Multi-Scale Mapping and Monitoring of Permafrost Conditions in a Canadian High Arctic Polar Desert

Abstract:

With Arctic temperatures rapidly increasing, permafrost landscapes are faced with the potential for accelerated thawing and thermokarst (ground subsidence). In the face of these changes, it is critical to continually monitor Artic landscapes. However, the remoteness and vastness of Arctic sites can make high spatial and spectral resolution data collection difficult, and therefore monitoring problematic. To comprehensively evaluate the ground surface conditions in such environments, we propose a multi-scale remote sensing framework, combining medium resolution multispectral satellite imagery (Landsat 7 and 8) with high resolution unmanned aerial vehicle (UAV) RGB and thermal imagery, and field observations. This nesting of scaled data sources helps validate changes identified in the coarser satellite imagery, and also provides clues to future changes and potential thermokarst hot spots.

This methodological framework was applied to the Eureka Sound Lowlands (ESL) of Ellesmere Island, Nunavut, a landscape dominated by vast networks of ice wedge polygons (IWP), to characterize past and current permafrost conditions. Google Earth Engine was used to compute various indices (NDVI, NDMI, SAVI, TCG, TCW, TCB) over 20 years of Landsat imagery on a regional scale, showing trends of increasing greenness and wetness in the ESL. UAV RGB images were used to create high resolution (<1 cm) orthophotos and DEMs on a local or landform scale to inform and validate the coarser satellite imagery. Finally, the high resolution thermal images were used to assess the surface thermal dynamics of IWP. Overall, the combination of large area, coarse resolution satellite imagery with small area, high resolution UAV imagery provides a thorough assessment of permafrost conditions. This methodology effectively bridges the gap between ground data collection, which is typically accurate, but limited in scope and satellite remote sensing, which covers large spatial extents, but is limited in accuracy.