Scholarlink

<p>Rock glaciers serve as vital indicators and reservoirs within mountain permafrost systems, but their mapping is challenging due to complex terrain and limited labelled data. In this study, we present a novel automated detection method that combines Google DeepMind’s pre-trained 64-dimensional satellite embeddings with machine learning classifiers within Google Earth Engine. Using expert-labelled and partially ground-truthed polygons, we evaluated classifiers including Random Forest, k-Nearest Neighbours, Gradient Tree Boosting, and Support Vector Machine over an area of 3900 Km<sup>2</sup> in the Southwestern Maritime Alps. The Support Vector Machine achieved the highest accuracy at 85.7% and demonstrated balanced performance with an F1-score of 0.80. The resulting rock glacier classified raster pixels covers 32&#xa0;km², with a post-processing inventory of 621 rock glaciers, consistent with regional knowledge and field validation. The embedding-based Support Vector Machine minimises the need for extensive feature engineering and large training datasets. It offers a data-efficient, reproducible framework for regional rock-glacier screening, with broader transferability to be validated through specific cross-region tests. This open-source method supplies an operational tool to enhance permafrost monitoring and climate change impact evaluations.</p> Graphical Abstract <p></p> <p>This study presents an automated rock glacier detection framework in Google Earth Engine, utilising Google DeepMind’s pre-trained satellite embeddings to locate periglacial landforms in the Southwestern Maritime Alps. The workflow employs 64-dimensional embedding vectors (Bands A00–A63), which integrate multi-sensor optical, radar, and topographic data as the primary dataset to train and evaluate four machine learning classifiers: Random Forest, k-Nearest Neighbours, Gradient Tree Boosting, and Support Vector Machine. The Support Vector Machine (SVM) achieved the highest accuracy (85.7%) and balanced performance (F1-score 0.80) and produced an inventory of 621 rock glaciers using the post-processing pipeline. By leveraging pre-trained embeddings, this data-efficient and reproducible framework minimises the need for extensive feature engineering, providing an operational tool for regional-scale permafrost monitoring and a reusable template for transfer tests in other mountain regions.</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 Glacier Detection Using Satellite Embeddings and Machine Learning on Google Earth Engine

  • 期刊论文

  • 发表日期：

    2026年4月9日

查看全文
  • Varun Khajuria,
  • Shaktiman Singh,
  • Luca Paro,
  • Matteo Spagnolo,
  • Adriano Ribolini

摘要

Rock glaciers serve as vital indicators and reservoirs within mountain permafrost systems, but their mapping is challenging due to complex terrain and limited labelled data. In this study, we present a novel automated detection method that combines Google DeepMind’s pre-trained 64-dimensional satellite embeddings with machine learning classifiers within Google Earth Engine. Using expert-labelled and partially ground-truthed polygons, we evaluated classifiers including Random Forest, k-Nearest Neighbours, Gradient Tree Boosting, and Support Vector Machine over an area of 3900 Km2 in the Southwestern Maritime Alps. The Support Vector Machine achieved the highest accuracy at 85.7% and demonstrated balanced performance with an F1-score of 0.80. The resulting rock glacier classified raster pixels covers 32 km², with a post-processing inventory of 621 rock glaciers, consistent with regional knowledge and field validation. The embedding-based Support Vector Machine minimises the need for extensive feature engineering and large training datasets. It offers a data-efficient, reproducible framework for regional rock-glacier screening, with broader transferability to be validated through specific cross-region tests. This open-source method supplies an operational tool to enhance permafrost monitoring and climate change impact evaluations.

Graphical Abstract

This study presents an automated rock glacier detection framework in Google Earth Engine, utilising Google DeepMind’s pre-trained satellite embeddings to locate periglacial landforms in the Southwestern Maritime Alps. The workflow employs 64-dimensional embedding vectors (Bands A00–A63), which integrate multi-sensor optical, radar, and topographic data as the primary dataset to train and evaluate four machine learning classifiers: Random Forest, k-Nearest Neighbours, Gradient Tree Boosting, and Support Vector Machine. The Support Vector Machine (SVM) achieved the highest accuracy (85.7%) and balanced performance (F1-score 0.80) and produced an inventory of 621 rock glaciers using the post-processing pipeline. By leveraging pre-trained embeddings, this data-efficient and reproducible framework minimises the need for extensive feature engineering, providing an operational tool for regional-scale permafrost monitoring and a reusable template for transfer tests in other mountain regions.