基于高光谱数据的水体叶绿素a指数反演模型的建立

水体叶绿素a含量是反映水体质量的重要指标之一,利用遥感技术监测其含量具有众多优势。该研究利用2012年7月在广西壮族自治区桂林市漓江流域实地采集的水体高光谱数据和实验室化验分析数据,借鉴陆表植被叶绿素a的遥感反演模型,发展了一种新的水体叶绿素a提取指数(water chlorophyll-a index,WCI)。通过与反射率敏感波段法、波段比值法和半分析方法对比分析发现,新提出的WCI指数使用650、685、696 nm波段,波段稳定,决定系数R2可达0.58,均方根误差最小为0.24,受水体悬浮物影响小,在天津海河区域的验证效果也表明了该模型可以有效地提取水体叶绿素a含量。该方法扩展了水体叶绿素a监测的建模思路,对水体叶绿素a监测建模有一定的指导作用。

Water chlorophyll-a retrieval index based on hyperspectral data

Abstract: Water chlorophyll-a is one of the most important indices for water quality monitoring. Remote sensing technology has strong advantages in monitoring both water and vegetation chlorophyll-a concentrations. Most of the current study on water chlorophyll-a monitoring chose the sensitive band based on the water chlorophyll-a spectral characteristics, and then established the inversion model. Some researchers established the water parameters inversion model based on an analytical physical mechanism, which are more complex and difficult to use in practice. And we also noticed that a vertical comparative analysis was needed for all these different inversion methods in the same area, and a few researchers used the water chlorophyll-a absorption similarity with leaf to build the water chlorophyll-a retrieval model. In this paper, a new water chlorophyll-a retrival index WCI (Water Chlorophyll-a Index) was built from the land surface vegetation chlorophyll retrieval index MTCI (MERIS terrestrial chlorophyll index), based on the in-situ water hyperspectral data and water chlorophyll-a content results in the laboratory in July 2012 in the Lijiang River, Guangxi Zhuang Autonomous Region. The MTCI was based on the fast climbing vegetation reflectance in 680-750 nm also called the "red edge." The MTCI was easy to calculate, and had a strong correlation with leaf chlorophyll-a content. From the beginning of 2004, the MTCI has became the core algorithm of the land chlorophyll-a product on ESA Envisat satellite's MERIS sensor. This index is now widely used in land leaf chlorophyll-a retrival and net primary productivity (NPP) estimation. The WCI index also uses the different ratio of characteristic bands to represent the water chlorophyll-a content. The WCI index uses hyperspectral water reflectance at 650, 685, and 696 nm. We used the traditional method at the same location to verify all these models's effect. The traditional methods consist of the reflectance model, reflectance ratio model, and the semi-analytical model (three bands model). The three traditional methods directly selected the water spectral reflectance at certain bands. Spectral smoothing can reduce the band noise at certain extent, but is easy to select the wrong band for the measured hyperspectral data because the absorption and reflection peak of water spectrum has big differences in different water spectrum curves. Our research also moticed that the change information of water spectrum was more useful compared with the water spectrum itself. Our results indicated that the new WCI index showed the best coefficient of determination 0.58 and the least RMSE 0.24 compared with the reflectance model, reflectance ratio model, and semi-analytical model. The test results also showed that the WCI model can retrieve the water chlorophyll-a content effectively at Tianjin City Haihe River. This method extended the idea of water chlorophyll-a content modeling from the view of the terrestrial vegetation chlorophyll-a monitoring, and has certain instructive effect on water chlorophyll-a content monitoring. More situ data of different water bodies is needed to verify the new model's robustness and effectiveness.