Vol17 No.11: 2707-2723
【Title】Evaluation of effective spectral features for glacial lake mapping by using Landsat-8 OLI imagery
【Author】ZHANG Mei-mei1; ZHAO Hang1*; CHEN Fang1,2,3; ZENG Jiang-yuan2
【Addresses】1 Key Laboratory of Digital Earth Science, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100094, China; 2 State Key Laboratory of Remote Sensing Science, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100094, China; 3 Hainan Key Laboratory of Earth Observation, Aerospace Information Research Institute, Chinese Academy of Sciences, Sanya 572029, China
【Corresponding author】ZHAO Hang
【Citation】Zhang MM, Zhao H, Chen F, et al. (2020) Evaluation of effective spectral features for glacial lake mapping by using Landsat-8 OLI imagery. Journal of Mountain Science 17(11). https://doi.org/10.1007/s11629-020-6255-4
【Abstract】Glacial lake mapping provides the most feasible way for investigating the water resources and monitoring the flood outburst hazards in High Mountain Region. However, various types of glacial lakes with different properties bring a constraint to the rapid and accurate glacial lake mapping over a large scale. Existing spectral features to map glacial lakes are diverse but some are generally limited to the specific glaciated regions or lake types, some have unclear applicability, which hamper their application for the large areas. To this end, this study provides a solution for evaluating the most effective spectral features in glacial lake mapping using Landsat-8 imagery. The 23 frequently-used lake mapping spectral features, including single band reflectance features, Water Index features and image transformation features were selected, then the insignificant features were filtered out based on scoring calculated from two classical feature selection methods - random forest and decision tree algorithm. The result shows that the three most prominent spectral features (SF) with high scores are NDWI1, EWI, and NDWI3 (renamed as SF8, SF19 and SF12 respectively). Accuracy assessment of glacial lake mapping results in five different test sites demonstrate that the selected features performed well and robustly in classifying different types of glacial lakes without any influence from the mountain shadows. SF8 and SF19 are superior for the detection of large amount of small glacial lakes, while some lake areas extracted by SF12 are incomplete. Moreover, SF8 achieved better accuracy than the other two features in terms of both Kappa Coefficient (0.8812) and Prediction (0.9025), which further indicates that SF8 has great potential for large scale glacial lake mapping in high mountainous area.
【Keywords】Glacial lake mapping; Landsat-8 OLI; Water Index; Spectral features