Improving Very Short-Term Rainfall Prediction by Improving Ensemble Initialization for Wlras

The WRF-Local Ensemble Kalman Filter with Radar data Assimilation System (WLRAS) has beenestablished for improving the very short-term heavy rainfall prediction in Taiwan. Although WLRAS canprovide significant improvement on heavy rainfall prediction, there are still many challenging issues due tothe complex terrain, complicated mechanisms for strong convections and heavy rainfall at different timescales and unknown errors in model physics. This proposal aims to improve the rainfall prediction byproving a better model moisture field before conducting WLRAS. This will be achieved by assimilatingmoisture-related observations, such as the satellite-derived temperature and moisture profiles and GPS-ZTDdata, to improve the model condition in the parent domain. Also, we would like to improve the representationof model errors by using the stochastic multi-scale perturbation generator in order to increase the ensemblespread for radar data assimilation. The optimal goal is to improve very short-term rainfall prediction andprovide a reliable ensemble production, such as the PQPF.