Construction And Cases Application Of Source Inversion Model Of Sudden River Water Pollution Based On Bayesian Inference

The problem of water pollution in our country is still serious.How to effectively supervise the river water quality and pollution discharge has become a new problem for the relevant management departments.Therefore,it is of great significance to develop a new and effective monitoring,early warning and source inversion technology of river water pollution and to establish an information system for scientific and effective management of river water environment.In this context,this dissertation studies the quantitative analysis method of source parameter uncertainty in the process of water pollution source inversion based on Bayesian inference,taking the problem of sudden river water pollution source inversion as the starting point.In the aspect of model construction,firstly,the research "localizes" the Bayesian inference formula with the environmental language in the field of water pollution source inversion,sets the parameters of source term as the uniform distribution,establishes the likelihood function according to the different distribution forms that the error term obeys,introduces the AM-MCMC sampling method to obtain the target posterior distribution of the parameters of source term,and completes the construction of source inversion model.Then a performance evaluation system matching the model is established.Finally,according to the specific evaluation indexes,the key parameters of the model are calibrated,analyzed and determined,and the model whose error term is subject to the assumption of uncorrelated homoscedastic distribution is used to infer the source.In addition,three improvement schemes,fixed step length,memory matrix and mixed update method,are put forward for the covariance matrix updating mode in the posterior distribution sampling.The calculation time can be saved by 15 % to 93 % for different discharge types,and the calculation burden can be significantly reduced while satisfying the results.In the aspect of system development,according to the existing research results,design and complete the Source Inversion Management System of Sudden River Water Pollution with the core function modules of data collection,data filtering,pollution source inversion,result visualization,resulting data recording and system management.The system is simple,practical and easy to operate,which is convenient for environmental management departments and relevant personnel in the field of water pollution source inversion to carry out source inversion research.The development of the system has positive significance for the application of the research results and the expansion of the audience of the water pollution source inversion model based on Bayesian inference.In the aspect of cases application,based on the established Source Inversion Management System of Sudden River Water Pollution,a series of fake source inversion case analyses are carried out for three common pollutant discharge types: instantaneous,continuous and intermittent,and the Truckee River with instantaneous point source characteristics and the West Hobolochto Creek with intermittent(N_c=1)point source characteristics are used respectively as river tracer experiments to verify the model and system.Overall,the accuracy of the model can be within 10 %,and the results of source inversion analyses and forward concentration simulation are in an acceptable range.The source inversion model and management system of sudden river water pollution constructed in this dissertation can describe the uncertainty of parameters of source term quantitatively,and reconstruct the history of pollutant discharge accurately,which can provide strong technical support and enhance the decision-making confidence for relevant water environment managers and decision-makers.The established technical method is universal and has the potential to be popularized in other regions.