The Potential of Artificial Intelligence and Remote Sensing for Cryospheric Research

Abstract

Recent advances in artificial intelligence, especially in the field of deep learning, have allowed new insights into cryospheric systems. Nowadays, an abundance of satellite imagery, new developments in deep learning algorithms and easy accessibility to computational power enable new potentials for data processing and analysis. Here, we present a variety of deep learning applications for cold and polar regions providing new possibilities for observing and monitoring the cryosphere. The presented examples cover a wide range of applications such as mapping retrogressive thaw slumps in Arctic permafrost regions with high-resolution satellite imagery based on a UNet++ or the automated identification of the firn line in L-Band SAR data. Furthermore, methodologies for glacial lake mapping in the Himalayas with the GLNet and the detection of supraglacial lake dynamics in Antarctica based on optical and SAR satellite data will be introduced. Additionally, we address the automated extraction of calving fronts in Greenland and Antarctica providing new understandings of glacier and ice shelf front dynamics in an unprecedented spatial and temporal resolution. Taking together these new potentials of artificial intelligence for cold and polar regions, we welcome discussions on how these techniques can be applied to other areas in cryospheric science and what challenges and limitations this might involve.