亚洲高山区冰川编目数据集（2020）

亚洲高山区发育着全球中低纬度面积最大，范围最广的山地冰川，是“亚洲水塔”重要的组成部分。已有区域内的冰川编目所使用的遥感数据源时间跨度会影响冰川学研究中的定量估算。为了得到一个最新的且时间集中的冰川编目，我们使用了2020年7-9月份覆盖亚洲高山区所有的无云Sentinel-2和Landsat-8光学数据以及Sentinel-1 SAR数据。利用冰川特有光谱特性、极化特性和地形特征，基于VGG16-Unet卷积神经网络进行冰川自动化提取，精度可高达92.8%。对自动化结果基于已有的冰川编目进行分割，并且在光学影像覆盖区域以Sentinel-2为主、Landsat-8为辅逐个进行人工修正，共修正了93043条冰川。最终得到2020年亚洲高山区大于0.01km2冰川共97934条，面积为91837.77 km2 ± 2527.7 km2，误差占总冰川面积的2.8%。并且提供了每条冰川的几何和地形属性。该数据集将为研究全球变化大背景下亚洲高山区冰川学与气候变化、水资源管理等研究热点提供重要的数据支撑。

The High Mountain Asia (HMA) hosts the largest and most extensive mountain glaciers in the middle and low latitudes globally, and is a critical component of the "Asian Water Tower". The temporal span of remote sensing data sources used in existing regional glacier inventories can affect quantitative estimations in glaciological research. To obtain an up-to-date and temporally concentrated glacier inventory, we utilized all cloud-free optical Sentinel-2 and Landsat-8 data, as well as Sentinel-1 SAR data, covering High Mountain Asia from July to September 2020. Leveraging the unique spectral, polarization and topographic characteristics of glaciers, we conducted automated glacier extraction based on the VGG16-Unet convolutional neural network, achieving an accuracy as high as 92.8%. The automated extraction results were segmented based on existing glacier inventories, and manual corrections were carried out individually in optical image coverage areas, prioritizing Sentinel-2 data and supplemented by Landsat-8 data, with a total of 93,043 glaciers corrected. Ultimately, we obtained a total of 97,934 glaciers with an area larger than 0.01 km² in High Mountain Asia in 2020, with a total area of 91837.77 km² ± 2527.7 km², with the error accounting for 2.8% of the total glacier area. Geometric and topographic attributes for each individual glacier are also provided. This dataset will provide critical data support for key research topics such as glaciology, climate change and water resource management in High Mountain Asia under the context of global change.

搜集汇总

背景与挑战

背景概述

该数据集是2020年亚洲高山区冰川的详细编目，基于2020年7-9月的多源遥感数据（包括Sentinel-2、Landsat-8和Sentinel-1），通过自动化提取（精度达92.8%）和人工修正，共收录97934条面积大于0.01 km²的冰川，总面积约91837.77 km²，误差率为2.8%，并提供了每条冰川的几何和地形属性，为冰川变化、气候变化和水资源管理研究提供关键数据支撑。