Measuring the snowpack depth with Unmanned Aerial System photogrammetry: comparison with manual probing and a 3D laser scanning over a sample plot

Abstract

Abstract.Photogrammetric surveys using Unmanned Aerial Systems (UAS) may represent an alternative to existing methods for measuring the distribution of snow, but additional efforts are still needed to establish this technique as a low-cost, yet precise tool. Importantly, existing works have mainly used sparse evaluation datasets that limit the insight into UAS performance at high spatial resolutions. Here, we compare a UAS-based photogrammetric map of snow depth with data acquired with a 5MultiStation and with manual probing over a sample plot. The relatively high density of manual data (135 pt over 6700 m 2 , i.e., 2 pt/100 m 2 ) enables to assess the performance of UAS in capturing the marked spatial variability of snow. The use of a MultiStation, which exploits a scanning principle, also enables to compare UAS data on snow with a frequently used instrument in high-resolution applications. Results show that the Root Mean Square Error (RMSE) between UAS and MultiStation data on snow is equal to 0.036 m when comparing the two point clouds. A large fraction of this difference may be, however, due 10 to spurious differences between datasets due to simultaneous snowmelt, as the RMSE on bare soil is equal to 0.02 m. When comparing UAS data with manual probing, the RMSE is equal to 0.31 m, whereas the median difference is equal to 0.12 m. The statistics significantly decrease up to RMSE = 0.17 m when excluding areas of likely water accumulation in snow and ice layers. These results suggest that UAS represent a competitive choice among existing techniques for high-precision, high-resolution remote sensing of snow.

Monitoring snow depth change across a range of landscapes with ephemeral snowpacks using structure from motion applied to lightweight unmanned aerial vehicle videos

“…Key point density decreased by almost 1 order of magnitude when comparing missions flown with snow more than 1 day old and missions with either deep fresh snow or smooth icy snowpacks. Previous studies have identified the drop in both elevation and SD accuracy due to deep fresh snow (Nolan et al 2015;Avanzi et al, 2017) and icy conditions (Gindraux et al, 2017). Here we demonstrate that D may be a useful indicator of such conditions and hence an indicator of the quality of SD estimates.…”

“…1), with a tendency for underestimation in areas with substantial ground thatch layer. The underestimate in these conditions was approximately the same magnitude of the thatch height, leading us to hypothesize that they are related to an overestimate in the local DSM height as previously suggested (Nolan et al, 2015;Avanzi et al, 2017). This hypothesis could be tested in future studies using supplementary in situ elevation measurements (e.g.…”

“…This should be performed in future studies using reference measurements from surface instruments (e.g. Avanzi et al, 2017).…”

“…to map SD with an accuracy (precision) of ±10 cm (8 cm at 1 standard deviation) in comparison to individual probe measurements over relatively flat surfaces. Similar results were subsequently reported using UAV systems, with a GSD ranging from ∼ 2 to ∼ 10 cm and altitude from 60 to 130 m a.g.l., over prairies (Harder et al, 2016), alpine shrub lands (Bühler et al, 2016;De Michele et al, 2016;Harder et al, 2016;Avanzi et al, 2017) and glaciers (Gindraux et al, 2017). Even greater accuracy (1.5 to 3.8 cm) and precision (4.2 to 9.8 cm at 1 standard deviation) have been reported for SD mapping over tundra (Cimoli et al, 2017) and alpine terrain (Vander Jagt et al, 2015) when using very low (10-30 m a.g.l.)…”

“…The temporal and spatial pattern of snow depth (SD) is of importance to hydrological, ecological and climate studies (GCOS, 2016). Together with representative estimates of snow density, time series of SD are indicative of changes in snow water equivalent that in turn are of importance to streamflow forecasting and management of hydroelectric resources (Clyde, 1939;Barnett et al, 2005;DeWall and Rango, 2008). In many ecosystems, SD is an important determinant of winter habitat in terms of range and access to forage (Bokhorst et al, 2016).…”