Bedfast and floating-ice dynamics of thermokarst lakes using a temporal deep-learning mapping approach: case study of the Old Crow Flats, Yukon, Canada

Abstract

Abstract. In light of the recent climate warming, monitoring of lake ice in Arctic and subarctic regions is becoming increasingly important. Many shallow Arctic lakes and ponds of thermokarst origin freeze to the bed in the winter months, maintaining the underlying permafrost in its frozen state. However, as air temperatures rise and precipitation increases, fewer lakes are expected to develop bedfast ice. In this work, we propose a novel temporal deep-learning approach to lake ice regime mapping from synthetic aperture radar (SAR) and employ it to study lake ice dynamics in the Old Crow Flats (OCF), Yukon, Canada, over the 1992/1993 to 2020/2021 period. We utilized a combination of Sentinel-1, ERS-1 and ERS-2, and RADARSAT-1 to create an extensive annotated dataset of SAR time series labeled as either bedfast ice, floating ice, or land, which was used to train a temporal convolutional neural network (TempCNN). The trained TempCNN, in turn, allowed us to automatically map lake ice regimes. The classified maps aligned well with the available field measurements and ice thickness simulations obtained with a thermodynamic lake ice model. Reaching a mean overall classification accuracy of 95 %, the TempCNN was determined to be suitable for automated lake ice regime classification. The fraction of bedfast ice in the OCF increased by 11 % over the 29-year period of analysis. Findings suggest that the OCF lake ice dynamics are dominated by lake drainage events, brought on by thermokarst processes accelerated by climate warming, and fluctuations in water level and winter snowfall. Catastrophic drainage and lowered water levels cause surface water area and lake depth to decrease and lake ice to often transition from floating to bedfast ice, while a reduction in snowfall allows for the growth of thicker ice. The proposed lake ice regime mapping approach allowed us to assess the combined impacts of warming, drainage, and changing precipitation patterns on transitions between bedfast and floating-ice regimes, which is crucial to understanding evolving permafrost dynamics beneath shallow lakes and drained basins in thermokarst lowlands such as the OCF.

Using RADARSAT constellation mission imagery to support talik mapping, Rankin Inlet, Nunavut, Canada

“…In addition, the waterbody mask may not align perfectly with the current lakeshores and contribute to land being classified as FI or BFI. As for the isolated patches of BFI classed in the middle of deeper lakes, these have been observed in other studies(Duguay and Wang 2019;Shaposhnikova et al 2023). Overestimation of BFI in these areas remains a challenge and is possibly due to ice cracking or local thinning(Pointner et al 2019; Pointner and Barsch 2020).…”

“…Airborne and satellite SAR imagery have been applied to shallow lakes to get knowledge on lake ice regime and thickness, lake depth, and indirectly to assess permafrost and thermokarst processes, methane gas emissions, overwintering fish habitat, and freshwater availability for regions in Alaska, northern Manitoba, Mackenzie Delta, and Russia, and at the circumpolar scale (Murfitt and Duguay 2021). Threshold-based methods using C-band co-polarized signals (HH and VV) are the most common approaches used to classify FI and BFI (van der Sanden et al 2012;Engram et al 2018;Murfitt and Duguay 2021;Shaposhnikova et al 2023). These methods use the difference in backscatter between FI and BFI to identify threshold(s) capable of differentiating between the two types of ice cover.…”

X- and Ku-Band SAR Backscattering Signatures of Snow-Covered Lake Ice and Sea Ice

“…For sea ice, the total σ o typically consists of surface scattering from the snow surface, volume scattering from the snow layer, surface scattering from the ice surface and scattering from the air bubbles and brine pockets in the ice medium. For lake ice, there is more volume scattering from the lower layers of the ice; surface scattering from internal layers in the ice, scattering from the ice-water surface and double-bounce scattering from the bottom bubbled ice layer are included as well [56,57]. In recent studies, the scattering caused by the roughness of the ice-water interface has been found to be the primary contribution to the total scattering over lake ice and not the double bounce from air bubbles, as has been assumed previously [46,58].…”

X- and Ku-Band Sar Backscattering Signatures of Snow-Covered Lake Ice and Sea Ice

“…For sea ice, the total σ o typically consists of surface scattering from the snow surface, volume scattering from the snow layer, surface scattering from the ice surface and scattering from the air bubbles and brine pockets in the ice medium. For lake ice, there is more volume scattering from the lower layers of the ice, surface scattering from internal layers in the ice, scattering from the ice-water surface and double-bounce scattering from the bottom bubbled ice layer are included as well [42,43]. In recent studies, the scattering caused by the roughness of ice/water interface has been found to be the primary contribution to the total scattering over lake ice and not the double bounce from air bubbles, as has been assumed previously [29,44].…”