Search for a command to run...
Abstract Sublimation is an important yet uncertain component of the water balance in snow-dominated regions. Models that parameterize sublimation as a function of temperature show that sublimation has increased over recent decades, and this has been identified as a possible reason for declining Colorado River streamflow. However, sublimation increases with wind speed and water vapor gradients and cannot be well explained by temperature alone. Thus, we hypothesize that most accumulation season sublimation occurs during distinct weather events. To test this hypothesis, we combined two years of meteorological observations spanning site-to-synoptic scales from three field campaigns in a Colorado mountain valley and identified the events that produced the mostsublimation, classifying them by duration and intensity. We found that approximately 60% of winter sublimation occurred during a small number of discrete events (14% of the season). These events coincided with either short sunny, dry periods or long storm events that brought new snowfall followed by blowing snow. Using these classified events and synoptic-scale reanalysis, we trained a random forest classifier to estimate long sublimation event occurrences and test whether they increased over the past four decades. The model performed well (average precision = 0.73; balanced accuracy = 79%), and results were most strongly associated with winter mean 500-hPa wind speed and total observed precipitation. Event occurrence varied considerably year to year with no significant long-term trend. In all, our findings show that accurately representing seasonal sublimation requires capturing discrete events driven by identifiable synoptic and local meteorological conditions.