B14. EAMv1 Cloud Evaluation

Title

Evaluation of Clouds in the E3SM Atmosphere Model Version 1 with Satellite and Ground-Based Simulators

Authors

Yuying Zhang, Shaocheng Xie, Wuyin Lin, Stephen A. Klein, Mark Zelinka, Po-Lun Ma, Philip J. Rasch, Yun Qian, Qi Tang, Hsi-Yen Ma

Abstract

This study systematically evaluates clouds simulated by the Energy Exascale Earth System Model (E3SM) Atmosphere Model version one (EAMv1) against satellite cloud observations. The simulator package, COSP, is used to facilitate a meaningful “apples-to-apples” comparison between model and observation by considering the different definitions of geophysical quantities among models and observations and the limitations/features of the observing process. EAMv1 is configured at two horizontal resolutions (1º and 0.25º ) and one vertical resolution of 72 layers for different scientific applications. To provide a more complete picture of the model performance in simulating clouds and insights into modeled cloud biases, the evaluation is performed by utilizing unique features of individual instrument contained in COSP in observing different aspects of clouds.

Both low (1°) and high (0.25°) resolution EAMv1 configurations generally underestimate clouds in low and midlatitudes and overestimate clouds in the Arctic although the error is smaller in the high-resolution model. The underestimate of clouds is due to the underestimate of optically thin to intermediate clouds, as EAMv1 generally overestimates optically intermediate to thick clouds. Other model errors include the largely under-predicted marine stratocumulus along the coasts and high clouds over the tropical deep convection regions. The underestimate of thin clouds results in too much LW radiation being emitted to space and too little SW radiation being reflected back to space while the overestimate of optically intermediate and thick clouds leads to too little LW radiation being emitted to space and too much SW radiation being reflected back to space. EAMv1 shows better skill in reproducing the observed distribution of clouds and their properties and has smaller radiatively relevant errors in the distribution of clouds than most of the CFMIP1 and CFMIP2 models. It produces more supercooled liquid cloud fraction than CAM5 and most CMIP5 models primarily due to a new ice nucleation scheme and secondarily due to a reduction of the ice deposition growth rate.