Probabilistic Short-term Distributed Flood Forecasting

Abstract

本研究では，流域における洪水流出量の確率的短期間予測を提供するため，分布型流出モデルに適合する降雨入力データを発生させる確率論的降雨パターンシミュレーション過程を開発する。移流ベクトルによる降雨シミュレーションモデルを降雨パターンの発達による不確実性を考慮し修正することで，6時間にわたる予測降雨パターンの発生を可能にする。分布型でのアンサンブル短期間洪水流出予測を提供するために，モンテカルロシミュレーションを用いる。

A framework is proposed for using distributed rainfall-runoff models for real-timeprobabilistic flood forecasting. A stochastic rainfall pattern simulation model capable ofgenerating input for a distributed rainfall-runoff model is developed to facilitate a short-termprobabilistic forecast of river discharge at multiple locations in a watershed. Generation ofrainfall patterns over a 6-hour period is achieved using a translation vector rainfall forecastingprocess modified to account for uncertainties in rainfall pattern development. The stochasticrainfall generation model is coupled with a distributed rainfall-runoff model in a Monte-Carlosimulation to provide a short-term ensemble forecast of distributed flood discharge. Anadaptive updating procedure for the real-time reduction of forecast error suitable for use witha distributed rainfall-runoff model is developed for the system. An example application of theproposed probabilistic flood stage forecasting system is provided for a typhoon event thatoccurred in the vicinity of the Nagara River watershed located in Gifu, Japan.