ABSTRACT

In this study, bio-optical parameters were derived from hyperspectral data of optically complex waters using spectrum matching technique (SMT). Models for inherent optical properties (IOPs) of the water column were tuned using in-situ dataset from the study area (i.e., Chilika lagoon). Constructed IOPs were used to simulate remote sensing reflectance ($R_{\mathrm{r}\mathrm{s}}$) spectra at 60 wavelengths (equally spaced between 400 and 700 nm) using radiative transfer solution provided by Hydrolight-Ecolight software (HE53). Results of the simulation were stored as a $R_{\mathrm{r}\mathrm{s}}$ – IOP look up table (LUT). To check the accuracy, in-situ measured $R_{\mathrm{r}\mathrm{s}}$ were compared with those from the LUT. Retrieved values of two bio-optical parameters i.e., chlorophyll-a concentration (Chl-a) and coloured dissolved organic matter (CDOM)+Detritus absorption coefficient at 440 nm ($a_{\mathrm{d}\mathrm{g}}(440)$) were compared with corresponding in-situ measurements to get good statistical match. Coefficient of determination ($R^2$) and root mean squared error (RMSE) were 0.80 and 2.66 mg m${ }^{-3}$ respectively for Chl-a, whereas 0.77 and 0.23 m${ }^{-1}$ respectively for $a_{\mathrm{d}\mathrm{g}}(440)$. These parameters were also retrieved using two commonly used semi-analytical inversion algorithms (SAA)- (a) Linear matrix inversion (LMI) and (b) Garver-Siegal Maritorena (GSM). Both the SAA showed poor performance. $R^2$ for Chl-a from GSM and LMI were 0.13 and 0.41, respectively, with RMSE of 6.85 mg m${ }^{-3}$ and 4.82 mg m${ }^{-3}$ respectively. For $a_{\mathrm{d}\mathrm{g}}(440)$, the value of $R^2$ from GSM and LMI were 0.87 and 0.71, respectively, but with a high RMSE of 0.91 m${ }^{-1}$ and 0.81 m${ }^{-1}$ respectively. SMT was applied to airborne hyperspectral AVIRIS-NG (Airborne Visible/Infrared Imaging Spectrometer Next Generation) dataset of Chilika lake to derive pixel-wise chlorophyll-a concentration and the magnitude of CDOM+Detritus absorption coefficient at 440 nm ($a_{\mathrm{d}\mathrm{g}}(440)$). Spatial variability of these parameters in its different domains (i.e. Northern-, Central- and Southern-region of the lake) have been addressed.

摘要

在这项研究中，生物光学参数是使用光谱匹配技术（SMT）从光学复杂水域的高光谱数据中得出的。使用研究区域（即Chilika泻湖）的原位数据集，对水柱的固有光学特性（IOP）模型进行了调整。构造的IOP用于模拟遥感反射率（${\mathrm{[R}}_{\mathrm{[R}\mathrm{s}}$）使用Hydrolight-Ecolight软件（HE53）提供的辐射转移溶液在60个波长（等间距在400和700 nm之间）处的光谱。模拟结果存储为${\mathrm{[R}}_{\mathrm{[R}\mathrm{s}}$– IOP查找表（LUT）。要检查精度，就地测量${\mathrm{[R}}_{\mathrm{[R}\mathrm{s}}$与来自LUT的进行了比较。两种生物光学参数的检索值，即叶绿素a浓度（Chl- a）和有色溶解有机物（CDOM）+碎屑在440 nm处的吸收系数（${\mathrm{一种}}_{\mathrm{d}\mathrm{G}}（440）$）与相应的原位测量结果进行比较，以获得良好的统计匹配。测定系数（${\mathrm{[R}}^2$）和均方根误差（RMSE）分别为0.80和2.66 mg m${ }^{-3}$Chl- a分别为0.77和0.23 m${ }^{-\mathrm{1个}}$ 分别为 ${\mathrm{一种}}_{\mathrm{d}\mathrm{G}}（440）$。还使用两种常用的半分析反演算法（SAA）检索了这些参数-（a）线性矩阵反演（LMI）和（b）Garver-Siegal Maritorena（GSM）。两种SAA均表现不佳。${\mathrm{[R}}^2$GSM和LMI中的Chl- a分别为0.13和0.41，RMSE为6.85 mg m${ }^{-3}$ 和4.82毫克米${ }^{-3}$分别。对于${\mathrm{一种}}_{\mathrm{d}\mathrm{G}}（440）$， 的价值 ${\mathrm{[R}}^2$ 来自GSM和LMI的数据分别为0.87和0.71，但均方根误差（RMSE）高达0.91 m${ }^{-\mathrm{1个}}$ 和0.81 m${ }^{-\mathrm{1个}}$分别。将SMT应用于Chilika湖的机载高光谱AVIRIS-NG（下一代机载可见/红外成像光谱仪）数据集，以得出像素级叶绿素a浓度以及440 nm处CDOM +碎屑吸收系数的幅度（${\mathrm{一种}}_{\mathrm{d}\mathrm{G}}（440）$）。这些参数在其不同区域（即湖的北部，中部和南部区域）的空间变异性已得到解决。