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

3.4 Warming effects on biomass responses to changes in NPP In the case of an expected decrease in NPP, biomass response types are correlated with the magnitude of warming under GFDL and IPSL (Pearson’s p-value<0.05, R = 0.79 ± 0.1 and R = 0.64 ± 0.4, respectively) (Figs 7A and S7A). Under GFDL forcing, trophic amplification is observed for all MEMs, with a decrease in total consumer biomass larger than the decrease in NPP (and thus, R, the ratio of change is greater than 1). ZooMSS, Macroecological, BOATS and EcoTroph project trophic amplifica- tion even where no increase in SST is projected, and amplification increases as warming inten- sifies. Thus, the decrease in biomass for these models is about two and a half to four times that of NPP when SST increases by more than 2˚C. In contrast, EcoOcean, FEISTY, and DBEM, project an attenuation of the NPP signal when SST does not change or increase a little (0<R<1, ΔSST = 0, Fig 7). But these three models also exhibit an increasing ratio of change according to the magnitude of warming, highlighting trophic amplification for all increases in SST larger than 0.8˚C. Thus, warming appears to be a factor strengthening trophic amplification. In the case of an expected increase in NPP (NPP change > 1%), a negative inversion (R < 0) is observed, with an effect of the warming intensity which offsets the increase in NPP and induces a decrease in total consumer biomass (Fig 7B). DBEM is an exception with a pro- jected positive attenuation and positive amplification (R>1) in response to temperature increases.

The same analysis with IPSL, projecting a SST increasing up to 5˚C compared to 3˚C under GFDL, exhibits similar qualitative results (S7B Fig), except that only ZooMSS and Macroecolo- gical project negative amplification without SST increases, APECOSM, BOATS, DBEM, DBPM, EcoOcean, EcoTroph, and FEISTY project negative amplification when SST increases by 1˚C. Furthermore, the R ratio tend to reach asymptotic values at these high temperatures compared to GFDL forcing. 3.5 Propagation of trophic amplification across the food web When disaggregating trophic amplification along the food web by size, MEMs exhibit different responses but an enhancement of observed patterns as SST increases (Fig 8). There is a damp- ening of the loss of consumer biomass between each size class of the food web for EcoOcean. While an amplification is projected for the smaller size classes, the magnitude of the decrease in biomass of individuals >90cm is similar to that of NPP under GFDL forcing. Under IPSL, all size classes show a negative inversion, but this inversion is larger for the small size class and close to zero for individuals >90cm. APECOSM exhibits similar patterns with less negative inversion toward the larger consumers. For BOATS, although the negative amplification (GFDL) or negative inversion (IPSL) with temperature across size classes is comparable, mid- size classes (100g-1kg, 1-10kg) show the largest response. Macroecological shows identical amplifications or inversions at all weight classes under GFDL and IPSL forcing respectively.

EcoTroph and FEISTY show greater negative changes in biomass (amplification or inversion) with increasing size. DBPM exhibits an increase in the magnitude of negative inversion of con- sumer biomass with increasing size as EcoTroph and ZooMSS under IPSL, except the highest- size class (>100kg) has the smallest response. PLOS ONE Trophic amplification revealed PLOS ONE | https://doi.org/10.1371/journal.pone.0287570 August 23, 2023 12 / 23