Sensitivities of Very Short-Term Numerical Prediction to Polarimetric Radar Data Assimilation: Typhoon Soudelor (2015)

A radar forward operator for reflectivity (Z_H), differential reflectivity (Z_DR) and specific differential phase (K_DP) that follows the CAPS Polarimetric Radar Data Simulator has been incorporated into a regional data assimilation system employing the local ensemble transform Kalman filter. Utilizing this system and the polarimetric radar data from Central Weather Bureau’s RCWF WSR-88D radar, several assimilation experiments as well as a control experiment with no data assimilation (NoDA) have been carried out for a heavy rainfall event in Taiwan associated with Typhoon Soudelor (2015). The sensitivities of the analyses and very short-term forecasts to different assimilated observation variables have been investigated. For the final-cycle analyses, assimilating Z_H or K_DP greatly corrects the over-forecasts of Z_H and K_DP in NoDA but not that of Z_DR; on the other hand, assimilating Z_DR corrects the over-forecast of Z_DR in NoDA to a certain extent but over-corrects those of Z_H and K_DP. For the 1-hour rainfall forecasts, assimilating any of Z_H, Z_DR and K_DP results in similar and significant improvements compared with NoDA. Further analyses and more experiments with various cloud microphysical schemes and assimilation strategies are in progress.