Articles | Volume 18, issue 22
https://doi.org/10.5194/gmd-18-8751-2025
© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
https://doi.org/10.5194/gmd-18-8751-2025
© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Development and technical paper
| 
20 Nov 2025
Development and technical paper |
| 20 Nov 2025
| 20 Nov 2025
Autoencoder-based feature extraction for the automatic detection of snow avalanches in seismic data
WSL Institute for Snow and Avalanche Research SLF, ETH Zurich, Davos, Switzerland
Cristina Pérez-Guillén
WSL Institute for Snow and Avalanche Research SLF, ETH Zurich, Davos, Switzerland
Michele Volpi
Swiss Data Science Center, ETH Zurich and EPFL, Zurich, Switzerland
Christine Seupel
WSL Institute for Snow and Avalanche Research SLF, ETH Zurich, Davos, Switzerland
Alec van Herwijnen
WSL Institute for Snow and Avalanche Research SLF, ETH Zurich, Davos, Switzerland
Related authors
Pascal Edme, Cristina Pérez-Guillén, Alec van Herwijnen, Johannes Aichele, Andri Simeon, Patrick Paitz, Fabian Walter, and Andreas Fichtner
EGUsphere, https://doi.org/10.5194/egusphere-2026-2373, https://doi.org/10.5194/egusphere-2026-2373, 2026
This preprint is open for discussion and under review for Natural Hazards and Earth System Sciences (NHESS).
Short summary
Short summary
We present a near real-time snow avalanche detection method using Distributed Acoustic Sensing (DAS) on a ~10 km telecom cable following a road pass in the Swiss Alps. Avalanches are discriminated from vehicle traffic through distinct dual frequency spatio-temporal signatures. The proposed workflow achieved low false alert rates and detected 73 avalanches over one winter, demonstrating robust and cost-effective monitoring potential.
Pascal Edme, Cristina Pérez-Guillén, Alec van Herwijnen, Johannes Aichele, Andri Simeon, Patrick Paitz, Fabian Walter, and Andreas Fichtner
EGUsphere, https://doi.org/10.5194/egusphere-2026-2373, https://doi.org/10.5194/egusphere-2026-2373, 2026
This preprint is open for discussion and under review for Natural Hazards and Earth System Sciences (NHESS).
Short summary
Short summary
We present a near real-time snow avalanche detection method using Distributed Acoustic Sensing (DAS) on a ~10 km telecom cable following a road pass in the Swiss Alps. Avalanches are discriminated from vehicle traffic through distinct dual frequency spatio-temporal signatures. The proposed workflow achieved low false alert rates and detected 73 avalanches over one winter, demonstrating robust and cost-effective monitoring potential.
Gwendolyn Dasser, Alessandro Maissen, Michele Volpi, Jordan Aaron, Florian Denzinger, Hugo Raetzo, and Andrea Manconi
EGUsphere, https://doi.org/10.5194/egusphere-2026-375, https://doi.org/10.5194/egusphere-2026-375, 2026
Short summary
Short summary
We aim to improve early landslide detection and monitoring by automating signal segmentation of Sentinel-1 radar interferograms. We first compared how different experts map mass-movement related signals on interferograms using ten case studies, identifying dataset uncertainties and limitations. We then trained and evaluated deep learning approaches to perform segmentation. Finally showing that models can achieve results comparable to expert variability while improving efficiency.
Michael Lombardo, Amelie Fees, Anders Kaestner, Alec van Herwijnen, Jürg Schweizer, and Peter Lehmann
The Cryosphere, 19, 4437–4458, https://doi.org/10.5194/tc-19-4437-2025, https://doi.org/10.5194/tc-19-4437-2025, 2025
Short summary
Short summary
Water flow in snow is important for many applications including snow hydrology and avalanche forecasting. This work investigated the role of capillary forces at the soil-snow interface during capillary rise experiments using neutron radiography. The results showed that the properties of both the snow and the transitional layer below the snow affected the water flow. This work will allow for better representations of water flow across the soil–snow interface in snowpack models.
Francois Kamper, Fabian Walter, Patrick Paitz, Matthias Meyer, Michele Volpi, and Mathieu Salzmann
EGUsphere, https://doi.org/10.5194/egusphere-2025-3864, https://doi.org/10.5194/egusphere-2025-3864, 2025
Short summary
Short summary
We use anomaly detection to automatically find patterns in seismic data that may signal dangerous mass-movement events such as landslides, glacier collapses, or debris flows. Because such movements are rare, our approach reduces the amount of data that must be analyzed to find them, whether by experts or clustering procedures. We demonstrate the usefulness of our approach by mining for mass movements in Switzerland and Greenland.
Grégoire Bobillier, Bertil Trottet, Bastian Bergfeld, Ron Simenhois, Alec van Herwijnen, Jürg Schweizer, and Johan Gaume
Nat. Hazards Earth Syst. Sci., 25, 2215–2223, https://doi.org/10.5194/nhess-25-2215-2025, https://doi.org/10.5194/nhess-25-2215-2025, 2025
Short summary
Short summary
Our study investigates the initiation of snow slab avalanches. Combining experimental data with numerical simulations, we show that on gentle slopes, cracks form and propagate due to compressive fractures within a weak layer. On steeper slopes, crack velocity can increase dramatically after approximately 5 m due to a fracture mode transition from compression to shear. Understanding these dynamics provides a crucial missing piece in the puzzle of dry-snow slab avalanche formation.
Jakob Boyd Pernov, William H. Aeberhard, Michele Volpi, Eliza Harris, Benjamin Hohermuth, Sakiko Ishino, Ragnhild B. Skeie, Stephan Henne, Ulas Im, Patricia K. Quinn, Lucia M. Upchurch, and Julia Schmale
Atmos. Chem. Phys., 25, 6497–6537, https://doi.org/10.5194/acp-25-6497-2025, https://doi.org/10.5194/acp-25-6497-2025, 2025
Short summary
Short summary
Particulate methanesulfonic acid (MSAp) is vital for the Arctic climate system. Numerical models struggle to reproduce the MSAp seasonal cycle. We evaluate three numerical models and one reanalysis product’s ability to simulate MSAp. We develop data-driven models for MSAp at four Arctic stations. The data-driven models outperform the numerical models and reanalysis product and identified precursor source-, chemical-processing-, and removal-related features as being important for modeling MSAp.
Philipp L. Rosendahl, Johannes Schneider, Grégoire Bobillier, Florian Rheinschmidt, Bastian Bergfeld, Alec van Herwijnen, and Philipp Weißgraeber
Nat. Hazards Earth Syst. Sci., 25, 1975–1991, https://doi.org/10.5194/nhess-25-1975-2025, https://doi.org/10.5194/nhess-25-1975-2025, 2025
Short summary
Short summary
Avalanche formation depends on crack propagation in weak snow layers, but the conditions that stop a crack remain unclear. We show that slab touchdown reduces the energy driving crack growth, which can halt propagation even under static conditions. This suggests that crack arrest is influenced not only by snowpack variability or dynamics but also by mechanical interactions within the snowpack. Our findings refine avalanche prediction models and improve hazard assessment.
Cristina Pérez-Guillén, Frank Techel, Michele Volpi, and Alec van Herwijnen
Nat. Hazards Earth Syst. Sci., 25, 1331–1351, https://doi.org/10.5194/nhess-25-1331-2025, https://doi.org/10.5194/nhess-25-1331-2025, 2025
Short summary
Short summary
This study assesses the performance and explainability of a random-forest classifier for predicting dry-snow avalanche danger levels during initial live testing. The model achieved ∼ 70 % agreement with human forecasts, performing equally well in nowcast and forecast modes, while capturing the temporal dynamics of avalanche forecasting. The explainability approach enhances the transparency of the model's decision-making process, providing a valuable tool for operational avalanche forecasting.
Amelie Fees, Michael Lombardo, Alec van Herwijnen, Peter Lehmann, and Jürg Schweizer
The Cryosphere, 19, 1453–1468, https://doi.org/10.5194/tc-19-1453-2025, https://doi.org/10.5194/tc-19-1453-2025, 2025
Short summary
Short summary
Glide-snow avalanches release at the soil–snow interface due to a loss of friction, which is suspected to be linked to interfacial water. The importance of the interfacial water was investigated with a spatio-temporal monitoring setup for soil and local snow on an avalanche-prone slope. Seven glide-snow avalanches were released on the monitoring grid (winter seasons 2021/22 to 2023/24) and provided insights into the source, quantity, and spatial distribution of interfacial water before avalanche release.
Jan Svoboda, Marc Ruesch, David Liechti, Corinne Jones, Michele Volpi, Michael Zehnder, and Jürg Schweizer
Geosci. Model Dev., 18, 1829–1849, https://doi.org/10.5194/gmd-18-1829-2025, https://doi.org/10.5194/gmd-18-1829-2025, 2025
Short summary
Short summary
Accurately measuring snow height is key for modeling approaches in climate science, snow hydrology, and avalanche forecasting. Erroneous snow height measurements often occur when snow height is low or changes, for instance during snowfall in summer. We prepare a new benchmark dataset with annotated snow height data and demonstrate how to improve the measurement quality using modern deep-learning approaches. Our approach can be easily implemented in a data pipeline for snow modeling.
Bastian Bergfeld, Karl W. Birkeland, Valentin Adam, Philipp L. Rosendahl, and Alec van Herwijnen
Nat. Hazards Earth Syst. Sci., 25, 321–334, https://doi.org/10.5194/nhess-25-321-2025, https://doi.org/10.5194/nhess-25-321-2025, 2025
Short summary
Short summary
To release a slab avalanche, a crack in a weak snow layer beneath a cohesive slab has to propagate. Information on that is essential for assessing avalanche risk. In the field, information can be gathered with the propagation saw test (PST). However, there are different standards on how to cut the PST. In this study, we experimentally investigate the effect of these different column geometries and provide models to correct for imprecise field test geometry effects on the critical cut length.
Stephanie Mayer, Martin Hendrick, Adrien Michel, Bettina Richter, Jürg Schweizer, Heini Wernli, and Alec van Herwijnen
The Cryosphere, 18, 5495–5517, https://doi.org/10.5194/tc-18-5495-2024, https://doi.org/10.5194/tc-18-5495-2024, 2024
Short summary
Short summary
Understanding the impact of climate change on snow avalanche activity is crucial for safeguarding lives and infrastructure. Here, we project changes in avalanche activity in the Swiss Alps throughout the 21st century. Our findings reveal elevation-dependent patterns of change, indicating a decrease in dry-snow avalanches alongside an increase in wet-snow avalanches at elevations above the current treeline. These results underscore the necessity to revisit measures for avalanche risk mitigation.
Alessandro Maissen, Frank Techel, and Michele Volpi
Geosci. Model Dev., 17, 7569–7593, https://doi.org/10.5194/gmd-17-7569-2024, https://doi.org/10.5194/gmd-17-7569-2024, 2024
Short summary
Short summary
By harnessing AI models, this work enables processing large amounts of data, including weather conditions, snowpack characteristics, and historical avalanche data, to predict human-like avalanche forecasts in Switzerland. Our proposed model can significantly assist avalanche forecasters in their decision-making process, thereby facilitating more efficient and accurate predictions crucial for ensuring safety in Switzerland's avalanche-prone regions.
Amelie Fees, Alec van Herwijnen, Michael Lombardo, Jürg Schweizer, and Peter Lehmann
Nat. Hazards Earth Syst. Sci., 24, 3387–3400, https://doi.org/10.5194/nhess-24-3387-2024, https://doi.org/10.5194/nhess-24-3387-2024, 2024
Short summary
Short summary
Glide-snow avalanches release at the ground–snow interface, and their release process is poorly understood. To investigate the influence of spatial variability (snowpack and basal friction) on avalanche release, we developed a 3D, mechanical, threshold-based model that reproduces an observed release area distribution. A sensitivity analysis showed that the distribution was mostly influenced by the basal friction uniformity, while the variations in snowpack properties had little influence.
Gwendolyn Dasser, Valentin T. Bickel, Marius Rüetschi, Mylène Jacquemart, Mathias Bavay, Elisabeth D. Hafner, Alec van Herwijnen, and Andrea Manconi
EGUsphere, https://doi.org/10.5194/egusphere-2024-1510, https://doi.org/10.5194/egusphere-2024-1510, 2024
Short summary
Short summary
Understanding snowpack wetness is crucial for predicting wet snow avalanches, but detailed data is often limited to certain locations. Using satellite radar, we monitor snow wetness spatially continuously. By combining different radar tracks from Sentinel-1, we improved spatial resolution and tracked snow wetness over several seasons. Our results indicate higher snow wetness to correlate with increased wet snow avalanche activity, suggesting our method can help identify potential risk areas.
Stephanie Mayer, Frank Techel, Jürg Schweizer, and Alec van Herwijnen
Nat. Hazards Earth Syst. Sci., 23, 3445–3465, https://doi.org/10.5194/nhess-23-3445-2023, https://doi.org/10.5194/nhess-23-3445-2023, 2023
Short summary
Short summary
We present statistical models to estimate the probability for natural dry-snow avalanche release and avalanche size based on the simulated layering of the snowpack. The benefit of these models is demonstrated in comparison with benchmark models based on the amount of new snow. From the validation with data sets of quality-controlled avalanche observations and danger levels, we conclude that these models may be valuable tools to support forecasting natural dry-snow avalanche activity.
Mathieu Le Breton, Éric Larose, Laurent Baillet, Yves Lejeune, and Alec van Herwijnen
The Cryosphere, 17, 3137–3156, https://doi.org/10.5194/tc-17-3137-2023, https://doi.org/10.5194/tc-17-3137-2023, 2023
Short summary
Short summary
We monitor the amount of snow on the ground using passive radiofrequency identification (RFID) tags. These small and inexpensive tags are wirelessly read by a stationary reader placed above the snowpack. Variations in the radiofrequency phase delay accurately reflect variations in snow amount, known as snow water equivalent. Additionally, each tag is equipped with a sensor that monitors the snow temperature.
Bastian Bergfeld, Alec van Herwijnen, Grégoire Bobillier, Philipp L. Rosendahl, Philipp Weißgraeber, Valentin Adam, Jürg Dual, and Jürg Schweizer
Nat. Hazards Earth Syst. Sci., 23, 293–315, https://doi.org/10.5194/nhess-23-293-2023, https://doi.org/10.5194/nhess-23-293-2023, 2023
Short summary
Short summary
For a slab avalanche to release, the snowpack must facilitate crack propagation over large distances. Field measurements on crack propagation at this scale are very scarce. We performed a series of experiments, up to 10 m long, over a period of 10 weeks. Beside the temporal evolution of the mechanical properties of the snowpack, we found that crack speeds were highest for tests resulting in full propagation. Based on these findings, an index for self-sustained crack propagation is proposed.
Stephanie Mayer, Alec van Herwijnen, Frank Techel, and Jürg Schweizer
The Cryosphere, 16, 4593–4615, https://doi.org/10.5194/tc-16-4593-2022, https://doi.org/10.5194/tc-16-4593-2022, 2022
Short summary
Short summary
Information on snow instability is crucial for avalanche forecasting. We introduce a novel machine-learning-based method to assess snow instability from snow stratigraphy simulated with the snow cover model SNOWPACK. To develop the model, we compared observed and simulated snow profiles. Our model provides a probability of instability for every layer of a simulated snow profile, which allows detection of the weakest layer and assessment of its degree of instability with one single index.
Cristina Pérez-Guillén, Frank Techel, Martin Hendrick, Michele Volpi, Alec van Herwijnen, Tasko Olevski, Guillaume Obozinski, Fernando Pérez-Cruz, and Jürg Schweizer
Nat. Hazards Earth Syst. Sci., 22, 2031–2056, https://doi.org/10.5194/nhess-22-2031-2022, https://doi.org/10.5194/nhess-22-2031-2022, 2022
Short summary
Short summary
A fully data-driven approach to predicting the danger level for dry-snow avalanche conditions in Switzerland was developed. Two classifiers were trained using a large database of meteorological data, snow cover simulations, and danger levels. The models performed well throughout the Swiss Alps, reaching a performance similar to the current experience-based avalanche forecasts. This approach shows the potential to be a valuable supplementary decision support tool for assessing avalanche hazard.
Frank Techel, Stephanie Mayer, Cristina Pérez-Guillén, Günter Schmudlach, and Kurt Winkler
Nat. Hazards Earth Syst. Sci., 22, 1911–1930, https://doi.org/10.5194/nhess-22-1911-2022, https://doi.org/10.5194/nhess-22-1911-2022, 2022
Short summary
Short summary
Can the resolution of forecasts of avalanche danger be increased by using a combination of absolute and comparative judgments? Using 5 years of Swiss avalanche forecasts, we show that, on average, sub-levels assigned to a danger level reflect the expected increase in the number of locations with poor snow stability and in the number and size of avalanches with increasing forecast sub-level.
Antoine Guillemot, Alec van Herwijnen, Eric Larose, Stephanie Mayer, and Laurent Baillet
The Cryosphere, 15, 5805–5817, https://doi.org/10.5194/tc-15-5805-2021, https://doi.org/10.5194/tc-15-5805-2021, 2021
Short summary
Short summary
Ambient noise correlation is a broadly used method in seismology to monitor tiny changes in subsurface properties. Some environmental forcings may influence this method, including snow. During one winter season, we studied this snow effect on seismic velocity of the medium, recorded by a pair of seismic sensors. We detected and modeled a measurable effect during early snowfalls: the fresh new snow layer modifies rigidity and density of the medium, thus decreasing the recorded seismic velocity.
Sebastian Landwehr, Michele Volpi, F. Alexander Haumann, Charlotte M. Robinson, Iris Thurnherr, Valerio Ferracci, Andrea Baccarini, Jenny Thomas, Irina Gorodetskaya, Christian Tatzelt, Silvia Henning, Rob L. Modini, Heather J. Forrer, Yajuan Lin, Nicolas Cassar, Rafel Simó, Christel Hassler, Alireza Moallemi, Sarah E. Fawcett, Neil Harris, Ruth Airs, Marzieh H. Derkani, Alberto Alberello, Alessandro Toffoli, Gang Chen, Pablo Rodríguez-Ros, Marina Zamanillo, Pau Cortés-Greus, Lei Xue, Conor G. Bolas, Katherine C. Leonard, Fernando Perez-Cruz, David Walton, and Julia Schmale
Earth Syst. Dynam., 12, 1295–1369, https://doi.org/10.5194/esd-12-1295-2021, https://doi.org/10.5194/esd-12-1295-2021, 2021
Short summary
Short summary
The Antarctic Circumnavigation Expedition surveyed a large number of variables describing the dynamic state of ocean and atmosphere, freshwater cycle, atmospheric chemistry, ocean biogeochemistry, and microbiology in the Southern Ocean. To reduce the dimensionality of the dataset, we apply a sparse principal component analysis and identify temporal patterns from diurnal to seasonal cycles, as well as geographical gradients and
hotspotsof interaction. Code and data are open access.
Nora Helbig, Michael Schirmer, Jan Magnusson, Flavia Mäder, Alec van Herwijnen, Louis Quéno, Yves Bühler, Jeff S. Deems, and Simon Gascoin
The Cryosphere, 15, 4607–4624, https://doi.org/10.5194/tc-15-4607-2021, https://doi.org/10.5194/tc-15-4607-2021, 2021
Short summary
Short summary
The snow cover spatial variability in mountains changes considerably over the course of a snow season. In applications such as weather, climate and hydrological predictions the fractional snow-covered area is therefore an essential parameter characterizing how much of the ground surface in a grid cell is currently covered by snow. We present a seasonal algorithm and a spatiotemporal evaluation suggesting that the algorithm can be applied in other geographic regions by any snow model application.
Bastian Bergfeld, Alec van Herwijnen, Benjamin Reuter, Grégoire Bobillier, Jürg Dual, and Jürg Schweizer
The Cryosphere, 15, 3539–3553, https://doi.org/10.5194/tc-15-3539-2021, https://doi.org/10.5194/tc-15-3539-2021, 2021
Short summary
Short summary
The modern picture of the snow slab avalanche release process involves a
dynamic crack propagation phasein which a whole slope becomes detached. The present work contains the first field methodology which provides the temporal and spatial resolution necessary to study this phase. We demonstrate the versatile capabilities and accuracy of our method by revealing intricate dynamics and present how to determine relevant characteristics of crack propagation such as crack speed.
Cited articles
Barandas, M., Folgado, D., Fernandes, L., Santos, S., Abreu, M., Bota, P., Liu, H., Schultz, T., and Gamboa, H.: TSFEL: Time Series Feature Extraction Library, SoftwareX, 11, 100456, https://doi.org/10.1016/j.softx.2020.100456, 2020. a
Bengio, Y., Courville, A., and Vincent, P.: Representation Learning: A Review and New Perspectives, IEEE Transactions on Pattern Analysis and Machine Intelligence, 35, 1798–1828, https://doi.org/10.1109/TPAMI.2013.50, 2013. a, b
Bessason, B., Eiríksson, G., Thórarinsson, Ó., Thórarinsson, A., and Einarsson, S.: Automatic detection of avalanches and debris flows by seismic methods, J. Glaciol., 53, 461–472, https://doi.org/10.3189/002214307783258468, 2007. a, b, c
Biescas, B., Dufour, F., Furdada, G., Khazaradze, G., and Suriñach, E.: Frequency content evolution of snow avalanche seismic signals, Surv. Geophys., 24, 447–464, https://doi.org/10.1023/B:GEOP.0000006076.38174.31, 2003. a
Breiman, L.: Random Forests, Machine Learning, 45, 5–32, https://doi.org/10.1023/A:1010933404324, 2001. a
Breiman, L.: Classification and regression trees, Routledge, https://doi.org/10.1201/9781315139470, 2017. a
Bründl, M., Etter, H.-J., Steiniger, M., Klingler, Ch., Rhyner, J., and Ammann, W. J.: IFKIS – a basis for managing avalanche risk in settlements and on roads in Switzerland, Nat. Hazards Earth Syst. Sci., 4, 257–262, https://doi.org/10.5194/nhess-4-257-2004, 2004. a
Bühler, Y., Bebi, P., Christen, M., Margreth, S., Stoffel, L., Stoffel, A., Marty, C., Schmucki, G., Caviezel, A., Kühne, R., Wohlwend, S., and Bartelt, P.: Automated avalanche hazard indication mapping on a statewide scale, Nat. Hazards Earth Syst. Sci., 22, 1825–1843, https://doi.org/10.5194/nhess-22-1825-2022, 2022. a, b
Caliński, T. and Harabasz, J.: A dendrite method for cluster analysis, Communications in Statistics, 3, 1–27, https://doi.org/10.1080/03610927408827101, 1974. a, b
Chmiel, M., Walter, F., Wenner, M., Zhang, Z., McArdell, B. W., and Hibert, C.: Machine Learning Improves Debris Flow Warning, Geophysical Research Letters, 48, https://doi.org/10.1029/2020GL090874, 2021. a, b
Clinton, J., Cauzzi, C., Fäh, D., Michel, C., Zweifel, P., Olivieri, M., Cua, G., Haslinger, F., and Giardini, D.: The Current State of Strong Motion Monitoring in Switzerland, Springer Netherlands, 219–233, ISBN 978-94-007-0152-6, https://doi.org/10.1007/978-94-007-0152-6_15, 2011. a
Dammeier, F., Moore, J. R., Hammer, C., Haslinger, F., and Loew, S.: Automatic detection of alpine rockslides in continuous seismic data using hidden Markov models, Journal of Geophysical Research: Earth Surface, 121, 351–371, https://doi.org/10.1002/2015JF003647, 2016. a
EAWS: European Avalanche Size Scale, https://www.avalanches.org/standards/avalanche-size/ (last access: 3 September 2024), 2021. a
EMSC-CSEM: Euro-Mediterranean Seismological Centre (EMSC), https://www.emsc-csem.org/#2 (last access: May 2023), 2023. a
Gu, S., Kelly, B., and Xiu, D.: Autoencoder asset pricing models, Journal of Econometrics, 222, 429–450, https://doi.org/10.1016/j.jeconom.2020.07.009, 2021. a
Hammer, C., Ohrnberger, M., and Fäh, D.: Classifying seismic waveforms from scratch: a case study in the alpine environment, Geophysical Journal International, 192, 425–439, https://doi.org/10.1093/gji/ggs036, 2013. a
Harris, C. R., Millman, K. J., van der Walt, S. J., Gommers, R., Virtanen, P., Cournapeau, D., Wieser, E., Taylor, J., Berg, S., Smith, N. J., Kern, R., Picus, M., Hoyer, S., van Kerkwijk, M. H., Brett, M., Haldane, A., del Río, J. F., Wiebe, M., Peterson, P., Gérard-Marchant, P., Sheppard, K., Reddy, T., Weckesser, W., Abbasi, H., Gohlke, C., and Oliphant, T. E.: Array programming with NumPy, Nature, 585, 357–362, https://doi.org/10.1038/s41586-020-2649-2, 2020. a
Heck, M., Hammer, C., van Herwijnen, A., Schweizer, J., and Fäh, D.: Automatic detection of snow avalanches in continuous seismic data using hidden Markov models, Nat. Hazards Earth Syst. Sci., 18, 383–396, https://doi.org/10.5194/nhess-18-383-2018, 2018a. a, b, c, d
Heck, M., van Herwijnen, A., Hammer, C., Hobiger, M., Schweizer, J., and Fäh, D.: Automatic detection of avalanches combining array classification and localization, Earth Surf. Dynam., 7, 491–503, https://doi.org/10.5194/esurf-7-491-2019, 2019. a
Hendrick, M., Techel, F., Volpi, M., Olevski, T., Pérez-Guillén, C., van Herwijnen, A., and Schweizer, J.: Automated prediction of wet-snow avalanche activity in the Swiss Alps, Journal of Glaciology, 1–14, https://doi.org/10.1017/jog.2023.24, 2023. a
Hibert, C., Mangeney, A., Grandjean, G., Baillard, C., Rivet, D., Shapiro, N. M., Satriano, C., Maggi, A., Boissier, P., Ferrazzini, V., and Crawford, W.: Automated identification, location, and volume estimation of rockfalls at Piton de la Fournaise volcano, Journal of Geophysical Research: Earth Surface, 119, 1082–1105, https://doi.org/10.1002/2013JF002970, 2014. a
Hinton, G. E. and Salakhutdinov, R. R.: Reducing the Dimensionality of Data with Neural Networks, Science, 313, 504–507, https://doi.org/10.1126/science.1127647, 2006. a
Hochreiter, S. and Schmidhuber, J.: Long Short-Term Memory, Neural Comput., 9, 1735–1780, https://doi.org/10.1162/neco.1997.9.8.1735, 1997. a
Ioffe, S. and Szegedy, C.: Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift, CoRR, arXiv [preprint], https://doi.org/10.48550/arXiv.1502.03167, 2015. a
Ismail Fawaz, H., Forestier, G., Weber, J., Idoumghar, L., and Muller, P.-A.: Deep learning for time series classification: a review, Data Mining and Knowledge Discovery, 33, 917–963, https://doi.org/10.1007/s10618-019-00619-1, 2019. a
Kingma, D. P. and Ba, J.: Adam: A Method for Stochastic Optimization, arXiv [preprint], https://doi.org/10.48550/arXiv.1412.6980, 2014. a
Kiranyaz, S., Avci, O., Abdeljaber, O., Ince, T., Gabbouj, M., and Inman, D. J.: 1D convolutional neural networks and applications: A survey, Mechanical Systems and Signal Processing, 151, 107398, https://doi.org/10.1016/j.ymssp.2020.107398, 2021. a
Kong, Q., Chiang, A., Aguiar, A. C., Fernández-Godino, M. G., Myers, S. C., and Lucas, D. D.: Deep convolutional autoencoders as generic feature extractors in seismological applications, Artificial Intelligence in Geosciences, 2, 96–106, https://doi.org/10.1016/j.aiig.2021.12.002, 2021. a
Lacroix, P., Grasso, J. R., Roulle, J., Giraud, G., Goetz, D., Morin, S., and Helmstetter, A.: Monitoring of snow avalanches using a seismic array: Location, speed estimation, and relationships to meteorological variables, J. Geophys. Res.-Earth, 117, 1–15, https://doi.org/10.1029/2011JF002106, 2012. a, b, c
Längkvist, M., Karlsson, L., and Loutfi, A.: A review of unsupervised feature learning and deep learning for time-series modeling, Pattern Recognition Letters, 42, 11–24, https://doi.org/10.1016/j.patrec.2014.01.008, 2014. a, b
Leprettre, B., Navarre, J.-P., Taillefer, A., Danielou, Y., organisms Panel, J. T. M., and Touvier, F.: Reliable Estimation of Avalanche Activity Using Seismic Methods, https://api.semanticscholar.org/CorpusID:54986472 (last access: March 2024), 1996. a
Li, Z., Meier, M. A., Hauksson, E., Zhan, Z., and Andrews, J.: Machine Learning Seismic Wave Discrimination: Application to Earthquake Early Warning, Geophysical Research Letters, 45, 4773–4779, https://doi.org/10.1029/2018GL077870, 2018. a
Lima, F. T. and Souza, V. M.: A Large Comparison of Normalization Methods on Time Series, Big Data Research, 34, 100407, https://doi.org/10.1016/j.bdr.2023.100407, 2023. a
Lin, G. W., Hung, C., Chien, Y. F. C., Chu, C. R., Liu, C. H., Chang, C. H., and Chen, H.: Towards automatic landslide-quake identification using a random forest classifier, Applied Sciences (Switzerland), 10, https://doi.org/10.3390/app10113670, 2020. a
Marchetti, E., Ripepe, M., Ulivieri, G., and Kogelnig, A.: Infrasound array criteria for automatic detection and front velocity estimation of snow avalanches: towards a real-time early-warning system, Nat. Hazards Earth Syst. Sci., 15, 2545–2555, https://doi.org/10.5194/nhess-15-2545-2015, 2015. a
Mayer, S., van Herwijnen, A., Ulivieri, G., and Schweizer, J.: Evaluating the performance of an operational infrasound avalanche detection system at three locations in the Swiss Alps during two winter seasons, Cold regions science and technology, 173, 102962, https://doi.org/10.1016/j.coldregions.2019.102962, 2020. a
Mayer, S., Techel, F., Schweizer, J., and van Herwijnen, A.: Prediction of natural dry-snow avalanche activity using physics-based snowpack simulations, Nat. Hazards Earth Syst. Sci., 23, 3445–3465, https://doi.org/10.5194/nhess-23-3445-2023, 2023. a
McClung, D. and Schaerer, P. A.: The avalanche handbook, The Mountaineers Books, ISBN 978-1-68051-539-8, 2006. a
McKinney, W.: Data Structures for Statistical Computing in Python, in: Proceedings of the 9th Python in Science Conference, edited by: van der Walt, S. and Millman, J., 56–61, https://doi.org/10.25080/Majora-92bf1922-00a, 2010. a
Meier, L., Jacquemart, M., Blattmann, B., and Arnold, B.: Real-time avalanche detection with long-range, wide-angle radars for road safety in Zermatt, Switzerland, in: Proceedings of the International Snow Science Workshop, Breckenridge, CO, 304–308, https://api.semanticscholar.org/CorpusID:59408658 (last access: March 2024), 2016. a
Mohammed, A. and Kora, R.: A comprehensive review on ensemble deep learning: Opportunities and challenges, Journal of King Saud University – Computer and Information Sciences, 35, 757–774, https://doi.org/10.1016/j.jksuci.2023.01.014, 2023. a
Mousavi, S. M. and Beroza, G. C.: Deep-learning seismology, Science, 377, eabm4470, https://doi.org/10.1126/science.abm4470, 2022. a
Mousavi, S. M., Zhu, W., Ellsworth, W., and Beroza, G.: Unsupervised Clustering of Seismic Signals Using Deep Convolutional Autoencoders, IEEE Geoscience and Remote Sensing Letters, 16, 1693–1697, https://doi.org/10.1109/LGRS.2019.2909218, 2019. a, b
Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., Desmaison, A., Köpf, A., Yang, E., DeVito, Z., Raison, M., Tejani, A., Chilamkurthy, S., Steiner, B., Fang, L., Bai, J., and Chintala, S.: PyTorch: An Imperative Style, High-Performance Deep Learning Library, arXiv [preprint], https://doi.org/10.48550/arXiv.1912.01703, 2019. a, b
Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M., and Duchesnay, E.: Scikit-learn: Machine Learning in Python, Journal of Machine Learning Research, 12, 2825–2830, https://doi.org/10.5555/1953048.2078195, 2011. a
Pérez-Guillén, C., Sovilla, B., Suriñach, E., Tapia, M., and Köhler, A.: Deducing avalanche size and flow regimes from seismic measurements, Cold Regions Science and Technology, 121, 25–41, https://doi.org/10.1016/j.coldregions.2015.10.004, 2016. a, b, c, d
Pérez-Guillén, C., Tsunematsu, K., Nishimura, K., and Issler, D.: Seismic location and tracking of snow avalanches and slush flows on Mt. Fuji, Japan, Earth Surf. Dynam., 7, 989–1007, https://doi.org/10.5194/esurf-7-989-2019, 2019. a, b, c
Rakthanmanon, T., Campana, B., Mueen, A., Batista, G., Westover, B., Zhu, Q., Zakaria, J., and Keogh, E.: Searching and mining trillions of time series subsequences under dynamic time warping, in: Proceedings of the 18th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD '12, Association for Computing Machinery, New York, NY, USA, 262–270, ISBN 9781450314626, https://doi.org/10.1145/2339530.2339576, 2012. a
Rousseeuw, P. J.: Silhouettes: A graphical aid to the interpretation and validation of cluster analysis, Journal of Computational and Applied Mathematics, 20, 53–65, https://doi.org/10.1016/0377-0427(87)90125-7, 1987. a, b
Schimmel, A., Hübl, J., Koschuch, R., and Reiweger, I.: Automatic detection of avalanches: evaluation of three different approaches, Natural Hazards, 87, https://doi.org/10.1007/s11069-017-2754-1, 2017. a
Schweizer, J., Mitterer, C., Techel, F., Stoffel, A., and Reuter, B.: On the relation between avalanche occurrence and avalanche danger level, The Cryosphere, 14, 737–750, https://doi.org/10.5194/tc-14-737-2020, 2020. a, b
Swiss Seismological Center (SED) at ETH Zurich: National Seismic Networks of Switzerland, http://www.seismo.ethz.ch/de/home/ (last access: May 2023), 2023. a
Seydoux, L., Balestriero, R., Poli, P., Hoop, M. D., Campillo, M., and Baraniuk, R.: Clustering earthquake signals and background noises in continuous seismic data with unsupervised deep learning, Nature Communications, 11, https://doi.org/10.1038/s41467-020-17841-x, 2020. a
Sielenou, P. D., Viallon-Galinier, L., Hagenmuller, P., Naveau, P., Morin, S., Dumont, M., Verfaillie, D., and Eckert, N.: Combining random forests and class-balancing to discriminate between three classes of avalanche activity in the French Alps, Cold Regions Science and Technology, 187, 103276, https://doi.org/10.1016/j.coldregions.2021.103276, 2021. a
Simeon, A.: Autoencoder-based feature extraction for the automatic detection of snow avalanches in seismic data, Zenodo [code], https://doi.org/10.5281/zenodo.15001358, 2025. a
Simeon, A., Seupel, C., Pérez-Guillén, C., and van Herwijnen, A.: Seismic Avalanche Signals recorded in Dischma Valley above Davos, Switzerland, Zenodo [data set], https://doi.org/10.5281/zenodo.14892926, 2025. a
Sola, J. and Sevilla, J.: Importance of input data normalization for the application of neural networks to complex industrial problems, IEEE Transactions on Nuclear Science, 44, 1464–1468, https://doi.org/10.1109/23.589532, 1997. a, b
Suriñach, E., Furdada, G., Sabot, F., Biescas, B., and Vilaplana, J.: On the characterization of seismic signals generated by snow avalanches for monitoring purposes, Ann. Glaciol., 32, 268–274, https://doi.org/10.3189/172756401781819634, 2001. a, b
Suriñach, E., Vilajosana, I., Khazaradze, G., Biescas, B., Furdada, G., and Vilaplana, J. M.: Seismic detection and characterization of landslides and other mass movements, Nat. Hazards Earth Syst. Sci., 5, 791–798, https://doi.org/10.5194/nhess-5-791-2005, 2005. a
Techel, F., Jarry, F., Kronthaler, G., Mitterer, S., Nairz, P., Pavšek, M., Valt, M., and Darms, G.: Avalanche fatalities in the European Alps: long-term trends and statistics, Geogr. Helv., 71, 147–159, https://doi.org/10.5194/gh-71-147-2016, 2016. a
Turner, R. J., Latto, R. B., and Reading, A. M.: An ObsPy Library for Event Detection and Seismic Attribute Calculation: Preparing Waveforms for Automated Analysis, Journal of Open Research Software, https://doi.org/10.5334/jors.365, 2021. a
van Herwijnen, A. and Schweizer, J.: Monitoring avalanche activity using a seismic sensor, Cold Regions Science and Technology,, 69, 165–176, https://doi.org/10.1016/j.coldregions.2011.06.008, 2011. a, b, c
van Herwijnen, A., Heck, M., and Schweizer, J.: Forecasting snow avalanches using avalanche activity data obtained through seismic monitoring, Cold Regions Science and Technology, 132, 68–80, https://doi.org/10.1016/j.coldregions.2016.09.014, 2016. a
van Herwijnen, A., Heck, M., Richter, B., Sovilla, B., and Techel, F.: When do avalanches release: investigating time scales in avalanche formation, in: International snow science workshop proceedings 2018, ISSW, Innsbruck, Austria, https://www.dora.lib4ri.ch/wsl/islandora/object/wsl:18935 (last access: March 2024), 2018. a
Vilajosana, I., Khazaradze, G., Suriñach, E., Lied, E., and Kristensen, K.: Snow avalanche speed determination using seismic methods, Cold Regions Science and Technology, 49, 2–10, https://doi.org/10.1016/j.coldregions.2006.09.007, 2007. a
Virtanen, P., Gommers, R., Oliphant, T. E., Haberland, M., Reddy, T., Cournapeau, D., Burovski, E., Peterson, P., Weckesser, W., Bright, J., van der Walt, S. J., Brett, M., Wilson, J., Millman, K. J., Mayorov, N., Nelson, A. R. J., Jones, E., Kern, R., Larson, E., Carey, C. J., Polat, İ., Feng, Y., Moore, E. W., VanderPlas, J., Laxalde, D., Perktold, J., Cimrman, R., Henriksen, I., Quintero, E. A., Harris, C. R., Archibald, A. M., Ribeiro, A. H., Pedregosa, F., van Mulbregt, P., and SciPy 1.0 Contributors: SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python, Nature Methods, 17, 261–272, https://doi.org/10.1038/s41592-019-0686-2, 2020. a
Wang, T., Trugman, D., and Lin, Y.: SeismoGen: Seismic Waveform Synthesis Using GAN With Application to Seismic Data Augmentation, Journal of Geophysical Research: Solid Earth, 126, e2020JB020077, https://doi.org/10.1029/2020JB020077, 2021. a
Wenner, M., Hibert, C., van Herwijnen, A., Meier, L., and Walter, F.: Near-real-time automated classification of seismic signals of slope failures with continuous random forests, Nat. Hazards Earth Syst. Sci., 21, 339–361, https://doi.org/10.5194/nhess-21-339-2021, 2021. a
Xu, B., Wang, N., Chen, T., and Li, M.: Empirical Evaluation of Rectified Activations in Convolutional Network, arXiv [preprint], https://doi.org/10.48550/arXiv.1505.00853, 2015. a
Xugang, L., Yu, T., Shigeki, M., and Chiori, H.: Speech Enhancement Based on Deep Denoising Autoencoder, arXiv [preprint], https://doi.org/10.48550/arXiv.2001.01538, 2013. a
Zhu, W., Mousavi, S. M., and Beroza, G. C.: Chapter Four – Seismic signal augmentation to improve generalization of deep neural networks, in: Machine Learning in Geosciences, edited by: Moseley, B. and Krischer, L., Advances in Geophysics, Elsevier, 61 151–177, https://doi.org/10.1016/bs.agph.2020.07.003, 2020. a, b
Short summary
Avalanche detection systems are crucial for forecasting, but distinguishing avalanches from other seismic sources remains a challenge. We propose novel autoencoder models to automatically extract features and compare them with engineered seismic features. These features are then used to classify avalanches and noise events. The autoencoder feature classifiers exhibit the highest sensitivity in detecting avalanches, while the engineered seismic classifier performs better overall.
Avalanche detection systems are crucial for forecasting, but distinguishing avalanches from...
