干旱区研究 ›› 2024, Vol. 41 ›› Issue (6): 940-950.doi: 10.13866/j.azr.2024.06.04 cstr: 32277.14.j.azr.2024.06.04

• 水土资源 • 上一篇    下一篇

基于Sentinel-2的依连哈比尔尕冰川变化监测

李若楠1,2,3(), 李均力1,3(), 李爽爽1,4, 刘帅琪1,2,3, 都伟冰4   

  1. 1.中国科学院新疆生态与地理研究所,荒漠与绿洲生态国家重点实验室,新疆 乌鲁木齐 830011
    2.中国科学院大学,北京 100049
    3.新疆遥感与地理信息系统应用重点实验室,新疆 乌鲁木齐 830011
    4.河南理工大学测绘与国土信息工程学院,河南 焦作 454003
  • 收稿日期:2023-10-23 修回日期:2024-01-30 出版日期:2024-06-15 发布日期:2024-07-03
  • 作者简介:李若楠(2001-),女,硕士研究生,主要从事干旱区冰川变化研究. E-mail: 1595499117@qq.com
  • 基金资助:
    新疆维吾尔自治区自然科学基金(2023D01E18);天山英才科技创新团队(2022TSYCTD0006);第三次新疆综合科学考察(2021xjkk1400)

Monitoring the glacier changes in Yilian Habirga Mountain using Sentinel-2 data

LI Ruonan1,2,3(), LI Junli1,3(), LI Shuangshuang1,4, LIU Shuaiqi1,2,3, DU Weibin4   

  1. 1. State Key Laboratory of Desert and Oasis Ecology, Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences, Urumqi 830011, Xinjiang, China
    2. University of Chinese Academy of Sciences, Beijing 100049, China
    3. Key Laboratory of GIS & RS Application Xinjiang Uygur Autonomous Region, Urumqi 830011, Xinjiang, China
    4. College of Survey and Territory Information Engineering, Henan Polytechnic University, Jiaozuo 454003, Henan, China
  • Received:2023-10-23 Revised:2024-01-30 Published:2024-06-15 Online:2024-07-03

摘要:

高分时序遥感对于监测冰川变化具有重要作用。本文利用2016—2022年Sentinel-2多时相卫星影像和D-UNet语义分割模型提取依连哈比尔尕冰川变化信息,并与时相相近的Landsat遥感数据的提取结果对比,比较Sentinel-2和Landsat在冰川制图的精度差异,在此基础上选择75条典型冰川分析近期研究区冰川总面积和冰川末端的变化特征。结果表明:(1) Sentinel-2冰川制图总体精度为95.0%,相同条件下比Landsat-8提高5%~10%。(2) 2016—2022年研究区冰川年平均面积退缩率为0.75%±0.69%,其中,海拔4600 m以下的区域为冰川面积减少的区域,海拔越低面积退缩率越大。(3) 近6 a 75条典型冰川末端的平均高度上升了17.75 m,长度平均退缩了11.39±2.36 m·a-1,其中,偏西、东北和南的退缩最为显著,分别为15.49±2.36 m·a-1、13.95±2.36 m·a-1和13.14±2.36 m·a-1,冰川末端退缩速率随海拔的升高而降低。

关键词: 冰川末端, 深度学习, 时空特征, Sentinle-2, 依连哈比尔尕山

Abstract:

High-resolution time-series remote sensing plays a vital role in monitoring glacier changes. In this paper, Sentinel-2 multitemporal satellite images from 2016-2022 were used along with the D-UNet semantic segmentation model to extract the glacier change information of Yilian Habirga. These results were compared with the Landsat remote sensing data of the similar temporal phase to ascertain any differences in the accuracies of Sentinel-2 and Landsat for glacier mapping. Based on these findings, 75 typical glaciers were selected to analyze the change-related characteristics of the total glacier area and glacier end in the recent study area. The results show that (1) The overall accuracy of Sentinel-2 glacier mapping was 95.0%, which is 5%-10% higher than Landsat-8 under the same conditions. (2) The average area retreat rate of glaciers in the study area from 2016 to 2022 was 0.75%±0.69%·a-1, in which the region<4600 m above sea level was that of glacier area reduction; the lower the altitude, the greater the area retreat rate. (3) In the last 6 years, the average heights of the 75 typical glacier ends rose by 17.75 m, and the average lengths reduced by 11.39 ± 2.36 m·a-1. Among these, the retreats in the west, northeast, and south were the most significant, which were 15.49 ± 2.36 m·a-1, 13.95 ± 2.36 m·a-1, and 13.14 ± 2.36 m·a-1, respectively; the rate of the glacier end retreated with an increase in the elevation and the decreased.

Key words: glacier terminus, deep learning, spatiotemporal variation, Sentinle-2, Yilian Habirga Mountain