Flood forecasting and uncertainty assessment with sequential data assimilation using a distributed hydrologic model

Accurate flood forecasting is essential for mitigating flood damage and addressing operational flood scenarios. In recent years, data assimilation methods have drawn attention due to their potentials to handle explicitly the various sources of uncertainty in hydrologic models. In this study, we implement sequential data assimilation for short-term flood forecasting and parameter uncertainty assessment using grid-based spatially distributed hydrologic models. The lag-time window is introduced to consider the response times of internal hydrologic processes. Results show improvement of flood predictions via particle filtering. For uncertainty assessment, parameters in both radar rainfall estimates and hydrologic models are estimated using kernel smoothing and a lag-time window via particle filtering. Results show that the proposed DA method can be used as a framework to estimate parameters and their predictive uncertainty in an integrative way.