Detection Of Snow Avalanches Research Articles

Performance of manual and automatic detection of dry snow avalanches in Sentinel-1 SAR images

Radar satellite-borne snow avalanche detection has rapidly grown into an important method for monitoring of avalanche activity over large spatial and long temporal scales. With increased application of Sentinel-1 data for avalanche detection, the need for improved performance evaluation arises. In this study, we make use of a unique dataset of field-based avalanche observations from eight adjacent avalanche paths at lake Tyin in central Norway. The dataset is a complete record of 318 dry slab avalanches that released during the winters 2016–2020, with information on release timing and extent. The dataset thus allows for detailed evaluation of the performance of automatic and manual avalanche detection in Sentinel-1 images, were both techniques make use of relative temporal backscatter intensity increase in case of an avalanche. Both automatic and manual detection underperform compared to previous studies with a probability of detection (POD) of 5.9% (false alarm rate (FAR) of 5.9%) and 11.3% (FAR of 19.27%) respectively. From the low relative backscatter intensity contrast of field-observed dry slab avalanches, it becomes evident that a higher backscatter contrast between avalanches and surrounding is needed for detectability in Sentinel-1 images. Neither the lag time between avalanche release and detection in a Sentinel-1 image, nor local incidence angle can explain the low POD's. Moreover, an analysis of meteorological conditions prior a during avalanche release and/or detection can explain the low POD's, given that differing snow conditions influence radar backscatter intensity. Finally, we cannot rule out that the small dataset of field-observed avalanches and the one-directional slope aspect of the study area influence our results, however, we believe that a physical limit of detectability of dry snow avalanches in C-band radar satellite data is reached.

Snow Avalanche Detection and Source Constraints Made Using a Networked Array of Infrasound Sensors

AbstractWe studied a triggered snow avalanche (∼60 s in duration and with ∼1,100 m run‐out) using a network of infrasound arrays and time‐synced video, with the objective of understanding the relationship between infrasound generation and flow dynamics. Using standard array processing techniques, we compared the infrasound source back azimuths with the avalanche flow path identified by frame‐differenced, geo‐referenced video. Results show that infrasound records begin with direct arrivals followed by echoes from the avalanche‐triggering explosions and these decay within 35 s of the detonations. Subsequent infrasound, which lasts 20–30 s, could then be attributed exclusively to the avalanche. These infrasound detections, and their triangulated source locations, progress downhill over time and the most intense infrasound appears to originate from a steep, mid‐path cliff band, where the avalanche reached speeds in excess of 30 m/s and accelerations of more than 5 m/s2. The recorded infrasound was compared to two candidate source models extracted from video: total flow motion and advancing flow motion. Advancing source locations were compared to acoustic intensity time series using a nonnegative least squares inversion to solve for, and to quantify, time‐varying infrasound source intensity. We observed that certain portions of the flow, most notably the early stages and the end stages (when the powder cloud was expanding and settling) were infrasonically quiet.

Automatic detection of snow avalanches in continuous seismic data using hidden Markov models

Abstract. Snow avalanches generate seismic signals as many other mass movements. Detection of avalanches by seismic monitoring is highly relevant to assess avalanche danger. In contrast to other seismic events, signals generated by avalanches do not have a characteristic first arrival nor is it possible to detect different wave phases. In addition, the moving source character of avalanches increases the intricacy of the signals. Although it is possible to visually detect seismic signals produced by avalanches, reliable automatic detection methods for all types of avalanches do not exist yet. We therefore evaluate whether hidden Markov models (HMMs) are suitable for the automatic detection of avalanches in continuous seismic data. We analyzed data recorded during the winter season 2010 by a seismic array deployed in an avalanche starting zone above Davos, Switzerland. We re-evaluated a reference catalogue containing 385 events by grouping the events in seven probability classes. Since most of the data consist of noise, we first applied a simple amplitude threshold to reduce the amount of data. As first classification results were unsatisfying, we analyzed the temporal behavior of the seismic signals for the whole data set and found that there is a high variability in the seismic signals. We therefore applied further post-processing steps to reduce the number of false alarms by defining a minimal duration for the detected event, implementing a voting-based approach and analyzing the coherence of the detected events. We obtained the best classification results for events detected by at least five sensors and with a minimal duration of 12 s. These processing steps allowed identifying two periods of high avalanche activity, suggesting that HMMs are suitable for the automatic detection of avalanches in seismic data. However, our results also showed that more sensitive sensors and more appropriate sensor locations are needed to improve the signal-to-noise ratio of the signals and therefore the classification.

Regional snow-avalanche detection using object-based image analysis of near-infrared aerial imagery

Abstract. Snow avalanches are destructive mass movements in mountain regions that continue to claim lives and cause infrastructural damage and traffic detours. Given that avalanches often occur in remote and poorly accessible steep terrain, their detection and mapping is extensive and time consuming. Nonetheless, systematic avalanche detection over large areas could help to generate more complete and up-to-date inventories (cadastres) necessary for validating avalanche forecasting and hazard mapping. In this study, we focused on automatically detecting avalanches and classifying them into release zones, tracks, and run-out zones based on 0.25 m near-infrared (NIR) ADS80-SH92 aerial imagery using an object-based image analysis (OBIA) approach. Our algorithm takes into account the brightness, the normalised difference vegetation index (NDVI), the normalised difference water index (NDWI), and its standard deviation (SDNDWI) to distinguish avalanches from other land-surface elements. Using normalised parameters allows applying this method across large areas. We trained the method by analysing the properties of snow avalanches at three 4 km−2 areas near Davos, Switzerland. We compared the results with manually mapped avalanche polygons and obtained a user's accuracy of > 0.9 and a Cohen's kappa of 0.79–0.85. Testing the method for a larger area of 226.3 km−2, we estimated producer's and user's accuracies of 0.61 and 0.78, respectively, with a Cohen's kappa of 0.67. Detected avalanches that overlapped with reference data by > 80 % occurred randomly throughout the testing area, showing that our method avoids overfitting. Our method has potential for large-scale avalanche mapping, although further investigations into other regions are desirable to verify the robustness of our selected thresholds and the transferability of the method.

A complete snow avalanche activity record from a Norwegian forecasting region using Sentinel-1 satellite-radar data

Collecting a complete dataset of snow avalanche activity in a given avalanche forecasting region throughout an entire winter is a seemingly simple task that is very difficult to accomplish. Traditional field-based methods are too time and cost consuming and limited to good visibility and accessible terrain. Different remote sensing methods have been proposed to assist in monitoring of snow avalanche activity, with satellite-borne radar remote sensing allowing for all-weather and all-light conditions monitoring. Here we present a complete, two-year snow avalanche activity record from a Norwegian forecasting region using Sentinel-1A radar satellite imagery. We used a change detection method to manually identify avalanche debris based on localized backscatter increase in avalanche runout zones and assess these detections using multi-temporal and multi-sensor datasets. We then used the complete dataset of snow avalanche activity to exemplify different applications from validating issued avalanche bulletins, to generating avalanche path maps, and creating a snow climate classification. Comparing manual identification in Sentinel-1A images and a very high resolution Radarsat-2 image suggest overall underestimation of avalanche activity, also when considering the non-detectability of small avalanche debris. We foresee that with the further increase in satellite-radar data availability and the improvement of automatized snow avalanche detection, the presented methodologies will be able to assist in operational avalanche forecasting worldwide in the near future.

Automatic detection of avalanches: evaluation of three different approaches

Automated detection of snow avalanches is an important tool for avalanche forecasting and for assessing the effectiveness of avalanche control measures at bad visibility. Avalanche detection systems are usually based on infrasound, seismic, or radar signals. Within this study, we compared three different types of avalanche detection systems: one avalanche radar, one infrasound array system consisting of four infrasound sensors, and a newly developed single sensor infrasound system. A special focus is given to the new single sensor system, which is a low cost, easy to install system, originally designed for the detection of debris flows and debris floods. Within this work, we analysed how this single sensor system could be adapted to detect also snow avalanches. All three systems were installed close to a road near Ischgl (Tyrol, Austria) at the avalanche-exposed Paznaun Valley. The valley is endangered by two avalanche paths which are controlled by several avalanche towers. The radar system detected avalanches accurately and reliably but was limited to the particular avalanche path towards which the radar beam was directed. The infrasound array could detect avalanches from all surrounding avalanche paths, however, with a higher effort for installation. The newly tested single infrasound sensor system was significantly cheaper and easier to install than the other two systems. It could also detect avalanches form all directions, although without information about the direction. In summary, each of the three different systems was able to successfully detect avalanches and had its particular strengths and weaknesses, which should be considered according to the specific requirements of a particular practical application.

Automatic detection of wet-snow avalanche seismic signals

Avalanche activity is an important factor when estimating the regional avalanche danger. Moreover, a complete and detailed picture of avalanche activity is needed to understand the processes that lead to natural avalanche release. Currently, information on avalanche activity is mainly obtained through visual observations. However, this involves large uncertainties in the number and release times, influencing the subsequent analysis. Therefore, alternative methods for the remote detection of snow avalanches in particular in non-observed areas are highly desirable. In this study, we use the excited ground vibration to identify avalanches automatically. The specific seismic signature of avalanches facilitates the objective detection by a recently developed classification procedure. A probabilistic description of the signals, called hidden Markov models, allows the robust identification of corresponding signals in the continuous data stream. The procedure is based upon learning a general background model from continuous seismic data. Then, a single reference waveform is used to update an event-specific classifier. Thus, a minimum amount of training data is required by constructing such a classifier on the fly. In this study, we processed five days of continuous data recorded in the Swiss Alps during the avalanche winter 1999. With the restriction of testing large wet-snow avalanches only, the presented approach achieved very convincing results. We successfully detect avalanches over a large volume and distance range. Ninety-two percentage of all detections (43 out of 47) could be confirmed as avalanche events; only four false alarms are reported. We see a clear dependence of recognition capability on run-out distance and source–receiver distance of the observed events: Avalanches are detectable up to a source-receiver distance of eight times the avalanche length. Implications for analyzing a more comprehensive data set (smaller events and different flow regimes) are discussed in detail.

Infrasound array criteria for automatic detection and front velocity estimation of snow avalanches: towards a real-time early-warning system

Abstract. Avalanche risk management is strongly related to the ability to identify and timely report the occurrence of snow avalanches. Infrasound has been applied to avalanche research and monitoring for the last 20 years but it never turned into an operational tool to identify clear signals related to avalanches. We present here a method based on the analysis of infrasound signals recorded by a small aperture array in Ischgl (Austria), which provides a significant improvement to overcome this limit. The method is based on array-derived wave parameters, such as back azimuth and apparent velocity. The method defines threshold criteria for automatic avalanche identification by considering avalanches as a moving source of infrasound. We validate the efficiency of the automatic infrasound detection with continuous observations with Doppler radar and we show how the velocity of a snow avalanche in any given path around the array can be efficiently derived. Our results indicate that a proper infrasound array analysis allows a robust, real-time, remote detection of snow avalanches that is able to provide the number and the time of occurrence of snow avalanches occurring all around the array, which represent key information for a proper validation of avalanche forecast models and risk management in a given area.

Robust snow avalanche detection using supervised machine learning with infrasonic sensor arrays

Automated detection of snow avalanches is crucial to assess the effectiveness of avalanche control by explosions, and to monitor avalanche activity in a given area in view of avalanche forecasting. Several automated or semi-automated detection technologies have been developed in the past among which infrasound-based detection is the most promising for regional-scale avalanche monitoring. However, due to significant ambient noise content in infrasonic signals, e.g. from atmospheric processes or airplanes, fully automated and reliable avalanche detection has been very challenging. Signal processing is highly critical and strongly affects detection accuracy. Here, a robust detection method by using supervised machine learning is introduced. Machine learning algorithms can take into account multiple signal features and statistically optimize the classification task. We analyzed infrasound data with concurrent visual avalanche observations from the test site Lavin (Eastern Swiss Alps) for the winter of 2011–2012. A support vector machine was trained by using training data from the first half of the winter season and the accuracy was tested on data from the second half of the season. A significant reduction of false detections, from 65% to 10%, was achieved compared to a threshold-based classifier provided by the sensor manufacturer. The proposed method enables reliable assessment of the avalanche activity in the surroundings of the system and paves the way towards robust and fully automated avalanche detection using infrasonic systems.

Rural Intelligent Transportation System Natural-Hazard Management on Low-Volume Roads

The growth of winter travel on alpine roads in the western United States has increased the risk to motorists and highway maintenance personnel owing to a variety of natural hazards. Hazards include snow and ice, avalanching snow, and blowing and drifting snow. The conditions call for attendant need for incident response. A substantial number of affected routes are low-volume rural winter roads. Configurations have been developed for rural intelligent transportation system (ITS) technology that can detect hazards and provide, autonomously and in real time, warnings to and traffic control actions for motorists, highway maintainers, and incident responders for roadway natural hazards. These warnings include on-site traffic control signing and road closure gates, in-vehicle audio alarms for agency maintenance and patrol vehicles, and notification to highway agency maintenance facilities or centralized multiagency dispatchers. These actions and notifications are initiated automatically from the remote rural sites and via manual intervention from off-site personnel, well removed from the rural roadway corridor itself. About 5 years of experience have been accumulated in using these rural ITS natural-hazard reduction systems, including snow avalanche detection and warning systems on Loveland Pass, Colorado; Hoback Canyon, Wyoming; and Banner Summit, Idaho. Automated road closure gates on the Teton Pass in Idaho and Wyoming now allow for remote road closure during heavy snow events. These cost-effective ITS natural-hazard systems are highly exportable for other processes that affect rural low-volume roadways, including landslide, flooding, high surf, high winds, loss of visibility, wildlife, and other natural hazards of this type.

Rural Intelligent Transportation System for Snow Avalanche Detection and Warning

The growth of winter travel on alpine roads in the western United States, a result of the demand for reliable winter access, has increased the hazard to motorists and highway maintenance personnel from snow avalanches. Configurations are presented for systems that can detect and provide, in real time, warnings to motorists and highway maintainers of roadway avalanches. These warnings include on-site traffic control signing, in-vehicle audio alarms for winter maintenance vehicles, and notifying maintenance facilities or centralized agency dispatchers. These avalanche detection and warning systems can detect an existing avalanche and use the avalanche’s remaining time of descent to initiate on-site alarms. Alternatively, real-time knowledge and notification of the onset of avalanching may be used to proactively manage the evolving hazard over an affected length or corridor of highway. These corridors can be several tens of kilometers in length and may be very remote, low-volume rural highways. As a consequence, these detection and warning systems must be cost-effective alternatives to existing avalanche hazard reduction technology. Results and experiences from the winters of 1997–1998 and 1998–1999 are presented, along with recommendations and criteria for future deployment of these automated natural hazard reduction systems for rural transportation corridors.