Study On Hyperspectral Remote Sensing Estimation Of Chlorophyll A Concentration In Inland Water

Lake water resources are one of the most precious water resources in my country,and have an important impact on national economic construction and people’s lives.Strengthening the dynamic monitoring of lake water quality is of great significance to achieve its refined and scientific management.Chlorophyll a is an important water quality safety evaluation index.The As one of the most precious inland water resources in my country,lake water resources have a real impact on the quality of life of our people and the rapid development of the national economy.Carrying out accurate dynamic monitoring of lake water quality is very important for the realization of pollution prevention and scientific control of inland lakes.Water resource management is of great importance.As an important water quality parameter of inland lake water quality,chlorophyll a is also one of the main components of phytoplankton organisms in the water body.Its concentration is an important biological indicator that objectively reflects the degree of eutrophication of inland lake water quality.Accurate monitoring of the concentration is also of great significance to the development of the aquatic industry,the balance of the water ecosystem,and the safety of human drinking water.The application of remote sensing methods to quickly and accurately determine the concentration of chlorophyll a in water is an important research direction in the current water ecological environment monitoring.The current research on remote sensing monitoring of chlorophyll a concentration in inland lake water is generally divided into two categories,one is based on measured spectra,and the other is based on satellite sensor data.Among them,the research based on measured hyperspectral mostly focuses on the mining of sensitive spectral features or the construction of a single link of regression model.Most of the research based on satellite sensor data uses multi-spectral remote sensing data,and the accuracy needs to be improved.Therefore,this study used hyperspectral remote sensing data.Based on the inversion research of chlorophyll a concentration in inland lake water,in the overall research process of inversion of chlorophyll a concentration in water based on measured spectral data to image data,focus on the introduction of spectrum analysis algorithms and innovations combined with intelligent inversion models,etc.The main research conclusions of this thesis include the following aspects:（1）In the study of chlorophyll a concentration in Nansi Lake water based on measured hyperspectral inversion,the Empirical Mode Decomposition（EMD）method could achieve effective noise reduction of the measured water spectral signal,and was different from the traditional signal preprocessing method.Compared with the chlorophyll a concentration prediction model established by the acquired spectral parameters,the three-band parameters obtained from the first-order differential spectral reflectance of the measured water body after EMD noise reduction,combined with the XGBoost intelligent algorithm model,obtained the best prediction result of chlorophyll a concentration,with R~2 of 0.902 and RMSE of 1.131μg·L^-1.The research in this thesis shows that the frequency domain method can effectively broaden the way of preprocessing the measured water spectral signal.（2）Based on the"Zhuhai-1"OHS hyperspectral imagery in 2021,the empirical formula model,the three-band linear model,the three-band one-dimensional quadratic model and the three-band e Xtreme Gradient Boosting（XGBoost）intelligent model were used to invert the chlorophyll a concentration of the whole lake in Nansi Lake.The inversion effect of the band XGBoost intelligent algorithm was the best.It was verified that the inversion accuracy of chlorophyll a concentration R~2 was 0.8197,which can provide an important reference for the subsequent hyperspectral remote sensing monitoring of water pollution in inland lakes.（3）The Savitzky-Golay（SG）filtering method based on"ZY1-02D"noise reduction and the inversion of the water chlorophyll a concentration intelligent algorithm showed that the SG filtering method could achieve effective noise reduction for"ZY1-02D"satellite hyperspectral images,the noise reduction effect of the image can be improved by adjusting the filtering parameters,and the inversion accuracy of chlorophyll a concentration can be significantly improved.Used the denoised image to invert the chlorophyll a concentration,the highest precision R~2 can reach 0.8737,and the RMSE was 4.2259μg·L^-1.The research in this thesis shows that the"ZY1-02D"satellite has innate excellent performance,and the hyperspectral images obtained by it have been processed by noise reduction,which can provide more accurate and convenient monitoring methods and methods for water quality remote sensing monitoring research.（4）In the summer of 2021,the areas with high chlorophyll a concentration are mainly distributed on the banks of the upper lake of Nansi Lake,where the river flows into the lake,and the areas with weak lake water mobility,and the areas with low chlorophyll a concentration are mainly located in the lower lake Weishan Lake.More open lake areas and canals.The spatial distribution of chlorophyll a concentration in the water bodies of Nansi Lake showed non-uniformity,and the areas with higher chlorophyll a concentration in the Nanyang Lake water body were scattered,and the overall value was high.There is a large area and concentrated high-value areas on the north side of Dushan Lake,which is generally higher than that of Zhaoyang Lake,and the concentration of chlorophyll a in Weishan Lake is lower.