Assessing Fire-Induced Permafrost Ground Deformation and Environmental Drivers Using Interferometric Synthetic Aperture Radar and Machine Learning

Arctic fires in permafrost landscapes have become more frequent in recent years with record fire years in Eastern Siberia in 2019, 2020 and 2021. The consumption of organic layer and charring of the landscape can deepen active layer and long-term permafrost ground deformation. However, the factors driving post-fire ground deformation are not well understood. Here we applied the interferometric synthetic aperture radar (InSAR) technique on Sentinel-1 SAR imagery to map post-fire line of sight (LOS) ground deformation of five tundra fires in northeastern Siberia and used a machine learning XGBoost regression to better understand the environmental drivers. Applying InSAR technique on Sentinel-1 SAR imagery proved robust for mapping fire-affected permafrost ground deformation. For instance, in 2021, the permafrost ground affected by 2020 fires exhibited four times greater deformation in the growing season (5.42 cm) compared to adjacent unburned areas (1.42 cm). Surface albedo alone accounted for more than half of total explained variability, and albedo was correlated with fire severity and land surface temperature. Thus, variables that quantified fire impacts showed greater explanatory power than topographic variables. Finally, our remote sensing analysis was compared with some numerical modeling experiments. These simulations further affirmed the influence of surface albedo and burned organic layer on post-fire ground deformation. Our approach successfully quantified post-fire deformation in Yedoma-dominated permafrost landscapes and revealed environmental drivers that govern spatial variability in deformation within burned areas. Our methods could be applied at circumpolar scale to better understand the impacts of emerging fires in permafrost landscapes.