面向浑浊水体叶绿素a浓度遥感反演的光谱基线校正

根据浑浊水体中悬浮物的光学性质，提出使用线性基线校正方法来削弱水体光谱中的悬浮泥沙贡献。线性基线定义为450和750 nm的反射率值连线，基线校正为光谱反射率减去基线。2010年3月和2011年4月太湖梅梁湾水体的实测数据验证结果表明，光谱线性基线校正可以较好的提高叶绿素a浓度的反演精度，改进反演模型的诊断性。3月数据建立的波段比值反演模型中，校正前模型RMSE为4.11 mg·m-3，基线校正后模型RMSE为3.58 mg·m-3，同时，基线校正后模型残差的方差齐性和正态分布均有明显的改善。4月模型有类似结果，数据校正后的模型具有更小的误差。在没有水华发生的浑浊水体中，线性基线校正可以作为提高水体叶绿素a浓度反演精度的光谱数据处理方法。

Baseline Correction of Spectrum for the Inversion of Chlorophyll-a Concentration in the Turbidity Water

Suspended particle material is the main factor affecting remote sensing inversion of chlorophyll-a concentration (Chla) in turbidity water. According to the optical property of suspended material in water, the present paper proposed a linear baseline correction method to weaken the suspended particle contribution in the spectrum above turbidity water surface. The linear baseline was defined as the connecting line of reflectance from 450 to 750 nm, and baseline correction is that spectrum reflectance subtracts the baseline. Analysis result of field data in situ of Meiliangwan, Taihu Lake in April, 2011 and March, 2010 shows that spectrum linear baseline correction can improve the inversion precision of Chl a and produce the better model diagnoses. As the data in March, 2010, RMSE of band ratio model built by original spectrum is 4.11 mg·m-3, and that built by spectrum baseline correction is 3.58 mg·m-3. Meanwhile, residual distribution and homoscedasticity in the model built by baseline correction spectrum is improved obviously. The model RMSE of April, 2011 shows the similar result. The authors suggest that using linear baseline correction as the spectrum processing method to improve Chla inversion accuracy in turbidity water without algae bloom.