High-resolution regional climate modeling of warm-season precipitation over the Tibetan Plateau: Impact of grid spacing and convective parameterization

Abstract

The Tibetan Plateau (TP) features high elevation and complex terrain, which together shape the region's unique precipitation with high spatial heterogeneity and strong diurnal variations. Due to low spatial resolution, deficient convective parameterization (CP) and uncertainties in other physical assumptions, global climate models and traditional regional models tend to produce excessive precipitation with an earlier diurnal timing. In this work, two sets of numerical experiments with 4- and 10-km grid spacing, respectively, were performed using the Weather Research and Forecasting model to simulate the seasonal precipitation for an average warm season (April–September 2017) over the TP. Results showed that the 4-km experiment with microphysics only (i.e., without CP) reproduced the precipitation amount, frequency, intensity and diurnal cycle reasonably well on regional and local scales, as well as the nocturnal propagating convection at the lee side. The 10-km experiments with scale-aware CP schemes significantly overestimated the precipitation amount and frequency, underestimated the intensity, and failed to capture the diurnal propagation signature. The 10-km experiment with microphysics only had the poorest skill in representing the spatial distribution of precipitation. The scale-aware CP schemes on the 4-km grid caused a too-frequent occurrence of light precipitation even though their overall contribution to the total precipitation amount was rather small, and thus had little benefit to the model performance. A common deficiency in all experiments was excessive afternoon precipitation across the high elevation of the central Himalaya possibly due to the excessive surface insolation, but the satellite-based data may not accurately capture the afternoon precipitation. This research highlights the value of the 4-km model (the so-called convection-permitting model) and the limited benefit of the tested convective parameterizations over the TP.