Dynamic downscaling based on weather types classification: An application to extreme rainfall in south-east Japan

Abstract

Optimisation physical scheme for weather forecast systems is an essential part of the development of efficient real-time weather forecasts. This research examines the use of weather type to cluster and to optimise the physical scheme of numerical weather prediction systems in the south-east coast of Japan in the Tokyo area. In this study, we calibrated and validated physical schemes by the observed rainfall at gauge stations around Tokyo, we have classified these per weather circulation. We used 24 ensemble members and 20 heavy rainfall events classified in 4 weather types to validate the scheme. The physical scheme ensemble construct by the association of micro-physics, cumulus, planetary boundary and radiative scheme and then, simulated with the weather research and forecasting model. We observed limited physical scheme variability for a pool of station within a weather type cluster. However, it shows large variations among stations. Then, we computed the rainfall Cumulative Probability Distribution Function curves which indicated wide differences. It suggests that clustering presents interesting properties. The results permitted (a) the selection of an optimal physical scheme per weather type, (b) the development of bias correction curve specific to weather type and (c) the evaluation the spatial distribution of regional bias correction.