Combating the Challenges of Real-Time Flood Forecasting by Using Physically Based Distributed Hydrological Model for Large River Basin (PAP014342)

Código

PAP014342

Autores

Yangbo Chen

Tema

Flood forecasting and early warning systems

Resumo

Physically based distributed hydrological models(PBDHM), which grid the studied terrain into very fine cells, thus having the potential to better represent the hydrological processes, and to couple the high resolution radar estimated precipitation, are considered to be able to improve the flood forecast capability, are regarded as the new generation flood forecast models. But there are great challenges in river basin real-time flood forecasting, particularly if the river basin is bigger than 1000km2. The first challenge is the computation efficiency, as the terrain is divided into very fine cells, so the computation resources needed for real-time flood forecasting is awesome, and the general computer and algorithm could not be used. Secondly, how to calibrate the model parameters is also a big challenge as the PBDHMs have too many parameter, and the conventional method of parameter calibration used for lumped model could not be employed directly. Furthermore, there are also uncertainties associated with PBDHMs that need to be addressed. Liuxihe Model is a physically-based distributed hydrological model that divides the studied basin into a number of cells horizontally and three layers vertically. All cells are classified as one of the three types, including hillslope cells, river cells and reservoir cells according to their flow accumulation which have their own properties and model parameters. The saturation excess mechanism is employed to determine the surface runoff while the interflow is calculated using Campbell's equation. The runoff routing is divided into hillslope routing and river routing. In this study, a parallel algorithm is presented for Liuxihe Model simulation, and a shell computation module is coded based on a high performance computer with multi-processors, and the validation results show that it can make the real-time flood forecasting computation at minute scale for the large river basin with a drainage area of 60 thousands km2 at the cell size of 100m. Based on this achievement, a model parameter calibration method by using Particle Swale Optimization is presented and a cloud system for Liuxihe Model is developed and based on deployed at a super fast supercomper installed in the author's university, which is now a public facility for research and development. Several river basins in southern China with drainage areas ranging from several thousands to tens thousands were case studied, and the model parameter were optimized, and several real-time operational flood forecasting system are also developed and have been put into operation, which can couple with the radar estimated precipitation and quantitative precipitation forecasting. The results show that the challenges facing the PBDHMs real-time flood forecasting in large river basins could be conquered.