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

Accurate knowledge of snow depth distributions in mountain catchments is critical for applications in hydrology and ecology. Recently, a method was proposed to map snow depth at meter-scale resolution from very-high-resolution stereo satellite imagery (e.g., Pléiades) with an accuracy close to 0.5 m. However, the validation was limited to probe measurements and unmanned aircraft vehicle (UAV) photogrammetry, which sampled a limited fraction of the topographic and snow depth variability. We improve upon this evaluation using accurate maps of the snow depth derived from Airborne Snow Observatory laser-scanning measurements in the Tuolumne river basin, USA. We find a good agreement between both datasets over a snow-covered area of 138 km2 on a 3 m grid, with a positive bias for a Pléiades snow depth of 0.08 m, a root mean square error of 0.80 m and a normalized median absolute deviation (NMAD) of 0.69 m. Satellite data capture the relationship between snow depth and elevation at the catchment scale and also small-scale features like snow drifts and avalanche deposits at a typical scale of tens of meters. The random error at the pixel level is lower in snow-free areas than in snow-covered areas, but it is reduced by a factor of 2 (NMAD of approximately 0.40 m for snow depth) when averaged to a 36 m grid. We conclude that satellite photogrammetry stands out as a convenient method to estimate the spatial distribution of snow depth in high mountain catchments.

Snow Depth Distribution Patterns and Consistency from Airborne Lidar Time Series

Snow provides fresh meltwater to over a billion people worldwide. Snow dominated watersheds drive western US water supply and are increasingly important as demand depletes reservoir and groundwater recharge capabilities. This motivates our inter- and intra-annual investigation of snow distribution patterns, leveraging the most comprehensive airborne lidar survey (ALS) dataset for snow. Validation results for ALS from both the NASA SnowEx 2017 campaign in Grand Mesa, Colorado and the time series dataset from the Tuolumne River Basin in the Sierra Nevada, in California, are presented. We then assess the consistency in the snow depth patterns for the entire basin (at 20-m resolution) and for subbasin regions (at 3-m resolution) from a collection of 51 ALS that span a six-year period (2013-2018) in the Tuolumne Basin. Strong correlations between ALS from different years near peak SWE confirm that spatial patterns exist between snow seasons. Year-to-year snow depth differs in absolute magnitude, but relative differences are consistent spatially, such that deep and shallow zones occur in the same location. We further show that elevation is the terrain parameter with the largest correlation to snow depth at the basin scale, and we map the expected pattern distribution for periods with similar snow-covered extents. Lastly, we show at a subbasin scale that distribution patterns are more consistent in vegetation-limited areas (bedrock dominated terrain and open meadows) compared to vegetation-rich zones (valley hillslopes and dense canopy cover). The maps of snow patterns and their consistency can be used to determine optimal locations of new long-term monitoring sites, design sampling strategies for future snow surveys, and to improve high resolution snow models.

Estimation of Daily Spatial Snow Water Equivalent from Historical Snow Maps and Limited In-Situ Measurements

We present a scheme aimed at estimating daily spatial snow water equivalent (SWE) maps in real time and at high spatial resolution from scarce in-situ SWE measurements from Internet of Things (IoT) devices at actual sensor locations and historical SWE maps. The method consists of finding a background SWE field, followed by an update step using ensemble optimal interpolation to estimate the residuals. This novel approach allowed for areas with parsimonious sensors to have accurate estimates of spatial SWE without explicitly discovering and specifying the spatial-interpolation features. The scheme is evaluated across the Tuolumne River basin on a 50 m grid using an existing LiDAR-based product as the historical dataset. Results show a minimum RMSE of 30% at 50 m resolutions. Compared with the operational SNODAS product, reduction in error is up to 80% with historical LiDAR-measured snow depth as input data.

Snow depth mapping from stereo satellite imagery in mountainous terrain: evaluation using airborne lidar data

An accurate knowledge of snow depth distribution in mountain catchments is critical for applications in hydrology and ecology. A recent new method was proposed to map the snow depth at meter-scale resolution from very-high resolution stereo satellite imagery (e.g., Pléiades) with an accuracy close to 0.50 m. However, the validation was mainly done using probe measurements which sampled a limited fraction of the topographic and snow depth variability. We deepen this evaluation using accurate maps of the snow depth derived from ASO airborne lidar measurements in the Tuolumne river basin, USA. We find a good agreement between both datasets over a snow-covered area of 137 km2 on a 3 m grid with a positive bias for Pléiades snow depth of 0.08 m, a root-mean-square error of 0.80 m and a normalized median absolute deviation of 0.69 m. Satellite data capture the relationship between snow depth and elevation at the catchment scale, and also small-scale features like snow drifts and avalanche deposits. The random error on snow depth can be reduced by a factor two (up to approximately 0.40 m) when the snow depth map is spatially averaged to a ~ 20 m grid. The random error at the pixel level is lower on snow-free areas than on snow-covered areas, but errors on both terrain type converge at coarser resolutions, which is important for further applications of the method in areas without snow depth reference data. We conclude that satellite photogrammetry stands out as an efficient method to estimate the spatial distribution of snow depth in high mountain catchments.