Using high resolution Lidar data from SnowEx to characterize the sensitivity of snow depth retrievals to point-cloud density and vegetation

The NASA SnowEx campaign conducted in 2016 and 2017 provides a rich source of high-resolution Lidar data from JPL's Airborne Snow Observatory (ASO - http://aso.jpl.nasa.gov) combined with extensive in-situ measurements in two key areas in Colorado: Grand Mesa and Senator Beck. While the uncertainty in the 50m snow depth retrievals from NASA's ASO been estimated at 1-2cm in non-vegetated exposed areas (Painter et al., 2016), the impact of forest cover and point-cloud density on ASO snow lidar depth retrievals is relatively unknown. Dense forest canopies are known to reduce lidar penetration and ground strikes thus affecting the elevation surface retrieved from in the forest. Using high-resolution lidar point cloud data from the ASO SnowEx campaigns (26pt/m2) we applied a series of data decimations (up to 90% point reduction) to the point cloud data to quantify the relationship between vegetation, ground point density, resulting snow-off and snow-on surface elevations and finally snow depth. We observed non-linear reductions in lidar ground point density in forested areas that were strongly correlated to structural forest cover metrics. Previously, the impacts of these data decimations on a small study area in Grand Mesa showed a sharp increase in under-canopy surface elevation errors of -0.18m when ground point densities were reduced to 1.5pt/m2. In this study, we expanded the evaluation to the more topographically challenging Senator Beck basin, have conducted analysis along a vegetation gradient and are considering snow the impacts of snow depth rather than snow-off surface elevation. Preliminary analysis suggest that snow depth retrievals inferred from airborne lidar elevation differentials may systematically underestimate snow depth in forests where canopy density exceeds 1.75 and where tree heights exceed 5m. These results provide a basis from which to identify areas that may suffer from vegetation-induced biases in surface elevation models and snow depths derived from airborne lidar data, and help quantify expected spatial distributions of errors in the snow depth that can be used to improve the accuracy of ASO basin-scale depth and water equivalent products.