On the evolution of crystallographic texture in snow
Maurine Montagnat, Neige Calonne, Henning Löwe, Martin Schneebeli, Matthias Jaggi
Corresponding author: Maurine Montagnat
Corresponding author e-mail: maurine.montagnat@univ-grenoble-alpes.fr
Snow on the ground is made of sintered ice crystals whose crystallographic orientation can be described by the orientation of the c-axis of the hexagonal crystal lattice. Crystallographic orientation has an impact on ice properties such as mechanical response, crystal growth and optical birefringence. To measure c-axis orientations, thin sections of ice are observed under cross-polarized light. The method is now automatized to enable high spatial resolution (5 μm) and high angular resolution (1–3°). Commonly used for ice, such measurements were only recently applied to snow in a way that enables a statistical analysis of the distribution of the crystallographic orientation of snow samples, referred as the crystallographic fabric. The first results obtained from laboratory experiments showed that temperature–gradient metamorphism could lead to the development of a specific fabric, depending on the density. This result was further supported by observations along 0–3 m depth of the snowpack at Point Barnola, a very cold, very low-precipitation site in Antarctica, where variations of the crystallographic fabric seem to match with the density and microstructure layering. However, similar measurements along a Greenlandic snowpack, from the central location of EastGRIP, show an unexpected type of fabric that brought more complexity to our previous fabric analysis. Indeed, some observed fabrics at this site, where temperatures are higher than at Point Barnola, may be explained by a slow but non-negligible component of compression. In an attempt to clarify the observations from natural snowpacks, we performed a temperature-gradient metamorphism experiment under conditions that complement the previous lab work. In contrast to previous results, the snow fabric did not evolve in time. This questions the role of metamorphism and rather opens the question of the impact of the conditions within which the snow evolves. Overall, all these studies highlight the complexity of the mechanisms at play that prevent to draw, yet, a clear picture of the fabric evolution in snow. In this presentation we will provide an overview of the methods and their limitations, and focus on the results obtained and their diversity. The aim of the presentation will be to open discussions on the relevance of snow fabric characterization to help understand densification mechanisms.
A new method to visualize liquid distribution in snow by superimposing MRI and X-ray CT images
Satoru Yamaguchi, Satoru Adachi, Sojiro Sunako
Corresponding author: Satoru Yamaguchi
Corresponding author e-mail: yamasan@bosai.go.jp
X-ray computed tomography (X-ray CT) is a nondestructive technique that can be used to obtain detailed images of porous media with high resolution. However, it cannot be used to visualize water in snow because the X-ray absorption coefficient of snow particles is similar to that of water, thus differentiation of snow particles and on X-ray images is difficult. Magnetic resonance imaging (MRI) is also a nondestructive tool for obtaining detailed images of porous media. As MRI acquires nuclear magnetic resonance (NMR) signals of protons in liquid water and visualizes the NMR signal intensity, distinguishing water signals from those of snow particles on the MR images of snow samples is possible. However, the differentiation of snow particles and air gap on MR images is not possible because they do not produce NMR signals. To investigate the relationship between structures and water distribution in snow, we developed a new method to combine X-ray CT and MR images to compensate for the disadvantages associated with each of the techniques. Here, we summarize the use of the novel method in a preliminary experiment. A sample holder made of acrylic was used on account of the strong magnetic field of MRI. The sample holder consisted of a pipe fixed on the inside wall, which was used to introduce liquid into the sample and as a marker for aligning the CT and MR images. In the preliminary experiment, the sample holder was first filled with dry snow particles and imaged to obtain the microstructure of the sample by X-ray CT at –15°C . Then, the sample holder was moved to another cold room at –5°C. When the temperature of the sample was the same temperature of the cold room, a small amount of liquid was introduced through the pipe in the sample holder. In this study, dodecane (C12H26) , which has a melting point of –10°C, was used instead of water. Then an MR image was acquired, which provided liquid distribution information. Finally, the MR image was superimposed on the X-ray CT image using ImageJ (image analysis software). The superimposed image provided three-dimensional information on three components in the sample: the distribution of snow particles, liquid, and air gap. Our novel method is effective in characterizing the properties of wet snow.
Lateral flow of liquid water in snowpack
Hiroyuki Hirashima, Hikaru Osawa, Francesco Avanzi, Satoru Yamaguchi
Corresponding author: Hiroyuki Hirashima
Corresponding author e-mail: hirasima@bosai.go.jp
In snow-covered areas, water transport in snowpack is an important process that must be considered in the prediction of wet snow disasters such as full-depth avalanches, snowmelt landslides and snowmelt floods. In slope snow, it has been reported that water movement on the snow surface is greater down-slope than along the vertical direction. Consequently, water supply to the ground surface is concentrated in the concave terrain. Therefore, the lateral flow process is important in wet snow research. Slope flow remains poorly understood; the vast majority of snow studies have focused on flat-terrain conditions. In this study, lateral flow was examined using tracer experiments in slope snow, analysis of multi compartment lysimeter (MCL) data, and a multi-dimensional water transport model. Tracer experiments were conducted on snow slopes to obtain the data of water infiltration in slope snow. The data was used for analysis of a detailed lateral flow process. Occurrence of lateral flow for various weather and snowpack conditions were analyzed using MCL data. Data from both tracer experiments and MCL were used for the validation of a reproduction simulation using a multi-dimensional water transport model. Overall, lateral flow occurred when infiltrated water ponded at the capillary barrier or ice layer. The water transport model exhibited good lateral flow at the capillary barrier. Additionally, discrepancies in lateral flow on the ice layer were observed. Currently, water infiltration through ice layer is not modeled sufficiently. When the capillary barrier formed, lateral flow occurred even if the gentle slope was less than 2°. This downward movement of water on a slight slope was confirmed by the water transport model. Consequently, simulation showed the concentration of discharge at the downslope area. MCL data also showed the concentration of discharge when the snow profiles had a fine over coarse layer boundary which could make capillary barriers. This project was supported by JSPS KAKENHI Grant JP20K04068 and JP20K14562.
Bulk snow density vs snow density profiles at Weissfluhjoch
Charles Fierz, Christoph Marty
Corresponding author: Charles Fierz
Corresponding author e-mail: fierz@slf.ch
There exist quite a few studies comparing measurements of snow density taken with different snow samplers. On the other hand, there are fewer comparisons between measurements of bulk snow density, i.e. the density of the full snowpack, and layer-based, continuous snow density profiles recorded concurrently at the same site. On the study plot Weissfluhjoch, located at 2536 m a.s.l. above Davos, bulk snow density has been measured since winter 1937. Starting in the late 1980s, however, about one-fifth of these measurements are accompanied by parallel measurements of snow-density profiles with snow samplers such as density cutters of various shapes, 0.5 litre cylinders, the electronic Denoth device, or the SnowMicroPen. Sometimes even more than one parallel measurement is available. In this paper we present the full data set and conduct a thorough comparison of (a) bulk snow-density measurements vs layer-based, continuous snow-density profiles to assess the difference statistics between them and (b) intercompare snow-density profiles to unveil advantages and disadvantages of both different samplers and different sampling methods.
Kevin Ostapowich, Joseph Shea, Brian Menounos
Corresponding author: Kevin Ostapowich
Corresponding author e-mail: k.ostapowich@gmail.com
Alpine snowpacks provide critical ecosystem functions and are an important freshwater resource in many downstream communities and industries. During spring and summer, snowmelt is driven primarily by incoming shortwave and longwave radiation fluxes, with a negligible ground heat flux contribution. However, mid-winter heat exchange from the ground may play key roles in basal snowmelt, groundwater recharge, and reduced cold content in Alpine snowpacks. In this study we measure the ground thermal regime with thermistors installed across the ground–snow interface at 29 locations in the Conrad basin, located in the Purcell Mountains of British Columbia, Canada. With digital elevation models and lidar-derived snow depths, we examine physical controls on the ground thermal regime, and estimate ground heat fluxes for a range of snow and bedrock thermal conductivities. Early snow-cover onsets, longer snow-cover durations and deeper end-of-winter snow depths were positively related to midwinter ground temperatures and temperature gradients, while other physical controls (slope, aspect, elevation, ruggedness) showed little to no relation. Modeled ground heat fluxes varied by an order of magnitude between shallow (1.4 m) and deep (2.6 m) snow sites, with shallow snow sites subject to greater overall total energy transfer, though this is likely lost directly to the atmosphere. Snow thermal conductivity has a large influence on the modeled ground heat flux, however, the true snow thermal conductivity will vary throughout the winter and was not measured in this study. Snow redistribution by wind plays an important role in the ground thermal regime across the Conrad basin, and deep snowpack sites that cover 51% of the non-glacierized terrain may be a significant contributor to winter baseflow.
Assessment of science readiness for a new snow mass radar mission
Chris Derksen, Joshua King, Stephane Belair, Camille Garnaud, Vincent Vionnet, Vincent Fortin, Juha Lemmetyinen
Corresponding author: Chris Derksen
Corresponding author e-mail: Chris.Derksen@ec.gc.ca
Environment and Climate Change Canada (ECCC) and the Canadian Space Agency (CSA) continue to advance a new satellite Ku-band radar mission focused on providing moderate resolution (500 m) information on seasonal snow mass. Like many regions of the northern hemisphere, estimates of the amount of water stored as seasonal snow are highly uncertain across Canada. To address this gap, a technical concept capable of providing dual-polarization (VV/VH), moderate resolution (500 m), wide swath (~250 km), and high duty cycle (~25% SAR-on time) Ku-band radar measurements at two frequencies (13.5 and 17.25 GHz) is under development. Parallel to engineering studies to address the technical readiness, a range of activities are in progress to advance scientific readiness. In this presentation, we will review how recent progress within the mission science team and across the snow community has provided a sound science foundation for the mission, and identify risks to meeting the required level of readiness within the required timeline for full mission implementation.
Key areas include:
Implementation of computationally efficient SWE retrieval techniques, including parameterized forward model simulations for prediction of snow volume scattering, physical snow modeling to provide initial estimates of snow microstructure, and consideration of background characteristics;
Incorporation of land surface model SWE estimates to infill gaps with no radar-derived SWE information due to dense forest, wet snow, and swath gaps;
Direct assimilation of Ku-band backscatter into environmental prediction systems (analogous to how SMOS and SMAP data have improved soil moisture analysis through radiance-based assimilation);
Segmentation of wet from dry snow;
Continued advancement of the understanding of the physics of Ku-band backscatter response to variations snow through new experimental tower and airborne measurements.
Ku-band radar is a viable approach for a terrestrial snow mass mission because of sensitivity to SWE (through the volume scattering properties of dry snow) and the wet versus dry state of snow cover. To justify investment in such a mission, however, the scientific pieces must be in place. Balanced and honest assessments of the state of scientific readiness, the likelihood for emerging capabilities, and the level of engagement across the snow community along the mission timeline are essential to ensure a healthy mission development process.
Consideration of slab ground contact after weak layer collapse in closed-form models
Johannes Schneider, Philipp Weißgraeber, Philipp Rosendahl
Corresponding author: Johannes Schneider
Corresponding author e-mail: j.schneider@ismd.tu-darmstadt.de
The Weak Layer Anticrack Model (WEAC) is a tool for the analysis of snow slabs and the underlying weak layer, a system prone to release dry snow slab avalanches. Based on first-order shear deformation beam theory, it provides a closed-form solution for the deformation of the snowpack and the stress and energy release rate of the weak layer. Using a mixed-mode failure criterion for the weak layer, the model can predict anticrack nucleation preceding the release of an avalanche. In the current state, it allows unbounded vertical slab deformation. To limit this deformation to the collapse height of the weak layer, contact between the snow slab and the substratum must be considered. The present work proposes an extension of the model to take this effect into account. The propagation of a crack in the weak layer increases the deflection of the slab until it is finally in contact with the substratum. This corresponds to a change of boundary conditions and introduces a structural nonlinearity, which is resolved using compatibility assumptions between the supported and unsupported sections of the snow slab. From these assumptions, a closed-form solution for the span of the unsupported section is derived, which is then employed to calculate the energy release rate of a self-similarly propagating anticrack and the stresses in the remaining intact weak layer. A comparison with a finite element reference model shows good agreement with the analytical model and thus validates the assumptions made. Parametric investigations on propagation saw tests address the effect of slab–substratum contact on the crack propagation conditions in snowpacks or experiments such as propagation saw tests.
Damian Steffen, André Burkard, Pascal Venetz, Andreas Tegethoff
Corresponding author: Damian Steffen
Corresponding author e-mail: damian.steffen@geoformer.ch
Snowdrift is a well known problem, but one that receives only little attention in practice. Practitioners involved in the assessment of snowdrift and related risks are dealing with a variety of uncertainties. These uncertainties are found in, but not limited to, the presence of meteorological data and the interactions between the local terrain features, snow and wind. To reduce risks such as rear-end collisions and the cost of snow clearing from traffic routes such as railways or highways different strategies can be considered to deal with the snowdrift problem. Due to the complexity of the process and often missing data, in practice classical measures such as snow fences or wind-deflection systems are usually proposed and implemented. The use of measures adapted to the local problem and based on a deeper analysis of the local topography is often unusual. An example of the spatial occurrence of drifting snow and the associated problems is the Arlas area on the Bernina Pass. Drifting snow leads to snow deposits higher than 5 m on the road with an average occurrence of once every 5 years. This causes long and expensive snow-removal operations as well as road closures. To plan an efficient and economic measure, the complex terrain, different snowdrift exposures and the interaction between the road and the railway line must be considered. For this, various approaches reducing local effects of snowdrift were planned. In order to investigate the effectiveness of the different mitigation measures, wind simulations based on adapted terrain models were carried out and compared. By illustrating the difficulties in dealing with snowdrift and planning of different mitigation measures related to the case study at the Bernina Pass, we highlight the need for deeper situation analysis as well as the cooperation of various specialists from practice and science. We present the approach used and thus address some of today’s major challenges with snowdrift in a practical context.
C- and K-band microwave penetration into snow studied with off-the-shelf tank radars
Arttu Jutila, Christian Haas
Corresponding author: Arttu Jutila
Corresponding author e-mail: arttu.jutila@awi.de
Snow cover on sea ice poses a challenge for radar measurements as microwave penetration into snow is not yet fully understood. Here we investigate microwave penetration into snow on Arctic sea ice using commercial C (6 GHz) and K (26 GHz) band tank radars. Radar measurements collected at nine study locations over first-year and multi-year landfast sea ice in the Lincoln Sea in May 2018 are analysed together with detailed measurements of the physical properties of the snow cover to determine the dominant scattering horizons at both frequencies and evaluated for the feasibility to determine snow depth. The results show that in 39% of the measurements and only on first-year ice major fraction of the C band radar backscatter originated closer to the snow–ice interface potentially enabling snow depth retrieval. At K band, 81% of the radar returns originated from the snow surface. However, the analysis was potentially hampered by relatively warm air temperatures (up to –0.9°C) during the study period and morphological features found in the snow cover, partly confirming the findings of previous studies.
How much snow is there across the Mediterranean region?
Francesco Avanzi, Hans Lievens, Christian Massari, Lorenzo Alfieri, Fabio Delogu, Lorenzo Campo, Andrea Libertino, Paolo Filippucci, Hamidreza Mossafa, Pere Quintana-Seguì, Simone Gabellani, Gabrielle De Lannoy
Corresponding author: Francesco Avanzi
Corresponding author e-mail: francesco.avanzi@cimafoundation.org
The seasonal snowpack plays an important role for the water budget of the Mediterranean Sea, but an exact quantification of this contribution is still elusive. This is particularly true if one compares this preliminary understanding with previous work in other semi-arid regions of the world such as the western USA, where both the scientific community and, importantly, the water-resources management sector have already achieved consensus estimates on this matter, with snow supplying up to 80% of annual runoff. In order to provide such figures for the Mediterranean Sea region, we are developing a 6-year (2015–21) reanalysis of snow water equivalent (SWE) at 1 km resolution and daily granularity for the whole basin of the Mediterranean Sea (Nile excluded). The reanalysis uses ERA5 meteorological data and satellite-based precipitation as input for a snow model, S3M, which then assimilates daily snapshots of snow depth from the C-SNOW Sentinel-1 product. These simulations are being validated using in-situ snow depth measurements across the Mediterranean Sea region. Maps of SWE from this reanalysis are being spatially aggregated for each water basin of the Mediterranean Sea region, with specific attention for the southern Alps, the Apennines, the Pyrenees, the southern Balkans, the Middle East, the Sierra Nevada and the Atlas Mountains. The 6 year period of record of the reanalysis provides a preliminary estimate of central tendencies and standard deviations of SWE for these basins, as well as an estimate of snow depletion curves and snowmelt timing. These estimates demonstrate the added value of remote-sensing products to tackle societally relevant questions in the 21st century.
Combining snow physics an machine learning to predict avalanche activity
Léo Viallon-Galinier, Pascal Hagenmuller, Nicolas Eckert
Corresponding author: Léo Viallon-Galinier
Corresponding author e-mail: leo.viallon@meteo.fr
Predicting avalanche activity is critical in mountainous areas to support warning services. For this purpose and to complement field observations, snow-cover modeling provides information otherwise unavailable on the present and future state of the snowpack. Combining machine-learning methods and a knowledge of the mechanical processes leading to avalanches, we develop and evaluate a method to assess snowpack stability and expected avalanche activity from simulated snow profiles. We combine extensive snowpack and snow stability simulations with avalanche observations within a tandom forest approach to predict avalanche days at a spatial resolution corresponding to elevations and aspects of avalanche paths in a given mountain range. We develop a rigorous leave-one-out evaluation procedure including an independent test set, confusion matrices, and receiver operating characteristic curves. In a study case (Haute-Maurienne, French Alps, 1960–2018), we show the added value within the statistical model of considering advanced snow-cover modeling and mechanical stability indices instead of using only simple meteorological and bulk information. Specifically, using mechanically based stability indices and their derivatives in addition to simple snow and meteorological variables increases the true positive rate from 65% to 76%. However, due to the scarcity of avalanche events and the possible misclassification of days without avalanche observation into non-avalanche days, the precision remains low, around 3.4%. We also show that the model is robust and works on other studied areas in the French Alps and Pyrenees. The obtained scores illustrate the difficulty of predicting avalanche occurrence with a high spatio-temporal resolution, even with the current cutting-edge data and modeling tools. Nevertheless, our study opens perspectives to improve modeling tools supporting operational avalanche forecasting.
Development of Handheld Integrating Sphere Snow Grain Sizer (HISSGraS)
Teruo Aoki, Akihiro Hachikubo, Masashi Niwano, Sumito Matoba, Tomonori Tanikawa, Motoshi Nishimura, Hiroshi Ishimoto, Rigen Shimada, Ryo Inoue, Jean-Charles Gallet, Masahiro Hori, Satoru Yamaguchi
Corresponding author: Teruo Aoki
Corresponding author e-mail: aoki.teruo@nipr.ac.jp
Snow grain size is an important physical parameter that governs the albedo and describes the snow metamorphism process. Recently, various techniques including an optical method, X-ray microtomography, gaseous adsorption (Brunauer–Emmett–Teller: BET) method and so on have been proposed to measure the specific surface area (SSA) as a measure of snow grain size. Among them, the IceCube (A2 photonics, France) is an accurate instrument for field measurement. However, it requires snow sampling and is too heavy (8 kg) to measure the different snow surfaces for validation of satellite measurements. We have developed the Handheld Integrating Sphere Snow Grain Sizer (HISSGraS) for field use. The basic measurement principle of the HISSGraS is the same as that of the IceCube, but it can directly measure the snow surface or snow pit face by installing a glass window at the front of an integrating sphere. The weight is 0.5 kg, and the output signals are recorded on a memory card. There are two types of HISSGraS: Version 1 needs to be calibrated by the spectralon reflectance standards before and/or after the measurements to convert the output signal to the reflectance of the target snow the same as the IceCube. This is mainly because a laser diode of the light source is dependent on temperature. We have improved this by monitoring the laser temperature on Version 2, which does not need simultaneous calibration with every field measurement. Furthermore, the two kinds of conversion from reflectance to SSA were formulated by a radiative transfer model calculation using spherical and non-spherical snow grain shape models. We observed SSAs using the HISSGraS, IceCube and BET method from 27 February to 4 March 2022 on the snow field in eastern Hokkaido, Japan. The BET method is in principle the most accurate among the three techniques. The SSAs measured with the HISSGraS Ver. 2 agreed better with those by the BET method than the HISSGraS Ver. 1. For small grain sizes, the HISSGraS Ver. 2 derived SSAs using a non-spherical grain shape model agreed better with those by the BET method than the HISSGraS Ver. 2 using a spherical grain shape model. The IceCube-derived SSAs agreed well with those by the BET method, but the agreement between the HISSGraS and the IceCube was relatively lower than that between the IceCube and the BET method due to the opposite biased error between the HISSGraS and the IceCube for smaller grain sizes.
Matthieu Vernay, Matthieu Lafaysse, Clotilde Augros
Corresponding author: Matthieu Vernay
Corresponding author e-mail: matthieu.vernay@meteo.fr
Reliable estimation of precipitation fields is a key issue for snow-cover modelling. Most existing snow-modelling systems use precipitation-gauge observations either directly or combined with a Numerical Weather Prediction model simulation through an assimilation algorithm. While rain-gauge observations are sufficient for local simulations, this is unsatisfactory for large-scale and high-resolution simulations because the spatial variability of precipitation is much higher than the density of precipitation observation networks, especially in mountainous areas. The use of radar observations could overcome this limitation by providing spatialized estimates of precipitation at high spatial and temporal resolutions. However, even when they use observations from a patchwork of radar sources, existing radar products are known to have a lower skill in complex terrain due to various radar beam elevation issues and other processes and interactions. This is the case for the French PANTHERE product, which combines information from various radar sources covering mainland France and has a significantly reduced quality over the French Alps. Products combining radar information and surface observations of precipitation such as the French ANTILOPE product hopefully mitigate this deficiency but few evaluations of such products have been carried out specifically over mountainous areas. ANTILOPE uses kriging to spatialize rain-gauge observations and mixes this with radar data to produce precipitation estimates at 1 km resolution and an hourly time step. This contribution evaluates the added value of merging precipitation gauges and radar data in ANTILOPE compared to a simple kriging of gauge observations and the radar-only PANTHERE precipitation estimates over the French Alps. Independent 24-hour precipitation observations from ski resorts at elevations between 800 m and 2400 m were used as reference for the evaluation. This evaluation focuses on the potential to use ANTILOPE precipitation estimates as observations in a high resolution assimilation system dedicated to improve snow modelling. Thus, a focus on solid precipitation is made with the underlying goal to quantify the observation errors associated with the ANTILOPE product and their spatial variations. This evaluation shows a promising benefit to assimilate such a product for high-resolution snow-cover modelling, provided that relief-dependent observation errors are prescribed due to increased errors near the ridges.
Large-diameter trees affect snowpack duration in post-fire old-growth forests
Michaela Teich, Kendall M. L. Becker, Mark S. Raleigh, James A. Lutz
Corresponding author: Michaela Teich
Corresponding author e-mail: michaela.teich@bfw.gv.at
Snow duration is the net result of accumulation and ablation and can be lengthened or diminished, depending on forest vegetation. Forest fire alters vegetation cover, leading to reduced snow interception, increased light transmission and decreased snow albedo from charred woody debris, which together modify accumulation and ablation rates. Low- to moderate-severity fire typically kills smaller-diameter trees while leaving some larger trees alive that are fire-resistant. Snowpack duration in post-fire forests is therefore influenced by the neighbourhoods of trees, snags and deadwood of various sizes, affecting water provision into summer months when demands increase. We used annually resolved, spatially explicit data on tree mortality, snow depth and snow duration collected at the Yosemite Forest Dynamics Plot (California, USA) in an old-growth, mixed-conifer forest that burned at low to moderate severity. We calculated snow disappearance timing from 63 HOBO temperature loggers that were installed on a fixed grid. We then generated 10 tree neighbourhood metrics at scales up to 40 m from snow depth and snow disappearance sampling points. Litter and woody debris were collected adjacent to each HOBO location, and incoming solar radiation was simulated with the Alpine surface process model ALPINE3D. We developed two linear mixed models, predicting snow disappearance timing as a function of tree neighbourhood, litter density and simulated incoming solar radiation, and two multiple regression models explaining variation in snow depth as a function of tree neighbourhood. Higher densities of post-fire large-diameter snags within 10 m of a sampling point were related to higher snow depth (indicating reduced snow interception). Higher densities of large-diameter trees within 5 m of a sampling point and larger amounts of litter were associated with shorter snow duration (indicating increased longwave radiation emittance and/or decreased snow albedo). However, live trees with diameters >60 cm within 10 m of a snow disappearance sampling point were associated with a longer-lasting spring snowpack. This suggests that, despite the local effects of canopy interception and emitted longwave radiation from the boles of large trees, shading from their canopies may prolong snow duration over a larger area. Therefore, conservation of widely spaced, large-diameter trees is important in old-growth forests because they are resistant to fire and can extend the seasonal duration of snowmelt.
Lab experiments to quantify the effect of aeolian snow transport on the surface snow microstructure
Benjamin Walter, Henning Löwe
Corresponding author: Benjamin Walter
Corresponding author e-mail: walter@slf.ch
The microstructural evolution of surface snow under the influence of wind is hardly understood and poorly quantified, but crucial for polar and Alpine snowpacks. Only a few field studies have addressed the process of wind affecting surface snow at the snow–atmosphere interface. Available descriptions are based on empirical relations between snow density, wind velocity and air temperature. A microstructural picture discerning independent controls of snow crystal fragmentation, abrasion and sublimation is still missing. The goal of this project is to quantify wind induced microstructural modifications, to analyze the relevant physical processes and to develop parametrizations from controlled wind-tunnel experiments. A ring-shaped wind tunnel (RWT) with an infinite fetch was used in a cold lab to quantify the snow microstructure through systematic variations of flow, snow and temperature conditions. For the experiments, dendritic fresh snow was produced in the WSL/SLF snowmaker and slowly added to the RWT during the experiments simulating precipitation. Measurement techniques such as X-ray tomography, SnowMicroPen, density cutter and IceCube were applied to characterize the snow density, specific surface area (SSA), particle size and vertical layering before and after the highly dendritic new snow was exposed to the wind. Densification rates were found to increase from ΔΦi/(ΦiΔt) = 2 h–1 to 5 h–1 with increasing wind velocity (5–7.3 ms 1, Φi = ice volume fraction) which are two to three orders of magnitude higher than those measured for isothermal metamorphism. Corresponding densification rates derived from state-of-the-art snow physical models span an order of magnitude, significantly deviating from the measured values, thus underlining the importance of accurately understanding wind induced microstructural modifications. The SSA was found to decreases with a rate of change of ΔSSA/(SSAΔt) = –0.1 h–1 to –0.15 h–1, which is one order of magnitude higher than the SSA rate of change for isothermal metamorphism. Our results provide a basis for an improved, fundamental understanding of the effect of aeolian snow transport on optically and mechanically relevant microstructural properties of surface snow. Refining existing or developing new models based on our RWT results may have an impact on various cryospheric topics like avalanche formation, exchange of chemical species with the atmosphere, Alpine and polar mass balances, or radiative transfer.
Evaluation of ICESat-2 derived snow depth estimates in Alpine watersheds
Karina Zikan, Ellyn Enderlin, Hans-Peter Marshall, Shad O’Neel
Corresponding author: Karina Zikan
Corresponding author e-mail: karinazikan@u.boisestate.edu
Quantification of seasonal snow depth, density and water equivalent is an ongoing challenge for snow research and watershed management. Snow distribution is controlled by numerous factors, including wind–snow interactions, terrain and orographic precipitation patterns creating a dynamic snowpack where snow depths can vary by orders of magnitude over ranges of tens to hundreds of meters. In-situ observations collected at fixed continuous monitoring stations (e.g. snotel sites) and by manual snow observation campaigns provide detailed and accurate measurements of snow depth but are spatially and temporally limited and cannot capture spatial variability. Airborne lidar techniques have greatly expanded our ability to study snow, providing snow depth maps at larger spatial scales. Satellite lidar can further expand observation capabilities, providing repeat observations throughout the snow accumulation and melt seasons that can improve estimates of snowpack evolution as we gain capacity to interpret the observations. In this presentation, we focus on evaluation of ICESat-2-derived snow depths in Alpine watersheds in the western continental United States. The objective of our work is to develop a general method to produce snow-depth estimates from ICESat-2 elevations in any watershed where a high-resolution snow-free digital elevation model (DEM) is available. Since ICESat-2 tracks do not perfectly repeat in the mid-latitudes, separately collected snow-free reference DEMs are required to calculate snow depths from ICESat-2. Preliminary snow depth mapping with ICESat-2 suggests that accuracy and precision of ICESat-2 elevations are slope-dependent. We have refined our method to better account for this slope dependence and ICESat-2 photon count distribution patterns. The performance of our current revised method is analyzed in multiple experimental watersheds where high-resolution snow-free DEMs and in-situ snow-depth data are available. The impacts of various terrain features including vegetation, slope and aspect are investigated, giving insight into where this method will provide meaningful snow depth data. This research is funded by the NASA EPSCoR award 80NSSC20M0222.
A novel mechanism for long run-out in dry-snow avalanches
Dieter Issler
Corresponding author: Dieter Issler
Corresponding author e-mail: di@ngi.no
Under suitable conditions such as copious cold, dry and light snow, run-out angles as low as 14–15° have been observed for the non-suspended part of snow avalanches. This corresponds to an effective friction coefficient of the avalanche head of only 0.25 – smaller than appears possible in a purely granular flow. This indicates the presence of a fluidizing mechanism that helps support the weight of the flowing mass without creating significant friction. We propose that air expulsion from the snow cover under the weight of the avalanche flowing over it provides a bed-normal pore-pressure gradient in the avalanche head, which reduces the normal and shear stress transmitted by snow particles. Furthermore, the density of the head can be significantly reduced by the air flow, allowing the avalanche to reach a higher speed for a given value of the bed shear stress. This mechanism resembles that of fluidized-bed reactors and suggests that different fluidization regimes such as bubbling or spouting might also occur in snow avalanches – possibly explaining the intermittency observed in recent full-scale avalanche experiments at Vallée de la Sionne. We outline a minimal 1D depth-averaged mathematical model for dense/fluidized avalanche flow. The mass and momentum balance equations of the flow must be supplemented with evolution equations for the densities of the snow cover and the avalanche as well as an equation describing pore-pressure dissipation. To close these equations, one needs to specify (i) a relation for the compaction of the snow cover under a given load, (ii) a density-dependent rheology for the moving avalanche, and (iii) the effective permeability of the avalanche as a function of density, particle size distribution and air flow velocity. For (i), we fit data from laboratory studies of relatively rapid compression of snow. For (ii), we provisionally use the extended Norem–Irgens–Schieldrop model. For (iii), empirical relations such as the Ergun equation provide guidance. The excess pore pressure may be up to 50% or more of the total normal stress, crucially depending on the properties of the snow cover. This appears to explain the observed wide range of effective friction coefficients and the conditions favoring very long run-out. Typical pore-pressure dissipation times in dry-snow avalanches are estimated as 0.5–5 s (with large uncertainty), corresponding to a fluidized head 5–250 m in length, in good agreement with observations.
Dynamically aggregating avalanche forecast regions based on simulated snow profiles
Simon Horton, Florian Herla, Pascal Haegeli
Corresponding author: Simon Horton
Corresponding author e-mail: shorton@avalanche.ca
A potential application for avalanche forecasting with snowpack models is to group gridded simulations into areas with similar snowpack conditions. We present and test a method to aggregate small forecast regions into larger regions by analyzing simulated snow profiles from a weather–snowpack model chain. The study focuses on the Columbia Mountains of Canada, where avalanche forecasters have split the range into 20 small regions. Representative snow profiles were produced for each region at treeline elevations by averaging gridded SNOWPACK simulations over each day of the 2021/22 winter. Then a clustering algorithm was applied to dynamically group the small regions into large regions each day based on the similarity of the 20 representative profiles. The workflow applies recently developed snow profile averaging and clustering methods that focus on features relevant to avalanche hazard (e.g. new snow, weak layers). We test and illustrate the capabilities of this approach for several cases where local hazard assessments revealed variable avalanche conditions across the Columbia Mountains. Avalanche danger ratings and problems were taken from roughly 50 professional operations to find examples where regional-scale variability was caused by factors such as variable snowfall amounts or the presence/absence of specific weak layers. The ability of the model-based groups to reproduce these patterns is discussed to show the potential value and limitations of using this approach to identify variability in avalanche conditions at regional scales.
Modified propagation saw tests for analyses of weak-layer fracture properties
Valentin Adam, Bastian Bergfeld, Alec van Herwijnen, Philipp Weißgraeber, Philipp L. Rosendahl
Corresponding author: Valentin Adam
Corresponding author e-mail: valentin.adam@ymail.com
To improve our understanding of the release of dry slab avalanches, a failure criterion that is derived from and can be validated by field studies is essential. In this work, we present a novel experimental setup capable of measuring the fracture envelope of weak layers. The driving force of crack initiation and propagation is controlled by mixed mode anticrack phenomena. Here, mode I denotes a deformation normal to the crack faces and mode II an antisymmetric separation due to relative horizontal displacements. Using a state-of-the-art coupled stress and energy criterion, displays a promising agreement with mechanical observations of this specific failure process. While a principal weak-layer failure envelope in terms of strength was already established, a study of the interaction of mode I and mode II energy release rates is missing. For this purpose, modified propagation saw tests were performed to draw conclusions about this so-called fracture envelope. Extracted snow beams including the weak layer were fixed on a tilting device and loaded by additional weight. This provided the possibility of tuning the mode-mixity while performing the tests. To analyse the data, a closed-form mechanical model based on a beam on elastic foundation was used. The model is capable of describing the complex layered system and especially deformations, stresses and energy release rates within the weak layer, and is referred to latest developments in fracture mechanics. With this experimental setup, it was possible to cover the whole range from pure mode I to almost pure mode II crack propagation. About 100 tests enable the construction of a fracture envelope for weak layers that describes the interaction of the two crack propagation modes. Although the results only apply to the specially assessed weak layer, the methodology is applicable to all weak-layer types. The experiments mark an important step to better describe and validate failure models in order to improve the hazard assessment of snow slab avalanches.
Measurement and modelling of snow in Arctic tundra and taiga biomes
Georgina Woolley, Nick Rutter, Leanne Wake, Vincent Vionnet, Camille Garnaud, Chris Derksen, Richard Essery, Gabriel Hould Gosselin
Corresponding author: Nick Rutter
Corresponding author e-mail: georgina.j.woolley@northumbria.ac.uk
Seasonal snow cover is a defining aspect of the Arctic region that has a strong impact on numerical weather prediction, climate modelling and hydrological forecasting. Simulation of snow properties (e.g. depth, density, snow water equivalent, specific surface area) differ due to variability in snowfall, wind redistribution, the presence of vegetation and slope angle. This complex variability across increasing spatial scales is extremely hard to capture and replicate within land surface models (LSMs) leading to uncertainties in current simulations. Most snow components of LSMs have been developed within Alpine regions and fail to incorporate key Arctic processes causing errors in simulation of snowpack layering and properties within each layer. Evaluating the ability of current LSMs in simulating Arctic snowpack properties is therefore essential. To identify uncertainties in simulating Arctic snowpack properties, we evaluated the Soil, Vegetation and Snow version 2 (SVS2) LSM at Trail Valley Creek (TVC) and Havikpak Creek (HPC), NWT, Canada. Measurements of snowpack properties, from a field campaign in March 2022, show spatial variability along a 50 km transect between TVC and HPC is largely controlled by differences in vegetation. Preliminary model results show SVS2 underestimates the magnitude of snow depth at TVC due to uncertainties in the parametrisation of blowing snow sublimation. The timing of melt onset within SVS2 simulations is early in comparison to measured snow depths likely caused by an increase in simulated sublimation around February–April. A better understanding of these processes within SVS2 will aid future model modifications to improve simulation of snowpack properties within the Arctic region without compromising model performance in other biomes.
Julien Meloche, Melody Sandells, Nick Rutter, Henning Löwe, Ghislain Picard, Alexandre Langlois, Richard Essery
Corresponding author: Julien Meloche
Corresponding author e-mail: julien.meloche@usherbrooke.ca
Sea-ice thickness is essential for climate studies and numerical weather prediction. Radar altimetry has provided sea-ice thickness measurement since the launch of ERS-1 and currently through CryoSat-2, but uncertainty arises from interactions between the emitted signal and snow cover on the ice surface. Snow and ice have been shown to affect the altimeter waveform mainly due to their microstructure properties, surface roughness and salinity. Therefore, modelling the scattering of the electromagnetic waves with the snowpack and ice is necessary to retrieve the sea-ice thickness accurately. The Snow Microwave Radiative Transfer (SMRT) model can be used to simulate both snow cover and ice with salinity, which makes it suitable for modelling of the altimeter waveform echo from the snow-covered sea ice. This work focuses on measurements made as part of the Altimetric Ku-band Radar Observations Simulated with SMRT (AKROSS) project, which took place in Cambridge Bay in the Canadian Arctic in April 2022. This field campaign included traditional snowpack measurements from snowpit observations and SSA from both laser reflectance within an integrating sphere (IceCube instrument) and snow penetration force measured with a snow micropenetrometer. Additional microstructure information was also derived from X-ray tomography of snow and ice samples, which allowed for full evaluation of all microstructural models within SMRT in altimeter mode for the first time. Roughness measurements with a photogrammetric technique allowed reconstruction of the 3-D surfaces of snow and sea ice to understand the impact of this important parameter. These field observations were used to evaluate SMRT simulated waveforms against CryoSat-2 measurements and investigate the effect of the snow cover on the radar echo.
Investigating meteorological and snowpack drivers for glide-snow avalanching
Moritz Altenbach, Amelie Fees, Alec van Herwijnen
Corresponding author: Moritz Altenbach
Corresponding author e-mail: moritz.altenbach@icloud.com
Predicting glide-snow avalanche activity is very challenging for avalanche warning and safety services. This is largely due to our limited knowledge of the driving physical processes, as glide-snow avalanches have received limited scientific attention. To identify the main drivers associated with glide-snow avalanche release, we conducted a comprehensive investigation of meteorological and snowpack parameters for 883 glide-snow avalanche events. The glide snow events were documented with time-lapse photography at the Dorfberg field site above Davos, Switzerland, from 2009 to 2022. We obtained meteorological and snowpack parameters for every event by performing simulations at several representative locations across the Dorfberg with the 1-D snow cover model SNOWPACK. Modeled results were validated by comparing the simulated snow stratigraphy to observed snow profiles from the 2021/22 winter season. For our analysis, we separated cold- and warm-temperature events, as previously suggested. Results show that snow height, mean snow temperature, liquid water content and snow density, as well as air temperature and radiation fluxes, can help distinguish glide-snow avalanche days from days with no avalanches. For instance, in spring, when water production at the snow surface due to snow melt is an important driver, high incoming solar radiation and liquid water content of the snowpack were indicative of glide-snow avalanche activity. The number of glide-snow avalanches used in this study allowed us to establish robust statistical evidence for our findings. In the future, we therefore plan to develop a model to predict glide-snow avalanche activity using the most relevant parameters associated with glide-snow avalanche days.
Michael Lombardo, Peter Lehmann, Anders Kaestner, Amelie Fees, Alec van Herwijnen, Jürg Schweizer
Corresponding author: Michael Lombardo
Corresponding author e-mail: michael.lombardo@slf.ch
Liquid water at the soil–snow interface is thought to play a critical role in the release of glide-snow avalanches, which can be massive and threaten infrastructure in Alpine regions. Several mechanisms have been postulated to explain the formation of this interfacial water. However, these mechanisms remain poorly understood, in part because suitable experimental techniques are lacking. We demonstrate, for the first time, the use of neutron transmission radiography for investigating water transport across the soil–snow interface. The soil, vegetation and snowpack layering found in the field is simulated with a column of sand, gravel and snow. The column is connected to a water reservoir and placed in a climatic chamber within the neutron beam. We show that neutron transmission radiography is capable of measuring changes in the liquid water content distribution within the snow and soil phases during snow melt processes. The results also suggest that a porous interface between the soil and snow can induce the formation of an interfacial water layer. Improved understanding of the water transport across the soil–snow interface should lead to better prediction of glide-snow avalanche releases in the future and could also benefit other fields such as snow hydrology.
Using time-lapse photography to investigate glide-snow avalanches across time scales
Amelie Fees, Moritz Altenbach, Michael Lombardo, Alec van Herwijnen, Jürg Schweizer
Corresponding author: Amelie Fees
Corresponding author e-mail: amelie.fees@slf.ch
Glide-snow avalanches release due to a loss of friction at the snow–soil interface, which can result in large avalanches that endanger infrastructure in Alpine regions. The underlying processes leading to this loss of friction are still relatively poorly understood, in part due to the limited data available on glide processes. We have developed a new, pixel-based algorithm to automatically detect glide cracks from time-lapse photographs under changing illumination and shadow conditions. The algorithm is applied to 14 years of time-lapse photography at the Dorfberg (Davos, Switzerland), resulting in more than 900 detected glide-snow events. The high temporal resolution (2–15 min) of the resulting dataset allows us to perform the first comprehensive investigation of glide events across multiple time scales (from seasonal to diurnal). The data show that glide-snow avalanche activity varied greatly inter- and intraseasonally with more than half of recorded avalanches releasing without prior glide-crack appearance. In the course of the day most avalanches released in the afternoon hours (12–4 pm) with no significant difference between cold and warm temperature events. The extraction of glide rates through georeferencing shows that most (89%) glide cracks that eventually released as an avalanche did so within 48 hours after initial crack formation. An overall exponential evolution of glide distance was a good indicator for an avalanche release. In the future, we will expand this evaluation to include morphological, snowpack and soil parameters. By improving our understanding of the drivers of glide-snow avalanche release, we will be able to assist avalanche forecasters in predicting their release.
Unraveling the complexity of Saharan dust deposition in snow using citizen science
Marie Dumont, Simon Gascoin, Marion Reveillet, Didier Voisin, Collectif neige orange
Corresponding author: Marie Dumont
Corresponding author e-mail: marie.dumont@meteo.fr
In the beginning of February 2021, a large dust plume traveled from the Sahara across the Mediterranean Sea and deposited a colorful layer of particles on the snow-covered slopes of the Pyrenees and the Alps. The event was widely reported in the media due to the surprising color of the sky and of the snow cover and was followed by a second event that reached Scandinavia. To characterize the amount of dust deposited on the ground during this remarkable event, we organized a citizen science campaign. We collected 150 snow samples from which the deposited dust mass was measured over the Pyrenees, the French and the Swiss Alps. The analysis of all samples shows a robust deposition gradient from the Pyrenees to the Alps and enhanced deposition rates on south facing slopes in agreement with satellite data. The samples were used in combination with detailed snow modeling to evaluate the dramatic impact of the dust deposition on the melt and duration of the snow cover. This presentation will present the results of the sample analysis as well as the estimated impact on the snow cover. Collectif neige orange: Marie Dumont 1, Simon Gascoin 2, Marion Réveillet 1, Didier Voisin 6, François Tuzet 1, Laurent Arnaud 6, Mylène Bonnefoy 8, Carlo Carmagnola 1, Alexandre Deguine 9,10, Lukas Dürr 7 , Olivier Evrard 3, Frédéric Flin 1, Firmin Fontaine 8, Laure Gandois 4, Pascal Hagenmuller 1, Hervé Herbin 9, Béatrice Josse 5, Bruno Jourdain 6, Matthieu Lafaysse 1, Irene Lefevre 3, Gaël Le Roux 4, Quentin Libois 5, Samuel Morin 5, Pierre Nabat 5, Denis Petitprez 10, Ghislain Picard 6, Alvaro Robledano 6,1, Martin Schneebeli 7, Delphine Six 6 , Emmanuel Thibert 8 ,Jürg Trachsel 7, Matthieu Vernay 1, Léo Viallon-Galinier 1, Céline Voiron 6. 1: Université Grenoble Alpes, Université de Toulouse, Météo-France, CNRS, CNRM, Centre d’études de la Neige, France 2: CESBIO, Université de Toulouse, CNES/CNRS/INRA/IRD/UPSFrance 3: LSCE: Laboratoire des Sciences du Climat et de l’Environnement (Gif-sur-Yvette) , CEA, CNRS,UVSQ, Université Paris-Saclay, France 4: Laboratoire Ecologie Fonctionnelle et Environnement, Université de Toulouse, CNRS, INPT, UPS, Toulouse, France 5: Université de Toulouse, Météo-France, CNRS, CNRM France 6: Université Grenoble Alpes, CNRS, IRD, G-INP, IGE, France 7: WSL Institute for Snow and Avalanche Research SLF, Davos, Switzerland 8: Université Grenoble Alpes, INRAE, UR ETGR, Grenoble, France
Betty Sovilla, Cristina Pérez- Guillén, Anselm Köhler, Pierre Huguenin, Michael Kyburz, Camille Ligneau, Emma Suriñach
Corresponding author: Betty Sovilla
Corresponding author e-mail: sovilla@slf.ch
Powder snow avalanches (PSAs) are a major threat to people and infrastructure in many mountainous regions of the world. One of the main problems with PSAs is that they can develop a large turbulent suspension that can overflow any protective structure (e.g. dams), or reach very great heights potentially threatening, for example, power transmission cables. Understanding the mechanisms that control the growth and decay of the turbulent suspension in PSAs and assessing the amount of energy this layer can carry are therefore of paramount importance in order to assess the danger they can pose. Unfortunately, little knowledge is currently available to answer this open question. In this contribution, we explore the potential of infrasound measurements to monitor the growth and decay of the turbulent suspension layer of PSAs. To this end, we use infrasound recordings of hundreds of avalanches collected since 2008 at the Vallée de la Sionne test site, covering a wide range of sizes and various degrees of suspension layer development. Here we correlate the evolution of the amplitude and energy of the infrasonic signals in the time-frequency domain along with other dynamic data collected at the site, including, but not limited to, avalanche flow regimes and front velocity inferred from GEODAR radar data, as well as snowpack properties at the time of release calculated with the SNOWPACK model. This analysis shows that infrasound measurements clearly capture the various stages of evolution (growth, steady-state and decay) of the turbulent suspension. The differences in the dominant frequencies of the signals enable us to distinguish between dense suspensions carrying many particles and more dilute suspensions. This information, supplemented with the snowpack properties along the path, allows us to identify some of the key parameters that control the development of the suspension layer.
Impact of hectometric modelling of wind-induced snow transport on the soil thermal regime
Matthieu Baron, Ange Haddjeri, Matthieu Lafaysse, Rafife Nheili, Philippe Choler
Corresponding author: Matthieu Baron
Corresponding author e-mail: matthieu.baron@univ-grenoble-alpes.fr
Snowcover variability at small and meso-scale in mountainous environment has a huge influence on hydrology, soil temperature and plant spatial distribution, and is strongly driven by snow transport by wind. Taking advantage of growing computational resources, operational systems simulating snowpack evolution at mountain range scale currently tend to move toward hectometric resolution in order to enhance the representation of this local variability, which makes necessary to represent processes causing it. We present here a model predicting snow transport by wind coupled to the state-of-the-art snow model Crocus and designed to allow the simulation of snowpack evolution at a resolution of 250 m over the entire French Alps during several years and the effect the representation of this process has on the quality of soil temperature simulation. The intermediate resolution that is used and the need for computational efficiency led to the choice of a simple scheme decoupled from the atmosphere. Semi-empirical parametrizations are evaluated and used to quantify snow occurrence and blowing snow flux. Comparison between soil thermal regimes with and without representation of wind-induced snow transport is then performed, and both simulation set-ups are evaluated against measurement at a large number of sites in contrasted topographic situations.
Crack propagation in weak snowpack layers: do faster cracks propagate further?
Alec van Herwijnen, Bastian Berg
Corresponding author: Alec van Herwijnen
Corresponding author e-mail: vanherwijnen@slf.ch
Crack propagation in weak snowpack layers is a fundamental process in avalanche formation. It is now widely accepted that, before an avalanche releases, a crack propagates along a buried weak layer. Yet, relatively little is known about how far cracks in weak layers will propagate, and how this can be predicted. As for other materials, we hypothesize that crack propagation distance could be linked to crack speed. We therefore measured crack speed in field experiments with wireless accelerometers that we placed on the snow surface before triggering whumpfs or avalanches on small and safe slopes. We collected data in several field experiments, ranging from small whumpfs with crack propagation of less than 10 m to avalanches with crack propagation exceeding 100 m. In our experiments, crack speeds ranged from 10 m s–1 to more than 50 m s–1. Pooling our data with other published crack-speed estimates indicates that larger mean crack speeds are typically associated with greater crack-propagation distances. This clearly indicates that snowpack conditions favouring rapid crack propagation are less prone to crack arrest. Overall, our results thus show that faster cracks propagate further, and the key to predicting crack-propagation distance might well be predicting crack speed.
Michael L. Kyburz, Betty Sovilla, Yves Bühler, Alessandro Cicoira, Johan Gaume
Corresponding author: Michael L. Kyburz
Corresponding author e-mail: michael.kyburz@slf.ch
Since the advent of modern avalanche research in the first half of the twentiet century, avalanche modelling – first by means of hand calculations, later using computers – has been an essential tool to assess and mitigate snow-avalanche hazard. To date the most popular numerical approaches for simulating snow avalanches mostly adopt depth-averaged equations. These models are computationally efficient but limited in capturing processes occurring in the flow depth direction, e.g. granulation or densifications. In some cases, however, these processes cannot be neglected when simulating the avalanche flow, as they influence the flow dynamics and runout. In particular, the mass exchanges between the avalanche and the snowpack, such as snow erosion or deposition, are not yet fully understood and are therefore often neglected or introduced into the models using ad hoc parameterizations. These issues can be addressed by using a new three-dimensional (3D) model, based on the material point method (MPM) and elasto-plasticity theory. With this method, the mass-exchange processes between the avalanche and the snowpack can be simulated realistically by explicitly implementing a layered snowpack with appropriate snow properties for each layer. To assess the new possibilities and challenges associated with these highly detailed but computationally expensive calculations, we simulated the well documented ‘Salezer’ snow avalanche that released in Davos, Switzerland, in 2019. In order to reproduce the event in our simulations, we use the original release area mapped by a photogrammetric drone survey. While demonstrating how the snow conditions and the snowpack on the day of the event can be represented in the numerical model, we also show the limits of the degree of physical detail that can currently be represented in the MPM simulations. Finally, we compare macroscopic flow features, such as runout and deposition, of the simulated avalanche to the field observations and discuss what are the major challenges that still need to be addressed.
Andreas Stoffel, David Liechti, Yves Bühler
Corresponding author: Andreas Stoffel
Corresponding author e-mail: buehler@slf.ch
Automated snow weather stations are the backbone of avalanche warning. They are the most important sources of information for snow depth and the accumulation of new snow in poorly accessible Alpine terrain, also available under bad weather conditions. This is crucial to decide, for example, when a road has to be closed and when it can be re-opened. Because the spatial variability of the snow-depth distribution is extremely high in the mountains, the positioning of the automated measurements is critical. Spatial coherent snow-depth measurements with high spatial resolution (better than 2 m) from drones, aeroplanes and satellites demonstrate that very different snow-depth characteristics occur within very short distances of a few metres. Therefore, it is important to place the weather stations at locations with snow depths representing the mean snow depth of the surrounding region. Areas where the snowpack is strongly influenced by wind or avalanches, either removing or depositing large amounts of snow, have to be excluded. Also, regions endangered by avalanches need also to be removed. We develop an automated approach combining remotely sensed snow-depth maps with terrain characteristics (e.g. slope or homogenity) and simulated avalanche scenarios to identify optimal positions for automated weather stations. We demonstrate how this approach can help to improve safety relevant information within the Dischma valley close to Davos, Switzerland. This approach could be applied worldwide, where high-quality digital elevation models are available and spatial coherent snow-depth maps can be acquired. Today, the positioning of automated weather stations strongly relies on expert judgment. Our tool could help to make these decisions more comprehensible and serve as a second opinion for the final decision on the optimal location for automated weather stations.
Variational formulation of weak layer stresses in stratified snowpacks
Florian Rheinschmidt, Philipp Weißgraeber, Philipp Rosendahl
Corresponding author: Florian Rheinschmidt
Corresponding author e-mail: rheinschmidt@ismd.tu-darmstadt.de
Dry-snow slab avalanches are a severe hazard for infrastructure and back country and free ride skiers. Avalanche release strongly depends on the stratification of the snow cover and the mechanical properties of the individual snow layers. The condition of so-called weak layers in the snow pack is of particular importance. Overloading of these layers results in anticracks, which propagate as collapse until an avalanche is triggered. To provide an efficient stability assessment of stratified snowpacks, we present an analytical model for snow-cover deformations and stresses within the weak layer for arbitrarily layered snowpacks. In particular, the model covers the effect of the layering order on both the extensional and bending stiffness of the slab. The model can be used for externally loaded slopes and for stability tests such as the propagation saw test. It is highly efficient and can easily be used for parameter studies and implementation into other toolchains. While early approaches to modelling anticracks disregarded the weak layer and more recent models treat it as a simplified continuum, the present work accounts for the full weak-layer kinematics. It includes the influences of longitudinal stresses and allows more complex deformations of the weak layer. This results in an improved rendering of shear stresses in the weak layer and subsequent better predictions of the energy release rate.
Energy dissipation of anticrack propagation in a weak snowpack layer
Bastian Bergfeld, Alec van Herwijnen
Corresponding author: Bastian Bergfeld
Corresponding author e-mail: bastian.bergfeld@slf.ch
For a slab avalanche to release, a weak layer buried below a cohesive snow slab is required, and the system of weak layer and slab must facilitate crack propagation over large distances. This process, called dynamic crack propagation, is still rather poorly understood even though it is highly relevant for avalanche release. While models are nowadays able to simulate crack propagation over increasingly larger distances, validation data from field experiments are not yet available. We therefore performed a series of flat field propagation saw test (PST) experiments, up to 10 m long, over a period of 10 weeks on the same weak layer. Within this period, PST results evolved from crack arrest to full propagation and back to crack arrest. All PST experiments were analyzed using digital image correlation to derive high-resolution displacement fields. From these we determined the static specific fracture energy at the onset of crack propagation. In addition, we computed a dynamic fracture energy of the weak layer. To do so, we separated the work done in the weak layer during dynamic crack propagation in two parts. One part is the energy required to advance the crack ahead of the crack tip, namely the dissipation of dynamic fracture; the second part is used for weak-layer compaction, the elastic–plastic compaction part. Results showed that, in our leveled propagation saw tests, the dissipation due to compaction was around 30 times higher than the dissipation of dynamic fracture. The latter was in the range of 5 mJ m–2 to 0.43 J m–2 and therefore somewhat lower than the static specific fracture energy prior to crack propagation, which ranged from 0.1 to 1.5 J m–2. The dissipation of dynamic fracture alone is insufficient for self-sustained crack propagation in leveled terrain. The separation of the two energy dissipations can help distinguish between stable (small whumpfs) and unstable crack growth (remote triggering of avalanches) occurring in layered snowpacks. Overall, our dataset provides new insight into the dynamics of crack propagation and provides valuable data to validate models used to study this process.
Franziska Koch, Christian Voigt, Simon Gascoin, Karl-Friedrich Wetzel, Karsten Schulz
Corresponding author: Franziska Koch
Corresponding author e-mail: franziska.koch@boku.ac.at
Snow water equivalent (SWE) is an essential climate variable and has vital importance for the water cycle and the wellbeing of billions of people living in and downstream of mountain catchments. However, estimating its amount and the spatiotemporal distribution in complex high-Alpine terrain is currently considered as one of the most important challenges in Alpine hydrology. Besides, it is extremely difficult to measure or estimate further Alpine water storage components, e.g. karst reservoirs, and to examine the relationship between precipitation, evapotranspiration, storages, internal fluxes and discharge. Hydro-gravimetry is the method of observing temporal gravity variations as the integral of all hydrological mass variations on a wide spectrum from 1 s to several years after reduction of all other geophysical signals. So far, the terrestrial hydro-gravimetric method has been applied successfully for the direct, integral and non-invasive monitoring of water storage variations at several test sites. The Zugspitze Geodynamic Observatory Germany (ZUGOG), with its worldwide unique installation of a superconducting gravimeter at Mount Zugspitze on top of a well instrumented, snow-dominated high-Alpine catchment, is applied as a novel snow-hydrological sensor system within a radius of 4 km. In general, we want to investigate to what extent such a snow-hydro-gravimetric approach contributes to a better understanding and quantification of processes and storages in high-Alpine catchments. In this study, we will use this unique instrumental setup in synthesis with in situ measured data, detailed physically based snowpack and hydrological modelling as well as airborne and satellite-based data including high-resolution snow-depth maps derived by stereo photogrammetry. We will give an introduction into the novel sensor setup and will show first results, especially regarding the integrative gravimetric signal describing the seasonal high-Alpine snowpack.
NAKSIN: a system for creating high-quality avalanche hazard indication maps for large areas
Dieter Issler, Kjersti Gleditsch Gisnås, Peter Gauer, Sylfest Glimsdal, Ulrik Domaas
Corresponding author: Dieter Issler
Corresponding author e-mail: di@ngi.no
In sparsely populated avalanche-prone areas, hazard indication maps (AHIM) are a cost-effective tool for land-use planning. The Norwegian AHIM from 2010, based on a purely topographic–statistical approach, is too conservative in forested areas with maritime climate but too optimistic in regions with continental climate. To create a more reliable AHIM accounting for local climate, forest cover and path topography, NGI developed the modular Python 3 program package NAKSIN for the Norwegian Waterways and Energy Directorate (NVE). For a given mapping area of typically about 1000 km2, the first step is to identify all potential release areas (PRAs) based on topographic criteria such as local slope angle, terrain curvature, area and shape. For each identified PRA, several million synthetic ‘days’ are created from 60 years of daily data of air temperature, precipitation, snow depth and snow water equivalent on a 1 km2 grid, using a Gaussian kernel distribution function with a random component after interpolating to the altitude of the release area. For each sample ‘day’, the shear strength of a randomly placed weak layer is estimated from the weather and snow data and compared to the shear stress from the overlying slab in the infinite-slope approximation. Snow depth, shear strength and snow-cover support depend on the forest properties (tree density and diameter leaf-area index). Counting the number of ‘unstable days’, the release probability P_rel can be estimated and the fracture depth calculated. If P_rel exceeds 10–3 a–1, the run-out is calculated with the 2-D depth-averaged code MoT-Voellmy. The friction parameters are chosen on the basis of the values recommended for RAMMS::AVALANCHE, with mean winter temperature used instead of altitude zones and interpolation across the categories of avalanche release volume, terrain curvature and winter temperature. In the final step, all run-out zones in the mapping area are superposed to produce the AHIM. NAKSIN has been run for all of Norway and has produced an AHIM – to be published in the near future – whose quality often rivals that of hazard maps elaborated by experts. Ongoing work aims at correcting a tendency to underestimate the release probability in continental-climate areas and overestimate it in maritime zones, and to include the effects of wind transport, powder-snow avalanches and snow entrainment.
Explicitly resolving high-resolution, ridge-scale snow deposition patterns over long time periods
Dylan Reynolds, Michael Lehning, Michael Haugeneder, Rebecca Mott
Corresponding author: Dylan Reynolds
Corresponding author e-mail: dylan.reynolds@slf.ch
Snow models rely on accurate meteorological input data at the spatial scales at which they operate. However, even the highest-resolution operational atmospheric models often run at horizontal resolutions at least an order of magnitude coarser than most snow models. Different downscaling techniques can be employed to bridge this scale gap, typically being sorted into either statistical or dynamical techniques. Recent efforts have been made to optimize dynamic downscaling techniques, reducing computational demand while maintaining physical accuracy of predicted variables as well as the interdependency of downscaled variables such as winds and precipitation. Here we demonstrate an application of the High-resolution Intermediate Complexity Atmospheric Research Model (HICAR), a new variant of the ICAR model developed for spatial resolutions as high as 50 m. Relative to a traditional atmospheric model such as WRF, HICAR runs up to two orders of magnitude faster than WRF, while resolving terrain-induced effects on the wind field not seen in ICAR. This is achieved through a novel combination of adjustments to a background wind field based on terrain descriptors with a wind solver. The solver enforces a mass-conservation constraint on the 3-D wind field. These modifications successfully mimic dynamic effects such as flow blocking, ridge-crest speed up and lee-side re-circulation, which can be captured in the resulting wind field. These features are of particular importance for resolving snow-deposition patterns, where the snow particles are particularly susceptible to advection by the near-surface flow field. We validate the accuracy of HICAR’s flow features using a wind LiDAR deployed in complex terrain and show a comparison between flow fields from HICAR and WRF at a horizontal resolution of 50 m. These comparisons demonstrate HICAR’s ability to resolve terrain-induced modifications to the flow field that result in increased heterogeneity of ridge-scale snowfall patterns. On this point, preliminary comparisons of snow-deposition patterns in complex terrain between the HICAR model and snow accumulation during the 2021/22 winter are presented. With this new model, physically based downscaling of precipitation and other atmospheric variables which preserves their interdependencies is made available for high resolutions (100 m) and large spatial extents (10 000 km2), which are often demanded by operational land-surface models.
A two-layer depth-averaged model to simulate mixed snow avalanches on three-dimensional terrains
Hervé Vicari, Dieter Issler
Corresponding author: Hervé Vicari
Corresponding author e-mail: herve.vicari@ngi.no
Most of the existing depth-averaged avalanche models only consider a single layer of constant density and thus can only simulate wet or dense snow avalanches. Depending on the snow properties and terrain topography, a turbulent suspension layer may form from the dense flow and generate a mixed snow avalanche (MSA), which can have long runouts. Despite its significance, very few depth-averaged models have been developed to model MSAs. In the 1980s, Eglit et al. proposed a two-layer MSA mode l– further developed by Nazarov – with depth-averaged equations for a dense bottom layer and a dilute top layer. We extend the Eglit–Nazarov model to a three-dimensional topography. The dilute layer is generated by the dense layer through turbulent entrainment and grows in volume by entraining ambient air. The density of the powder snow cloud is therefore variable and computed by the model. The closure assumptions generally follow those in Eglit’s original model but differ from it in a few respects: At the avalanche–air interface, we apply an approximate expression for pressure drag. The bed-entrainment rate is given by the tangential jump entrainment model. Air entrainment at the top surface is governed by the bulk Richardson number, similar to the three-equation model of Parker and co-workers. The equations are solved using a forward Euler method and the simplest version of the Method of Transport, which is close to an upwind scheme. Exploratory simulations on simple topographies serve to study the behaviour of the model and to find reasonable ranges of key model parameters such as the settling velocity of suspended snow particles and the coefficient of the mass flux from the denser layer to the more dilute one. We observe a complex interplay between the two exchange rates: If settling dominates, e.g. due to large snow grains, or the avalanche moves slowly due to high bed friction and/or gentle terrain, the formation of the snow cloud is suppressed. Conversely, if the dense layer has relatively low density and consists of small or strongly dendritic particles, and the avalanche has high speed on a steep slope, turbulent entrainment is rapid and particle settling slow, which can lead to depletion of the denser layer, i.e. a nearly pure PSA. In the run-out zone, the powder snow cloud may continue a fair distance beyond the stopping point of the dense core. This wide spectrum of behaviour closely resembles what is observed in real snow avalanches.
Forcing the snow-cover model SNOWPACK with CaPA precipitation data for avalanche risk assessment
William Durand
Corresponding author: William Durand
Corresponding author e-mail: william.durand@usherbrooke.ca
Avalanches have long been one of Canada’s greatest geohazards. Today, outdoor recreationists are increasing in number and represent most avalanche victims in Canada. Organizations such as Avalanche Canada and Avalanche Quebec have the responsibility to inform and educate the public wishing to engage in this type of activity, which carries significant risks. To do so, public avalanche safety organizations issue daily avalanche forecasts throughout the winter season based on avalanche observations, punctual snowpack stability tests and weather forecasts to estimate the risks. However, manual snow-cover observations are time-consuming, expensive and lack spatially distributed information to precisely cover the entire study area. Thus, the improvement of avalanche forecasting and management programs is the subject of several current studies, but few are based on the emergence of new technologies such as meteorological modelling to simulate the snow cover in order to provide forecasters with decision-support tools for issuing public avalanche bulletins. Weather conditions in Alpine environments are highly variable spatially and temporally due to orographic effects, which results in significant variations in the snow cover. Snow simulation models such as SNOWPACK are used to study snow-cover stability. However, the lack of spatial coverage of the weather stations used to force such a model remains the main limiting factor in the mountains. Since 2011, Environment and Climate Change Canada has been developing the Canadian Precipitation Analysis (CaPA). The High-Resolution Deterministic Precipitation Analysis (HRDPA) generated by the CaPA system allows a quantitative estimation of precipitation in near real time with a spatial resolution of 2.5 km for most of Canada. The main objective of this study is to evaluate the potential of numerical simulations of snow cover by the SNOWPACK model when used with CaPA precipitation data, corrected using data from meteorological stations, in the perspective of avalanche forecasting. The validation and calibration of precipitation simulations from the CaPA system should lead to improved precipitation modelling in Alpine environments. The operationalization of such a snow cover modelling method would allow avalanche forecasters to cover vast territories used by outdoor enthusiasts daily and therefore reduce the number of avalanche casualties in Canada.
Local snow avalanche monitoring based on automotive lidar and radar sensors
Stefan Muckenhuber, Thomas Goelles, Birgit Schlager, Kathrin Lisa Kapper, Alexander Prokop, Wolfgang Schöner
Corresponding author: Stefan Muckenhuber
Corresponding author e-mail: stefan.muckenhuber@uni-graz.at
Comprehensive data of local snow avalanche releases are indispensable for several purposes, e.g. monitoring locations with high-risk potential, verifying artificial avalanche triggering, and improving algorithms for satellite-based avalanche detection. Existing lidar and radar technologies for monitoring local avalanche activity are costly and require closed-source commercial software. These systems are often inflexible for exploring new purposes and too expensive for large-scale applications, e.g. 100–1000 slopes. Therefore, developing reliable and inexpensive measurement and monitoring techniques with cutting-edge lidar and radar technology is highly required. Today, the automotive industry is a leading technology driver for lidar and radar sensors, because the greatest challenge for achieving the next level of vehicle automation is to improve the reliability of its perception system. Automotive lidar sensors record high-resolution point clouds with very high acquisition frequencies of 10–20 Hz and a range of up to 250 m. High costs of mechanically spinning lidars (5–20k €) are still a limiting factor, but prices have already dropped significantly during the last decade and are expected to drop by another order of magnitude in the coming years. Modern automotive radar sensors operate at 24 GHz and 77 GHz, have a range of up to 300 m, and provide raw data formats that allow the development of algorithms for detecting changes in the backscatter caused by avalanches. To exploit the potential of these newly emerging, cost-effective technologies for geoscientific applications, a stand-alone, modular sensor system called MOLISENS (MObile LIdar SENsor System) was developed in a cooperation between Virtual Vehicle Research Center and University of Graz. MOLISENS allows the modular incorporation of cutting-edge radar and lidar sensors. The open-source Python package ‘pointcloudset’ was developed for handling, analyzing and visualizing large datasets that consist of multiple point clouds recorded over time. This Python package is designed to enable the development of new point-cloud algorithms, and it is planned to extend the functionality to radar cluster data. Based on MOLISENS and pointcloudset, a strategy for their operational use in local avalanche monitoring will be developed. Test data will be collected before, during and after avalanche blasting in collaboration with the Skiresort Lech and with the SnowScan GmbH.
What are the chances of surviving in a building hit by an avalanche?
Kate Robinson, Nezam Bozorgzadeh, Dieter Issler, Zhongqiang Liu
Corresponding author: Kate Robinson
Corresponding author e-mail: kate.robinson@ngi.no
A recent study in Norway (the FOSS2040 project) showed that, in order to fulfil current building-code requirements, nearly 50 000 existing buildings across Norway may require protection measures against natural hazards from steep terrain, such as avalanches. This amounts to a cost estimate in the order of 2.0–4.5 billion euros. In addition, municipalities with expanding populations in mountainous terrain may wish to expand areas zoned for new building construction into regions potentially exposed to mountain hazards. Therefore, two challenges involved in risk assessment by decision makers are prioritizing existing buildings for security measures, and quantifying the risk of expanding acceptable new building locations into potentially hazardous areas. Risk is a function of hazard, exposure and vulnerability; this study focuses on assessment of vulnerability of people present in buildings when an avalanche strikes as a function of degree of damage to the buildings. To this end, a dataset of historical avalanches in Norway (starting from the year 1900) is gathered and filtered to 76 avalanches impacting 154 buildings which include complete information on the number of people present in the building and whether they survived the avalanche, together with sufficient information to allow an assessment of the degree of damage to the building. The degree of damage is assigned based on a damage scale that reflects, from 0% to 100%, the extent to which the building was able to maintain not only structural integrity but also open spaces free of snow and debris. Vulnerability curves are then established using Bayesian logistic regression models relating the degree of structural damage to probability of survival. A considerable component of uncertainty in developing such vulnerability curves from the available data arises from the assessment of building damage, which is subjective and based on qualitative information such as newspaper articles and personal documentation. To better understand this uncertainty, multiple individuals provided independent assessments of the degree of damage levels. Furthermore, each individual was asked to report a 95% confidence range associated with their best assessment. When possible, the damage was assessed on a per-floor basis for each building. The challenges in including these uncertainties in the regression models are discussed, along with their effect and ramifications for the suggested vulnerability curves.
Michael Haugeneder, Michael Lehning, Dylan Reynolds, Tobias Jonas, Rebecca Mott
Corresponding author: Michael Haugeneder
Corresponding author e-mail: michael.haugeneder@slf.ch
Snow–atmosphere interactions drive the mass and energy balance of the snow pack throughout the whole snow season. In late spring, when the snow cover gets patchy, the extreme surface heterogeneity induces complex atmospheric processes as the lateral advection of heat or the development of thin stable internal boundary layers (SIBL) over the leading edge of snow patches. We aim at a better understanding of these near-surface atmospheric processes. Therefore, we conducted a comprehensive field campaign at an Alpine research site. We measured meteorological parameters, snow ablation patterns, and turbulence characteristics using eddy-covariance sensors. Additionally, we applied a novel experimental method. A high resolution thermal infrared camera records a 30 Hz sequence of infrared frames. The camera points at vertically deployed thin, synthetic screens. The screens cover a horizontal distance of 6 m across the transition from bare ground to snow. The surface temperature of the screens serves as a proxy for local air temperature. The recorded air temperature fields capture the dynamics of turbulent eddies adjacent to the surface depending on different parameters such as wind speed or the snow coverage. A thin SIBL develops above the leading edge of a snow patch possibly protecting the snow surface from warmer air above. However, sometimes the warm air entrains into the SIBL and reaches down to the snow surface, adding further energy to the snow pack. To quantify exchange processes from observed spatio-temporal dynamics, we developed a method to estimate a near-surface 2-D wind field from tracking temperature pattern on the screens. Resulting vertical profiles of air temperature, horizontal and vertical wind speeds can be evaluated with a high spatial (0.01 m) and temporal (30 Hz) resolution. Combining the screen measurements with data from eddy-covariance sensors enables us to gain an extensive overview of the (sub)meter-scale heat exchange processes. For example, we can investigate the influence of laterally advected heat on vertical turbulent sensible heat fluxes within the atmospheric layer adjacent to the surface. With the data we aim to better understand and quantify small-scale energy transfer processes over patchy snow covers and their dependency on the atmospheric conditions. This will ultimately allow the improvement of parameterizations of these processes in coarser-resolution snow-melt models.
A multi-scale constitutive model for snow engineering
Marie Miot, Pit Polfer, Antoine Wautier, François Nicot, Pierre Philippe, Tibor Fulop
Corresponding author: Marie Miot
Corresponding author e-mail: marie_miot@goodyear.com
Snow is a granular material built of ice grains, which implies two important points: its mechanical response depends on both the microstructure and the ice behavior. Multiple numerical approaches can be used to model snow-like granular materials. The Discrete Element Method is explicitly accounting for the microstructure and its evolution over loading. A drawback is that explicitly modeling each ice grain can be computationally very expensive. The continuum approaches such as the Finite Element Method can reduce the computation cost, but the existing constitutive models do not precisely account for the microstructure. The current work proposes a multiscale constitutive modelling approach, based on the 3-D H-model initially developed for geomaterials. The 3-D H-model is a multiscale constitutive law that accounts for the microstructure of a granular material. The core ingredient is the description of a representative elementary volume (REV) as a statistical distribution of 3-D grain cells with different orientations. The use of a bi-hexagonal structure to describe the grain cells allows the derivation of an analytical relationship between the strains and the stresses acting within each cell. The response also includes the complex contact interaction between the grains. To account for the particularities of snow, the grain–grain interaction is updated in the H-model. The contacting grains are initially linked by ice bonds, described as a constant volume elasto-viscoplastic beam. When a bond between two grains fails, a residual elasto-frictional contact between those grains applies. The upscaling process allows the snow behavior to be modeled at the specimen scale (REV). Unconfined compression tests and triaxial tests at the REV scale were numerically performed to demonstrate the capability of the H-model to describe the complex snow behavior.
Fast tomography of snow microstructure
Pascal Hagenmuller, Neige Calonne, Marie Dumont, Julien Brondex, Kévin Fourteau, Jacques Roulle
Corresponding author: Pascal Hagenmuller
Corresponding author e-mail: pascal.hagenmuller@meteo.fr
Characterizing snow microstructure is essential to predict snow evolution and its properties. On the one hand, tomography can be used to fully describe the three-dimensional arrangement of ice and pores at a micron resolution. However, the scanning time generally hinders its routine use in the field. Indeed, a desktop tomograph requires 1 hour to scan a centimetric sample at a resolution of 10 μm. On the other hand, snow profiles of penetration resistance or reflectance can be obtained in the field to quickly capture the whole snowpack stratigraphy. However, the inversion of snow profiles into microstructural proxies relies on error-prone empirical relations and partly captures the variety of microstructural patterns. For instance, while high-resolution penetration profiles provide reliable snow-density profiles, their translation in terms of specific surface area or grain shape remains somewhat problematic. We developed a fast tomography method for snow to bridge the gap between an explicit characterization of snow microstructure and in-situ measurements of the whole snowpack. We reduced the nominal voxel size to 42 μm to scan snow at a rate of about 1 cm of height per minute scanning time. At this resolution, the voxel size is close to the typical size of snow microstructural features, which, consequently, cannot be fully resolved. In particular, most of the voxels are mixed, i.e. they represent a region composed of air and ice. With this partial volume effect, the segmentation of the voxels into the air and ice phases becomes ambiguous. Therefore, we developed a statistical model that directly interprets the greyscale tomographic images into microstructural descriptors, skipping the step of segmentation into binary images. We evaluated the approach by comparing results from the fast tomography method and the standard one. We applied it to data collected at Col de Porte, France, during the winter season 2021/22. The presented methodology opens the way to monitoring the seasonal evolution of snow microstructure in the field.
Renée Mie Fredensborg Hansen, Eero Rinne, Knut Vilhelm Høyland, René Forsberg, Henriette Skourup
Corresponding author: Renée Mie Fredensborg Hansen
Corresponding author e-mail: rmfha@space.dtu.dk
The orbit manoeuvre, known as CRYO2ICE, that periodically aligned CryoSat-2 with ICESat-2, allows for unprecedented near-coincident radar and lidar observations currently targeting the polar regions of the Northern hemisphere. This is of particular interest to sea-ice thickness studies, since snow on sea ice remains the largest contributor to sea-ice thickness uncertainties. Snow-depth estimates from space have been acquired from passive microwave radiometers and by using dual-frequency observations (Ku- and Ka-band, or laser and Ku-band). However, dual-frequency observations have until now only been based on monthly averaged estimates at basin scales. CRYO2ICE presents the possibility of investigating along-track snow depth using observations at various wavelengths along with an opportunity for further investigation of penetration capabilities and footprint-related issues. We examine near-coincident radar and laser freeboards from CryoSat-2 and ICESat-2 (CRYO2ICE observations) and the resulting snow-depth observations. Our particular focus is on how the CryoSat-2- and ICESat-2-derived freeboards respond along-track to various conditions and how this affects snow-depth retrieval. This study will investigate the freeboards and derived snow depth in relation to changes in surface roughness, sea-ice concentration and sea-ice lead identifications, and show comparisons of radar and laser freeboards with auxiliary data. Some of the most noticeable differences between CryoSat-2 and ICESat-2 are measurement configuration and sampling rates. This difference in measurement configuration between retrieving surface elevation using conventional methods such as re-tracking the SAR radar waveform of CryoSat-2 in comparison to re-tracking the surface from high-density photon clouds of ICESat-2, as well as the difference in sampling rates, presents additional challenges. This study will discuss the challenge of binning these very different types of observation strategy into comparable observations and what we currently can expect to get from the CRYO2ICE observations over sea ice. This is especially interesting in preparation for the upcoming CRYO2ICE Antarctic campaign initiated in June 2022, where the CRYO2ICE observations will be targetingd the Southern hemisphere – a much more challenging environment when it comes to space-derived freeboards and snow-depth observations.
Applying morphological indexes to describe snow-cover patterns in a high Alpine area
Lucia Ferrarin, Franziska Koch, Karsten Schulz, Daniele Bocchiola
Corresponding author: Lucia Ferrarin
Corresponding author e-mail: lucia.ferrarin@mail.polimi.it
The spatiotemporal distribution of snow cover has a large impact on many processes at different scales, including the Earth’s energy balance, the hydrological cycle and ecosystem functions. Studying the variability of snow cover in space and time is therefore fundamental to understanding the implications that it has on many aspects of human life. Although it is known that the occurrence of snow-cover patterns is mainly driven by topography and wind, they are still complex to describe in their spatial and temporal dynamics. Moreover, it is difficult to obtain reliable and systematic field measures of their features, as many regions influenced by seasonal snow are difficult to access. However, remote-sensing techniques allow study ofsnow-cover patterns even in complex terrain and testing of new approaches to describe such patterns. In this study, we evaluate the effectiveness of different morphological indexes in quantitatively describing Sentinel-2-derived snow-cover maps of the high Alpine catchment Zugspitze. The indexes used in this study are the three Minkowsky measures, representing area, perimeter and Euler characteristic, and the average value of the chord-length distribution, computed on snow-cover patterns. To ourknowledge, these measures have never been applied in the field of snow monitoring before. The goal of this study is to evaluate whether they can provide useful information towards a better understanding of processes that control snow cover distribution, which could help in a later step to improve modelling. In this context, we investigate how these indexes are affected by topographic features (e.g. aspect, slope) and meteorological variables, and how they are correlated to snow-related variables (e.g. snow water equivalent, snow depth). The topographic features and the meteorological variables that most affect snow deposition, redistribution and melting, will be shown through the analysis of correlation indexes. As expected, the correlation between snow-pattern features and snow water equivalent and snow depth was high. In general, this study demonstrates that Minkowsky indexes and the average chord length retrieve useful information on the dynamics of snow-cover patterns, although further investigations are needed to evaluate whether such pattern descriptors can be used to improve the accuracy of the understanding and modelling of snow-related processes.
The effects of precipitation particle shape on snowpack
Takahiro Tanabe
Corresponding author: Takahiro Tanabe
Corresponding author e-mail: tnabe@bosai.go.jp
Snowflakes reflect meteorological information in the sky such as the temperature and degree of supersaturation, and their shape is formed by the interaction between them. These shapes affect not only how the snow falls (e.g. snow-fall speed) but also how the snowpack is formed. Although these shape differences of precipitation particle are negligible after mechanically and thermally induced metamorphosis in snowpack, it causes differences in the mechanical properties of snowpack in their early stages (new snow). In addition, it is said that some kinds of precipitation particle from cyclones form a weak layer within the snowpack, which causes the potential occurrence of surface avalanches over a broad area simultaneously. Our interest is to understand the snowpack properties caused by differences in precipitation-particle and environmental conditions from the granular scale, e.g. the effects on snowpack with/without rimed snow crystal and with/without wind. In nature, it is difficult to observe the desired precipitation particles freely in a controlled environmental setting. In order to solve these problems, numerical granular sedimentation experiments with non-spherical grains were conducted. Clump particles constructed by spheres are employed to resemble precipitation particles in a discrete element method (DEM). To make granular packing, free fall of clumps is simulated with changing the shape of clumps and initial conditions (initial orientation and initial velocity/rotational velocity) as parameters.
The relationship between flow regimes and mass exchange processes in snow avalanches
Camille Ligneau, Betty Sovilla, Xingyue Li, Johan Gaume
Corresponding author: Camille Ligneau
Corresponding author e-mail: camille.ligneau@slf.ch
The mass of an avalanche is a crucial parameter for hazard assessment because it significantly affects its run-out distance and thus its danger potential. After the release, the avalanche interacts with the snow cover along its run, dragging initially static snow into the flow and depositing snow at the same time. Dense avalanches usually take in snow through frontal or basal entrainment, or also when secondary slabs release along the path and merge with the existing flow. Snow deposition mostly occurs at the tail but sometimes also at the side edges of the flow, forming levees that channel the avalanche and increase the run-out distance. Entrainment and deposition govern the mass change of the avalanche, but a gap still exists in the quantitative assessment of erosion/deposition fluxes, especially when considering the different mechanical properties of the flowing snow and erodible bed (e.g. dry vs wet snow avalanches). With a view to filling this gap, the present work aims to investigate how dense snow avalanches in different flow regimes interact with an initially static and metastable snow cover. Two-dimensional discrete element modeling (DEM) is used to simulate snow as a cohesive granular material. The particles interact through a contact model that can be tuned in terms of cohesion and friction to match a realistic range of mechanical properties typical for dry and wet snow. We naturally reproduce various erosion mechanisms including frontal ploughing and basal erosion and we describe how the particles are taken in and mixed into the flow in various combinations of flow regime and snow cover properties. In particular, frontal ploughing and basal erosion are naturally reproduced in the simulations. Further, we examine how the erosion, entrainment and deposition mechanisms are impacted by the snow cohesion, slope angle and depth of the erodable bed. Finally, we discuss the relevance of such 2-D meso-scale findings for full-scale 3-D models that use depth-averaged equations.
Modelling the simultaneous heat conduction, liquid water transport and melt/refreezing in a snowpack
Kévin Fourteau, Julien Brondex, Pascal Hagenmuller, Matthieu Lafaysse, Laurent Oxarango, Marie Dumont
Corresponding author: Kévin Fourteau
Corresponding author e-mail: kevin.fourteau@meteo.fr
Current snowpack models treat heat transport, liquid water transport and melt/refreeze processes in a sequential manner: heat and liquid water are transported independently, as if no phase changes occurred, and then thermodynamical equilibrium is re-established by locally melting or refreezing water. This approach leads to nonphysical heat and water transport characterized by local overshoots, since transport is not impeded by phase changes. To overcome this limitation, we propose a new mathematical formulation of energy conservation in snow in which the heat diffusion, liquid water transport and phase change are solved in a fully coupled manner. The formulation is chosen to be valid over the entire range of possible snow physical states, from dry to water-saturated, and this in order to effortlessly transition from dry to humid snow. For numerical simulations, the resulting physical model is translated as a nonlinear algebraic system. As this system is strongly non-linear, due to the built-in Richards water transport equation, numerical stabilization techniques are explored in order to ensure convergence even in the case of constant 15-minute time steps. First results show that this approach is able to handle the complex situation of a snowpack with dry and humid zones, locally exchanging heat and liquid water in the presence of phase changes.
Kathrin Lisa Kapper, Stefan Muckenhuber, Thomas Goelles, Andreas Trügler, Muhamed Kuric, Jakob Abermann, Eirik Malnes, Jakob Grahn, Alexander Prokop, Birgit Schlager, Wolfgang Schöner
Corresponding author: Kathrin Lisa Kapper
Corresponding author e-mail: kathrinlisa.kapper@uni-graz.at
Knowledge of the snow avalanche coverage and activity of a larger region is essential for a variety of applications, e.g. avalanche warnings and hazard management. Recent advances in automated machine-learning-based algorithms set the scene for fast and comprehensive detection of avalanches from satellite images. In the framework of this project, we develop an automated state-of-the-art avalanche-detection system for the Austrian Alps, including a best-practice data-processing pipeline and a learning-based approach applied to synthetic aperture radar (SAR) satellite images. For this purpose, we make use of the openly available Copernicus Sentinel-1 SAR images that have successfully been used for avalanche detection, together with a variety of published training data sets. In a first step, the labelled training data, which comprise around 26 000 manually detected avalanche outlines from Switzerland and Greenland, were downloaded and preprocessed. The SAR images were selected to correspond to the regions and time slots of the training data and were preprocessed to yield optimum detection results. In addition, SAR images from the Austrian Alps from an avalanche-rich winter season will be used to evaluate how well our detection algorithm generalizes to this independent data that is potentially differently distributed from the training data. Furthermore, selected ground-truth data from Switzerland, Greenland, and Austria will allow us to validate the accuracy of the detection approach. As a novel approach to improve detection performance, we propose to include encoded weather data into the avalanche detection pipeline. The weather data include several meteorological parameters, such as precipitation and wind speed, over a certain time range that are downscaled to fit the corresponding pixels of the SAR image. In this way, simple averages of meteorological parameters but also full snowpack models can be aggregated over time.
Kévin Fourteau, Henning Loewe, Johannes Freitag
Corresponding author: Kévin Fourteau
Corresponding author e-mail: kevin.fourteau@meteo.fr
Slow, viscous deformation is one of the key processes for snow, firn and bubbly ice. The relevant material property is the compactive viscosity, which must be parametrized in any snow or firn model for calculating the densification. Accurate predictions of densification rates in snow or firn have major implications, ranging from the stabilization of persistent weak layers (depth hoar) responsible for avalanches to the understanding of the enclosure of atmospheric gases in ice cores. Existing approaches for the compactive viscosity in literature show a vast diversity of conceptual starting points, free parameters, deformation mechanisms, relevant microstructural parameters, and considered density ranges. This prevents a conclusive, commonly accepted picture of compactive viscosity. In contrast to other effective properties (thermal conductivity, elastic moduli), the compactive viscosity of snow firn and bubbly ice has never been addressed within microstructure-based simulations over the entire density range by simply invoking stated literature values for the ice material parameters. To this end we conducted massively parallel finite element simulations with ELmer/Ice for computing the effective compactive viscosity from X-ray tomography images from the Alps and Antarctica. First we address the celebrated picture of ‘snow as a foam of polycrystalline ice’ and employ an isotropic rheology (Glen’s law) from literature as ice constitutive law. Second we address the immediate extension of ‘snow as an aggregate of mono-crystalline grains’ where we employ geometrical segmentation for disassembling the microstructure into grains that are equipped with a (randomly oriented) transverse isotropic constitutive law to mimic the single-slip rheology of ice monocrystals. Within the unavoidable uncertainties in existing observations, our simulations for compactive viscosity yield good agreement in deep firn. In contrast, in snow the different employed material models are not able to reproduce the observed compactive viscosities. In particular, none of the non-linear viscous (Glen-type) material models could fully explain the much higher compactive viscosity of snow that underwent temperature gradient metamorphism (depth hoar) compared to rounded grains, a distinctive characteristic of snow densification.
Paul Billecocq, Alexandre Langlois, Benoît Montpetit, Jean-Benoît Madore
Corresponding author: Paul Billecocq
Corresponding author e-mail: paul.billecocq@usherbrooke.ca
Snow cover in the mountain ranges of Canada has great social economical and environmental impacts on Canadians. It is one of the main drivers for winter tourism, attracting tourists both nationwide and internationally to visit ski resorts and National Parks, and snowmobile various areas. However, these mountains also include avalanche terrain. Avalanches can strongly alter the mountain’s landscape, naturally clearing timber and severely damaging human infrastructure such as roads and railways, and are a deadly hazard to recreationists. Moreover, this seasonal snow will eventually melt, feeding rivers and the ecosystems that depend on them, filling reservoirs for hydropower, and act as freshwater resources for cities and agriculture. As such, monitoring the snowpack dynamics through the winter season would be highly beneficial, especially in an era where extreme weather events that lead to both positive and negative anomalies are becoming increasingly common. Satellite remote sensing can help to provide information, but snow-property inversion techniques are still being developed and they need validation data over a wide spectrum of snow conditions. Collecting in-situ measurements over such a wide and difficult-to-access terrain remains seriously challenging. However, snow models are becoming increasingly reliable, although they need accurate weather data to run properly, and the Canadian mountains are too vast to be equipped with a dense enough array of weather stations. Furthermore, the Canadian High-Resolution Deterministic Prediction System (HRDPS) currently has a resolution of 2.5 km × 2.5 km, which is too coarse to capture the spatial variability of snow in a complex topography. In this study, we investigate whether the HRDPS forecast can be subgridded and used as input for snow modelling over two winter seasons in Glacier National Park, British Columbia, Canada. Air temperature, precipitation and wind speed were first statistically parametrized as regards elevation using six automatic weather stations (AWS). Alpine3D was then used to spatialize weather parameters and radiation input accounting for terrain reflections, and perform the snow modelling. Modelled profiles were compared to profiles generated with AWS data and in-situ profiles when available with a Distance Time Warping algorithm designed in a remote-sensing perspective. Similarity was computed on layer density and optical grain size, two key parameters for radiative transfer in the snowpack.
Victor Nussbaum, Neige Calonne, Matthieu Lafaysse, Léo Viallon-Galinier
Corresponding author: Matthieu Lafaysse
Corresponding author e-mail: matthieu.lafaysse@meteo.fr
Snow models simulate the physical processes of snow to predict the state of the snowpack in 1-D under given boundary conditions. They are useful for various applications such as avalanche hazard forecasting, water resource management, climate change impact studies, or snow–atmosphere coupling in atmospheric modelling. Among the state variables of snow models, the two main structural properties are density (proportion of ice and air in a given snow volume) and specific surface area (SSA), which is related to snow microstructure and is defined as the ice–air interface surface area over the snow mass. Accurate simulations of these variables are crucial for most physical and mechanical snow processes. However, only coarse evaluations of density and SSA simulations have been performed up to now because of insufficient data and the challenge to define an appropriate methodology. This contribution presents a detailed evaluation of the density and SSA simulations from the French model Crocus based on the RHOSSA dataset, which provides a unique time series of density and SSA from different instruments over the 2015/16 winter season at the Weissfluhjoch research site, Swiss Alps. It offers a high vertical (cm) and temporal resolution (weekly to daily depending on the sensors). First, we use a layer-based evaluation methodology to compare various Crocus configurations (with different physical options and sets of parameters) and outputs of the Swiss SNOWPACK model in their performance to simulate the density and SSA temporal evolution for specific snow layers. Then, the full simulated and observed property profiles are compared using a matching data algorithm. This method generalizes some results obtained from the layer-based evaluation, but with intrinsic limitations of this automatic process. Our results emphasize the potentials and limitations of detailed snow models to simulate internal snow properties, and allows discrimination of different physical parameterizations with more accurate criteria than classical evaluations of bulk properties. The results also identify some critical processes requiring future model improvements, including metamorphism-induced compaction. Such evaluations will be extended to more sites in the future with similar field campaigns in progress at Col de Porte (French Alps) and planned in Alaska for winter 2023/24.
Microstructure-based modelling of snow visco-plasticity
Louis Védrine, Antoine Bernard, Pascal Hagenmuller, Lionel Gélébart, Maurine Montagnat, Kévin Fourteau
Corresponding author: Louis Védrine
Corresponding author e-mail: louis.vedrine@ens-paris-saclay.fr
The porous structure of snow densifies with time under gravity. Knowledge of snow settlement is essential for many applications, such as paleoclimatology and avalanche forecasting. Snow densification is mainly driven by the creep of the ice matrix undergoing viscoplastic deformation. Although macroscopic behavior has been previously explored through experiments, little is known about the driving mechanism at the microscale and the influence of the microstructure and crystalline texture on snow viscosity. In this study, we modeled the viscoplastic behavior of snow based on 3-D images of its microstructure and the mechanical properties of ice. We conducted the simulations with a Fast Fourier Transform-based numerical solver (AMITEX_FFTP) to avoid the complex meshing procedure and benefit from high-performance computing facilities. We used two different viscoplastic models for ice at the microscale. The ice matrix in the snow is considered as either (i) a homogeneous structure of polycrystalline ice modeled with a 3-D Norton–Hoff law or (ii) a heterogeneous structure composed of sintered crystals for which deformation by dislocation glide on slip systems is modeled. We compared our numerical experiments to oedometric compression tests captured by tomography. We showed that snow could not be considered a foam of polycrystalline ice but rather an ensemble of sintered ice crystals. We also explored the continuum of ice texture between single crystals and polycrystalline ice by investigating the model sensitivity to the contact area between ice crystals. Overall, this study contributes to developing improved formulations of snow settlement in detailed snowpack simulation tools. It provides a tool to perform mechanical numerical experiments on various snow microstructural patterns. Most of the computations presented in this paper were performed using the GRICAD infrastructure, which is supported by CNRS, University Grenoble Alpes and INRIA.
Influence of slab depth spatial variability on skier triggering probability and avalanche size
Francis Meloche, Johan Gaume, Louis Guillet, Francis Gauthier, Alexandre Langlois
Corresponding author: Francis Meloche
Corresponding author e-mail: francis.meloche@uqar.ca
Snow avalanches represent a major risk for backcountry recreationists. Spatial variability of snowpack properties adds uncertainty to the challenging task of route finding and decision-making in avalanche terrain. To better understand this issue, we propose a combined mechanical–statistical approach to study how spatial variation of slab properties affects the skier triggering probability and possible release size. First, we generate multiple slab depth maps over a 200 m × 100 m slope based on Gaussian random fields (GRFs) for a specific set of mean, variance and correlation length. For each GRF, we found analytically the skier crack length, critical crack length and a new metric, the skier propagation index (SPI). We then simulate multiple skier tracks using a sinusoidal trajectory with different values of spacing between skiers and track radius. A positive skier hit is recorded if the SPI is below 1. The probability of skier triggering is then computed based on the number of hits. Finally, we use the depth-averaged material point method (DAMPM) to evaluate the possible avalanche size for given slab-depth variations. We present a sensitivity analysis of the skier triggering probability by varying the mean, variance and correlation length. The results of this analysis show some clear stable and unstable regimes for all skier tracks as well as more complex intermediate regimes. In the latter, large correlation lengths and small variances lead to a low probability of skier triggering as they reduce the size and the number of areas with low slab depth. Because the force induced by a skier at a given depth is inversely proportional to the depth, this leads to fewer areas to trigger the weak layer. Then, we show the effect of skier radius, spacing and skier group size on skier triggering probability. Increasing the skiing radius helps to reduce the skier triggering probability following an exponential relationship. Furthermore, we explore the conditional probability of skier triggering with respect to skier group size and preexisting track with no hits. For cases with non-negligible global skier triggering probability (one or two weak spots in the system, typically), results suggest that the probability of triggering dramatically increases for a group with more than five skiers. Finally, we present preliminary results regarding the size of avalanches and how it relates to the skier triggering probability.
Ryo Inoue, Teruo Aoki, Shuji Fujita, Shun Tsutaki, Hideaki Motoyama, Fumio Nakazawa, Kenji Kawamura
Corresponding author: Ryo Inoue
Corresponding author e-mail: inoue.ryo@nipr.ac.jp
The grain size of surface snow is a dominant factor of surface albedo for the Antarctic ice sheet, in which impurity concentration is extremely low. In addition, snow grain size affects the rates of deformation and compaction of firn such that the densification rate of coarse-grained layers is larger than that of fine-grained layers. Thus, it is important to observe the spatial and temporal distribution of the surface snow grain size and understand how depositional environment such as snow temperature, accumulation rate and wind controls the grain size in the near-surface snow. However, only a few field observations for wide-area distribution of snow grain size have been performed in Antarctica. For example, snow grain size on a pit wall was measured at 21 sites between Dumont D’Urville (near the coast) and Dome C (inland), and 36 sites between S16 (near the coast) and Dome Fuji (inland) using the IceCube (A2 photonics, France) instrument. In this study, we observed the surface snow grain size along two round-trip traverse routes between S16 and Dome Fuji from November 2021 to February 2022 during the 63rd Japanese Antarctic Research Expedition, to reveal the spatial distribution of snow grain size at higher measurement intervals. We used a newly developed Handheld Integrating Sphere Snow Grain Sizer (HISSGraS), which has the same measurement principle as the IceCube and has the advantage of rapid measurements and fewer calibrations. These improvements greatly shorten the time spent for the measurements at each observation site and allowed us to collect unprecedented amount of in-situ data from Antarctica. We performed measurements at a total of ~200 sites along the traverse routes, at 20–30 km intervals during each leg of two round trips. At each site, we measured 10 surfaces at intervals of ~2 m. Thus, a total of ~2000 snow surfaces were measured. We find that the surface snow grain size gradually decreases from S16 to Dome Fuji as air temperature and accumulation rate become lower. In addition, the grain size of glazed surfaces is larger than that of non-glazed surfaces. Our data provide crucial information for studying surface albedo and the initial stage of the densification process.
Florian Herla, Simon Horton, Pascal Haegeli
Corresponding author: Florian Herla
Corresponding author e-mail: fherla@sfu.ca
Numerical models that simulate the evolution of the snowpack can be an important source of information for avalanche forecasting when forecast regions are large and direct observations are sparse. In Canada, the public avalanche warning service forces the model SNOWPACK with output from a weather model to simulate the evolution of the snowpack on a 2.5 km grid across western Canada. For forecasters to have confidence in these simulations, it is critical to validate them for their practical value and assess their capabilities to meaningfully represent the conditions over large areas. While snowpack simulations have been validated in detail at point locations, regional-scale evaluations are more challenging due to the distributed nature of the simulations and the lack of direct observations at the relevant scale. Hence, regional-scale evaluations have so far primarily been restricted to bulk properties, such as snow depth or layer-independent stability indices, which are easier to monitor and validate. However, to be truly useful for avalanche forecasting, it is critical that snowpack simulations accurately resolve hazardous weak layers. To provide insight into the ability of the Canadian weather and snowpack model chain to accurately simulate hazardous weak layers, we investigate if and how individual layers of concern that were tracked by public avalanche forecasters are represented in the simulations. Our human assessment dataset includes all weak layers that were of operational concern in the 2011/12–2020/21 winter seasons in three data-rich forecast regions in British Columbia. To quantify the ability of the model chain to reliably identify layers of concern, we compute confusion matrices that show the odds of whether weak layers of concern are captured by the model and whether simulated weak layers actually were of concern. Furthermore, we explore how well the layers of concerns are captured by comparing summary statistics on the simulated stability, the spatial distribution and the timing of the first identification of the layer with forecaster assessments of the layers. Besides quantifying the added value of the simulations for operational avalanche forecasting in Canada, our study also aims to demonstrate how the implementation of operational validation suites can support forecasters’ real-time confidence in simulations in a meaningful way.
Antoine Bernard, Guillaume Chambon, Pascal Hagenmuller, Maurine Montagnat
Corresponding author: Antoine Bernard
Corresponding author e-mail: antoine.bernard@univ-grenoble-alpes.fr
Snow mechanical behavior is highly strain-rate dependent: ductile at low strain rates and brittle at high strain rates. The ductile-to-brittle transition has been shown to occur in two stages, with an intermediate regime of intermittent brittle failures, assumed to result from a competition between two time-dependent processes. In this work, we investigate the brittle-to-ductile transition in snow by conducting displacement-controlled compression tests and monitoring the evolution of the microstructure with X-ray micro-computed tomography (microCT). Samples (14 mm in diameter, 14 mm in height) were prepared from natural snow and sieved directly into sample holders. Uniaxial compression tests were performed at controled strain rates of 10–6 s–1 to 10–2 s–1 , up to a peak stress of 250 kPa and at a constant temperature of –18.5 °C. The 3-D microstructures were acquired with microCT prior to and after the tests but also during loading for the experiments at the lowest strain rate. For all tests, simple radiographs were acquired during loading at a rate of 5 frames per second. The obtained time series comprises one of the most-resolved (8.5 μm, 1 h) and complete dataset on snow microstructure evolution near the ductile-to-brittle transition to date. Results confirm the sensitivity of snow mechanical response to the strain rate. At strain rates smaller than about 10–4 s–1, stress increases smoothly with strain. At strain rates larger than about 10–3 s–1, snow samples display heterogeneous deformations with the formation of compaction bands, while the stress–strain curve shows a serrated behavior. In between, stress seems to evolve in a mixed way with an increase interrupted regularly by abrupt drops. To relate this macroscopic behavior to the microstructure evolution, quantitative investigation of the density and specific surface area changes, as well as of bond network evolution, were made thanks to microCT scans. This study therefore contributes to the identification of the micro-scale mechanisms at play during deformation of snow through both ductile and brittle ranges.
Deriving spatial snowfall deposition in mountainous terrain
Nora Helbig, Rebecca Mott, Yves Bühler, Perry Bartelt, Michael Lehning
Corresponding author: Nora Helbig
Corresponding author e-mail: norahelbig@gmail.com
During snowfall, the spatial snow depth distribution on the ground is determined by wind–precipitation–terrain interactions leading to preferential deposition. While preferential deposition can be modelled reasonably well, fine scale 2-D or 3-D wind fields are required and existing wind modelling approaches need careful numeric model pre-adjustments and have high computational costs. The required detailed wind fields are generally not available for large regions or over longer periods and broad model application is thus not feasible. Our goal was therefore to develop parameterizations to spatially downscale modelled or measured point-scale snowfall information to the topography using as little fine-scale meteorological input as possible. To do so, we used the physically based snow transport module of Alpine3D, driven with 3-D wind fields to produce a large ensemble of spatial new snow distributions on complex topography. The wind fields were computed using the non-hydrostatic, compressible atmospheric model ARPS (Advanced Regional Prediction System) on simulated mountains. By systematically investigating thousands of modelled spatial snowfall distributions, consisting of millions of data points of snow depth, wind components, and terrain slope and aspect, we then developed two statistical downscaling schemes: (1) a scheme using a combination of local surface vertical wind speed and local terrain slope; and (2) a scheme based on spatial mean horizontal surface wind speed, coarse surface wind direction and local terrain slope and aspect. Thus, highly resolved 3-D wind fields are not necessary any more. We evaluated our downscaling schemes using an independent spatial snow depth data set acquired after the first significant snowfall in October 2020 for a small Alpine catchment above Davos. The measured spatial patterns of preferential snowfall deposition using a photogrammetric drone survey were well reproduced by the Alpine3D preferential deposition module driven with ARPS wind fields as well as by the two downscaling schemes. The largest discrepancies were observed on the steepest slopes, which were probably due to avalanches that released during the snow storm. Overall, our results show that the proposed downscaling schemes can be used to reliably reproduce spatial snow depth patterns with a computationally very efficient method and could be used for various model applications such as hydrological, avalanche or weather forecasts.
In pursuit of open-access accurate global snow albedo observations
Karl Rittger, Timbo Stillinger, Ned Bair, McKenzie Skiles, Mark Raleigh, Keith Musselman, Mary J. Brodzik
Corresponding author: Karl Rittger
Corresponding author e-mail: karl.rittger@colorado.edu
Snow mapping was one of the first applications of satellite technology in hydrologic science, and global snow maps from MODIS were among the first group of standard products from NASA’s Earth Observing System, starting in 2000. Refinements since then include spectral mixing models to estimate fractional snow cover in pixels where vegetation and exposed soil are also present, and the estimate of the albedo of that snow based on its inferred grain size and the degree to which the albedo has been degraded by absorbing impurities such as dust or soot. All independent evaluations of these estimates from spectral mixing models using finer-resolution sensors and in-situ observations show their improvement over the standard EOS snow products, especially in mountainous and forest regions. The accuracy of spectral mixture algorithms across sensors such as MODIS, VIIRS, Landsat and Sentinel allow data fusion using machine-learning techniques to combine the strengths of each, specifically temporal and spatial resolution. We present observations of snow albedo from a suite of satellite sensors and from multiple models in pursuit of accurate global snow albedo. In addition, we provide a science-based forum for monitoring, discussing and contextualizing the state of snow conditions through Snow Today at the National Snow and Ice Data Center (http://nsidc.org/snowtoday), a website providing (1) near-real time updates of multivariate snow conditions, and (2) regular narratives that analyze the presented snow data from a climate and hydrologic perspective. The website provides a monthly insight into snow conditions and daily near-real-time data across the western USA, told with a combination of satellite data from MODIS and surface observations from the Natural Resources Conservation Services and California Department of Water Resources. Snow Today now provides the near-real-time gridded data of snow extent, snow persistence and snow albedo in easily usable GEOTIFF format. Growing support from the NASA research community and water resources managers aims to expand data distribution and analyses to all of North America and Greenland for the coming snow season.
Rand Decker, Thomas Whelan
Corresponding author: Rand Decker
Corresponding author e-mail: rand.decker@nau.edu
A Lagrangian–Eulerian mixture theory is used to model the downrange transport or ‘throw’, and ultimately the fate or ground snow accumulation distribution for a variety of manmade (technical) snow particle sizes being shot from a typical modern snowmaking ‘fan gun’. Ground snow accumulation, as a function of position downrange from the snow gun, is sensitive to both the snow particle size distribution and the power of the snow gun fan being used. The resulting modeled accumulation of ground snow ‘whales’ is compared to available measured data for annual and limited duration snowmaking sessions. The modeled snow distribution whales compare well in extent and morphology, including peak snow depths and mass center, with those that are measured. Both the measured and model results compare well with manufacturers’ declared snow accumulation ‘throw’ or downrange extent from their snowmaking fan guns. The sensitivity of the snow accumulation whales to both terrain sloping away from the snowmaking gun and tailwinds approaching the gun from behind is modeled. As expected, both a downslope and tailwind, and especially in combination, can extend the furthest accumulation point of the smallest size fractions of snow particles in the snow whale. A slope of 0.20, falling away from the snow gun, will throw the smallest snow particles 22% further downrange than over flat ground. A 12.5 m s–1 tailwind will have a similar effect, extending the furthest downrange (and downwind) extent of the snow accumulation by 20% over that of a calm day. The two combined, a 20% downslope and a 12.5 m s–1 tailwind, will double the furthest extent of the snow whale over the flat ground/calm day scenario. These model results for snow accumulation extent are for spherical ice particles (manmade/technical snow) ranging in size from 30 μm to 400 μm, distributed about a Volume Mean Diameter (VMD) of 200 μm. Despite being a small fraction of the total snow mass and soon doomed to loss if they’re still in the air, it is the smallest-sized snow particles that cause the greatest increase in the modeled downrange extent of the snow whales. The larger snow-particle sizes, at and around the 200 μm VMD, account for most of the snowmaking mass. As expected, their throw and accumulation is less impacted by downslope and tailwind, hence the relatively smaller impact downslopes and tailwinds have on the location of the snow whale mass center.
Cristina Pérez-Guillén, Frank Techel, Martin Hendrick, Michele Volpi, Alec van Herwijnen, Tasko Olevski, Guillaume Obozinski, Fernando Pérez-Cruz, Jürg Schweizer
Corresponding author: Cristina Pérez-Guillén
Corresponding author e-mail: cristina.perez@slf.ch
Avalanche forecasting plays a crucial role in ensuring safety and mobility in the Alps. The Swiss avalanche warning service twice daily issues a bulletin to inform the public and local authorities about snow and avalanche conditions. In the bulletins, the danger is described according to the five-level European avalanche danger scale. So far, the assessment of the danger has been entirely a human expert decision-making process. Avalanche forecasters analyse heterogeneous data sources and dray conclusions about the avalanche scenario based on expert judgment. Today, with the increase in computing power and data volumes, there are new opportunities to develop automated decision support tools for avalanche forecasting. We evaluate the performance of a fully data-driven approach based on a Random Forest (RF) classifier to assess the regional avalanche danger level for dry-snow conditions in Switzerland. We rely on over 20 years of meteorological data, snow cover simulations, and danger level estimates to train our RF algorithm. During the last two winter seasons, the RF model was tested in an operational setting providing a ‘nowcast’ and a ‘24-hour forecast’ prediction in real-time. The nowcast predictions were computed with input data from SNOWPACK simulations and weather data provided by a network of automated weather stations (AWS) in the Swiss Alps. In forecast mode, the snow cover simulations were driven with the numerical weather prediction model COSMO1 operated by MeteoSwiss and downscaled to the AWS locations. In addition, we tested individual danger level predictions for virtual slopes of the main four aspects. Preliminary results showed that the overall predictive performance of the regional danger level in forecast and nowcast mode and per aspect was similar, ranging between 70% and 75%, which is comparable to the accuracy of Swiss avalanche forecasts. Furthermore, the RF classifier predicted not only the most likely danger level but also the class probabilities for each danger level. Finally, we evaluated the performance of the model for different avalanche situations. This is the first time that a data-driven decision support tool has been used in real time to provide a ‘second opinion’ about avalanche danger for operational avalanche forecasting. Overall, avalanche forecasters provided positive feedback on its performance and usefulness.
Daniela Brito Melo, Armin Sigmund, Varun Sharma, Michael Lehning
Corresponding author: Daniela Brito Melo
Corresponding author e-mail: daniela.britomelo@epfl.ch
Drifting snow events lead to snow redistribution and enhance snow sublimation. Taking into account the importance of snow sublimation to the correct prediction of the surface mass balance, a general consensus has emerged within the scientific community on the importance of correctly representing the aeolian transport of snow in large scale models. Nevertheless, few attention has been given to the description of snow saltation – the motion of snow particles in approximately the first 10 cm above the ground. The saltation layer feeds the upper regions of the atmosphere with suspended particles and, therefore, considerably influences the predictions of drifting snow concentration. In order to improve our understanding of snow saltation dynamics and the quality of the parameterizations used in large scale models, numerical simulations combining Large Eddy Simulations with a full description of the particle–fluid–bed interactions were performed. In particular, we focused on the exponential decay of the particle mass flux vertical profile in the saltation layer – a commonly observed feature in both wind tunnel and field experiments and a widely used assumption in saltation models. The inverse of the exponential decay constant is expected to be proportional to the height of the saltation layer, which is generally assumed to increase with the square of the fluid friction velocity, following wind tunnel experiments developed with snow. However, recent field measurements developed over sand-covered surfaces revealed that the saltation layer height is invariant with the fluid friction velocity. From numerical simulations of snow saltation, we found that these conclusions are not contradictory. In fact, they reveal two distinct dynamic regimes of the saltation system: while, at low wind velocities, typical of wind tunnel experiments, a quadratic increase of the saltation layer height with the fluid friction velocity is seen, at high wind velocities, frequently found in the field, the saltation layer height does not vary with the fluid friction velocity. Following these findings, adjustments to the exponential decay constant can be proposed. In addition, this work shows the potential of detailed models to improve our understanding of the saltation system and foster the development of more physically based snow saltation parameterizations.
Amy R. Macfarlane, Ruzica Dadic, Madison M. Smith, Bonnie Light, Marcel Nicolaus, Hannula Henna-Reetta, Melinda Webster, Felix Linhardt, Stefan Hämmerle, Martin Schneebeli
Corresponding author: Amy R. Macfarlane
Corresponding author e-mail: amy.macfarlane@slf.ch
The microstructure of the uppermost portions of a melting Arctic sea ice cover largely determines how incident sunlight is reflected and absorbed in the ice/ocean system. The surface scattering layer (SSL) effectively backscatters solar radiation and keeps the surface albedo of melting ice relatively high. Measurements of albedo, light transmissivity and extinction provide information on how incoming shortwave radiation is partitioned by the SSL and have been pivotal to improving model parameterizations. However, the relationship between the physical and optical properties of the SSL is still poorly constrained. Until now, radiative transfer models (RTMs) have been the only way to infer the microstructure of the SSL. During the MOSAiC expedition, for the first time, we directly measured the microstructure of the SSL on bare sea ice using a micro-computed tomograph (MicroCT). We show that the SSL has a highly anisotropic, coarse and porous structure, with a small optical diameter and density at the surface that increases with depth. It regenerates during melt and maintains its microstructure throughout the season. This is in contrast to the grain coarsening due to metamorphism typically implemented in RTMs for snow. We use the microstructure measurements as inputs into the two-stream radiative transfer in snow (TARTES) to improve our understanding of the relation of optical properties with the physical properties of the SSL at 850 nm. Direct comparisons between the RTM and in-situ measurements show that spatial variability of the ice surface has a large role in the measurements. The micrometre variability of the geometric properties is not apparent in the larger footprint of the in situ optical measurements, and when this microstructure is used directly in the RTM we see an overestimation of the reflectance at 850 nm.
Julien Brondex, Kevin Fourteau, Marie Dumont, Pascal Hagenmuller, Neige Calonne, Francois Tuzet, Henning Löwe
Corresponding author: Julien Brondex
Corresponding author e-mail: julien.brondex@meteo.fr
Faithful, dynamical modeling of the physical properties of snow is a prerequisite for a wide range of applications impacting human activities, such as avalanche hazard forecasting, water resources management or projections of future climate evolution. To address these needs, several one-dimensional snowpack models with various levels of complexity have been developed over the years. Yet, even the most detailed models still suffer from numerous limitations. Notably, some important processes with commonly accepted mathematical representation, such as water vapor diffusion, liquid water percolation, and melting/refreezing, remain inadequately, if at all, implemented. This limitation is due to the numerical difficulties arising as the resulting mathematical model turns into a system of highly non-linear and highly coupled partial differential equations. The choice of the numerical scheme and the representation of couplings between processes is crucial to ensure an accurate and robust solution while allowing time steps in the order of 15 minutes or more. To freely explore the impact of the specificities of numerical schemes (time discretization approach, linearization strategy, etc.), we have developed a highly modular finite-element program in which every step of the numerical formulation and resolution is coded internally. Specifically, only the actual resolution of the linearized system of equations is performed through an external linear algebra library. In order to illustrate the capabilities of this preliminary version of the code, we consider the problem of the coupled heat conduction, vapor diffusion and settlement within a dry snowpack. We run the model on several test cases proposed in recently published literature. We take particular care to verify that mass and energy are conserved. We show that a fully coupled and fully implicit time stepping approach enables the acquisition of solutions consistent with previously published ones with little restriction on the time-step size. The present solver combines advantages from different recent numerical studies and may serve as a support to the development of the next generation of detailed snowpack models, associating sound and universal physics to a robust numerical treatment.
Snow instability in a changing climate
Stephanie Mayer, Alec van Herwijnen, Jürg Schweizer
Corresponding author: Stephanie Mayer
Corresponding author e-mail: stephanie.mayer@slf.ch
Several studies have shown that snow depths in the Alps will decline in the future. Yet the influence of climate change on avalanches remains less clear. The formation of avalanches is determined by snow instability in avalanche starting zones, resulting from a complex interplay between local weather and terrain. For snow instability, the layering of the snowpack is more decisive than the amount of snow on the ground. Modelling future snow instability thus requires climate scenarios with high temporal and spatial resolution to resolve the processes leading to weak layers within the snowpack. To investigate changes in snow instability due to climate change, we therefore used data from downscaled climate projections as input for the numerical snow-cover model SNOWPACK. This allowed us to simulate snow stratigraphy at the Weissfluhjoch (2540 m a.s.l.) field site above Davos, Switzerland, for complete winter seasons between 1980 and 2100. We then applied a random forest classifier trained on past snow-instability data to detect the weakest layers in each stratigraphic profile, assess the degree of instability on a daily basis, and indicate potential avalanche size using the depth of the weak layer. After validating this model chain for the reference period, we analyzed changes in snow instability by comparing past and future periods. Our analysis showed a significant decrease in days with poor dry-snow instability by the end of the century, with the highest emission scenario leading to a decrease of about 50%. This decrease was mostly linked to warmer temperatures resulting in a snowpack with an increased prevalence of melt forms and fewer faceted weak layers. With the analysis of future snow instability based on simulated snow stratigraphy, our study represents a valuable step towards a better understanding of the influence of climate change on snow avalanches.
Grégoire Bobillier, Provence Mahjoub Bonnaire, Bastian Bergfeld, Johan Gaume, Alec van Herwijnen, Jürg Schweizer
Corresponding author: Grégoire Bobillier
Corresponding author e-mail: gregoire.bobillier@slf.ch
Most avalanche fatalities are caused by dry-snow slab avalanches. Their release is a multi-scale process, starting with the formation of a localized failure in a highly porous weak snow layer underlying a cohesive snow slab. The process ends with a tensile fracture through the slab leading to its detachment. The dynamics of crack propagation across a snow slope, and in particular the interaction between crack propagation in the weak layer and slab fractures, are believed to influence avalanche size, yet remain poorly understood. Here, we use the discrete element method (DEM) to numerically simulate crack propagation in weak snowpack layers in 3-D and analyze the influence of slab fractures using a micro-mechanical approach. Our DEM simulations show that there is a competing effect between crack propagation in the weak layer and slab fracture. In particular, it appears that the crack tips in the slab and the weak layer are attracted to each other and tend to merge. This resulted in different simulated slab-fracture morphologies that were in line with observations in field experiments. Our analysis further shows that the tensile strength of the slab will in large part determine whether crack propagation in the weak layer will be arrested or not. Furthermore, on sloped terrain slab fractures can limit weak-layer crack propagation speed. For low slab-strength values, slow en-echelon fractures appear. In contrast, hard slabs promote the transition into a fast supershear propagation regime. Overall, our results lay the foundation for a comprehensive study on the effect of the snowpack mechanical properties on the fundamental processes for avalanche release.
How entrainment-induced dispersive pressures enhance avalanche mobility
Perry Bartelt
Corresponding author: Perry Bartelt
Corresponding author e-mail: bartelt@slf.ch
We investigate how mass entrainment enhances avalanche mobility. Entrainment is considered the plastic part of the mechanical interaction between the avalanche and snowcover. As such, mechanical energy is dissipated and mass entrainment must slow the avalanche down, especially at the leading edge. If, however, the process of entrainment changes the flow configuration of the avalanche core, a significant reduction in frictional resistance is possible. In this paper we relate slope-perpendicular dispersive pressures, produced during snow-cover entrainment, to changes in particle configuration and therefore changes in flow resistance. We show how entrainment can change the natural frequency of dynamic flow pulsations associated with the leading edge of the avalanche. We apply the theory to model recent avalanche case studies to quantify the entrainment parameters that govern the mechanical energy fluxes of the collisional/erosional interaction with the snowcover. The results indicate that snowcover entrainment can produce large configurational changes in the avalanche core, fundamentally changing the structure, dynamics and mobility of the avalanche.
Todd Redpath, Pascal Sirguey, Nicolas Cullen, Aubrey Miller, Christian Zammit
Corresponding author: Todd Redpath
Corresponding author e-mail: todd.redpath@otago.ac.nz
Situated in the Southern Hemisphere mid-latitudes, the Southern Alps/Kā Tiritiri o te Moana are characterized by high relief, a maritime climate featuring strong and persistent westerly winds, and a notable west–east precipitation gradient. Freshwater runoff from the Southern Alps, to which snowmelt provides a substantial contribution, is important ecologically, culturally and economically. The New Zealand Water Model (NZWaM) includes a semi-distributed snow model, within which snow water equivalent (SWE) is distributed hypsometrically across 100 m elevation bands. Within each elevation band, spatial variability in the SWE distribution is parameterized statistically. While NZWaM is validated against discharge within gauged catchments, there is uncertainty associated with the snow model due to a historic lack of observations. This is addressed by implementing a satellite photogrammetric mapping (SPM) workflow using a series of Pléiades tri-stereo images captured during the 2012 austral winter. For the first time, the resulting maps resolve the snow depth distribution across the main divide of the Southern Alps at high spatial resolution. Focussing on the gauged Jollie Catchment, which has been the site of historic snow-measurement campaigns, these maps demonstrate the extent and nature of re-distribution by wind and avalanching, with maximal snow depths occurring at mid-elevations. The non-linear relationship between snow depth and elevation departs from previous knowledge established from field measurements. Applying snow density estimates allows for the first spatially widespread assessment of SWE estimates produced by NZWaM. Furthermore, the 2 m resolution and spatially continuous nature of the snow-depth maps enables the spatial variability of snow depth to be characterized in unprecedented detail. These results underpin the assessment and refinement of the statistical parameterization of spatial variability in SWE distribution within NZWaM. Ultimately, this work demonstrates the step-change offered by SPM for collecting observations to support efforts to reduce hydrological model uncertainty by improving the representation of seasonal snow.
A historical reanalysis of SWE time-series: towards a near real-time reconstruction
Valentina Premier, Carlo Marin, Giacomo Bertoldi, Riccardo Barella, Claudia Notarnicola, Lorenzo Bruzzone
Corresponding author: Valentina Premier
Corresponding author e-mail: valentina.premier@eurac.edu
Snow water equivalent (SWE) is a key variable for several applications (e.g. hydrology, ecology, agriculture, or hydropower production). The typical high spatial variability of SWE linked to snow-distribution processes requires its monitoring at a proper spatial and temporal scale. However, spatialized high-resolution (HR) time series (TS) of SWE maps are rarely available. We explore the use of multisource remote-sensing data to derive information about snow-depletion curves (SDC), i.e. the relationship with the snow cover area (SCA) and SWE. We propose a fusion approach that merges different sensors, if available, as optical HR and low-resolution (LR) data, synthetic aperture radar (SAR), and in-situ temperature and snow depth observations to reconstruct SWE TS. The method does not require precipitation data as input, which could be a relevant advantage in poorly monitored mountain regions. The approach allows historical reconstruction obtaining spatial reanalyses SWE TS, with daily sampling time and HR spatial detail (i.e. a few dozens metres). The approach, designed for a hydrological catchment, assumes that similar snow patterns repeat interannually due to the climate and geomorphology of the area. Moreover, the catchment is characterized by a state: it is either subjected to an accumulation (SWE increase) or ablation (SWE decrease). The state is identified by mean of in-situ snow depth observations (if available) and/or SAR data, and is used to regularize the SCA as well as to decide whether SWE is added to or removed from the reconstruction. The potential melting is calculated through a degree day model. The output SWE TS present an unprecedented spatial and temporal detail. The proposed approach has been tested in reanalysis in few mountainous catchments with different climates and when evaluated against a reference product (i.e. Airborne Snow Observatory), shows a bias of –40 mm and a RMSE of 216 mm for a catchment of 970 km2 in the Sierra Nevada (California, USA). The power of the method is represented by a good precision for capture of snow persistence on the ground, which in turn is linked to the SWE amount of the pixel. This is achieved using HR spatialized data as input, which allow proper sampling of the phenomenon at the correct spatial scale resulting in a good detection of the typical SWE spatial patterns. The availability of a long TS of accurate spatial reanalysis products opens the door to possible near-real-time predictions that exploit SDC.
Riccardo Barella, Carlo Marin, Claudia Notarnicola
Corresponding author: Riccardo Barella
Corresponding author e-mail: riccardo.barella@eurac.edu
Mapping snow-cover fraction (SCF) from optical remote sensing is a widely studied topic that still presents some challenges, especially in mountainous areas when using high-resolution time series. The generation of reliable maps under all acquisition conditions is limited by topographic effects such as shadows, sunglint on snow and atmospheric disturbances. If these effects are not corrected or taken into account even the most advanced methods fail. The errors usually present in these maps are a limitation to the full exploitation of these SCF for model assimilation or retrieval of snow parameters, e.g. snow water equivalent (SWE) reconstruction. In this work, we propose a new unmixing method for SCF retrieval based on the Vapnik–Chervonenkis (VC) theory. The general idea is that is easier to find a different model that locally adapts to the specific problem than a unique model able to generalize the problem in all situations. The proposed method starts from two sets of endmembers of the pure class ‘snow’ and ‘snowless’ that are identified in an unsupervised manner considering the spectral signature adapted to the scene. A further endmember filtering is performed by considering the cosine similarity among the trainings. Once the endmembers have been identified, the method finds the best separation hyperplane by maximizing the distance between the two classes. When an unknown pixel is considered, it is mapped into that same space and predicted to belong to a category based on which side of the gap it falls. If it falls in the tube defined by the endmembers, then it is a mixed pixel with an abundance that is proportional to the SCF. To handle the non-linearity of the problem, an RBF (radial basis function) kernel was considered instead of simply using the scalar product between trainings. The RBF kernel relies on a parameter, γ, which is usually selected performing an optimization based on the classification accuracy. In our specific case the classification accuracy is not sensitive enough to identify the best γ for the SCF retrieval. For that reason, we propose to find the best γ as the one that maximize the kernel interclass standard deviation of the class ‘snow’. The method was validated with the three WV3 images and acquired in several sites in the Alps, showing promising results both in terms of robustness under challenging situations and in terms of SCF sensitivity, reaching for the test sites an RMSE lower than 20% and bias close to 0.
Henning Löwe, Ghislain Picard, Florent Domine, Laurent Arnaud, Fanny Larue, Vincent Favier, Emmanuel Le Meur, Eric Lefebvre, Alain Royer
Corresponding author: Henning Löwe
Corresponding author e-mail: loewe@slf.ch
Remote sensing is one of the few methods that has the potential to characterize the global state of terrestrial snow on climatologically relevant time scales. A key prerequisite is the understanding of how the snow microstructure controls scattering, in order to inform the development of retrievals by in-situ measurements and radiative transfer models. For optical wavelenghts, this problem is conceptually and satisfactorily solved by using an optical grain size that is defined through the inverse specific surface area of snow. An equivalent, stringent, and predictive concept for a microwave grain size has hitherto been missing. As a remedy, we propose a definition of the microwave grain size that is purely geometrical (depends only on the microstructure) but naturally emerges from the Born approximation in the low-frequency limit. It follows that this quantity is optimal to calculate the scattering coefficient. Our non-parametric definition allows us to express the microwave grain size in terms of the optical grain size (accessible through in-situ measurements) and a novel polydispersity parameter of snow. To demonstrate this concept, we retrieve the polydispersity from microwave satellite observations from 18 Antarctic and 86 Canadian sites via objective optimization using the model SMRT and in-situ measurements of the optical grain size. We validate these estimates through independent polydispersity calculations from the chord length distribution of 3-D micro-computed tomography images from lab and Alpine snow. For convex grains, the polydispersity values (0.6–0.8) compare favorably between the two calculations, while depth hoar shows higher variations (1.2–1.9), and the polydispersity retrieved from satellite observations in Canada exceeds those of available depth hoar tomography samples. Nonetheless, using only a single value for each grain type allows to predict the microwave observations in Antarctica and Canada from optical in-situ measurements with good accuracy within a fully physical model. In addition, the novel concept also explains the different values of empirical factors that were previously used to scale measured optical diameters to the size parameters of common autocorrelation models (exponential, sticky hard spheres) in the microwave scattering coefficient. Our findings unify important aspects of microwave scattering towards a more accurate use of satellite observations in hydrological, meteorological and climatological applications.
Callum Tregaskis, Peter Gauer, Dieter Issler
Corresponding author: Callum Tregaskis
Corresponding author e-mail: callum.tregaskis@ngi.no
Recent advances in the continuum modelling of granular materials have led to significant development of rheological models, such as the regularized μ(I) rheology, where μ is the internal friction and I is the inertial number. Regularization has extended the applicable range of the rheology into regions of rapid highly collisional flow and also at the limit between the inertial and quasi-static regimes, where the original formulation of the μ(I) rheology is ill posed. With this adaption such rheologies may now be applicable to dry fluidized snow avalanches which have regions of rapid collisional flow alongside a dense core. Theories for particle segregation, where the material species that make up the granular mixture separate during the flow, have also been advanced with work coupling advanced rheologies to segregation theory. This paper details theoretical and numerical frameworks that utilize these models to describe a fluidization process in dry-snow avalanches driven by escaping air from the snow pack. Including a fluidization process is important in order to describe the density variation observed in powder snow avalanches, which affects the velocities of the differing flow regimes and thus the avalanche path and runout of the flow. A depth-averaged approach to modelling is taken in order to maintain a feasible processing time for computations while allowing the model to be applied at geophysical scales. The advantage of such models is that the evolving concentration of the bulk flow can be captured, so it is possible to separate each phase of material (from dense to fluidized) and allow each phase to have disparate frictional properties. Further, inclusion of an excess-air material phase, which segregates up through the avalanche, allows the complex process of fluidization alongside the evolution of the free surface to be modelled in a depth-averaged framework. By including additional considerations for varying topography and erosion, entrainment and deposition of material, the model is of direct relevance for hazard zoning of snow avalanches in mountainous areas.
Influence of environmental factors on snow gliding prior to and after a first snowfall event
Zorin Ivanov, Erich Tasser, Reinhard Fromm, Francesco Comiti, Georg Leitinger, Peter Höller
Corresponding author: Zorin Ivanov
Corresponding author e-mail: zivanov1993@gmail.com
Abstract Despite the fact that snow gliding and glide avalanches have been thoroughly studied, they remain fairly unpredictable by the current methodology. There is a growing body of evidence that the environmental conditions in late autumn affect gliding activity in the following winter. In this study we analyze the effect of 67 independent variables, including variables that characterize the autumn conditions, on snow gliding by using ordinary least square regression. The analysis shows that the cumulative soil temperature (at 10 cm depth) summed for 3 weeks before a permanent snow-cover formation, the soil temperature at 10 cm depth, the soil water content at 0 cm depth, the cumulative air temperature summed for 1 month before a permanent snow cover formation and the daily minimum snow liquid-water content (LWC) significantly influence snow gliding. These results confirm our hypothesis that conditions preceding the formation of a permanent snow cover affect snow gliding. Furthermore, our findings suggest that stored heat in the soil is responsible for inducing water phase changes which are a source of liquid water at the snow–soil interface and promote gliding. Based on this and previous studies, we conclude that it is beneficial for further investigations to gather snow-gliding data for multiple seasons and differentiate between gliding and non-gliding events, as well as between cold- and warm-temperature gliding events.
Evolution and spatial variability of small-scale surface roughness of snow and sea ice during MOSAiC
Ruzica Dadic, Martin Schneebeli, Henna-Reetta Hannula, Roberta Pirazzini, Amy Macfarlane, Oliver Wigmore, Huw Horgan, Henning Löwe, Lauren Vargo, David Wagner, Stefanie Arndt, Matthias Jaggi, Daniela Krampe, Julia Regnery, David Brus, Gunnar Spreen, Julienne Stroeve
Corresponding author: Ruzica Dadic
Corresponding author e-mail: ruzica.dadic@vuw.ac.nz
Small-scale surface roughness (mm to cm scale) affects remote sensing retrieval algorithms and radiative transfer models, yet data on snow and ice surface roughness over sea ice are scarce. Radar backscatter and microwave emissivity, important for satellite retrievals of sea-ice concentration, type, thickness, etc., is a function of ice–snow and snow–air interface roughness (among other factors). During MOSAiC, we collected data to determine the small-scale surface roughness of the surface (snow and surface scattering layer) and the snow–sea-ice interface using a low-cost photogrammetric method. We collected surface roughness data with most snow pits and on some transects to account for temporal and spatial variability. Here, we present the surface roughness evolution for the year-long snow-pit observations and discuss the applicability and relevance of this dataset for remote sensing and radiative transfer studies. We will consider the spatial variability and anisotropy of surface roughness for different snow and surface types. We will also discuss other potential applications for this dataset, such as correlations between the small-scale surface roughness and specific surface area (from near-infrared photography, SnowMicroPen, and microCT) or shear strength (from SnowMicroPen).
Evolution of the snow cover during the sea-ice freeze-up phase on the MOSAiC expedition
Ruzica Dadic, Martin Schneebeli, Amy Macfarlane, Henna-Reetta Hannula, Roberta Pirazzini, Melinda Webster, Michael Gallagher, Marcel Nicolaus, Mario Hoppmann, Julia Regnery, Ola Persson, Henning Löwe, Huw Horgan, Lucille Gimenes, Matthias Jaggi, David Wagner, Linda Thielke
Corresponding author: Ruzica Dadic
Corresponding author e-mail: ruzica.dadic@vuw.ac.nz
Snow cover dominates the variability of the thermal and optical properties of sea ice and is essential to understanding sea-ice growth and decay. It governs the energy and mass fluxes between the ocean and the atmosphere, sea-ice thickness, bottom-water formation, and ocean circulation. During the freeze-up phase of sea ice in autumn, the thickness of the snow cover and its physical properties are key to understanding the sea-ice growth rate. Snow thermally insulates the sea ice, inhibiting its growth, or in case of warming events, it can prevent the intermittent warming of the sea ice. We observed the physical properties of snow during the transition from summer melt to autumn freeze-up (August–October 2020) of the MOSAiC (Multidisciplinary Drifting Observatory of the Study of Arctic Climate) expedition. Here, we will present the evolution of the snow cover during that transition using a suite of unprecedented state-of-the-art measurements (μm–cm scale) of physical snow properties, including snow microstructurespecific surface area, density, thermal structure, water percolation observations, and surface roughness. Our measurements show a highly spatially and temporally variable snow cover affected by a) the different underlying ice surfaces (existing ice, frozen melt ponds, and new ice), b) highly irregular topography, and c) typical temporally varying autumn atmospheric conditions (including positive and negative net energy fluxes; snowfall; and rain on snow after the freeze onset). This complex snow cover forms the base of the winter snow cover over sea ice, influencing sea-ice evolution.
Opportunities and challenges for hectometric scale simulations of Alpine snow cover
Ange Haddjeri, Matthieu Baron, Louis Le Toumelin, Rafife Nheili, César Deschamps-Berger, Hans Lievens, Simon Gascoin, Marie Dumont, Matthieu Lafaysse
Corresponding author: Ange Haddjeri
Corresponding author e-mail: ange.haddjeri@meteo.fr
Accurate snowpack predictions are decisive for anticipating natural hazards related to snow and improving hydrological forecasts. In mountainous terrain, resolving the high and complex snow spatial variability is hardly feasible without a hectometric or higher-resolution simulation. At this scale, there is the need to account for new simulated processes such as wind-blown snow transport. SnowPappus is a new hectometric wind-blown snow transport model coupled to the state-of-the-art multilayer Crocus snowpack model. In this study snow-cover simulations are computed with this new modeling system using different atmospheric forcing and different simulation resolutions. These simulations show changes in snow cover caused by the snow-transport module mainly localized around mountain crests. Evaluations are made by comparing results to on-site measurement stations and satellite imagery: Sentinel 2 snow cover Presence (SCP), C-Snow snow height product derived from Sentinel 1 backscattered radar imagery and Pléiades stereo imagery snow height retrieval. The first evaluation of simulated snow depth reveals the recurrent dominance of precipitation uncertainty over simulation errors at the hectometric scale. Thus these evaluations underline the need for Alpine hectometric snow-forecasting models of assimilating snow-related observations.
Ruzica Dadic, Lauren Vargo, Nora Helbig, Rebecca Mott, Brian Anderson
Corresponding author: Ruzica Dadic
Corresponding author e-mail: ruzica.dadic@vuw.ac.nz
Snow accumulation is a crucial part of the water balance – on both seasonal time scales and annual to decadal time scales, in the case of glaciers. Previous work shows that some catchments have a very high contribution of snowmelt to stream flow, but also that snow accumulation variability is critical for stream-flow simulations, as deep, late-lying snow significantly influences the hydrograph in late summer. Snow melt is reasonably well understood, but significant deficiencies in modelling snow accumulation exist. The high spatial variability in snow accumulation results from two processes: preferential deposition and redistribution of snow by wind, and avalanches. These processes interact, and it is ultimately the wind-driven distribution of snow accumulation that may influence avalanching. To improve snow-accumulation models, at least 2-D wind speed and direction data are needed. Until recently, it was not possible to derive wind-speed data in steep mountain terrain without high-resolution atmospheric modelling – and we present a new empirical wind downscaling scheme combined with an existing parameterization of preferential snow deposition. We show snow distribution’s effects on glacier mass balance and runoff for Haut Glacier d’Arolla in Switzerland and Brewster Glacier in New Zealand. We also discuss the relevant topography scales for modeling wind-driven snow distribution for glacier mass balance.
Operational snow water resources monitoring at SLF
Jan Magnusson, Rebecca Mott, Bertrand Cluzet, Louis Queno, Tobias Jonas, Nora Helbig, Adam Winstral, Guilia Mazzotti
Corresponding author: Jan Magnusson
Corresponding author e-mail: jan.magnusson@slf.ch
The SLF’s operational snow-hydrological service (OSHD) provides information crucial for flood forecasting and reservoir management in Switzerland. This service has now been operating for over 10 years and products have been constantly improved and further developed. For generating these products, the OSHD has developed a modular framework to conduct ongoing analyses of snow water resources and snow climatology. Furthermore, the OSHD analyzes spatio-temporal dynamics of new snow depths for snowfall events in the fore- and hindcast which is used for avalanche warning and weather predictions. The modular framework consists of models with different complexity, which are dedicated to different modelling strategies and tasks. The physics-based OSHD model chain is based on a snow model (FSM2: flexible snow model) solving the complete mass and energy balance of the snowpack for open, forested and glacierized areas at a horizontal resolution of 250 m and an hourly temporal resolution. Results from this model chain is useful for predicting, for example, rain-on-snow-related floods. Furthermore, conceptual models were developed within the OSHD model chain for climatological runs to compare snow water resources in Switzerland of water years since 1998 at a horizontal resolution of 1000 m. Such comparisons are of high relevance for reservoir management throughout Switzerland. We will present strategies for data assimilation, downscaling techniques and sub-grid parametrizations of relevant processes in order to provide an optimal balance between process representation and computational requirements in an operational snow hydrological model setting. While the model framework is operationally used to forecast snow cover dynamics in Switzerland, model components are also applied for many different research applications, as will be shown.
Mapping the spatio-temporal evolution of wet snow from space
Gwendolyn Dasser, Valentin T. Bickel, Andrea Manconi, Mylène Jacquemart, Yves Bühler, Alec van Herwijnen, Martin Hendrick
Corresponding author: Gwendolyn Dasser
Corresponding author e-mail: 8c3pvb.ge@dasser.ch
During spring and midwinter warm spells, wet-snow avalanches pose a threat to life and infrastructure in Alpine environments. Risk assessment and early warning efforts require continuous and high-resolution monitoring of wet snow on large spatial scales and in complex terrain. Spaceborne synthetic aperture radar (SAR) can be used to provide this information by exploiting temporal changes in the backscattered radar signal, but publicly available products currently have several drawbacks: They lack continuous, weather-independent coverage, are limited to a resolution of 60 m, are prone to geometric artifacts, depend on multi-platform data fusion (optical and radar data), and/or have not been thoroughly validated. Here, we attempt to address the mentioned limitations by optimizing existing Sentinel-1 (S1) data processing workflows, deploying them on regional-scale, multi-year SAR data using local resolution weighting (LRW) composites, and validating them with weather station data. We analyze two S1 LRW composite time series (August 2018–August 2021) over the areas of Davos (900 km2) and Interlaken (3042 km2), Switzerland. We compare LRW backscatter to a range of snow parameters measured by in-situ stations and/or modeled using SNOWPACK. Radar signal variations show a negative cross-correlation with snowpack water content (lower backscatter – higher water content), and a positive correlation with snow height (lower backscatter – lower snow height). We analyze the dynamics of snowmelt by calculating the wet snow ratio (total number of pixels/wet pixels in the area of interest) as a function of time and terrain (slope angle and aspect). In the period from August 2018 to August 2019, extensive snowmelt (wet snow ratio >0.75) begins in late April at altitudes of ~2200 m a.s.l. (above sea level), proceeding to ~2200 to ~2700 m a.s.l. in May, eventually reaching >2700 m a.s.l. in early June. The analysis will be extended until summer 2022 to identify potential interannual variations of this behavior. We will continue to study the spatio-temporal relation between wet-snow avalanches, wet-snow distribution, snowpack water content and radar backscatter. In addition, we will assess the capabilities of our approach to detect and characterize midwinter warm spells. Associated products will include, for example, regional-scale, temporospatial diagrams and maps of wet-snow distribution that can be used to enhance current wet-snow avalanche prediction capabilities.
SMP5: electronics upgrade of the SnowMicroPen®
Matthias Jaggi, Michael Hohl, Henning Löwe, Rafael Ottersberg, Benjamin Walter, Martin Schneebeli
Corresponding author: Matthias Jaggi
Corresponding author e-mail: jaggi@slf.ch
The SnowMicroPen® (SMP) is a snow penetrometer that was developed and introduced more than 20 years ago. The high-resolution, penetration force-distance signal can be used to recover microstructural parameters, which allows for the derivation of stratigraphic and macroscopic snow properties such as snow density and specific surface area. In comparison to a traditional snow profile, the time requirement to perform a measurement is short and the measurement is not biased by the observer. The instrument consists of two main pieces: motor and controller. The motor unit drives a high-sensitivity force sensor attached to a metal rod through the snowpack. The motor unit is controlled by the controller. Because several electrical components of the controller have recently reached their end-of-life, a major upgrade of the SMP was needed. Upgrades were made to the signal processing, power supply, usable temperature range, and safety measures, with the addition of diagnostic tools and modularity for maintenance on expeditions. To prove consistency between the old and new SMP versions (SMP4 to SMP5), laboratory and field experiments were performed. The tests compared the derived quantities of snow density and specific surface area. For the laboratory tests, the force sensor was replaced with a signal generator and an analogue signal was used to simulate a real snow profile. For the field tests, fully functioning SMP units were tested on the SLF’s flat-field measurement site on Weissfluhjoch. Results show that there are variations between instruments, but not related to the aforementioned upgrades. Therefore, we recommend upgrading existing SMP4 units.
Is a Snow Monitoring Competence Centre still needed?
Christoph Marty, Nadine Salzmann, Charles Fierz, Rodica Nitu, Patricia de Rosnay, Kari Luojus
Corresponding author: Charles Fierz
Corresponding author e-mail: fierz@slf.ch
In recent years, the idea of a Snow Monitoring Competence Centre came up within the Global Cryosphere Watch (GCW) community, in particular the Expert Team on Snow Watch. First ideas were presented at scientific conferences and WMO workshops and meetings. The concept attracted interest but it was not possible to come up with a concrete ‘business plan’ for such a centre. With the recent integration of the activities of GCW under the remit of the WMO Infrastructure Commission (INFCOM), the question of assuming a lead role in sustaining the availability of quality snow data arose again. The proposal is to build on the existing mature framework of WMO Measurement Lead Centres and to propose establishing knowledge-based competence hubs that would assume functions in support of sustaining the quality of snow observations and the quality of data, and including capacity-development activities. Additionally, these hubs would link to the GCW data portal to facilitate access to data sets and data providers. Once the framework is established, one or more centres could apply to obtain this designation The proposal of these centres it timely, since this coincides with the launch of the newly established Joint Body on the Status of the Mountain Snow Cover (JB-SMSC), a joint venture between the International Association of Cryospheric Sciences (IACS), the Mountain Research Initiative (MRI), and GCW. A strong mutual benefit is expected, including the proposed centres as long-term repositories for the knowledge developed within the scope of JB-SMSC. It is proposed that Snow Monitoring Competence Centres become specialized WMO Measurement Lead Centres providing, for example, answers to questions concerning in situ measurement practices of common snow variables, and standards for snow-data quality checks, as well as connecting both measurement and modelling experts to raise awareness of each other’s needs. This contribution will outline the benefits of establishing these centres and the path forward.
Saoirse R. Goodwin, Thierry Faug, Guillaume Chambon
Corresponding author: Thierry Faug
Corresponding author e-mail: srgoodwin@protonmail.com
Wet-snow avalanches are becoming a progressively more commonplace hazard in mountainous areas, as a consequence of climate change hastened by anthropogenic activities. This is increasing the risk to mountain-dwellers and associated infrastructure. Wet-snow avalanches tend to be denser and pastier than their dry-snow counterparts. Notably, current designs for mitigation structures against dry avalanches may be suboptimal for mitigating wet cases. However, numerical studies into wet-snow avalanches are rather limited. Furthermore, there is currently little consensus regarding what types of constitutive model might be appropriate for modelling wet-snow avalanches. Other difficulties pertain to the model resolution required when modelling various combinations of flow rheology and topographical complexity; as well as developing objective criteria that can determine when flows have come to rest, especially when avalanches are traversing complex topographies. In this work, we first detail an extended Voellmy model that includes a term for cohesion. This model was implemented into a finite volume method solver for the incompressible Saint-Venant equations. We hence carried out a systematic sensitivity study to investigate the interplay between apparent cohesion, the mesh resolution and the complexity of the topography. The numerical domains all included a steep slope and a horizontal runout zone. We used a Perlin noise generator to procedurally superimpose complex topographies with features characteristic of real mountains. This allowed us to study similar topographies, which differ only in terms of the spatial distribution of specific features. Results reveal that flow dynamics can be very sensitive to topographical complexities; more so at higher resolutions, as well as for lower values of the cohesion term. The relative lack of sensitivity of highly cohesive flows to complex topographies is partly due to their tendency to maintain a thicker flow depth, relative to the characteristic size of the topographical features. Furthermore, we found that the cohesion term governed the way in which flows became arrested: highly cohesive flows become arrested from the top down, whereas less cohesive flows could only become arrested on the runout zone. Finally, a robust set of arrest criteria remains elusive, although we suggest several possible criteria that can be used complementarily.
An elastic-viscoplastic constitutive model for snow avalanche dynamics
Lars Blatny, Johan Gaume
Corresponding author: Lars Blatny
Corresponding author e-mail: lars.blatny@epfl.ch
Accurate modeling of snow avalanches is crucial to hazard mapping and mitigation in mountainous regions. Current models are mostly based on depth-averaged flow equations with viscoplastic laws, e.g. the Voellmy fluid law considering the avalanche rheology as a result of Coulomb and turbulent-type friction. Mesh-free continuum schemes, such as the particle finite element method (PFEM), smoothed particle hydrodynamics (SPH) and the material point method (MPM), have become increasingly popular in simulating full 3-D granular avalanches as they can readily handle large deformation problems. In particular, an elastoplastic approach based on critical state soil mechanics (CSM) in an MPM scheme was recently proposed, enabling the solid–fluid phase transition needed to capture the release and flow of an avalanche, taking into account the snow elasticity and its mixed-mode failure. However, while it is known that snow avalanches can exhibit strong dependency on the rate of the flow, this constitutive model is rate-independent. Here, we suggest the introduction of rate-dependency to a CSM-based model by considering Perzyna’s overstress approach and μ(I) rheology, the latter originally developed for dense granular flow. Our model facilitates the treatment of dilatancy and compressibility through a hardening law of a Cam-Clay yield criterion and the associative plastic flow rule. With the model implemented in a B-spline material point method (BS-MPM) scheme with finite strain hyperelasticity, our approach is well suited to large deformation problems without suffering from mesh-distortion issues. The model is demonstrated on various setups of different scales, including full-scale avalanche dynamics simulations.
Nicolas Eckert, Florie Giacona, Matthieu Lafaysse, Samuel Morin
Corresponding author: Nicolas Eckert
Corresponding author e-mail: nicolas.eckert@irstea.fr
The cryosphere is highly sensitive to warming, which may dramatically modify snow on the ground and avalanche activity. As snow is a major component of the water cycle and snow avalanches represent a significant threat, assessing and understanding their changes is of major societal importance. However, because of the lack of sufficiently comprehensive observation records and biases in projections of weather conditions in mountain terrain, decadal to centennial changes remain poorly documented so far. Here we analyse how snow on the ground, avalanche activity and their climate drivers have evolved over the 1865–2015 period in the Vosges Mountains, a low-to-medium (>1500 m) mountain range of northeast France. We base our analysis on i) a combination of weather data and proxy reconstructions and ii) avalanche activity from historical records homogenized with Bayesian techniques over the entire period, iii) snow-on-the-ground measurements since the late 19th century, iv) snow-cover reanalyses since 1958. We first evaluate the different data series, and demonstrate how homogenized avalanche activity and reanalyses improve the mutual consistency of the different data sets with regard to raw records. We then provide a baseline local snow and avalanche climatology for the 1985–2015 ‘current’ period. Eventually, we demonstrate the massive shrinkage of snow conditions that has occurred over one and a half centuries with warming. However, the latter was modulated by elevation, being much more marked at the lower elevations of the massif. Also it was very irregular, with notable sub-periods of harsher winter conditions and, by contrast, periods of a sharp drop in snow conditions as functions of climate variability. Regarding avalanches, a major decrease had already occurred between the end of the Little Ice Age and the early 20th that caused the disappearance of avalanche activity at low elevations of the massif. Since then, it has more or less remained stable, but with modulation of avalanche type, size and timing as functions of changes in snow conditions. These results confirm that snow and avalanche activity react strongly to changes in climate conditions, but in a complex and non-linear way. They indicate that long-range analyses of past changes in low-to-medium-elevation mountains may serve to anticipate future changes in higher environments, and could thus help in the design of efficient adaptation strategies under ongoing warming.
Elisabeth D. Hafner, Frank Techel, Yves Bühler
Corresponding author: Elisabeth D. Hafner
Corresponding author e-mail: elisabeth.hafner@slf.ch
Where avalanches occurred and how big they were, is crucial information for avalanche forecasting. Furthermore, the accurate mapping of observed avalanches is important for risk management, hazard mitigation measures, hazard zonation and numerical simulations. Currently, in Switzerland, observers provide an assessment of avalanche size together with the approximate location of the avalanche (a point) or they map the outlines of avalanches, while avalanche forecasters recording avalanches primarily use photographs provided by third parties for these tasks. In recent years, avalanches have increasingly been mapped from remote-sensed imagery such as satellite images or ortho-photos derived from aeroplanes or drones. Even though assessments by human experts have often been treated as the ‘gold standard’ in the past, numerous studies in various fields have found differences among expert assessments. The aim of this work is therefore to investigate consistency in the assessment of avalanche size as well as interobserver variability in the mapping of avalanche outlines from photographs or from remote-sensed imagery. We conducted three investigations to achieve this objective: a) We carried out a survey among avalanche forecasters requesting an assessment of avalanche size for ten avalanches depicted on photographs. b) We asked human experts to independently map six avalanches from photographs. c) We requested human experts to map all avalanches they could identify on remote-sensed imagery in an area of interest. We then performed statistical analysis for those three use cases quantifying consistency in the assessment of avalanche size and the variability in the mapped avalanche outline. With this investigation, we want to draw attention to this often-neglected variability and quantify the uncertainty for avalanche mapping, not only raising awareness among practitioners concerning the variability of the data they are relying on, but also providing benchmark values for applications incorporating such information. Moreover, we hope that these benchmark values will serve as a standard to compare to for future automation, for instance with deep learning.
Snowpack, soil and forest energy budget and flux partitioning in boreal ecosystems
Jari-Pekka Nousu, Matthieu Lafaysse, Giulia Mazzotti, Pertti Ala-aho, Hannu Marttila, Samuli Launiainen, Bertrand Cluzet, Mika Aurela, Pasi Kolari, Annalea Lohila
Corresponding author: Jari-Pekka Nousu
Corresponding author e-mail: jari-pekka.nousu@oulu.fi
Snowpack, as situated between the atmosphere and the ground, has a major influence on the winter surface energy budget. Accurate simulation of the snowpack energy budget is challenging due to i) landscape properties that complicate the radiation budget (e.g. vegetation and topography), and ii) general limitations in the theory of turbulent flux modelling over snow. In fact, turbulent fluxes are known to be one of the major sources of uncertainty in snow modelling, and forest snow has been identified as a critical issue in climate modelling. Despite recognition of these factors, and the known limitations of the theory, detailed studies that evaluate snow and land surface models against observations of all surface energy components are rare. In this study, we compare SURFEX/ISBA-MEB-Crocus model simulations with energy flux observations from eddy covariance stations located in Finland. With four independent stations, we are able to cover two different climate and snow conditions, southern and northern subarctic zone, and two contrasting landscape types typical for the boreal region, forest and peatland. In particular, we test the sensitivity of energy fluxes simulated with different process parameterizations included in the model. In addition, due to the complex nature of the surface energy fluxes and their interconnection, we compare different ways to conceptualize the soil and the vegetation and assess their performance in simulating snow conditions and soil thermal regimes in subarctic climates. Our results show that the turbulent fluxes simulated by the classical theory do not satisfactorily represent the observed values, so it is necessary to increase the simulated turbulent exchange under stable conditions. Although the energy budget of a forest remains challenging to simulate accurately, we demonstrate the major improvements on forest snow simulations when the vegetation energy budget is explicitly incorporated into the model. Moreover, we found the inclusion of soil organic carbon in soil parameterization to have a major positive impact on simulated soil thermal properties.
Climate change impacts on snow avalanche hazard: A climate scenario-based approach
Gregor Ortner, David N. Bresch, Michael Bründl, Adrien Michel, Yves Bühler
Corresponding author: Gregor Ortner
Corresponding author e-mail: gregor.ortner@slf.ch
Several recent studies have shown that changes in climate, such as higher temperatures and more intense heavy precipitation, strongly influence gravity-induced natural hazards. As part of the Climate Change Impacts on Alpine Mass Movements (CCAMM) research program, we are developing a framework for modeling mass movement hazard and risk including the impact of climate change. In a first approach, we modeled avalanche hazard and risk in central Switzerland for the current climate situation for three return periods. For each of these return periods, we used3-day snow depth derived from meteorological stations for defining avalanche formation. The RAMMS::LSHIM Large Scale Hazard Indication Mapping algorithm was used for the avalanche hazard modelling. It combines an automatic delineation of potential release areas with a high-resolution forest layer to represent the spatial distribution of avalanche hazard within the selected hazard setting. For modelling potential climate change impacts on avalanche hazard, we use hourly data downscaled from daily CH2018 data. With these data we try to generate diurnal cycle of temperature and precipitation for the RCP 8.5 scenario simulations and thus represent possible changes of snowfall by using the multi-purpose snow and land-surface model SNOWPACK. We applied extreme value statistics to investigate possible changes in the 3-day snowfall accumulation and thus the avalanche fracture depth at different return periods. Future climate snowfall and avalanche release depth are derived for the first and the second half of the 21st century. These snow accumulations can be used for avalanche simulations and thus for estimating possible changes in avalanche hazard and risk in an extreme climate change scenario.
Influence of attachment kinetics inclusion on modeling the specific surface area evolution in snow
Anna Braun, Kévin Fourteau, Henning Löwe
Corresponding author: Anna Braun
Corresponding author e-mail: anna.braun@slf.ch
Vapor fluxes in snow are often inferred from the temperature field by assuming vapor concentrations to be in local thermodynamic equilibrium with the temperature. However, this assumption is in contradiction with the observed evolution of the specific surface area (SSA) in temperature gradient metamorphism. Here, we demonstrate that the inclusion of attachment kinetics due to the coupling of vapor and heat transport leads to an agreement between pore-scale simulations and 4-D experimental data. The coupled pore-scale temperature and vapor fields are calculated with the Finite Element (FE) Software Elmer on microtomography images. Subsequently, the rigorous evolution equation for the surface area per unit volume is utilized to calculate the relevant microstructure parameters by volume averaging the numerical FE solution. This procedure facilitates a rigorous validation of the surface area dynamics without the need to explicitly track or integrate the moving ice–air interface in 3D space. Assuming isotropic, homogeneous attachment kinetics, the best estimate for the kinetic coefficient is obtained through parameter optimization by matching numerical and experimental SSA curves. Sensitivity studies are carried out to assess the impact of experimental errors, FE mesh quality and the assumption of homogeneous attachment kinetics on the surface.
SPASS: a spatial snow climatology for Switzerland
Adrien Michel, Stefanie Gubler, Sven Kotlarski, Johannes Aschauer, Christoph Marty, Tobias Jonas
Corresponding author: Adrien Michel
Corresponding author e-mail: adrien.michel@epfl.ch
The SPASS research project aims to develop a daily gridded snow dataset for Switzerland on a climatological scale (i.e. since 1961) based on the modelling system currently used by SLF’s Operational Snow Hydrological Service (OSHD). This project is a collaboration between SLF and MeteoSwiss. Snow water equivalent (SWE) maps over Switzerland are produced at a resolution of 1 km and at a daily resolution using the OSHD model. This model uses sequential data assimilation methods to account for available snow measurements. However, the temporally consistent use of data assimilation is not possible over a long time scale as the availability of snow observations is constantly changing over time. For this reason, the present project uses a climatological version of the OSHD model. This version is forced by daily temperature and precipitation gridded data sets starting in 1961 developed by MeteoSwiss and runs without snow data assimilation. Then, we use a spatialized quantile mapping procedure to correct the biases in these offline simulations by using the data-assimilated simulations (since 1999) as training data for the quantile mapping algorithm. We first present the developed method, which is available as an open-source fast R package. The package can be used for other similar studies. We then present our first analysis of spatial trends in SWE since 1961 at different elevations in Switzerland.
Thermal conductivity of snow on sea ice
Amy R. Macfarlane, Lucille Gimenes, Matthias Jaggi, Henning Löwe, David Wagner, Martin Schneebeli
Corresponding author: Martin Schneebeli
Corresponding author e-mail: schneebeli@slf.ch
The thermal conductivity of snow on sea ice has been challenging to measure directly. The MOSAiC-expedition allowed us to measure this in situ using the most precise measurement technique, X-ray micro-computed tomography. Direct numerical simulation of the heat flow in the porous structure of snow results in its thermal conductivity. We present a new set of thermal conductivity snow measurements for the deep winter period January–March 2020. We subsampled ten cubes from 10 cm long snow cores for the calculation. We found that the conductivity on sea ice is about 30% higher than for a typical Alpine snowpack at a density of 300 kg m–3. The thermal anisotropy of the samples varies between 0.75 and 2. This large anisotropy shows that density alone is insufficient for precise parameterization. Using the vertically resolved thermal conductivities from complete snow profiles, we show that thermal resistance increases with increasing snow depth. This work was funded by the Swiss Polar Institute, the Swiss National Science Foundation andHorizon Europe 2020 ARICE, and supported by Scanco Medical AG.
Seismo-acoustic signal characterization and array processing for dry- and wet-snow avalanches
Christine Seupel, Cristina Pérez-Guillén, Alec van Herwijnen
Corresponding author: Christine Seupel
Corresponding author e-mail: christine.seupel@slf.ch
Seismic and infrasound systems are well suited to monitor avalanche activity independent of weather conditions. It is also possible to characterize avalanches and their dynamics, providing complementary information. Whereas seismic waves are generated by the friction and impact of the avalanche with the terrain, infrasound waves are produced by the interaction of fast-flowing snow with air. It is therefore generally assumed that infrasound signals contain less information about the avalanche than seismic signals. To investigate this in more detail, we installed a combined seismo-acoustic array in the Dischma valley above Davos. The array consisted of five co-located seismic and infrasound sensors, and we collected data during two winter seasons. During the second winter season, we also installed a Doppler radar at our test site to obtain independent validation data and accurate release times. Finally, data from several automatic cameras, field surveys and airplane photogrammetry were used to characterize the type and size of avalanches. These data allowed us to comprehensively compare signal characteristic and array processing results of the seismo-acoustic array for both dry- and wet-snow avalanches. Compared to dry-snow avalanches, wet-snow avalanches generate higher-amplitude seismic signals but lower-amplitude infrasound signals. Using array processing techniques, we obtained source parameters from both seismic and infrasound signals and compared those with the radar measurements. Results of seismic signals show more scatter than for infrasound signals, as the propagation of the seismic waves in the ground is more complex. Furthermore, beamforming results suggest that seismic source duration of wet-snow avalanches is longer and more energetic, confirming that seismic detection is more effective for these avalanche types. Overall, our results show that combining seismic and infrasound sensors provides more detailed information on avalanches, and could improve the automatic detection and characterization of avalanches.
A depth-averaged material point method for the initiation and dynamics of snow-slab avalanches
Louis Guillet, Thomas Pauze, Bertil Trottet, Lars Blatny, Denis Steffen, Johan Gaume
Corresponding author: Johan Gaume
Corresponding author e-mail: louis.guillet@epfl.ch
A snow-slab avalanche releases due to the failure of a weak snow layer buried below a cohesive snow slab. Once in motion, the avalanche transits from a solid-like snowpack to a fast gravitational flow which poses a threat to snow enthusiasts and infrastructure in mountainous regions. Recently, large-scale simulations based on the material point method (MPM), a hybrid Eulerian–Lagrangian approach, allowed researchers to gain new insights into the initiation and dynamics of snow avalanches. However, the high computational cost of such simulations prevents their automatic deployment for hazard-mapping purposes. In view of keeping the advantages of MPM such as its ability to handle large deformations and collisions, its simplicity of implementation and parallelization, we developed a depth-averaged MPM (DAMPM) in the framework of shallow water equations. This model was then applied separately to study the release and the dynamics of snow avalanches. For avalanche release simulations, the weak layer is treated as an external shear force acting at the base of the slab and is modelled as an elastic quasi-brittle material with residual friction. We first validate the model based on simulations of the so-called propagation saw test (PST) by comparing numerical results to analytical solutions and three-dimensional simulations. Second, we perform large-scale simulations and analyse the shape and size of avalanche release zones, before applying the model to more complex topographies. For avalanche dynamics simulations, different rheological models are implemented. The model is validated by comparing DAMPM simulations of granular collapse and dam break to three-dimensional simulations, experimental results and analytical solutions. Furthermore, an adaptation of the Savage-Hutter model is proposed to perform simulations over complex topographies. We show preliminary results of such simulations and discuss challenges related to different assumptions. Due to the low computational cost compared to three-dimensional MPM, we expect our work to have important operational applications for the evaluation of avalanche release sizes and runout distances required in hazard-mapping model chains. Finally, the model can be easily adapted to simulate other types of shallow mass movements such as debris flows, landslides, glacier processes and outburst floods.
Lisa Bouvet, Neige Calonne, Frédéric Flin, Christian Geindreau
Corresponding author: Lisa Bouvet
Corresponding author e-mail: lisa.bouvet@univ-grenoble-alpes.fr
Although temperature-gradient metamorphism experiments on snow layers have been widely studied, several micro-scale processes remain unclear, such as the vertical mass transfer and the influence of crystal-growth regimes on the development of the snow grains. Both processes involve the observation of micro-scale properties on the vertical dimension. Two new experiments of a snow layer evolving at 100 K m–1 for 3–4 weeks have been conducted to further investigate the temporal evolution of microstructural properties, as well as their evolution in the vertical direction. Vertical snow columns were sampled at regular time intervals and scanned by X-ray tomography to monitor the microstructural features at resolutions of 8 and 21 μm. The analysis of those profiles led to interesting results, such as the quantification of the local mass loss in the lowest and warmest part of the snow layer. Another notable result is the development of larger grains in the middle of the snow layer compared to the top and bottom, highlighting higher grain-growth rates at that location. This outcome can be linked to the different crystalline growth regimes depending on temperature conditions, which are classically observed for the growth of crystals in the atmosphere.
Bicontinuous microstructures from level-cut Gaussian random fields as numerical snow analogs
Henning Löwe, Simon Frei
Corresponding author: Henning Löwe
Corresponding author e-mail: loewe@slf.ch
To develop physical snow models from microstructure-based simulations it is indispensable to utilize 3D model media for testing numerical procedures. As a minimal requirement for snow analogs, these model media should not be made of spheres and show realistic variations for common microstructural parameters. Level-cut Gaussian random fields (LC-GRF), originally developed in the context of micro-emulsions, fulfill these requirements and are indeed already used for snow-related simulations such as microwave scattering and micro-mechanics. To facilitate a wider availability and common understanding of LC-GRF for snow studies, we implemented an open-source microstructure generator in Python for creating anisotropic LC-GRF random media. Our code implements a well established algorithm based on Fast Fourier transformation and a prescribed GRF covariance which is generalized to anisotropic media with spheroidal symmetry. This type of anisotropy is commonly used for the analysis of anisotropic thermal, dielectric and elastic properties of snow. We demonstrate the correctness of our implementation by an extensive comparison of parameters inferred by 3D image analysis with their analytical values for LC-GRF which are available for volume fraction, specific surface area, first and second moment of the mean curvature, Gaussian curvature, geometrical fabric tensor, and the two-point correlation function. Specifically, we explore the conditions of image resolution and image size under which the inferred parameters converge to the exact results. Extensions and limitations of the implemented method are discussed in view of diverse applications of microstructure-based snow simulations.
Evaluation of the InfraSnow: a handheld device to measure snow specific surface area
Fabian Wolfsperger, Silvio Ziegler, Martin Schneebeli, Henning Löwe
Corresponding author: Fabian Wolfsperger
Corresponding author e-mail: wolfsperger@slf.ch
Describing the snow microstructure quantitively is crucial to investigate any physical process within the snowpack. Whereas valid and reliable laboratory measurement exists, e.g. micro-CT scanning or gas absorption methods, measuring snow structural quantities on the field is still challenging. So far, the only field-ready and market-available device has been the IceCube (IC), which deduces the snow’s specific surface area (SSA) from its diffuse reflectance at near-infrared (NIR) wavelengths of 1320 nm in combination with a radiative transfer model. The InfraSnow (IS) uses a similar measurement principle (at 945 nm), but was realized as a compact handheld instrument. This allows non-destructive measurements on the snow surface or along snow profiles as well as on snow surfaces where sampling is undesirable, e.g. ski slopes. To make the advantages of the IS available, SLF and FPGA Company GmbH collaborated to launch a small-volume production of 25 devices, which are available in a bundle with a dielectric sensor for density and LWC measurements. The new IS devices were evaluated on-field and in the laboratory using micro-CT and IC reference measurements. In contrast to the prototype validation of Gergely, the conducted comparison with CT-measurements included also new snow samples with SSA up to 63 mm–1. The laboratory data revealed the IS method systematically overestimating the CT-SSA from about 2% to 15% at SSA between 3 and 61 mm–1. In contrast, the relative random error was found to decrease with increasing SSA from about 20% to 1%. IC-SSAs lay in between the IS and CT-SSA. For both, the IS and the IC, sampling and the surface preparation, respectively, had a significant influence on the results. To compensate for the systematic error of the IS, an adjustable correction factor was implemented to calibrate the IS-SSA to the CT-SSA. A reason for the SSA overestimation is possibly the underlying optical snow model, which is limited to represent high-SSA snow, as a simplified snow microstructure of ice spheres is assumed. However, as the measured IS reflection depends on the SSA and on the snow density, the latter can contribute significant uncertainty to the measurement result. An uncertainty in density of ±20 kg m–3 can lead to an absolute SSA error of up to 20 mm–1 for low density snow. For snow with a density above 150 kg m–3 the error propagation due to density remains below about 8 mm–1.
Bertil Trottet, Ron Simenhois, Grégoire Bobillier, Alec van Herwijnen, Johan Gaume
Corresponding author: Bertil Trottet
Corresponding author e-mail: bertil.trottet@epfl.ch
The release of snow-slab avalanches starts with the failure of highly porous weak layer buried beneath a cohesive slab leading to mixed-mode crack propagation along the slope. The first modelling attempt of the process dates back to 1979 with a pure shear weak layer fracture assumption proposed by McClung. Later, Heierli et al. extended the anticrack concept to slab-avalanche release, accounting for weak layer volumetric collapse and subsequent slab bending. Recent advances reconciled these approaches and reported, above a critical slope angle, the existence of a transition from slow anticrack propagation to a supershear regime with intersonic propagation speeds. Here, based on the material point method, finite strain elastoplasticy and critical state theory, we highlight that in specific cases of rather soft slabs, slab fracture can prevent the supershear transition from occurring. We specifically investigate the conditions for crack arrest or so-called ’en echelon’ slab fracture and their interplay with the propagation regimes. In particular, the anticrack to supershear transition arising from the increase in slab strength leads to a transition in the characteristic size of released blocks. Furthermore, three-dimensional simulations reveal interesting propagation and release patterns related to the interplay between cross-slope, down/up-slope propagation and slab tensile failure. This enables the analysis of slab fracture modes at the crown, flanks and staunchwall of the avalanche. These findings are eventually compared to an analysis of several in situ avalanche videos treated using the Euler video magnification (EMV) technique. It allow us to reach a next step in our understanding of the release mechanics towards the prediction of avalanche release shapes and sizes.
Thomas Nagler, Gabriele Schwaizer, Riccardo Barella, Jose-Luis Bueso-Bello, Edoardo Cremonese, Richard Essery, Markus Hetzenecker, Tobias Jonas, Lars Keuris, Roland Koch, Carlo Marin, Nico Mölg, Claudia Notarnicola, Marc Olefs, Paola Rizzoli, Helmut Rott, Benedita Santos
Corresponding author: Thomas Nagler
Corresponding author e-mail: thomas.nagler@enveo.at
AlpSnow (2020–22) is a science activity within ESA’s Alpine Regional Initiative, addressing the development of novel Earth observation (EO) techniques and algorithms for the generation of snow products optimized for specific scientific and operational applications in mountain areas. The AlpSnow products cover several physical parameters of the seasonal snowpack, namely snow area extent, surface albedo, grain size, snow water equivalent (SWE), snow depth, snowmelt area extent and liquid water content. The usability of the products will be demonstrated within six scientific and operational use cases in meteorology, snow science, hydrology and water management. Algorithm development and validation activities are performed in five test areas in different Alpine regions, two of which are part of the INARCH project. The test areas are equipped with operational field stations recording meteorological and snow parameters. Additional snow reference data are collected during field surveys. In the first phase of the project several candidate algorithms for each snow parameter were implemented and evaluated. A set of preferred algorithms was selected to be further optimized for complex topography and diverse surface cover. For high-resolution snow extent two algorithms are applied using Sentinel-2 and Landsat multispectral data, a linear multispectral unmixing approach and a method using machine learning techniques. For surface albedo and grain size, algorithms adopted from Painter et al. (2009) and Kokhanovsky (2015) are applied. A critical issue for both algorithms is the correction of topographic effects. The monitoring of snowmelt extent is based on dense radar image (SAR) time series of Sentinel-1. For SWE mapping two approaches are explored, the assimilation of EO snow extent products into the physical SNOWGRID model of the Austrian meteorological service (ZAMG), and the use of repeat-pass interferometry using long wavelength (L-band) satellite SAR data for high-resolution SWE retrieval. For snow depth mapping we investigate the suitability of differencing DEMs from TanDEM-X interferometric SAR data acquired during snow-free conditions and close to the maximum snow accumulation at the start of the snowmelt season. We will present results of the experiments dedicated to the selection of retrieval algorithms for the different snow parameters and report on the accuracy of the prototype algorithms and products developed within the project.
Measuring snow and avalanche properties using acoustic and seismic distributed fiber optic sensing
Alexander Prokop, Nicola P. Agostinetti, Bernhard Grasemann
Corresponding author: Nicola P. Agostinetti
Corresponding author e-mail: alex@snowscan.at
Since 2012 we monitor avalanche activity using distributed acoustic and seismic fiber optic sensing at our avalanche test area at Lech am Arlberg, Austria. The method is based on an optical time domain reflectometer system that detects seismic vibrations and acoustic signals on a fiber optic cable that can have a length of up to 30 km in 80 cm resolution. While in the first years we focused on successfully developing an operational avalanche detection system that is able to tell in real time reliably when an avalanche was triggered and what the size of the avalanche is, we now present our investigations of the seismic signals to measure snow properties such as snow depth and avalanche properties such as flow behavior. Our test in winter 2022 recorded by blasting triggered avalanches and during data post processing we extracted seismic guided waves. We discuss methods for extracting information from guided waves for measuring snow depth, which was verified against spatial snow depth measurements from terrestrial laser scanning. Analyzing the seismic signals of avalanches with run-out distances ranging from a few metres to approximately 250 m allows us to differentiate between wet and snow avalanches, which is discussed in the context of avalanche dynamics.
Comparison of AMSR2 SWE data with ground observations over the Kars region in eastern Türkiye
Mehmet Ali Tor, Ahmet Emre Tekeli, Senayi Dönmez
Corresponding author: Mehmet Ali Tor
Corresponding author e-mail: tor.mehmet1@gmail.com
Rivers located in eastern Türkiye are mainly fed by snowmelt. Therefore, discharges in those rivers increase with rising air temperatures during spring months. Dams located in eastern Türkiye are filled with incoming snowmelt and enable the regulation of the downstream flow. Thus, accurate estimation of snowmelt is crucial for optimum operation of these dams, which depends on knowledge of the water available in the basin. The main parameter determining water potential of the basin in eastern Türkiye’s watersheds is snow water equivalent (SWE). SWE, indicating the water that will be released as snow melts, is affected by meteorological, topographical and/or land use parameters such as snow depth, wind, surface temperature, precipitation, snow covered area, surface albedo, slope and vegetation cover, and may show large variation within the basin. SWE can be estimated by satellite remote sensing, ground penetrating radar and/or by ground measurements, the last method being considered the most accurate. Unfortunately, performing frequent site visits and getting ground measurements is not that easy in regions of rough topography such as eastern Türkiye. Advanced Microwave Scanning Radiometer 2 (AMSR2) on board the JAXA GCOM-W1 satellite provides daily SWE observations at 25×25 km2 spatial resolution. This study compares AMSR2 SWE data with ground SWE observations performed with Mount Rose snow sampler during site visits to the Kars region in eastern Türkiye.
Simone Schauwecker, Álvaro Ayala, Gonzalo Cortés, Katerina Goubanova, Shelley MacDonell, Nicole Schaffer, Eduardo Yáñez
Corresponding author: Simone Schauwecker
Corresponding author e-mail: simoneschauwecker@gmail.com
In semi-arid Chile, river discharge and freshwater supply is mainly dominated by melting snow from high-elevation areas of the Andes (>2000 m a.s.l.). Since the seasonal snow cover depends on few winter events, there is a large year-to-year variability in the snow water equivalent (SWE). Typically, there are some dry years with very low annual precipitation, which are compensated by wet years. However, extraordinarily dry conditions have been experienced almost continuously since 2010 together with increased water consumption in the region, which has resulted in considerable water stress and increasing public concern. There is a growing need to know the actual SWE stored in the mountain snowpack for efficient water allocation and water management. Currently decisions are based on a few point measurements of the SWE and snow-area estimations from MODIS. In the Coquimbo region (nearly the size of Switzerland, with ~35% of its area over 2000 m a.s.l..) there are only about 5–10 automatic weather stations registering snow depth. The SWE estimates are therefore very uncertain and this inhibits efficient water allocation with important implications for water security impacting sectors such as hydropower, agriculture and domestic use. To improve SWE estimates during the winter season, we are developing a new operational SWE estimation tool for water resources decision-making in the Coquimbo region (SWEET-Coquimbo). The SWE will be estimated in near real time (every 2 weeks) during the winter using a data assimilation framework that combines meteorological forcing ensembles from ERA5 reanalysis data, hydrological modelling and satellite observations of the snow-covered area. Along with the operational SWE estimates, retrospective SWE estimates from 1985 to the present will be used to improve our understanding and modelling of the spatial distribution and temporal evolution of the SWE in the region. The results will be published on an open-access web platform and will be available as inputs for downstream applications such as hydrological models and hydrological assessments. We will present first results of the downscaling of meteorological forcing from ERA5 (e.g. bias corrections and precipitation ensemble generation) and the data assimilation framework and discuss the main challenges in estimating SWE in near-real time.
NASA SnowEx mission: towards mnitoring snow mass globally from space
Hans-Peter Marshall, Carrie Vuyovich, Kelly Elder, Mike Durand
Corresponding author: Hans-Peter Marshall
Corresponding author e-mail: hpmarshall@boisestate.edu
The goal of the NASA SnowEx mission is to prepare the snow remote-sensing community for a dedicated spaceborne snow mission. During 2020, we performed an intensive 3-week airborne and field campaign on Grand Mesa, Colorado, USA, which included over 100 participants from over 20 institutions, five different aircraft and seven airborne instruments. Many different snow remote-sensing approaches were tested, including SAR, InSAR, passive microwave, thermal IR, lidar, and gamma sensors. These datasets are being used to test mission concepts and algorithms for retrieval of snow depth, SWE and snow surface temperature. During both 2020 and 2021, we performed a time series experiment with airborne L-band InSAR, with weekly to biweekly overflights from January to March, at 14 sites across the western USA. We are using this campaign to test SWE and depth-change retrieval approaches to prepare for the upcoming launch of NISAR in 2024. The NASA SnowEx mission is preparing us for a new chapter in snow remote sensing, which shows promise for monitoring SWE from space at biweekly intervals and spatial resolutions less than 500 m.
Spaceborne snow-depth measurements from ICESat-2 laser altimetry
Désirée Treichler, Marco Mazzolini, Livia Piermattei, Clare Webster, Luc Girod
Corresponding author: Désirée Treichler
Corresponding author e-mail: desiree.treichler@geo.uio.no
Snow-depth measurements such as from weather stations or targeted remote sensing campaigns mainly exist in easily accessible places in wealthy regions of the Earth. In remote, mountainous areas, the lack of inexpensive methods to measure snow amounts strongly limits the monitoring of supplies for drinking water, irrigation, hydropower or flood hazards for a large part of the Earth’s population. We show that data from the spaceborne laser altimeter ICESat-2 can provide high-quality snow-depth measurements when combined with an accurate digital terrain model (DTM). ICESat-2 measures the surface elevation of the Earth along profiles with a revisit time of 3 months, whereby ground tracks are shifted in space between overpasses in the mid-latitudes. Data from three field studies in Norway and Finland show that snow depths derived from the freely available ICESat-2 ATL03 data product match the vertical accuracy of snow depth data acquired with a drone-based lidar sensor and snow probe measurements. This is also true for forested areas where snow amounts are difficult to measure. Given cloud-free conditions and a very-high-resolution DTM, one ICESat-2 overpass yields six parallel snow depth profiles with ~1 m along-track spacing and a snow-depth accuracy at centimeter- to decimeter-scale. Lower-resolution DTMs or the presence of moderate elevation bias in the DTM can still allow for regional snow-depth averages but require more extensive data processing and spatial binning. The method is thus limited by the availability of a high-quality DTM, whereby the absence of elevation bias is far more important than the spatial resolution of the DTM. A further limitation is the satellite’s low repeat time combined with the sparse ground track pattern at mid-latitudes, resulting in few snow-depth profiles over the course of one winter – in particular in cloudy areas such as western Norway. However, an ICESat-2-derived snow-depth profile can provide valuable information about the spatial distribution of snow depths within a region that is not captured by point-based snow-depth time series from a weather station. With the quality and accuracy of global DTMs ever improving, we expect that the accuracy of snow-depth measurements derived from spaceborne laser altimetry will continue to improve in the future and provide currently lacking snow-depth data in remote and mountainous areas.
Spatial distribution and controls of snowmelt hotspots in the Desert Andes of Chile
Álvaro Ayala, Simone Schauwecker, Shelley MacDonell
Corresponding author: Álvaro Ayala
Corresponding author e-mail: alvaro.ayala@ceaza.cl
Snow plays an important role in the water cycle of high-elevation mountain regions. For example, the western side of the Desert Andes (20–32°S) is a high-elevation mountain range (with peaks >6000 m a.s.l.) where the snowpack is formed by only a few winter events and this snow is the main water source for streams and groundwater in the melt season. Previous research in the Desert Andes has shown that snow accumulation patterns are shaped by the interaction between strong winds and topography, whereas the partition between melt and sublimation is defined by solar shading and wind exposure. In this work we hypothesize that the meteorological and topographic conditions conducive to sublimation define large geographical areas, compared to relatively few sites where snowmelt is dominant over sublimation losses. To test this hypothesis, we study snow accumulation and melt patterns in Los Corrales basin (30.16° S, 69.88° E), Chile, a 79 km2 well monitored catchment with an elevation range between 3967 and 5823 m a.s.l. Our main objectives are to identify sites with large snow accumulation and melt and characterize the dominant processes that define these sites. To this end we use snow-presence indices calculated using a set of optical satellite images and results from SnowModel, a process-based snow evolution model that was applied in the study area for the period 2019–21. Results show that 60% of the accumulated snow in the catchment sublimates and that southeast-oriented slopes below 4500 m a.s.l. are the largest snowmelt contributors. There is little snow accumulation on the northwest-oriented slopes, and strong winds yield large sublimation amounts (>70% of the total accumulation) that largely reduce snowmelt runoff. Preliminary estimates show that at least 50% of the total melt volume originates from only 30% of the catchment area. Sensitivity tests show that the partition between sublimation and melt is most sensitive to the surface-roughness model parameter. Our results suggest that a detailed monitoring of these melt-favourable sites would improve our understanding of the hydrology and ecosystems of the Desert Andes. Model results are also being used as a base for an operational SWE monitoring system in the study region.
Eunsang Cho, Colleen Mortimer, Lawrence Mudryk, Chris Derksen, Carrie Vuyovich, Mike Brady
Corresponding author: Eunsang Cho
Corresponding author e-mail: eunsang.cho@nasa.gov
Terrestrial snow research in hydrological and climate sciences at continental and global scales typically rely on spatially distributed snow water equivalent (SWE) products. Recently, many gridded SWE datasets have been developed by the snow, climate and hydrological sciences community, which must be validated with independent and reliable observations in various environments. However, such validation is challenging due to limited spatial coverage of available reference data (e.g. in-situ measurements and airborne flight lines), so validation with multiple sources of reference data is required. As part of the European Space Agency (ESA) Satellite Snow Product Intercomparison and Evaluation Exercise (SnowPEx) efforts, we evaluate 14 gridded SWE products using two reference datasets, airborne gamma radiation SWE and snow-course transect observations. This presentation will consist of two parts 1) prior evaluation of spatial representativeness of airborne gamma radiation SWE measurements through comparisons at different distances from snow course transects across United States and 2) comprehensive performance of 14 gridded SWE products over North America and the Northern Hemisphere.
Exploring alternative possibilities for sublimation measurements over snow and ice surfaces
Hendrik Huwald, Armin Sigmund, Franziska Gerber, Michael Lehning
Corresponding author: Hendrik Huwald
Corresponding author e-mail: hendrik.huwald@epfl.ch
Sublimation of snow is a major depletion mechanism, particularly in dry and windy environments such as high mountains or polar regions. Yet the quantification of the latent heat flux is a difficult task, and both measurements and estimates from models still have large uncertainties. In particular, remote areas in high mountain terrain and on the large polar ice sheets are known to have insufficient spatial coverage through measurements and insufficient capabilities of models to accurately estimate snow sublimation. Additionally, power requirements for eddy covariance (EC) systems are often beyond supply in extreme environments. We present latent heat flux measurements obtained using standard EC instrumentation from several high-Alpine and Antarctic field sites and compute corresponding sublimation rates. Where possible, these quantities are compared to sublimation rates derived from measurements of meteorological variables along a vertical profile (bulk approach). This study further explores the suitability of (low-cost) alternative sensors and instrumentation to determine latent heat fluxes over snow and ice-covered surfaces. Specifically, inexpensive fast-response humidity sensors are in the focus of the approach. If successful, this may lead to denser networks of surface stations in the Alps and potentially in polar regions, capable of measuring sublimation from snow. A similar approach is tested for fast-frequency air temperature measurements, investigating whether such a methodology is viable in typically stably stratified boundary layers over cold surfaces. In its early stages, this project is largely exploratory but may pave the road for interesting and affordable sensing solutions for measuring turbulent heat fluxes in cold, snow- and ice-covered environments as standard components on existing traditional automatic weather stations.
Trends in snow avalanche climatology in the French Alps
Benjamin Reuter, Pascal Hagenmuller, Nicolas Eckert
Corresponding author: Benjamin Reuter
Corresponding author e-mail: reuter@slf.ch
The impact of climate change on snow is manifold and has been demonstrated on snow amounts or on solid precipitation frequency, for instance. As our knowledge on snow avalanche activity derives from observations, we know little about changes at higher elevation where observations are sparse. To study conditions at higher elevation, where most release areas are located, a simulation approach is needed. We demonstrate a recently developed mechanics-based universal simulation approach to create a snow climatology and study changes. The regional snow avalanche climates can be described by occurrence of avalanche type problems. Over the past 62 years avalanche problem occurrences underwent marked changes, in particular after a period between 1988 and 1991, when turning points appear in French Alpine regions. The observed changes refer to an increase in new snow situations in northern regions at high elevation, a general shift of the wet-snow cycles to earlier dates and an overall slight increase of days when natural avalanche can be expected. At lower elevation the known retreat of the snow cover can be confirmed to limit the days with natural release. The presented study is important for understanding ongoing change, but more importantly the methodology we demonstrate paves the road to studying future change.
Double-scale modelling of avalanche release
Olivier Ozenda, Guillaume Chambon, Vincent Richefeu, Pascal Hagenmuller
Corresponding author: Guillaume Chambon
Corresponding author e-mail: guillaume.chambon@inrae.fr
When deforming at large strain rates, the snowpack behaves as a loose, cohesive granular material. In this quasi-brittle regime, recent numerical simulations using the discrete element method (DEM) have highlighted a complex mechanical response, including strong strain-softening and volumetric collapse. Further, this mechanical response appears to be sensitive to the wide variety of microstructural patterns that can be observed in snow. On the other hand, avalanches at the scale of a mountain slope involve large deformations and can propagate over hundreds or thousands of metres. To tackle the challenge of modelling such large scale phenomena, while preserving the possibility of accounting for the influence of snow microstructure on mechanical behavior, a double-scale MPM× DEM approach is proposed. The MPM (material point method) solver, which is used to compute the evolution of the flow at large scale, relies on a homogenized numerical constitutive law stemming from the resolution of numerous individual DEM problems at each material point. Hence, each macroscopic lumping point embeds its own microstructure that can evolve independently. At the micro-scale, a cohesive contact law between particles is considered, whose parameters are adjusted to recover the macroscopic stiffness and strength of the material. This approach is expected to capture the complex mechanical behavior of snow in a more robust manner than any empirical analytical constitutive model. The two-way coupling strategy between the two numerical methods will be explained. Benchmark simulations of column collapses will then be presented to evaluate the approach and discuss the sensitivity to mechanical and numerical parameters. In particular, the evolution of local damage and granular deformations in each DEM cell is monitored for different initial densities and related to macroscopic flow features. Lastly, preliminary results concerning the influence of snow microstructure on avalanche release, using real 3-D microtomographic images of snow samples as inputs for the DEM simulations, will be shown.
Mickaël Lalande, Martin Ménégoz, Gerhard Krinner, Catherine Ottlé
Corresponding author: Mickaël Lalande
Corresponding author e-mail: mickael.lalande@univ-grenoble-alpes.fr
Snow cover strongly modulates the energy fluxes between the atmosphere and the Earth’s surface. Indeed, snow has generally a much higher albedo compared to other surfaces and therefore reduces the amount of solar radiation absorbed by the surface. General circulation models (GCMs) usually compute the snow-cover fraction (SCF) as a diagnostic variable derived from other snow quantities, as for instance, the snow water equivalent (SWE) or the snow depth (SD). The relationship between SWE and SCF varies from simple linear relationships to more advanced parameterizations taking into account further parameters, such as snow density, ground roughness lengths, etc. However, only a few GCMs take into account the topography, while Swenson and Lawrence (2012) highlighted strong differences in snow-cover extent between plains and mountainous areas, which may be explained by the persistence of snow on the summits whereas a faster melting occurs in the valleys. In this study, we designed three new snow parameterizations that include the impact of the sub-grid topography on the SCF in the ORCHIDEE land surface model (LSM) coupled to the LMDZ atmospheric model (part of the French GCM of IPSL). This model shows a strong cold bias and an excess of SCF over the High Mountains of Asia (HMA). The new SCF parameterizations are based on the following existing ones: Swenson and Lawrence (2012; hereafter SL12), Roesch et al. (2001; hereafter R01), and a modified version of Niu and Yang (2007; hereafter NY07). These new parameterizations were calibrated over HMA using a high-resolution snow reanalysis, and compared to a deep learning model trained on the reanalysis dataset. The calibrated parameterizations SL12, R01, and the modified version of NY07 were then tested in coupled ORCHIDEE/LMDZ simulations. Preliminary results show improvements in simulated snow cover in HMA but slight deterioration in other areas depending on the model resolution. They suggest also that calibration should be extended to other snow-covered areas and should include other parameters such as the type of vegetation in particular.
Moritz Oberrauch, Bertrand Cluzet, Jan Mangusson, Tobias Jonas
Corresponding author: Moritz Oberrauch
Corresponding author e-mail: moritz.oberrauch@slf.ch
On one hand, models provide continuous estimates over space and time, but they inherently introduce errors given the simplifications made in their implementation and any uncertainties in the input data. On the other hand, observations are stationary in either space or time. Data assimilation combines the two sources of information into the best estimate. In this work we make use of the information provided by a dense snow-station network across Switzerland to improve upon estimates of snow height and snow water equivalent computed by a spatially distributed setup of the Flexible Snow Model (FSM) for the entire country. This is done in a two step approach of (a) using a particle filter to assimilate snow monitoring observations from 450 stations and (b) distributing the gained information in space to non-observed grid points. The particle filter combines ensemble simulations at a point scale with observations. This allows the weighting of individual ensemble members based on their proximity to the observed value following Bayesian statistics, and provide a weighted ensemble average as best estimate. Since each ensemble member is a ‘closed system’, this approach does not introduce any physical inconsistency into the modeling chain. To improve the particle filter’s performance we test different assimilation frequencies, as well as different methods of ensemble generation (stochastic perturbation of meteorological input, ensemble input from numerical weather prediction models and different snow-model parameters). Knowledge on the spatial and temporal error structure of the model ensemble gained from the point runs is then used to come up with a spatialization strategy. This includes testing the performance of different localization radii, i.e., including different numbers of neighboring observations. The incorporation of the particle filter shows promising results, both from an operational perspective by having a more accurate forecast, and from a scientific perspective by identifying model shortcomings and error/compensation patterns.
Giulia Mazzotti, Jari-Pekka Nousu, Tobias Jonas, Clare Webster, Matthieu Lafaysse
Corresponding author: Giulia Mazzotti
Corresponding author e-mail: giulia.mazzotti@slf.ch
Rising temperatures and forest disturbances are causing rapid change in seasonally snow-covered forest environments. Accurate predictions of forest snow are thus relevant for a variety of disciplines such as biogeochemistry, ecohydrology and climate sciences. Research in each of these fields relies on process-based models that are usually discipline-specific, e.g. snow hydrology and land-surface models. These models include canopy and snowpack process representations of varying complexity. It remains largely unexplored how such model differences affect the simulated exchange of water end energy between the forest-snow system, the atmosphere and the ground. We present simulations generated with alternative model configurations at subAlpine and boreal forested sites, combining concepts from two established forest-snow models: 1) FSM2, an intermediate-complexity snow model with explicit representation of canopy structure heterogeneity, developed for meter-scale snow-hydrological applications; and 2) MEB-Crocus, which comprises a complex snow physics scheme and a big-leaf-type canopy representation and constitutes the most sophisticated snow–vegetation scheme within the SURFEX land surface modelling system. Contrasting results from these configurations allows us to assess how effective land-surface properties (e.g. temperature, albedo) and ground thermal regimes are influenced by the representation of canopy structure heterogeneity and internal snowpack processes, and whether the relevance of these representations varies with forest structure and climatic setting. Understanding these dependencies is an important step towards tailoring model complexity to specific applications. Our work thus aims to inform and further promote the use of process-based modelling tools in interdisciplinary ecosystem research, and in support of environmental change impact studies, management practices and mitigation/adaptation strategies.
Michael Neuhauser, Anselm Köhler, Anna Wirbel, Jan Frederik Höller, Wolfgang Fellin, Jan-Thomas Fischer
Corresponding author: Anselm Köhler
Corresponding author e-mail: anselm.koehler@bfw.gv.at
A non-invasive in-flow measurement device that tracks particle motion within snow avalanches is currently being developed. These so-called AvaNodes reveal the internal dynamic processes that determine the flow mobility of avalanches and run-out behavior as well as the influence of snow and particle properties, such as temperature or density. Our current understanding of flow dynamics on a particle level is mostly based on laboratory experiments and computer simulations. Thus, the tracking of particle motions and the respective flow trajectories within field-scale avalanches is believed to increase the understanding on processes such as segregation, phase separation and eventually flow regimes evolution and transitions. Here, we combine the different measurement approaches of radar remote sensing and in-flow sensor nodes with numeric simulations of the avalanche by comparing tracked particle trajectories with their simulated counterparts. For that, we identify the optimal model parameters by fitting to the mGeodar radar data that offers the possibility to evaluate numerical simulation tools with respect to the temporal evolution of the flow variables. These radar data are typically displayed in a range-time diagram along the antenna line of sight and the distance to the approaching avalanche front is extracted for each time step. Transforming the computed avalanche front into the radar coordinate system yields a synthetic range–time radargram that facilitates a direct comparison. From a probabilistic ensemble of simulations covering multiple dimensions of the parameter space, the best-fitted numerical simulation is identified for a detailed comparison with the real-world data from the AvaNodes. The thickness-integrated model com1dfa implemented in the simulation toolbox AvaFrame produces surface parallel trajectories and utilizes a particle-grid solver that allows direct tracking of discretized numerical avalanche particles. By combining all data, we show difficulties and benefits of the sensor fusion from remote radar sensing and in-flow measurements. We compare the simulated and measured particle trajectories and find an overall reasonable agreement during the evolution of the flow. However, the acceleration during the avalanche release and the deceleration of the AvaNode sensors is not exactly reproduced by the simulations, which indicates that certain flow dynamic aspects and processes are not well represented in the underlying flow models.
Using backscattered laser signal intensity to measure snow surface properties automatically
Alexander Prokop
Corresponding author: Alexander Prokop
Corresponding author e-mail: alex@snowscan.at
Knowledge about the condition of the snow surface right before snowfall is important for avalanche forecasting. Weak layers that form on the snow surface such as surface hoar can pose a serious hazard when loaded with additional snow on top. In this context standard 1-D laser measurement devices are applied that measure automatically snow depth by sending a laser signal that is back-scattered by the snow surface and received by the sensor. While the snow depth is measured via the time-of-flight of the laser signal, the sensor that is mounted on standard meteorological measurement stations records also the intensity of the back-scattered signal. Next to other parameters the intensity is dependent on surface properties such as snow crystal type and size, and the moisture content of the snow. Starting in winter 2019, time series of snow-depth and received signal intensity values were collected and analyzed against traditional manual snow stratigraphy investigations that verify snow- crystal type and size and moisture content. In this work a description of the theoretical background, the system implementation, the field installation, realization of tests and an investigation of the recorded data is presented. Analysis of the robust dataset of three winters allows the classification of weak layers forming and disappearing on the snow surface and measurement of the snow moisture content. With the help of automatic snow surface- and air-temperature measurements it is possible to determine snow-surface properties with a high degree of confidence, which is important not only for avalanche forecasting but also for, for example, snow hydrology and the verification of satellite remote-sensed data of the snow surface.
Natural convection in snow: when it occurs and how it influences snow properties
Mahdi Jafari, Michael Lehning
Corresponding author: Mahdi Jafari
Corresponding author e-mail: mahdi.jafari@epfl.ch
Convection in seasonal snow cover has been hypothesized to importantly influence snow cover, in particular in the Arctic. However, direct observations of convection in snow have not been made to date and convection is not considered in current snow models. Our knowledge about convection in snow comes from observations of secondary parameters such as temperature and density distribution (in time and space). Using the recent development of a numerical implementation in OpenFOAM that allows the simulation of natural convection in two-dimensional (2-D) snow cover, including the effect on snow density, we present simulations of conditions that lead to strong effects of convection on snow properties. We show that initial heterogenous density distribution, in particular low-density patches close to the ground in seasonal snow cover, enhances the onset and magnitude of natural convection. Such patches are typically caused by shrubs or rocks. Coupling the 2-D convection scheme to the snow model SNOWPACK, we further discuss how natural convection changes typical snow cover from Arctic environments and influences density profiles and layering. The results qualitatively confirm observations of snow properties. We conclude that models of thin Arctic snow cover need to consider the effect of convection.
Detection and evaluation of snow surface properties with a 24GhZ FMCW radar
Jean-Benoit Madore, Alexandre Langlois
Corresponding author: Jean-Benoit Madore
Corresponding author e-mail: jean-benoit.madore@usherbrooke.ca
Forecasting avalanche danger in mountainous areas where data observations are sparse is challenging, especially in a climate change reality. Changes in seasonal patterns and augmentation in extreme events underline the need to have new tools to evaluate snow properties throughout the season. Knowing the snow surface state (i.e. dry, moist, wet or refrozen) is of the utmost importance for forecasters and mountain users since it affects avalanche risk and safety decisions while traveling. Here, we present a 24 GhZ FMWC radar in a fixed position that evaluated the snow continuously (15 min timestamp) throughout the spring of 2019 at the Fidelity station, in Glacier National Park, Canada. A 2-month field campaign was conducted from March to May 2019 to evaluate snow layers and surface properties. Snow surface was detected by radar, and automated snow-height measurement was elaborated from the radar signal. SWE evaluation gave satisfactory results during the winter months (January–March) using a previously developed algorithm for that instrument on thinner snowpack. Spring melt/freeze diurnal events were identified with signal lost at maximum surface wetness. The depth of surface refreezing was measured by the instrument and compared with field measurements. This radar setup showed great potential for multiple snow properties that could be used by snow practicians as well as snow modeling studies.
Snow interception processes and prediction in a Windy subalpine environment
Alex Cebulski, John Pomeroy
Corresponding author: Alex Cebulski
Corresponding author e-mail: alex.cebulski@usask.ca
Snowmelt from mountains provides an important freshwater resource and vital headwaters to many rivers across the globe. Ablation of intercepted snow through sublimation, unloading, wind transport and meltwater drip exerts a strong control on the spatial and temporal variability of snow accumulation and melt in forests. This research aims to investigate the influence of redistribution, sublimation and unloading processes in a windy subalpine environment. New observation data has been collected for a forested site in the Fortress Mountain Research Basin (FMRB), Alberta, Canada (115° W, 51° N, 2100 m a.s.l.). Snow interception was quantified using combination of weighed tree lysimeter observations, snow water equivalent (SWE) measurements from station instruments, manual snow surveys and pre- and post-storm airborne laser scanning (lidar)-derived snow depth. Snow unloading and melt of intercepted snow was measured using sub-canopy snow scales, tipping-bucket rain gauges, and time-lapse cameras. Sublimation of intercepted snow was assessed using eddy covariance systems located above and below the forest canopy. Wind-induced redistribution was quantified using a disdrometer. The role of various snow ablation processes on intercepted snow was quantified towards developing models appropriate for cold and windy mountain environments.
How much snow sublimates in extreme environments: quantifying blowing snow sublimation
Michael Lehning, Franziska Gerber, Guang Li, Armin Sigmund, Daniela Brito Melo, Varun Sharma, Hendrik Huwald
Corresponding author: Michael Lehning
Corresponding author e-mail: lehning@slf.ch
The mass balance of snow and ice has large uncertainties in high mountains and polar regions. While snow accumulation on the ground can be measured with increasing accuracy and improved spatial and temporal coverage, losses through melt and sublimation remain poorly constrained. When sublimation is the main ablation mechanism, such as in inner Antarctica, the uncertainty in the mass balance is particularly high as current models and observations disagree on the magnitude of the effect. We present simulations with the new CRYOWRF model, which has an improved snow representation based on SNOWPACK and a drifting and blowing snow sublimation routine. CRYOWRF more reliably reproduces (scarce) observations compared to previous models. We demonstrate, how model resolution influences the results and present simulations across Antarctica, High Mountain Asia and the European Alps. The results suggest that sublimation has typically been underestimated in previous models, partly because near-surface processes such as adiabatic heating in downdrafts and turbulent mixing need to be adequately resolved. The results are not only important for a better understanding of the surface mass balance of snow and ice but also serve as a bridge to test model precipitation estimates against snow accumulation measurements. This is seen as a major contribution to the improvement of precipitation models in extreme environments.
Assessing avalanche problems for operational avalanche forecasting based on different model chains
Martin Perfler, Michael Binder, Ben Reuter, Christoph Mitterer
Corresponding author: Christoph Mitterer
Corresponding author e-mail: chris.mitterer@tirol.gv.at
Snow-cover models were originally developed to assist operational avalanche forecasting. Up to now, it has mainly been researchers who have used the models while for forecasting services great potential remains to be unlocked. To advance snow-cover modelling for forecasting, the models needs to tie in with traditional forecasting strategies, which typically build on observations and measurements. To that end, an algorithm for identifying avalanche problems from snow-cover simulations was developed. In a first step, it was used to create a snow avalanche climatology based on the simulated snow instabilities. Now we have adapted the existing algorithm to provide an objective first guess for avalanche problems in forecasting regions based on the available data. Our algorithm combines observed snow profiles or measured weather data with numerical weather prediction data to provide a short-term forecast of avalanche problems, which are the common starting point to assess regional avalanche danger in many forecasting services. To meet the needs of operational avalanche forecasting we have refined the assessment of wet snow and wind slab problems by including slope simulations for different aspects. We also show how simulations can be initialized from observed snow profiles to update the avalanche problem identification in the algorithm. Moreover, by switching input data, i.e. using observations, measurements or model data, to initialize the short-term forecast we can assess the uncertainty of automatically identified avalanche problems in contrast to forecast avalanche problems. The enhancements we present streamline traditional forecasting with the modelling world and pave the way for early integration of automatic avalanche problem identification into operational snow cover simulations.
Influence of the snow cover and test persons on snow stability tests
Silke Griesser, Christine Pielmeier, Ingrid Reiweger
Corresponding author: Silke Griesser
Corresponding author e-mail: silke.griesser@students.boku.ac.at
The compression test (CT) is a mechanical test within the snow cover to assess fracture initiation and hence obtain an index of snow-cover instability. The CT is used by both scientists and practitioners, as it is fast and easy to perform and needs no additional equipment besides the avalanche shovel and a cord or snow saw. During a CT an isolated column of snow is exposed to increasing stresses by placing a shovel blade on the snow surface and tapping on the blade manually. The test result is the number of taps it takes to produce a fracture within a weak layer or weak interface in the snow column. Naturally those test results vary, even for similar weak layers, as the stress transmitted to the weak layer from the tapping depends on the depth of the weak layer, the energy dissipation in the snow layers above the weak layer, as well as on the test person (force applied, shovel placement). To assess the influence of the depth of the weak layer, the snow cover above the weak layer, and the variations from different test persons on CT results, we performed force measurements during the CTs. The device was placed at defined snow depths in varying snow covers and during experiments with numerous different test persons. Preliminary results show that changing the test person has the strongest influence on CT results, particularly during high-loading steps. Yet, during low-loading steps, which are crucial to observe snow-cover instability, the variations due to different test persons were smaller. As expected, the force measured due to tapping decreased with depth of the measuring device We also observed a decrease of force variations with increasing depth. Measurements in the snow cover showed an increase of force for consecutive taps with the same kind of loading, presumably due to snow compaction due to tapping. The study provides valuable insight into possible variations and reproducibility of CT results.
The value of synthetic high-resolution daily snow-cover maps for hydrological modeling
Pau Wiersma, Fatemeh Zakeri, Grégoire Mariéthoz
Corresponding author: Pau Wiersma
Corresponding author e-mail: pau.wiersma@unil.ch
To study long-term changes in the hydrology of snow-fed catchments, there is a need for long-term time series of snow-cover data. Satellite imagery and climate reanalysis can both be used to quantify past snow cover, but they lack baseline period length and spatial resolution respectively. In this study, we apply a similarity-based method to generate synthetic high-resolution daily snow-cover maps, and then combine these maps with hydrological model outputs in a data-assimilation framework. The results are benchmarked against a case without synthetic snow-cover forcing and a case with only observed snow-cover maps. The study is performed on the Thur catchment in eastern Switzerland, a meso-scale catchment covering a wide elevation range and experiencing different degrees of snow intermittency. With the currently implemented direct insertion data assimilation, the streamflow simulation performance slightly deteriorates but the SWE estimates improve significantly. To further explore the robustness of the methodology and the validity of these findings we will consider other data-assimilation algorithms and hydrological models in the near future.
Synthesizing daily 30 m snow-cover maps based on climate and satellite data
Fatemeh Zakeri, Gregoire Mariethoz
Corresponding author: Fatemeh Zakeri
Corresponding author e-mail: fatemeh.zakeri@unil.ch
Knowledge of the snow-covered area as a rapidly changing process is essential for studying the timing and duration of the melting season, snowmelt run-off, etc. Remote sensing is a valuable tool for observing snow-cover changes. However, satellite data are restricted due to clouds, cloud shadows and revising time. Moreover, due to the trade-offs among temporal and spatial resolutions, accessing high temporal and spatial resolution snow-cover observation is still a challenge. On the other hand, when there is a resemblance between two states of the atmosphere, the meteorological phenomena associated with them also resemble each other. Accordingly, snow patterns from previous years could be repeatable in many regions between years with similar meteorological phenomena. Based on these repeatable patterns, this study proposes a methodology to estimate daily, 30 m-resolution snow-cover maps based on Landsat, Sentinel-2, and climate data sets and auxiliary data such as MODIS if available. The proposed method has two main steps: selection and estimation. In the selection step, for each day that there is no data available, called a test day, it computes a similarity metric between the test day and the training day for which the clear-sky satellite data is available. Then select k best training days, and estimate the test day’s snow cover based on the k selected images using the mode operation for categorical and the average operation for continuous data such as normalized difference snow index. The proposed method is tested based on 11 km global reanalyzed climate data (ERA5) and 1 km MeteoSwiss data. The proposed methodology was applied to the Jonschwil sub-basin in Switzerland and to a subset of the Alpine belt called the western Swiss Alps. The estimated snow-cover maps have been generated for 20 years since 2000. The leave-one-out statistical assessment, comparison with MODIS data, comparison with PlanetScope images, comparison with in situ snow depth data, and comparison with the degree-day snow model have been used to analyze the results. The results indicate that the proposed method has the ability to generate daily snow-cover maps with good agreement with the actual snow cover maps, even using 11 km ERA5 climate data. This highlights the potential of the method to be used in other areas using freely available data.
Antoine Rolland, Jean-Benoit Madore, Alexandre Langlois
Corresponding author: Antoine Rolland
Corresponding author e-mail: antoine.rolland@usherbrooke.ca
A multitude of factors can affect the structure of snowpack in Alpine environments such as weather conditions, topography and forest density. Weather conditions are the most important because they govern the formation of weak layers within the snowpack. Indeed, a weak layer can develop in the presence of kinetic metamorphism, surface hoar or ice crusts. With climate change, the consequential increase in temperatures leads to a change in the precipitation phase. With this change in patterns, it is expected that Alpine environments will see a decrease in snow cover, particularly at lower elevations, and an increase of rain-on-snow events (ROS) creating ice crusts in the snowpack. These ice crusts open the way for the formation of a weak layer, thus increasing the instability of the snow cover as well as the risk of avalanches. In recent years, the number of visits to avalanche terrain has been steadily increasing as the popularity of backcountry sports continues to grow, thus increasing the level of risk. With manual avalanche risk assessment methods being very localized, it becomes difficult to meet the increasing demand to assess the stability of the snowpack over large areas. Therefore, it becomes imperative to develop new methods of quantifying the microstructure and instability of the snow cover, deployable at the slope scale that would be quicker to use (compared to existing methods) and less expensive. As such, with advances in technologies such as frequency-modulated continuous wave (FMCW) radar in recent years, it is now possible to assess snow-cover structure. More specifically, the 24 GHz FMCW radar is a compact, inexpensive tool that can non-destructively extract information on snowpack stratigraphy. This type of sensor could overcome the problem of lack of observations due to the complexity and duration of a traditional measurement (e.g. snowpit) and become an important factor in decision making for agencies producing avalanche bulletins.
Matthias Tonnel, Anna Wirbel, Felix Oesterle, Anselm Koehler, Jan-Thomas Fischer
Corresponding author: Jan-Thomas Fischer
Corresponding author e-mail: JT.Fischer@bfw.gv.at
Thickness-integrated flow models are well established in operational avalanche simulation tools, scientific investigations and academic education. Testing avalanche simulation tools is a challenging but crucial step. Ready-to-use tools for model verification and validation and corresponding testing remain rare. Within AvaFrame – the open avalanche framework – we try to address this by providing both model verification tests that facilitate analytical solutions for the implementation of the flow model, as well as tools to perform simulation evaluation to check if the model applies to the problem at hand. We introduce two different testing approaches to verify the correct implementation of the thickness-integrated flow model and the convergence and robustness of the numerical model code. The testing approaches are based on analytical solutions and the investigation of total energy balance. The flow variable tests allow us to investigate the local spatio-temporal evolution of flow thickness and velocity. In contrast, the energy line test focuses on the accuracy of the global kinetic energy (velocity) along the path and the corresponding centre of mass run out. We also explore and explain the limitations of these approaches. We present the Thalweg-Time diagram to evaluate flow-model behaviour and to visualize the temporal evolution of flow variables. It allows flow evolution to be investigated at a glance by displaying flow thickness or velocity in a coordinate system along the avalanche path. This presentation highlights AvaFrame’s environment for model testing and result visualization. Using its thickness-integrated, computational dense flow module (com1DFA), the testing workflow and the corresponding results are demonstrated. We invite everybody to apply this workflow to other avalanche simulation tools and contribute to developing existing or new tests.
Rebecca Mott, Bert Kruyt, Joel Fiddes, Varun Sharma, Franziska Gerber, Dylan Reynolds
Corresponding author: Rebecca Mott
Corresponding author e-mail: mott@slf.ch
We compare three models of different complexity in their ability to downscale and represent wind and snowfall patterns at the ridge- and slope-scale. We present a methodical comparison of three downscaling methods of varying complexity that are used to downscale data from the numerical weather prediction model COSMO-1 at 1.1 km horizontal resolution to 250 and 50 m in highly complex terrain. We compare WRF, a dynamical atmospheric model; ICAR, a model of intermediate complexity; and TopoSCALE, an efficient topography-based downscaling scheme. Point-scale validation at meteorological stations shows similar results for all three models. Spatial snow deposition patterns are validated against lidar data and indicate that WRF is able to capture preferential deposition of snow, while ICAR shows a weak signal. Qualitative comparison of 3-D ridge–flow interactions shows reasonable agreement between ICAR and WRF at 250 m resolution, yet at 50 m resolution WRF simulates complex flow patterns that ICAR cannot reproduce. Based on these findings and the significant reduction in computational costs, ICAR is a cost-efficient alternative to WRF at the 250 m resolution. TopoScale performs very well in point-scale comparisons, but it is unclear if this can be attributed to the model itself or to the forcing data and the observations assimilated therein. These finding motivated the recent development of a new model variant of the ICAR model (HICAR) which will considerably improve the model performance at high resolution and in complex terrain.
Expansion of operational Airborne Snow Observatories programs through tech transfer from NASA
Thomas Painter, Jeffrey Deems, Kat Bormann
Corresponding author: Thomas Painter
Corresponding author e-mail: painter@airbornesnowobservatories.com
The spatial distribution of snow in mountain basins exerts a first-order control on runoff timing and volume, with direct influence on other snow-dependent mountain systems. Conventional snow monitoring and modeling methods rely on indirect extrapolations or assumptions of spatial relationships, and as such are unable to accurately depict the spatiotemporal variations in snow depth or water equivalent. Since 2013, the Airborne Snow Observatory (ASO) – a coupled scanning lidar and imaging spectrometer system with integrated physical modeling – has produced measurements of snow depth, snow water equivalent, and snow albedo across full mountain basins on operational time scales. In turn, the ASO data have driven marked increases in the accuracy of snowmelt runoff forecasting in a wide range of mountain systems. The ASO program commenced at the NASA Jet Propulsion Laboratory (JPL) in 2013 and grew to implementations in the western USA and Switzerland, providing unprecedented data for science and water management. In 2019 NASA supported a successful technology transfer to the public-benefit Airborne Snow Observatories, Inc., extending the reach beyond the capacity of a research institute such as JPL. ASO data sets have shown critical improvements in snowmelt runoff forecasting and snow–ecosystem interactions. The ASO program has now become a part of the State of California water measurement infrastructure and is growing throughout the western USA and internationally. While the remote sensing and modeling products are focused primarily on water management and utility customers, the freely available data remain deeply relevant for scientific investigations on snow physics, hydrology, glaciology, ecosystem function, and geology. In this presentation, we present an overview of ASO, Inc. data, activities to date, and upcoming implementations.