Research On Dynamic Assessment Method Of Non-point Source Pollution Risk In Watershed

Panjiakou reservoir,as a major water source "Yin Luan Ru Jin" project,in recent years due to the upstream agriculture,urban non-point source pollution,lead to the reservoir water quality deterioration,has seriously affected the reservoir water quality,thus to carry out the panjiakou reservoir basin research on water pollution prevention and control of agricultural non-point source pollution is of important theoretical and practical significance.Non-point source pollution has the characteristics of wide source,random spread and uncertainty,which makes it difficult to monitor it.At present,model simulation is one of the main ways to study non-point source pollution.Non-point source pollution output risk research at home and abroad is given priority to with output coefficient model,but most of the study did not consider years rainfall spatial and temporal heterogeneity and the influence of topography on non-point source pollution,therefore,this article introduced rainfall factor and terrain factor to improve the output coefficient model,year nonpoint source pollution output dynamic risk assessment method is put forward.However,the existing spatial resolution and land use classification accuracy of precipitation products cannot meet the requirements of fine non-point source pollution risk assessment model construction.Therefore,in this paper,precipitation downscaling model was firstly constructed based on satellite inversion precipitation products(GPM)to obtain precipitation data with spatial resolution.Then,the land use classification method was studied based on Gaofen-6 satellite,and the refined land use data were obtained.Finally,based on the downscaling precipitation data and high-precision land use data,a dynamic assessment model of non-point source pollution output risk was constructed,and the model was verified in the Panjiakou Reservoir watershed.The main conclusions are as follows:(1)The downscaling research results of precipitation show that the original GPM precipitation products have a relatively high estimation of precipitation in the study area on different time scales.And the use of PSO and BP model after downscaling precipitation data space distribution is consistent with the lower dimensions of data,and in a month time scale,after downscaling of rainfall data and the measured rainfall correlation coefficients of above0.8,the relative deviation control within 10%,space scale and accuracy can meet the demand of non-point source pollution output risk assessment study.(2)Land use in the Panjiakou Reservoir watershed was extracted by using multiple classifier combination method,support vector machine and maximum likelihood method respectively.The results showed that the overall classification accuracy was 94.54%,85.25%and 79.89% respectively,and the Kappa coefficients were 0.91,0.79 and 0.72,respectively.Through comparison,it is found that the spatial distribution of each land use type in the classification results of the three classifiers is relatively consistent.However,in terms of details,the multi-classifier combination method has better extraction effect and higher accuracy,which can be used in the risk assessment research of non-point source pollution output.(3)Considering the influence of precipitation and terrain on the formation of non-point source pollution,the study introduced precipitation and terrain factors to improve the classical output coefficient model,and established a dynamic assessment model for the output risk of non-point source pollution.The Panjiakou Reservoir Basin was taken as the study area to verify the model.The simulation results show that the output risk of nitrogen pollution is higher than that of phosphorus pollution on the whole.In terms of time,the risk of non-point source pollution output is higher in flood season,especially in July and August.Spatially,the risk of non-point source pollution output in the upper and lower reaches of Panjiakou watershed is higher,while the risk in the middle reaches is lower.Compared with the existing research results of non-point source pollution load estimation in Panjiakou Basin,the model constructed in this study can accurately estimate the dynamic output risk of non-point source pollution in the basin.