Understanding CWRF Ability to Simulate U.S. Extreme Precipitation Characteristics

A long-standing model problem is to realistically simulate extreme precipitation characteristics. Most global models tend to underestimate the extreme precipitation intensity, accompanying with more frequent drizzling events. To better understand this problem, we have conducted an ensemble of 24 regional climate simulations over the contiguous U.S. at 30-km grid spacing for 1980-2016 using the Climate-Weather Research and Forecasting (CWRF) model, each with a different physics configuration combining alternate schemes among cumulus, microphysics, cloud, aerosol, radiation, planetary boundary layer, and surface processes. These simulations were driven by the ECMWF-Interim reanalysis and thus best represent the CWRF ability in downscaling U.S. regional climate variations. Our intercomparison among the CWRF mesoscale simulations and the driving reanalysis against observations offers a unique opportunity to determine the sensitivities of model physics configurations and the key mechanisms that affect the ability in representing extreme precipitation variations in regions and seasons.

We found that the cumulus parameterization is the most sensitive model component for extreme precipitation simulation. Among the five major schemes tested, the multiple-closure ensemble cumulus parameterization (ECP) can well capture the spatial patterns, seasonal variations and interannual trends of U.S. extreme precipitation, with better performance even than the driving reanalysis that has assimilated pseudo-observation of daily rainfall distributions. Our analysis of model biases indicated that an accurate representation of cloud amount and convective to total rainfall ratio is essential to a realistic simulation of extreme precipitation. The ECP produces a more accurate cloud amount and thus surface solar radiation as well as a medium convective rainfall ratio as compared with other cumulus schemes. Inadequate radiation causes reduced evapotranspiration and total precipitable water in the Central to Midwest, and insufficient convective potential available energy in the Gulf States, both of which result in underestimation of the convective rainfall ratio and so extreme precipitation. The opposite conditions lead to overestimation of extreme precipitation.