Study And Application Of Data Assimilation Method For Nitrate Transport

Recently,serious non-point source contaminant in the world poses a threat to ecological environment and human health.In developed countries such as Europe and US,groundwater pollution caused by agriculture is prevalent.In order to management agricultural water resources and water environment,scholars have established many models to characterize transport and transformation of nitrate in groundwater based on different theories.Generally,it requires a large number of parameters(dispersion,maximum reaction rate,half saturation constant,and temperature coefficient)that represent transport and transformation properties of nitrate when running those models.However,there exists strong temporal and spatial variability in the biochemical reaction parameters for reactive solutes,and it is nontrivial to obtain accurate estimation on concentration and parameters.Thus,it is necessary for modelers to estimate parameters by limited historical data or select the parameters based on experience,and apply to model without considering temporal and spatial variability of parameters.Therefore,parameters involved in the numerical simulation usually contains strong uncertainties,which makes parameters cannot accurately reflect the migration and transformation characteristics of solutes in groundwater.Simultaneously,transport and transformation model cannot integrate observed data into numerical model to improve model prediction.It is of great significance to introduce new numerical techniques into transport and transformation modeling for more effective parameter estimation and model prediction.This work reviews inversion algorithms widely used in groundwater,atmospheric prediction,ocean forecasting and oil reservoir inversion,such as Markov chain Monte Carlo algorithm(MCMC),ensemble Kalman filter algorithm(EnKF)and the nonlinear regression algorithm.Then models applied to simulate transport and transformation of solutes in groundwater are summarized.On this basis,migration and transformation process of nitrate nitrogen is firstly described by a coupling hydrodynamic and biochemical reaction kinetic model.In this paper,performance of MCMC,EnKF and nonlinear regression are analyzed in detail,and factors which may affect parameter estimation are discussed.Then,EnKF is applied to complex biochemical reaction system to.Finally,assimilation model is tested by soil column experiment.The main contents and conclusions are as follows:(1)The migration and transformation of nitrate nitrogen in groundwater are described by coupling hydrodynamic model(HYDRUS)and chemical reaction kinetic model(PHREEQC).Temporal and spatial distribution of nitrate nitrogen is predicted.(2)Establish a data assimilation model based on the above coupled model.Observed nitrate concentration data are introduced into the one-dimensional saturated-flow nitrate transformation and transformation data assimilation scheme.Six cases are designed to verify performance of parameter estimation for the afore-mentioned three inversion algorithms,and to study factors influencing parameter estimation.Numerical results show that MCMC algorithm gives good estimation of parameters,but the calculation cost is high.Sensitivity,dimension and correlation of parameters affect the performance of MCMC algorithm.Nonlinear regression algorithm is a local optimization algorithm whose parametric estimation performance is affected by starting point,range and sensitivity of parameters.We consider this method only if global optimal solution is not required and parameter value needs to be obtained quickly.EnKF algorithm is cost-friendly,and presents good estimation of parameters.Simultaneously,it weights uncertainly of prior variance and observation error.(3)In view of complex biochemical reaction,the influence of dispersion,observation type,number of observations and location on data assimilation is considered firstly.The numerical study shows that the observation type has a great influence on the complex biochemical reaction.Combing different types of observations can achieve better assimilation results,which is useful for monitoring network design.(4)Carry out soil column experiment in laboratory to test performance of EnKF algorithm in practice.Result shows that EnKF significantly improves estimation of nitrate parameter,and prediction of spatial and temporal distribution of NN concentration by assimilating real-time observations.The existence of denitrification makes the spatial distribution of nitrate more uniform,which overestimates estimation of dispersivity.It is feasible to estimate nitrate dispersivity with bromide observations.Finally,the major works and contributions are summarized and several suggestions for future work are proposed.