High resolution distributed hydrological modeling for river flood forecasting

Abstract:

High resolution distributed hydrological model can finely describe the river basin hydrological processes, thus having the potential to improve the flood forecasting capabilities, and is regarded as the next generation flood forecast model. But there are great challenges in deploying it in real-time river flood forecasting, such as the awesome computation resources requirement, parameter determination, high resolution precipitation assimilation and uncertainty controls. Liuxihe Model is a physically-based distributed hydrological model proposed mainly for catchment flood forecasting, which is a process-based hydrological model. In this study, based on Liuxihe Model, a parallel computation algorithm for Liuxihe model flood forecasting is proposed, and a cloudy computation system is developed on a high performance computer, this largely improves the applicability of Liuxihe Model in large river. Without the parallel computation, the Liuxihe Model is computationally incapable in application to rivers with drainage area bigger than 10,000km² at the grid size of 100m. With the parallel computation, the Liuxihe Model is used in a river with a drainage area of 60,000km², and could be expended indefinitely. Based on this achievement, a model parameter calibration method by using Particle Swale Optimization is proposed and tested in several rivers in southern China with drainage areas ranging from several hundreds to tens thousands km², and with the model parameter optimization, the model performance has been approved largely. The modeling approach is also tested for coupling radar-based precipitation estimation/prediction for small catchment flash forecasting and for coupling quantitative precipitation estimation/prediction from meteorological model for large river flood forecasting.