Linking Coupled Model Errors in Simulating East African Climatological Rainfall to Model Biases in SSTs

A well-known feature of the CMIP5 generation of coupled climate models is their inability to properly simulate the annual cycle of rainfall in East Africa. In observations, much of the region has two rainy seasons, the short rains from September-November and the long rains from March-May. Despite their names, what distinguishes these seasons is related more to the amount of rainfall received than the duration of the respective seasons. Climatological rainfall is greater during the long rains than in the short rains. The CMIP5 models generally reverse these amplitudes, making the short rains too wet and the long rains too dry. The models also exhibit systematic errors in the phase of the rainfall seasons, particularly during the long rains. In this study two atmospheric GCMs were forced with observed monthly SSTs as a control and observed SSTs plus the mean monthly CMIP5 SST bias from 31 models. The control runs are found to capture the annual cycle in East Africa quite well while the bias runs exhibit many of the same errors in the rainfall annual cycle that are found in the CMIP5 model output, pointing to the important role of the latter in simulation errors of East African climate. Some physical reasons for the CMIP5 model behavior will be discussed.