Study On Inversion Of Water Quality Parameters Of Surface Water Source Based On Back Propagation Neural Network Technology

As the main water supply source of Hunchun,the surface water source of Hunchun Laolongkou reservoir has special geographical location and great social impact.Its water quality has an important impact on drinking water safety of Hunchun.Taking the surface water source of Laolongkou reservoir in Hunchun as the research area,taking the measured water quality parameters in September 2019 and the Landsat 8 image data of that day as the basic data,this paper establishes a BP neural network model for water quality parameter inversion,applies the model to the long-time series water quality parameter inversion,and analyzes the temporal and spatial distribution characteristics of water quality parameters and eutrophication.The conclusions are as follows:(1)The correlation analysis between the measured contents of TN,TP and Chla and band 1-7,NDWI and MNDWI of Landsat 8 OLI remote sensing image shows that the input factors of TN model are B1,B2 and NDWI;The input factors of TP model are B5,B6,B7 and MNDWI;The input factors of Chl-a model are B5,B6 and B7.(2)The BP neural network model is established by using the single band and band combination with the highest correlation with water quality parameters,and the accuracy is tested.The correlation coefficient between the estimated value and the measured value of the validation test in the TN inversion model is 0.7106;In the TP inversion model,the correlation coefficient between the estimated value of the validation test and the measured value is 0.7509;The correlation coefficient between the estimated value and the measured value of the validation test of Chl-a inversion model is 0.985,which shows that the estimation results of the three models have good fitting accuracy with the measured value,and can be applied to the inversion of water quality parameters of surface water source of Laolongkou reservoir in Hunchun.(3)Using the established water quality inversion model,the water quality parameters of the surface water source of Hunchun Laolongkou reservoir in 2013,2015,2017 and 2019 are inversed,and the temporal and spatial variation characteristics are analyzed.It is obtained that the overall spatial distribution characteristics of TN concentration are higher in the Northeast than in the southwest,showing an increasing decreasing trend;The overall spatial distribution of TP concentration is high in the southwest and low in the northeast,showing an upward trend;The overall spatial distribution characteristics of Chl-a concentration showed that the concentration in the reservoir area was higher than that in other areas,showing a decreasing trend.(4)The water eutrophication status of the surface water source of Hunchun Laolongkou reservoir in 2013,2015,2017 and 2019 is evaluated by using the water quality parameter concentrations obtained from the inversion model.The surface water source of Laolongkou reservoir in Hunchun is in mesotrophic state.The results of principal component analysis show that TN and Chl-a are the main influencing factors,and TP is the secondary factor.Based on the long-time series data and the whole water source,this paper establishes the BP neural network water quality parameter inversion model,and discusses the temporal and spatial distribution characteristics and main influencing factors of the water quality of the surface water source of Hunchun Laolongkou reservoir,which has a certain reference significance for the water quality inversion research,protection and management of other water sources.