Empirical classification of dry-wet snow status in Antarctica using multi-frequency passive microwave observations

Abstract. Passive microwave satellite observations are commonly used to detect liquid water in the snowpack on the ice sheet. Typically, algorithms yield a binary dry-wet indicator limiting the information. Theoretical analyses have been demonstrated that these dry-wet indicators correspond to different levels in the snowpack depending on the frequency: from surface to ∼ 0.2 m at 37 GHz, from surface to ∼ 1 m at 19 GHz and from surface to depths exceeding 1 m at 1.4 GHz. In this study, our objective is to enhance understanding of melting and refreezing processes in Antarctica. For this, we proposed an empirical method that combines several binary dry-wet indicators computed at three frequencies (1.4, 19, and 37 GHz) and for two acquisition times (afternoon/night). We also introduced another indicator to estimate if most of the pixel (> 80 %) is subject to melt. By combining these six binary indicators, we obtained 64 possible daily “dry-wet signatures”, which were interpreted to infer whether the snowpack was dry, actively melting, or only wet below the surface, if night refreezing was occurring, and if a large proportion of the pixel was impacted. 99.6 % of the examined pixels show a consistent and physically meaningful daily dry-wet signature across Antarctica during the 2012–2023 considered period. To synthesise the 64 dry-wet signatures, we grouped the signatures conveying similar information into 10 qualitative classes of “snowpack status”. This new classification reveals a clear relationship between the various snowpack status and average surface temperature from ERA5 reanalysis, demonstrating the reliability of the empirical definition of the 10 classes. Furthermore, the classification captures the expected seasonal melt evolution: night refreezing is frequent at the beginning of the melt season, while sustained melting is observed in the middle of the summer, and remnant liquid water at depth features the end of the melt season. In the Antarctic Peninsula, over 11 years, we found an increasing trend in melting, significantly related to an increase in remnant liquid water at depth and a decrease in nighttime refreezing. This new classification offers deeper insights in melt processes for investigating extreme events and climate variations compared to previous binary indicators.