Improvement in Prediction Characteristics of Landfalling Tropical Cyclone Using Multi-Domain Radar Data Assimilation

It is always challenging for operational agencies to accurately forecast tropical cyclone (TC) track, intensity, and rainfall, particularly during the landfall. This work aims to address this issue by assimilating Doppler Weather Radar (DWR) reflectivity (R_f) and radial velocity (V_r) observations at different model resolutions (9 km/3km) and assessing its performance with a lead time up to 72 h for TC Vardah (2016). A total of five experiments are conducted. The control experiment (CNTL) i.e., without assimilation. The 2nd and 3rd experiments, R_f (DA_RF) and V_r (DA_RV) has been assimilated in both domains (9/3km) followed by 4th and 5th experiments, carried out for exclusively assimilating R_f (DA_RF_IN) and V_r (DA_RV_IN) at innermost domain (3 km) respectively. Maximum improvement in TC track, intensity, structure, rainfall, and inner core dynamics are noted while assimilating radar data only in the inner domain (3 km). DA_RV_IN provided the best result with track compared to all other experiments. Overall, the direct position error of track has been reduced in the DA_RV_IN and DA_RF_IN by 30.33% and 8.5%, respectively, compared to DA_RV and DA_RF. Furthermore, DA_RF_IN has shown a 46.2% improvement in rainfall forecast compared to DA_RF. In general, it is found that V_r assimilation improved the intensity in terms of minimum central pressure of TC and captured the tendency of intensity and structure well, particularly at the inner domain. Therefore, this study suggests that assimilating radar observation especially in higher model resolution is the way forward to improve the forecasting of TC characteristics during landfall. This result has direct implications for improving the early warning system over the Indian region.