Abstract

Radar and surface rainfall observations are two sources of operational data crucial for heavy rainfall prediction. Their individual values on improving convective forecasting through data assimilation have been examined in the past using convection-permitting numerical models. However, the benefi... Show moreRadar and surface rainfall observations are two sources of operational data crucial for heavy rainfall prediction. Their individual values on improving convective forecasting through data assimilation have been examined in the past using convection-permitting numerical models. However, the benefit of their simultaneous assimilations has not yet been evaluated. The objective of this study is to demonstrate that, using a 4D-Var data assimilation system with a microphysical scheme, these two data sources can be assimilated simultaneously and the combined assimilation of radar data and estimated rainfall data from radar reflectivity and surface network can lead to improved short-term heavy rainfall prediction. In our study, a combined data assimilation experiment is compared with a rainfall-only and a radar-only (with or without reflectivity) experiments for a heavy rainfall event occurring in Taiwan during the passage of a mei-yu system. These experiments are conducted by applying the Weather Research and Forecasting (WRF) 4D-Var data assimilation system with a 20-min time window aiming to improve 6-h convective heavy rainfall prediction. Our results indicate that the rainfall data assimilation contributes significantly to the analyses of humidity and temperature whereas the radar data assimilation plays a crucial role in wind analysis, and further, combining the two data sources results in reasonable analyses of all three fields by eliminating large, unphysical analysis increments from the experiments of assimilating individual data only. The results also show that the combined assimilation improves forecasts of heavy rainfall location and intensity of 6-h accumulated rainfall for the case studied. Show less