First Accuracy Assessment of Ground-Based GNSS-R for Coastal Arctic Sea Ice Freeze-Thaw Cycles Monitoring: A Five-Year Study (2018–2022) in Tuktoyaktuk, Canada

Under global warming, high-precision and rapid monitoring of Arctic sea ice freeze-thaw cycles has become increasingly critical for understanding polar climate dynamics and predicting global climate impacts. Ground-based Global Navigation Satellite System-Reflectometry (GNSS-R) is emerging as a promising technique for such monitoring, yet prior research has primarily focused on distinguishing sea ice from open water, with limited validation of its ability to capture continuous freeze-thaw transitions. To address this gap, this study presents a novel multi-frequency combination strategy that integrates spectral area factors (SAF) derived from multi-frequency (L1, L2, L5) GNSS-R observations using a Bayesian classifier. The method enhances detection by leveraging both state-dependent differential signatures and inter-frequency correlations. Using five years of observations (2018–2022) from the coastal station TUKT in Tuktoyaktuk, Canada, we trained prior probability distributions with data from 2018–2020 and tested the approach on independent data from 2021–2022. The results demonstrate that the proposed method effectively captures the dynamic progression of freeze-thaw cycles. It achieves a sample-level classification accuracy of 92.72% and a daily accuracy of 98.49% during the test period. This performance meets practical application requirements, confirming the potential of ground-based GNSS-R as a reliable, cost-effective tool for the sustained monitoring of coastal Arctic sea ice freeze-thaw processes. This study thereby bridges the critical gap between theoretical research and operational environmental decision-making in polar regions.