基于太湖、巢湖、滇池和三峡水库水体组分生物光学特性, 根据辐射传输模型和神经网络优化算法建立悬浮颗粒物和叶绿素浓度优化生物光学反演模型.利用野外实测数据对优化生物光学算法进行检验, 结果表明, 该优化生物光学反演模型在一定程度上可以减少测量噪音对反演精度的影响.反演结果表明, 受悬浮颗粒物和叶绿素生物光学特性时空差异影响, 该优化生物光学反演模型在太湖、巢湖、滇池和山峡反演精度具有一定的差异, 但总体上能够较为准确地反演悬浮颗粒物和叶绿素浓度.其中悬浮颗粒物反演精度(平均绝对误差: MAPE, 均方根误差: RMSE)分别能够达到23%和15.13mg/L(样本数N=228), 叶绿素反演精度(MAPE和RMSE)分别能够达到26%和17.68μg/L(样本数N=228).

Based on the bio-optical properties of water constitute in the Taihu, Chaohu, Dianchi lake and Sanxia reservoir, bio-optical retrieval model for the suspended particle matter and chlorophyll-a were established according to the radiance transfer model and genetic optimization algorithm. The measured data in situ was used to test this optimized model. The results indicate that, the model can remove the influence of noise on the final retrieval precision in a certain degree. The retrieval precision of this model in the Taihu, Chaohu, Dianchi lake and Sanxia reservoir is different due to the variation of suspended particle matter and chlorophyll-a bio-optical properties. However, in general, this model can retrieve the concentration of suspended particle matter and chlorophyll-a with a preferable precision. Two parameters, mean absolutely percentage error (MAPE) and root mean square error (RMSE) which represent the retrieval precision of suspended particle matter, can reach to a value of 23% and 15.13mg/L, respectively (the number of sampling points is 228). The retrieval precision of chlorophyll-a (MAPE and RMSE) can reach to 26% and 17.68μg/L, respectively with the same number of sampling points.