基于实测光谱与Landsat8影像的白洋淀COD遥感反演

Remote sensing inversion of COD in Baiyang Lake based on actually-measured spectra and Landsat8 image

以白洋淀为研究区域,结合实测光谱、Landsat8影像和化学需氧量(COD)实测值,使用BP神经网络构建COD反演模型。结果显示,光谱反射率可对白洋淀水体COD污染程度进行区分,BP神经网络模型预测值与实测值的平均相对误差为16.5%,模型精度较好。基于模型反演2017年10月30日白洋淀水体COD浓度空间分布,白洋淀水体存在一定程度的有机污染,部分水体达到劣V类,污染中心主要位于东南部的村镇和旅游景点,区域内的生活、生产污水可能是造成COD升高的主要原因。

By taking Baiyang Lake as the research area,the chemical oxygen demand(COD)inversion model is constructed by using BP neural network,combining actually.measured spectrum,Landsat8 image and COD actually-measured values.The results show that the spectral reflectivity can distinguish the pollution level of COD in water of Baiyang Lake,the average relative error of the value predicted by BP neural network and actually-measured value is 16.5%,and the model has high accuracy.The space distribution of COD concentration in water of Baiyang Lake on 30th October,2017,is inversed on the basis of the model.The inversion results show that the Baiyang Lake has organic pollution to a certain extent,and the pollution level of part of the water can reach up to V-level,the center of the pollution area locates around villages and scenic spots at the southeast of Baiyang Lake.The main reason to cause COD rising is domestic and industrial sewage in the areas.