Articles https://doi.org/10.1038/s41558-021-01173-9 Next-generation ensemble projections reveal higher climate risks for marine ecosystems Derek P. Tittensor 1,2 ✉, Camilla Novaglio 3,4, Cheryl S. Harrison 5,6, Ryan F. Heneghan 7, Nicolas Barrier 8, Daniele Bianchi 9, Laurent Bopp 10, Andrea Bryndum-Buchholz 1, Gregory L. Britten 11, Matthias Büchner 12, William W. L. Cheung 13, Villy Christensen 13, Marta Coll 14,15, John P. Dunne 16, Tyler D. Eddy 17, Jason D. Everett 18,19,20, Jose A. Fernandes-Salvador 21, Elizabeth A. Fulton 4,22, Eric D. Galbraith 23, Didier Gascuel 24 , Jerome Guiet 9, Jasmin G. John 16, Jason S. Link 25, Heike K. Lotze 1, Olivier Maury 8, Kelly Ortega-Cisneros 26, Juliano Palacios-Abrantes 13,27, Colleen M. Petrik 28, Hubert du Pontavice 24,29, Jonathan Rault8, Anthony J. Richardson 18,19, Lynne Shannon 26, Yunne-Jai Shin 8, Jeroen Steenbeek 15, Charles A. Stock 16 and Julia L. Blanchard 3,4 Projections of climate change impacts on marine ecosystems have revealed long-term declines in global marine animal biomass and unevenly distributed impacts on fisheries. Here we apply an enhanced suite of global marine ecosystem models from the Fisheries and Marine Ecosystem Model Intercomparison Project (Fish-MIP), forced by new-generation Earth system model outputs from Phase 6 of the Coupled Model Intercomparison Project (CMIP6), to provide insights into how projected climate change will affect future ocean ecosystems. Compared with the previous generation CMIP5-forced Fish-MIP ensemble, the new ensemble ecosystem simulations show a greater decline in mean global ocean animal biomass under both strong-mitigation and high-emissions scenarios due to elevated warming, despite greater uncertainty in net primary production in the high-emissions scenario. Regional shifts in the direction of biomass changes highlight the continued and urgent need to reduce uncertainty in the projected responses of marine ecosystems to climate change to help support adaptation planning. A nthropogenic climate change is a growing threat to marine Projecting the magnitude and impacts of climate change through ecosystems1, with impacts projected to intensify a suite model intercomparison projects (MIPs) produces ensemble projec- of organismal responses, including increased mortality, tions that quantify inter-model spread (the range of projections reduced calcification and changes to species distributions, inter- from low to high)9 and stimulate long-term efforts to develop and actions, abundance and biomass2,3. Furthermore, climate change improve models. The most prominent MIP, the Coupled Model can interact with other stressors such as overfishing4,5, which can Intercomparison Project (CMIP)10, is currently on its sixth phase threaten marine conservation6 and societal benefits derived from of Earth system model (ESM) simulation experiments, forming a the ocean7,8. Thus, understanding the risks of climate change for core contribution to the sixth Intergovernmental Panel on Climate marine ecosystems and the benefits of mitigation is paramount. Change Assessment Report (IPCC AR6). These simulation outputs 1 Department of Biology, Dalhousie University, Halifax, Nova Scotia, Canada. 2United Nations Environment Programme World Conservation Monitoring Centre, Cambridge, UK. 3Institute for Marine and Antarctic Studies, University of Tasmania, Hobart, Tasmania, Australia. 4Center for Marine Socio-ecology, University of Tasmania, Hobart, Tasmania, Australia. 5School of Earth, Environmental and Marine Science, University of Texas Rio Grande Valley, Port Isabel, TX, USA. 6Department of Ocean and Coastal Science and Centre for Computation and Technology, Louisiana State University, Baton Rouge, LA, USA. 7School of Mathematical Sciences, Queensland University of Technology, Brisbane, Queensland, Australia. 8MARBEC, IRD, Univ Montpellier, Ifremer, CNRS, Sète/Montpellier, France. 9Department of Atmospheric and Oceanic Sciences, University of California Los Angeles, Los Angeles, CA, USA. 10LMD/IPSL, CNRS, Ecole Normale Supérieure, Université PSL, Sorbonne Université, Ecole Polytechnique, Paris, France. 11Program in Atmospheres, Oceans, and Climate, Massachusetts Institute of Technology, Cambridge, MA, USA. 12Potsdam-Institute for Climate Impact Research (PIK), Potsdam, Germany. 13Institute for the Oceans and Fisheries, The University of British Columbia, Vancouver, British Columbia, Canada. 14Institute of Marine Science (ICM-CSIC), Barcelona, Spain. 15Ecopath International Initiative Research Association, Barcelona, Spain. 16NOAA/OAR Geophysical Fluid Dynamics Laboratory, Princeton, NJ, USA. 17Centre for Fisheries Ecosystems Research, Fisheries and Marine Institute, Memorial University of Newfoundland, St. John’s, Newfoundland and Labrador, Canada. 18School of Mathematics and Physics, The University of Queensland, St. Lucia, Queensland, Australia. 19 Commonwealth Scientific and Industrial Research Organisation (CSIRO) Oceans and Atmosphere, Queensland Biosciences Precinct, St Lucia, Brisbane, Queensland, Australia. 20Centre for Marine Science and Innovation, The University of New South Wales, Sydney, New South Wales, Australia. 21 AZTI, Marine Research, Basque Research and Technology Alliance (BRTA), Sukarrieta (Bizkaia), Spain. 22Commonwealth Scientific and Industrial Research Organisation (CSIRO) Oceans and Atmosphere, Hobart, Tasmania, Australia. 23Department of Earth and Planetary Science, McGill University, Montreal, Quebec, Canada. 24UMR Ecology and Ecosystems Health (ESE), Institut Agro, Inrae, Rennes, France. 25NOAA Fisheries, Woods Hole, MA, USA. 26Department of Biological Sciences, University of Cape Town, Cape Town, South Africa. 27Center for Limnology, University of Wisconsin, Madison, WI, USA. 28Department of Oceanography, Texas A&M University, College Station, TX, USA. 29Atmospheric and Oceanic Sciences Program, Princeton University, Princeton, NJ, USA. ✉e-mail: derek.tittensor@dal.ca Nature Climate Change | VOL 11 | November 2021 | 973–981 | www.nature.com/natureclimatechange 973 Articles NaTuRE ClIMaTE CHangE Table 1 | List of global MEMs participating in the study and differences relative to CMIP5 analysis MEM Model type11 CMIP5 CMIP6 Key forcing variables Summarized Taxonomic groups Key used differences between included reference(s) CMIP5 and CMIP6 MEM structure (if any) APECOSM Composite (size- and x x Carbon concentrations To avoid problems with Epipelagic 39,40 trait-based; functional (small phytoplankton, the 3D interpolation fish, migratory group structure) large phytoplankton, of forcings, APECOSM mesopelagic fish, small zooplankton, large was run on the native resident mesopelagic zooplankton), particulate ORCA1 grid using the fish organic matter (small and native IPSL-CM set large), zonal and meridional of forcing fields. The currents, turbulent 3D outputs were then mixing, temperature, vertically integrated and water density, dissolved interpolated on the 2D oxygen concentration, light Fish-MIP 1° x 1° grid. irradiance. All fields 3D and Minor improvements monthly. include fine tuning of some parameters and bug fixes with minor impacts on the outputs. BOATS Size-based x x Mean temperature 0–75 m, None All commercially 41,42 NPP fished species, both finfish and invertebrates DBEM Species distribution x x Surface and bottom O2, pH, None 956 species of 43,44 model salinity and temperature. exploited fishes and Ice cover, current velocity, invertebrates NPP, NPP pico and NPP diat. All variables on a yearly basis. DBPM Composite (size- and x x Surface and bottom None All benthic and 45 trait-based) temperature, phytoplankton pelagic marine carbon groups animals weighing between 1 mg and 1 tonne EcoOcean Composite x x SST, seafloor temperature, (1) Improved Includes 51 functional 29,46 (trophodynamic and column average representation of groups representing species distribution temperature, phytoplankton species contributions to the whole spectrum model) carbon groups ecosystem dynamics, (2) of marine organisms improved responses of from bacteria the marine food web to to whales, and stratified environmental integrates explicit drivers information for 3,400 species of vertebrates, invertebrates and primary producers EcoTroph Trophic-level-based x NPP, SST, integrated None Implicitly all groups, 37,47 mesozooplankton carbon including pelagic and demersal fishes and invertebrates FEISTY Composite x Seafloor temperature, None Small pelagic fish, 48 seafloor detritus flux, large pelagic fish, mean temperature demersal fish, benthic 0–100 m, integrated invertebrates mesozooplankton carbon 0–100 m Macroecological Size-based x x NPP, SST None Implicitly all marine 49 organisms from 1 gram to 1 tonne ZooMSS Composite (size- and x Chlorophyll-a, SST None Flagellates, cilliates, 50 trait-based; functional omnivorous group structure) copepods, carnivorous copepods, larvaceans, salps, chaetognaths, euphausiids, jellyfish, fish All models produce monthly outputs on a 1° × 1° grid (except DBEM, which produces yearly outputs on a 0.5° × 0.5° grid). 2D, two-dimensional; 3D, three-dimensional. 974 Nature Climate Change | VOL 11 | November 2021 | 973–981 | www.nature.com/natureclimatechange NaTuRE ClIMaTE CHangE Articles a b 5 5 Scenario Scenario Historical Historical 4 RCP2.6 4 SSP1−2.6 RCP8.5 SSP5−8.5 Change in SST (°C) 3 3 2 Earth system model 2 Earth system model GFDL−ESM2M GFDL−ESM4 IPSL−CM5A−LR IPSL−CM6A−LR 1 1 0 0 −1 −1 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 c d 15 15 10 10 Change in NPP (%) 5 5 0 0 −5 −5 −10 −10 −15 −15 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 Year Year e f 5 5 Change in phytoplankton carbon (%) 0 0 −5 −5 −10 −10 −15 −15 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 g h 2100 5 5 Change in zooplankton carbon (%) 0 0 −5 −5 −10 −10 −15 −15 −20 −20 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 Year Year Fig. 1 | Projected mean global change in oceanographic properties from IPSL and GFDL ESMs. a–h, Rows depict SST (a,b), NPP (c,d), phytoplankton carbon (e,f) and zooplankton carbon (g,h) for GFDL and IPSL CMIP5 and CMIP6 under strong-mitigation (blue) and high-emissions (red) scenarios. Historical values (1970–2005 for CMIP5; 1970–2014 for CMIP6) are shown in black, and projections (2006–2099 for CMIP5; 2015–2099 for CMIP6) are coloured. All values are normalized relative to the period 1990–1999. Vertical grey shaded area indicates reference decade, and vertical grey line indicates first year of projection (subsequent to historical period). RCP, representative concentration pathway. Nature Climate Change | VOL 11 | November 2021 | 973–981 | www.nature.com/natureclimatechange 975 Articles NaTuRE ClIMaTE CHangE a b SST CMIP6 c SST change between CMIPs SST CMIP5 Difference Change in (°C) (°C) change 4 2 0° 0° 3 0° 1 2 0 1 −1 0 −2 0° °E 12 ° W 0° °W E 0° °E 12 ° 0° °E W 0° 12 ° W 0° °W °W E 0 E 0 0 18 18 0° 0° 18 18 18 18 60 0° 0° 0° 0° 60 60 60 12 60 60 12 12 d e f NPP change between CMIPs NPP CMIP5 NPP CMIP6 Difference Change in (%) (%) change 50 30 0° 0° 25 0° 15 0 0 −25 −15 −50 −30 0° °E 12 ° W 0° °W E 0° °E 12 ° W 0° °W E 0° °E 12 ° W 0° °W E 0 0 0 18 18 0° 18 18 0° 0° 18 18 60 0° 0° 60 0° 60 60 60 12 60 12 12 g h i Phytoplankton carbon change between CMIPs Phytoplankton carbon CMIP5 Phytoplankton carbon CMIP6 Difference Change in (%) (%) change 50 30 0° 0° 25 0° 15 0 0 −25 −15 −50 −30 0° °E 0° W 0° °W E 0° °E 0° W 0° °W E 0° °E 0° W 0° °W E 18 18 0° 0° 18 18 60 0° 18 18 0° 0° 0° 60 60 60 12 60 60 12 12 12 12 12 j Zooplankton carbon CMIP5 k Zooplankton carbon CMIP6 l Zooplankton carbon change between CMIPs Difference Change in (%) (%) change 50 30 0° 0° 25 0° 15 0 0 −25 −15 −50 −30 0° °E 0° °E 0° W 0° 0° °E 0° W 0° °W 0° W 0° °W E °W E E 18 18 18 18 18 18 0° 0° 0° 0° 0° 0° 60 60 60 60 60 60 12 12 12 12 12 12 Fig. 2 | Projected mean spatial changes in oceanographic properties. a–l, Rows depict SST (a–c), NPP (d-f), phytoplankton carbon (g–i) and zooplankton carbon (j–l) for 1990–1999 and 2090–2099. Maps represent mean change for GFDL and IPSL models under a high-emissions scenario. Columns depict mean change under CMIP5 (a,d,g,j), mean change under CMIP6 (b,e,h,k) and the difference in these century changes between CMIP6 and CMIP5 (c,f,i,l), with a positive value indicating a stronger increase (or weaker decrease) in CMIP6, a negative value indicating a weaker increase (or stronger decrease) in CMIP6 and zero representing an equal change projected in both CMIP5 and CMIP6. can also be used to drive impact models to explore how climate simulation has a higher spatial resolution. The bias of CMIP6 mod- change will affect specific human sectors or natural processes. els was reduced by 20–70% for surface nitrate, phosphate and sili- While individual marine ecosystem models (MEMs) have cate relative to CMIP5, and the root mean squared error for surface explored climate impacts on ocean ecosystems, the Fisheries and temperature and chlorophyll improved17,18. CMIP6 models exhibit Marine Ecosystem Model Intercomparison Project (Fish-MIP) increased climate sensitivity (the equilibrium response of mean compares models produced by different modelling groups into surface air temperature to a doubling of atmospheric CO2) over standardized ensemble projections11. Fish-MIP has explored a range CMIP519, resulting in generally stronger marine ecosystem forcings of topics, including global12 and regional13–15 changes over the com- (for example, ocean warming) under the high-emissions scenario, ing century and their potential socioeconomic consequences8. although with more variation in terms of impacts on net primary All Fish-MIP contributions to date have been driven by CMIP5 productivity (NPP)17,20. ESM outputs. However, the ‘next-generation’ CMIP6 ESMs provide Three global MEMs have been added to the Fish-MIP ensem- an updated suite of oceanographic drivers16,17. In this article, we com- ble, bringing the total to nine (Table 1). In this article, we present pare MEM ensembles forced with CMIP5 and CMIP6 variants of the results from the newly expanded suite of MEMs under CMIP6 Geophysical Fluid Dynamics Laboratory (GFDL) and the Institut high-emissions and strong-mitigation scenarios. We focus on tem- Pierre-Simon Laplace (IPSL) ESMs. CMIP6’s next-generation perature and productivity as key drivers of marine ecosystem change ESMs capture improved representations of marine biogeochem- as these variables are used by all MEMs (Supplementary Table 1). istry18, sea ice and other oceanographic properties, and the GFDL We also provide a direct comparison between the subset of MEMs 976 Nature Climate Change | VOL 11 | November 2021 | 973–981 | www.nature.com/natureclimatechange NaTuRE ClIMaTE CHangE Articles a b 5 5 0 0 Change in biomass (%) −5 −5 −10 −10 −15 −15 Scenario Scenario Historical Historical −20 −20 RCP 2.6 SSP1−2.6 RCP 8.5 SSP5−8.5 −25 −25 −30 −30 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 Year Year Fig. 3 | Multimodel mean change in marine animal biomass under strong-mitigation and high-emissions scenarios. a, CMIP5. b, CMIP6. Blue colouring represents strong mitigation, and red represents high emissions. Coloured dots indicate years with higher ensemble means for CMIP5. All values relative to standardized reference period (1990–1999). Solid coloured lines indicate ensemble means; shaded areas indicate inter-model standard deviation. Vertical grey shaded area and line indicate reference decade and first year of projection after historical period, respectively. The full ensemble of MEMs is included for CMIP5 (7 models using IPSL and 5 models using GFDL, n = 12) and CMIP6 (9 models using IPSL and 7 models using GFDL, n = 16). that used both CMIP5 and CMIP6 forcings to evaluate the con- to CMIP5 (Fig. 2d,e). The Arctic shows a blanket increase under sequences of improved next-generation climate models for global CMIP6, seemingly driven by IPSL differences relative to CMIP5 marine ecosystem projections and to assess mitigation benefits. The (Extended Data Figs. 1 and 2). Changes in phytoplankton and zoo- revised Fish-MIP simulation protocol is kept as similar as possible plankton biomass are more consistently negative, decreasing every- to the CMIP5 protocol to enable direct comparison, including using where except the poles (Fig. 2g-l), and more regionally variable for the same two ESMs (albeit from different generations) and focusing zooplankton (Fig. 2j–l). The spatial congruence between CMIP5 on climate impacts on total unfished marine animal biomass. and CMIP6 in the direction of changes for plankton biomasses, however, was greater than for NPP (Fig. 2e,h,k). ESM projections The IPSL and GFDL CMIP6 ESM simulations show a stronger MEM projections mean surface warming of the global ocean from the 1990s to the Mean projected global marine animal biomass from the full MEM 2090s relative to the CMIP5 ESM simulations in the high-emissions ensemble shows no clear difference between the CMIP5 and CMIP6 scenario but a reduced difference in the strong-mitigation scenario simulations until ~2030 (Fig. 3). After 2030, CMIP6-forced mod- (Fig. 1a,b). By contrast, there is more diversity in projected changes els show larger declines in animal biomass, with almost every year in NPP (Fig. 1c,d) and plankton biomass (Fig. 1e-h). Under CMIP5, showing a more pronounced decrease under strong mitigation only IPSL under high emissions shows an overall decline in global and most years from 2060 onwards showing a more pronounced NPP, of ~10% by 2090–2099 (Fig. 1c). All other simulations (GFDL decrease under high emissions (Fig. 3). Both scenarios have a for both scenarios and IPSL for strong mitigation) exhibit little significantly stronger decrease in 2090–2099 under CMIP6 than overall global change in NPP. However, under both CMIP6 sce- CMIP5 (two-sided Wilcoxon rank-sum test on annual values; narios, GFDL shows up to 5–10% decline in global NPP, whereas n = 160 for CMIP6, 120 for CMIP5; W = 12,290 and P < 0.01 for IPSL projects an increase of the same magnitude, with stronger strong mitigation, W = 11,221 and P = 0.016 for high emissions). responses under high emissions in both cases (Fig. 1d). The NPP For the comparable MEM ensemble (Extended Data Fig. 3), only increase in IPSL CMIP6 is especially prominent in subtropical gyres the strong-mitigation scenario is significantly different (n = 120 (Extended Data Fig. 1), and preliminary analysis suggests this could for both CMIPs; W = 6,623 and P < 0.01). The multiple consecu- be linked to a warming-induced increase in di-nitrogen fixation17. tive decades in which CMIP6 projections are more negative than Importantly, despite increasing NPP, the phytoplankton and zoo- CMIP5 (Fig. 3b and Extended Data Fig. 3b) suggest that these plankton biomasses in IPSL show marked declines for CMIP6, as results are not due simply to decadal variability in the selected ESM was the case in CMIP5, although with greater differences between ensemble members. Under high emissions, the mean marine animal the two ESMs (Fig. 1e-h). biomass for the full MEM ensemble declines by ~19% for CMIP6 Spatially, IPSL and GFDL CMIP6 ESMs project sea surface tem- by 2099 relative to 1990–1999 (~2.5% more than CMIP5), and the perature (SST) increases across almost the entire global ocean under mitigation scenario declines by ~7% (~2% more than CMIP5). the high-emissions scenario (Fig. 2b), with the largest changes in the Similar declines were observed for the comparable MEM ensemble. Arctic, northern temperate regions (except the centre of the North Notably, the ensemble inter-model standard deviations show a total Atlantic subpolar gyre) and a belt around the Equator. Polar oceans separation between high-emissions and strong-mitigation scenar- show a generally broader surface warming in CMIP6 simulations ios under CMIP6 after the 2080s for both the full and comparable than in CMIP5 (Fig. 2c and Extended Data Figs. 1 and 2) as well as MEM ensembles, whereas they overlap under CMIP5 (Fig. 3b and the highest NPP increases over the twenty-first century under the Extended Data Fig. 3). This was due primarily to a narrowing of the high-emissions scenario (Fig. 2e and Extended Data Figs. 1 and 2). inter-model standard deviation for the high-emissions scenario; all The Pacific and North Atlantic oceans contain large areas of pro- the CMIP6 GFDL-forced MEMs project a higher rate of biomass jected NPP decrease (Fig. 2e), although of reduced extent relative decrease in the latter half of the twenty-first century (Fig. 4c,d and Nature Climate Change | VOL 11 | November 2021 | 973–981 | www.nature.com/natureclimatechange 977 Articles NaTuRE ClIMaTE CHangE a b CMIP5 IPSL−CM5A−LR CMIP6 IPSL−CM6A−LR 5 5 0 0 −5 −5 Change in biomass (%) Ecosystem model −10 Ecosystem model −10 APECOSM APECOSM BOATS −15 −15 BOATS DBEM DBEM DBPM −20 DBPM −20 EcoOcean EcoOcean EcoTroph −25 MACROECOLOGICAL −25 FEISTY ZooMSS MACROECOLOGICAL −30 −30 ZooMSS −35 −35 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 c d CMIP5 GFDL−ESM2M CMIP6 GFDL−ESM4 5 5 0 0 −5 −5 Change in biomass (%) −10 −10 Ecosystem model Ecosystem model BOATS −15 BOATS −15 DBEM DBEM EcoOcean −20 EcoOcean −20 EcoTroph MACROECOLOGICAL FEISTY −25 ZooMSS −25 MACROECOLOGICAL ZooMSS −30 −30 −35 −35 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 Year Year Fig. 4 | Projected global change in marine animal biomass from individual MEMs driven by IPSL and GFDL under the high-emissions scenario. a, CMIP5 IPSL-CM5A-LR. b, CMIP6 IPSL-CM6A–LR. c, CMIP5 GFDL-ESM2M. d, CMIP6 GFDL-ESM4M. A different set of MEMs is included for CMIP5 (7 models using IPSL and 5 models using GFDL, n = 12) and CMIP6 (9 models using IPSL and 7 models using GFDL, n = 16). All values are relative to the standardized reference period of 1990–1999. Vertical grey shaded area indicates reference decade, and vertical grey line indicates first year of projection subsequent to historical period. Extended Data Fig. 4c,d), which brings those models’ projections Notably, the MEM model agreement (Extended Data Fig. 7e,f) closer to those of IPSL-forced models. is fairly high (>80%) in most oceanic regions under CMIP6, except Biomass changes across the full and comparable MEM ensembles for the subtropical gyres where the agreement drops to 50% (maxi- show similar global spatial patterns between CMIP5 and CMIP6 mum disagreement among models), probably due to the marked under both scenarios (Fig. 5a,b and Extended Data Figs. 5a,b and differences in CMIP6 NPP and the various ways MEMs incorpo- 6a,b). However, there are substantial regional differences (Fig. 5a–c rate lower-trophic-level forcing21. Regardless of whether the full or and Extended Data Figs. 5a–c and 6a–c). In all cases, CMIP6 MEM comparable MEM models are considered (Extended Data Figs. 8, 9 ensemble means project an increase in animal biomass essentially and 10), there is no clear spatial improvement in model agreement everywhere in the Arctic by the 2090s, whereas CMIP5 projects compared with CMIP5. both increases and decreases in the region. The pattern is spatially heterogeneous for other regions, although equatorial regions do Discussion and conclusions show a consistent biomass decline for both CMIP5 and CMIP6 Our comparison of GFDL and IPSL climate drivers for MEM projec- (Fig. 5a–c and Extended Data Figs. 5a–c and 6a–c). In total, 71% tions revealed a substantial spatial reshuffling of projected marine of grid cells indicate the same direction of change for the ensemble animal biomass change in the global ocean between CMIP5 and model biomass under CMIP5 and CMIP6, with 15% switching from CMIP6. Overall, these changes suggest that when averaged across decreases to increases (Fig. 5d) and another 14% vice versa (Fig. 5e). the ensemble, total marine animal biomass will decline more steeply Among the areas that change, large parts of the Arctic Ocean show when forced by CMIP6 ESMs than by CMIP5 (Fig. 3), with a greater positive changes, and a large area of temperate latitudes exhibits separation between high-emissions and strong-mitigation scenarios negative changes (Fig. 5d–e). emphasizing the benefits of mitigation (Fig. 3 and Extended Data 978 Nature Climate Change | VOL 11 | November 2021 | 973–981 | www.nature.com/natureclimatechange NaTuRE ClIMaTE CHangE Articles a CMIP5 RCP 8.5 b CMIP6 SSP5−8.5 Biomass Biomass change change (%) (%) 50 50 0° 25 0° 25 0 0 −25 −25 −50 −50 0° 0° °E 0° °W E 0° W 0° °E 0° °W E W 18 0° 18 18 0° 18 0° 0° 60 60 60 60 12 12 12 12 c d Change between CMIPs Same direction of change between CMIPs Difference Difference in (%) in (%) change change 50 50 0° 0° 25 25 0 0 −25 −25 −50 −50 0° 0° °E 0° °W E W 0° 0° °E 0° °W E W 18 0° 18 18 0° 18 0° 60 0° 60 60 60 12 12 12 12 e Shift to positive change f Shift to negative change Difference Difference in (%) in (%) change change 50 0 0° 0° 25 −25 0 −50 0° 0° 0° °E 0° °W 0° °E 0° E °W W E W 18 0° 18 18 0° 18 0° 0° 60 60 60 60 12 12 12 12 Fig. 5 | Ensemble mean change in marine animal biomass under the high-emissions scenario. The full ensemble of MEMs is included for CMIP5 (7 models using IPSL and 5 models using GFDL, n = 12) and CMIP6 (9 models using IPSL and 7 models using GFDL, n = 16). a,b, Maps represent mean percentage change between 1990–1999 and 2090–2999 under CMIP5 (a) and CMIP6 (b). c, Difference in percentage change between CMIP5 and CMIP6. d, Difference in percentage change between CMIP5 and CMIP6 for grid cells showing same direction of change. e, Difference in percentage change between CMIP5 and CMIP6 for grid cells changing from CMIP5 decrease to CMIP6 increase. f, Same as in e for grid cells changing from CMIP5 increase to CMIP6 decrease. Fig. 3). Differences in projected biomass changes appeared to be forcing variables, particularly NPP, were substantially altered under caused primarily by increased climate sensitivity of CMIP6 simula- the high-emissions scenario in CMIP6, with directional differences tions22, specifically the IPSL and GFDL simulations used here. This between the two ESMs at the global scale (Fig. 1d), although the phy- is supported by the comparable MEM ensemble results (Extended toplankton and zooplankton biomasses that support upper trophic Data Figs. 3–5 and 8). levels decrease in a very similar way to CMIP5. Four of nine MEMs Warming can affect metabolic costs, rates of biomass produc- use NPP as their primary input (BOATS, DBEM, EcoTroph and tion, mortality and species distributions and interactions in indi- MACROECOLOGICAL), while others are forced by phytoplankton vidual MEMs; a detailed study of the response of individual MEMs and/or zooplankton biomass or a proxy thereof (DBPM, EcoOcean to warming in isolation21 revealed a substantial variability in mecha- and ZooMSS) or combine plankton biomass with particulate organic nisms and responses, although a broadly consistent negative impact matter (APECOSM and FEISTY) to generate new animal produc- on biomass. Warming is also accompanied by increased ocean tion (Table 1 and Supplementary Table 1). These differences within stratification and a marked biomass decrease of non-nitrate-fixing Fish-MIP represent structural variability in the MEMs, and the sen- phytoplankton23. Combining these mechanisms, the pattern is one sitivity of the results, as well as the general agreement, should be seen of consistent global animal biomass decline under both CMIP5 and as a test of how robust the result of declining MEM biomass under CMIP6, with a strengthened decline under CMIP6. climate change is to the ecological and other assumptions of the The impact of changing productivity and biomass in the lower MEMs. Variability in how lower trophic levels are included in MEMs trophic levels modelled in ESMs remains complex. Biogeochemical can result in a range of directional changes even under the same Nature Climate Change | VOL 11 | November 2021 | 973–981 | www.nature.com/natureclimatechange 979 Articles NaTuRE ClIMaTE CHangE climate simulation experiment, highlighting the need for improve- exploration of a greater range of climate impacts. With increasing ment in the coupling of MEMs with biogeochemical variable maturity of MEMs and communication between climate simulation outputs21. However, the fact that all MEMs projected declining and impact communities, such as through the CMIP6 Vulnerability, biomass under high emissions despite these differences around Impacts, Adaptation and Climate Services Advisory Board16, these lower-trophic-level forcing suggests a robustness to the results and opportunities should arise in the future. perhaps the particular importance of temperature effects21. In addition, and importantly, the MEM outputs explored While the ESMs exhibited generally consistent trends in NPP here focus on total animal biomass without the inclusion of fish- and export in response to warming24, disagreement in the regional ing impacts, which can act synergistically with climate change4. patterns of these responses and their global signature is an ongo- Consequently, projected changes in an exploited ocean may be ing issue25. The inclusion of strong temperature-dependent rem- larger. While restoring overfished stocks and limiting exploitation ineralization, for example, can enhance recycling and NPP despite to sustainable levels may help with climate change adaptation at the increasing stratification under ocean surface warming, yet this is regional and global levels32, our CMIP6 results suggest that there are treated unevenly across ESMs. Variations in light harvesting and larger challenges ahead for future fisheries potential than previously nutrient uptake parameters within broad uncertainty bounds can anticipated. Fisheries production potential has remained essentially shift latitudinal thresholds between nutrient- and light-limited flat for the past 40 years despite an increase in fishing effort, with regimes in biogeochemical models, with the latter likely to be posi- marine capture fisheries landings on the order 0.1 Gt yr–1 (ref. 33). tively impacted by increasing stratification under climate change If CMIP6 projections hold, wild-capture fishery contributions to and the former negatively26. Regional circulation changes, which global food security may be further challenged. often vary strongly between models, can shape regional NPP As per previous Fish-MIP studies8,12, our results focus on total trends27. Efforts to constrain factors controlling the NPP response (potential) ecosystem biomass, rather than the ‘edible’ biomass to climate change28 or leverage emergent constraints to reduce available for fisheries, in part because we are interested in the NPP projection uncertainty23 are ongoing but did not result in con- overall ecosystem response to environmental change. In addition, vergence for CMIP6. We also caution that none of the MEMs (or given the heterogeneous nature of the MEMs, only a proportion of ESMs for plankton) yet includes the potential for adaptation or evo- which are species-based, total ecosystem biomass is a variable that lution of individual taxa and that there remains the lack of a full is consistently comparable across models. The ensemble outputs bi-directional coupling between higher- and lower-trophic-level that we provide here complement organismal34 and spatial35 studies (biogeochemical) processes in the MEMs. Furthermore, many spe- of climate impacts and vulnerabilities, both in terms of using such cies interactions are not well captured by global-scale MEMs, and studies to further inform the development of individual MEMs and the potential for nonlinear tipping points that cause rapid ecological to provide ecosystem-level projections of trends and uncertainties deterioration remains unclear21. due to changes in both temperature and productivity that provide Projected mean changes in animal biomass had a smaller spread additional context to the more finely resolved studies of thermal across CMIP6-forced Fish-MIP models under both scenarios niches and life-history responses. For robust projections of edible (Fig. 3). However, combining opposing NPP signals from ESMs fish biomass, we would also need scenarios of future fleet behav- that have also effectively become more similar in terms of warm- iour, economics and changes in target fishery species that are not ing response could create an illusion of model improvement and yet available (and not included in Shared Socioeconomic Pathways reduced uncertainty, although, as described, not all MEMs used (SSPs)). Projected decreases of global animal biomass do not nec- NPP. The reduced spread of MEM projections is evident only when essarily imply that global fisheries catches would be reduced in ESMs are averaged; within each ESM the spread of MEM biomass proportion; changes in animal production, where explored in projections is similar under CMIP5 and CMIP6 (Fig. 4). Individual individual MEMs, suggest a similar climate response to biomass, improvements in two MEMs between CMIP5 and CMIP6 could although lower in magnitude36,37. We recognize that the inclusion of also have impacted results. For example, EcoOcean has recently fishing impacts, which can act synergistically with climate change4, been updated by improving its representation of specific spe- remains an important concern, and Fish-MIP is in the process of cies contributions to ecosystem dynamics and the response of the developing scenarios to enable such comparisons. marine food web to different environmental drivers29. However, the Our CMIP6 projections of twenty-first-century climate change majority of MEMs were unchanged (Table 1), so this is likely to have impacts show steeper global biomass declines and thus greater had a limited impact relative to the changes in ESM forcings. climate risks for marine ecosystems than their CMIP5 counter- While here we explore uncertainties in future animal biomass parts forced by the same two ESMs, and emphasize the benefits of through generating an ensemble of impact models (MEMs) forced strong mitigation. Marked shifts in directional differences for many by standardized inputs from two contrasting ESMs and low and high regions of the global ocean, probably driven by differences in ESM carbon emission scenarios, we recognize that our treatment of uncer- forcing, and in particular NPP, highlight the large uncertainties that tainty associated with internal climate variability is limited. However, still exist, suggesting that the readiness of ESM-forced global-scale this is generally a smaller source of uncertainty at the multi-decadal MEMs to support country-level adaptation policies is still nascent, to century time scales explored here than model and scenario although these capabilities may be more advanced for regional mod- uncertainty for most fisheries-relevant biogeochemical drivers30,31. els38. There remains an urgent need for model refinement to tackle Nonetheless, future efforts to fully characterize these uncertainties uncertainty at all levels, including both climate and marine ecosys- associated with internal climate variability will be important for inte- tem projections. Only once these uncertainties have been addressed, grating an additional source of uncertainty into the ensembles, as will so that climate-to-ecosystem modelling is improved, will the projec- the addition of further ESMs, although here our experimental design tions of climate impacts on marine organisms and fisheries be more aimed to use the same two ESMs as in the CMIP5 experiment11,12. robust and thus more strategically useful. A more ambitious model Sensitivity to climate responses in MEMs could also be examined evaluation of the whole ocean including ecological-to-human cou- across a range of past modelled and observed biogeochemical and/ pled systems is required to deliver the rigorous projections urgently or ecological variables that might constrain responses to climate needed to advance climate adaptation and mitigation. change, so-called emergent constraints23, as a means to improve both suites of ensembles. In addition, the greater integration of Online content other changing biogeochemical components (such as oxygen) to Any methods, additional references, Nature Research reporting more MEMs (Supplementary Table 1) would enable an ensemble summaries, source data, extended data, supplementary information, 980 Nature Climate Change | VOL 11 | November 2021 | 973–981 | www.nature.com/natureclimatechange NaTuRE ClIMaTE CHangE Articles acknowledgements, peer review information; details of author 28. Laufkötter, C., John, J. G., Stock, C. A. & Dunne, J. P. Temperature and oxygen dependence of the remineralization of organic matter. Glob. contributions and competing interests; and statements of data Biogeochem. Cycles 31, 1038–1050 (2017). and code availability are available at https://doi.org/10.1038/ 29. Coll, M. et al. Advancing global ecological modeling capabilities to simulate s41558-021-01173-9. future trajectories of change in marine ecosystems. Front. Mar. Sci. 7, 741 (2020). Received: 2 June 2021; Accepted: 1 September 2021; 30. Hawkins, E. & Sutton, R. The potential to narrow uncertainty in regional Published online: 21 October 2021 climate predictions. Bull. Am. Meteorol. Soc. 90, 1095–1107 (2009). 31. Frölicher, T. L., Rodgers, K. B., Stock, C. A. & Cheung, W. W. L. Sources of uncertainties in 21st century projections of potential ocean ecosystem References stressors. Glob. Biogeochem. Cycles 30, 1224–1243 (2016). 1. IPCC Special Report on the Ocean and Cryosphere in a Changing Climate 32. Gaines, S. D. et al. Improved fisheries management could offset many (eds Pörtner, H.-O. et al.) (IPCC, 2019). negative effects of climate change. Sci. Adv. 4, eaao1378 (2018). 2. Doney, S. C. et al. Climate change impacts on marine ecosystems. Ann. Rev. 33. The State of World Fisheries and Aquaculture 2020 (FAO, 2020). Mar. Sci. 4, 11–37 (2012). 34. Dahlke, F. T., Wohlrab, S., Butzin, M. & Pörtner, H.-O. Thermal bottlenecks 3. Bindoff, N. L. et al. in Special Report on the Ocean and Cryosphere in a in the life cycle define climate vulnerability of fish. Science 369, 65–70 (2020). Changing Climate (eds Pörtner, H.-O. et al.) Ch. 5 (IPCC, 2019). 35. Stuart-Smith, R. D., Edgar, G. J. & Bates, A. E. Thermal limits to the 4. Griffith, G. P., Fulton, E. A., Gorton, R. & Richardson, A. J. Predicting geographic distributions of shallow-water marine species. Nat. Ecol. Evol. 1, interactions among fishing, ocean warming, and ocean acidification in a marine 1846–1852 (2017). system with whole-ecosystem models. Conserv. Biol. 26, 1145–1152 (2012). 36. Carozza, D. A., Bianchi, D. & Galbraith, E. D. Metabolic impacts of climate 5. Fu, C. et al. Risky business: the combined effects of fishing and changes in change on marine ecosystems: implications for fish communities and primary productivity on fish communities. Ecol. Modell. 368, 265–276 (2018). fisheries. Glob. Ecol. Biogeogr. 28, 158–169 (2019). 6. Tittensor, D. P. et al. Integrating climate adaptation and biodiversity 37. du Pontavice, H., Gascuel, D., Reygondeau, G., Stock, C. & Cheung, W. W. L. conservation in the global ocean. Sci. Adv. https://doi.org/10.1126/sciadv. Climate-induced decrease in biomass flow in marine food webs may aay9969 (2019). severely affect predators and ecosystem production. Glob. Change Biol. 27, 7. IPBES: Summary for Policymakers. In Global Assessment Report on Biodiversity 2608–2622 (2021). and Ecosystem Services (eds Díaz, S. et al.) (IPBES Secretariat, 2019). 38. Piroddi, C. et al. Effects of nutrient management scenarios on marine food 8. Boyce, D. G., Lotze, H. K., Tittensor, D. P., Carozza, D. A. & Worm, B. Future webs: a pan-European assessment in support of the marine strategy ocean biomass losses may widen socioeconomic equity gaps. Nat. Commun. framework directive. Front. Mar. Sci. 8, 179 (2021). 11, 2235 (2020). 39. Maury, O. An overview of APECOSM, a spatialized mass balanced ‘Apex 9. Payne, M. R. et al. Uncertainties in projecting climate-change impacts in Predators ECOSystem Model’ to study physiologically structured tuna marine ecosystems. ICES J. Mar. Sci. 73, 1272–1282 (2016). population dynamics in their ecosystem. Prog. Oceanogr. 84, 113–117 (2010). 10. Eyring, V. et al. Overview of the Coupled Model Intercomparison Project 40. Maury, O. & Poggiale, J. C. From individuals to populations to communities: Phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev. a dynamic energy budget model of marine ecosystem size-spectrum including 9, 1937–1958 (2016). life history diversity. J. Theor. Biol. 324, 52–71 (2013). 11. Tittensor, D. P. et al. A protocol for the intercomparison of marine fishery and 41. Carozza, D. A., Bianchi, D. & Galbraith, E. D. The ecological module of ecosystem models: Fish-MIP v1.0. Geosci. Model Dev. 11, 1421–1442 (2018). BOATS-1.0: a bioenergetically-constrained model of marine upper trophic 12. Lotze, H. K. et al. Global ensemble projections reveal trophic amplification of levels suitable for studies of fisheries and ocean biogeochemistry. Geosci. ocean biomass declines with climate change. Proc. Natl Acad. Sci. USA 116, Model Dev. 9, 1545–1565 (2016). 12907–12912 (2019). 42. Carozza, D. A. et al. Formulation, general features and global calibration of a 13. Bryndum-Buchholz, A. et al. Twenty-first-century climate change impacts on bioenergetically-constrained fishery model. PLoS ONE 12, e0169763 (2017). marine animal biomass and ecosystem structure across ocean basins. 43. Cheung, W. W. L. et al. Building confidence in projections of the responses of Glob. Change Biol. 25, 459–472 (2019). living marine resources to climate change. ICES J. Mar. Sci. https://doi. 14. Bryndum-Buchholz, A. et al. Differing marine animal biomass shifts under org/10.1093/icesjms/fsv250 (2016). 21st century climate change between Canada’s three oceans. Facets 5, 44. Cheung, W. W. L., Dunne, J., Sarmiento, J. L. & Pauly, D. Integrating 105–122 (2020). ecophysiology and plankton dynamics into projected maximum fisheries 15. Bryndum-Buchholz, A. et al. Climate-change impacts and fisheries catch potential under climate change in the Northeast Atlantic. ICES J. Mar. management challenges in the North Atlantic Ocean. Mar. Ecol. Prog. Ser. Sci. 68, 1008–1018 (2011). 648, 1–17 (2020). 45. Blanchard, J. L. et al. Potential consequences of climate change for primary 16. Ruane, A. C. et al. The vulnerability, impacts, adaptation and climate services production and fish production in large marine ecosystems. Phil. Trans. R. advisory board (VIACS AB v1.0) contribution to CMIP6. Geosci. Model Dev. Soc. B 367, 2979–2989 (2012). 9, 3493–3515 (2016). 46. Christensen, V. et al. The global ocean is an ecosystem: simulating marine life 17. Kwiatkowski, L. et al. Twenty-first century ocean warming, acidification, and fisheries. Glob. Ecol. Biogeogr. 24, 507–517 (2015). deoxygenation, and upper-ocean nutrient and primary production decline 47. Gascuel, D., Guénette, S. & Pauly, D. The trophic-level-based ecosystem from CMIP6 model projections. Biogeosciences 17, 3439–3470 (2020). modelling approach: theoretical overview and practical uses. ICES J. Mar. Sci. 18. Séférian, R. et al. Tracking improvement in simulated marine 68, 1403–1416 (2011). biogeochemistry between CMIP5 and CMIP6. Curr. Clim. Change Rep. 6, 48. Petrik, C. M., Stock, C. A., Andersen, K. H., van Denderen, P. D. & Watson, 95–119 (2020). J. R. Bottom-up drivers of global patterns of demersal, forage, and pelagic 19. Meehl, G. A. et al. Context for interpreting equilibrium climate sensitivity fishes. Prog. Oceanogr. 176, 102124 (2019). and transient climate response from the CMIP6 Earth system models. 49. Jennings, S. & Collingridge, K. Predicting consumer biomass, size-structure, Sci. Adv. 6, eaba1981 (2020). production, catch potential, responses to fishing and associated uncertainties 20. Tebaldi, C. et al. Climate model projections from the Scenario Model in the world’s marine ecosystems. PLoS ONE 10, e0133794 (2015). Intercomparison Project (ScenarioMIP) of CMIP6. Earth Syst. Dyn. 12, 50. Heneghan, R. F. et al. A functional size-spectrum model of the global 253–293 (2021). marine ecosystem that resolves zooplankton composition. Ecol. Modell. 435, 21. Heneghan, R. F. et al. Disentangling diverse responses to climate change 109265 (2020). among global marine ecosystem models. Prog. Oceanogr. 198, 102659 (2021). 22. Zelinka, M. D. et al. Causes of higher climate sensitivity in CMIP6 models. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in Geophys. Res. Lett. 47, e2019GL085782 (2020). published maps and institutional affiliations. 23. Kwiatkowski, L. et al. Emergent constraints on projections of declining primary Open Access This article is licensed under a Creative Commons production in the tropical oceans. Nat. Clim. Change 7, 355–358 (2017). Attribution 4.0 International License, which permits use, sharing, adap- 24. Cabré, A., Marinov, I. & Leung, S. Consistent global responses of marine tation, distribution and reproduction in any medium or format, as long ecosystems to future climate change across the IPCC AR5 Earth system as you give appropriate credit to the original author(s) and the source, provide a link to models. Clim. Dyn. 45, 1253–1280 (2015). the Creative Commons license, and indicate if changes were made. The images or other 25. Laufkötter, C. et al. Drivers and uncertainties of future global marine third party material in this article are included in the article’s Creative Commons license, primary production in marine ecosystem models. Biogeosciences 12, unless indicated otherwise in a credit line to the material. If material is not included in 6955–6984 (2015). the article’s Creative Commons license and your intended use is not permitted by statu- 26. Doney, S. C. Plankton in a warmer world. Nature 444, 695–696 (2006). tory regulation or exceeds the permitted use, you will need to obtain permission directly 27. Rykaczewski, R. R. & Dunne, J. P. Enhanced nutrient supply to the California from the copyright holder. To view a copy of this license, visit http://creativecommons. Current Ecosystem with global warming and increased stratification in an org/licenses/by/4.0/. Earth system model. Geophys. Res. Lett. 37, L21606 (2010). © The Author(s) 2021 Nature Climate Change | VOL 11 | November 2021 | 973–981 | www.nature.com/natureclimatechange 981 Articles NaTuRE ClIMaTE CHangE Methods 53. Dufresne, J.-L. et al. Climate change projections using the IPSL-CM5 Earth The full ensemble approach used for Fish-MIP model intercomparisons, MEM system model: from CMIP3 to CMIP5. Clim. Dyn. 40, 2123–2165 (2013). architectures and the selection of ESM simulation outputs used as forcing variables 54. Dunne, J. P. et al. The GFDL Earth System Model Version 4.1 (GFDL-ESM is previously described11, but briefly, monthly ocean physical and biogeochemical 4.1): overall coupled model description and simulation characteristics. outputs from ESMs run under prescribed scenarios are interpolated in space and J. Adv. Model. Earth Syst. 12, e2019MS002015 (2020). time to a regular monthly 1° grid. These variables are then used under a common 55. Krasting, J. P. et al. NOAA-GFDL GFDL-ESM4 Model Output Prepared for simulation protocol to force individual MEMs, with each modelling group using MIP6 CMIP Historical Version 20190726 (Earth System Grid Federation, all relevant variables for their model (Supplementary Table 1). MEM outputs are 2018); https://doi.org/10.22033/ESGF/CMIP6.8597 then archived in a standardized 1° grid format for a common range of ecosystem 56. John, J. G. et al. NOAA-GFDL GFDL-ESM4 Model Output Prepared for variables. CMIP6 ScenarioMIP ssp585 Version 20180701 (Earth System Grid Federation, For the Fish-MIP simulation round here, to enable the best possible 2018); https://doi.org/10.22033/ESGF/CMIP6.8706 comparison between CMIP5 and CMIP6, we utilized the same two ESMs: 57. Boucher, O. et al. Presentation and evaluation of the IPSL-CM6A-LR climate GFDL-ESM2M51,52 and IPSL-CM5A-LR53 for CMIP5 and the equivalent new model. J. Adv. Model. Earth Syst. 12, e2019MS002010 (2020). generations for CMIP6, GFDL-ESM4.154–56 and IPSL-CM6A-LR57–59. These 58. Boucher, O. et al. IPSL IPSL-CM6A-LR Model Output Prepared for CMIP6 were originally chosen on the basis of key criteria, including that they spanned CMIP Version 20180727 (Earth System Grid Federation, 2018); https://doi. a substantial fraction of the range of CMIP5 ESM projections in relevant org/10.22033/ESGF/CMIP6.1534 oceanographic variables11. We compared projections under the strong-mitigation 59. Boucher, O. et al. IPSL IPSL-CM6A-LR Model Output Prepared for CMIP6 RCP 2.6 (CMIP6 SSP1–2.6) and the high-emissions RCP 8.5 (CMIP6 SSP5–8.5) CMIP Historical Version 20180103 (Earth System Grid Federation, 2018); scenarios. For the CMIP5-forced Fish-MIP model runs, the historical simulations https://doi.org/10.22033/ESGF/CMIP6.5195 spanned 1970–2005, and the RCP scenarios spanned 2006–2099; for CMIP6, the historical simulations spanned 1970–2014, and the SSP scenarios spanned Acknowledgements 2015–2099. To enable a standardized comparison, a historical baseline period We thank ISIMIP for assistance with processing ESM inputs and outputs and the of 1990–1999 was used, with changes in MEM outputs evaluated relative to this ESM community for enabling these impact analyses. This work was supported by the period. To further enable direct comparison and isolation of impacts, climate Jarislowsky Foundation (D.P.T.), the Natural Sciences and Engineering Research Council change was the only stressor imposed on marine ecosystems, without fishing of Canada Discovery Grant programme (D.P.T., H.K.L., T.D.E., W.W.L.C., J.P.-A. and or other anthropogenic pressures superimposed. The primary output variable V.C.); Australian Research Council (ARC) Discovery Projects DP170104240 (J.L.B. and examined was total global marine animal biomass, with the specific range of C.N.), DP190102293 (J.L.B., C.N., A.J.R., J.D.E. and D.P.T.) and DP150102656 (J.D.E.); marine animals represented (for example, species or functional groups) varying the European Union’s Horizon 2020 research and innovation programme under grant among MEMs11 but generally including major taxonomic groups or major fish taxa agreements 817578 (TRIATLAS) (M.C., J.S., L.S., O.M., L.B., Y.-J.S., N.B. and J.R.), (Table 1). Given this, we examined trajectories of relative animal biomass change 869300 (FutureMARES) (J.A.F.-S.,Y.-J.S. and M.C.) and 862428 (MISSION ATLANTIC for each model over the twenty-first century rather than absolute values. For details (J.A.F.-S, Y.-J.S. and M.C.); the Spanish National Project ProOceans (PID2020- on model calibration and validation, see ref. 12; for key MEM reference papers, 118097RB-I00) (M.C. and J.S.); the Open Philanthropy Project (C.S.H.); the United see Table 1. The complete updated Fish-MIP protocol for CMIP6 forcings can be Kingdom Research and Innovation (UKRI) Global Challenges Research Fund (GCRF) found at https://bit.ly/3jhWH7c. One Ocean Hub (NE/S008950/1) (K.O.-C. and L.S.); the Simons Foundation (nos. 54993, In addition to the refined CMIP6 ESMs, two of the original six global 645921) (G.L.B.); the Belmont Forum and BiodivERsA under the BiodivScen ERA-Net Fish-MIP MEMs have also undergone improvements, reflecting the further COFUND programme (SOMBEE project, ANR-18-EBI4-0003-01) (Y.-J.S. and N.B.); the development of parameterizations, additional processes and underlying MEOPAR Postdoctoral Fellowship Award 2020–2021 and the Ocean Frontier Institute hypotheses for the structure and dynamics of marine ecosystems (Table 1). (Module G) (A.B.-B.); the French ANR project CIGOEF (grant ANR-17-CE32-0008-01) Furthermore, three additional MEMs were added to the ensemble (EcoTroph37,47, (O.M., L.B. and J.R.); the California Ocean Protection Council Grant C0100400, the FEISTY48 and ZooMSS50; Table 1). EcoTroph is a trophic-level-based model Alfred P. Sloan Foundation and the Extreme Science and Engineering Discovery that implicitly includes all marine species of vertebrates and invertebrates; Environment (XSEDE) allocation TG-OCE170017 (D.B. and J.G.); the National FEISTY is a composite model that includes both pelagic and demersal fish species Oceanographic and Atmospheric Association (NA20OAR4310441, NA20OAR4310442) as well as benthic invertebrates; ZooMSS is a composite model that includes (C.M.P.). M.C. acknowledges the Severo Ochoa Centre of Excellence accreditation fish but also focuses on resolving zooplankton taxa. In addition to examining (CEX2019-000928-S) to the Institute of Marine Science (ICM-CSIC). ZooMSS was run projections across the entire expanded CMIP6 ensemble of nine MEMs (‘full using the computational cluster Katana, supported by Research Technology Services MEM ensemble’), we also performed a more direct standardized comparison at UNSW Sydney; O.M., N.B. and J.R. acknowledge the Pôle de Calcul et de Données by using outputs from the CMIP5-driven ensemble of six MEMs against results Marines (PCDM) for providing DATARMOR storage and computational resources; from only these six MEMs for CMIP6 forcings (‘comparable MEM ensemble’). A W.W.L.C. and J.P.-A. acknowledge advanced computing support from Compute Canada. large number of regional MEMs also contribute to Fish-MIP and warrant detailed investigation at the local scale, but they are not reported here for as straightforward comparison as possible. Author contributions D.P.T. developed the simulation protocol; D.P.T. and J.L.B. conceived the study; C.N. and Reporting Summary. Further information on research design is available in the D.P.T. developed the analysis code and C.N. analysed the results; C.N., R.F.H., N.B., D.B., Nature Research Reporting Summary linked to this article. M.C., J.D.E., D.G., J.G., O.M., J.P.-A., C.M.P., H.d.P., A.J.R., J.S. and J.L.B. ran the global models; C.S.H., R.F.H. and M.B. processed the data; D.P.T. and J.L.B. wrote the paper; C.N., C.S.H., R.F.H., N.B., D.B., L.B., A.B.-B., G.L.B., M.B., W.W.L.C., V.C., M.C., J.P.D., Data availability T.D.E., J.D.E., J.A.F.-S., E.A.F., E.D.G., D.G., J.G., J.G.J., J.S.L., H.K.L., O.M., K.O.-C., All standardized forcing variables from the ESMs are available at https://doi. J.P.-A., C.M.P., H.d.P., J.R., A.J.R., L.S., Y.-J.S., J.S. and C.A.S. contributed to and critically org/10.48364/ISIMIP.575744.1; all outputs from the MEMs are available via ISIMIP revised the manuscript. (https://www.isimip.org/gettingstarted/data-access/). Competing interests Code availability The authors declare no competing interests. All code used to analyse simulations is available at https://github.com/Fish-MIP/ CMIP5vsCMIP6. Additional information Extended data is available for this paper at https://doi.org/10.1038/s41558-021-01173-9. References Supplementary information The online version contains supplementary material 51. Dunne, J. P. et al. GFDL’s ESM2 global coupled climate–carbon Earth system available at https://doi.org/10.1038/s41558-021-01173-9. models. Part I: physical formulation and baseline simulation characteristics. Correspondence and requests for materials should be addressed to Derek P. Tittensor. J. Clim. 25, 6646–6665 (2012). 52. Dunne, J. P. et al. Carbon Earth system models. Part II: carbon system Peer review information Nature Climate Change thanks Sheila Heymans and the other, formulation and baseline simulation characteristics. J. Clim. 26, anonymous, reviewer(s) for their contribution to the peer review of this work. 2247–2267 (2013). Reprints and permissions information is available at www.nature.com/reprints. Nature Climate Change | www.nature.com/natureclimatechange NaTuRE ClIMaTE CHangE Articles Extended Data Fig. 1 | Spatial change in oceanographic properties under CMIP6 SSP5-8.5. Change in sea -surface temperature (a-c), net primary production (d-f), phytoplankton carbon (g-i) and zooplankton carbon (j-l) between 1990–1999 and 2090–2099. Left column shows the mean change for IPSL-CM6A-LR. Middle column shows the mean change for GFDL-ESM4. Rightmost column shows the mean change averaged across GFDL and IPSL. Nature Climate Change | www.nature.com/natureclimatechange Articles NaTuRE ClIMaTE CHangE Extended Data Fig. 2 | Spatial change in oceanographic properties under CMIP5 RCP8.5. Change in sea surface temperature (a-c), net primary production (d-f), phytoplankton carbon (g-i) and zooplankton carbon (j-l) between 1990–1999 and 2090–2099. Left column shows the mean change for IPSL-CM5A-LR. Middle column shows the mean change for GFDL-ESM2M. Rightmost column shows the mean change averaged across GFDL and IPSL. Nature Climate Change | www.nature.com/natureclimatechange NaTuRE ClIMaTE CHangE Articles Extended Data Fig. 3 | Multi-model change in marine animal biomass across comparable set of six marine ecosystem models (MEMs) run using both CMIP5 and CMIP6 forcings. CMIP5 (a) and CMIP6 (b) results averaged across GFDL and IPSL under strong-mitigation (blue) and high-emissions (red) scenarios. Solid lines indicate ensemble model means; shaded areas indicate + /- inter-model standard deviation (n = 10; two MEMs only used IPSL). All values are relative to the standardized reference period of 1990–1999 (vertical grey shaded area). Vertical grey shaded area indicates reference decade and vertical grey line indicates first year of projection (subsequent to historical period). For the suite of MEMs considered, see Fig. 4. Coloured dots indicate CMIP5 values in years in which the ensemble mean values were higher for CMIP5 than for CMIP6. Nature Climate Change | www.nature.com/natureclimatechange Articles NaTuRE ClIMaTE CHangE Extended Data Fig. 4 | Projected global change in marine animal biomass across comparable set of six marine ecosystem models (MEMs) run using both CMIP5 and CMIP6 forcings under the high-emissions scenario. MEM outputs using CMIP5 (a & c) and CMIP6 (b & b) forcings (n = 10; two MEMs only used IPSL). All values are relative to the standardized reference period of 1990–1999. Vertical grey shaded area indicates reference decade and vertical grey line indicates first year of projection (subsequent to historical period). Nature Climate Change | www.nature.com/natureclimatechange NaTuRE ClIMaTE CHangE Articles Extended Data Fig. 5 | Ensemble mean change in marine animal biomass across comparable set of six marine ecosystem models (MEMs) that used both CMIP5 and CMIP6 forcings under the high-emissions scenario. MEM outputs using CMIP5 (a) and CMIP6 (b) forcings from GFDL and IPSL (n = 10; two MEMs only used IPSL). Maps represent mean percentage change between 1990–1999 and 2090–2999 under a) CMIP5 and b) CMIP6; c) difference in percentage change between CMIPs; d) difference in percentage change between CMIPs for grid cells that showed the same direction of change; e) difference in percentage change between CMIPs for grid cells that changed from a decrease in CMIP5 to an increase in CMIP6 and f) vice versa. Nature Climate Change | www.nature.com/natureclimatechange Articles NaTuRE ClIMaTE CHangE Extended Data Fig. 6 | Ensemble mean change in marine animal biomass across comparable set of six marine ecosystem models (MEMs) that used both CMIP5 and CMIP6 forcings under the high-mitigation scenario. MEM outputs using CMIP5 (a) and CMIP6 (b) forcings from GFDL and IPSL (n = 10; two MEMs only used IPSL). Maps represent mean percentage change between 1990–1999 and 2090–2999 under a) CMIP5 and b) CMIP6; c) difference in percentage change between CMIPs; d) difference in percentage change between CMIPs for grid cells that showed the same direction of change; e) difference in percentage change between CMIPs for grid cells that changed from a decrease in CMIP5 to an increase in CMIP6 and f) vice versa. Nature Climate Change | www.nature.com/natureclimatechange NaTuRE ClIMaTE CHangE Articles Extended Data Fig. 7 | Ensemble model results from the full set of global marine ecosystem models under the high-emissions scenario. Total consumer biomass change (a, b), inter-model standard deviation (c, d), and model agreement (e, f) where 100% represents all models indicating the same direction of change, and 50% indicates half the models indicating one direction of change and half indicating the opposite. Nature Climate Change | www.nature.com/natureclimatechange Articles NaTuRE ClIMaTE CHangE Extended Data Fig. 8 | Ensemble model results from the comparable set of global marine ecosystem models that ran both CMIP5 and CMIP6 simulations under the high-emissions scenario. Total consumer biomass change (a, b), inter-model standard deviation (c, d), and model agreement (e, f) where 100% represents all models indicating the same direction of change, and 50% indicates half the models indicating one direction of change and half indicating the opposite. Nature Climate Change | www.nature.com/natureclimatechange NaTuRE ClIMaTE CHangE Articles Extended Data Fig. 9 | Ensemble model results from the full set of global marine ecosystem models under the strong-mitigation scenario. Total consumer biomass change (a, b), inter-model standard deviation (c, d), and model agreement (e, f) where 100% represents all models indicating the same direction of change, and 50% indicates half the models indicating one direction of change and half indicating the opposite. Nature Climate Change | www.nature.com/natureclimatechange Articles NaTuRE ClIMaTE CHangE Extended Data Fig. 10 | Ensemble model results from the comparable set of global marine ecosystem models that ran both CMIP5 and CMIP6 simulations under the strong-mitigation scenario. Total consumer biomass change (a, b), inter-model standard deviation (c, d), and model agreement (e, f) where 100% represents all models indicating the same direction of change, and 50% indicates half the models indicating one direction of change and half indicating the opposite. Nature Climate Change | www.nature.com/natureclimatechange nature portfolio | reporting summary Corresponding author(s): Derek P. Tittensor Last updated by author(s): Aug 25, 2021 Reporting Summary Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist. Statistics For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section. n/a Confirmed The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one- or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section. A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted Give P values as exact values whenever suitable. For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated Our web collection on statistics for biologists contains articles on many of the points above. Software and code Policy information about availability of computer code Data collection Data were simulation outputs from individual marine ecosystem models, all cited in the main manuscript. Data analysis R version 4.0.3 was used to conduct the analysis of this data. For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information. Data Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: - Accession codes, unique identifiers, or web links for publicly available datasets - A description of any restrictions on data availability March 2021 - For clinical datasets or third party data, please ensure that the statement adheres to our policy All standardized forcing variables from the ESMs are available at doi: 10.48364/ISIMIP.575744.1; all outputs from the MEMs are available via ISIMIP (see: https:// www.isimip.org/gettingstarted/data-access/). 1 nature portfolio | reporting summary Field-specific reporting Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection. Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf Ecological, evolutionary & environmental sciences study design All studies must disclose on these points even when the disclosure is negative. Study description This study evaluates simulation outputs from marine ecosystem models forced by standardized earth-system models outputs from the CMIP5 and CMIP6 projects. Research sample The research sample is a set of marine ecosystem models from the FishMIP project (www.fishmip.org). Sampling strategy N/A Data collection ESM data were provided by ISIMIP; marine ecosystem model outputs were provided by individual modellers and coordinated by FishMIP. Timing and spatial scale Simulation outputs of the period 1970-2099 were analyzed. The spatial scale was global. Data exclusions No data were excluded. Reproducibility Models are reproducible in the sense that they can be re-run and outputs re-analyzed; all outputs are freely available. Randomization This is not relevant since we were not testing covariates. Blinding Blinding was not relevant to this study since we were not evaluating individual models versus one another but the whole set as an ensemble. Did the study involve field work? Yes No Reporting for specific materials, systems and methods We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material, system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response. Materials & experimental systems Methods n/a Involved in the study n/a Involved in the study Antibodies ChIP-seq Eukaryotic cell lines Flow cytometry Palaeontology and archaeology MRI-based neuroimaging Animals and other organisms Human research participants Clinical data Dual use research of concern March 2021 2