Combined Use of Sentinel-1 and Sentinel-2 for Glacier Mapping: An Application Over Central East Alps

Abstract

The systematic monitoring of glaciers is essential to both evaluate water resource availability and better understand the effects of climate change. The increased speed of glacier changes observed in the past few years requires a more frequent update of the glacier inventories than in the past; however, the high human supervision required by the state-of-the-art techniques is discouraging their systematic application over large areas. This article proposes a novel approach to exploit the large volume of data provided by Copernicus Sentinel missions for detecting glacier outlines, including debris-covered glaciers. In detail, our method exploits the Sentinel-1 and Sentinel-2 multitemporal images to build a composite image representing the glacier conditions during the yearly maximum ablation period. The Sentinel-2 multispectral images are classified with a support vector machine and composed to a mosaic that represents the information of the maximum glacier ablation. At the same time, the Sentinel-1 time series are exploited to build a multitemporal coherence composite that represents all the snow-covered and glaciated areas together with all the moving surfaces. This information is used together with the Sentinal-2 composite to detect the debris-covered part of the glaciers. The proposed method was tested in the Central East Alps and presented an overall accuracy of 92% with respect to a reference inventory over South Tyrol and an agreement of 90% with respect to the latest glacier inventory of the Alps from Sentinel-2. The proposed approach enables to assist glacier experts in identifying glacier outlines over large areas and in short time

Keywords

• Copernicus
• data assimilation
• debris-covered glaciers
• glacier monitoring
• machine learning
• Sentinel-1