A multi-frequency SAR polarimetric analysis of British Columbia seasonal snowpack

Snow cover in the mountain ranges of Canada has great social economical and environmental impacts on Canadians. It is one of the main drivers for winter tourism, attracting tourists both nationwide, and internationally to ski resorts, visit National Parks, and snowmobile various areas. However, the mountains also include avalanche terrain, and if the conditions are met, snow accumulations can often result in avalanches, which can be either naturally or humanly triggered. Avalanches are a deadly hazard to recreationists, but they can also severely damage infrastructures, such as roads, railways, or even habitats (Naaim et al. 2013). Moreover, this seasonal snow will melt during spring, where the resulting runoff will feed rivers, and eventually will be used by a variety of services, from hydroelectricity to freshwater resource for cities and agriculture (Viviroli et al. 2007). Finally, fast melting events can create a rapid level rise of the catchment stream network, causing flash floods in the connected valleys (Wever et al. 2017). Monitoring the snowpack dynamics though the winter season would be highly beneficial, in an era of constant warming that leads to observed negative anomalies of snow cover extent, duration and mass.
\n
\nCollecting snow information through manual observations over such a wide and difficult terrain to access remains seriously challenging and logistically expensive. To meet this gap, modeling efforts are in progress in our lab, but the implementation of a modeling approach strongly depends on our ability to provide information across various terrain characteristics and climate types. Remote sensing techniques could provide valuable information about the snowpack, and Synthetic Aperture Radar (SAR) satellites have been successfully used to retrieve snow parameters already (Shi and Dozier 2000; Rott et al. 2009; Lievens et al. 2019). As a result, the German and Canadian space agencies SAR satellites TerraSAR-X (TSX), RADARSAT-2 (RS2), and RADARSAT Constellation Mission (RCM) could provide some valuable high-resolution information regarding the snowpack.
\n
\nIn this study, a time series of 10 TerraSAR-X images in orbit stripNear_007 mode, 9 RS2 SLC Wide Fine Quad-Pol images at two beam angles (FQ10W and FQ12W), and a time series of 12 RCM images at two beam angles (5MCP10 and 5MCP16) were acquired over Glacier National Park, BC, between January and May 2020. Images from both TSX (orbits stripNear_007 stripFar_001, 11 days return period) and RCM (5MCP16, 12 days return period) are being acquired for the 2020-2021 winter season as well. A bulk processing framework was implemented using the polarimetric toolbox provided by PCI Geomatica algorithms, and several polarimetric discriminators were computed. First, images were radiometrically calibrated, then an adaptive Lee Filter with a window of 7x7 pixels was applied. From there, Copolar Coherence was computed to retrieve its magnitude and phase (e.g. Copolar Phase Difference). For RCM images, wave coherence, relative phase, ellipticity, degree of polarization (DoP), degree of Linear Polarization (DoLP), and degree of circular polarization (DoCP) were also computed. All image processing was done in the SAR topology, end-products were then reprojected in WGS84 coordinates using a Digital Elevation Model of the area.
\nSimultaneously to satellite measurements, in-situ data was acquired at the Mt Abbott Automatic Weather Station (2084m a.s.l.), enabling simulations of the seasonal evolution of the snowpack using the SNOWPACK model (Bartelt and Lehning 2002; Lehning et al. 2002). From there, anisotropy of the snowpack was computed, using the anisotropy model of Leinss et al. (2020). The polarimetric response of both the compact-pol signal and quad-pol signal was put in perspective with snow properties and SAR acquisition parameters at Mt Abbott study site. For reference, the study site is in alpine environment (e.g. no trees on the site) and has a very gentle slope.
\n
\nThe modelling of the snowpack revealed that the layers were primarily horizontally structured; vertical structures appearing at the bottom of the snowpack towards the end of the season. The depth of the snowpack, associated with relatively mild temperatures resulted in a low temperature gradient, causing almost no temperature gradient metamorphism. Thus, available data suggests that gravitational settling was then the principal driver for snow metamorphism, hence a vast majority of horizontal structures in the snowpack.
\n
\nFirst, the magnitude of the Copolar Coherence (CCOH) at the study site was explored with regards to the local incidence angle (LIA) for the RS2 and TSX data, as well as the received wave coherence for RCM data. Overall, CCOH showed an important dependency to LIA, with magnitudes around 0.7 at LIA = 22° for RS2 FQ12W data, and around 0.4 at LIA = 44° for RS2 FQ10W and TSX data. This suggests strong volume depolarization, even at C-Band, even though scatterers are significatively smaller than the wavelength. Furthermore, the relationship between Copolar Phase Difference (CPD) with snow depth and modeled snowpack anisotropy was analyzed. For FQ12W data (LIA = 22°), CPD showed a good correlation with snow depth (R2 = 0.94, p-value = 0.04) over the season. However, the relationship is stronger with the total height of horizontally structured layers in the snowpack (R2 = 0.96, p-value = 0.04). For TSX, data showed a rather good relationship with the height of new snow which fell between acquisitions (R2 = 0.62, p-value = 0.1).
\nPreliminary results from RCM data showed that the received wave presents a coherence magnitude around 0.25 for 5MCP16 beam mode, and 0.4 for 5MCP10 beam mode. Moreover, coherence for 5MCP10 data is, on average, slowly decreasing throughout the season. This is probably due to an augmentation of depolarization with the increase of the depth of the snowpack. Finally, the study of the different DoPs over the season is suggesting that the snowpack is acting like an imperfect linear polarizer, half of the received wave being linearly polarized, the other half being unpolarized.
\n
\nOverall, this study aims to provide insights on the processing, analysis, and challenges of using polarimetric SAR imagery for observing snow in a real mountainous environment.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.