Evaluating the Influence of Resolution and Cumulus Parameterization at 4km on Spatial Precipitation Patterns of NU-WRF in Eastern Kansas and Western Missouri

Dynamic downscaling using Regional Climate Models (RCMs) has become a common approach for obtaining better predictions at local and regional scales. However, there is still much debate whether turning off the cumulus parameterization at convective permitting scales (1-5km) can provide significant improvement in prediction. The uncertainties on resampling and model skills that may not able to fully resolve all convection processes on these gray scales, could overtake the performance of the convective permitting simulations and sometimes provide even better representations. This can be found in some precipitation researches which have more complicate physical processes and larger uncertainty to predict rather than temperature. For precipitation analysis, many researches prove the high skill of the model to capture the large-scale system, however, it’s still a challenge to predict local scale precipitation. To test these questions above, this study focuses on a small region in Central US (mainly include eastern Kansas and western Missouri) during a five-year July from 2000 to 2004. We analyze the uncertainty of resampling (a common preprocess applied before analyzing data in many researches) as a benchmark together with NU-WRF runs at 4km (both cumulus parameterization on and off), 12km and 24km to answer the main questions: (1) Does the 4km run shows significant better spatial pattern of precipitation than 12km and 24km runs? (2) Does turning off the cumulus parameterization at 4km shows better spatial pattern of precipitation than turning it on? These answers will help to evaluate the NU-WRF skill of predicting precipitation on a gray scale of 4km and to help understanding the uncertainty of model prediction.