Can we estimate snow accumulation and melt across climates using simple temperature-index modelling?

Abstract

There are two main limitations to understanding large-scale impacts of environmental change on snow resources, 1) observational snow data at the point scale is highly limited, and 2) extrapolation using models can be challenging due to data availability and performance. This study seeks to address these limitations using widely available climate network data combined with a temperature-index snow model to derive large-scale estimates of mean snow water equivalent conditions across the Northern Hemisphere. Temperature-index modelling is a common approach for simulating snow accumulation and melt in hydrological models. Many studies use this method because of its simplicity, efficiency, and generally good performance if properly calibrated. The approach relies on three assumptions and parameters, namely the snowfall and snowmelt temperature thresholds and the degree-day factor. At scales beyond single gauged catchments, the estimation of these parameters was difficult to date due to a lack of observations on snowmelt. Using the new Northern Hemisphere snow water equivalent dataset (NH-SWE) and co-located climate network observations of temperature and precipitation, this work provides the first large-scale evaluation of temperature-index melt model assumptions and parameters across a diverse range of snow climates. Our study reveals the 0°C as snowfall air temperature threshold captures most snowfall events, especially in cold climates, but risks missing 13% of snowfall events, especially in climates hovering at near-freezing temperatures. Similarly, a snowmelt air temperature threshold of 0°C performs well for most daily snowmelt observations but may incorrectly identify the onset of the melt season too early. Estimated degree-day factors converge towards 3-5 mm/°C/day for deeper snowpack climates (> 300 mm), but their estimation may be more challenging for colder climates with shallower snowpacks (< 300 mm), conditions where the degree-day factors have much higher interannual variability. For estimating mean values of seasonal snow onset and snowmelt season onset and mean snow accumulation at a given location, the temperature-index melt model performs consistently well on average despite its simplicity, but challenges may arise due to warm biases in temperature records or solid precipitation undercatch, mainly over higher elevation areas. This study provides valuable insights into temperature-index melt modelling for large-scale applications, and the results should help refine modelling approaches to enhance our understanding of snowpack responses to global warming.