Studying On The Best Band Choice And Model Of Chlorophyll-a Concentration In Taihu Lake Using In Situ Spectrum Dat

The compositions of the water spectra signal and the shortcomings of traditional spectra measurement above-water were analyzed in this paper at first, and then the method of above-water spectra measurement for inland Class- â…¡ water presented by Tang et al. (2004) was adopted. Through repeated measurements in Taihu Lake, a large quantity of hyperspectral data and synchronous water quality data of the Lake were obtained from June to August of 2004. Since the destruction of the direct sunlight reflection and hull for water light field was avoided effectively and the impact of skylight on water spectra signal was eliminated, we got the comparatively pure water spectra information. Those data acquired in unfavorable or abnormal monitoring conditions were removed from the dataset. The remaining ones were analyzed to calculate water-leaving radiance and reflectance of water in Taihu Lake.Based on three months' reflectance data in Taihu Lake, the data were processed by the methods of normalization and first derivative respectively. The inversion models of chlorophyll-a concentration were analyzed detailedly in the choice of sensitive bands, the building of models and the accuracy of models. Then, through calculating the water normalized reflectance in different bandwidth, the sensitive bands and estimation level of chlorophyll-a concentration inversion in different bandwidth were analyzed. The main conclusions of this paper are as follows:(1)Through comparative analysis between the remote sensing reflectance spectra from slantwise measurement and the relative reflectance spectra from vertical measurement, it was found that there were two peaks at 719nm and 725nm in remote sensing reflectance spectra, and the one at 719nm was more evident and stable and didn't drift with chlorophyll-a concentration level;(2)Based on the normalized remote sensing reflectance data, through comparative analysis it was found that hyperspectral reflectance at these two peaks can improve the accuracy of estimating chlorophyll-a. Moreover, the accuracy of estimation was the highest using the ratio R719nm/R670nm as the independent variable in the linear model.(3)Based on the first derivative results of remote sensing reflectance data, through comparative analysis of inversion models of chlorophyll-a, it was found that the exponential model was better in the models built by single band, while the linear model was better in the models built by two bands radio. Moreover, the accuracy of estimation is the highest using the ratio D700nm/D477nm as the independent variable in the linear model.(4)Through studying on the correlation coefficient curves between normalized remote sensing reflectance data at 25 kinds of bandwidths and chlorophyll-a concentration, it was found that the maximum correlation coefficient showed a decreasing tendency with the increasing of bandwidth, the maximum positive correlation bands having a trend of moving to the long wave while the maximum negative correlation bands having a trend of moving to the short wave. Moreover,...