Research On The Inverse Problems Of Sudden Water Pollution Source Identification In Rivers

Continuously accelerated urbanization and booming industrialization have also brought us increasingly frequent sudden water pollution.Once the pollution incident occurs,the first task is to quickly locate the pollution source,thus to reconstruct the release history and transport trajectory and to predict the spread process of pollutants.Focusing on the pollution source identification,this research develops a set of numerical models to solve the inverse problems in relation to upstream inflow,pollution source,tributary inflow and model parameter,using the adjoint and data assimilation methods.The research systematically inveatigates the primary inverse problems of sudden water pollution source identification in rivers.And the main contents and results are summarized as follows:(1)The proposed inverse flow and water quality models formulated on the rotated x-t plane can address the instability issue caused by inverse iteration in time and space.After an ideal and then a natural river case study,the results demonstrate the effectiveness and stability of our proposed model,with the simulated inflow highly fitted to the observations.(2)With the adjoint method which calculates pollutant concentration by solving the advection-diffusion adjoint equation,the inverse problem of pollution source identification is converted to linear regression analysis and data assimilation,respectively.Then two efficient models for pollution source identification are proposed,the LR-BLP corresponding to linear regression and the BLP-EnKF corresponding to data assimilation.Two natural river cases,i.e.,the Ganjiang River(its tributary)and Xijiang River(Xunjiang reach),are selected for numerical experiments.And the influences from number of pollution source,flow state and observation error are investigated.The results show that:both LR-BLP and BLP-EnKF present great advantages in efficiency,and their computational time is much smaller than that of previous models;the LR-BLP achieves higher accuracy and stability,but relys more on the observation data,while BLP-EnKF supports online analysis,performs more efficiently and obtains more reliable results when observation error is large,but presents worse applicability to the multi-point case;in unsteady flow case,both LR-BLP and BLP-EnKF can obtain source identification results with accuracy reaching to 5%,satisfying the requirements in practice.(3)In consideration of the tributary pollution,we propose another inverse model for inversely simulate the interval inflow and pollutant discharge,on the basis of flow and water quality models and data assimilation with EnKF.The model is also applied in an ideal and then a natural river case,where its effectiviness and efficiency are demonstrated.This model can obtain the interval inflow and pollutant discharge with high accuracy as the error indicator R~2and NSE are larger than 0.9.This part of research extends the scope of pollution source identifictaion to the tributary situations.(4)When model parameters are unknown,the EnKF are used for parameter identification and correction,after the uncertainty partition method for sensitivity analysis,thereby solving the inverse parameter problem in pollution source identification.The natural river case of Xijiang River is used again for model validation,and the estimation errors of roughness and diffusion coefficient in this case are smaller than 3%.In addition,the model results are sensitive to roughness error while the sensitivity of diffusion coefficient is relatively low.This research provides a relatively complete set of inverse models of pollution source identification for the emergency treatment of sudden water pollution,including the numerical models for both forward and reverse flow and water quality simulation,two efficient pollution source identification models applicable to those cases with offline or online data,single-point or multi-point sources,steady or unsteady flow,and the EnKF method to inversely estimate interval pollution discharge and model parameters.The achievements in this research can be directly used in the emergency treatments of sudden water pollution in rivers.