基于高分五号卫星遥感数据的长江河口叶绿素a浓度反演

Inversion of chlorophyll-a concentration in the Yangtze River estuary based on remote sensing data from GF-5 satellite

采用高分五号卫星遥感监测长江河口水质,通过星地同步取样测得水体叶绿素a浓度的实际值,并从相应水体的卫星高光谱影像中提取水体表面反射率,构建8种不同波段组合参数的指数型叶绿素a浓度反演模型。从统计学参数、反演结果影像、误差稳定性3方面对反演模型进行评价。结果表明:反演模型普遍具有较高的决定系数R~2和较低的平均相对误差,增加使用的波段数时,R~2有所增大,平均相对误差略减小。结合反演结果图像和抗噪性能图像,可判断各模型受光照条件与随机噪声干扰的程度。基于简单参数组合的多波段算法模型具有最高的反演精度,R~2为0.983,平均相对误差为14.33%。证实高分五号卫星的高光谱遥感技术应用于河口水质反演与水体富营养监测具有可行性。

The water quality of the Yangtze River estuary was monitored by remote sensing using GF-5 satellite.The actual values of chlorophyll-a concentration in water bodies were measured by simultaneous satellite-ground sampling.The surface reflectance of water bodies was extracted from the satellite hyperspectral images of the corresponding water bodies to construct an exponential chlorophyll-a concentration inversion model with eight different waveband combination parameters.The inversion models were evaluated from three aspects:statistical parameters,inversion result images and error stability.The results show that the inversion models generally have a high coefficient of determination R~2 and low mean relative error,and the R~2 increases and the mean relative error decreases slightly when the number of bands used is increased.Combining the images of inversion results and noise-robust image,the interference degree of each model by illumination conditions and random noise can be judged.The multi-band algorithm model based on simple parameter combination has the highest inversion accuracy with R~2 of 0.983 and an average relative error of 14.33%.It confirms the feasibility of the application of hyperspectral remote sensing from GF-5 satellite to estuarine water quality inversion and eutrophication monitoring.