喀斯特地区的高分一号与Landsat8遥感影像归一化水体指数定量比较

以百色市田阳县右江河谷为研究样区,定量比较喀斯特地区的高分一号与Landsat8影像归一化水体指数 (Normalized difference water Index, NDWI),以期为喀斯特地区的水资源研究提供科学支撑。选择喀斯 特河谷盆地和喀斯特石山两类不同地貌类型为研究区,分别计算两类地区两类影像的NDWI,结合目视解译, 对两类影像的水体识别阈值和精度进行比较,进而分析两类影像的归一化水体指数的相互定量关系。 结果表明:1)从水体识别阈值看,高分一号影像的水体混合像元NDWI 值多在0.6$\sim$0.8之间,水体的NDWI值 大多为1.0,而Landsat8的水体混合像元NDWI值在-0.2$\sim$0之间,水体的NDWI值多为0以上; 2)从识别 精度而言,在河谷地带和山区,高分一号影像都明显优于Landsat8,特别是对混合水体混合像元的识别, 高分一号影像有比较好的优势; 3)两类影像NDWI值具有一定的线性关系,但定量回归结果不太理想。 高分一号和Landsat8在喀斯特地区水体识别的山地地带和河谷地带都有较好的效果,但高分一号的的估算精度和效果都优于Landsat8。

Quantitative Comparison of Normalized Difference Water Index of GF-1 and Landsat8 images in Karst Region

Youjiang River Valley in Tianyang County of Baise City, Guangxi, China, is taken as the research sample area, and the normalized differencewater index (NDWI) of remote sensing images of GF-1 and Landsat8 for this Karst area is quantitatively analyzed to provide important scientific support for the study of Karst water resources. Specifically, the valley basins and Karst rocky mountains in the Karst valley are selected as the study areas. Firstly, the normalized water index (NDWI) of the two types of remote sensing images are calculated respectively. Then combined with visual interpretation, the accuracy of water body identification and normalized water body index of the two types of images are compared. At last,the quantitative relationship between normalized water index of images is quantitatively analyzed and compared. The results show that the high-resolution images of GF-1 are superior to those of Landsat8 in valley and mountain areas, especially for the recognition of mixed directional elements. As for the threshold of water identification, the mixed pixels of high-resolution images are mostly between 0.6 and 0.8, and the NDWI values of water bodies are mostly around 1.0. The mixed pixels of Landsat8 are in the range of $-0.2\sim$0, and the NDWI value of water body is more than 0. In addition, the NDWI values of the two images have a certain linear relationship, but the quantitative regression results are not so satisfactory. It seems that both GF-1 and Landsat8 are qualified for identifying mountain topography and valley zones in Karst area, however, the estimafion accuracy and effect of GF-1 are better than Landsat8.