Scholarlink

Mountain glaciers, crucial components of Earth’s climate system, are experiencing an accelerated loss of mass primarily due to global warming, leading to notable environmental and socioeconomic effects. In Peru, glaciers have shrunk by 53% in the last 50 years. This underscores the importance of continuous glacier monitoring, and deep-learning techniques offer innovative possibilities to enhance precision and effectiveness. The study sought to develop a comprehensive methodological framework for creating a new data set specifically designed to map glaciers in the Cordillera Blanca and Cordillera Vilcabamba of the Peruvian Andes. This method integrates Sentinel-2 multispectral satellite imagery, spectral indices, and ASTER GDEM digital elevation models, used to segment clear and debris-laden glaciers. In an initial demonstration, various established deep learning models were evaluated, confirming the new dataset’s effectiveness in glacier segmentation. This research advances the application of remote sensing techniques related to glacier studies, providing advanced tools to observe the impacts of climate change in tropical mountain regions.

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

Methodological Approach for the Development of a Mountain Glacier Dataset Using Satellite Imagery for Deep Learning Applications

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • William Condori-Quispe,
  • Jose Herrera-Quispe,
  • Oscar Jimenez-Flores,
  • Juan Jimenez-Castilla

摘要

Mountain glaciers, crucial components of Earth’s climate system, are experiencing an accelerated loss of mass primarily due to global warming, leading to notable environmental and socioeconomic effects. In Peru, glaciers have shrunk by 53% in the last 50 years. This underscores the importance of continuous glacier monitoring, and deep-learning techniques offer innovative possibilities to enhance precision and effectiveness. The study sought to develop a comprehensive methodological framework for creating a new data set specifically designed to map glaciers in the Cordillera Blanca and Cordillera Vilcabamba of the Peruvian Andes. This method integrates Sentinel-2 multispectral satellite imagery, spectral indices, and ASTER GDEM digital elevation models, used to segment clear and debris-laden glaciers. In an initial demonstration, various established deep learning models were evaluated, confirming the new dataset’s effectiveness in glacier segmentation. This research advances the application of remote sensing techniques related to glacier studies, providing advanced tools to observe the impacts of climate change in tropical mountain regions.