Research On Uncertainty Water Quality Modeling Based On WASP Model

As water quality modeling is further explored and perfected all the time, its structure becomes more complex with an increasing number of parameters getting involved, leading to great uncertainty in modeling application. It demands immediate attention to the parameter calibration of water quality modeling through the uncertainty method and establish uncertainty water quality modeling.Taking Water Quality Analysis Simulation Program(WASP) model as its research object, the paper discussed the methodology and specific steps of parameter calibration with uncertainty methods. With Delayed Rejection Adaptive Metropolis(DRAM)as the principal algorithm and incorporated with sensitivity analysis of model parameters and the comprehensive diagnose of Markov chain convergence, parameter uncertainty analysis was carried out under the guidance of Bayesian theorem and the posterior distribution of model parameters were efficiently obtained. It realized stochastic water quality simulation on the basis of parameter uncertainty analysis and established uncertainty water quality modeling based on WASP model.DRAM algorithm was applied innovatively in the research of uncertainty water quality modeling in this paper. DRAM is able to implement highly efficient parameter searching strategy to obtain effective sampling of target distribution. The results of numerical experiments in the paper proved DRAM to be an effective way of getting posterior distribution samplings of parameters in water quality modeling. No relevant references on its application have been seen in this research field. The paper formulated a methodology for DRAM algorithm to apply WASP parameter uncertainty analysis and redeveloped the source code of WASP model to employ the methodology.A comprehensive method of parameter sensitivity analysis in WASP model was proposed in this paper. Based on a systematic analysis about WASP model structure, sensitivity analysis of single parameter and that of parameters in orthogonal test was carried out. It enhanced the observation of how the change of model outputs corresponds to that of model parameters and consequently confirmed the parameters that significantly affected water quality variation.In order to verify that the parameter samplings resulting from DRAM converge into the posterior distribution, a systematic method of convergence diagnose for Markov chain was proposed in the paper. After qualitative analysis of convergence, two kinks of convergence diagnostics were calculated, Monte Carlo standard error was also assessed. By analyzing the parameter samplings of numerical experiments, it was proved that DRAM was able to seek posterior distribution of parameters efficiently and obtain rational calibration results of parameters.Based on the parameter posterior distribution and uncertainty analysis, a stochastic simulation was performed on WASP model and the uncertainty water quality model based on WASP was established. The model output samplings underwent statistic analysis and an investigation was done on the uncertainty of the output of water quality variables in the model.