Extended Data Fig. 9: Signal-to-noise paradox in the CMIP6 models.
From: Tropical Atlantic multidecadal variability is dominated by external forcing
a, distribution of the post-1950 AMV variance for ensemble member (gray), forced response (EM, blue), total response of model (MOE, black), and observation (red). b, c, d as in a, but for VWS, Sahel rainfall, and AMM. e, ensemble mean of AMV in models with strong aerosol-cloud interaction (blue) and weak aerosol-cloud interaction (red). f, g, h as in e, bur for VWS, Sahel rainfall, and AMM. i, distribution of the post-1950 AMV variance in models with strong aerosol-cloud interaction (blue) and weak aerosol-cloud interaction (red). j, k, l as in i, but for VWS, Sahel rainfall, and AMM. In e-l, models are divided into two composites based on the strength of aerosol forcings19. Models with strong aerosol forcings are represented in blue and they are: TaiESM1, CESM2-FV2, SAM0-UNICON, CESM2-WACCM, CESM2-WACCM-FV2, CESM2, NorESM2-LM, NorESM2-MM, ACCESS-CM2, CNRM-CM6-1, MIROC6. Models with weak aerosol forcings are represented in red, and they are: GFDL-ESM4, MIROC-ES2L, BCC-CSM2-MR, CNRM-ESM2-1, GFDL-CM4, CanESM5, EC-Earth3-Veg, IPSL-CM6A-LR, BCC-ESM1, FGOALS-g3, MPI-ESM1-2-HR, MPI-ESM1-2-LR, INM-CM4-8, CAMS-CSM1-0.
