Sensitivity studies and comprehensive evaluation of RegCM4.6.1 high-resolution climate simulations over the Tibetan Plateau

This study provides the first comprehensive assessment of Regional Climate Model version 4.6.1 (RegCM4) for the Tibetan Plateau (TP) region. A wide range of model configurations were analyzed with different parameterizations employed to represent cumulus convection (Kuo, Grell, Emanuel, Kain, and Tiedtke), land surface processes (BATS and CLM), planetary boundary layer turbulence (Holtslag and UW PBL), and radiation (CCM3 and RRTM). In addition to the above experiments at a 30-km horizontal resolution, another experiment was conducted based upon the use of a double-nested dynamic downscaling method to construct a simulation at a 10-km resolution to study the sensitivity to the model resolution. We evaluated a 20-year simulation for precipitation, cloud cover, surface radiation budget, 2-m air temperature, and the surface atmospheric circulation against ground and satellite-based observations during the period 1989-2008. Among the factors analyzed regarding sensitivity, precipitation was unsurprisingly found to be sensitive to the cumulus parameterization scheme, and the CLM is found to reduce rainfall compared with BATS, which is satisfactory for both the Emanuel and Tiedtke schemes. Compared with the cumulus convection schemes, the cloud cover and surface radiation budget are sensitive to the land surface, PBL, and radiation schemes. Generally, the CLM is characterized by reduced mean cloud cover and enhanced surface longwave and shortwave radiation compared with BATS. Conversely, the UW PBL and RRTM radiation schemes result in increased cloud cover and less surface radiation compared with the default options in RegCM4. All experiments, except those employing the Kuo scheme, represent the mean 2-m air temperature and regional circulation patterns reasonably well. At the basin scale, the seasonal cycle and interannual variations of precipitation are found to be not well depicted by most model configurations, although the temperature field was well reproduced. Considering all the analyzed variables collectively, the Tiedtke scheme combined with the CLM land surface model is demonstrated to provide the best performance over the TP. However, the higher-resolution version of the model improves the precipitation simulation significantly, particularly in the Brahmaputra river basin, which is located in the north of the Himalayas.