Deviation Distribution Features of CMA-GFS Cloud Prediction

Abstract

Clouds play a vital role in weather, climate system and the atmospheric water cycle. The diagnosis and evaluation of numerical model prediction results is important for numerical model research and development. Reasonable diagnosis and evaluation methods can not only provide references for model researchers to optimize model schemes, but also help users understand the performance of model prediction. The cloud characteristics of different regions and seasons should be considered for evaluation because the attributes in different regions are markedly different. The performance and deviation characteristics of the operational CMA-GFS of four seasons are evaluated, based on the reanalysis of ERA5 reanalysis data from March 2021 to February 2022. The frequency bias of cloud occurrence, cloud fraction, integrated cloud hydrometeors from various levels, and the bias and root mean square error of those variables are carefully diagnosed and evaluated via different methods. The deviation characteristics of cloud are emphatically analyzed according to different regions. The possible causes for the significant difference of cloud prediction deviation characteristics at different levels in different regions are preliminarily discussed. The results show that the overall distribution of cloud predicted by CMA-GFS is reasonable, which can describe the meridional peak and valley distribution characteristics of global cloud and reflect the seasonal trend. The cloud amount deviation of high cloud and medium cloud by CMA-GFS is greater than that of low cloud. The root mean square error of cloud amount of high cloud and medium cloud is smaller than that of low cloud. And the model stability for high cloud and medium cloud prediction is also better. The liquid-phased hydrometers integration is mainly negative deviation, and the ice-phased hydrometers integration is mainly positive deviation. The causes of the deviation of cloud predicted by CMA-GFS are different in different regions. In tropical region the deviation is related to the incongruity of convective parameterization and microphysical schemes, while in middle and high latitudes regions the deviations are related to the bias of relative humidity. It also shows that the diagnosis of model cloud features will cover up the actual problems only by a single method, and it needs to be evaluated comprehensively by combining a variety of methods.

Keywords

• cloud occurrence
• cloud fraction
• cloud hydrometeor integration