Trophic amplification: A model intercomparison of climate driven changes in marine food webs

production and biomass. Nevertheless, the MEMs ensemble projections driven by GFDL sug- gest that a global negative trophic amplification, characterised by a decrease in total consumer biomass much larger than the decrease in NPP, is projected throughout the 21st century under a high emissions SSP5-8.5 scenario. In addition, the global increase in NPP projected by IPSL does not lead to a global increase in consumer biomass but a decrease that is even larger than for GFDL in some cases. These outcomes suggest the projected decrease in biomass may be at least partially independent of the change in NPP since it occurs even when a positive change in NPP is projected. The amplification of the input NPP signal, what is commonly called “trophic amplification” appears to be the result of additional processes leading to a decrease in con- sumer biomass. Thus, with or without NPP decrease, based on SSP5-8.5 GFDL and IPSL simu- lations, the average additional projected decrease in consumer biomass in the 2090s is on the same order of magnitude (16.7% ± 9.5%). All MEMs we considered are forced by temperature (SST, epipelagic temperature, and/or bottom temperature), primary producers (production/ biomass depending on MEMs) and/or secondary producers (S1 Table and S9 Fig; [5]), which implies that the projected decrease in consumer biomass occurring throughout the food web is most likely induced, at least partly, directly or indirectly by temperature-related rates or processes. Spatial variability of biomass responses was observed, highlighting different types of

responses throughout the food web. On the one hand, where NPP is projected to decrease, 61.4% and 31.4% of ocean area under GFDL and IPSL, respectively (S3 and S5 Figs), a "nega- tive amplification" response is projected by a majority of MEMs (89% and 82% of the ocean surface on average, respectively). Previous studies [22, 24] pointed out similar areal coverage of negative amplification at low trophic levels, but here we show that amplification has a simi- lar areal coverage when considering the whole food web. In all MEMs, temperature seems to amplify the energy losses when the NPP is projected to decrease. On the other hand, where NPP is projected to increase, various responses are projected but the main response is an inversion of the climate signal from primary producers to higher trophic levels for all MEMs leading on average to “negative inversion” on 70% of the considered ocean surface, according to both IPSL and GFDL. This negative inversion also strengthens with increasing temperature. 4.2 Limitations of the model intercomparison First, the use of two ESM forcings highlights one major caveat that is now well known: the uncertainties associated with the trends of NPP and lower trophic level response to climate change in the ESMs. Furthermore, while agreement between ESMs over SST has been improved from CMIP5 to CMIP6, the differences in projected NPP increased: NPP projec- tions differ ±30% between CMIP5 and CMIP6 [5, 12]. Specifically in our study, the NPP trends over the late 21st century are different under GFDL and IPSL (Figs 1B and S2B). Given the

importance of NPP and low trophic levels on the functioning of ecosystems and their influence on the climate response projected by the MEMs [5, 18], there is a strong need for more research on their response to climate change to constrain uncertainty. Secondly, to enable the best possible comparison between models, we do not consider fish- ing effects because not all Fish-MIP MEMs represent fishing and also because plausible future scenarios of fishing were not available. Future Fish-MIP simulations that include fishing may allow us to investigate the fishing effects on trophic amplification. By inducing additional bio- mass losses at mid- and high-trophic levels, fishing in the MEMs may exacerbate the projected trophic amplification, or alternatively cause a top-down trophic cascade of alternating amplifi- cation and attenuation or inversion (e.g., [50]). PLOS ONE Trophic amplification revealed PLOS ONE | https://doi.org/10.1371/journal.pone.0287570 August 23, 2023 14 / 23