A Bayesian-SWMM coupled stochastic model developed to reconstruct the complete profile of an unknown discharging incidence in sewer networks

Unknown illicit discharges from manufactories often contain toxic chemical matters that are detrimental to the receiving waterbody by deteriorating the performance of wastewater treatment plants. Numerical models that identify these sources and reconstruct the discharging profiles are highly desired for environment management purpose. In this study, a stochastic source identification model that couples Bayesian inference with SWMM is developed to reconstruct the profile of an instantaneous dumpling incidence in sewer networks. The unknown source parameters include location, dumping rate and time of the dumping incidence. Key factors that impact the convergence and performance of the model including walking step size, numbers of unknown source parameters and numbers of monitoring sites are investigated. Results show that the Bayesian-SWMM coupled model is effective and accurate in identifying the unknown sources parameters in an instantaneous dumping event. It is also found that walking step size is crucial for the results to converge to true solutions. Furthermore, it shows that the identified results are highly dependent on the numbers of unknown source parameters. More unknowns result to unsatisfying results. However, the study shows that this limitation could be significantly reduced by using more monitoring site data. One contribution of the study is that errors from measurements and numerical simulation are considered in the identification while results are presented in probabilities with all possible values revealed. This feature is highly practical and efficient when it comes to assist further field screening efforts to pinpoint the true sources.