Convection-permitting climate simulations over South America: Experimentation during different phases of ENSO☆

This paper presents the first-ever continental-scale convection-permitting simulations over South America for three water years of different ENSO phases, corresponding to an ENSO neutral year (2018/19), an EI Nino year (2015/16), and a La Nina year (2010/11), using the Weather Research and Forecasting (WRF) model at 4-km grid spacing. The model performance has been validated against precipitation derived from satellite, surface observations, and surface air temperature from reanalysis. The evaluation shows a promising skill at reproducing the observed multi-scale spatiotemporal characteristics of precipitation and temperature, such as the seasonal and sub-seasonal variability, the diverse patterns of diurnal cycle, and deep convective clouds. Sensitivity simulations quantify the impacts of cumulus parameterization, grid spacing, and spectral nudging. Results indicate that a tested scale-aware convection scheme has little benefit, and the model performance degrades as horizontal resolution decreases. Spectral nudging can reduce the precipitation bias over some tropical and subtropical regions but exacerbates the wet bias over the Andean Mountains. A noteworthy model deficiency shared in all simulations is the excess orographic precipitation, a problem in association with the overly active afternoonevening convection possibly resultant from under-representation of clouds and missing cloud-aerosol interaction, though the uncertainty of observational data might contribute to the wet bias as well. These results provide useful guidance for improving the model physics. The overall encouraging agreement between the 4-km model simulations and observations provides confidence in the usage of the established model configuration for regional climate downscaling and climate change projections over South America.