The Research On SWAT Model Coupled With Ensemble Precipitation Forecast And Its Application In Flood Forecasting

The flood disaster seriously restricts the economic and social development in the occurrence area.Timely and accurately flood forecasting can earn valuable time for flood prevention and disaster reduction.Precipitation prediction error is an important source of uncertainty in flood forecasting.In order to improve the reliability of flood forecasting,ensemble precipitation forecasting technology is introduced in this thesis.And the precipitation ensemble forecasting data is used to drive the SWAT model for flood forecasting in Xijiang River.The main research results are as follows:(1)The SWAT model which based on the historical rainfall data of the Xijiang River is built to analyze the forecast effect.When the model is built,the local sensitivity analysis method and the global sensitivity analysis method which based on the LH-OAT algorithm are used to analyze the sensitivity of the parameter.And the SWAT model is calibrated and validated by using the SUFI-2 algorithm.The simulation results show that the average qualification rate of the floods is 75%,and the accuracy of simulation results reaches grade B.(2)The rainfall data from the three ensemble systems which are National Centers for Environmental Prediction(NCEP),European Center for Medium-Range Weather Forecasts(ECMWF),and China Meteorological Administration(CMA)are used as the basic data source for ensemble precipitation forecast.And the basic data as the input data of the SWAT model are tested by different reliability test methods.The dispersion analysis method are used to compare the probabilistic forecasting effect and the control forecasting effect in the ensemble systems.The results show that the probabilistic forecasting result is more advantageous.The four test methods,such as Relative Operating Characteristic(ROC),the Threat Score(TS),the Brier Score(BS)and Talagrand distribution diagram,are used to test the reliability of the forecast rainfall data.The results show that the forecast rainfall data for all ensemble members and the average forecast rainfall data in the three ensemble systems have certain forecasting skills.The forecast rainfall data can better reflect the actual rainfall process.The data can be used as the input data to SWAT model for flood forecasting.In addition,the results also show that the reliability of ensemble average rainfall data is better than the reliability of individual ensemble member.(3)Ensemble precipitation forecast data are used as the input data of the SWAT model to analyze the forecast effect.The ensemble average method,multi-mode ensemble average method,and multi-mode factor analysis method are used to modify the basic data.The modified precipitation data are used to drive the SWAT model for flood forecasting.The simulation results of the three correction methods are better than the result which based on historical rainfall data.In addition,the average qualification rate of the multi-mode factor analysis correction method is 91.7%,and the accuracy of simulation results reaches grade A.The average qualification rate of the multi-model ensemble average method is 87.5%,and the accuracy of simulation results reaches grade A.The average qualification rates of ECMWF,NCEP and CMA using the ensemble average method are 83.2%,79.2%,79.2%,and the accuracy of simulation results all reaches grade B.