Large-scale glacier area expansion or shrinkage is among the most conspicuous indicators of regional and global climate change. Glacier extraction from satellite images is an essential step in large-scale area change monitoring. However, glacier extraction is usually influenced by cloud cover and seasonal snow, and thus far, there is no effective and automatic method to monitor glacier area change in large regions. In this study, a new method ‘multi-temporal minimum NDSI composite’ is proposed for clean glacier and nonseasonal snow extraction, and the results agree well with the glacier inventory (excluding the debris part), except in Himalayan regions where non-seasonal snow exists widely out of glaciers. This method is also used to estimate the glacier and nonseasonal snow area changes in High Mountain Asia (HMA) over the past 29 years. With this method, we can use abundant Landsat images of summer seasons from continuous years to composite a cloud-free and seasonal snow-free glacial image pixel by pixel. Glacial maps for the whole HMA are produced every 10 years beginning in 1990, so glacier changes are presented in different subregions and periods. The number of images which is used to composite a glacial map in each pixel is also given for uncertainty analysis. According to the results, the clean glacier and nonseasonal snow area in HMA decreased by −0.43 ± 0.19%/a during 1990–2018 but had high temporal and spatial heterogeneity. Clean glacier and nonseasonal snow area decreased rapidly during 1990–2000, slowed down during 2000–2010, and then sped up during 2010–2018. The fastest reductions were in Hengduan Shan at rates of approximately −1.05 ± 0.09%/a, and the only area with growth (at a rate of 0.50 ± 0.11%/a) was in West Kun Lun during 1990–2018.

大规模冰川面积的扩大或缩小是区域和全球气候变化最明显的指标之一。从卫星图像中提取冰川是大规模区域变化监测中必不可少的步骤。然而，冰川提取通常受云层覆盖和季节性降雪的影响，因此，到目前为止，还没有有效且自动的方法来监测大区域的冰川面积变化。在这项研究中，提出了一种用于清洁冰川和非季节性积雪的新方法“多时间最小NDSI复合”，其结果与冰川存量（不包括碎片）非常吻合，除了喜马拉雅地区非季节性积雪广泛存在于冰川之外。该方法还用于估算过去29年亚洲高山（HMA）的冰川和非季节性降雪面积的变化。通过这种方法，我们可以使用连续多年的夏季大量Landsat图像来逐像素合成无云和无季节的冰川图像。从1990年开始，每10年制作一次整个HMA的冰川图，因此在不同的次区域和时期都显示了冰川变化。还给出了用于在每个像素中合成冰川图的图像数量，以进行不确定性分析。根据结果​​，HMA的清洁冰川和非季节性降雪面积在1990–2018年期间下降了-0.43±0.19％/ a，但具有较高的时空异质性。清洁的冰川和非季节性降雪面积在1990–2000年期间迅速减少，在2000–2010年期间减慢，然后在2010–2018年加速。降幅最快的是横断山，降幅约为-1.05±0.09％/ a，