., Finlayson, M., Halpern, B.S., Jorge, M.A., Lombana, A., Lourie, S.A., et al. (2007). Marine ecoregions of the world: A bioregionalization of Supplemental Information coastal and shelf areas. Bioscience 57, 573–582. 14. Nye, J.A., Link, J.S., Hare, J.A., and Overholtz, W.J. (2009). Changing Supplemental Information includes four figures, two tables, and Supple- spatial distribution of fish stocks in relation to climate and population mental Experimental Procedures and can be found with this article online size on the Northeast United States continental shelf. Mar. Ecol. Prog. at doi:10.1016/j.cub.2011.08.016. Ser. 393, 111–129. 15. Froese, R., and Pauly, D., eds. (2010). FishBase. (http://www.fishbase. Acknowledgments org). 16. Charnov, E.L., and Gillooly, J.F. (2004). Size and temperature in the We thank staff of the Centre for Environment Fisheries and Aquaculture evolution of fish life histories. Integr. Comp. Biol. 44, 494–497. Science UK, Agri-Food and Biosciences Institute Northern Ireland, Marine 17. Branch, T.A., Watson, R., Fulton, E.A., Jennings, S., McGilliard, C.R., Scotland, International Council for the Exploration of the Sea, and the Pablico, G.T., Ricard, D., and Tracey, S.R. (2010). The trophic fingerprint Marine Biological Association UK who collected and provided fish and of marine fisheries. Nature 468, 431–435. environmental survey data and Jason Holt (Proudman Oceanographic 18. Food and Agriculture Organization of the United Nations (2010). Global Laboratory) for providing modeled SBT data. We thank two anonymous Capture Production 1950–2008. (http://www.fao.org/fishery/statistics/ referees for valuable comments. This work was supported by a Natural software/fishstat/en). Environment Research Council (NERC)/Department for Environment Food 19. Genner, M.J., Sims, D.W., Wearmouth, V.J., Southall, E.J., Southward, and Rural Affairs (Defra) Sustainable Marine Bioresources program award A.J., Henderson, P.A., and Hawkins, S.J. (2004). Regional climatic (NE/F001878/1), with additional support from Great Western Research warming drives long-term community changes of British marine fish. (M.J.G.), Defra (J.L.B. and S.J.), NERC Oceans 2025 (D.W.S. and M.J.G.), Proc. Biol. Sci. 271, 655–661. The Worshipful Company of Fishmongers (D.W.S.), and a Marine Biological 20. Cheung, W.W.L., Lam, V.W.Y., Sarmiento, J.L., Kearney, K., Watson, R., Association Senior Research Fellowship (D.W.S.). and Pauly, D. (2009). Projecting global marine biodiversity impacts under climate change scenarios. Fish Fish. 10, 235–251. Received: July 5, 2011 21. Beare, D.J., Burns, F., Greig, A., Jones, E.G., Peach, K., Kienzle, M., Revised: August 5, 2011 McKenzie, E., and Reid, D.G. (2004). Long-term increases in prevalence Accepted: August 5, 2011 of North Sea fishes having southern biogeographic affinities. Mar. Ecol. Published online: September 15, 2011 Prog. Ser. 284, 269–278. Please cite this article in press as: Simpson et al., Continental Shelf-Wide Response of a Fish Assemblage to Rapid Warming of the Sea, Current Biology (2011), doi:10.1016/j.cub.2011.08.016 Current Biology Vol 21 No 18 6

22. Jennings, S., Greenstreet, S.P.R., and Reynolds, J.D. (1999). Structural change in an exploited fish community: A consequence of differential fishing effects on species with contrasting life histories. J. Anim. Ecol. 68, 617–627. 23. Intergovernmental Oceanographic Commission, International Hydrographic Organization, and British Oceanographic Data Centre (2003). Centenary Edition of the GEBCO Digital Atlas. (http://www. gebco.net). 24. Met Office (2010). HadISST 1.1 - Global sea-Ice coverage and SST (1870–Present). (http://badc.nerc.ac.uk). 25. Proudman Oceanographic Laboratory Ocean Modelling System (2010). Ocean Monitoring and Forecasting. (http://catalogue.myocean.eu.org/ external/en/NWS-NERCPOL-NWS-AMM_PHYS-RAN_long.html). 26. International Council for the Exploration of the Seas (2010). Stock Assessment Summary/Standard Graph Database. (http://www.ices. dk/datacentre/StdGraphDB.asp). 27. Daan, N., Gislason, H., Pope, J.G., and Rice, J.C. (2005). Changes in the North Sea fish community: Evidence of indirect effects of fishing? ICES J. Mar. Sci. 62, 177–188. 28. Achcar, F., Camadro, J.M., and Mestivier, D. (2009). AutoClass@IJM: A powerful tool for Bayesian classification of heterogeneous data in biology. Nucleic Acids Res. 37, W63–W67. Current Biology, Volume 21

Supplemental Information Continental Shelf-Wide Response of a Fish Assemblage to Rapid Warming of the Sea Stephen D. Simpson, Simon Jennings, Mark P. Johnson, Julia L. Blanchard, Pieter-Jan Schön, David W. Sims, and Martin J. Genner

Supplemental Inventory

1. Supplemental Figures and Tables

Figure S1, related to Figure 2A

Figure S2, related to Figure 2A

Figure S3, related to Figure 2A

Figure S4, related to Figures 2B and 2C and Figure 4

Table S1

Table S2

2. Supplemental Experimental Procedures

3. Supplemental References

80-84

28%

85-89

22%

90-94

12%

95-99

12%

00-04

12%

05-08

Figure S1, Related to Figure 2A. Distributions of 12 biogeographic clusters identified by Bayesian analysis of presence-absence data, showing the cells and the percentage of total cells that changed between consecutive 5-year periods in the panels on the right.

1980-1984 1985-1989 1990-1994

1995-1999 2000-2004 2005-2008

Figure S2, Related to Figure 2A. Structuring of nine biogeographic clusters, identified by Bayesian analysis of presence-absence data for persistent species. Analysis included only species that occurred in 50% or more of the hauls in a cell in a 5-year block.

1980-1984 1985-1989 1990-1994

1995-1999 2000-2004 2005-2008

1980-1984 1985-1989 1990-1994

1995-1999 2000-2004 2005-2008

1985-1989 1990-1994 1995-1999 2000-2004

1985-1989 1990-1994 1995-1999

2000-2004 2005-2008

Figure S3. Structuring of Biogeographic Clusters, Identified by Bayesian Analysis of Presence-Absence Data from Single Surveys, Related to Figure 2A (A) Eight clusters identified within the North Sea spring International Bottom- Trawl Survey. (B) Four clusters identified within the North Sea autumn Cefas survey. (C) Four clusters identified within the Celtic Sea Cefas survey. (D) Two clusters identified within the NW Scotland spring FRS survey. This analysis demonstrates that geographically coherent clusters occur within single surveys, matching those from the analysis of all 11 surveys combined, rather than being driven by differences between surveys, and that overlapping surveys (A & B) produce similar qualitative patterns despite being conducted by multiple agencies in different seasons.

Sea-surface temperature

Sea-bottom temperature

Figure S4, Related to Figures 2B and 2C and Figure 4. Strength of correlation of the first axis of community assemblage-level change (PC1) with time series for different temperature metrics. Mean ± se for 252 separate time series, which for the 18 different temperature measures were restricted in duration to include all the same years, enabling direct comparison of association. Sea-surface temperature with a one-year time lag (circled in red) showed the closest association and was used for subsequent species- and community-level analyses.

Table S1. Details of the 11 Fisheries Surveys Used in This Study, Showing Numbers of Cells and Trawls Used for the Final Analysis

Survey Season Gear # Years Years covered # 1º Cells # Trawls

AFBI Irish Sea Q3 autumn otter: rockhopper 16 1992-2007 6 726

AFBI Irish Sea Q1 spring otter: rockhopper 17 1992-2008 6 742

otter: rockhopper 1984-1993, CEFAS Celtic Sea winter (spring pre-2003) 20 27 1312 or GOV 1995-2004

CEFAS Eastern Channel summer beam 19 1989-2007 9 1775

CEFAS Irish Sea autumn beam 18 1990-2007 13 2038

CEFAS North Sea autumn otter: Granton pre '92, then GOV 26 1982-2007 63 1804

CEFAS Western Channel autumn beam 18 1989-2006 4 1100

FRS NW Scotland Q1 spring otter: GOV 24 1985-2008 22 1024

FRS NW Scotland Q4 winter otter: GOV 18 1990-2007 18 580

ICES IBTS North Sea Q1 spring otter: GOV 29 1980-2008 96 11341

MBA Western Channel throughout year otter trawls 31 1967-'79, '83-'94, 2001-'07 1 427

Table S2. Extended Matrix of Species-Level Abundance Responses to Warming Showing Traits, Life-History Parameters, and Values of Significant Indices (Mean Correlations for All Cells within a Geographic Area)

colder sub-regions warmer sub-regions

Mean sea-surface temperature (°C) 11.12 9.78 10.00 10.04 10.16 10.50 10.94 11.02 11.06 11.92 12.69 13.11 13.93 Mean rise in temperature (°C y-1) 0.044 0.043 0.061 0.043 0.057 0.065 0.026 0.066 0.064 0.038 0.022 0.031 0.031

Centre of Maximum Max size in Preferred Northern Skagerrak / Central West Central Central East Northwest Southeast Southwest Preferred All Regions Irish Sea Channel Celtic Sea Shelf-edge Common name Scientific Name Family latitudinal body size database temperature North Sea Kattegat North Sea North Sea North Sea Scotland North Sea North Sea depth (m) (n=169) (n=13) (n=12) (n=16) (n=9) range (°N) (mm) (mm) (°C) (n=22) (n=7) (n=18) (n=16) (n=14) (n=22) (n=7) (n=13) red mullet Mullus surmuletus Mullidae 38 400 420 80 11.6 0.28 0.32 * 0.27 0.50 0.46 0.27 - 0.37 * 0.32 0.28 - 0.13 * 0.19 * grey gurnard Eutrigla gurnardus Triglidae 45 600 530 96 10.7 0.26 0.45 0.42 0.50 0.60 0.56 0.18 - 0.15 * - - - - splendid alfonsino Beryx splendens Berycidae 1 † 700 310 246 14.0 0.26 x x x x x x x x x x x 0.26 red bandfish Cepola macrophthalma Cepolidae 35.5 800 670 64 10.8 0.24 * x x - x x - x x - - - - red gurnard Aspitrigla cuculus Triglidae 37.5 500 530 117 12.1 0.19 0.26 * x 0.31 0.29 * - 0.25 0.14 0.12 * 0.31 - - - higher John dory Zeus faber Zeidae 13 † 900 600 127 12.2 0.18 0.24 0.19 * 0.29 0.22 0.21 0.32 - - 0.20 -0.25 * - 0.22 * abundance scaldfish Arnoglossus laterna Bothidae 39.5 250 300 59 11.4 0.18 - 0.38 0.17 * 0.26 * 0.50 - - 0.39 0.15 * - - - when shorthorn sculpin Myoxocephalus scorpius Cottidae 60 600 390 45 9.8 0.17 - 0.41 * 0.32 0.23 * 0.21 * - - 0.17 * - - x x temperatures hagfish Myxine glutinosa Myxinidae 48.5 800 790 116 9.8 0.17 0.19 * 0.51 - 0.12 * - x x x x x x x are warm lemon sole Microstomus kitt Pleuronectidae 57.5 650 490 97 10.7 0.17 0.34 0.25 * 0.24 0.48 0.28 - - 0.17 * - - - - small-spotted catshark Scyliorhinus canicula Scyliorhinidae 37.5 1000 800 107 11.7 0.17 0.20 - 0.29 0.18 * 0.20 * 0.26 - 0.29 * 0.20 * - - - dab Limanda limanda Pleuronectidae 57.5 400 450 76 10.4 0.15 0.23 0.50 0.32 0.41 0.29 - - 0.15 * - -0.22 * - - boarfish Capros aper Caproidae 36 300 190 161 13.0 0.15 0.17 * x - - x - x x - - 0.24 - thickback sole Microchirus variegatus Soleidae 37.5 350 270 118 12.3 0.15 0.12 * x 0.20 x - 0.25 x - - - 0.19 - hake Merluccius merluccius Merlucciidae 47 1400 1150 133 11.5 0.14 0.37 0.08 * - 0.19 - 0.29 x - - - - - lesser weever Echiichthys vipera Trachinidae 39.5 150 200 42 11.5 0.14 x - - 0.38 * 0.21 - 0.27 0.29 - - 0.13 * -0.14 * plaice Pleuronectes platessa Pleuronectidae 46.5 1000 660 70 10.6 0.13 0.15 0.40 0.35 0.20 0.21 - - - 0.21 * - - x spotted dragonet Callionymus maculatus Callionymidae 38.5 160 300 121 11.3 0.12 0.36 0.36 * - 0.25 0.12 * -0.10 * 0.22 0.24 - - - - solenette Buglossidium luteum Soleidae 30.5 150 220 50 11.0 0.11 - 0.41 - - 0.47 -0.26 0.38 * 0.46 0.15 * - x x tub gurnard Chelidonichthys lucerna Triglidae 37.5 750 650 58 11.8 0.11 - 0.26 0.21 - 0.17 * 0.16 * - - 0.16 * - - - dragonet Callionymus lyra Callionymidae 40.5 300 380 80 11.2 0.11 0.18 0.37 0.17 0.25 0.35 -0.10 0.31 * 0.36 - - - - poor cod Trisopterus minutus Gadidae 47 400 350 98 11.1 0.10 0.41 - 0.24 - - 0.21 ------blue whiting Micromesistius poutassou Gadidae 52.5 500 440 147 12.2 0.08 - - - -0.15 * -0.24 * 0.15 * x - 0.23 * - 0.22 * - long-rough dab Hippoglossoides platessoides Pleuronectidae 59 826 710 98 10.2 0.07 * 0.33 0.44 0.21 * 0.27 - -0.25 - - -0.30 * - - - pogge Agonus cataphractus Agonidae 60 210 260 49 11.0 0.06 * 0.14 - - 0.18 - -0.22 * 0.25 * 0.16 * - - x x fourbeard rockling Enchelyopus cimbrius Lotidae 46.5 410 450 86 10.0 0.05 * - - - - 0.15 * -0.15 * 0.15 * 0.28 - - - - cuckoo ray Leucoraja naevus Rajidae 37.5 710 800 119 11.3 0.05 * 0.10 * x - - 0.17 0.23 ------silvery cod Gadiculus argenteus argenteus Gadidae 49 150 190 158 11.2 - 0.18 - - - - - x x - x - - spotted ray Raja montagui Rajidae 38.5 800 820 80 11.0 - - x - 0.16 * 0.19 * - - - - -0.16 * - -0.23 * greater weever Trachinus draco Trachinidae 46.5 530 440 41 12.0 - - - x 0.11 * - x - - - - x x pouting Trisopterus luscus Gadidae 43.5 460 460 44 11.7 ------megrim Lepidorhombus whiffiagonis Scophthalmidae 47.5 600 650 145 12.0 - 0.22 - - - 0.28 - x 0.14 * - - -0.21 - anglerfish Lophius piscatorius Lophiidae 52.5 2000 1360 120 11.2 - 0.28 - - - - - x - -0.19 * - - - Dover sole Solea solea Soleidae 42 700 530 47 11.8 - - 0.27 * ------pearlsides Maurolicus muelleri Sternoptychidae 8.5 † 80 170 121 11.1 ------0.19 * x -0.12 * - x 0.13 * - blackbelly rosefish Helicolenus dactylopterus dactylopterus Sebastidae 12 † 470 450 170 11.6 ------x - x x - - thorny skate Amblyraja radiata Rajidae 52.5 1050 800 95 10.0 - - - -0.22 - 0.16 * -0.09 * - 0.14 * x x x x flounder Platichthys flesus Pleuronectidae 51 600 610 38 10.1 ------0.27 - - - - - x x witch Glyptocephalus cynoglossus Pleuronectidae 54.5 600 540 113 10.3 - - 0.21 * ------three-spined stickleback Gasterosteus aculeatus aculeatus Gasterosteidae 48.5 110 130 42 10.7 ------x x saithe Pollachius virens Gadidae 55 1300 1170 139 10.4 - 0.16 ------0.20 * -0.22 * - x -0.19 * - haddock Melanogrammus aeglefinus Gadidae 57 1120 820 107 10.4 -0.04 * -0.22 - - - -0.15 * -0.15 * -0.27 * -0.16 * 0.21 * - 0.26 - lower pollack Pollachius pollachius Gadidae 54 1300 1010 121 11.7 -0.05 * - -0.41 * ------abundance whiting Merlangius merlangus Gadidae 53.5 700 690 88 10.5 -0.06 * - 0.13 * -0.12 * - -0.15 * -0.35 ------when ling Molva molva Lotidae 55 2000 1570 128 11.5 -0.06 * ------x - - -0.19 * -0.19 * - temperatures thornback ray Raja clavata Rajidae 49.5 1050 1120 82 11.5 -0.08 -0.21 - -0.17 -0.19 * -0.12 * 0.11 * 0.16 * - - - -0.19 - are warm Norway pout Trisopterus esmarkii Gadidae 63.5 350 300 110 10.3 -0.09 ------0.29 x - -0.16 * - - - Norway redfish Sebastes viviparus Sebastidae 60.5 350 510 136 10.1 -0.10 -0.15 - - - - -0.14 * x - x x x x Atlantic cod Gadus morhua Gadidae 57.5 2000 1370 91 10.4 -0.12 -0.16 - -0.15 -0.18 * -0.12 * - -0.26 * -0.11 * -0.21 * - - - spurdog Squalus acanthias Squalidae 46.5 1600 1210 110 11.2 -0.13 -0.14 - -0.15 -0.18 -0.22 * - x -0.13 * - - -0.26 -

† denotes species which are found in northern and southern hemispheres. !"!#$%&'$(!)$*%!+$(,&%($(!'-*'!*+$!(./%.0.1*%'!2-$%!345657!89'!%&'!*0'$+!:&;)<(!($=9$%'.*;!>&%0$++&%.!1&++$1'.&% - denotes species that did not vary significantly with temperature x denotes species that were not present in a sub-region Supplemental Experimental Procedures Survey Data We compiled data from 11 demersal surveys conducted by the Centre for Environment Fisheries and Aquaculture Science UK (Cefas), Agri-Food and Biosciences Institute Northern Ireland (AFBI), Fisheries Research Services Scotland (FRS, now Marine Scotland), International Council for the Exploration of the Sea (ICES) and the Marine Biological Association UK (MBA) between 1980 and 2008. Individual hauls covered an area of 15° latitude and 25° longitude and were allocated to a grid of 218 1x1° cells (Table S1). Eighteen pelagic species were removed prior to analysis (of these, >99% by numbers were sprat (Sprattus sprattus), herring (Clupea harengus), mackerel (Scomber scombrus) and horse mackerel (Trachurus trachurus)) since they are more likely to be captured during the shooting and retrieving of nets than during the period when the net is in contact with the seabed [7], and data from 1x1° cells where <70% of the years were sampled within a time series were removed, as were very rare trawls with misreported locations (<0.01% of hauls). The resulting dataset consisted of 25,612 hauls comprising an estimated 105,304,825 fish from 177 species/taxa in 172 1x1° cells.

Environmental and Fishing Data We obtained bathymetry data from the GEBCO Digital Atlas [S1], Sea-Surface Temperature (SST) data from the UK Meteorological Office Hadley Centre global ocean surface temperature database (HadISST1.1) [S2] and Sea-Bottom Temperature (SBT) data from the Proudman Oceanographic Laboratory [S3]. These SST and SBT data products closely matched in situ measurements taken during survey hauls in the 22% and 11% of hauls for which they were available respectively (Pearson’s correlation; SST: n = 5666, r = 0.94, P < 0.001; SBT: n = 2789, r = 0.78, P < 0.001). The weaker correlation for the SBT data was largely due to 270 hauls for which the modelled SBT data were >5°C below the in situ measurements. These hauls were predominantly late summer (mean late-August, median month September), so are likely to be for cells where the thermocline persisted slightly later than was predicted by the model. If excluded from the analysis, the correlation for the remaining 2519 cells was greatly improved (r = 0.91, P < 0.001), suggesting that the SBT data product generally provided a good prediction of local conditions. Data from both sources were provided as monthly means for 1x1° cells, from which we calculated winter (January-March), summer (July-September) and annual mean temperatures. We used ICES stock assessment data [S4] for demersal species in our dataset to calculate standardised multispecies estimates of fishing mortality weighted by spawning-stock biomass [S5] for each ICES region. We used FAO commercial landings data [S6] to calculate catch (tonnage) per species for the northeast Atlantic fishing area during each decade. The trend in landings was calculated for each species as the slope of the line through the decadal means of the log (x+1) annual landings for the 1980s, 1990s and 2000s.

Dealing with Issues of Variable Catchability Between Surveys For some species, it is likely that estimates of abundance in the dataset vary not only due to spatial and temporal factors, but also due to differences in catchability between the surveys used in our analyses. This issue perhaps explains why previous studies have focussed on data from a single dataset, at the cost of determining a more regional understanding of the effects of recent warming. Potential discrepancies may be caused by surveys being conducted in different seasons, with contrasting gear types, or the counting and identification being conducted by different teams of observers. We addressed this issue of variable catchability explicitly throughout our analyses.

For assessment of overall changes to species occurrence, species presence- absence was quantified over 5-year periods. This was used in place of estimates of abundance, which would be more likely to vary due to season and gear. Additionally, the resultant spatial patterns of ”biogeographic” clusters over time were confirmed by repeating analyses independently using data from each of the four surveys with greatest spatial coverage (>20 1x1° cells). For assessment of species- and assemblage-level responses, we used either a Eulerian (1x1° cell grid-cell) based approach, or derived indices robust to differences in the magnitude of catches. In theory, if a fish population is declining or increasing, even if the species differs in susceptibility to different fishing gears, we expect identical temporal correlation coefficients of abundance with a predictor variable (for example temperature). Thus, by generating this coefficient for each cell in each different survey independently, the derived indices are standardised both within and between surveys. Similarly, we can use this approach of using a derived correlation coefficient from each cell in each different survey separately to quantify the change in local assemblage composition explained by a predictor variable (Assemblage-Level Responses below).

Spatial Patterns of Species Presence-Absence To identify biogeographic clusters and to test for temporal shifts in their distributions we employed an unsupervised Bayesian clustering approach [S7] to determine the most likely number of clusters, with presence-absence data for six 5-year time periods analysed in a single analysis to allow direct comparison of the distributions of clusters between time periods. Additionally, to ensure that the distribution of clusters was not driven by differences in data quality among surveys, we repeated this multiple-survey analysis excluding species with a frequency-of-occurrence in hauls of less than 50%. Further, to ensure that the clusters were not simply demarcations of different surveys, we analysed presence-absence data from the four geographically large surveys (>20 1x1° cells) separately to test whether clusters persisted in these more localised, but methodologically standardised, single datasets.

Species-Level Responses To characterise the abundance response of species to temperature, we used Pearson’s correlations to quantify the association of mean annual log (x+1) abundance with temperature data for each 1x1º cell. Separate analyses were conducted on data from each survey in cells where surveys overlapped, and a mean species response was calculated for such cells. We focussed on the 50 most abundant species which comprised 99.9% of all individuals sampled, since low abundance compromises the power to detect population changes in surveys [S8]. In parallel with our assemblage-level analysis (below), we used mean annual SST from the previous year for the temperature data (see Figure S4 for a comparison of associations of community-level change with different temperature time series). One- sample t-tests were used to identify species where the sub-regional and regional correlations were significantly different to zero, indicating a positive or negative !"#$%&'()*&+,-&%*,%".+"!$%/!"0,)&1(&2&3$(3",-$),4"%"!.&("4,/)&(1,5'%*,6,7,898:,$(4, $4;/)%"4, 6-values using Holm’s sequential Bonferroni correction [S9]. Finally, we derived a species-level index of response in abundance to temperature, using the mean correlation coefficient for all cells where a species was present, which was positive if a species was more abundant in warm years and negative where a species declined. For each species biogeographic affinities were determined using the central latitudinal range from FishBase [S10] (Table S2), which provides information on species distributions beyond the range of this study but may be susceptible to

variable reporting by different countries, and also calculated the central latitudinal tendency of each species within the compiled trawl dataset. These two measures of latitudinal preference were significantly associated when compared for the 50 most abundant species in the database (Pearson’s r = 0.64, P < 0.001). The maximum body size was determined for each species from length records in our compiled data. Where the maximum size in the dataset was >1.2 times greater than the maximum length in FishBase, the length-distribution of the species was analysed, and if the maximum was clearly an outlier (probably due to misidentification or erroneous recording), the largest credible size was used. Central latitudinal tendency, and also preferred temperature and depth, were derived for each species from the dataset using frequency-of-occurrence (foo) in each cell and year rather than abundance, combined with cell latitude and measures of in situ temperature and depth. E.g. for preferred temperature of a species:

survey ,cell ,year "1 foo #tempcell , year Tpref " ! survey ,cell ,year "n foo

Thus the estimated parameters account for different numbers of hauls and gears used in different cells across the dataset.

Assemblage-Level Responses We derived an assemblage-level response to warming for each cell by calculating the association between trends in the reorganisation of community composition with trends in temperature above the level of association that could be expected by chance. Following previous studies [S11-13], we used annual mean log (x+1)- transformed catch data for each species in each cell, so each year was a single sample, and analysed data from overlapping surveys separately to avoid issues of variable catchability of species due to differences in the trawl gears employed and the season of sampling in different surveys. Data were reduced using Principal Component Analysis from 177 taxa to the major axis of variation (PC1), which captured a mean of 34.1% of variation in the data (236 survey/cell PCAs, range: 20.3-73.6%). Although PC2 captured a mean of 18.3% of the variation (range 6.9- 33.2%), we focussed our analysis on the main trend captured in the PC1 scores. We compared PC1 scores for each cell in each survey with SST and SBT data, using annual, summer and winter means with a time lag of 1-3 years (Figure S4). To 2 determine the strength of the influence of temperature in each cell, the R actual-value (derived from r, the Pearson’s correlation coefficient for PC1 with temperature) was 2 compared with the mean R random-value from 100 randomisations of the same data for 2 2 each cell in each survey. Where the difference between R actual and R random is positive, this represents the amount of variation in PC1 above random that associates with the temperature trend. When a given cell was sampled by more than one survey, a mean measure of assemblage response to temperature was calculated for the cell using values calculated from data from each survey separately, thereby avoiding direct comparison of catches from different gears or seasons.

Supplemental References S1. IOC, IHO, and BODC. (2003). Centenary Edition of the GEBCO Digital Atlas. (http://www.gebco.net; British Oceanographic Data Centre). S2. Met Office. (2010). HadISST 1.1 - Global sea-Ice coverage and SST (1870- Present). (http://badc.nerc.ac.uk; British Atmospheric Data Centre). S3. Proudman Oceanographic Laboratory Ocean Modelling System (POLCOMS). (2010). (http://catalogue.myocean.eu.org/external/en/NWS-NERCPOL-NWS- AMM_PHYS-RAN_long.html; Proudman Oceanographic Laboratory). S4. ICES. (2010). Stock Assessment Summary/Standard Graph Database. (http://www.ices.dk/datacentre/StdGraphDB.asp; International Council for the Exploration of the Seas). S5. Daan, N., Gislason, H., Pope, J. G., and Rice, J. C. (2005). Changes in the North Sea fish community: evidence of indirect effects of fishing? ICES J. Mar. Sci. 62, 177-188. S6. FAO. (2010). Global Capture Production 1950-2008. (http://www.fao.org/fishery/statistics/software/fishstat/en; Food and Agriculture Organization of the United Nations). S7. Achcar, F., Camadro, J. M., and Mestivier, D. (2009). AutoClass@IJM: a powerful tool for Bayesian classification of heterogeneous data in biology. Nucleic Acids Res. 37, W63-W67. S8. Blanchard, J. L., Maxwell, D. L., and Jennings, S. (2008). Power of monitoring surveys to detect abundance trends in depleted populations: the effects of density-dependent habitat use, patchiness, and climate change. ICES J. Mar. Sci. 65, 111-120. S9. Holm, S. A simple sequential rejective multiple test procedure. (1979). Scand. J. Stat. 6, 65-70. S10. Froese, R., and Pauly, D. (2010). FishBase. (http://www.fishbase.org; WorldFish Center). S11. Genner, M. J., Sims, D. W., Wearmouth, V. J., Southall, E. J., Southward, A. J., Henderson, P. A., and Hawkins, S. J. (2004). Regional climatic warming drives long-term community changes of British marine fish. Proc. R. Soc. London, Ser. B 271, 655-661. S12. Mahon, R., Brown, S. K., Zwanenburg, K. C. T., Atkinson, D. B., Buja, K. R., Claflin, L., Howell, G. D., Monaco, M. E., O'Boyle, R. N., and Sinclair, M. (1998). Assemblages and biogeography of demersal fishes of the east coast of North America. Can. J. Fish. Aquat. Sci. 55, 1704-1738. S13. Genner, M. J., Sims, D. W., Southward, A. J., Budd, G. C., Masterson, P., McHugh, M., Rendle, P., Southall, E. J., Wearmouth, V. J., and Hawkins, S. J. (2010). Body size-dependent responses of a marine fish assemblage to climate change and fishing over a century-long scale. Global Change Biol. 16, 517-527.