Study On Remote Sensing Inversion Model Of Water Quality In Qinghe Reservoir Based On Spectral Characteristics

Qinghe Reservoir is one of the seven most famous reservoirs in liaoning province,and it was officially designated as one of the provincial reserve water sources in 2010.During the"12th Five-Year Plan"period of our country,the annual mean value of total nitrogen monitored exceeded seriously the allowing standard,and the annual mean value of permanganate index was also close to the minimum limit value.At the same time,the comprehensive nutritional index came close to the moderate nutritional level,which was at the potential risk of eutrophication.Generally speaking,conventional fixed-point sampling used in water quality monitoring not only costs a lot of manpower,material and financial resources,but also can not reflect the overall water quality of the whole reservoir.Hence,the real-time monitoring,evaluation and management for the water quality is indispensable before and after the reservoir was constructed.Remote sensing technology has the advantages of wide monitoring area,low cost and strong dynamics,which could provides important support for real-time modern water quality monitoring of Qinghe Reservoir.Researchers at home and abroad have obtained a lot of research results by using remote sensing methods to invert water quality.However,how to adopt appropriate remote sensing data and inversion methods through limited samples is still a difficult point in current research.In this study,the water quality parameters of Qinghe Reservoir were taken as the research object.Through analysis of the measured hyperspectral data and interpreting the environmental satellite data,the reflectivity spectral characteristics of typical inland water bodies in Qinghe Reservoir were revealed,and a model suitable for water quality inversion of Qinghe Reservoir was constructed.The main research contents are as follows:(1)The establishment of the high spectral data model of the Qinghe ReservoirIn this paper,the spectral data was obtained using the SVCHR-1024 high-spectrometer above the water surface measurement method.Based on the spectral response of the band,a high-spectral remote sensing model of the water quality parameters of the Qinghe Reservoir was established using sensitive bands and combinations.The unary quadratic model established by normalizing the spectrum at 762.4 nm is the best model for retrieving chlorophyll-a from the Qinghe Reservoir,but the maximum coefficient of determination is relatively low at 0.554.The combination of wave bands improves the correlation with transparency and suspended matter,and the unary quadratic models constructed by selecting the band combination corresponding to the maximum correlation coefficient can realize the daily monitoring and inversion of the transparency and suspended matter of Qinghe Reservoir.The unary quadratic models of permanganate were constituted based on the first-order differential index at 506 nm and the combination of bands at 622.8 nm and 628.4 nm,which was the best model with a fitting degree of R~2up to 0.837 for reverseing the permanganate index.Because the total nitrogen and total phosphorus have no significant optical properties,the binary quadric polynomial of total nitrogen and quaternion quadric polynomial of total phosphorus are finally constructed through the analysis of the first-order differential and wave band combination inversion models,which obtained the best fitting effect with R~2up to 0.954 and 0.966,respectively.(2)The Establishment of the classical regression inversion model of the environmental satellite dataIn order to make the sensitive band of multispectral image data more directional in Qinghe Reservoir,and at the same time to provide a reference for the selection of water quality parameter bands for satellite remote sensing data retrieval in the future,the measured spectral data was resampled to the corresponding band of the environmental satellite sensor to construct the water quality Parametric resampling inversion model,The correlation coefficient between suspended matter and total nitrogen was not significant at the level of 0.05,which failed to establish the model.The fitting degree R~2of the permanganate index inversion model was 0.65 at the maximum.Therefore,whether the model can be directly applied to the inversion of water quality of Qinghe Reservoir after resampling needs further research.The paper focuses on the analysis of the single-band data of the pre-processed environmental satellite and the combination of 85 common band combinations with the chlorophyll-a,the transparency,the total nitrogen,the total phosphorus,the permanganate index and the suspended matter water quality parameters,and the optimal wave band and its combination were selected for water quality inversion of Qinghe Reservoir.Field water quality sampling data in summer,autumn of 2015 and spring of 2016 were used to establish water quality parameter inversion models of different seasons with classical regression method.(3)Establishment of inversion model based on least squares methodConsidering the multiple correlation between water quality parameters and wave bands,a partial least squares regression estimation model was established,and the accuracy of the model inversion is higher than the single-band and band-combined models;Considering that there was no physical correlation and nonlinearity between water quality parameters and satellite wave band,the model of least square support vector machine was established.The results show that in summer and spring,the band combination has a better inversion effect on total nitrogen,suspended matter and chlorophyll-a,and the partial least squares model is superior to the inversion of total phosphorus,permanganate index and transparency band combination model,single band model and least squares support vector machine model;In autumn,the inversion precision of partial least squares regression model is better than the above three models.Base on the above research,the multi-band model and the partial least squares regression models of the six water quality parameters,which were established in this paper can be directly applied to the daily water quality monitoring of Qinghe Reservoir through the verification and improvement of a large number of data.