Research On The Inversion Of Water Quality Parameters Of Dianchi Lake Based On Landsat8 OLI Image

Dianchi is not only the largest shallow lake on the Yunnan-Guizhou plateau,which is important for maintaining the ecological environment of Kunming and the surrounding areas,but also a giant water body storing more than 1.5 billion cubic meters of five types of highly eutrophic poor-quality water.The pollution and eutrophication management of Dianchi will be a long-term arduous task,and the catastrophic spread of its highly eutrophic poor-quality water will have significant ecological and environmental impacts.Usually,routine spot sampling of water quality in monitoring is not only time-consuming and laborious,but also cannot provide comprehensive monitoring of water quality conditions.Therefore,it is necessary to improve the realtime assessment and monitoring of water quality.Remote sensing technology has the characteristics of real-time,large scale and low cost.It compensates for the lack of traditional peak sampling and has great potential to provide more specific data for real-time monitoring of Dianchi water quality.This paper takes Dianchi waters as the study area,and through the correlation analysis between the measured water quality parameters and remote sensing image data of Dianchi from 2013-2021,the optimal single band and band combinations of different water quality parameters are derived,based on which regression analysis and neural network analysis are conducted for permanganate,dissolved oxygen,acidity and ammonia nitrogen,to select the most appropriate inversion model for the water quality parameters.Using the inversion model for the water quality parameters,the water quality of Dianchi was evaluated.The results of this work are as follows:(1)Regression analysis was conducted for permanganate,dissolved oxygen,p H and ammonia nitrogen,and inverse models such as one-dimensional linear,polynomial function and exponential function were established.The single band with the highest correlation with permanganate was B5 with a coefficient of 0.85,which satisfied the linear relationship;the best combination was B5+B2 with a coefficient of 0.86,which satisfied the linear relationship;the single band with the highest correlation with dissolved oxygen content was B7 with a coefficient of 0.80,which satisfied the cubic polynomial relationship,and the best combination band was(B7-B1)/(B7+B1).83,which satisfies the cubic polynomial relationship;the highest correlation with p H content is B3 with a coefficient of0.76,which satisfies the logarithmic relationship;the best band combination is B3+B4 with a coefficient of 0.78,which satisfies the logarithmic relationship;the highest correlation with ammonia nitrogen content is B6 with a coefficient of 0.75,which satisfies the linear relationship;the best band combination is B6+B7.The best band combination is B6+B7with the coefficient is 0.77,which satisfies the logarithmic relationship.The accuracy of the single-band water quality inversion model and the combined water quality inversion model was analyzed,and the overall accuracy of the combined inversion model was found to be higher.(2)The use of correlation optimal single-band and band combination to establish a neural network water quality parameter model for each water quality parameter,and accuracy testing,the R,MRE,RMSE between the estimated and measured values of the validation sample in the permanganate inversion model are 0.97,0.01,0.21;dissolved oxygen inversion model between the estimated and measured values of the validation sample R,MRE,R,MRE,and RMSE between the estimated and measured values of the validation samples in the inversion model for dissolved oxygen were 0.92,0.02,and 0.17,respectively;R,MRE,and RMSE between the estimated and measured values of the validation samples in the inversion model for acidity were 0.95,0.02,and 0.23,respectively;R,MRE,and RMSE between the estimated and measured values of the validation samples in the inversion model for ammonia nitrogen were The R,MRE and RMSE of the inversion model for ammonia nitrogen were 0.89,0.19 and 0.046,respectively;except for the poor inversion results of ammonia nitrogen model,the estimation results of the other three models fit well with the measured values and can be used in the inversion study of water quality parameters in Dianchi water body.(3)The water quality parameters of Dianchi water bodies in 2013,2018 and 2021 were inverted using neural network water quality parameter inversion model and regression analysis water quality parameter inversion model,and the images of 2015 and2020,which were not involved in the experiment,were used to verify the spatial and temporal variation characteristics of Dianchi and the most applicable water quality parameter inversion model.It is concluded from the analysis that the permanganate concentration and acidity concentration exceeded the water quality standard of Dianchi,which is one of the main reasons for the water quality of Dianchi.The neural network water quality parameter inversion model and regression analysis water quality parameter inversion model are not applicable to the ammonia nitrogen in Dianchi water body.Neural network water quality parameter inversion model and regression analysis water quality parameter inversion model can invert permanganate,dissolved oxygen,acidity and alkalinity very well,and the neural network effect is better.The experiments show that,except for ammonia nitrogen,the multiple regression model and neural network have certain guiding significance for water quality inversion of Dianchi,and have certain reference significance for water quality inversion research of other water sources as well as protection and management.