Rain-on-snow (ROS) events occurring in the Arctic can form dense ice layers, preventing ungulates such as muskoxen and caribou from foraging throughout the winter. Such snow conditions are known as locked pastures as the vegetation is trapped beneath the dense ice layer. These locked pastures can have severe consequences for Arctic wildlife populations and Indigenous communities that depend on them. This study (1) assessed how well two meteorological reanalyses, ERA5 and the Canadian Surface Reanalysis (CaSR) v3.2, represent the meteorological conditions driving ROS events, and (2) evaluated whether the multi-physics snowpack model Crocus can reproduce the resulting locked-pasture stratigraphy. The reanalyses were evaluated against three local weather stations on Banks Island in the Canadian Arctic, for near-surface air temperature, wind speed, relative humidity, and precipitation. Four methods to determine the precipitation phase were also compared against weather observations. CaSRv3.2 paired with an atmospheric-profile-based partitioning method reproduced the timing and phase of a ROS event in 2003 most accurately and was used to drive SVS2-Crocus in an ensemble composed of four water holding capacity equations, three snow-compaction schemes, three falling snow density representations, and three wind packing variations. The ensemble reproduced a locked pasture consistent with reported snow conditions in 80 of 108 members of the ensemble for the 2003–2004 winter. Sensitivity experiments that removed either the rainfall or the accompanying warm temperatures showed that above-freezing temperatures driving melt-freeze events were the dominant driver of snowpack densification, although rainfall increased the probability of a locked pasture. A 35-year (1985–2020) simulation validated against remote sensing ROS detections identified accurately four other winters that were subject to a ROS event in late fall that resulted in a dense basal snow layer. These results show that physically based snowpack models can capture and help explain the formation of ROS-driven locked pastures once meteorological reanalysis biases are identified.