Application of GNSS Interferometric Reflectometry for the Estimation of Lake Ice Thickness

Abstract:

Global Navigation Satellite Systems-Interferometric Reflectometry (GNSS-IR) has shown potential in the remote sensing of geophysical and environmental parameters. However, this technique has not yet been practically evaluated for the retrieval of lake ice thickness, a sensitive indicator of climate change that has wide-ranging impacts of socio-economic activities and under-ice ecosystem dynamics. In this study, we report on results from GNSS-IR experiments conducted at sub-Arctic and mid-latitude lake locations during the ice season. GPS antennas were set up to receive reflected signals to estimate ice thickness based on the dominant frequency of signal-to-noise ratio (SNR) of reflected signals. For the sub-Arctic experiment, which took place in the Northwest Territories, Canada, the antenna was set up directly on the ice surface at 14 lake sites (March 2017 and 2019), whereas for the mid-latitude lake location (MacDonald Lake, Ontario, Canada) a GPS tower (5m height) was installed to provide time series of reflected signals over the ice season of 2019-2020. In the case of the sub-Artic lake sites, the main reflective surface was determined to be from the ice-water interface and, therefore, the antenna phase centre height could be used to estimate ice thickness. Ice thickness estimates compared favourably with in-situ measurements at these sites (correlation of 0.66, root mean square error (RMSE) of 0.07 m, and mean average error (MAE) of 0.05 m). In contrast, at the mid-latitude location, the main reflective surface was found not to always be from the ice-water interface due to air temperature fluctuations conducive to the formation of wet layers such as slush and wet snow. Here, less dominant frequencies of the SNR were obtained using least-squares harmonic estimation (LS-HE) to retrieve antenna heights from multiple reflective layers. Ice thickness estimates were in general agreement with manual measurements and output from a thermodynamic lake ice model (correlation of 0.68 and RMSE of 0.07 m).