Scholarlink

<p>Rock glaciers are distinctive geomorphological features acting as indicators for climatic changes and permafrost occurrence. They store significant amounts of ice and are widespread in mountain ranges, making them potentially important for regional hydrology. However, existing inventories for the Southeastern Tibetan Plateau (STP) exhibit significant discrepancies and varying levels of completeness, highlighting the need for a systematic reassessment to overcome the limitations of previous manual or coarse-resolution mapping efforts. Here, we present a detailed rock glacier inventory covering 545,400 km<sup>2</sup> of STP, derived from 4.7-m resolution Planet Basemap imagery. The dataset was generated using a SegFormer deep learning model trained on a globally constructed dataset of 19,096 rock glacier samples, designed to maximize generalization across diverse terrains. The inventory was refined and quality-controlled through an inventory-wide human-in-the-loop workflow, including manual inspection of raw polygons, expert review of uncertain cases, and cross-verification with high-resolution imagery and existing datasets. In addition, Uncrewed Aerial Vehicle (UAV)-supported field surveys at three sites provided independent high-resolution reference examples for assessing boundary reliability. The final inventory contains 47,916 rock glaciers, significantly expanding previous inventories by over 20,000 entries. Each record includes comprehensive geometric attributes and topographic information. This dataset provides an important baseline dataset for monitoring permafrost dynamics, hydrological modeling, and geohazard assessment in the context of climate change.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Rock glaciers of the Southeastern Tibetan Plateau: A large-scale inventory derived from deep learning and high-resolution imagery

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年8月5日

查看全文
  • Xin Guo,
  • Xiaowen Wang,
  • Siqi Zhang,
  • Zhangyu Sun,
  • Lin Liu,
  • Guoxiang Liu,
  • Tobias Bolch

摘要

Rock glaciers are distinctive geomorphological features acting as indicators for climatic changes and permafrost occurrence. They store significant amounts of ice and are widespread in mountain ranges, making them potentially important for regional hydrology. However, existing inventories for the Southeastern Tibetan Plateau (STP) exhibit significant discrepancies and varying levels of completeness, highlighting the need for a systematic reassessment to overcome the limitations of previous manual or coarse-resolution mapping efforts. Here, we present a detailed rock glacier inventory covering 545,400 km2 of STP, derived from 4.7-m resolution Planet Basemap imagery. The dataset was generated using a SegFormer deep learning model trained on a globally constructed dataset of 19,096 rock glacier samples, designed to maximize generalization across diverse terrains. The inventory was refined and quality-controlled through an inventory-wide human-in-the-loop workflow, including manual inspection of raw polygons, expert review of uncertain cases, and cross-verification with high-resolution imagery and existing datasets. In addition, Uncrewed Aerial Vehicle (UAV)-supported field surveys at three sites provided independent high-resolution reference examples for assessing boundary reliability. The final inventory contains 47,916 rock glaciers, significantly expanding previous inventories by over 20,000 entries. Each record includes comprehensive geometric attributes and topographic information. This dataset provides an important baseline dataset for monitoring permafrost dynamics, hydrological modeling, and geohazard assessment in the context of climate change.