Research And Application Of Parameter Estimation Method In Hydrological Model Based On Ensemble Kalman Filter

The occurrence of flood causes great harm to human production and life.Flood control is a major problem that people are facing because of the frequent occurrence of flood in our country.With the development of science and technology,flood forecasting technology has become an important means of flood control,and hydrological model is the main tool for flood forecasting.To simulate streamflow with high precision,accurate hydrological model parameters must be obtained.Although field measurements and using long hydrological data series for parameter calibration provide a reliable method for obtaining these parameters,changes in the underlying basin surface and lack of hydrometeorological data may affect parameter accuracy.The realtime parameter estimation method is helpful to solve the above problems.Previous studies have shown that ensemble Kalman filter has higher accuracy than other commonly used parameter estimation methods such as least square method and variational method.In this study,the most commonly used improved form of Kalman filter in hydrological model at this stage,the dual ensemble Kalman filter,is selected for parameter estimation of hydrological model.In this paper,based on the results of global sensitivity analysis of parameters,five representative parameters of Xin'anjiang model are selected to analyze the characteristic of parameter estimation of ensemble Kalman filter in hydrological model,At the same time,the feasibility and practicability of ensemble Kalman filter in hydrological model are evaluated.The research contents include: In the case of single parameter and multi parameter,the effects of distribution of the initial ensemble of parameters and the different functions of parameters in the model on the parameter estimation process are studied;In the case of multi parameter,the influence of the interaction between parameters on the parameter estimation process of different parameters is studied;using the method of numerical simulation,whether the filtering algorithm can find the parameter value correctly under different conditions is used to judge its feasibility in the hydrological model;Whether it can improve the forecast effect of Jinjiang River basin in different scenarios after parameter estimation is used to judge its practicability in hydrological model.At present,there are few researches on the practicability of ensemble Kalman filter in distributed hydrological model.Based on the research results of Xin'anjiang model and the global sensitivity analysis of parameters,this paper selects three parameters of Wetspa model,and studies its practicability in the hydrological model by setting different forecast scenarios in Jinjiang River basin,China.The numerical simulation results in the case of single parameter estimation show that the efficiency of parameter estimation can be improved by determining the initial distribution mean through parameter calibration,and the most favorable initial distribution variance for the parameter ensemble stabilization depends on the difference between the ensemble mean and the true value of the parameter,and the phenomenon that parameters at a certain time not only affect the traffic at that time,but also affect the traffic at other times has a negative impact on the parameter estimation process.The results of numerical simulation in the case of multi parameter estimation show that there is mutual interference between the parameters in the estimation process.The degree of interference is related to the sensitivity of the parameters to be estimated and the time required for the single parameter estimation to reach the stable state,and some conclusions in the case of single parameter estimation are no longer applicable.The results of numerical simulation in the case of multi parameter estimation show that there is mutual interference between the parameters in the estimation process.The degree of interference is related to the sensitivity of the parameters to be estimated and stabilization time step for single-parameter estimation,and some conclusions in the case of single parameter estimation are no longer applicable.The results of numerical simulation also show that the ensemble Kalman filter method is feasible for hydrological model.The application of Xin'anjiang model and Wetspa model in Jinjiang River Basin shows that the ensemble Kalman filter is always practical in hydrological model,and it can improve the accuracy of flood prediction in the basin where the accuracy of streamflow simulation decreases due to the change of underlying surface or the lack of hydrological and meteorological data.