Cloud responses in the AMIP simulations of CMIP5 models in the southeastern Pacific marine subsidence region

摘要

The sensitivity of regimes dominated by low clouds has been identified as the largest contributor to uncertainties in tropical cloud feedback estimates in climate models. The Atmospheric Model Intercomparison Project simulations of the low cloud response to sea surface temperature (SST) are compared with satellite observations in the southeastern Pacific subsidence region. The model ensemble annual average cloud fraction is only about 10% lower than Moderate Resolution Imaging Spectroradiometer observations; however, many models compensate by overestimating the cloud liquid water path (LWP), especially in areas typically associated with shallow cumulus. Analysis of the monthly distribution also shows that models have considerable difficulty in simulating the annual cycle in the cloud radiative effect (CRE), cloud fraction, and LWP likely due in part to underestimation of the strength of lower tropospheric stability and the depth of the boundary layer. The interannual sensitivity of CRE to SST agrees with observations in about half of the models, with the other half generally underestimating the cloud radiative forcing sensitivity. Model-observational differences are driven by the varying interannual responses in cloud fraction and LWP. Most models, including those that capture the mean interannual sensitivity of CRE to SST, have lower sensitivity in cloud fraction that is compensated by oversensitivity in the cloud LWP, especially in areas of more frequent shallow cumulus. Results presented here also highlight the possibility of using the vertical gradient of moist static energy (MSE) to test the fidelity of a model's representation of clouds and cloud sensitivity. Models that reproduce the observed distribution of cloud fraction with the lower tropospheric MSE gradient not only show better regional distribution and annual cycle in clouds and radiative forcing, but also demonstrate cloud and radiative sensitivities to SST that are more well correlated with the observed cloud sensitivities.