Development of a Flood Forecasting Method Integrating Particle Filter-Based Data Assimilation With Gaussian Process Regression

Key Points

• Develop a flood forecasting method leveraging data assimilation and regression to improve prediction accuracy.
• Integrated particle filter for data assimilation
• Applied gaussian process regression for statistical modeling
• Conducted case study in the Naruse River system
• Compared performance with existing flood forecasting models
• Achieved improved prediction accuracy compared to existing models
• Extended lead time by approximately 1 hour for flood warning
• Properly modeled uncertainty information from particle filter helped maintain accuracy in later forecasting periods