Global ensemble projections reveal trophic amplification of ocean biomass declines with climate change Heike K. Lotzea,1, Derek P. Tittensora,b, Andrea Bryndum-Buchholza, Tyler D. Eddya,c, William W. L. Cheungc, Eric D. Galbraithd,e, Manuel Barangef, Nicolas Barrierg, Daniele Bianchih, Julia L. Blanchardi,j, Laurent Boppk, Matthias Büchnerl, Catherine M. Bulmanm, David A. Carozzan, Villy Christenseno, Marta Collg,p, John P. Dunneq, Elizabeth A. Fultonj,m, Simon Jenningsr,s,t, Miranda C. Jonesc, Steve Mackinsonu, Olivier Mauryg,v, Susa Niiranenw, Ricardo Oliveros-Ramosx, Tilla Royi,y, José A. Fernandesz,aa, Jacob Schewel, Yunne-Jai Shing,bb, Tiago A. M. Silvar, Jeroen Steenbeekp, Charles A. Stockq, Philippe Verleycc, Jan Volkholzl, Nicola D. Walkerr, and Boris Worma a Department of Biology, Dalhousie University, Halifax, NS B3H 4R2, Canada; bScience, UN Environment World Conservation Monitoring Centre, CB3 0DL Cambridge, United Kingdom; cNippon Foundation-Nereus Program, Institute for the Oceans and Fisheries, University of British Columbia, Vancouver, BC V6T 1Z4, Canada; d Institució Catalana de Recerca i Estudis Avançats, 08010 Barcelona, Spain; eInstitut de Ciència i Tecnologia Ambientals, Universitat Autònoma de Barcelona, 08193 Barcelona, Spain; fFisheries and Aquaculture Department, Food and Agriculture Organization of the United Nations, 00153 Rome, Italy; gMarine Biodiversity, Exploitation and Conservation (MARBEC) Research Unit, Institut de Recherche pour le Développement, University of Montpellier, Institut Français de Recherche pour l’Exploitation de la Mer, Centre National de la Recherche Scientifique, 34203 Sète cedex, France; hDepartment of Atmospheric and Oceanic Sciences, University of California, Los Angeles, CA 90095-1565; iInstitute for Marine and Antarctic Studies, University of Tasmania, Hobart, TAS 7001, Australia; jCentre for Marine Socioecology, University of Tasmania, Hobart, TAS 7004, Australia; kInstitut Pierre-Simon Laplace, Laboratoire de Météorologie Dynamique, Centre National de la Recherche Scientifique, Ecole Normale Supérieure, Paris Sciences and Lettres Research University, Sorbonne University, Ecole Polytechnique, 75231 Paris, France; l Potsdam Institute for Climate Impact Research, 14473 Potsdam, Germany; mCommonwealth Scientific and Industrial Research Organization, Marine and Atmospheric Research, Hobart, TAS 7001, Australia; nDepartment of Earth and Planetary Sciences, McGill University, Montreal, QE H3A 0E8, Canada; oInstitute for the Oceans and Fisheries, University of British Columbia, Vancouver, BC V6T 1Z4, Canada; pInstitute of Marine Science, Ecopath International Initiative, 08003 Barcelona, Spain; ECOLOGY q Geophysical Fluid Dynamics Laboratory, National Oceanic and Atmospheric Administration, Princeton University, Princeton, NJ 08540; rLowestoft Laboratory, Centre for Environment, Fisheries and Aquaculture Science, NR33 0HT Lowestoft, United Kingdom; sSchool of Environmental Sciences, University of East Anglia, NR4 7TJ Norwich, United Kingdom; tScience Committee, International Council for the Exploration of the Sea, 1553 København V, Denmark; uHeritage House, Scottish Pelagic Fishermen’s Association, AB43 9BP Fraserburgh, United Kingdom; vDepartment of Oceanography, Marine Research Institute, University of Cape Town, 7701 Rondebosch, South Africa; wStockholm Resilience Centre, Stockholm University, 11419 Stockholm, Sweden; xInstituto del Mar del Perú, 07021 Callao, Perú; yEcosystem, Climate and Ocean Analysis, 75019 Paris, France; zMarine Research Division, AZTI Tecnalia, 20110 Pasaia, Spain; aaPlymouth Marine Laboratory, PL13 DH Plymouth, United Kingdom; bbDepartment of Biological Sciences, Marine Research Institute, University of Cape Town, 7701 Rondebosch, South Africa; and ccAMAP Research Unit, Institut de Recherche pour le Développement, 34398 Montpellier, France Edited by James A. Estes, University of California, Santa Cruz, CA, and approved May 22, 2019 (received for review January 5, 2019) While the physical dimensions of climate change are now routinely assessed through multimodel intercomparisons, projected impacts on Significance the global ocean ecosystem generally rely on individual models with a specific set of assumptions. To address these single-model limitations, Climate change can affect the distribution and abundance of marine we present standardized ensemble projections from six global marine life, with consequences for goods and services provided to people. ecosystem models forced with two Earth system models and four emis- Because different models can lead to divergent conclusions about sion scenarios with and without fishing. We derive average biomass marine futures, we present an integrated global ocean assessment trends and associated uncertainties across the marine food web. of climate change impacts using an ensemble of multiple climate and Without fishing, mean global animal biomass decreased by 5% (±4% ecosystem models. It reveals that global marine animal biomass will decline under all emission scenarios, driven by increasing tempera- SD) under low emissions and 17% (±11% SD) under high emissions by ture and decreasing primary production. Notably, climate change 2100, with an average 5% decline for every 1 °C of warming. Projected impacts are amplified at higher food web levels compared with biomass declines were primarily driven by increasing temperature and phytoplankton. Our ensemble projections provide the most com- decreasing primary production, and were more pronounced at higher prehensive outlook on potential climate-driven ecological changes in trophic levels, a process known as trophic amplification. Fishing did not the global ocean to date and can inform adaptive management and substantially alter the effects of climate change. Considerable regional conservation of marine resources under climate change. variation featured strong biomass increases at high latitudes and de- creases at middle to low latitudes, with good model agreement on the Author contributions: H.K.L., D.P.T., T.D.E., W.W.L.C., E.D.G., M. Barange, J.L.B., L.B., V.C., direction of change but variable magnitude. Uncertainties due to vari- E.A.F., S.J., O.M., and J. Schewe designed research; H.K.L., D.P.T., A.B.-B., T.D.E., W.W.L.C., ations in marine ecosystem and Earth system models were similar. E.D.G., M. Barange, N.B., D.B., J.L.B., M. Büchner, C.M.B., D.A.C., V.C., M.C., E.A.F., S.J., M.C.J., S.M., O.M., S.N., R.O.-R., J.A.F., Y.-J.S., T.A.M.S., J. Steenbeek, P.V., N.D.W., and B.W. performed Ensemble projections performed well compared with empirical data, research; L.B., J.P.D., T.R., J. Schewe, C.A.S., and J.V. contributed new reagents/analytic tools; emphasizing the benefits of multimodel inference to project future H.K.L., D.P.T., A.B.-B., T.D.E., T.R., J.V., and B.W. analyzed data; and H.K.L., D.P.T., A.B.-B., outcomes. Our results indicate that global ocean animal biomass con- T.D.E., W.W.L.C., E.D.G., M. Barange, N.B., D.B., J.L.B., L.B., M. Büchner, C.M.B., D.A.C., V.C., M.C., J.P.D., E.A.F., S.J., M.C.J., S.M., O.M., S.N., R.O.-R., T.R., J.A.F., J. Schewe, Y.-J.S., T.A.M.S., sistently declines with climate change, and that these impacts are am- J. Steenbeek, C.A.S., P.V., J.V., N.D.W., and B.W. wrote the paper. plified at higher trophic levels. Next steps for model development The authors declare no conflict of interest. include dynamic scenarios of fishing, cumulative human impacts, and Downloaded at Universitat Autonoma de Barcelona (UAB) on April 14, 2021 This article is a PNAS Direct Submission. the effects of management measures on future ocean biomass trends. This open access article is distributed under Creative Commons Attribution-NonCommercial- NoDerivatives License 4.0 (CC BY-NC-ND). | climate change impacts marine food webs | global ecosystem modeling | | model intercomparison uncertainty Data deposition: All data reported in this paper are archived and publicly available at http://dataservices.gfz-potsdam.de/pik/showshort.php?id=escidoc:2956913. 1 To whom correspondence may be addressed. Email: hlotze@dal.ca. C limate change is altering the abundance and distribution of marine species (1–5), with consequences for ocean ecosystem structure and functioning, seafood supply, and marine management This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1900194116/-/DCSupplemental. Published online June 11, 2019. www.pnas.org/cgi/doi/10.1073/pnas.1900194116 PNAS | June 25, 2019 | vol. 116 | no. 26 | 12907–12912 and conservation (6–8). Quantifying future trends and uncertainties is Results and Discussion critical to inform ongoing global assessments (1), including the Ensemble Projections of Global Ocean Animal Biomass. Our ensemble Intergovernmental Panel for Climate Change (IPCC) and Inter- projections revealed consistent declines in global marine animal governmental Science-Policy Platform on Biodiversity and Eco- biomass from 1970 to 2100 across all emission scenarios (RCPs, Fig. system Services, and guide viable pathways toward achieving key 1A). Almost all scenarios and MEM−ESM combinations predicted policy objectives, such as the United Nations Sustainable Devel- decreasing animal biomass (SI Appendix, Table S3), although the opment Goals (SDGs). Various modeling approaches exist to magnitude of decline varied among models. This general trend can assess current and future impacts on marine ecosystems (8–12), be explained by warming causing increased ocean stratification, yet each individual model is necessarily an incomplete simplifi- which reduces nutrient availability in the upper ocean, leading to cation of the natural world, with different assumptions, structures, decreased primary production and lower energy supply for higher and processes (13). One approach to overcoming any single- trophic levels (14, 18), and changes in metabolic rates, among others model limitations is to force a suite of models with standardized (19). Without fishing, mean total biomass declines ranged from climate change scenarios and combine them into ensemble pro- 4.8% (±3.5% SD) under low emissions (RCP2.6) to 17.2% (±10.7% jections to estimate mean future trends and associated intermodel SD) under high emissions (RCP8.5) by 2090–2099 relative to 1990– spread (13). Such model intercomparison projects (MIPs) have 1999 (Fig. 1A). All four emission scenarios projected similar de- become a “gold standard” in climate science and have proven clines by 2030, the target year of many SDGs, and through to critical for enhancing credibility and understanding of the physical midcentury, after which they began to diverge. Projected mean and biochemical climate change projections (14) and associated biomass declines were similar for animals of >10 cm and >30 cm impacts on Earth’s terrestrial biosphere (15–17), yet can only now (Fig. 1C), albeit slightly lower and more variable for those of >30 cm be attempted for the global ocean ecosystem (13). (SI Appendix, Fig. S5 and Table S3). Thus, the consequences of Over the past decade, a number of global fisheries and marine different emission scenarios may not be distinguishable over the next ecosystem models (MEMs) have been developed (13). Some of two to three decades but differ markedly in the long term. these have been used individually to project future changes in species distribution, biomass, or potential fisheries catch (8–12), Climate Change Effects in a Fished and Unfished Ocean. Three MEMs but it remains unclear how consistent and comparable these results were also able to run simulations with fishing, including time-varying are, and thus how applicable for providing robust insight and ad- historical and constant future fishing pressure (SI Appendix, SI vice. The Fisheries and Marine Ecosystem Model Intercomparison Methods), which we used to compare projected climate change ef- Project (Fish-MIP; ref. 13) was created to bring these various fects (RCP8.5 vs. RCP2.6) with and without fishing. The magnitude models and modeling groups together to produce ensemble pro- and variability of the climate change effect were similar (Fig. 1D), jections under standardized climate change scenarios. suggesting that fishing, at least under current levels of intensity, may Here we assess projected changes in global marine animal not substantially alter the relative effect of climate change. The biomass over the 21st century through ensemble projections with six slightly weaker climate change effects with fishing (mean difference 2 published global MEMs from Fish-MIP, forced with standardized to 3%; Fig. 1D and SI Appendix, Fig. S6A) may be due to an indirect outputs from two contrasting Earth system models (ESMs) and effect: Warming enhances both growth and predation rates, yet four emission scenarios [Representative Concentration Pathways predation rates are reduced due to selective fishing of larger animals (RCPs)]. The MEMs range from size-structured [Bioeconomic and lower predator abundance (20, 21) which may indirectly en- Marine Trophic Size-spectrum (BOATS), Macroecological] and hance prey biomass and weaken the relative climate change effect trait-based [Dynamic Pelagic Benthic Model (DPBM), Apex (19). This is a relatively small effect, however, compared with the Predators EcoSystem Model (APECOSM)] to species distribution large direct effect of fishing itself, which resulted in 16 to 80% lower [Dynamic Bioclimate Envelope Model (DBEM)] and troph- biomass for animals of >10 cm and 48 to 92% for animals of >30 cm odynamic models (EcoOcean) (SI Appendix, Tables S1 and S2). compared with unfished conditions in 2100 under RCP2.6, and The ESMs span the range of available Coupled Model Intercom- slightly lower values under RCP8.5. We note that the absolute parison Project Phase 5 (CMIP5) projections, from low [Geophysical magnitude of the fishing effect is not directly comparable across Fluid Dynamics Laboratory Climate Model (GFDL-ESM2M)] to MEMs, due to inherent differences in how fishing pressure and high Institute Pierre Simon Laplace Climate Model (IPSL-CM5A- commercial versus noncommercial taxa are incorporated (SI Ap- LR) increases in sea surface temperature (SST) and associated pendix, SI Methods). We also caution that our future constant fishing changes in net primary production (NPP), while other drivers were scenario is simplistic and does not incorporate potential changes in more similar (ref. 14 and SI Appendix, Figs. S1 and S2). The RCPs effort, technology, management, and conservation (11, 21–23), which range from a low-emission strong mitigation scenario (RCP2.6) to a are likely to strongly affect future biomass trends. Nevertheless, a high-emission business-as-usual scenario (RCP8.5), with two inter- possible consistent climate change effect is an important consider- mediate scenarios (RCP4.5 and RCP6.0). All climate change sce- ation in the context of marine management and conservation. narios were run for historical (1970–2005) and future (2006–2100) periods without fishing to isolate the climate signal, and with fishing Variability among Model Projections. Although ensemble means to evaluate how climate responses differ in an ocean fished at revealed global biomass declines across all emission scenarios (Fig. current levels of intensity (13). The six MEMs generated stan- 1A), there was considerable variation among MEMs and ESMs dardized outputs of total animal biomass (except zooplankton) (Fig. 1B). The latter largely reflects differences in SST increases and biomass of animals of >10 cm and >30 cm. Since not all and NPP reductions among ESMs (SI Appendix, Fig. S1), in addi- MEMs could run the full set of scenarios, due to MEM or ESM tion to other physical and biochemical drivers (14), and reinforces limitations, we analyzed all available runs for each scenario, and previous work highlighting the importance of ESM and scenario performed sensitivity analyses on subsets, which revealed similar uncertainty in future projections of fish biomass or fisheries pro- Downloaded at Universitat Autonoma de Barcelona (UAB) on April 14, 2021 results (SI Appendix, Table S3). duction (24, 25). Interestingly, the variability in projected biomass The goals of this study were to examine the consistency of changes among ESMs was of similar magnitude to that among projections across MEMs over a range of climate change sce- MEMs (Fig. 1E), suggesting similar levels of uncertainty associated narios globally and regionally from 1970 to 2100. We also eval- with physical and biological models. Since the field of global eco- uated how ocean animal biomass changes correspond with those system modeling is relatively new compared with Earth system in the physical environment and the extent to which projected modeling (13, 14), one might have expected higher variability climate impacts on primary producers and zooplankton (18) are across MEMs. Other MIPs also found that uncertainties in both transmitted to higher food web levels. ESMs and climate impact models contribute to overall projection 12908 | www.pnas.org/cgi/doi/10.1073/pnas.1900194116 Lotze et al. A 10 B10 5 5 Change in biomass (%) 0 0 -5 -5 -10 -10 -15 -15 -20 -20 Fig. 1. Ensemble projections of global ocean animal -25 -25 biomass with climate change. (A) Multimodel mean -30 -30 change in total biomass without fishing for historical -35 -35 and four future emission scenarios (RCPs) relative to -40 1990–1999 with 1 SD (±SD, n = 10); thin vertical line -40 indicates target year for achieving SDGs. (B) Individ- 1970 1990 2010 2030 2050 2070 2090 1970 1990 2010 2030 2050 2070 2090 ual model projections for RCP8.5 without fishing C RCP2.6 RCP8.5 D unfished fished E Ecosystem models Earth-system models showing the spread across different ecosystem model and ESM combinations. (C) Projected biomass declines Variability in biomass change Change in biomass (%) for three size groups in 2090–2099 relative to 1990– (% standard deviation) Difference in biomass 1999 under low (RCP2.6) and high (RCP8.5) emissions change (%) without fishing (n = 10). (D) Projected climate change effect (RCP8.5 vs. RCP2.6) in 2090–2099 for three size groups with and without fishing (n = 6). (E) Variability in biomass projections (expressed as SD of % change) due to different ESMs and different MEMs under RCP2.6 and RCP8.5 without fishing for three size groups (n = 10). Box plots display the median (hori- ECOLOGY RCP2.6 RCP8.5 RCP2.6 RCP8.5 zontal line), mean (x), and interquartile range (boxes). uncertainty, with similar contributions from ESMs and global hy- temporal trends and correlations for the ensemble mean showed a drological models (15), yet much higher uncertainty in global crop better fit (R2 = 0.96) than individual models (0.80 ≤ R2 ≤ 0.94). and vegetation models than ESMs (16, 17). These analyses suggest that our ensemble projections reflect ob- The variability among MEMs can be attributed to differences served trends for assessed fish stocks, providing confidence in in fundamental structures, taxonomic groups, and ecological pro- our historical and future projections. cesses. Generally, some MEMs respond strongly to temperature changes affecting metabolic rates (BOATS, Macroecological), while Trophic Amplification of Marine Biomass Declines. Although we did others are strongly driven by NPP changes affecting trophic dy- not find major differences in biomass changes among our three size namics (EcoOcean), or a combination of temperature, NPP, and groups (Fig. 1C), the combined biomass of higher trophic levels from additional drivers (e.g., pH, oxygen, ice cover) affecting habitat our MEMs declined more strongly than that of lower trophic levels niches and species distribution (DBEM) or food web dynamics from ESMs across all RCP scenarios (Fig. 3). This trophic amplifi- (DPBM, APECOSM) (13, 26). The MEMs also differ in whether cation of biomass declines has been previously shown for phyto- species interactions (e.g., predation, competition), movement, or plankton and zooplankton across a range of ESMs (18), and our dispersal is included (SI Appendix, Table S1), which can alter their results suggest this effect may extend to higher food web levels. Such response and result in compensatory changes (27, 28). Some MEMs amplification of the climate signal from primary producers to higher also allow for adaptive responses of species or communities to trophic levels arises from multiple factors that vary among ESMs changing environments (e.g., size-structured and species distribution and MEMs, including changes in phytoplankton size composition, models), while others do not. Notably, the variability of projected lengthening of food chains, reduced trophic efficiencies, and higher changes among both MEMs and ESMs was higher under RCP8.5 metabolic costs with increased body size (18). In addition to larger than RCP2.6 (Fig. 1E), and the variability among MEMs was higher mean biomass declines, we also observed larger variability for higher in animals of >30 cm than other size groups, suggesting greater trophic levels, particularly at higher RCPs (Fig. 3). Mean NPP, projection uncertainty with stronger warming and for larger animals. phytoplankton, and zooplankton declines and variability from our two ESMs were comparable to those of other ESM ensembles (14, Empirical Validation. All MEMs included in this study have been 18). We caution that many MEMs only use NPP or phytoplankton individually tested across a range of physical and biogeochemical biomass as forcing variables directly influencing higher trophic levels variables, and outputs have been compared with empirical data (SI Appendix, Table S2), and cannot resolve all underlying food web across multiple temporal and spatial scales (SI Appendix, SI mechanisms. Nevertheless, the consistency of the response across Methods and Fig. S3). In addition, we compared our ensemble diverse ESMs and MEMs does suggest a general pattern of higher projections against biomass trends of scientifically assessed fish trophic levels being more likely to show larger biomass declines than stocks representing 25 to 33% of the global fisheries catch (29, 30). lower trophic levels. This raises concerns about wider impacts of First, we extracted results of a recent analysis (30) hindcasting the climate change on the structure, function, and stability of ocean effect of warming on the maximum sustainable biomass yield (MSY) ecosystems (31, 32), especially in combination with other human of 235 assessed fish populations globally in the absence of fishing stressors, such as fishing, that disproportionally affect higher trophic Downloaded at Universitat Autonoma de Barcelona (UAB) on April 14, 2021 (Fig. 2). Temporal trends matched those of our ensemble projec- levels, a process called trophic downgrading (33). tions without fishing (n = 10 MEM−ESM combinations), and cor- relations suggest a good fit for the ensemble mean (R2 = 0.44), Trends in Relation to Global Temperature Changes. Many policy which was higher than for most individual models (0.13 ≤ R2 ≤ 0.47), processes use the change in global air temperature since pre- reflecting the strength of the ensemble approach. We also compared industrial times as a reference for the effects of climate change (34, biomass projections with fishing (n = 6 MEM−ESMs) to average 35). For our ensemble projections, this revealed a consistent linear biomass relative to biomass at MSY (B/BMSY) across 331 assessed relationship with an average 5% drop in total animal biomass with and exploited fish stocks (ref. 29 and SI Appendix, Fig. S4). Again, every 1 °C of Earth surface warming in the absence of fishing Lotze et al. PNAS | June 25, 2019 | vol. 116 | no. 26 | 12909 A B Projected biomass (% relative to 1990s) 20 4 R2 = 0.44 Biomass (% relative to 1990s) 15 3 10 2 Fig. 2. Empirical validation of ensemble projections. 5 Shown is the relative biomass change (animals of >10 1 cm) for our multimodel mean without fishing (orange 0 line, ±95% CI, n = 10) compared with temperature- 0 dependent hindcasts of MSY for 235 assessed fish -5 stocks that are independent of fishing effects (gray -10 -1 lines, ±95% CI; ref. 30). Both projected and empirical Assessed stocks Ensemble model biomass trends are displayed as (A) time series and (B) -15 -2 yearly scatter plot with a linear regression fit. A similar 1970 1975 1980 1985 1990 1995 2000 2005 36 35 34 33 analysis of biomass trends with fishing is shown in SI Year Biomass of assessed stocks (MSY million tons) Appendix, Fig. S4. (Fig. 4). Similar declines were found for animals of >10 cm Conclusions and >30 cm (SI Appendix, Fig. S7). These relationships may be Our ensemble projections demonstrate that global ocean animal slightly less negative in a fished ocean, given the dampening effect biomass consistently declines with climate change, and that impacts described above, yet overall biomass reductions remain substantial are amplified at higher trophic levels. Our hindcasts support recent (i.e., 15% instead of 17% decline under highest warming; SI Ap- empirical work that shows ongoing climate impacts on fish biomass pendix, Fig. S6B). Individual ecosystems may well show more- (30, 37), and project elevated climate-driven declines in ocean complex responses to warming (3, 36), but these simple relation- ecosystems, with magnitudes dependent on emission pathways. ships represent well-founded approximations for a global average Amplification of biomass declines for higher trophic levels repre- response. Recent results hindcasting temperature effects on bio- sents a particular challenge for human society, including meeting mass of 235 assessed fish stocks found a 4.1% drop in MSY from the SDGs for food security (SDG2), livelihoods (SDG1), and well- 1930 to 2010, a period that saw an average 0.6 °C of warming (30). being (SDG3) for a growing human population while also sustaining This reconstructed rate of change is consistent with our 5% drop life below water (SDG14). Our ensemble projections indicate the in biomass per 1 °C of warming throughout the 21st century. largest decreases in animal biomass at middle to low latitudes, According to these results, limiting future warming to 1.5 °C to where many nations depend on seafood and fisheries, and where 2.0 °C above preindustrial levels would limit biomass declines to 4 marine biodiversity is already threatened by multiple human activ- to 6% by 2100, underscoring the potential impact of climate change mitigation according with the Paris Agreement (34, 35). ities (6, 38). In turn, the largest increases are projected at high latitudes, highlighting new opportunities for—and potential conflict Spatial Patterns of Biomass Change. Not all ocean regions respond over—resource use, but also an urgent need for protecting similarly to climate warming (Fig. 5 and SI Appendix, Fig. S8). Our sensitive species and rapidly changing ecosystems. Overall, our ensemble projections revealed strong increases in total animal results clearly highlight the benefits to be gained from climate biomass in polar regions and widespread declines in temperate to change mitigation, as all impacts were substantially reduced under tropical regions under RCP8.5 (Fig. 5B), with qualitatively similar but less pronounced patterns under RCP2.6 (Fig. 5A). The climate change effects were spatially similar with and without fishing (SI 5 Appendix, Fig. S9). However, our ensemble projections differed spatially from previous single-model results highlighted in the 0 IPCC’s Fifth Assessment Report (1): We found strong biomass declines (not increases) in many temperate to subtropical regions and increases (not declines) around Antarctica. The magnitude of -5 Relative change (%) regional changes also varied from other single-model results (8–11). Generally, warming waters and enhanced primary production are -10 expected to facilitate species expansions and biomass increases in polar regions, while tropical areas may experience pronounced -15 species losses as thermal thresholds are exceeded. In temperate regions, warming is expected to change species composition, and reduced primary production due to enhanced stratification will -20 result in biomass declines (3, 4, 30). Our ensemble projections showed high model agreement on the direction of change in many -25 ocean regions (75 to 100%; Fig. 5 E and F), providing confidence in our multimodel results that combine different ecosystem structures -30 Downloaded at Universitat Autonoma de Barcelona (UAB) on April 14, 2021 and processes. Many models also agreed relatively well on the RCP2.6 RCP4.5 RCP6.0 RCP8.5 magnitude of projected changes in temperate to tropical regions but showed considerable intermodel variability in many polar and Fig. 3. Trophic amplification of marine biomass declines. Shown is the coastal regions (Fig. 5 C and D), reflecting differences among magnitude of projected mean changes in 2090–2099 relative to 1990–1999 (percent, ±SD) for NPP, phytoplankton, and zooplankton biomass derived ESMs and MEMs (SI Appendix, Figs. S10 and S11). These results from ESMs (n = 2) and higher trophic levels from MEMs across four RCPs (n = again underscore the importance of model intercomparison in 10 for RCP2.6 and RCP8.5; n = 8 for RCP4.5 and RCP6.0). All ecosystem models identifying uncertainties and constraining expected outcomes of use either NPP or phytoplankton biomass as forcing variables, and some also ecosystem changes in the ocean. use zooplankton biomass. 12910 | www.pnas.org/cgi/doi/10.1073/pnas.1900194116 Lotze et al. 5 is to address these uncertainties and improve our ability to predict marine ecosystem responses to climate change at different tem- poral and spatial scales. Projections based on global models are 0 often less certain for coastal and polar regions but may be im- proved through regional downscaling to incorporate higher- Change in total biomass (%) resolution climate and ecosystem features (7, 39). The next round of CMIP6 projections with improved representation of biochemical -5 parameters may also advance future ensemble projections (13, 40). The expansion of global observational datasets provides further opportunities to better constrain and validate models. The in- -10 corporation of additional MEMs based on novel paradigms or reflecting alternative structures and processes may also be in- formative (13), along with regional ecosystem or fish stock models that more accurately capture processes at management-relevant -15 scales (21, 41, 42). Future MEMs could also further explore how species interactions and potential acclimatization or adaptation of marine organisms modify projected distribution and abundance. -20 Finally, a large component of future change will depend on the 0 +0.5 +1 +1.5 +2 +2.5 +3 +3.5 +4 +4.5 +5 trajectories of fisheries, aquaculture, and other human impacts on Change in surface air temperature (ºC) the ocean (6, 11, 29, 38). Incorporating standardized temporally Fig. 4. Projected biomass trends in relation to global air temperature and spatially resolved scenarios of human activities and alternative changes. Shown is the relationship of the change in marine animal biomass management approaches will improve our understanding of the to increasing global surface air temperature since preindustrial times future of marine animals and ocean ecosystems (13, 23), and (1860s). Each dot represents an annual ensemble mean (n = 10) relative to identify the points of greatest leverage for mitigating impacts. ECOLOGY 1990–1999 across historical and future emission scenarios (RCPs) in the ab- Improved dynamic and adaptive ecosystem-based management sence of fishing. Vertical lines frame expected changes between 1.5 °C and 2 °C of warming. may mitigate some future climate change impacts and maintain ecosystem health and service provision (6, 21, 22, 43). However, this can only happen if the international community, including a strong mitigation (RCP2.6) compared with the business-as-usual national and regional bodies, strengthens the required institu- (RCP8.5) scenario. tions and management approaches (6, 44). By providing estimates of global biomass changes and associ- Methods ated uncertainties, our ensemble projections represent the most Model Selection. The six global MEMs varied in their model structure, pro- comprehensive outlook on the future of marine animal biomass to cesses, representation of functional groups, size classes or commercial taxa, and date. Our results are robust in terms of the direction of change, yet how they incorporated ESM-derived climate forcings and fishing (SI Appendix, the substantial spread in the magnitude of projections illustrates SI Methods and Tables S1 and S2). Our ensemble thus includes a greater variety considerable uncertainty in both ESMs and MEMs. The challenge of ecosystems components and processes than any single model. RCP2.6 RCP8.5 A B Biomass change (%) 50 25 0 -25 -50 C D Variability (SD) 50 38 25 12 0 E F Downloaded at Universitat Autonoma de Barcelona (UAB) on April 14, 2021 Model Fig. 5. Spatial patterns of projected biomass changes. agreement (%) Shown are global ensemble projections at a 1 × 1 de- 100 gree resolution for (A, C, and E) RCP2.6 and (B, D, and 88 F) RCP8.5. (A and B) Multimodel mean change (per- 75 cent, n = 10) in total marine animal biomass in 2090– 2099 relative to 1990–1999 without fishing. (C and D) 62 Variability among different ecosystem model and ESM 50 combinations expressed as 1 SD. (E and F) Model agreement (percent) on the direction of change. Lotze et al. PNAS | June 25, 2019 | vol. 116 | no. 26 | 12911 Climate Change Scenarios. The two ESMs from CMIP5 (https://pcmdi.llnl.gov/? with higher trophic level biomass from MEMs across RCPs. Mean biomass cmip5/) provided the necessary physical and biochemical outputs to force our changes were also compared with global air temperature changes since MEMs, particularly monthly depth-resolved fields of phytoplankton and preindustrial times (1861–1870) from ESMs. Spatial patterns were mapped as zooplankton size groups (SI Appendix, SI Methods and Table S2). Average trends mean % biomass changes in 2090–2099 vs. 1990–1999, the SD of the mean to of GFDL-ESM2M and IPSL-CM5A-LR reflect the multimodel mean and range of a assess intermodel variability in the magnitude of change, and the % model broad set of CMIP5 models (14). The four RCPs represent standard IPCC in- agreement on the direction of change (14). We also mapped the climate formed emission scenarios (SI Appendix, SI Methods). change effect with and without fishing and the variability of results across ESMs and MEMs. For further details, see SI Appendix, SI Methods. Simulations. All MEMs ran simulations without fishing, and three MEMs ran simulations with fishing for historical (1970–2005) and future (2006–2100) Data Accessibility. All data reported in this paper are archived (45) and periods. Historical fishing reflected time-varying effort or mortality rates publicly available at http://dataservices.gfz-potsdam.de/pik/showshort.php? depending on MEM requirements, and future fishing was kept constant at 2005 levels (SI Appendix, SI Methods). Not all MEMs performed all simulations, id=escidoc:2956913. and sensitivity analyses cross-checked subsets of results (SI Appendix, Table S3). ACKNOWLEDGMENTS. We thank the Inter-Sectoral Impact Model Intercom- parison Project (ISI-MIP) and CMIP5 for providing data and logistical support, Analyses and Validation. Annual outputs of total animal biomass density and we thank C. Free for sharing data. Financial support was provided by the (grams carbon per square meter) and animal biomass of >10 cm and >30 cm German Federal Ministry of Education and Research through ISI-MIP (Grant were derived on a 1 × 1 degree grid. We calculated time series of % biomass 01LS1201A1), the European Union’s Horizon 2020 Research and Innovation change from 1970 to 2100 relative to 1990–1999 (reference period), and % Program (Grant 678193), and the Ocean Frontier Institute (Module G). We biomass change in 2090–2099 vs. 1990–1999 for each simulation, as absolute acknowledge additional financial support as follows: to H.K.L., W.W.L.C., biomass densities were not strictly comparable across MEMs. Relative and B.W. from the Natural Sciences and Engineering Research Council changes were combined into ensemble means and SD. The climate change (NSERC) of Canada; to D.P.T. from the Kanne Rasmussen Foundation Den- effect [(RCP8.5 − RCP2.6)/RCP2.6] was calculated in a fished and unfished mark; to A.B.-B. from the NSERC Transatlantic Ocean Science and Technology ocean within and across MEMs. Empirical validation was achieved by com- Program; to W.W.L.C. and T.D.E. from the Nippon Foundation-Nereus Program; paring historical projections with biomass trends of assessed fish stocks in a to E.D.G., M.C. and J. Steenbeek from the European Union’s Horizon 2020 Re- fished ocean (B/BMSY; ref. 29) and temperature-dependent biomass hindcasts search and Innovation Program (Grants 682602 and 689518); to E.A.F., J.L.B., and (MSY) of assessed stocks without fishing (30), in addition to published indi- T.R. from Commonwealth Scientific and Industrial Research Organization and vidual MEM validations with empirical data (SI Appendix, Fig. S3). Trophic the Australian Research Council; to N.B., L.B., and O.M. from the French Agence amplification was evaluated by comparing mean (±SD) changes (2090s vs. Nationale de la Recherche and Pôle de Calcul et de Données pour la Mer; and to 1990s) in NPP and total phytoplankton and zooplankton biomass from ESMs S.J. from the UK Department of Environment, Food and Rural Affairs. 1. H.-O. Pörtner et al., Ocean Systems. Climate Change 2014: Impacts, Adaptation, and 23. O. Maury et al., From shared socio-economic pathways (SSPs) to oceanic system Vulnerability Part A: Global and Sectoral Aspects. Contribution of Working Group II to pathways (OSPs): Building policy-relevant scenarios for global oceanic ecosystems and the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, fisheries. Glob. Environ. Change 45, 203–216 (2017). C. B. Field et al., Eds. (Cambridge University Press, Cambridge, UK, 2014), pp. 411–484. 24. M. R. Payne et al., Uncertainties in projecting climate change impacts in marine 2. S. C. Doney et al., Climate change impacts on marine ecosystems. Annu. Rev. Mar. Sci. ecosystems. ICES J. Mar. Sci. 73, 1272–1282 (2016). 4, 11–37 (2012). 25. W. W. L. Cheung et al., Building confidence in projections of the responses of living 3. B. Worm, H. K. Lotze, “Marine biodiversity and climate change” in Climate Change: Ob- marine resources to climate change. ICES J. Mar. Sci. 73, 1283–1296 (2016). served Impacts on Planet Earth, T. M. Letcher, Ed. (Elsevier, Amsterdam, 2016), pp. 195–212. 26. A. Bryndum-Buchholz et al., Twenty-first-century climate change impacts on marine animal 4. M. L. Pinsky, B. Worm, M. J. Fogarty, J. L. Sarmiento, S. A. Levin, Marine taxa track biomass and ecosystem structure across ocean basins. Glob. Change Biol. 25, 459–472 (2019). local climate velocities. Science 341, 1239–1242 (2013). 27. C. J. Brown et al., Effects of climate-driven primary production change on marine food 5. J. García Molinos et al., Climate velocity and the future global redistribution of ma- webs: Implications for fisheries and conservation. Glob. Change Biol. 16, 1194–1212 (2010). rine biodiversity. Nat. Clim. Chang. 6, 83–88 (2016). 28. J. A. Fernandes et al., Modelling the effects of climate change on the distribution and 6. J. L. Blanchard et al., Linked sustainability challenges and trade-offs among fisheries, production of marine fishes: Accounting for trophic interactions in a dynamic bio- aquaculture and agriculture. Nat. Ecol. Evol. 1, 1240–1249 (2017). climate envelope model. Glob. Change Biol. 19, 2596–2607 (2013). 7. C. A. Stock et al., Reconciling fisheries catch and ocean productivity. Proc. Natl. Acad. 29. B. Worm, T. A. Branch, The future of fish. Trends Ecol. Evol. 27, 594–599 (2012). Sci. U.S.A. 114, E1441–E1449 (2017). 30. C. M. Free et al., Impacts of historical warming on marine fisheries production. Science 8. W. W. L. Cheung et al., Large-scale redistribution of maximum fisheries catch po- 363, 979–983 (2019). tential in the global ocean under climate change. Glob. Change Biol. 16, 24–35 (2010). 31. J. A. Estes, M. Heithaus, D. J. McCauley, D. B. Rasher, B. Worm, Megafaunal impacts on 9. J. L. Blanchard et al., Potential consequences of climate change for primary pro- structure and function of ocean ecosystems. Annu. Rev. Environ. Resour. 41, 83–116 (2016). duction and fish production in large marine ecosystems. Philos. Trans. R. Soc. Lond. B 32. G. L. Britten et al., Predator decline leads to decreased stability in a coastal fish Biol. Sci. 367, 2979–2989 (2012). community. Ecol. Lett. 17, 1518–1525 (2014). 10. M. Barange et al., Impacts of climate change on marine ecosystem production in 33. J. A. Estes et al., Trophic downgrading of planet Earth. Science 333, 301–306 (2011). societies dependent on fisheries. Nat. Clim. Chang. 4, 211–216 (2014). 34. Food and Agriculture Organization, The State of World Fisheries and Aquaculture 11. E. D. Galbraith, D. A. Carozza, D. Bianchi, A coupled human-Earth model perspective 2016 (Food and Agriculture Organization, Rome, Italy, 2016). on long-term trends in the global marine fishery. Nat. Commun. 8, 14884 (2017). 35. W. W. L. Cheung, G. Reygondeau, T. L. Frölicher, Large benefits to marine fisheries of 12. S. Lefort et al., Spatial and body-size dependent response of marine pelagic com- meeting the 1.5°C global warming target. Science 354, 1591–1594 (2016). munities to projected global climate change. Glob. Change Biol. 21, 154–164 (2015). 36. B. Planque et al., How does fishing alter marine populations and ecosystems sensi- 13. D. P. Tittensor et al., A protocol for the intercomparison of marine fishery and eco- tivity to climate? J. Mar. Syst. 79, 403–417 (2010). system models: Fish-MIP v1.0. Geosci. Model Dev. 11, 1421–1442 (2018). 37. G. L. Britten, M. Dowd, B. Worm, Changing recruitment capacity in global fish stocks. 14. L. Bopp et al., Multiple stressors of ocean ecosystems in the 21st century: Projections Proc. Natl. Acad. Sci. U.S.A. 113, 134–139 (2016). with CMIP5 models. Biogeosciences 10, 6225–6245 (2013). 38. B. S. Halpern et al., Spatial and temporal changes in cumulative human impacts on the 15. J. Schewe et al., Multimodel assessment of water scarcity under climate change. Proc. world’s ocean. Nat. Commun. 6, 7615 (2015). Natl. Acad. Sci. U.S.A. 111, 3245–3250 (2014). 39. J. Holt et al., Prospects for improving the representation of coastal and shelf seas in 16. F. Piontek et al., Multisectoral climate impact hotspots in a warming world. Proc. Natl. global ocean models. Geosci. Model Dev. 10, 499–523 (2017). Acad. Sci. U.S.A. 111, 3233–3238 (2014). 40. A. C. Ruane et al., The Vulnerability, Impacts, Adaptation and Climate Services Advisory 17. K. Nishina et al., Decomposing uncertainties in the future terrestrial carbon budget Board (VIACS AB v1.0) contribution to CMIP6. Geosci. Model Dev. 9, 3493–3515 (2016). associated with emission scenarios, climate projections, and ecosystem simulations 41. E. A. Fulton et al., “Decadal scale projection of changes in Australian fisheries stocks using the ISI-MIP results. Earth Syst. Dynam. 6, 435–445 (2015). under climate change” (CSIRO Rep. to FRDC Project No. 2016/139, Fisheries Research 18. L. Kwiatkowski, O. Aumont, L. Bopp, Consistent trophic amplification of marine and Development Corporation, Hobart, TAS, Australia, 2018). Downloaded at Universitat Autonoma de Barcelona (UAB) on April 14, 2021 biomass declines under climate change. Glob. Change Biol. 25, 218–229 (2019). 42. S. C. Anderson et al., Improving estimates of population status and trend with su- 19. D. A. Carozza, D. Bianchi, E. D. Galbraith, Metabolic impacts of climate change on perensemble models. Fish Fish. 18, 732–741 (2017). marine ecosystems: Implications for fish communities and fisheries. Glob. Ecol. Bio- 43. W. W. L. Cheung, M. C. Jones, G. Reygondeau, T. L. Frölicher, Opportunities for geogr. 28, 158–169 (2018). climate-risk reduction through effective fisheries management. Glob. Change Biol. 20. V. Christensen et al., A century of fish biomass decline in the ocean. Mar. Ecol. Prog. 24, 5149–5163 (2018). Ser. 512, 155–166 (2014). 44. Y. Ye, N. L. Gutierrez, Ending fishery overexploitation by expanding from local suc- 21. C. Costello et al., Global fishery prospects under contrasting management regimes. cesses to globalized solutions. Nat. Ecol. Evol. 1, 0179 (2017). Proc. Natl. Acad. Sci. U.S.A. 113, 5125–5129 (2016). 45. D. P. Tittensor et al., ISIMIP2a Simulation Data from Fisheries & Marine Ecosystems 22. C. M. Roberts et al., Marine reserves can mitigate and promote adaptation to climate (Fish-MIP; Global) Sector. Potsdam Institute for Climate Impact Research. GFZ Data change. Proc. Natl. Acad. Sci. U.S.A. 114, 6167–6175 (2017). Services. http://doi.org/10.5880/PIK.2018.005. Deposited 31 January 2018. 12912 | www.pnas.org/cgi/doi/10.1073/pnas.1900194116 Lotze et al.