Using An Ensemble Kalman Filter Method To Calibrate Parameters Of A Prediction Model For Chemical Transport From Soil To Surface Runoff

With the development of society,agricultural non-point pollution has more and more influence on the environment.Agricultural non-point pollution is mainly caused by excessive use of fertilizers and loss.The erosion of soil by rainfall can cause large amounts of chemicals transport from soil to the surface runoff and flow into rivers,lakes and other water bodies.Therefore,it is of great significance to study the process of chemical transport from soil to surface runoff.Many scholars have modeled the process of chemical transport from soil to the surface runoff,mainly developing the mixed layer theory and diffusion theory.Gao et al.(2004)proposed a model of chemical transport from soil to the surface runoff,which combined the process of chemical transport driven by raindrops and the diffusion of the chemical in the soil layer.This model of chemical transport from soil to the surface runoff is a numerical advection-diffusion-equation(ADE)model,and this research is based on it.The ensemble Kalman filter(EnKF)is a data assimilation method,which is easy to be combined with existing models,and has been widely used in the field of hydrology.In this paper,EnKF method is applied to the model proposed by Gao et al.(2004)to improve the prediction accuracy and inversion parameters.This is the first time that EnKF method is applied to the model of chemical transport from soil to surface runoff based on ADE.Data assimilation results under the static water transfer rate and the dynamic water transfer rate are considered and analyzed respectively,and it shows that the updated solute concentration under dynamic water transfer rate is better fitted to the observations.In order to further study the inversion of model parameters,observations were generated by an ideal model of static water transfer rate.Then,EnKF method is used to carry out parameter inversion for a model based on poor initial estimation parameters.The parameters updating situation was analyzed,and six potential factors affecting the performance of EnKF method were studied.The following conclusions were drawn:(1)considering the calculation cost and prediction accuracy,the ensemble size of 300 is most suitable for the study;(2)within a wide range of initial parameter estimation errors,EnKF is effective for updating parameters and improving model prediction accuracy;(3)key moment observation can provide more information;(4)EnKF can be applied to different permeable boundary conditions;(5)EnKF is suitable to update multiple parameters(E_r,d_e and h_w);(6)when the observation error and the parameter error which used to generate the initial ensemble both increase or decrease at the same multiple under standard conditions.There are still some deficiencies in the application of EnKF to the model of chemical transport from soil to the surface runoff,such as the problem of filtering divergence in the inversion of multiple parameters.The application of EnKF method to the real situations still need further study.