Development of a Japan Flood Risk Model for General Insurance (1) Development of a Stochastic Rainfall Event Model

Abstract

This paper presents a stochastic rainfall event model, which generates a long-term synthetic rainfall series (or distributions), as a component of a flood risk model for the whole Japan. Annual frequency of rainfall event, accumulated volume of rainfall, and spatiotemporal distribution of rainfall are statistically modelled, and then synthetic rainfall events are generated by a Monte Carlo simulation. The model was verified with the observed rain data in terms of frequency of heavy rain and return period of annual maximum 72-hour/24-hour rainfall. The annual maximum rainfalls at 100 year return period that was estimated from the synthetic rainfall events were in good agreement with those estimated from the observed rainfall data. Thus, the synthetic rainfall events were thought to be valid for a flood risk model.