Using the Annual Cycle to Understand Climate Model Biases in Trade-wind Clouds

Abstract:

This study investigates the clouds of the north Atlantic trade-wind region as represented in climate models. We focus on an area near Barbados, a site with long-term cloud observations. We begin by asking whether the annual cycle of cloud cover is properly represented in models compared to satellite observations and reanalysis. The models robustly overestimate the amplitude of the annual cycle in cloud cover compared to satellite estimates, but qualitatively capture the phase of the annual cycle with the cloudiest months in the summer wet season in association with deeper convection and higher clouds. We show that the wet season biases are associated with biases in the large-scale circulation, in particular the location of the ITCZ. During the dry season, however, cloud cover is underestimated. Comparison of simulated cloud fraction with the expected cloud-controlling factors shows only weak relationships and little correspondence between the models and observation-based estimates. We infer that the cloud cover variations within the dry season are controlled at least as much by local factors as large-scale ones; these local factors are determined by parameterized physics in the climate models. Like the cloud cover, the vertical structure of the the simulated clouds varies tremendously across models. Perhaps unsurprisingly, the shortwave cloud radiative effect (SWCRE) is relatively well-captured by the models during the dry season, signaling a common compensating bias among the models and reaffirming the ''too few, too bright'' error. Conditioning on dry season and SWCRE shows that models diverge as to the cause of SWCRE variation, some having more influence from shallow cumulus cloud variation while others show more dependence on middle and upper-level clouds. These higher clouds are usually considered unimportant for the overall cloudiness and the shortwave radiation budget in the trades, but this does not appear to be the case at least for some models. Since shortwave cloud feedback in trade-wind regions is a leading contributor to the spread in climate sensitivity, a better understanding of these differing cloud distributions and how they compare to observations could shed light on models' cloud feedbacks and provide an observational constraint with the potential to narrow the spread in climate sensitivity.