Autoencoder-based feature extraction for the automatic detection of snow avalanches in seismic data

Abstract. Monitoring snow avalanche activity is essential for operational avalanche forecasting and the successful implementation of mitigation measures to ensure safety in mountain regions. To facilitate and automate the monitoring process, avalanche detection systems equipped with seismic sensors can provide a cost-effective solution. Still, automatically distinguishing avalanche signals from other sources in seismic data remains challenging. This is mainly due to the complexity of seismic signals generated by avalanches, the complex signal transmission through the ground, the relatively rare occurrence of avalanches, and the presence of multiple sources in seismic data. To study and interpret the variety of these signals, we compiled a dataset of seismograms recorded with an array of five seismometers installed in an avalanche study site above Davos, Switzerland. For the winter seasons of 2020–2021 and 2021–2022, this dataset comprised 84 avalanches and 828 noise (unrelated to avalanches) events. An approach to automate the detection of avalanches in seismic data is by applying machine learning methods. So far, research in this area has mainly focused on extracting domain-specific signal attributes as input features for statistical models. In contrast, we propose a novel application of representation learning from seismograms using autoencoder models to automatically extract features from 10 s seismic signals of snow avalanches. On top of that, we applied random forest classifiers to evaluate whether these features facilitate the detection of avalanches. Concretely, we trained one random forest classifier each on a set of expert-engineered seismic attributes (baseline), temporal autoencoder features and spectral autoencoder features. The classifiers achieved an avalanche recall of 0.67 (±0.00) (baseline), 0.71 (±0.02) (temporal autoencoder) and 0.70 (±0.01) (spectral autoencoder) and macro average f1-scores of 0.78 (±0.00) (baseline), 0.70 (±0.01) (temporal autoencoder) and 0.77 (±0.01) (spectral autoencoder). The developed approach could be potentially used as an operational, near real-time avalanche detection system. Yet, the relatively high number of false alarms still needs further implementation of the current automated seismic classification algorithms for effective avalanche detection.