Application of Error-Ensemble prediction method to a short-term rainfall prediction model considering orographic rainfall

In order to improve the accuracy of short-term rainfall predictions, especially for orographic rainfall in mountainous regions, a conceptual approach and a stochastic approach were introduced into a radar image extrapolation using a Translation Model. In the conceptual approach, radar rainfall measurements are separated into orographic and non-orographic rain fields by solving physically-based equations, including additional atmospheric variables, such as vertical wind velocity. In the stochastic approach, mean bias of current prediction errors was estimated and used to adjust mean prediction bias. Furthermore, the vertical wind velocity was updated with the mean bias for convective rainfall. As a result, 1-h prediction accuracy in mountainous regions was much improved for the case study. In the future, improved updating procedures can be expected to allow more accurate predictions.