Quantifying contributions of model processes to the surface temperature bias in FGOALS-g2

Abstract

To quantify the annual mean surface temperature bias due to various processes in Flexible Global Ocean-Atmosphere-Land-System model, Grid point version 2 (FGOALS-g2), the climate feedback-response analysis method (CFRAM) is used to isolate contributions from both radiative and nonradiative processes in the model by comparing the model simulation with ERA-Interim reanalysis. The observed surface temperature bias is decomposed into seven partial temperature biases associated with surface albedo, water vapor, cloud, both surface sensible and latent heat fluxes, land/ocean heat transport processes, and atmospheric transport processes. The global mean cold bias (-1.39 K) is mostly attributed to surface albedo and land/ocean heat transport processes while surface latent heat fluxes tend to weaken this bias. Cloud-induced bias is dominated by shortwave cloud radiative effect (SWCRE) over low-latitudes and longwave cloud radiative effect (LWCRE) over high latitudes. The mixed layer depth (MLD) bias is consistent with the bias due to ocean heat transport over North Pacific, North Atlantic, and the Southern Ocean. On global scale, contributions of radiative processes and nonradiative processes to the total observed cold bias are comparable, but tend to compensate each other over most regions except for the northern high latitudes. We suggest that the improvements in tropical clouds in the model may significantly decrease the global temperature bias through the interaction between clouds and circulation.