基于2006-01-07~2006-01-09和2006-07-29~2006-08-01太湖地面实测高光谱数据以及同步水质参数数据,对比分析了三波段模型、两波段模型、反射峰位置法、一阶微分法4种方法用于估算太湖叶绿素a浓度的精度,并讨论其应用于遥感影像中估算叶绿素a浓度的可行性.2次采样3类水色参数总悬浮物、叶绿素a浓度和有色可溶性有机物在440 nm处吸收系数的变化范围分别为12.24~285.20 mg.L-1、4.83~155.11μg.L-1和0.27~2.36 m-1.前述4种方法在反演太湖水体的叶绿素a浓度时都取得较高的精度;决定系数分别为:0.813、0.838、0.872、0.819,均方根误差分别为:13.04、12.12、13.41、12.13μg.L-1;相对误差分别为:35.5%、34.9%、24.6%、41.8%.反射峰位置法估算精度最高,但应用到叶绿素a浓度遥感影像估算比较困难.三波段模型和两波段模型的反演结果优于传统的一阶微分法,且在卫星遥感反演中具有良好的应用前景.根据模拟MERIS数据,分别得到最优三波段模型[R-1(665)-R-1(709)]×R(754)和两波段模型R(709)/R(681),其决定系数、均方根误差、相对误差分别为0.788、13.87μg.L-1、37.3%和0.815、12.96μg.L-1、34.8%,反映了MERIS数据能非常好地应用于太湖这类浑浊二类水体叶绿素a浓度的精确估算.
Based on the measured remote sensing reflectance and concurrent chlorophyll a (Chl-a) concentration in Taihu Lake from January 7 to 9 and July 29 to August 1, 2006, this study comparatively analyzed the estimation precision of three-band-model, two-band-model, reflectance peak position method and first derivative method, and further discussed the feasibility of the four methods to estimate Chl-a using remote sensing image. The data set of two samplings contained widely variable total suspended matter (12.24-285.20 μg.L-1), Chl-a (4.83- 155.11 μg.L-1 ) and chromophoric dissolved organic matte absorption coefficient at 440 nm (0.27-2.36 m-1 ). The former four methods all got high precisions on Chl-a concentration estimation in Taihu Lake with determination coefficients (r2) of 0. 813, 0.838, 0. 872 and 0.819, respectively. The root mean square error (RMSE) between measured and estimated Chl-a concentrations using the four models was 13.04, 12.12, 13.41 and 12.13 μg.L-1, respectively, and the relatively error (RE) was 35.5%, 34.9%, 24.6% and 41.8%, respectively. Although the reflectance peak position method had the highest estimation precision, it was difficult to be applied on remote sensing image due to lacking spectral channel. The three-band-model and two-band-model had higher estimation precisions than the first order differential method and good application foreground in Chl-a retrieval using remote sensing image. The r2 , RMSE, RE of [ R-1(665) - R-1 (709)] × R (754) in three-band-model and R(709)/R(681) in two-band-model based on simulation MERIS data were 0.788, 13.87 μg.L-1, 37.3% , and 0.815, 12.96μg.L-1, 34.8%, respectively. The results in this study demonstrated MERIS data could be applied to retrieve Chl-a concentration in turbid Case- Ⅱ waters as Taihu Lake.