ENTIRE HYOGO PREFECTURE USING COVARIANCE MATRIX IMPROVING ACCURACY OF REAL-TIME FLOOD PREDICTION SYSTEM

Abstract

 The increase in heavy rains due to global warming in recent years is causing floods in various places. In Hyogo Prefecture, we operate a real-time flood prediction system for the entire Hyogo prefecture, with the aim of predicting water levels and flooding up to 6 hours in advance, taking into account flooding in small and medium-sized rivers. In this study, with the aim of further improving accuracy, we estimate a background error covariance matrix that takes into account the decline in correlation using four indicators such as river channel distance and straight line distance, and use an optimal interpolation method as a surface water level correction method. Water level data was assimilated. Cross-validation of prediction calculations was performed using the assimilated water level data as the initial calculation value. As a result of verifying the observation points by dividing them into assimilation points and verification points, it was confirmed that the accuracy had improved at the assimilation points and that the accuracy had not decreased at the verification points. It was applied to a real-time flood prediction system.