A regional forecast model evaluation of statistical rainfall properties using the CPOL radar observations in different precipitation regimes over Darwin, Australia

Abstract

This study evaluates the ability of the regional high‐resolution version of the Australian Community Climate and Earth‐System Simulator (ACCESS‐A) model (horizontal resolution of 12 km, 50 height levels with a 37.5 km top) to simulate statistical rainfall properties during three monsoon seasons using the Darwin C‐band polarimetric (CPOL) radar observation as a reference. The comparison consists of characteristics of precipitation, including the total‐domain accumulated daily rainfall, the daily rain intensity and frequency and the probability distribution function of hourly rain rate. These characteristics are then binned using five large‐scale atmospheric regimes that represent the main weather states in that region to assess the ability of the model to reproduce the variability of the statistical rainfall properties as a function of these defined regimes. The main results are that although the model reproduces the accumulated daily rainfall amount reasonably well, albeit tending to slightly overestimate it, the mean rain rate is too weak, compensated by too large a rain frequency of occurrence. In terms of intensity the model tends to perform worse in the dry regimes but improves in the wet regimes, especially during the ‘Deep Westerly’ regime, when most of the monsoonal precipitation is produced. Probability distribution functions of rain rates reveal that although there are small differences amongst regimes, marked discrepancies are found between model and observations. The ACCESS‐A overestimates light rain rates, underestimates moderate to heavy rain rates and produces spurious very heavy rain rates that are not observed at all by the radar. The evaluation of the diurnal cycle of rainfall properties also reveals a general underestimate of rain rates, compensated by an overestimated rain frequency over the whole cycle, resulting in an overestimated amplitude and a wrong timing of the diurnal cycle. In all regimes, the overestimation of total‐domain daily rainfall and rain frequency is found to be due to an overestimation of low rain rates produced by the ACCESS‐A convective scheme, while the underestimation of mean rain rate is due to an underestimate of high rain rates and high rain rates frequency by the ACCESS‐A convective scheme.