Model Errors in Precipitation, Surface Insolation and Cloudiness in the NARCCAP Hindcast Experiment

The relationship between the model errors in simulating precipitation, surface insolation, and cloudiness is examined for the conterminous U.S. from the multi-RCM NARCCAP hindcast experiment. A total of four RCMs and their ensemble are incorporated into this study in conjunction with reference data for evaluation from the surface-station based CRU3.1 precipitation data and satellite-based CERES surface insolation and cloudiness data. It has been found that for a majority of RCMs in the NARCCAP hindcast study, the spatial pattern of the insolation bias is negatively correlated with that of the precipitation and cloudiness biases. The relationship varies according to seasons as well with stronger relationship between the simulated precipitation and surface insolation during winter. It is also found that despite the well defined negative correlations among the biases in these three fields, the mean biases for the entire domain are not clearly correlated. This may suggest that RCMs possess higher skill in simulating spatial variability than in the mean fields.