Next-generation ensemble projections reveal higher climate risks for marine ecosystems

Anthropogenic climate change poses a significant threat to marine ecosystems, intensifying impacts like increased mortality, reduced calcification, and altered species distributions, interactions, abundance, and biomass. These climate change effects can interact with other stressors, such as overfishing, jeopardizing marine conservation and societal benefits derived from the ocean. Understanding these climate change risks and the benefits of mitigation is crucial. Model intercomparison projects (MIPs), such as the Coupled Model Intercomparison Project (CMIP), generate ensemble projections to quantify inter-model spread and improve models. CMIP6, currently in its sixth phase, provides updated oceanographic drivers for impact models. This study compares MEM ensembles forced with CMIP5 and CMIP6 outputs from GFDL and IPSL ESMs, leveraging the improved representations of marine biogeochemistry, sea ice, and other properties in CMIP6. The CMIP6 models exhibit increased climate sensitivity, leading to stronger marine ecosystem forcings under high-emission scenarios, although with greater variability in net primary productivity (NPP) impacts. Three global MEMs were added to the Fish-MIP ensemble, resulting in a total of nine models. The study focuses on temperature and productivity as key drivers of marine ecosystem change and compares a subset of models using both CMIP5 and CMIP6 forcings to assess the impact of improved climate models on global marine ecosystem projections and mitigation benefits.

Previous research using CMIP5 data and Fish-MIP has already shown global and regional changes in marine ecosystems over the coming century and their socioeconomic consequences. Studies have highlighted the combined effects of fishing and changes in primary productivity on fish communities, and the potential for widening socioeconomic equity gaps due to ocean biomass losses. However, uncertainties in projecting climate-change impacts remain a significant challenge. Individual marine ecosystem models (MEMs) have explored climate impacts, but Fish-MIP provides a standardized comparison of ensembles, offering more robust projections and insights into the overall consequences of climate change on marine ecosystems.

This study employs the Fish-MIP framework, which uses standardized protocols to compare multiple global marine ecosystem models (MEMs). Monthly oceanographic data from CMIP5 (using GFDL-ESM2M and IPSL-CM5A-LR) and CMIP6 (using GFDL-ESM4 and IPSL-CM6A-LR) Earth System Models (ESMs) under strong-mitigation (RCP2.6/SSP1-2.6) and high-emission (RCP8.5/SSP5-8.5) scenarios were used to force the MEMs. The models were run under a common simulation protocol, with slight modifications to accommodate differences in the CMIP6 outputs. The primary output of interest was the total unfished marine animal biomass. Nine global MEMs were included in the analysis, with three new models added to the previous six models from CMIP5. Each model uses different approaches, from simple size-based models to complex, composite models incorporating multiple ecological processes and species interactions. To allow a more precise comparison, a subset of the same six MEMs was analyzed separately. Changes in oceanographic properties, such as sea surface temperature (SST), net primary production (NPP), phytoplankton and zooplankton biomass, and marine animal biomass were assessed, calculated relative to a baseline period (1990-1999). Spatial and temporal variations in biomass changes were also analyzed. Statistical tests (Wilcoxon rank-sum tests) were conducted to identify significant differences in biomass changes between CMIP5 and CMIP6 scenarios and ensemble intermodel standard deviations were calculated to quantify model uncertainty.

The CMIP6 ESM simulations showed stronger mean surface warming of the global ocean compared to CMIP5 under the high-emission scenario, but less difference under strong mitigation. Projections of NPP showed more diversity, with CMIP6 GFDL projecting declines while IPSL projected increases, particularly in subtropical gyres. Despite increased NPP in some CMIP6 scenarios, phytoplankton and zooplankton biomasses showed declines, similar to CMIP5 but with greater differences between the ESMs. Spatially, CMIP6 projected SST increases across most of the global ocean under high emissions, with larger changes in polar regions. Polar regions showed NPP increases while other regions showed decreases in NPP under CMIP6 high emissions. Phytoplankton and zooplankton biomass changes were more consistently negative under CMIP6, except at the poles. The mean global marine animal biomass decline was larger under CMIP6 than CMIP5 after 2030 for both scenarios. The difference was statistically significant, with CMIP6 projecting a decline of ~19% under high emissions by 2099 (compared to ~16.5% under CMIP5) and ~7% under strong mitigation (compared to ~5%). The inter-model standard deviations were narrower under CMIP6 high emissions after 2080s, suggesting less uncertainty, although this was primarily driven by the high-emissions GFDL forced MEMs. Spatial patterns of biomass changes under CMIP5 and CMIP6 were similar globally, but notable regional differences were observed, particularly in the Arctic, where CMIP6 projected increases while CMIP5 showed both increases and decreases. Model agreement was generally high under CMIP6 except in subtropical gyres due to differences in CMIP6 NPP responses. The analysis of the subset of six MEMs showed similar trends.

The increased climate sensitivity in CMIP6 models, particularly in IPSL and GFDL simulations, explains the larger declines in marine animal biomass projected under CMIP6. Warming affects various ecological processes, impacting metabolic costs, biomass production, mortality, and species distributions. The inconsistent NPP signals from CMIP6 ESMs, despite increased warming, could lead to an illusion of improved model agreement. The narrowing of the inter-model standard deviation under CMIP6 high emissions is mainly due to the consistent decline in marine animal biomass among GFDL forced MEMs. Despite variations in the ways MEMs incorporate lower-trophic level forcing, the consistent decline in biomass across all MEMs under high emissions suggests that temperature effects have a significant impact. The study highlights that while there is an overall increase in model agreement (in the direction of change) under CMIP6, there are still significant regional differences, especially in areas such as subtropical gyres. Despite improvements in some MEMs, the dominant effect appears to be due to the changes in the climate forcings, reflecting the need for improved lower trophic level modeling within MEMs.

This study demonstrates that next-generation CMIP6 climate models, when used to drive marine ecosystem models, project steeper declines in global marine animal biomass compared to their CMIP5 counterparts. The results underscore the benefits of strong mitigation efforts to reduce climate change impacts on marine ecosystems. The significant regional variations in projected biomass changes and the remaining uncertainties highlight the need for further model refinement and improvement in lower trophic level representation in order to support robust adaptation and mitigation strategies. Future research should focus on refining MEMs, enhancing the coupling of ESMs and MEMs, and incorporating factors like fishing pressure to obtain more realistic and reliable projections of climate change impacts on marine ecosystems.

The study focuses on two ESMs (GFDL and IPSL) and two scenarios (strong mitigation and high emission). The lack of inclusion of internal climate variability as a source of uncertainty and the exclusion of fishing impacts in the models could underestimate the actual effects of climate change. Additionally, the models do not account for potential species adaptation or evolution and lack full bi-directional coupling between higher and lower trophic levels. The study's focus on total animal biomass may not fully represent the changes in edible biomass available for fisheries. Lastly, the number of MEMs included in the analysis, while improved relative to prior studies, still represents a relatively small ensemble of models.