Improvement of the WRF-LETKF Radar Data Assimilation System for Heavy Rainfall Prediction Involving Multiscale Interactions

The WRF-LETKF radar data system (WLRAS) has been established for the quantitative precipitation prediction and has demonstrated very useful skill in probability quantitative precipitation prediction. However, the heavy precipitation events in Taiwan often involve multi-scale interactions with synoptic-scale meiyu front and topography. Different strategies associated with multi-scale data assimilation are sought to improve the heavy rainfall prediction.

First, the ground-based GNSS zenith total delay data is assimilated, in addition to the use of WLRAS, to provide the meso to convective scale moisture corrections. Assimilating the ZTD data is very useful in correcting the locations of heavy rainfall. Also, the dual-localization method (Kondo and Miyoshi, 2013) is implemented in WLRAS to provide the corrections of wind and moisture in larger scale. Based on a heavy rainfall episode on 2^ndJune 2017, the dual-localization method can successfully improve the movement and orientation of the front as it approaches Taiwan. The interaction between front and the topography can be better represented and thus the low-level moisture convergence and local barrier jet are enhanced. As a result, the location and intensity of the heavy rainfall can be greatly improved.