Machine learning-based predictions of current and future susceptibility to retrogressive thaw slumps across the Northern Hemisphere

Abstract

Retrogressive thaw slumps (RTSs) caused by the thawing of ground ice on permafrost slopes have dramatically increased and become a common permafrost hazard across the Northern Hemisphere during previous decades. However, a gap remains in our comprehensive understanding of the spatial controlling factors, including the climate and terrain, that are conducive to these RTSs at a global scale. Using machine learning methodologies, we mapped the current and future RTSs susceptibility distributions by incorporating a range of environmental factors and RTSs inventories. We identified freezing-degree days and maximum summer rainfall as the primary environmental factors affecting RTSs susceptibility. The final ensemble susceptibility map suggests that regions with high to very high susceptibility could constitute (11.6 ± 0.78)% of the Northern Hemisphere's permafrost region. When juxtaposed with the current (2000–2020) RTSs susceptibility map, the total area with high to very high susceptibility could witness an increase ranging from (31.7 ± 0.65)% (SSP585) to (51.9 ± 0.73)% (SSP126) by the 2041–2060. The insights gleaned from this study not only offer valuable implications for engineering applications across the Northern Hemisphere, but also provide a long-term insight into the potential change of RTSs in permafrost regions in response to climate change.

Keywords

• Retrogressive thaw slump
• Machine learning
• Susceptibility map
• Permafrost
• Northern Hemisphere