The Utility of Optical Satellite Winter Snow Depths for Initializing a Glacio‐Hydrological Model of a High‐Elevation, Andean Catchment

Abstract

Information about end‐of‐winter spatial distribution of snow depth is important for seasonal forecasts of spring/summer streamflow in high‐mountain regions. Nevertheless, such information typically relies upon extrapolation from a sparse network of observations at low elevations. Here, we test the potential of high‐resolution snow depth data derived from optical stereophotogrammetry of Pléiades satellites for improving the representation of snow depth initial conditions (SDICs) in a glacio‐hydrological model and assess potential improvements in the skill of snowmelt and streamflow simulations in a high‐elevation Andean catchment. We calibrate model parameters controlling glacier mass balance and snow cover evolution using ground‐based and satellite observations, and consider the relative importance of accurate estimates of SDICs compared to model parameters and forcings. We find that Pléiades SDICs improve the simulation of snow‐covered area, glacier mass balance, and monthly streamflow compared to alternative SDICs based upon extrapolation of meteorological variables or statistical methods to estimate SDICs based upon topography. Model simulations are found to be sensitive to SDICs in the early spring (up to 48% variability in modeled streamflow compared to the best estimate model), and to temperature gradients in all months that control albedo and melt rates over a large elevation range (>2,400 m). As such, appropriately characterizing the distribution of total snow volume with elevation is important for reproducing total streamflow and the proportions of snowmelt. Therefore, optical stereo‐photogrammetry offers an advantage for obtaining SDICs that aid both the timing and magnitude of streamflow simulations, process representation (e.g., snow cover evolution) and has the potential for large spatial domains.

Monitoring Spatial and Temporal Differences in Andean Snow Depth Derived From Satellite Tri-Stereo Photogrammetry

“…Comparing records of precipitation at the Embalse Yeso and Termas del Plomo AWS in 2017 (not shown), we find that the former could not represent all of the same precipitation events seen by Termas del Plomo AWS in the study catchment (Shaw et al, 2020b). Therefore, we can infer that at least some of the differences in the Yeso Embalse precipitation records and Pléiades observations may be explained by orographic precipitation differences between these stations or by storms originating from Argentina.…”

“…( Figure 1C). There are three main glaciers in the basin: Bello (4.6 km 2 ), Yeso (2.9 km 2 ) and Pirámide (4.7 km 2 , debriscovered) which have been estimated to contribute between 3-32% of the basin's spring-summer streamflow through icemelt (Burger et al, 2019), though the hydrological regime is typically dominated by contributions from snowmelt (Ayala et al, 2016;Burger et al, 2019;Shaw et al, 2020b). The catchment is characterised by steep terrain (average slope of 27°) that supports small sclerophyllous vegetation.…”

“…The key challenge of all such modelling studies in mountainous regions to date, is the uncertainty in the distribution of solid precipitation in space and time (Ragettli et al, 2014;Ayala et al, 2016;Burger et al, 2019;Ayala et al, 2020;Shaw et al, 2020b), largely governed by the lack of high elevation data and/or the reliability of high elevation snow-precipitation records (e.g., Scaff et al, 2018). Shaw et al (2020a) addressed this challenge by using Pléiades DEM differencing to obtain an estimate of snow depth for the Rio Yeso catchment.…”

“…The very high (up to 5,000-6,000 m) elevation differences between the mountain peaks and populous lowland regions further emphasise the strong dipole of the source and sink of snow which is exacerbated by prolonged severe drought conditions for the region (Garreaud et al, 2017;Garreaud et al, 2019). Quantifying the snow volume of these high mountain catchments is therefore of great importance (Masiokas et al, 2006;Gascoin et al, 2013;Ragettli et al, 2014;Cortés et al, 2016;Masiokas et al, 2020;Shaw et al, 2020a;Shaw et al, 2020b), yet the lack of high elevation observations is a limiting factor in understanding the spatiotemporal behaviour of the snowpack (Alvarez-Garreton et al, 2018;Scaff et al, 2018;Masiokas et al, 2020). This is particularly evident when there are insufficient observations at very high elevations (>∼3,000 m a.s.l.)…”

“…This is particularly evident when there are insufficient observations at very high elevations (>∼3,000 m a.s.l.) (Scaff et al, 2018;Shaw et al, 2020a) or for observing precipitation events that originate from the eastern side of the Andes (Shaw et al, 2020b).…”