Automated Glacier Extraction Index by Optimization of Red/SWIR and NIR /SWIR Ratio Index for Glacier Mapping Using Landsat Imagery
College of Urban and Environmental Science, Northwest University, Xi’an 710127, China
Shaanxi Key Laboratory of Earth Surface System and Environmental Carrying Capacity, Northwest University, Xi’an 710127, China
Shaanxi Key Laboratory of Ecology and Environment of River Wetland, Weinan Normal University, Weinan 714099, China
Author to whom correspondence should be addressed.
Water 2019, 11(6), 1223; https://doi.org/10.3390/w11061223
Received: 9 May 2019 / Revised: 30 May 2019 / Accepted: 6 June 2019 / Published: 12 June 2019
(This article belongs to the Section Hydrology and Hydrogeology)
Glaciers are recognized as key indicators of climate change on account of their sensitive reaction to minute climate variations. Extracting more accurate glacier boundaries from satellite data has become increasingly popular over the past decade, particularly when glacier outlines are regarded as a basis for change assessment. Automated multispectral glacier mapping methods based on Landsat imagery are more accurate, efficient and repeatable compared with previous glacier classification methods. However, some challenges still exist in regard to shadowed areas, clouds, water, and debris cover. In this study, a new index called the automated glacier extraction index (AGEI) is proposed to reduce water and shadow classification errors and improve the mapping accuracy of debris-free glaciers using Landsat imagery. Four test areas in China were selected and the performances of four commonly used methods: Maximum-likelihood supervised classification (ML), normalized difference snow and ice index (NDSI), single-band ratios Red/SWIR, and NIR/SWIR, were compared with the AGEI. Multiple thresholds identified by inspecting the shadowed glacier areas were tested to determine an optimal threshold. The confusion matrix, sub-pixel analysis, and plot-scale validation were calculated to evaluate the accuracies of glacier maps. The overall accuracies (OAs) created by AGEI were the highest compared to the four existing automatic methods. The sub-pixel analysis revealed that AGEI was the most accurate method for classifying glacier edge mixed pixels. Plot-scale validation indicated AGEI was good at separating challenging features from glaciers and matched the actual distribution of debris-free glaciers most closely. Therefore, the AGEI with an optimal threshold can be used for mapping debris-free glaciers with high accuracy, particularly in areas with shadows and water features.