There lies much uncertainty in the environmental hydraulics system, such as the uncertainty of the measurement data, therefore the kind of pollution source identification problem is ill-posed, especially the non-unique. The classic regularization method and the optimization method can only get the “point estimation” of the parameter, so it is hard for them to solve the problem with more uncertainty. Since the water quality model is coupled with the flow field equation(Navier-Stokes equation), the direct problem is much nonliear. In order to settle the above difficulties, for the hydrodynamics-water quality coupled model, a pollution point source identification model is advanced based on Bayesian inference. Markov chain Monte Carlo sampling method is used to get the posterior probabilty distribution of the source’s position and intensity, thus solving the uncertainty and the nonlinearity well. Computational case’s result indicates the Bayesian inference with MCMC sampling can describe and solve the pollution source identification inverse problem for the hydrodynamics-water quality coupled model better.
