Alec van Herwijnen

Seismic sensor array for monitoring an avalanche start zone: design, deployment and preliminary results

In snow, acoustic emissions originate from the breaking of bonds between snow crystals and the formation of cracks. Previous research has shown that acoustic signals emanate from a natural snowpack. The relation between these signals and the stability of the snowpack has thus far remained elusive. Studies on other hazardous gravitational processes suggest that damage accumulation precedes major failure. If increased cracking activity could be detected in snow this might be used for avalanche prediction. We report on the development of a seismic sensor array to continuously monitor acoustic emissions in an avalanche start zone. During three winters, over 1400 sensor days of continuous acoustic data were collected. With the aid of automatic cameras and a microphone the main types of background noise were identified. Seismic signals generated by avalanches were also identified. Spectrograms from seismic signals generated by avalanches exhibit a unique triangular shape unlike any source of background noise, suggesting that automatic detection and classification of events is possible. Furthermore, discriminating between loose-snow and snow-slab avalanches is possible. Thus far we have not identified precursor events for natural dry-snow slab avalanche release. Detailed investigation of one dry-snow slab avalanche showed that signals observed prior to the release originated from background noise or small loose-snow avalanches.

Automatically Detecting Avalanche Events in Passive Seismic Data

During the 2010-2011 winter season, we deployed seven geophones on a mountain outside of Davos, Switzerland and collected over 100 days of seismic data containing 385 possible avalanche events (33 confirmed slab avalanches). In this article, we describe our efforts to develop a pattern recognition workflow to automatically detect snow avalanche events from passive seismic data. Our initial workflow consisted of frequency domain feature extraction, cluster-based stratified subsampling, and 100 runs of training and testing of 12 different classification algorithms. When tested on the entire season of data from a single sensor, all twelve machine learning algorithms resulted in mean classification accuracies above 84%, with seven classifiers reaching over 90%. We then experimented with a voting based paradigm that combined information from all seven sensors. This method increased overall accuracy and precision, but performed quite poorly in terms of classifier recall. We, therefore, decided to pursue other signal preprocessing methodologies. We focused our efforts on improving the overall performance of single sensor avalanche detection, and employed spectral flux based event selection to identify events with significant instantaneous increases in spectral energy. With a threshold of 90% relative spectral flux increase, we correctly selected 32 of 33 slab avalanches and reduced our problem space by nearly 98%. When trained and tested on this reduced data set of only significant events, a decision stump classifier achieved 93% overall accuracy, 89.5% recall, and improved the precision of our initial workflow from 2.8% to 13.2%.

Chapter 12 – Snow Avalanches

Snow avalanches are a major natural hazard in most snow-covered mountain areas of the world. They are rapid, gravity-driven mass movements and are considered a meteorologically induced hazard. Snow avalanches are one of the few hazards that can be forecast, and in situ measurements of instability are feasible. Advanced hazard-mitigation measures exist, such as land-use planning based on modeling avalanche dynamics. The most dangerous snow avalanches start as a dry-snow, slab avalanche that is best described with a fracture mechanical approach. How fast and how far an avalanche flows is the fundamental question in avalanche engineering. Models of different levels of physical complexity enable the prediction of avalanche motion. Although the avalanche danger (probability of occurrence) for a given region can be forecast—in most countries with significant avalanche hazard, avalanche warnings are issued on a regular basis—the prediction of a single event in time and space is not (yet) possible.