Arctic Sea Ice Seasonal Melt‐Freeze Transitions Observed by ASCAT From 2007 to 2022

Abstract Seasonal melt‐freeze transitions of Arctic sea‐ice play a vital role in Arctic climate patterns. They are detectable by satellite remote sensing. In this study, we present a sea ice type‐independent machine learning algorithm using Advanced Scatterometer (ASCAT) data at 4.45 km resolution, offering a means to complement melt‐freeze onset detection under uncertain sea ice type conditions. The underlying principle rests on the substantial variations in the backscatter coefficient during the melt season, enabling the identification of the melt onset (MO, the beginning of the melt season) and freeze onset (FO, the beginning of the freeze season). Comparisons of our ASCAT‐derived seasonal melt‐freeze transitions with those from passive microwave sensors (PMW) demonstrate that our results are robust, with the differences of three and two days in the average MO and FO, respectively, over the entire Arctic between the ASCAT and PMW. Incorporating buoy‐based air temperature data into the evaluation further provided supplementary insights and enhanced the understanding of microwave signal response to ice melt and freeze processes. Finally, our analysis reveals a pronounced interannual variability rather than statistically significant trends in the seasonal transition dates during the study period (2007–2022). Further investigation of atmospheric and oceanic circulation patterns is needed to understand the variability in Arctic sea‐ice melt and freeze onsets.