| Water quality affects human health and ecosystem operation,and due to the complex coupling relationship between suspended solids and water quality,the content of suspended solids in water is often used as one of the important indicators to measure water quality.Traditional sampling and monitoring methods are easily interfered by weather and other conditions and have high costs,and satellite remote sensing technology is widely used in water monitoring due to its advantages of all-weather and all-round monitoring.The correlation between the reflectivity of satellite images and the concentration of suspended solids in water is an important scientific issue in remote sensing water quality monitoring.Taking Chagan Lake as the research area,taking the measured hyperspectral data and the synchronized/quasi-synchronous Landsat 8 satellite,Sentinel-2 satellite and GF-6 satellite multispectral remote sensing data as the data source,this paper first explores the correlation between water reflectivity and suspended solids in different time phases and bands according to the spectral characteristics of suspended solids in different data sources,and determines the sensitive band for the concentration identification of suspended solids.On the basis of the sensitive band,the spectral index of suspended solids concentration was further constructed.Finally,regression analysis and neural network methods are used to study the correlation model of multi-source remote sensing data,suspended solids concentration in water at different time phases,sensitive bands,and spectral index of suspended solids concentration.The main conclusions of this paper are as follows:(1)The correlation between water reflectivity and suspended solids concentration found that there was a positive correlation between water reflectivity and suspended solids concentration,and with the increase of concentration,the reflection peak showed a tendency to move in a longer wavelength.The results of Pearson correlation analysis show that the water reflectivity of multi-source remote sensing data under multitemporal phase is correlated with the concentration of suspended solids in green light(500-560nm),red light(620-760nm)and near-infrared band(760-960nm),and the results show that the green band,red band and photographic infrared band are sensitive bands for detection of suspended solids concentration.(2)The study of suspended solids concentration spectral index finds that on the basis of band sensitivity analysis,the bands suitable for constructing suspended solids concentration spectral index are quantitatively screened through the interband confusion matrix,and the suspended solids concentration spectral index is constructed by band combination.The optimal band/spectral index differs between different remote sensing data sources,and the correlation between the spectral index and suspended solids concentration is slightly higher in some seasons than in the single band.The results show that the suspended solids concentration spectral index can be used to construct the suspended solids concentration inversion model.(3)The model study on the correlation between water reflectivity and suspended solids concentration found that the results of the models established by the three satellite remote sensing data all showed that the inversion effect of the seasonal model was better than that established in the whole season.The accuracy of the all-season inversion model increases with the decrease of image resolution,and the accuracy of the allseason inversion model of the Landsat 8 satellite(30m)with the lowest resolution is the closest to the accuracy of the seasonal inversion model.In addition,the accuracy of the inversion model of the neural network model is slightly higher than that of the regression model,and the regression model is better than the neural network model in some seasons.The results show that the inversion model should be established seasonally for high-resolution images,and the all-season inversion model can be considered as the image resolution gradually decreases.From the comprehensive consideration of inversion accuracy and inversion efficiency,the regression model can be used as the optimal model for the inversion of suspended solids concentration in multi-source remote sensing data. |
PDF Full Text Request