A mutlisensor C-band synthetic aperture radar (SAR) approach to retrieve freeze/thaw cycles: A case study for a low Arctic environment

Abstract. This study investigates the spatial variability of surface freeze/thaw (F/T) cycles in low arctic tundra retrieved from multisensor SAR backscatter time series. To increase the temporal resolution of SAR observations, we combined measurements from Sentinel-1 and RADARSAT-2. An incidence angle normalization was applied to the backscatter time series to remove the influence of the acquisition angle on backscatter. A seasonal threshold algorithm (STA) was used to detect F/T transitions and applied to HH, HV and HH+HV polarization datasets. The classification threshold was optimized using soil temperature measurements from spatially distributed sites. A detection accuracy of over 93 % was calculated with an optimized classification threshold of 0.62 for the HH+HV time series on those sites. We created surface F/T day of the year (DOY) maps of the study area for the 2018 and 2019 freezing transitions, and for the 2019 thawing transition using the HH+HV time series with the optimized classification threshold. Those maps were combined with a terrestrial ecosystem (ecotype) map to investigate the impact of ecotypes on the F/T transitions. Three generalized least squares (GLS) models were fitted on the coupling of the maps. Differences of about 2–3 days were observed between ecotype classes. Based on these differences, we hypothesize that differences during the freezing transition were probably due to the underlying soil moisture and during the thawing transition, to the influence of vegetation. Our study demonstrates the power of merging two C-band SAR time series to create near-daily F/T maps over arctic environment to allow for better understanding of surface F/T processes happening at small spatial scale in arctic environments.