This is the peer reviewed version of the following article: Carozza, D. A., Bianchi, D., & Galbraith, E. D. (2019). Metabolic impacts of climate change on marine ecosystems: Implications for fish communities and fisheries. Global Ecology and Biogeography, 28(2), 158-169, which has been published in final form at https:// doi.org/10.1111/geb.12832. This article may be used for non-commercial purposes in accordance with Wiley Terms and Conditions for Use of Self-Archived Versions. 1 Metabolic impacts of climate change on marine ecosys- 2 tems: implications for fish communities and fisheries 3 4 Short running title: Climate and marine ecosystem metabolism 5 6 David A. Carozza1,2 , Daniele Bianchi3 , Eric D. Galbraith1,4,5 7 Corresponding Author: David A. Carozza 8 1 9 Department of Earth and Planetary Sciences, McGill University, Montreal, Canada 10 (david.carozza@gmail.com) 2 11 Now at Département de mathématiques, Université du Québec à Montréal, Mon- 12 treal, Canada 3 13 Department of Atmospheric and Oceanic Sciences, University of California, Los An- 14 geles, 520 Portola Plaza, Los Angeles, California, 90095, USA (dbianchi@atmos.ucla.edu) 4 15 Institut de Ciència i Tecnologia Ambientals (ICTA), Universitat Autonoma de 16 Barcelona, 08193 Barcelona, Spain 5 17 Institució Catalana de Recerca i Estudis Avançats (ICREA), 08010 Barcelona, 18 Spain (eric.d.galbraith@gmail.com) 1 19 Acknowledgements: D.A.C was supported by the Social Sciences and Humanities 20 Research Council of Canada through a Joseph-Armand Bombardier Canada Gradu- 21 ate Scholarship, by the Marine Environmental Observation Prediction and Response 22 (MEOPAR) Network through a doctoral fellowship and operational support, and the 23 Birks Family Foundation through a doctoral bursary. Computational infrastructure 24 was provided by the Canadian Foundation for Innovation (grant no. 25402). This 25 project has received funding from the European Research Council (ERC) under the 26 European Unions Horizon 2020 research and innovation programme (grant agree- 27 ment No 682602). D.B. acknowledges support from NASA grant TunaScape award 28 number 80NSSC17K0290, California Department of Resources - Ocean Protection 29 Council grant C0100400, and a Faculty Research Grant from University of Califor- 30 nia, Los Angeles. We thank Elizabeth A. Fulton for comments on an earlier version 31 of this manuscript, and thank Ian A. Hatton for comments and discussions. 32 33 Author contributions: E.D.G. supervised the project and all authors contributed 34 equally to the model development and experimental design. D.B. and D.A.C con- 35 ducted simulations and analyses. D.A.C. and E.D.G. wrote the manuscript and 36 prepared the figures, with input from D.B. 2 37 Abstract 38 Aim: Climate change will reshape marine ecosystems over the 21st century through 39 diverse and complex mechanisms that are difficult to quantitatively assess. Here we 40 characterize expectations for how marine community biomass will respond to the en- 41 ergetic consequences of changes in primary production and temperature-dependent 42 metabolic rates, under a range of fishing/conservation scenarios. 43 Location: Global ocean. 44 Time period: 1950-2100. 45 Major taxa studied: Commercially-harvested marine ectotherms (’fish’). 46 Methods: We use a size-structured macroecological model of the marine ecosystem, 47 coupled with a catch model that allows for calibration with global historical data 48 and simulation of fishing. We examine the four energetic mechanisms that, within 49 the model framework, determine the community response to climate change: net 50 primary production, phytoplankton cell size, and the temperature dependencies of 51 growth and natural mortality. 52 Results: Climate change decreases the modeled global fish community biomass by 53 as much as 30% by 2100. This results from a diminished energy supply to upper 54 trophic levels as photosynthesis becomes more nutrient-limited and phytoplankton 55 cells shrink, and from a temperature-driven increase of natural mortality that, to- 56 gether, overwhelm the effect of accelerated somatic growth rates. Ocean circulation 57 changes drive regional variations of primary production, producing patterns of win- 58 ners and losers that largely compensate each other when averaged globally, whereas 59 decreasing phytoplankton size drives weaker but more uniformly-negative changes. 3 60 The climate impacts are similar across the range of conservation scenarios, but are 61 slightly amplified in the strong conservation scenarios due to the greater role of nat- 62 ural mortality. 63 Main conclusions: The spatial pattern of climate impacts is mostly determined by 64 changes in primary production. The overall decline of community biomass is at- 65 tributed to a temperature-driven increase of natural mortality, alongside an overall 66 decrease in phytoplankton size, despite faster somatic growth. Our results highlight 67 the importance of the competition between accelerated growth and mortality in a 68 warming ocean. 69 Keywords 70 conservation, fisheries, global climate change, marine communities, marine ecosystem 71 model, metabolic impacts, net primary production, temperature change 4 72 Introduction 73 Energy is supplied at the base of the marine ecosystem by Net Primary Produc- 74 tion (NPP), generally thought to be dependent on water temperature, sunlight, and 75 the availability of nutrient elements at the ocean surface (Moore et al., 2013). This 76 energy, embodied as organic matter, is then transferred to marine heterotrophic or- 77 ganisms, which span many orders of magnitude in size, through feeding relationships. 78 At each trophic step in the ecosystem, some portion of the biomass-energy is used 79 to construct the tissues of the consumer, while the remainder is either ejested or 80 respired. Thus, the fate of the available NPP, as it is distributed through trophic 81 links in the ecosystem and is ultimately respired, determines the abundance and 82 size-distributions of animals in the marine ecosystem. 83 Climate change is now altering both the total NPP and the trophic links in the 84 ecosystem. These alterations are brought about by multiple drivers, including warm- 85 ing the water, changing the distribution and composition of phytoplankton, altering 86 habitat, modifying ecosystem structure, reducing dissolved oxygen concentrations, 87 increasing acidification and shifting seasonality (Prtner et al., 2014). Although most 88 of these changes are difficult to predict, water temperature and net primary produc- 89 tion (NPP) are routinely projected by the current generation of Earth System Models 90 (Bopp et al., 2013), as emergent properties of physics and biogeochemistry in response 91 to atmospheric forcing. The temperature and NPP changes should have direct im- 92 pacts on ecosystem metabolism, since NPP plays a role in limiting whole ecosystem 93 biomass (Ware and Thomson, 2005; Chassot et al., 2010), and the metabolic rates 94 of growth and respiration depend strongly on temperature, as evident both at the 5 95 physiological level of individual animals (Kooijmann, 2000; Pörtner, 2002; Schulte, 96 2015) and at the level of whole ecosystems (Brown et al., 2004). Here, we quanti- 97 tatively estimate these metabolic consequences, as driven by the temperature and 98 NPP changes predicted by an Earth System Model for the 21st century, through the 99 lens of a model of global fish communities. 100 Earth System Models generally predict that as surface waters are warmed as a 101 result of rising atmospheric carbon dioxide, the nutrient supply to the sunlit surface 102 is reduced, and the rate of net primary production declines (Bopp et al., 2013). In 103 addition, observations have shown that warmer, less nutrient-rich waters tend to host 104 smaller phytoplankton cells (Daufresne et al., 2009; Dutkiewicz et al., 2004), which 105 are preferentially eaten by small zooplankton, leading to longer trophic chains (Ry- 106 ther, 1969). Because most of the energy consumed at a given trophic level is lost to 107 respiration, lengthening the trophic chain reduces the fraction of energy from primary 108 production that can be transferred to larger organisms (Ryther, 1969; Woodworth- 109 Jefcoats et al., 2012; Stock et al., 2017). Warming of waters also affects ectothermic 110 organisms by increasing their metabolic rates (Gillooly et al., 2001; Clarke and Fraser, 111 2004) and is commonly expected to produce more rapid growth alongside more rapid 112 respiration, activity, and predation (Pepin, 1991). Different species react differently 113 to changes in temperature, a process that can further depend on other physiological, 114 chemical, and ecological variables (Rall et al., 2012; Seebacher et al., 2014; Deutsch 115 et al., 2015). The net impact of temperature on variables such as production and 116 biomass at the species or ecosystem level is therefore difficult to ascertain. 117 The wild-capture fishery offers a perspective on the global marine ecosystem that 118 can help resolve these questions, while simultaneously playing a major role as the 6 119 dominant top predator in the oceans (Darimont et al., 2015). Although the high 120 cost of accessing and sampling most of the ocean, compounded by the mobility of 121 many marine organisms, has impeded the development of comprehensive scientific 122 assessments of global marine biomass, marine organisms are intensively sampled by 123 fishers in search of commercially marketable organisms. Fishing vessels are active 124 throughout most of the world ocean (Kroodsma et al., 2018), and the global catch has 125 recently approached, or slightly exceeded, the total production capacity for exploited 126 species (Pauly and Zeller, 2016; Galbraith et al., 2017). Historical fish catch records 127 therefore represent a valuable source of scientific information on the marine ecosystem 128 - but one which is filtered through the economic drivers of fisheries, and which has 129 also altered the marine ecosystem through time. Interpreting the output of this 130 filter requires a framework that can simultaneously take into account both the natural 131 ecosystem dynamics and the behaviour of fishers. At the same time, fisheries reshape 132 the ecosystem directly in a way that will interact with future climate change. 133 Here we apply a number of macroecological principles with broad empirical sup- 134 port to better understand some of the interactive impacts that climate change and 135 fishing activities could have on marine ecosystems. Specifically, we present a first- 136 order assessment of how changes in water temperature and NPP could affect the 137 global marine fish community through ecosystem metabolism, considering multiple 138 future fisheries regulation scenarios. We use BOATS, a bioenergetically-constrained 139 size-based global model that represents the harvested fraction of the marine ecosys- 140 tem with a generalized, low level of ecological detail (Carozza et al., 2016), integrated 141 with a simple prognostic representation of fisheries economics (Carozza et al., 2017). 142 The model does not explicitly resolve individual species, which are certain to migrate 7 143 and evolve as conditions change (Sunday et al., 2012). Instead, the model implicitly 144 assumes that, on a multi-decadal timescale, migration and evolution will adjust local 145 ecosystems to result in a stationary relationship with a given set of environmental 146 conditions. In other words, the model assumes that as environmental conditions 147 shift, the ecosystem shifts along with them, which is likely to be an optimistic as- 148 sumption. Nor do we resolve changes in species assemblage, which are likely to be 149 important additional consequences of both fisheries regulations and climate change 150 (Pecl et al., 2017), but focus instead on the total abundance of fish. 151 We use a model ensemble in which parameters are optimized against historical 152 fish catch and stock assessment data from ecosystems throughout the ocean, ensur- 153 ing a realistic rate of fish production as a function of NPP and water temperature 154 (Carozza et al., 2017). We subdivide the simulated effects of climate change into four 155 mechanistic elements, as represented within the model: 1) the total energy available 156 to the community from net primary production, 2) the impact of phytoplankton 157 size on trophic transfer, 3) the temperature dependence of somatic growth, and 4) 158 the temperature dependence of natural fish mortality. Although we use only one 159 model architecture in our ensemble, which contributes to unavoidable uncertainty in 160 the quantitative accuracy of our results, we focus the analysis on general patterns 161 and principles that are likely to apply to the real ocean, while identifying important 162 outstanding uncertainties in need of further investigation. 8 163 Methods 164 Macroecological model 165 The BiOeconomic mArine Trophic Size-spectrum model (BOATS) model is described 166 in detail by Carozza et al. (2016) and Carozza et al. (2017), which focus on the 167 ecosystem and the parameter optimization procedure, respectively. The model is 168 publicly available for download at https://github.com/davidcarozza/boats0d-review, 169 (see the Data Availability Statement below). Here we provide a general overview of 170 the model, focusing on the most relevant aspects for the current work. Supporting 171 Information Appendix S1 provides a descriptive example for the use of the BOATS. 172 BOATS is designed to run on a 2-dimensional horizontal grid of the ocean, and 173 evolves over time in response to environmental and human factors. It uses the shal- 174 low subsurface water temperature (top 75 meters) and vertically-integrated NPP in 175 each grid cell as inputs, which determine the flow of energy through the commu- 176 nity and its accumulation as biomass (Figure 1). The simulations here represent 177 all harvested marine ectotherms, which we refer to as fish, within three spectra of 178 size classes, i.e. continuous ranges of logarithmically-spaced size classes from 10 179 g (juveniles) to a spectrum-dependent maximum size. The model employs empiri- 180 cal parameterizations to describe phytoplankton community structure, the trophic 181 transfer of primary production from phytoplankton to fish, natural mortality, and 182 recruitment. Avoiding the need to compute feeding relationships simplifies model 183 dynamics and reduces computational expense, allowing extensive global-scale cali- 184 bration and ensemble simulations 185 In BOATS, the total energy input to growth (somatic and reproductive) of an 9 186 individual fish is determined by the local primary production and ecosystem trophic 187 transfer efficiency, to an upper limit that is the maximum rate at which a well-fed 188 fish can grow (von Bertalanffy, 1949; Andersen and Beyer, 2015). Water tempera- 189 ture modifies the upper limit growth rate through a van’t Hoff–Arrhenius tempera- 190 ture dependence, which is parameterised with a representative activation energy of 191 metabolism (Gillooly et al., 2001). The fraction of the resulting input energy that is 192 allocated to reproduction, as opposed to somatic growth, increases as fish approach 193 their maximum size (Andersen and Beyer, 2015). The somatic growth rate of a fish 194 within a given size spectrum therefore depends on the local energy source from NPP 195 (i.e. within the local grid cell), the local trophic transfer efficiency, and the local 196 temperature which determines the upper limit. 197 The trophic transfer of NPP to fish depends on the size structure of phytoplank- 198 ton, which we estimate using the empirical algorithm of Dunne et al. (2005). This 199 algorithm predicts the fraction of primary production that is generated by large phy- 200 toplankton in each grid cell from the in situ NPP and water temperature. We employ 201 this large fraction to estimate the average phytoplankton size. The trophic level of a 202 fish of a given size is then calculated from the mass ratio of that fish to the average 203 phytoplankton, and using an average predator-to-prey mass ratio for the community. 204 The fraction of NPP that can be taken up by fish of a given size is then given by its 205 trophic level and the average trophic efficiency. This simple approach captures the 206 basic size-dependence of energy distribution within the community, while avoiding 207 the complexity of explicit feeding relationships. Implicitly, it assumes that most fish 208 are opportunistic feeders, and that variations in predator-to-prey mass ratios tend 209 to be approximately compensated by opposing changes in trophic efficiency, leading 10 210 to constant efficiencies of total energy transfer to fish of a given size. 211 To represent natural mortality, defined here as all non-harvesting sources of fish 212 mortality and including losses to predation, parasitism, disease, old age, and star- 213 vation (Brown et al., 2004), we apply the empirical mortality rate of Gislason et al. 214 (2010). This formulation varies mortality as a function of individual fish mass and 215 asymptotic mass, and depends on temperature through a van’t Hoff–Arrhenius re- 216 lationship. To capture physiological differences between growth and predation rates 217 (Rall et al., 2012), we employ a different activation energy of metabolism parameter 218 in each of the van’t Hoff–Arrhenius temperature dependence relationships for growth 219 and mortality. 220 BOATS simulates fishing activity from simple economic principles, as described 221 in detail by Carozza et al. (2017). In brief, the fish caught in a grid cell is calculated 222 as the product of catchable biomass, effort (the fishing energy exerted per unit area), 223 and a catchability constant that represents the fraction of biomass that is caught for 224 a unit amount of effort. The effort is either imposed at the level that achieves the 225 Maximum Sustainable Yield, or allowed to vary independently in each cell according 226 to an Open Access dynamic. Using one of these two general frameworks for the 227 fishing rate, we consider four fishing scenarios that are described further below and 228 summarized in Table 1. 229 Model parameters were optimized using a Monte Carlo Approximate Bayesian 230 Computation approach (Csillry et al., 2010), using the global catch data of the Sea 231 Around Us Project and the RAM Legacy Stock Assessment database as observational 232 constraints (Ricard et al., 2011). Supporting Information Appendix S1 details the 233 parameter optimization approach (Carozza et al., 2017). Importantly, this procedure 11 234 includes comparing the modeled fish catches among all Large Marine Ecosystems 235 to those observed, to ensure a realistic production rate of fish biomass under the 236 global range of present-day NPP and water temperatures. From a subset of 100 237 acceptable parameter combinations, we chose a subsample of six different parameter 238 combinations. We refer to the collection of six parameter combinations as the model 239 ensemble, and to each of the individual parameter combinations as an ensemble 240 member. Supporting Information Appendix S2 details the parameter values and 241 global characteristics of the six ensemble members used in this study. 242 Like any model, BOATS provides a simplified representation of reality. The 243 model ignores a multitude of potential stressors, such as the impact of phenology 244 on recruitment (Asch, 2015), explicit inter-species interactions, decreased oxygen 245 concentrations (Cheung et al., 2013; Prtner and Peck, 2010) and ocean acidification 246 (Fabry et al., 2008; Briffa et al., 2012). BOATS also does not resolve movement 247 between oceanic grid cells (Watson et al., 2015), which could be important for the 248 adaptation of large predatory fish to changing food patterns, or changes to ecosystems 249 due to bottom-trawling (Puig et al., 2012). Nonetheless, the structural simplicity of 250 the model is necessary in order to allow the parameter optimization with historical 251 fishing observations, which ensures a well-calibrated response to water temperature 252 and NPP. In addition, the inclusion of prognostic fishing effort allow it to estimate 253 how basic macroecological impacts of long-term climate change could interact with 254 conservation efforts. 12 255 Simulation strategy 256 To estimate the relative roles of the energetic mechanisms that affect fish communities 257 under changing NPP and water temperature, we conducted six sets of simulations, 258 summarized in Table 1. Each of the first four sets isolates a specific aspect of the 259 macroecological response to climate change, by allowing anthropogenic changes in 260 NPP and/or water temperature to apply only to that aspect. The fifth set allows 261 all elements to change together, and the last represents a constant climate over the 262 period of analysis. In detail, the sets of simulations are: 263 1. NPP. Isolates the effects of changes in the input of energy to the base of the 264 food web by photosynthesis. In the model, an individual fish of size m is limited 265 by the proportion of NPP that is transferred to all fish of size m through the 266 local food web, divided by the number of fish in that size class. Because this 267 energy is partitioned uniformly among all fish of size m, the individual growth 268 rate will increase (up to a maximum physiological rate) when NPP increases 269 and/or the number of fish decreases. The fraction of NPP that can reach size 270 m depends on the trophic efficiency and the predator to prey mass ratio, both 271 of which are global constants that differ for each ensemble member. Greater 272 NPP also improves larval survival by increasing the flux of biomass from mature 273 individuals that enters the smallest size classes (recruitment). 274 2. PhytoSize. Isolates the effects of changes in phytoplankton size structure on 275 fish growth rates. The transfer of energy from NPP to fish of size m depends 276 on phytoplankton cell size, since this contributes to determining the trophic 277 distance (Ryther, 1969; Woodworth-Jefcoats et al., 2012; Stock et al., 2017). 13 278 In BOATS, the empirical model of Dunne et al. (2005) is used to estimate the 279 fraction of primary production that is attributed to large phytoplankton as a 280 function of temperature and NPP. Higher productivities and lower tempera- 281 tures favor larger phytoplankton sizes. 282 3. TempGrowth. Isolates the impact of temperature on the maximum physio- 283 logical growth rate of fish. In the model, individual fish cannot grow faster than 284 a maximum rate that follows the widely-used von Bertalanffy growth formu- 285 lation (von Bertalanffy, 1949; Hartvig et al., 2011; Andersen and Beyer, 2015) 286 and depends on the individual fish size (relative to its maximum size) as well 287 as temperature, following the van’t Hoff–Arrhenius equation. As temperature 288 increases, so does the maximum physiological growth rate. 289 4. TempMortality. Isolates the impact of temperature on natural fish mortal- 290 ity. BOATS represents the natural (i.e. non-fishing) mortality rate using the 291 empirical formulation of Gislason et al. (2010), as the product of a natural 292 mortality constant, a temperature-dependent term that is based on the van’t 293 Hoff–Arrhenius equation, individual mass, and the asymptotic mass (Carozza 294 et al., 2016). Note that temperature in the model affects fish growth and mor- 295 tality rates differently, consistent with the distinct physiological and ecological 296 processes controlling somatic growth vs. respiration and predation rates (Gis- 297 lason et al., 2010; Rall et al., 2012). The magnitudes of the two activation 298 energies are allowed to vary independently of each other in the Monte Carlo 299 procedure, so that the six-member ensemble includes six different combinations 300 of the activation energies. 14 301 5. Total. Simulates the overall climate impact by including all four of the above 302 mechanisms simultaneously. 303 6. Constant Climate. Simulates no climate change. Forces the model with 304 a constant climate of the monthly averages calculated from the preindustrial 305 period of 1851-1900. 306 Simulation design 307 We force the six optimal model ensemble members described above with net pri- 308 mary production and temperature output from the Institut Pierre Simon Laplace 309 IPSL-CM5A-LR global climate model (Dufresne, et al., 2013), which employs the 310 PISCES biogeochemical model (Aumont and Bopp, 2006), under the business as 311 usual RCP8.5 scenario (Moss, et al., 2010). Figure 2a,b show the preindustrial water 312 temperature and NPP (average of years 1851-1900) used for our constant climate 313 scenario, respectively, of the IPSL-CM5A-LR model output, while Figure 2c,d show 314 the change in water temperature and NPP between 1851-1900 and 2081-2100. Simi- 315 larly, Supporting Information Appendix S3 presents the preindustrial phytoplankton 316 size and its change compared to 2081-2100, respectively, estimated with the method 317 of Dunne et al. (2005). 318 For each of the six ensemble members, we conduct simulations under four idealized 319 regulation scenarios that span a broad range of possible futures (Table 1) following 320 Galbraith et al. (2017). These scenarios are intended to illustrate the bounds of 321 possibility, rather than being detailed attempts at future predictions. The base 322 scenario has no fishing effort, which we call the ’Perfect Conservation’ case, while 15 323 a second scenario allows fishing effort at the level ’Optimized for Human Food’ 324 production, commonly known as the Maximum Sustainable Yield (MSY) fishing 325 rate. The remaining two scenarios do not include regulations, but simulate profit- 326 driven fishing effort following the open-access principle (Carozza et al., 2017) under 327 the assumption of either a near-future stabilization (No Conservation scenario), or 328 a continuing intensification of fishing (Intense Overfishing scenario). Supporting 329 Information Appendix S1 further details the simulation protocol. 330 Results 331 Global changes of total biomass 332 Under Perfect Conservation, climate change reduces the globally-integrated marine 333 biomass by 32 % (lower estimate -33, upper estimate -29 %) by 2100 (Figure 3a; 334 Figure 4). The increase in the natural mortality rate (simulation TempMortality) 335 brought on by a warming ocean has the single greatest negative impact on biomass, 336 resulting in a decrease of 43 % (-46, -33 %) by 2100. Net primary production (simula- 337 tion NPP) has a minor negative or negligible impact on globally-integrated biomass, 338 reducing it by only 3 % (-5, 1 %), whereas the shift to smaller phytoplankton cells 339 (simulation PhytoSize) accounts for a more significant biomass reduction of 13 % (- 340 15, -9 %). Warming waters have a positive impact on biomass through their impact 341 on the growth rate upper limit (simulation TempGrowth), raising biomass by 18 % 342 (16, 28 %). 343 The impact of climate change on fish biomass in the Optimized for Human Food 344 scenario is similar to the Perfect Conservation scenario (Figure 3b; Figure 4), with 16 345 an overall decline of 32 % (-35, -29 %). As in the case with Perfect Conservation, 346 this decline is mostly driven by increasing mortality under warming, reinforced by 347 declines in productivity and phytoplankton size, which are only partially offset by 348 faster growth rates. Interestingly, the range of variability among ensemble members 349 is much larger for the TempMortality and TempGrowth simulations (Figure 4), re- 350 vealing a large sensitivity to the uncertain parameters. However, this sensitivity is 351 greatly reduced in the Total simulations, indicating that the temperature sensitivity 352 of growth is correlated with the temperature sensitivity of mortality in any given 353 ensemble member. This correlation is consistent with the constraint identified in 354 Carozza et al. (2017) that, for realistic global harvests to arise from the mdel param- 355 eters, temperature-driven increases in growth must be balanced by parallel increases 356 in mortality. 357 Under the No Conservation scenario (Figure 3c; Figure 4), the negative impacts 358 of climate change are significantly damped relative to the Perfect Conservation and 359 Optimized for Human Food cases. Here, climate change only results in a loss of 15 360 % (-20, -12 %) of biomass by 2100. The reduced climate impact is mainly driven 361 by a weakened negative impact of the mortality rate; because fishing and natural 362 mortality both act to reduce fish abundance, Intense Overfishing reduces the rela- 363 tive importance of natural mortality. Biomass changes due to primary production 364 (simulation NPP) and temperature-dependent growth (simulation TempGrowth) are 365 similar to those without fishing, but the impact of phytoplankton size (simulation 366 PhytoSize) is significantly damped, since when the number of fish is reduced, growth 367 rates are determined by the size-dependent physiological upper limit rather than by 368 primary production (Carozza et al., 2016). 17 369 In the extreme Intense Overfishing scenario, the impacts of the NPP and Phyto- 370 Size mechanisms disappear almost entirely, and the overall impact of climate change 371 to year 2100 is equivocal (Figure 4). The extremely intense fishing rate further damps 372 the negative impact of TempMortality, which causes a biomass fall of only 7 % (-37, 373 1 %), which is then entirely compensated by the more rapid growth rates. However, 374 we caution that this effect only occurs in the model under extreme, and likely un- 375 realistic values of harvesting technologies, at which point ecosystems are decimated 376 and the average global fish catch is very small (Supporting Information Appendix 377 S4g). 378 Spatial patterns of change 379 As shown in Figure 5a, the net reduction of global biomass under climate change 380 does not reflect a uniform global decrease, but a patchwork of increases and decreases 381 that largely compensate each other in the global sum. Reductions over the tropics 382 and mid- to high-latitudes are partially counteracted by increases in subtropical (e.g. 383 South Pacific and South Atlantic gyres) and polar regions (in particular the Southern 384 Ocean), and over eastern boundary upwelling systems (California, Chile, and Canary 385 Islands). 386 The pattern of net change closely resembles the responses driven by primary 387 production (simulation NPP, Figure 5b), and to a lesser extent by phytoplankton 388 size structure (PhytoSize, Figure 5c). However, the latter are generally shifted to- 389 ward more negative values due to the effect of warming, which tends to decrease 390 phytoplankton size everywhere. Thus, whereas NPP changes result in regional pat- 391 terns that largely cancel each other out, phytoplankton size changes produce weaker 18 392 regional contrasts but a more significant negative global impact. 393 The uniquely temperature-dependent impacts on growth and mortality (simula- 394 tions TempGrowth and TempMortality, Figure 5d,e) are more spatially homogeneous 395 than those driven by net primary production, due to the homogeneous distribution of 396 warming (Figure 2) and have opposite and nearly compensating effects on biomass. 397 The activation energy of mortality is more sensitive to temperature than that of 398 growth in all but one of our ensemble members (Supporting Information Appendix 399 S2). Increases of biomass relative to the constant climate scenario only occur in re- 400 gions where NPP increases enough to overcome the combined effect of enhanced mor- 401 tality and shrinking phytoplankton cells. The simulated changes in fisheries catches 402 (harvest) are qualitatively similar to the simulated biomass changes, as shown in 403 Supporting Information Appendix S7. 404 Discussion 405 Our results show a large negative impact of climate change on marine fish commu- 406 nities from metabolic effects, when summed at the global scale. The main ecological 407 mechanisms driving this decrease are the temperature-sensitivity of natural mortal- 408 ity, which reflects enhanced dissipation of biomass by respiration in warmer water, 409 and a decrease in phytoplankton size, which reduces the energy available to fish by 410 trophic transfer for a given rate of NPP. These deleterious effects are opposed by the 411 increase of growth rates at higher temperatures, but this is insufficient to compensate 412 for the negative effects in any of our six ensemble members. Under intense levels of 413 overfishing, the negative impacts of climate are lessened because of a reduced impor- 19 414 tance of natural mortality, a diminished competition for resources, and an increased 415 importance of growth when fish populations are greatly impoverished. 416 We find significant regional variability in the impacts of climate change, mostly 417 driven by the spatial patterns of NPP changes simulated by the Earth System Model. 418 While most of the tropics and mid-latitudes show a decline in fish biomass, some re- 419 gions actually show an increase of these quantities, particularly in the Southern 420 Ocean, South Pacific and South Atlantic gyres, and some Eastern Boundary Up- 421 welling Systems. The current generation of Earth System models indicates a sub- 422 stantial degree of uncertainty in projections for NPP (Bopp et al., 2013) and the 423 details of this mosaic of winners and losers should therefore be viewed with caution. 424 Nonetheless, the spatial heterogeneity of NPP changes typically simulated by mod- 425 els, compared to the much more homogeneous warming, suggests that the dominance 426 of NPP in determining spatial patterns is a robust result. 427 Our results in the Perfect Conservation set of simulations are generally consistent 428 with those simulated by Lefort et al. (2014), despite important structural differences 429 between the models employed. Simulated biomass in BOATS falls nonlinearly with 430 decreases in the fraction of large phytoplankton, with a global spatially-weighted 431 average decrease of 3% in the large fraction (Supporting Information Appendix S3) 432 resulting in a median biomass decrease of 13%. This is a much greater sensitiv- 433 ity to the phytoplankton size than that presented by Blanchard et al. (2012) and 434 Woodworth-Jefcoats et al. (2012), in which fish biomass varied linearly with phy- 435 toplankton size, and points to the important uncertainty regarding the response of 436 trophic efficiency to climate change. 437 Our simulations also agree with Cheung et al. (2010) in many parts of the world, 20 438 but whereas those authors estimated that fish would increase in high northern lati- 439 tudes, we find decreases over much of the high northern latitudes by 2100, in agree- 440 ment with Lefort et al. (2014). This contrast likely reflects the fact that the biocli- 441 mate envelope approach applied by Cheung et al. (2010) is designed to estimate catch 442 potential for particular species based on environmental factors such as temperature, 443 but does not explicitly simulate the metabolic effect of temperature on growth and 444 mortality. Given that these aspects produce the largest climate impact on BOATS, 445 it is not surprising that Cheung et al. (2010) simulate a different spatial pattern of 446 change, as well as a weaker globally-averaged response to climate. 447 One potentially-surprising aspect of the simulations is a reduction in the negative 448 impact of climate change in the total absence of fishery regulation, as fishing pressure 449 increases to extremely high levels. This reduction is mainly due to a reduction of 450 the temperature impact on natural mortality, with further important contributions 451 from temperature-dependent growth and phytoplankton size effects (Figure 4). In a 452 hypothetical future with Perfect Conservation, biomass is large and growth is there- 453 fore significantly limited by NPP, while increasing temperature tends to have a net 454 negative effect since the natural mortality effect is larger than the growth rate effect. 455 Essentially, if primary production is the limiting factor for growth of the overall com- 456 munity, then NPP changes are important for determining total biomass. Similarly, 457 if biomass production is balanced only by natural mortality, then the temperature 458 effect on mortality is important. In contrast, under Intense Overfishing, the reduc- 459 tion of fish biomass results in more energy availability per individual fish, and so 460 somatic growth becomes less dependent on NPP and the phytoplankton community 461 size structure. 21 462 In addition, as fishing becomes a major loss term for biomass, it reduces the 463 impact of natural mortality relative to the case without harvest. Instead, the impor- 464 tance shifts to the rate at which fish can grow from juveniles to adulthood, which 465 limits the replacement rate of harvested fish. As a result, the positive impact of 466 warmer temperatures on growth rates becomes increasingly significant as fishing in- 467 tensifies, counterbalancing the negative impacts of NPP and natural mortality. 468 The real world outcomes would undoubtedly be more nuanced than in this simple 469 model framework, but we hypothesize that the general weakening of climate-driven 470 bioenergetic impacts under intense fishing is likely to be a robust feature of marine 471 ecosystems. If true, this metabolic effect would be expected to reduce the relative 472 impact of climate change on biomass in heavily exploited ecosystems, all else being 473 equal. At the same time, it may offset some gains to be made from future conservation 474 efforts, as accelerated natural mortality may consume a significant portion of the 475 biomass saved from fishing. We would caution that this implied trade-off refers only 476 to the biomass, and does not consider the impacts on other aspects of the community 477 such as species diversity. In addition, the model does not include other impacts of 478 climate change, such as ocean deoxygenation, which may interact differently with 479 fishing pressure. 480 At the same time, it is important to emphasize that overfishing could have a 481 significantly more deleterious impact on the evolution of 21st century biomass than 482 climate change (Galbraith et al., 2017). In the Intense Overfishing scenario, biomass 483 is reduced by > 90 % by relative to that of Perfect Conservation (Supporting Infor- 484 mation Appendix S4), as opposed to an average climate-change-induced reduction of 485 30 % (Figure 3a). Thus, although the metabolic impacts of climate change may be 22 486 stronger for an ecosystem protected by effective conservation, this effect is dwarfed by 487 the much larger overall benefits to be achieved through conservation. It is also pos- 488 sible that additional ecosystem resilience may be provided by conservation measures 489 (Loreau et al., 2001), which could counteract the metabolic trade-off. 490 Finally we point out that, within our observationally-calibrated macroecological 491 model, the dominant effects of temperature on ecosystem metabolism are via growth 492 and natural mortality. Thus, if other unresolved temperature-dependences affect ma- 493 rine fish communities, we expect they would have biased the parameter selection by 494 masquerading as the temperature sensitivities of growth and mortality. For example, 495 it has been suggested that trophic efficiency varies with temperature (Stock et al., 496 2017), which would cause harvests to vary with temperature in a way not explic- 497 itly simulated by the model. Our parameter selection would implicitly ‘correct’ for 498 this by including the trophic efficiency contribution in one of the other temperature 499 dependences. Similarly, ‘natural mortality’ is a simplification of a complex web of 500 processes that ultimately results in the removal of biomass from the spectrum of up- 501 per trophic level organisms; explicit representation of these processes could modify 502 their environmental sensitivities to some degree. These are important uncertainties 503 that could be addressed in future work. 504 In summary, our model predicts that climate change will reduce the total supply 505 of energy to upper trophic levels, and will accelerate the rate at which energy flows 506 through ecosystems. These changes result in a large decrease of total fish abundance 507 under the strong warming of the RCP8.5 emissions scenario. The overall negative 508 impact reflects the net outcome of opposed, nearly-compensating accelerations of 509 mortality and growth rates under warming, coupled with a shrinking of phytoplank- 23 510 ton cells that lengthens trophic chains. Meanwhile, changes in primary production 511 determine the spatial patterns of simulated climate impacts but have relatively little 512 effect on globally-integrated responses, particularly under Intense Overfishing. Fur- 513 ther work should focus on improving the quantitative, mechanistic understanding 514 of the ecological processes behind this response, particularly the poorly-constrained 515 variations in natural mortality due to rising temperatures, and the impacts of com- 516 munity structure on the transfer of energy from producers to consumers. Our results 517 also emphasize the importance of preventing overfishing through effective regula- 518 tions, with or without climate change (Worm, B et al., 2009; Galbraith et al., 2017), 519 if further loss of wild fish abundance is to be prevented. 520 References 521 Andersen, K. H. and Beyer, J. E. (2015). Size structure, not metabolic scaling rules, 522 determines fisheries reference points. Fish and Fisheries, 16:1–22. 523 Asch, R. G. (2015). Climate change and decadal shifts in the phenology of larval 524 fishes in the california current ecosystem. Proceedings of the National Academy of 525 Sciences, 112:E4065–E4074. 526 Aumont, O. and Bopp, L. (2006). Globalizing results from ocean in situ iron fertil- 527 ization studies. Global Biogeochemical Cycles, 20:GB2017. 528 Blanchard, J. L., Jennings, S., Holmes, R., Harle, J., Merino, G., Allen, J. I., Holt, J., 529 Dulvy, N. K., and Barange, M. (2012). Potential consequences of climate change for 24 530 primary production and fish production in large marine ecosystems. Philosophical 531 Transactions of the Royal Society B: Biological Sciences, 367(1605):2979–2989. 532 Bopp, L., Resplandy, L., Orr, J. C., Doney, S. C., Dunne, J. P., Gehlen, M., Halloran, 533 P., Heinze, C., Ilyina, T., Séférian, R., Tjiputra, J., and Vichi, M. (2013). Multiple 534 stressors of ocean ecosystems in the 21st century: projections with CMIP5 models. 535 Biogeosciences, 10(10):6225–6245. 536 Briffa, M., de la Haye, K., and Munday, P. L. (2012). High CO2 and marine animal 537 behaviour: Potential mechanisms and ecological consequences. Marine Pollution 538 Bulletin, 64(8):1519–1528. 539 Brown, J., Gillooly, J., Allen, A., Savage, V., and West, G. (2004). Toward a 540 metabolic theory of ecology. Ecology, 85(7):1771–1789. 541 Carozza, D. A., Bianchi, D., and Galbraith, E. D. (2016). The ecological module 542 of boats-1.0: a bioenergetically constrained model of marine upper trophic levels 543 suitable for studies of fisheries and ocean biogeochemistry. Geoscientific Model 544 Development, 9(4):1545–1565. 545 Carozza, D. A., Bianchi, D., and Galbraith, E. D. (2017). Formulation, general fea- 546 tures and global calibration of a bioenergetically-constrained fishery model. PLOS 547 ONE, 12:1–28. 548 Chassot, E., Bonhommeau, S., and Dulvy, N. (2010). Global marine primary pro- 549 duction constrains fisheries catches. Ecology Letters, 13:495–505. 550 Cheung, W. W. L., Lam, V. W. Y., Sarmiento, J. L., Kearney, K., Watson, R., 551 Zeller, D., and Pauly, D. (2010). Large-scale redistribution of maximum fisheries 25 552 catch potential in the global ocean under climate change. Global Change Biology, 553 16(1):24–35. 554 Cheung, W. W. L., Sarmiento, J. L., Dunne, J., Frölicher, T. L., Lam, V. W. Y., 555 Deng Palomares, M. L., Watson, R., and Pauly, D. (2013). Shrinking of fishes 556 exacerbates impacts of global ocean changes on marine ecosystems. Nature Climate 557 Change, 3:254–258. 558 Clarke, A. and Fraser, K. P. P. (2004). Why does metabolism scale with temperature? 559 Functional Ecology, 18(2):243–251. 560 Csillry, K., Blum, M. G., Gaggiotti, O. E., and Franois, O. (2010). Approxi- 561 mate Bayesian Computation (ABC) in practice. Trends in Ecology & Evolution, 562 25(7):410 – 418. 563 Darimont, C. T., Fox, C. H., Bryan, H. M., and Reimchen, T. E. (2015). The unique 564 ecology of human predators. Science, 349(6250):858–860. 565 Daufresne, M., Lengfellner, K., and Sommer, U. (2009). Global warming benefits 566 the small in aquatic ecosystems. Proceedings of the National Academy of Sciences, 567 106(31):12788–12793. 568 Deutsch, C., Ferrel, A., Seibel, B., Pörtner, H., and Huey, R. (2015). Climate change 569 tightens a metabolic constraint on marine habitats. Science, 348:1132–1135. 570 Dufresne, et al. (2013). Climate change projections using the IPSL-CM5 Earth 571 System Model: from CMIP3 to CMIP5. Climate Dynamics, 40(9-10):2123–2165. 26 572 Dunne, J., Armstrong, R., Gnanadesikan, A., and Sarmiento, J. (2005). Empirical 573 and mechanistic models for the particle export ratio. Global Biogeochemical Cycles, 574 19(4). 575 Dutkiewicz, S., Scott, J. R., and Follows, M. J. (2004). Winners and losers: Ecological 576 and biogeochemical changes in a warming ocean. Global Biogeochemical Cycles, 577 27(2):463–477. 578 Fabry, V. J., Seibel, B. A., Feely, R. A., and Orr, J. C. (2008). Impacts of ocean 579 acidification on marine fauna and ecosystem processes. ICES Journal of Marine 580 Science, 65(3):414–432. 581 Galbraith, E. D., Carozza, D. A., and Bianchi, D. (2017). A coupled human-Earth 582 model perspective on long-term trends in the global marine fishery. Nature Com- 583 munications, 8:14884. 584 Gillooly, J. F., Brown, J. H., West, G. B., Savage, V. M., and Charnov, E. L. (2001). 585 Effects of size and temperature on metabolic rate. Science, 293(5538):2248–2251. 586 Gislason, H., Daan, N., Rice, J. C., and Pope, J. G. (2010). Size, growth, temperature 587 and the natural mortality of marine fish. Fish and Fisheries, 11(2):149–158. 588 Hartvig, M., Andersen, K. H., and Beyer, J. E. (2011). Food web framework for 589 size-structured populations. Journal of theoretical Biology, 272(1):113–122. 590 Kooijmann, S. A. L. M. (2000). Dynamic Energy Mass Budgets in Biological Systems. 591 Cambridge: Cambridge University Press. 27 592 Kroodsma, D. A., Mayorga, J., Hochberg, T., Miller, N. A., Boerder, K., Ferretti, 593 F., Wilson, A., Bergman, B., White, T. D., Block, B. A., Woods, P., Sullivan, 594 B., Costello, C., and Worm, B. (2018). Tracking the global footprint of fisheries. 595 Science, 359(6378):904–908. 596 Lefort, S., Aumont, O., Bopp, L., Arsouze, T., Gehlen, M., and Maury, O. (2014). 597 Spatial and body-size dependent response of marine pelagic communities to pro- 598 jected global climate change. Global Change Biology, 21(1):154–164. 599 Loreau, M., Naeem, S., Inchausti, P., Bengtsson, J., Grime, J., Hector, A., Hooper, 600 D., Huston, M., Raffaelli, D., Schmid, B., et al. (2001). Biodiversity and ecosystem 601 functioning: current knowledge and future challenges. science, 294(5543):804–808. 602 Moore, C. M., Mills, M. M., Arrigo, K. R., Berman-Frank, I., Bopp, L., Boyd, 603 P. W., Galbraith, E. D., Geider, R. J., Guieu, C., Jaccard, S. L., Jickells, T. D., 604 La Roche, J., Lenton, T. M., Mahowald, N. M., Maran, E., Marinov, I., Moore, 605 J. K., Nakatsuka, T., Oschlies, A., Saito, M. A., Thingstad, T. F., Tsuda, A., and 606 Ulloa, O. (2013). Processes and patterns of oceanic nutrient limitation. Nature 607 Geoscience, 6:701–710. 608 Moss, et al. (2010). The next generation of scenarios for climate change research and 609 assessment. Nature, 463(7282):747–756. 610 Pauly, D. and Zeller, D. (2016). Catch reconstructions reveal that global marine 611 fisheries catches are higher than reported and declining. Nature Communications, 612 7:10244. 28 613 Pecl, G. T., Araújo, M. B., Bell, J. D., Blanchard, J., Bonebrake, T. C., Chen, I.-C., 614 Clark, T. D., Colwell, R. K., Danielsen, F., Evengård, B., et al. (2017). Biodiversity 615 redistribution under climate change: Impacts on ecosystems and human well-being. 616 Science, 355(6332):eaai9214. 617 Pepin, P. (1991). Effect of temperature and size on development, mortality, and 618 survival rates of the pelagic early life history stages of marine fish. Canadian 619 Journal of Fisheries and Aquatic Sciences, 48(3):503–518. 620 Pörtner, H.-O. (2002). Climate variations and the physiological basis of temperature 621 dependent biogeography: systemic to molecular hierarchy of thermal tolerance in 622 animals. Comparative Biochemistry and Physiology Part A: Molecular & Integra- 623 tive Physiology, 132(4):739–761. 624 Puig, P., Canals, M., Company, J. B., Martn, J., Amblas, D., Lastras, G., Palan- 625 ques, A., and Calafat, A. M. (2012). Ploughing the deep sea floor. Nature, 626 489(7415):286–289. 627 Prtner, H.-O., Karl, D., Boyd, P., Cheung, W., Lluch-Cota, S., Nojiri, Y., Schmidt, 628 D., and PO, Z. (2014). Ocean systems. In Field, C., Barros, V., Dokken, D., Mach, 629 K., Mastrandrea, M., Bilir, T., Chatterjee, M., Ebi, K., Estrada, Y., Genova, R., 630 Girma, B., Kissel, E., Levy, A., MacCracken, S., Mastrandrea, P., and White, 631 L., editors, Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part 632 A: Global and Sectoral Aspects. Contribution of Working Group II to the Fifth 633 Assessment Report of the Intergovernmental Panel on Climate Change, chapter 6, 634 pages 411–484. Cambridge University Press, Cambridge, UK and New York, USA. 29 635 Prtner, H. O. and Peck, M. A. (2010). Climate change effects on fishes and fisheries: 636 towards a cause-and-effect understanding. Journal of Fish Biology, 77(8):1745– 637 1779. 638 Rall, B. C., Brose, U., Hartvig, M., Kalinkat, G., Schwarzmuller, F., Vucic-Pestic, 639 O., and Petchey, O. L. (2012). Universal temperature and body-mass scaling 640 of feeding rates. Philosophical Transactions of the Royal Society B: Biological 641 Sciences, 367(1605):2923–2934. 642 Ricard, D., Minto, C., Jensen, O. P., and Baum, J. K. (2011). Examining the 643 knowledge base and status of commercially exploited marine species with the RAM 644 Legacy Stock Assessment Database. Fish and Fisheries, 13(4):380–398. 645 Ryther, J. H. (1969). Photosynthesis and fish production in the sea. Science, 646 166(3901):72–76. 647 Schulte, P. M. (2015). The effects of temperature on aerobic metabolism: towards a 648 mechanistic understanding of the responses of ectotherms to a changing environ- 649 ment. Journal of Experimental Biology, 218(12):1856–1866. 650 Seebacher, F., White, C. R., and Franklin, C. E. (2014). Physiological plasticity 651 increases resilience of ectothermic animals to climate change. Nature Climate 652 Change, 5(1):61–66. 653 Stock, C. A., John, J. G., Rykaczewski, R. R., Asch, R. G., Cheung, W. W. L., 654 Dunne, J. P., Friedland, K. D., Lam, V. W. Y., Sarmiento, J. L., and Watson, 655 R. A. (2017). Reconciling fisheries catch and ocean productivity. Proceedings of 656 the National Academy of Sciences, 114(8):E1441–E1449. 30 657 Sunday, J. M., Bates, A. E., and Dulvy, N. K. (2012). Thermal tolerance and the 658 global redistribution of animals. Nature Climate Change, 2:686–690. 659 von Bertalanffy, L. (1949). Problems of Organic Growth. Nature, 163(4135):156–158. 660 Ware, D. M. and Thomson, R. E. (2005). Bottom-up ecosystem trophic dynamics 661 determine fish production in the northeast pacific. Science, 308(5726):1280–1284. 662 Watson, J. R., Stock, C. A., and Sarmiento, J. L. (2015). Exploring the role of 663 movement in determining the global distribution of marine biomass using a coupled 664 hydrodynamic size-based ecosystem model. Progress in Oceanography, 138, Part 665 B:521 – 532. 666 Woodworth-Jefcoats, P. A., Polovina, J. J., Dunne, J. P., and Blanchard, J. L. (2012). 667 Ecosystem size structure response to 21st century climate projection: large fish 668 abundance decreases in the central North Pacific and increases in the California 669 Current. Global Change Biology, 19(3):724–733. 670 Worm, B et al. (2009). Rebuilding global fisheries. Science, 325(5940):578–585. 671 Data Availability Statement 672 The BOATS model was written in MATLAB version R2012a and was also tested 673 in version R2010b. BOATS for a single patch of ocean, which includes the model 674 run script, required functions, and forcing data, is archived and available for down- 675 load under the reference Carozza, David A, Bianchi, Daniele, & Galbraith, Eric D. 676 (2015). 0-D BiOeconomic mArine Trophic Size-spectrum (BOATS) model (Version 31 677 v1.0) [Data set]. Zenodo. http://doi.org/10.5281/zenodo.27700, and is additionally 678 available at https://github.com/davidcarozza/boats0d-review. Supporting Informa- 679 tion Appendix S1 provides a descriptive example for the use of the model. The 680 model output and all MATLAB scripts required to reproduce the figures in this ar- 681 ticle have been archived and are available for download under the reference Carozza, 682 David A, Bianchi, Daniele, and Galbraith, Eric D. (2017). Metabolic impacts of 683 climate change on marine fish communities and fisheries - Dataset and figure plot 684 script [Data set]. Zenodo. http://doi.org/10.5281/zenodo.495487. 685 Biosketch 686 David A. Carozza has a background in mathematics, economics and climate science, 687 and led the development of the BOATS model as a PhD student. He is currently 688 a Postdoctoral Fellow at the University of Quebec at Montreal studying economic 689 vulnerability and climate change. Daniele Bianchi is an Assistant Professor at the 690 University of California, Los Angeles, where he studies ocean biogeochemistry and 691 marine ecosystems. Eric Galbraith is an ICREA Research Professor based at the 692 Autonomous University of Barcelona, where he is focused on improving the large- 693 scale understanding of coupling between global ecosystems and human activity. 32 694 Supporting Information 695 Appendix S1: Simulation methods detailing parameter optimization, experimental 696 design; choice of IPSL model; simulation protocol; analyses; BOATS example use 697 instructions. 698 Appendix S2: Ensemble member characteristics. 699 Appendix S3: Maps of preindustrial values of the large phytoplankton fraction 700 (average of 1851-1900) and its change between 2081-2100 and 1851-1900. 701 Appendix S4: Globally-integrated nonnormalized biomass and fish catch. 702 Appendix S5: Maps of median normalized change in biomass due to the impact of 703 climate change on metabolism for the Perfect Conservation scenario. 704 Appendix S6: Maps of median normalized change in biomass due to the impact of 705 climate change on metabolism for the No Conservation scenario. 706 Appendix S7: Globally-integrated fish catch change, relative to the constant cli- 707 mate simulation. 33 Table 1: Metabolic simulations and conservation scenarios. Each metabolic simula- tion was conducted with all conservation scenarios, for each of the six model ensemble members, leading to a total of 144 simulations. Metabolic Simulation Input(s) from warming scenario Resulting impacts on fish NPP NPP Trophic growth limit, recruitment PhytoSize NPP and water temperature Trophic growth limit, recruitment TempGrowth Water temperature Physiological growth limit TempMortality Water temperature Natural mortality rate Total NPP and Water temperature All Clim None None Conservation Scenario Characteristics Perfect Conservation Zero fishing effort everywhere Optimized for Human Food Maximum stable fish catch everywhere No Conservation Open access, stabilizing by 2036 Intense Overfishing Open access, increasing continuously 34 Environmental Growth rate limiting terms Forcings Net Primary Production (NPP) Trophic limit: food energy potentially available to fish Physiological limit: maximum possible somatic growth rate Fish biomass spectra small group Fishing Water temperature effort Fishing large group effort natural mortality harvest growth recruitment m∞,small m∞,large log(size) Figure 1: Schematic overview of the BOATS model. The red, green, and black arrows indicate dependencies of model components on external forcings (left panel). The top-right panel indicates the energetic limits of growth as a function of fish size, whereas the bottom-right panel illustrates the size spectra of fish groups, their internal dynamics, and link to economics via fish catch and the interactive effort. 35 Preindustrial 75−meter averagewater temperature climatological temperature Preindustrial ClimatologicalNPP NPP 80oN 30 80oN 750 (a) (b) (mg C m−2 d−1 ) 40oN 40oN 20 500 (°C) 0o 0o 10 250 40oS 40oS o 80 S 0 80oS 0 WaterTemperature temperaturechange change NPP NPP change change 80oN 8 (c) 80oN 250 (d) 40oN 6 (mg C m−2 d−1 ) o 40 N 100 (°C) 0o 4 0 o 0 2 −100 40oS 40oS o 80 S 0 80oS −250 Figure 1: Climatology (1850-1900) and change (average over 2081-2100 less clima- Figure 2: Preindustrial climate forcing variables and change in those variables due to tology) in temperature and net primary production in the IPSL-CM5A-LR global climate change in the IPSL-CM5A-LR global climate model used to force BOATS. (a) climate model used to force BOATS. (a) Climatological temperature averaged over Preindustrial water temperature. (b) Preindustrial net primary production (NPP). the upper 75 meters. (b) Vertically integrated climatological net primary production. (c) Water temperature change. (d) Net primary production change. Water temper- (c) Temperature change. (d) Net primary production change. ature is averaged over the upper 75 meters of the ocean, while NPP is vertically- integrated. Preindustrial climate variables are the average over 1851-1900, whereas change is measured as the average over 2081-2100 less the average over 1851-1900. The constant climate scenario employs the preindustrial climate variables. 28 36 (a) Perfect Conservation (b) Optimized for Human Food Global biomass change (%) 40 40 Global biomass change (%) 20 20 0 0 Constant climate -20 NPP -20 Phyto Size Temp Growth −40 Temp Mortality −40 Total Change 1950 2000 2050 2100 1950 2000 2050 2100 Year (c) No Conservation (d) Intense Overfishing Global biomass change (%) Global biomass change (%) 40 40 20 20 0 0 -20 -20 −40 −40 1950 2000 2050 2100 1950 2000 2050 2100 Figure 3: Globally-integrated fish biomass change, relative to the constant climate simulation. Each panel shows the six metabolic simulations (colour-coded), for one conservation scenario. Normalized quantities for each simulation are expressed in terms of the % change relative to the constant climate biomass for that simulation by taking the mean over the 6 ensemble members. Grey vertical triangles at years 2006 and 2036 in (c) represent years where the increase in fishing technology begins to slow and stops, respectively. 37 Perfect Optimized for No Intense Conservation Human food Conservation Overfishing 60 NPP PhytoSize 40 TempGrowth TempMort Global fish biomass change (%) Total 20 0 -20 -40 -60 38 change relative to the constant climate Figure 4: Globally-integrated fish biomass scenario from 2081 to 2100. Circles represent the median over the 6 ensemble mem- bers, whereas the lower and upper bars represent the 5th and 95th percentiles over the 6 ensemble members, respectively. Total Change (a) 80 o N 100 o 40 N 50 (% change) 0o 0 40 o S -50 80 o S -100 NPP PhytoSize (b) o 80 N 100 (c) o 80 N 100 o o 40 N 50 40 N 50 (% change) (% change) 0o 0 0o 0 40 o S -50 40 o S -50 80 o S -100 80 o S -100 TempGrowth TempMortality (d) 80 o N 100 (e) 80 o N 100 40 o N 50 40 o N 50 (% change) (% change) 0 o 0 0 o 0 o 40 S -50 o 40 S -50 80 o S -100 80 o S -100 Figure 5: Maps of median normalized change in biomass due to the impacts of climate change on the metabolic model components for the Optimized for Human Food sce- nario over 2081-2100. (a) All combined effects (Total). (b) Growth rate dependence on net primary production (NPP). (c) Growth rate dependence on phytoplankton size structure (PhytoSize). (d) Growth rate upper limit temperature dependence (Temp- Growth). (e) Natural mortality rate temperature dependence (TempMortality). For each metabolic simulation and set of ensemble members, changes are calculated rel- ative to the constant climate forcing scenario. For each scenario, we calculate the median change over the 6 ensemble members of the temporal average of the normal- ized biomass over 2081-2100. Biomass change in the Perfect Conservation scenario is presented in Supporting Information Appendix S5, whereas biomass change in the No Conservation scenario is detailed in Supporting Information Appendix S6. 39