Simulation And Uncertainty Analysis Of Non-point Source Pollution In Yitong River Watershed

With control of the point source pollution, the non-point source pollution has become the main problem of water pollution and the focus of the society. Due to the non-point source pollution has the characters, such as concealment, randomization etc, it is difficult to carry out monitoring and management. The non-point source pollution model is one of the major method to study non-point source pollution, and it is widely used by researchers. However the non-point source pollution model has many parameters, the parameter uncertainty analysis become the focus of research on non-point source pollutionYitong River is an important tributary of the second Songhua River, and it is significant for residents. Also, Yitong River Basin is main grain producing area. A large area of deforestation results in serious soil erosion in the basin for years, and a large number of pesticides and nutrients with surface runoff merge in the Yitong River. Non-point source pollution becomes a threat to the water environment of Yitong River.Studying in the Yitong River Basin, we set up non-point source pollution SWAT model based on spatial and attribute database of the study area and simulate non-point source pollutants in sediment, total nitrogen, total phosphorus load. We identify the key area of non-point source pollutants. Non-point source pollution control measures are proposed for key area, and simulate the effect of non-point pollution control measures. Based on the model of radial basis function neural network taking the place of simulation model, Using Monte- Carlo simulation method analyses the uncertainty of the model parameters. The research results show that :(1) the average annual output of sediment load is 308.151×104t, the average annual output of total nitrogen load is 1.516×104t, and the average total phosphorus load is 0.357×104t.(2) Non-point source pollutants,sediment, total phosphorus and total nitrogen load, are similar distribution in space, and the key pollution area are concentrated in the upstream region of Yitong River Basin in the south.(3) Two kinds of non-point source pollution control measures are used in study area, and they are returning cropland to forest and the setting vegetation buffer zone. The results show that there is a significant reduction of nonpoint source pollution in the study area.(4) The results of using radial basis function neural network model is fitted well with the SWAT mode, and it can be instead of the SWAT model for simulation.(5) The results of parameter uncertainty analysis indicate that, when the confidence level is 90% and 80%, the confidence interval of sediment load output are 82.08-540.71×104t and 149.25-473.55×104t, the confidence interval of sediment load output are 0.49-2.62×104t and 0.80-2.31×104t, the confidence interval of sediment load output are 0.17-0.55×104t and 0.22-0.49×104t; The confidence interval of sediment reduction of returning cropland to forest are 55.95%-60.44% and 57.19%-59.20%, total nitrogen reduction are 27.67%-31.20% and 28.51%-30.37%, and total phosphorus reduction are 27.22%-30.67% and 28.30%-29.50%;The confidence interval of sediment reduction of setting vegetation buffer zone are 30.88%-33.72% and 31.15%-32.45%, total nitrogen reduction are 21.27%-24.20% and 22.72%-23.74%, and total phosphorus reduction are 21.54%-24.10% and 22.52%-23.22%;(6) The results of parameter uncertainty analysis indicate that, total nitrogen load has the maximum uncertainty, followed by total phosphorus and sediment; Reduction of non-point source pollution control measures in total nitrogen has the maximum uncertainty, followed by total phosphorus, and sediment is the smallest. Non-point source reduction of setting vegetation buffer zone is more influenced by parameter uncertainty than that of returning cropland to forest.