The Utility of Infrequent Snow Depth Images for Deriving Continuous Space‐Time Estimates of Seasonal Snow Water Equivalent

Abstract

Snow water equivalent (SWE), particularly in mountains regions, has been an elusive hydrologic measurement. We examine the utility of a data assimilation approach to generate space‐time continuous estimates of SWE from more readily available snow depth (SD) measurements. A multitemporal lidar data set provides a unique opportunity to assimilate single SD images and verify posterior estimates against SD images at nonassimilation times. Application over three water years shows significant improvement in the posterior estimates with an average correlation between estimated and measured SD fields of 0.88 compared to 0.52 for prior estimates that do not benefit from the assimilated SD data. We also show that posterior estimates are consistent with independent in situ SWE and streamflow measurements. This work demonstrates that using high‐resolution/high‐accuracy, but infrequent, SD measurements combined with a data assimilation framework could make significant inroads toward the goal of spatially distributed SWE and snowmelt estimates at the global scale.

Snow depth mapping from stereo satellite imagery in mountainous terrain: evaluation using airborne laser-scanning data

“…These sets of options were empirically selected but do not cover all the options available in the ASP. The first set of options is the one used by Marti et al (2016). This set uses the local-search-window stereo algorithm and the normalized cross-correlation parametric cost function with windows of 25 px×25 px (these options are hereafter called local search).…”

“…Recently a method was introduced to retrieve HS maps from satellite data at meter-scale resolution, typically 1 to 4 m (Marti et al, 2016;McGrath et al, 2019;Shaw et al, 2019). The method is based on the differencing of snow-on (winter) and snow-off (in general end of summer) digital elevation models (DEMs) that are generated from very-highresolution satellite stereo imagery (e.g., Pléiades; Digital-Globe/Maxar WorldView-1, WorldView-2 and WorldView-3; and GeoEye-1).…”

“…The method is based on the differencing of snow-on (winter) and snow-off (in general end of summer) digital elevation models (DEMs) that are generated from very-highresolution satellite stereo imagery (e.g., Pléiades; Digital-Globe/Maxar WorldView-1, WorldView-2 and WorldView-3; and GeoEye-1). The method was first tested using two Pléiades stereo triplets over the Bassiès catchment in the Pyrenees (Marti et al, 2016). The snow-on and snow-off DEMs were generated using the Ames Stereo Pipeline (ASP; Shean et al, 2016;Beyer et al, 2018) and coregistered before differencing (Berthier et al, 2007).…”

“…The results showed that snow depth could be retrieved from Pléiades images with an accuracy of roughly ∼ 0.5 m (standard deviation of residuals of 0.58 m for a pixel size of 2 m), suggesting that the method had the potential to become a viable alternative to airborne campaigns in mountain catchments with the benefits of a space-based platform: access to any point on the globe and lower cost for the end user. HS maps from stereo satellite images were evaluated in two recent studies besides Marti et al (2016) against terrestrial laser-scanning data over a small area (< 1 km 2 ; Shaw et al, 2019) and against ground-penetrating radar measurements, which were limited to roughly 50 km 2 of relatively flat terrain (McGrath et al, 2019).…”

“…However, these works provided only a partial validation of the method since the reference data did not homogeneously sample the topographic and HS variability of the study area. For example, in Marti et al (2016), accumulation due to snow drifts on the lee side of high-elevation ridges was not surveyed for safety reasons. The sampling depth was also limited to 3.2 m, which was the length of the snow probes.…”