Application of Remote Sensing and Gauged Precipitation Information for Improving Hourly Typhoon Rainfall Forecasting of WRF
Jhong-Wei Chen

This study aims at developing a weighted WRF ensemble model (W-WRFEM) based on remote sensing precipitation information to improve hourly typhoon rainfall forecasting by the WRF ensemble model. Based on gauged precipitation information, the W-WRFEM is further coupled with two error correction models developed by Random Forest (RF) and Support Vector Machine (SVM), respectively, to increase the forecasting accuracy of W-WRFEM. The result indicates that (1) the W-WRFEM based on QPESUMS radar rainfall with the weighting method, Rank Reciprocal Method, performs better than the other weighting methods; (2) the W-WRFEM coupled with the error correction model by RF has a better performance than the error correction model by SVM and has the ability to improve 1-hour- and 2-hour-ahead rainfall forecasting.

Key words: weighted WRF ensemble model, remote sensing precipitation information, hourly typhoon rainfall forecasting, Random Forest, Support Vector Machine