利用高光谱反演模型评估太湖水体叶绿素a浓度分布

叶绿素a浓度是评价水体富营养化和初级生产力的一个重要参数,高光谱遥感是获取叶绿素a浓度的有效手段.为建立太湖水域叶绿素a的最佳高光谱估算模型,选取2015年5—7月共计60组同步实测高光谱数据和叶绿素a浓度数据,在地面光谱反射率和叶绿素a浓度相关性分析的基础上,使用2∶1的数据样本进行太湖水域叶绿素a的最佳高光谱估算模型的建立和验证,筛选模型分别为波段比值、三波段、荧光峰位置、峰谷距离、一阶微分、NDCI(Normalized Difference Chlorophyll Index)、峰面积、荧光峰高度、WCI(Water Chlorophyll-a Index)和四波段模型.结果表明,建模得到的四波段模型决定系数最高,峰面积模型的决定系数相对最低;四波段模型的反演精度最高,均方根误差(RMSE)为0.00376 mg·L~(-1),平均绝对误差(MAPE)为27.86%,而WCI模型的反演精度相对最低,RMSE为0.01231 mg·L~(-1),MAPE为45.11%.将反演精度最高的四波段模型应用于2015年8月3日的两景HSI(Hyperspectral Imaging Radiometer)高光谱影像数据,也得到较高精度,利用同步实测叶绿素a浓度验证的决定系数为0.7643,RMSE为0.00433 mg·L~(-1),MAPE为45.62%.在春、夏季叶绿素对水体光学特性占主导作用且叶绿素分布均匀的情景下,本研究可为太湖水域叶绿素a的高光谱反演和水环境监测提供有价值的参考,其它季节水体光谱特点的研究尚待进一步开展.

Evaluation on distribution of chlorophyll-a content in surface water of Taihu Lake by hyperspectral inversion models

Chlorophyll-a concentration, effectively obtained by using hyper-spectral remote sensing, is an important index for assessment of water eutrophication and primary productivity. In our study, 60 data samples were divided into 2:1 sub-samples to establish and validate the optimal hyperspectral estimation model of chlorophyll-a based on correlation analysis between ground spectral reflectance and chlorophyll-a synchronously measured from May to July in 2015. The screening models were band ratio, tri-waveband, fluorescence peak position, peak-valley distance, first-order differential, normalized difference chlorophyll index, peak area, fluorescence line height, water chlorophyll-a index and four-band, respectively. The results show that, R² of the four-band model was highest compared to that of the peak area model. The inversion accuracy of the four-band model was the highest, with RMSE=0.00376 mg·L^-1 and MAPE=27.86% and lowest in the WCI model with RMSE=0.01231 mg·L^-1 and MAPE=45.11%. The four-band model gave better performance applied in two scenes of Hyperspectral Imaging Radiometer data on April 3rd, 2015 and was validated with R²=0.7643, RMSE=0.00433 mg·L^-1 and MAPE=45.62% by synchronous chlorophyll-a. It can provide valuable references for hyperspectral inversion of chlorophyll-a and water environment monitoring in Taihu lake under the circumstances that chlorophyll plays an important role on water optical properties with relatively even distribution in spring and summer. Similar studies in other seasons should be conducted in future.