Increased Arctic winter Rain-on-snow-induced melt events detected by spaceborne passive microwave remote sensing

Abstract

As the Arctic warms, more rain is falling on snow in winter, a phenomenon known as rain-on-snow (ROS) events. The ROS-induced snowmelt can alter snow properties, potentially leading to destructive floods, avalanches, and mass die-offs of ungulates. Passive microwave (PMW) satellite data shows promise in providing a pan-Arctic view of melt events but faces challenges in differentiating ROS melt from non-ROS melt events. This study developed a machine-learning based method to detect rainfall-induced (> 10 mm) ROS melt events in winter (December- January-February, DJF) using PMW data at 19 and 37 GHz, along with ancillary in-situ rainfall datasets and MODIS snow products. The method efficiently classifies ROS melt events, non-ROS melt events, and frozen snowpack, achieving an F-score ≥ 0.74 and a Kappa coefficient ≥ 0.74. Our result suggests winter ROS melt events were rare across the pan-Arctic, averaging fewer than 2.5 days per season (with regional maxima of ~4.5 days) over 2000-2023, yet the frequency of winter ROS melt events has significantly increased by 75% (p < 0.05) over this period. Climate linkage analyses show that large-scale atmospheric circulations account for most of the variability in winter ROS melt events. In North America and Asia, the AMO primarily influenced the variability of winter ROS melt events, while in Europe, the AO was the dominant factor. In particular, atmospheric river events transport warm, moist air poleward, increasing cloud cover and downward longwave radiation, which can trigger extreme winter ROS events. Our research presents an efficient remote sensing method for identifying winter ROS melt events across the pan-Arctic, enabling the assessment of high-latitude environmental changes and the rapid tracking of extreme events.