中文摘要：

本文利用MODIS数据的可见光、近红外波段和准实时的地面采样数据，分别利用线性回归模型和神经网络模型反演了太湖的叶绿素a和悬浮物浓度。结果表明，利用MODIS数据的波段组合（M2／M8）和（M2／M9）可估算太湖的叶绿素a浓度；而MODIS数据的波段组合（M12／M17）、（M13／M17）及MODIS（M4）波段能定量估算太湖的悬浮物浓度，但估算精度仍不能满足实际需要。因此，构建了一个以MODIS可见光及近红外波段为输入，以太湖水质参数为输出的2层BP神经网络模型反演太湖的水质参数，大大提高了反演精度。

英文摘要：

MODIS has become a potential remote sensing data for monitoring inland lakes＇ water quality because of it＇s high spectral resolution, time resolution and radiance resolution. The correlation between water quality parameters and visible and near infra-red bands was analyzed and the liner model and BP neural net model were used to retrieve water quality concentration. It was demonstrated that SS in Taihu Lake has the highest correlation coefficient with MODIS band 13, chlorophyll-a in Taihu Lake has the highest correlation coefficient with MODIS band 2, in the hundreds of band combinations, SS in Taihu Lake has the highest correlation with M12/M17, chlorophyll-a has the highest correlation with M2/MS, so the chlorophyll-a concentration can be retrieved by band combination of M2/M8 and M2/M9, while solid suspended concentration can be estimated using band combination of M12/M17, M13/M17 and band 4, but the estimation accuracy is very low that can＇t be used in practical applications. So a two-layer BP neural net with 16 input nodes of visible near-infrared bands of MODIS and 1 output nodes of water quality, was constructed to infer the water quality concentration in Taihu Lake, the accuracy of this model was highly improved compared to the linear model. Water quality in Lake Taibu can be effectively retrieved via BP neural net and MODIS＇s visible and near-infrared bands.