High‐resolution simulations of tropical island thunderstorms: Does an increase in resolution improve the representation of extreme rainfall?

Abstract

AbstractRecent increases in computational resources have led to the application of kilometre‐ and sub‐kilometre‐scale simulations in research, numerical weather prediction, and climate modelling alike. Despite anticipated improvements with resolution, there is still considerable work needed to evaluate how well such models improve the representation of intense convection. In this study we conduct ensemble simulations with kilometre‐ and sub‐kilometre‐scale horizontal grids to investigate intense convective events in the tropical island thunderstorm system Hector, which frequently occurs over the Tiwi Islands in North Australia. To avoid losing information through spatio‐temporal averaging we apply a tracking algorithm to simulated and observed storms. When compared with observations, the model storms exhibit a lack of propagation across the study domain. In general, simulated storms are too intense but too small and too short‐lived. This is especially true for the sub‐kilometre simulations, where storms are more intense, smaller, and more numerous than in the kilometre‐scale counterparts. We argue that size and duration errors compensate for storm number and intensity errors, which could lead to misleading interpretations when only comparing time and space averages of rainfall fields. Investigating some properties of the simulated storms suggests that storms with high rainfall intensities have stronger updrafts in the sub‐kilometre model and are accompanied by an increase in cold pool intensity. The results and their resolution sensitivities highlight that the remaining parametrisations and their many tuning parameters in high‐resolution set‐ups influence the representation of convective storms in such models.