TerraSAR-X and time-lapse photography for snow on sea-ice monitoring in Deception Bay, Nunavik

• Main Authors: Dufour-Beauséjour, Sophie, Bernier, Monique, Wendleder, Anna, Poulin, Jimmy, Gilbert, Veronique, Tuniq, Juupi, Gauthier, Yves, Rouleau, Amélie
• Format: Conference Object
• Language: unknown
• Published: 2018
• Subjects: Dynamik der Landoberfläche; Arctic; Bare Rocks; Hudson; Hudson Strait; Kangiqsujuaq; Nunavik; Salluit; inuit; Sea ice
• Online Access: https://elib.dlr.de/124566/

Snow is an important element of safety on the land for Inuit living in the Arctic; land users may be stranded on bare rocks for lack of snow on the ground. Snow is necessary for snowmobile transport on sea ice trails, and is closely tied to the thickness ice may reach on average in a given winter. While an observation of less snow in recent years was reported by salluimmiut (J. Tuniq, ArcticChange 2017), climate projections for 2071-2100 feature an increase in precipitation over Nunavik (Ouranos, 2018). In this context, the Safe Passage research project (Polar Knowledge Research Program), which included Ice Monitoring in Salluit, Deception and Kangiqsujuaq, also focused on snow on sea ice. This presentation will report on snow information derived from three years of Monitoring in Deception Bay, Nunavik (Hudson Strait). Time-lapse cameras have been taking hourly pictures of the bay since December 2015 as well as measuring air temperature. This database was used to develop a proxy for estimating snow accumulation on the bay. TerraSAR-X high-resolution satellite images have been acquired every 11 days since December 2015 and were analyzed to identify seasonal trends in backscattering from snow over sea ice. Fieldwork was done twice per winter since 2016 and includes snow and ice thickness measurements. Time-lapse camera pictures (N ~ 26 000) were filtered by view (four different views are acquired each hour) and visibility using machine learning (python-Tensorflow). The foreground in the field-of-view is highly exposed, and the wind-swept rocks are regularly almost snow-free. Ground pixels were classified as either snow or no snow using computer vision (python-openCV) and an average of snow cover fraction was computed daily. Daily snow accumulation was modeled as the increase in snow pixels from one day to the next, or as zero in the absence of a snow cover increase. TerraSAR-X images were processed by the German space agency (DLR) with the MultiSAR software. Backscattering statistics were computed for land pixels ...