Improved workflow for customized ICESat-2 ATL06 elevations captures seasonal mountain snow depths at sub-kilometer scale
Abstract. Mountain snowpacks are a critical global water resource, providing freshwater to more than a billion people worldwide. Estimates of snow distribution across mountain watersheds are dependent on sparse in situ observations and spatio-temporally limited aerial surveys. Models expand snow depth estimation coverage, but are also limited by the amount of snow depth measurements available for data assimilation. Satellite remote sensing provides larger spatial and temporal coverage of observations. Recent studies have shown that NASA’s ICESat-2 satellite can resolve a mountain snow signal by comparing snow covered ICESat-2 returns to an independently collected snow-free digital terrain model (DTM). However, ICESat-2 snow depth uncertainties are estimated to be approximately on the order of snow depths in mountainous terrain, limiting their use. In this study we present a refined methodology for calculating ICESat-2 snow depth to minimize uncertainty in mountain environments and identify terrain with the most potential for ICESat-2 observation. We calculate snow depth by differencing a customized ICESat-2 product (ATL06_SR) and 1 m resolution DTMs of four mountainous sites in central and southern Idaho. Accuracy of ICESat-2 snow depths is assessed at 100 m–5000 m scales through comparison with in situ automatic weather station (AWS) data and six airborne lidar snow depth surveys. By co-registering to minimize geolocation errors, correcting for slope-related biases, and filtering negative snow depth observations, we find ICESat-2 snow depth uncertainties as low as 0.2 m and R2 correlation as high as 0.9 when compared to in situ observations. The median ICESat-2 snow depth over 100 m–1000 m lengths accurately captures seasonal snow depth variability and spatial patterns in snow distribution, including elevation-controlled orographic patterns. ICESat-2 snow depths are most accurate in regions where the majority of the terrain has slopes <20° and typical winter snow depths >0.5 m. Over 50 % of Idaho’s snow-covered area within this moderate slope, deep snowpack range and other mountain regions likely have sizable portions that meet these criteria. Thus by using available high-resolution snow-free DTMs, ICESat-2 applications can be dramatically expanded to provide valuable snow depth timeseries for observation or data assimilation.