Widespread loss of safe lake ice access in response to a warming climate

context because people often use lake ice in the early winter and late spring for recreation, when a majority of incidents take place [15]. As with ice phenology and duration, air temperature plays a large role in ice cover develop- ment and melt. Air temperature is the primary driver of lake ice formation, growth, and melt, and increasing global air temperatures will shift lake ice timing and diminish thickness [1,39,40]. While the magnitude of the trends in ice phenology and duration will vary based on the study, data used, and model type, the current most comprehensive collection of global lake ice records, extending from 1846 to 2019, indicates that warming air temperatures have resulted in delayed ice formation of 11.0 days and advanced breakup of 6.8 days [41]. Even while ice cover is diminishing in duration throughout the Northern Hemisphere, we found that warming lengthened the amount of time that ice cover is unsafe for use, even if it is present. While ice cover is thickened by snow and lakes in regions with increased snow cover can show signs of increasing ice thickness [42], we found that air temperature governed the dura- tion of the transition periods. When comparing CESM2-LE modeled results for air tempera- ture, snow depth, and lake ice thickness, we found that thickness is overwhelmingly affected by air temperature rather than snow cover (S2 Fig). In areas throughout the Northern Hemi- sphere that are warming more, ice thickness decreased. The spatial patterns of unsafe transi-

tion period anomalies followed that behavior (Fig 3). Transition periods tended to increase most in areas where warming is expected, such as lower latitudes and coastal regions [1,16]. In fact, our derived transition period anomalies followed similar spatial patterns to both studies Fig 4. Distributions of unsafe transition periods. A split violin plot that illustrates the distribution of the transition period (in days) for warming scenarios of 1˚C, 2˚C, and 4˚C. The yellow shading demonstrates the ice formation period and the red shading represents the melt period. The boxplot within each distribution shows the quantiles, median (black point), and outlier values. Each grouping of distributions represents the ice quality for each set of warming scenarios, from left to right: Black ice, a 50/50 ice scenario, and white ice. Each average value was statistically significant (Kruskal-Wallis, p < 0.05, see S2 and S3 Tables for more details) from one another within ice quality categories and across ice quality categories for both ice formation and melt. Additionally, each distribution was significantly different from one another (Kolmogorov-Smirnov, p < 0.05, see S6 and S7 Tables for more details) across all ice quality categories and warming scenarios. https://doi.org/10.1371/journal.pone.0313994.g004 PLOS ONE Changing ice conditions inhibit safe use PLOS ONE | https://doi.org/10.1371/journal.pone.0313994 December 11, 2024 11 / 19