Satellite-based evaluation of cloud cover through three generations of global atmosphere models

Abstract

Cloud cover products from multiple satellite projects are long enough to provide a robust evaluation of climate models. Using global atmosphere models forced by observed sea-surface temperature and employing satellite simulator software, three generations of the Community Atmosphere Model are evaluated. This inter-generational comparison shows how the cloud radiative effect has improved through time but cloud cover has shown only modest improvements over the past decade. Diagnostics are introduced that allow a decomposition of spatial biases to separately evaluate systematic errors in the mean from the spatial variability. Errors in cloud properties are evaluated using a dynamical regimes analysis to connect the climatological errors to the large-scale circulation. Two closely related, current-generation models, CESM2 and E3SM, are compared to show how slightly different model development and tuning decisions can impact the the cloud climatology. Leveraging multiple long-term satellite data sets suggest that despite improvements through time, there remain significant systematic errors in cloud cover. It is suggested that simultaneously constraining cloud radiative effect and cloud cover, and therefore reducing the longstanding "too few, too bright" bias, is feasible and could improve climate projections.