基于穷举法的鄱阳湖叶绿素a浓度高光谱反演模型与应用研究——以GF-5卫星AHSI数据为例

光学特性复杂的浑浊水体叶绿素a浓度反演一直是水环境遥感研究的难点，高光谱以其波宽窄、谱段多、获取地物信息近似连续等优势，有望成为解决该难题的有效手段.该文以鄱阳湖为研究水域，利用2009年10月、2011年7月及2016年10月获取的实测高光谱及叶绿素a浓度数据，基于穷举法建立了针对高分五号(Gaofen-5， GF-5)卫星可见短波红外高光谱相机(Advanced HyperSpectral Imager， AHSI)的叶绿素a浓度三波段反演模型(决定系数R2＝0.82；均方根误差RMSE＝1.468 μg·L－1).通过与2018年10月7日Sentinel-2A卫星多光谱成像仪(MultiSpectral Instrument， MSI)的叶绿素a浓度反演结果对比分析，两者的空间分布具有较高的一致性，超过80%的研究水域相对偏差小于8%，在一定程度上证明了GF-5卫星高光谱数据在浑浊水体叶绿素a浓度反演中的潜力.

Retrieval and application of chlorophyll-a concentration in the Poyang Lake based on exhaustion method：a case study of Chinese Gaofen-5 Satellite AHSI data

Retrieval of chlorophyll-a concentration in turbid water with complicated optical characteristics has always been a difficulty in remote sensing research on water environment， and hyperspectral is a promising method that is expected to solve this problem because of its advantages such as narrow wave width， multiple spectral bands and approximately continuous information acquisition of ground objects. This study establishes a three-band inversion model of chlorophyll-a concentration at the Poyang Lake for GF-5 Advanced HyperSpectral Imager (AHSI) using the exhaustive method. This model is validated based on the in-situ hyperspectrum and chlorophyll-a concentration data obtained in October 2009， July 2011， and October 2016， with a determination coefficient (R2) of 0.82 and a root mean square error (RMSE) of 1.468 μg·L－1. By comparing and analyzing the GF-5 AHSI and Sentinel-2A MultiSpectral Instrument (MSI) chlorophyll-a concentration inversion results at the Poyang Lake on October 7， 2018， we found that the spatial distribution of the two results were highly consistent， and the relative bias between them for more than 80% area of the Poyang Lake was less than 8%， which effectively demonstrated the potential of GF-5 AHSI in the monitoring of chlorophyll-a concentration for turbid waters.