The sensitivity of modeled snow accumulation and melt to precipitation phase methods across a climatic gradient

Abstract. A critical component of hydrologic modeling in cold and temperate regions is partitioning precipitation into snow and rain, yet little is known about how uncertainty in precipitation phase propagates into variability in simulated snow accumulation and melt. Given the wide variety of methods for distinguishing between snow and rain, it is imperative to evaluate the sensitivity of snowpack model output to precipitation phase determination methods, especially considering the potential of snow-to-rain shifts associated with climate warming to fundamentally change the hydrology of snow-dominated areas. To address these needs we quantified the sensitivity of modeled snow accumulation and melt to rain-snow partitioning at research sites in the western United States. Simulations using the physics-based SNOWPACK model and 12 different precipitation phase methods indicated maritime sites were the most sensitive to method selection. Relative differences between the minimum and maximum annual snowfall fractions predicted by the different methods sometimes exceeded 100 % at elevations less than 2000 m in the Oregon Cascades and California’s Sierra Nevada mountains. This led to ranges in annual peak snow water equivalent (SWE) typically greater than 200 mm, exceeding 400 mm in certain years. At the warmer sites, ranges in snowmelt timing predicted by the different methods were generally larger than 2 weeks, while ranges in snow cover duration approached 1 month and greater. Conversely, the three coldest sites in this work were relatively insensitive to the choice of a precipitation phase method with average ranges in annual snowfall fraction, peak SWE, snowmelt timing, and snow cover duration less than 18 %, 62 mm, 10 d, and 15 d, respectively. Overall, sites with a greater proportion of precipitation falling at air temperatures between 0 °C and 4 °C exhibited the greatest sensitivity to method selection. These findings have large implications for modeled snowpack water storage and land surface albedo at the warmer fringes of the seasonal snow zone.