Abstract
Low cloud feedbacks are at the heart of climate model uncertainty of projections of anthropogenic warming. In this work, I aim to quantify the role of local meteorology in controlling stratocumulus (Sc) and shallow trade cumulus (ShCu) cloud fraction throughout the tropics, evaluate regional differences in these relationships, test the robustness of the observed relationships across different datasets, and examine how these relationships are modeled in current state-of-the-art global climate models (GCMs). First, I analyze the covariability of Sc and ShCu with local meteorological cloud controlling factors (CCFs) and assess the responses of Sc, ShCu, and CCFs to interannual surface temperature changes. First, there is generally good agreement among datasets on the sign of the Sc sensitivities to CCFs across the tropics: strong temperature inversions, high free-tropospheric relative humidity (RH700) (with the exception of the northeast Pacific), low sea surface temperatures, strong surface winds, and cold temperature advection all support higher Sc cloudiness. All observational datasets suggest that EIS is the dominant control on Sc cloud fraction. On the other hand, no singular CCF dominantly controls ShCu cloud fraction across the tropics: substantial regional variability is observed in the ShCu-CCF relationships. In general, however, reductions in estimated inversion strength (EIS), relative humidity at 700 hPa (RH700), and subsidence, as well as increases in sea surface temperatures (SST) and warm temperature advection, support a Sc to ShCu transition on the western edges of Sc regions. A complementary analysis is then performed with GCM output to assess model performance in capturing these sensitivities. Compared with observations, tropical Sc and ShCu in GCMs are overly sensitive to EIS, SST, RH700, and vertical velocity and not sensitive enough to wind speed and temperature advection. The largest multi-model mean changes and largest inter-model spread in low cloud fraction (LCF) changes in the tropics occur in the equatorial Pacific in response to EIS, SST, and vertical velocity perturbations. The larger contribution from vertical motion in the models suggests that the modeled low cloud feedback will be more sensitive to future large-scale overturning (Hadley/Walker) circulation changes than the observations suggest they should be. Finally, the linear model derived from our multiple linear regression analysis of the modeled output does substantially better at predicting LCF changes than the linear model predicting LCF changes in the observations, suggesting that (a) nonlinearities in the real world that are non-trivial to predicting LCF changes are absent from the models and/or (b) the linear model is too limited by the six chosen CCFs. This work emphasizes the importance of using multiple observational datasets to observationally constrain low cloud feedbacks, highlights priorities for improving parameterization of low cloud processes, and underscores the need to expand our analysis of LCF sensitivity to CCFs to include additional predictors and/or nonlinear methods.
