Monitoring permafrost degradation in northern Sweden using InSAR technique

Sammanfattning

Arctic permafrost plays an important role in global carbon storage and has been thawing due to rising temperatures imparted by climate change. Monitoring the active layer thickness (ALT), the seasonally thawing and refreezing uppermost permafrost layer, is essential for understanding permafrost dynamics and their response to warming temperatures. Traditional ALT monitoring relies on ground-based measurements from networks like the Circumpolar Active Layer Monitoring (CALM) program, which provides valuable permafrost data but suffers from limited spatial coverage. To address this, spaceborne Interferometric Synthetic Aperture Radar (InSAR) has emerged as a promising tool and demonstrated feasibility for large-scale ALT estimation, particularly in homogenous permafrost zones (> 50%) of North America. This study evaluated the performance of using InSAR to estimate ALT near Lake Torneträsk in Abisko, Sweden, a sporadic permafrost zone where permafrost coverage is spatially heterogeneous (< 50%). The study area is approximately 2750 km2 and includes CALM site S2, which contains six sample locations for in-situ validation data. Each of the sample locations covers an area of approximately 100 x 100 m. Synthetic Aperture Rader (SAR) images from Sentinel-1 and soil measurement data taken from Stordalen mire, an area in the study region, were used to produce InSAR-derived ALT estimates. These estimates were evaluated against ground truth ALT measurements for 2018 – 2023. Overall, the InSAR technique was able to estimate ALT values in the Abisko region, with results showing that ALT has been increasing from 2018 to 2023. Results showed an absolute bias of 0.10–0.40 m and a root mean squared error (RMSE) below the European Space Agency (ESA) Permafrost_CCI threshold of 0.25 m for four out of six of the sample locations. However, the results were only able to be validated for areas close to Lake Torneträsk, due to the limited distribution of the sample locations. The InSAR technique was able to closely estimate ALT in areas with similar soil characteristics as the sample locations, such as the wetland subsets of the study area, but the current model lacks the complexity to generate accurate estimates for different terrains. Future research should focus on expanding the model, allowing for its usage in areas with variable soil characteristics and topography. Assessments of InSAR-derived ALT estimates would benefit from being validated on areas with both a large amount of in-situ measurement data and variation in terrain, vegetation, and permafrost landcover types. This study also compared InSAR-derived ALT estimates to the recently released ESA product, Permafrost_CCI. This product estimates ALT at a resolution of 1 x 1 km using a model based on altitude, temperature, snow depth, and other climate drivers. The comparison revealed that these two estimation techniques largely follow the same pattern, especially in areas with wetland land cover, as is the case with Stordalen mire and the sample sites. The InSAR-derived estimates showed greater deviations in locations outside of the peat wetland areas, such as the mountains in the south of the study area, likely due to the soil characteristics used in model equations not accurately representing the diverging regions.