APPLICATION OF ERROR-ENSEMBLE METHOD TO A SHORT-TERM RAINFALL PREDICTION WITH TRANSRATION MODEL CONSIDERING OROGRAPHIC RAINFALL

Abstract

To improve short term prediction accuracy of rainfall, especially for orographic rainfall over mountain areas, a physical approach and a stochastic approach were introduced into a radar image extrapolation method with translation model. In the physical approach, radar observed rain field is separated into orographic and non-orographic rain fields by solving physically-based equations. In the stochastic approach, a bias-modified prediction field was calculated from a lot of spatial error distributions of the pre-predicted rain fields. In order to improve the separation accuracy of the physically-based orographic model, vertical wind velocity W is updated so that the physically-based procedure could re-predict the rain filed which meets to the bias-modified deterministic prediction field. As a result, one-hour prediction accuracy especially over mountain areas were much improved. In the future, more accurate predictions can be expected by improving the W-updating procedure.