Cloud and Precipitation Prediction Within the GFDL SHiELD Model

Cloud and precipitation prediction poses considerable challenges in many numerical models. The System for High-resolution prediction on the Earth-to-Local Domains (SHiELD), developed at the Geophysical Fluid Dynamics Laboratory (GFDL), a Unified Forecast System (UFS) prototype atmospheric model, is used to evaluate the prediction accuracy of clouds and precipitation. In SHiELD, the complexities of cloud and precipitation prediction have resulted in noticeable biases in predicting the geographic distribution of precipitation, precipitation diurnal cycle, ice and liquid water path, and cloud fraction. We assessed the accuracy of the SHiELD model's predictions of clouds and precipitation and evaluated to what extent we could enhance the prediction accuracy by doubling the model's horizontal resolution from 13 km to 6.5 km. We found that the SHiELD prediction system exhibits the potential for improving cloud and precipitation prediction via higher horizontal resolutions. Significant tests help uncover and understand the true biases in the SHiELD system, with the goal of proposing solutions for improving cloud and precipitation prediction

Figure: Precipitation prediction of SHiELD in the global domain. From top to bottom are mean
precipitation, bias (difference between model and observation), and root mean square error
(RMSE). In these panels, C768 is the 13km-resolution model, C1536 is the 6.5km-resolution
model, and OBS is the MSWEP precipitation observation. The shaded areas are the
significance intervals: difference between model and observation in the top panel, difference
between two models in the middle and bottom panels. The numbers in the legends are the
time-averages.