Differential Credibility Analysis of Dynamical Downscaling Framework with a Focus on Precipitation Characteristics over Southern Great Plains

Credibility assessment of different dynamical downscaling approaches is important for effectively communicating model projection and associated uncertainty to stakeholders. To this end, we produced downscaled regional climate data by three atmospheric models: RegCM4 and Weather Research and Forecasting model (WRF) on limited-area grids of 12, 25, and 50 km grid spacing, and the Model for Prediction Across Scales coupled to the Community Atmosphere Model physics (CAM-MPAS) on global uniform 12 grids and variable-resolution (VR) grids with refinement resolutions similar to the two limited-area models. A variety of metrics are applied to the simulated and observed daily precipitation over the Southern Great Plains (SGP) to assess model credibility. CAM-MPAS VR demonstrates benefits of mesoscale (<=50 km) resolutions compared to 12 global grids by improving precipitation intensity and frequency. Within the mesoscale-resolution range, RegCM4 and WRF outperform CAM-MPAS. Particularly notable biases of CAM-MPAS compared to other models are overestimated frequency of low-intensity events, stronger negative correlation between precipitation and near-surface temperature, and larger seasonal variability in precipitation intensity spectra. In all three models, finer resolution does not consistently improve simulated precipitation. In RegCM4 and WRF, some precipitation characteristics are sensitive to different lateral boundary conditions provided by ERA Interim and a CMIP5 model, but neither horizontal resolution nor lateral boundary conditions explain the large gap between the limited-area and global VR simulations. Changing the default deep convection scheme of CAM-MPAS to the Grell-Freitas scheme largely alleviates the aforementioned biases and helps to close the gap with RegCM4 and WRF both featuring well-optimized physics. The result suggests a more prominent role of physics parameterizations than resolutions or large-scale forcing for better simulating the characteristics of observed precipitation over SGP. An implication is that modeling frameworks with more flexibility in optimizing physics parameterizations has an advantage for producing credible regional climate information through downscaling.