| In recent years,with the development of water pollution control work,the quality of water environment has been continuously optimized,but there are still problems such as inadequate governance boundaries and repeatability.The eutrophication of inland lakes and coastal waters has occurred from time to time,which has a serious impact on the healthy and sustainable development of the economy and the normal life of residents.Therefore,it is of great significance to carry out accurate dynamic monitoring of water quality for the overall optimization of water environment.The concentration of chlorophyll a is the basic index to measure the primary productivity of water.By monitoring the concentration of chlorophyll-a in water,the eutrophication of water can be effectively evaluated.Remote sensing technology has special advantages such as fast,accurate and periodic,and has been used in the field of water quality monitoring for a long time.In the monitoring of chlorophyll-a concentration,remote sensing technology has strong practical significance and application prospects.In this paper,Dianshan Lake in Qingpu District of Shanghai and the western coast of Hong Kong are taken as the research areas.The spectral reflectance characteristics of the water bodies in the two research areas are analyzed by using the hyperspectral remote sensing image of Zhuhai-1.Combined with the data of chlorophyll a concentration,the inversion models are constructed based on empirical method,semi-analytical method and artificial neural network respectively,and the spatial and temporal distribution characteristics of chlorophyll a concentration in the study area are obtained.The main research contents and conclusions are as follows:(1)Preprocessing was conducted on the Zhuhai-1 hyperspectral and Landsat-8multispectral satellite imagery data for different time periods in the study area.A comparative analysis was performed on the spectral reflection characteristics of different study areas and different images.The results show that the spectral resolution of Zhuhai-1 hyperspectral remote sensing imagery is higher compared to Landsat-8multispectral satellite imagery,providing more detailed spectral reflection curves with more pronounced peak and valley positions.The fluorescence peak near 700 nm in the Dianshan Lake area is more prominent than in the offshore waters of western Hong Kong,indicating a richer presence of algal substances in the water.(2)Based on the measured concentration data of chlorophyll-a and spectral reflectance,Pearson correlation analysis was used to calculate the correlation coefficients between single-band reflectance and band combinations with chlorophyll-a concentration.Band combinations with higher correlation were selected to construct chlorophyll-a concentration inversion models for the two study areas.Based on the analysis of the apparent optical properties of Dianshan Lake water using measured spectral data from an ASD spectrometer,and considering the spectral curve characteristics,three-band models and peak position models were separately constructed for the two study areas.In the 400-500 nm range,the reflectance is low due to the strong absorption of chlorophyll-a and yellow substances in the blue-violet wavelength range.The highest reflectance peak appears at 560-570 nm due to the scattering effect of algae and suspended matter.The reflectance valley at 660 nm is caused by the strong absorption of chlorophyll-a in the red wavelength range.The fluorescence effect of chlorophyll-a causes a reflectance peak at 700 nm.When the content of planktonic organisms and algae in the water increases and the concentration of chlorophyll-a rises,the peak value of the reflectance peak also increases and shifts towards longer wavelengths.The overall inversion accuracy of the three-band models and peak position models in the two study areas is unsatisfactory,indicating limited applicability of the semi-analytical method in the study areas.(3)A BP neural network model was constructed.In the Dianshan Lake area,the root mean square error(RMSE)between the inverted chlorophyll-a concentration and the measured values was 1.68 μg/L,with an average relative error of 13.27%.There were 8 validation points with relative errors within 10%,indicating slightly better inversion accuracy compared to the band combination model.In the offshore waters of western Hong Kong,the RMSE between the inverted chlorophyll-a concentration and the measured values was 1.91 μg/L,with an average relative error of 34.55%.The inversion accuracy was not high.For the offshore waters of western Hong Kong,grey relational analysis(GRA)was used to select other water quality parameters in the area as covariates to improve the BP neural network model.The improved GRA-BP neural network model showed significant improvements in determination coefficient and inversion accuracy,with an RMSE of 1.64 μg/L and an average relative error of19.23%.(4)There are spatial variations in the distribution of chlorophyll-a concentration in the Dianshan Lake area.The concentration distribution is uneven,with some small nearshore areas exhibiting higher local chlorophyll-a concentrations.These areas are mostly strongly influenced by surrounding human activities,and the chlorophyll-a concentration shows an upward trend over the years.In the offshore waters of western Hong Kong,there is a significant seasonal variation in chlorophyll-a concentration,with higher concentrations overall during summer compared to winter.High concentrations of chlorophyll-a are mostly observed in narrow nearshore areas,such as eastern Victoria Harbour,the northern part of Lantau Island,and the northeastern part of Hau Hoi Wan. |
PDF Full Text Request