Convection-Permitting Regional Climate Simulations over South America and the Peruvian Central Andes: Evaluation of Precipitation and MCS Simulations and Sensitivity to Physics Parameterizations

Resolutions of global climate models are too coarse to resolve local forcing and local weather, and some aspects of precipitation simulation are poor. Previous studies have indicated the added value of regional climate model (RCM) dynamic downscaling, especially at convection-permitting resolutions. In South America, the Andes have dynamic and thermodynamic impacts on circulation (e.g., the South American low-level jet) and precipitation systems. As the first step to evaluate and optimize simulations, a RCM based on the WRF model is run for recent 10-year historical period, with a 15-km grid covering the entire South America and a nested 3-km grid covering the Peruvian central Andes region, forced at the lateral boundaries by hourly ERA5 reanalysis data. Characteristics of simulated precipitation and mesoscale convective systems (MCSs) are evaluated using half-hourly GPM IMERG and CMORPH precipitation datasets. Results show that the 10-year, 3-km control simulation successfully captures the four precipitation hotspots along the east slope of the Peruvian Central Andes. However, the simulation generally overestimates the precipitation intensity. The 3-km simulation also reasonably captures the diurnal cycle of precipitation over the mountain, western foothills of Andes, and western Amazon Basin, as well as precipitation propagation. MCS characteristics are retrieved by applying an object tracking method to the hourly simulated and satellite precipitation data. The RCM successfully simulates the spatial distribution and the annual and diurnal cycles of MCS geneses, MCS propagation direction, duration, and precipitation volume; but it overestimates the MCS number in summer, MCS propagation speed, peak and mean precipitation, and underestimates MCS size. Sensitivity experiments on different microphysics, land surface models, and planetary boundary layer (PBL) schemes are conducted during short periods. The results indicate that simulations are most sensitive to PBL scheme, and simulation bias in precipitation intensity can be greatly reduced using the MYNN PBL scheme. These results provide guidance on the optimal configuration of RCM for future climate dynamic downscaling for the Peruvian Central Andes region, for statistical bias correction, and for impact assessments using hydrologic models.