Model Of Inland Water Remote Sensing Of Chlorophyll A Concentration Of Inversion

Chlorophyll-a is an important element of the aquatic algae, which can directly reflect the eutrophication degree of water body. How to get the accurate quantitative simulation between remote sensing reflectance and chlorophyll-a concentration is the key to inverse chlorophyll-a concentration.The Guan Ting reservoir is one of the most important water resources in Beijing, Guishui River is one of the three main water source of the Guan Ting reservoir. The water was in the state of eutrophication for a long time. Monitoring Chlorophyll-a of Guishui River can reflect the water quality of Guan Ting reservoir indirectly. Taking Guishui River as the study area, try to inverse chlorophyll-a concentration of water using both hyperspectral and multispectral data, attempt to improve the accuracy of using remote sensing data monitoring chlorophyll-a concentration, for the purpose of providing some reference of technology and method in monitoring water quality of Guishui River with remote sensing technology.In the study of chlorophyll-a hyperspectral inversion:we introduce the theory of baseline correction. Reflectance of750nm was defined as a kind of baseline, a connecting line of reflectance from500nm to750nm was defined as another baseline. Baseline correction was defined as spectral reflectance minus baseline values. We evaluated the inversion ability of three-band model and support vector regression model. First, analyzing the measured spectral data of six filed survey from2011to2013, we found spectral data measured in May、June and July has the following characteristics, the mean value of reflectance at450nm was1.16×10-3higher than that of750nm, which from450nm to750nm was5.89×10-4higher than that from710nm to760nm, the absorption peak at676nm and the fluorescence peak at700nm were weak. All of these characteristics made it clear that Guishui River has the characteristics of high concentration of suspended solids at the days. Second, In order to weaken the spectral interferences of suspended matter, to improve the inversion precision of chlorophyll-a concentration, we made five kinds of baseline correction on the original spectrum data. In order to weaken the spectral interference from aquatic environment factors, first order differential processing was taken on the original spectrum data. In the end, original spectral data was used as the evaluation standard of spectra pre-treatment technology. Three-band model was applied to the six kinds of hyperspectral data after pretreatment and the original spectral data, seven three-band models were made, and evaluate the accuracy of models. Do the related analysis between bands selected from three-band models and Chlorophyll-a concentration, bands were selected as the input variables of SVM models while the correlation coefficient greater than0.49, training the SVM model, we made the following conclusions.(1) It was not suitable to take the first order differential technique as the spectral pretreatment when we build three-band models, but this kind of technology was good at finding the bands which carry a lot of spectral information of chlorophyll-a.(2) Reasonable baseline correction can improve the inversion precision of chlorophyll-a concentration:we did the second baseline correction method on spectral data which was measured in May、June and July and had the characteristics of high concentration of suspended solids, the inversion results were much better. Inversion results of the rest kinds of baseline correction were poor. Probably because spectral information was removed while baseline correction.(3) The SVM has a good scalability, SVM model can be used to inverse chlorophyll-a concentration with hyperspectral data, results got from SVM model were close to the measured values, and there were no deflection value.In the study of chlorophyll-a multispectral inversion:we build linear empirical models and Support Vector Machine (SVM) models to retrieve chlorophyll-a concentration of Guishui River based on in situ collected chlorophyll-a concentration data and Multi-spectral data of HJ-1A. After check the correlation coefficient between chlorophyll-a and reflectance of bands, band which has the largest correlation coefficient was chose to build linear inversion model, and when we chose reflectance of four bands of HJ-1A and two kind of band combinations with high correlation coefficient the input variable. Based on the results of inversion chlorophyll-a concentration, we evaluated the degree of eutrophication of water bodies and got the following conclusions:(1) Results of SVM model were better than that of linear regression model, which are more close to the measured values. The SVM model was suitable to be used in the inversion of chlorophyll-a concentration with multispectral data image data.(2) The spatial distribution of Chlorophyll-a concentration shows that deep water areas present lower Chlorophyll-a concentrations values than the shallow water areas, the upstream areas present higher Chlorophyll-a concentrations values than the downstream areas. The Mean value of Chlorophyll-a concentration of Guishui River increased6.8603ug/L in2012compared to2011.(3) we calculated the nutrition state index based on chlorophyll-a of the study area which was inversion by SVM models, then we evaluated the eutrophication degree of water body, this research shows that most of the water in study area was in a state of eutrophication, the others was in a state of mild nutritional status.