Content uploaded by Jordy Hendrikx
Author content
All content in this area was uploaded by Jordy Hendrikx on Nov 12, 2018
Content may be subject to copyright.
INFRASOUND DETECTION OF AVALANCHES: OPERATIONAL EXPERIENCE FROM
28 COMBINED WINTER SEASONS AND FUTURE DEVELOPMENTS
Walter Steinkogler1*, Giacomo Ulivieri2, Sandro Vezzosi2, Jordy Hendrikx3, Alec Van Herwijnen4and Tore
Humstad5
1Wyssen Avalanche Control, Reichenbach, Switzerland
2GeCo s.r.l, Florence, Italy
3Snow and Avalanche Laboratory, Montana State University, Bozeman, MT, USA
4WSL Institute for Snow and Avalanche Research SLF, Davos, Switzerland
5Norwegian Public Roads Administration, Molde, Norway
ABSTRACT: We present an overview of multiple verification campaigns performed to evaluate the perfor-
mance of and experience with IDA®(Infrasound Detection of Avalanches) operational systems in Austria,
Switzerland, Canada, Norway and the USA. This work focuses on operationally relevant facts and recom-
mendations for the design of infrasound systems. The comprehensive dataset consists of 28 combined op-
erational winter seasons at 10 different locations, covering a wide range of avalanche sizes and types, snow
climates (stratigraphy and snow depth), topographies and site-specific characteristics. The IDA®systems
automatically detected natural avalanches, artillery gun shots and detonations as well as explosions from
different remote avalanche control systems. Results show that the operational reliability of IDA®is limited to
avalanches of size class > 2.5 (corresponding to ~ 500 m of run-out distance and 5 ha), both dry and wet,
within a distance of 3-4 km from the array, with a probability of detection (POD) between 40 and 90% and a
false alert ratio (FAR) between 0 and 20%. The POD increases with size and decreases with distance. Dif-
ferences in performance are mainly related to site-specific characteristics. Site-specific calibration of the
automatic algorithm as well as tuning of the thresholds is a key factor for the performance optimization. The
presence of local terrain features and complex topography can limit the monitoring of certain avalanche
paths. Preliminary results suggest the use of multiple arrays and adapted algorithms can be an effective
solution to mitigate this limitation. Wind noise, ice layers or a dense snowpack can significantly reduce the
detection capability and in extreme cases render the system inoperative. However, a detailed design study,
optimized site selection, properly installation solutions, hardware robustness improvements and a clear defi-
nition of the operational requirements of the local avalanche control team can help minimizee these limita-
tions.
KEYWORDS: Infrasound detection, avalanche, operational
1. INTRODUCTION
Snow avalanches pose a direct threat for people
and infrastructure during winter. Governmental
agencies protect settlements and traffic routes using
permanent measures (tunnels, steel structures, etc.)
and/or active and passive temporary measures (e.g.
road closures, evacuations, preventive avalanche
release, avalanche forecasting, etc.). In this context,
by providing more timely information on avalanche
activity, automatic avalanche detection systems
could help reduce closure times of roads and facili-
tate decision making for local avalanche control
services. In addition, knowledge on the occurrence,
frequency and size of avalanches could assist re-
gional safety services responsible for the control
and forecasting of avalanche hazard.
In recent years ground based operational avalanche
detection systems (Steinkogler et al. 2018), such as
infrasound (Marchetti et al. 2015), radar (Fischer et
al. 2014, Gauer et al. 2007) or seismic based tech-
nology (Heck et al. 2017) have emerged to support
avalanche safety personnel in their daily decision-
making. A variety of systems for the detection of
avalanches has been developed and evaluated and
Proceedings, International Snow Science Workshop, Innsbruck, Austria, 2018
621
partly transferred into operational use at traffic route
operations and ski resorts (Steinkogler et al. 2016).
Depending on the aim of the operation and the ob-
ject at risk, the most suitable system should be se-
lected. Operational systems, such as the presented
IDA®system, can provide automatic notifications of
avalanche activity over single or multiple paths,
regardless of visibility conditions.
The first operational infrasound array, incorporating
the presented technology, was installed in 2012 in
Ischgl, Austria. The aim was to gather data on ava-
lanche activity of a larger area, develop algorithms
and evaluate if such systems can be used in opera-
tional context to assist the local avalanche control
teams. Based on the experience gathered during
these first winters, additional systems were installed
and operationally used in Switzerland and Norway.
Since 2016 systems were also installed at Rogers
Pass in Canada and Alta, USA. In Switzerland,
Canada and Norway extensive verification cam-
paigns were carried out over the last years. The
main results of these 28 cumulative winter seasons
of operation of IDA®are summarized here.
2. METHODS
The infrasound detection systems installed at the
various sites for operational use consist of 4 to 5
element infrasound arrays of a small aperture (100-
180 m). The systems were powered with autono-
mous or grid power supply. Infrasound array data
were transmitted in real-time to a dedicated cloud
server for the near real-time data processing and
automatic notification (SMS, e-mail, publication in
databases) of alerts. Details on the implemented
hardware and processing algorithms can be found
in Ulivieri et al. 2011-2016 and Marchetti et al. 2015.
The installation sites cover a wide range of snow
climates and avalanche types (from dry to wet flow-
ing avalanches and all avalanche size classes),
terrain types (steep avalanche paths to relatively
avalanche release areas and paths), vegetation
characteristics (open alpine meadows to heavily
forested areas) and other potential sources of
acoustic noise (e.g. railway, highway, airports, etc.).
Figure 1: Overview of the automatic detections provided by 7 IDA systems, which were operated during the
winter season 2017-18: natural (green) and artificially triggered avalanches (red) as well as detonations
from RACS or artillery (yellow) are shown. More than 700 avalanches were detected.
Proceedings, International Snow Science Workshop, Innsbruck, Austria, 2018
622
In addition to operational applications of the pre-
sented infrasound systems we summarize and con-
solidate the information of the verification cam-
paigns conducted in Norway (Humstad et al. 2016),
Switzerland (Mayer S. 2017, Mayer et al. 2018) and
Canada (Hendrikx et al. 2017). To have an inde-
pendent verification data set avalanche occurrence
data were collected with traditional manual visual
observations and automatic cameras to evaluate
system performance and to find the boundary condi-
tions for its effective operational use.
3. RESULTS AND DISCUSSION
3.1 Operational needs and system design
The experience gained during the last 4 years using
the IDA®technology, clearly shown the system is an
operational tool supporting avalanche control ser-
vices for the forecasting of avalanche hazard, the
management of traffic routes during avalanche cy-
cles and the verification of avalanche control. How-
ever, matching between specific needs of the ava-
lanche control teams and the system performance
and limits play an important role for a proper system
design. The main hazards that should be monitored,
e.g. the avalanche sizes and types that are relevant
for forecasting, are some examples of site specific
needs, while sensor position and geometry as well
as custom calibration of the algorithm are some
example of system design.
Operational needs criteria:
Are the avalanche paths which are of, main
and secondary, interest within the physical
detection limits of IDA®? (Figure 2)
What are typical characteristics of the ava-
lanches (dry/wet, size, runout distance, etc.)
and paths in the area of interest?
3.2 IDA®physical limitations
The comparison between IDA®detections and inde-
pendent observations of avalanche occurrence col-
lected by visual observers and/or cameras was
used to the verify the performance of the IDAs.
However, due to limits and uncertainties of the clas-
sical avalanches observation (i.e. bad visibility,
night, uncertain observation time, etc.) several of
the IDA®detections could not be verified by the
observations during the various campaigns, thus
Figure 2: IDA system performances in term of Probability of Detection (POD) as a function of the distance
and the size of (dry) natural avalanches (simplified illustration). Yellow and green colors indicate the ex-
pected medium or high POD, respectively. The operational needs fit these limitations to ensure the best op-
erational benefit of the system.
Proceedings, International Snow Science Workshop, Innsbruck, Austria, 2018
623
limiting the validation analysis to a semi-quantitative
and qualitative estimate of the performance and
highlighting the difficulties to obtain independent
verification data to compare IDA®measurements to.
The overall semi-quantitative and qualitative results
of these verification campaigns showed that mid-
sized and large dry slab avalanches were detected
with a Probability Of Detection (POD) of 40 to 90%
within a 3 to 5 km radius (Figure 2). The POD is
decreasing with increasing distance and decreasing
of the avalanche size, with POD reducing to zero for
small (size 1) avalanches.
Although both dry-snow, wet-snow and “mixed” slab
avalanches as well as glide-snow avalanches were
detected by IDA, the data indicates the detection
capability for wet-snow avalanches is limited to
larger sized events flowing over steep terrain, i.e.
with larger accelerations, respect to dry one.
The experience gathered with the various opera-
tional installations and verification can be used as a
guideline for planning new installations and to en-
sure the defined operational needs (Section 3.1)
can be met.
3.3 IDA General and technical limitations
Site specific characteristics, e.g. large terrain fea-
tures such as ridges, can significantly limit the per-
formance of IDA®and should be avoided when
choosing an adequate installation site.
In addition to site selection, it is crucial to calibrate
the automatic algorithm criterion as well as tuning of
the alert thresholds to the local conditions. Hendrikx
et al. 2017 showed that the initial operational year 1
calibration can be increased for year 2 after the first
representative avalanches, i.e. avalanches that fit
the defined operational needs, were recorded.
Based on a test installation at Rogers Pass Summit
Hendrikx et al. 2018 showed an increase in POD
after the algorithm calibration.
To fulfil operational needs of avalanche safety ser-
vices, the system needs to work reliable, on a 24/7
basis and with minimal maintenance efforts during
the entire winter season. It has proven to be chal-
lenging to guarantee system reliability (both hard-
ware and software) due to the often-challenging
environmental conditions. Improvements and tests
of a variety of solutions have now resulted in a set-
up that minimized hardware related issues. Yet,
environmental effects, such as strong winds or thick
melt freeze crusts inside the snow cover can reduce
the performance of IDA®.
Especially in situations when the detection perfor-
mance is reduced, and automatic detections are
only partly possible or with reduced reliability manu-
al analysis had to be performed throughout the sea-
son.
In addition to general technical considerations (e.g.
power supply, communication) the following tech-
nical criteria should be considered:
Typical seasonal snow cover at IDA location
(temporal and spatial coverage, maximum
snow depth, crusts, density, etc.)
Potential noise from wind or other (natural
and artificial) sources
Tree cover (wind noise reduction, solar ra-
diation on-site)
Hazards for workers and system (flooding,
extreme avalanche runout, rockfall, wildlife,
access, etc.)
3.4 Notable operational events Winter 2017/18
The IDA®installations that had been used opera-
tionally in Winter 2017-18 recorded more than 700
avalanches (Figure 1). Some selected operational
examples include:
The avalanche control team of Quinto
(Switzerland) did not only use the detec-
tions of IDA®for verification of artificially re-
lease avalanche with their RACS but also
for monitoring the natural activity on the op-
posite side of the valley (Figure 1).
IDA®Frutigen (Switzerland): This installa-
tion was a test installation to investigate the
performance of IDA®in this (lower eleva-
tion) area. In a discussion with the local
avalanche safety service, it was decided
that the operational needs (detection of ava-
lanches of mainly size 2 at distances up to 7
km) did not match the design and perfor-
mance of IDA.
IDA®Goms (Switzerland) and Rogers Pass
(Canada): At both locations a large number
of infrasound detections (artificial/natural
avalanches and explosions) were recorded
(IDA®Rogers Pass: 332 detections and
IDA®Goms (Switzerland: (160 detections).
Avalanche control teams considered the
IDAs as a substantial operational benefit.
Proceedings, International Snow Science Workshop, Innsbruck, Austria, 2018
624
At most sites IDA®provided reliable infor-
mation on the onset and timing of ava-
lanche cycles as well as real-time verifica-
tion during ongoing avalanche control work
4. CONCLUSIONS AND OUTLOOK
Experience and independent verification campaigns
have shown that infrasound detection of avalanches
can be applied in an operational context –provided
the operational requirements lie within the physical
limitations of the system, the deployment site is
adequately chosen and the system (hardware and
software) works reliable throughout the season.
The presented infrasound system (IDA®)provided
valuable data for local avalanche safety services by
gathering information about avalanche activity of
multiple avalanche paths in a larger area as well as
reliable information for avalanche control verifica-
tion. Typically, larger dry snow avalanches within a
3 to 4 km radius of the system can reliably be de-
tected. Since it continuously monitors its surround-
ings it provides data on natural avalanche activity,
which can be very useful information for the local
avalanche control team. During major avalanche
cycles, the system often provided more timely in-
formation on the start of the cycle.
The presented simplified overview of research re-
sults and operational experiences highlight the
strengths and limitations of current IDA®systems
and provide a benchmark for the design and instal-
lation of future operational infrasound systems.
In the future multi array-processing, with a similar
setup as for the Goms installations, will allow to
increase the location accuracy of this technology.
Based on the experience gathered with the existing
installations in challenging environmental condi-
tions, and since most system interruptions were
hardware related, based on the field experience a
new hardware generation is currently under devel-
opment (i.e. wireless infrasonic array) to further
increase the system robustness and to facilitate its
installation.
Merging and integrating with other technologies,
such as other detection systems or (modelled) snow
cover information, could also allow increasing over-
all reliability and accuracy. Recent developments in
the algorithm of the presented infrasound system
now also allow for a better detection of wet-snow
avalanches.
ACKNOWLEDGEMENT
The authors thank the local avalanche control
teams for the excellent cooperation and data collec-
tion.
LIMITATIONS
The presented results are only valid for the applied
technology (IDA®) and not necessarily for other
radar, infrasound or seismic systems.
REFERENCES
Bedard, A. 1994. An evaluation of atmospheric infrasound for
monitoring avalanches. Proceedings, 7th International Sym-
posium on Acoustic Remote Sensing and Associated Tech-
niques of the Atmosphere and Oceans, 3–5 October, Boul-
der, CO.
Fischer J. T., Fromm R., Gauer P., Sovilla B. 2014 Evaluation of
probabilistic snow avalanche simulation ensembles with
Doppler radar observations. Cold Regions Science and
Technology, Volume 97, 151-158.
Gauer, P., Kern, M., Kristensen, K., Lied, K., Rammer,
L.,Schreiber, H. 2007. On pulsed Doppler radar measure-
ments of avalanches and their implication to avalanche dy-
namics, Cold Regions Science and Technology, 50, 55–71
Gubler H., Hiller M. 1984. The use of microwave FMCW radar in
snow and avalanche research. Cold Regions Science and
Technology, Volume 9, Issue 2, 109-119.
Heck, M., Hammer, C., Van Herwijnen, A., Schweizer, J., Faeh,
D. 2017. Automatic detection of snow avalnaches in continu-
ous seismic data using hidden Markow models, Natural Haz-
ards and Earth System Sciences.
Hendrikx, J., Dreier, L., Ulivieri, G., 2017. Evaluation of an infra-
sound detection system for avalanches, Rogers Pass, Cana-
da; Winter 2016-17. Report for McElhanney Consulting Ser-
vices Ltd
(http://www.montana.edu/earthsciences/facstaff/MSU_Evalua
tion_Infrasound_RogersPass_Final.pdf)
Hendrikx, J., Dreier, L., Ulivieri, G., Sanderons, J., Jones, A.,
Steinkogler, W. 2018. Evaluation of an infrasound detection
system for avalanche in Rogers Pass, Canada, Internation
Snow Science Workshop ISSW 2018, Innsbruck, Austria
Humstad, T., Soderblom, O., Ulivieri, G., Langeland, S., Dahle,
H., 2016. Infrasound Detection of Avalanches in Grasdalen
and Indreeidsdalen, Norway, International Snow Science
Workshop ISSW 2016, Breckenridge, USA.
Kogelnig, A., Suriñach, E., Vilajosana, I., Hübl, J., Sovilla, B.,
Hiller, M. and Dufour, F. 2011. On the complementariness of
infrasound and seismic sensors for monitoring snow ava-
lanches, Natural Hazards and Earth System Sciences, 11(8),
2355-2370.
Proceedings, International Snow Science Workshop, Innsbruck, Austria, 2018
625
Marchetti E., Ripepe M., Ulivieri G., Kogelnig A. 2015 Infrasound
array criteria for automatic detection and front velocity esti-
mation of snow avalanches: towards a real-time early-
warning system. Natural Hazards and Earth System Scienc-
es 3(4):2709-2737.
Mayer, S. 2017. Evaluation von Infraschall-Systemen zur
Lawinendetektion, SLF Report
Mayer, S., van Herwijnen, A., Ulivieri, G., Schweizer, J. 2018
Evaluation of the performance of operational infrasond ava-
lanche detection systems at three locations in the swiss Alps
during two winter seasons. International Snow Science
Workshop ISSW 2018, Innsbruck, Austria
Pérez-Guillén, C., Sovilla, B., E. Suriñach, E., Tapia, M., and
Köhler, A. 2016. Deducing avalanche size and flow regimes
from seismic measurements, Cold Regions Science and
Technology, 121, 25–41, 2016.
Steinkogler, W. , Meier, L., Langeland, S., Wyssen, S. 2016.
Avalanche detection system: A state-of-the art overview on
selected operational radar and infrasound systems, Inter-
praevent 2016, Lucerne, Switzerland.
Steinkogler, W., Vera, C., Langeland, S.. 2018. Operational
avalanche detection systems: Experiences, physical limita-
tions and user needs, Geohazards 7, Canmore, Canada
Ulivieri, G., Marchetti, E., Ripepe, M., Chiambretti, I., De Rosa,
G. and Segor, V. 2011. Monitoring snow avalanches in
Northwestern Italian Alps using an infrasound array, Cold
Regions Science and Technology, Volume 69, Issues 2–3,
December 2011, Pages 177–183P
Proceedings, International Snow Science Workshop, Innsbruck, Austria, 2018
626
