Technical Addendum

Technical Addendum

_________ Version du document: v0.9E

About the utility,

the potential and

the limits in...

... high infrasonic

environmental intelligent

surveillance & recognition

Arfang-S2R T e c h n i c a l A d d e n d u m

2/31

Arfang-S2R is a remote system allowing permanent monitoring of the noise immissions of a wide range of natural or human activities such as natural dangers, aviation, and other possible acoustic events such as transient events, subsonic or sonic booms or other.

This document aims to specify and discuss the utility, the possibilities and the limits of high-infrasonic acoustic goniometry on which the specific Arfang-S2R solution is based.

Additional documentation:

To quickly understand what the Arfang-S2R product is and how it can be implemented for snow avalanche monitoring, see the

Arfang-S2R –Product Notice.

To walk through the Arfang-S2R application features, refer to the

Arfang-S2R User Guide.

The Arfang-S2R technical platform can be used to monitor many other infra-sonic activities than snow avalanches. An example (experimental surveillance of helicopter activity around a helicopter base) is described in

Arfang-S2R Case Study.

IAV Technologies SARL www.arfang.com

ISR Products Division Chemin des Couleuvres 4A, 1295 Tannay, Switzerland Tel: +41 (0)22 960 11 04 E-mail: [email protected] © IAV Technologies


3/31

Table of content

Preamble ............................................................................................................................................................ 4

Infrasound .......................................................................................................................................................... 5

Infrasonic Event .................................................................................................................................................. 6

Acoustics & Seismics ........................................................................................................................................... 8

Sound Recording Qualities ............................................................................................................................... 10

Beamforming, identification and classification ................................................................................................ 13

Triangulation, range ......................................................................................................................................... 15

Optimal Basic Typology for Local Sensing ........................................................................................................ 17

Evaluation of a Detection Performance ........................................................................................................... 20

Technical Considerations .................................................................................................................................. 22

History and Latest Developments in Avalanche Surveillance.......................................................................... 23

Bibliography ...................................................................................................................................................... 30


4/31

Preamble

In general, the subject of applied infrasonic acoustic goniometry can be described as the cap-ture and recognition of acoustic waves with frequencies ranging from a thousandth Hz to about twenty Hz.

In this very wide frequency spectrum, the Arfang-S2R system is situated at the highest extremi-ty and even slightly above, targeting wavelengths of about 10 m to 340 m (i.e. from 1Hz to

40 Hz). At the origin of these wavelengths are sources like nat-ural phenomena or medium-power human activity, whose emissions propagate over kilometrical distances and circum-vent obstacles or reliefs of similar dimensions. As such, they create several possibilities of surveillance over long distances, ranging from one to about one thousand kilometres, depending on the situation.

This document aims to specify and discuss the utility, the possi-bilities and the limits of high-infrasonic acoustic goniometry on which the specific Arfang-S2R solution is based.

This solution is the culmination of 25 years of experience with sensing and recognition of infra-sonic events, especially aimed at avalanche surveillance, detection of pulse signals and general surveillance of certain aviation activities. In this period, experimental prototypes or pilot ver-sions were developed through a dozen long-term research and development projects. A list of related publications on the background of the Arfang-S2R system is available at the end of this document.

In our view, Arfang-S2R provides a simple, robust, versatile, progressive and high-performance solution for any infrasonic surveillance application, which can be centred on the frequency range from 1 to 40 Hz (or up to a decade lower and two decades higher by adapting the geom-etry of the antenna and the bandwidth of the sensors).

The system was essentially developed when applied to the sur-veillance of avalanches and aviation activities. It can easily be used in numerous other applications like characterisation of gravitational effects of terrain instabilities or glaciers, like fall-ing rocks or debris flow. Or else for applications characterising mechanical sources with rotative, impulsive or broadband components: aviation activities, heavy industry, mining; or else, and at eventually greater distances (and in this case clos-er to the low-frequency limit of 1 Hz), capturing of certain phenomena in the atmosphere or earthquakes.

Web image evoking the military origin of high infrasonic detection. Or when direct observation and electromagnetic radars are not operational, long-range acoustics can provide early and directional infor-mation

Infrasonic signature of helicopters captured by 2 ground Arfang-S2R substations


5/31

Infrasound

Infrasound is sound which extends below the range of human hearing, from 20 Hz down to 0.001 Hz. The Arfang-S2R system targets the highest extremity and even slightly above this frequency spectrum: from 1Hz to 40 Hz, i.e. aerial sound waves of 340 m to 1 m respectively. This restricted domain is suitable for the environmental acoustic surveillance of many medium-sized natural phenomena, or medium-powered human activities, like for example surveillance of avalanche activity or certain aviation activities or pulse phenomena like mining or explosions.

Half of these are therefore inaudible waves (1 to 20 Hz, or high infrasound) and the rest are low sounds (20 to 40 Hz, or very low audio). This spectrum of wavelengths, considered in terms of deca-kilometric de-tection ranges, corresponds with emission sources whose apparent size is established in approximately dozens to hundreds of metres. Their mass (for gravitational sources) varies from tons to millions of tons, while the power (for sources with a mechanical or chemical origin) ranges from several kW to thousands of kW. Characteristics, therefore, which often belong to small- to medium-sized natural phenomena, geotech-nical industrial activities or medium to heavy land- or air-transport.

Strictly speaking, infrasound is a frequency domain that extends much lower in the spectrum of acoustic waves (infinitely lower even, if one considers that the only real limit is 0Hz, i.e. a net air flow, like wind for example). In the rest of this document, though, we will speak of high infrasound or even simpler of infra-sound.

Examples of infrasonic sources: supersonic aircrafts, natural phenomena and human activity like natural or artificial high-infrasonic sources. To generate infrasound, the source must be of considerable power or dimension as per the above illustrations. By analogy with loudspeakers, the (probably fake) giant loudspeaker on the left (web source) shows a diameter of about 2.5 m which corresponds to a wavelength of about 130 Hz, so still far too small to generate infrasound (wavelength of 10 Hz = 34 m).


6/31

Infrasonic Event

An infrasonic event captured by a microphone or an array of acoustic, infrasonic surveillance microphones can be defined as a coherent, low frequency, sound wave which propagates in space. It is thus character-ized by its own duration and acoustic signature and can be distinguished from the infrasonic background noise by a propagation direction (or more precisely by an angle of incidence) seen by an observation sys-tem.

For example, the passages of the blade of a wind-turbine in front of its mast, just like the passage of a helicopter-blade past the wing of the machine, are typically two well-known sources of infrasound whose capture (passing of the helicopter at a long or short dis-tance, windy episode) could establish an infrasonic event seen from a distant observation system. In both cases, the frequency of the infrasound corresponds to the number of passages per second of the blades in front of the obstacle (that is approximately 10 to 30 Hz for the helicopter and 10 times less for the wind-turbine, or the

ratio between their dimensions and the number of blades). Their respective acoustic signatures, so-called tonal components, can easily be identified and recognized. For these two events, the duration of the signa-tures could extend to a couple of dozens of seconds for a helicopter passing, to a possibly much longer or even uninterrupted duration for the wind-turbine. Unless the helicopter is in stationary flight, the frequen-cy of its acoustic signature will vary according to the Doppler-effect, i.e. its radial speed, in relation to the source, whereas the frequency of the acoustic signature of the wind-turbine will vary according to its rota-tion speed, which in turn depends on the speed of the wind.

Moreover, this same wind could create another infrasonic element, originating from its interaction with the wind-turbine. This is unrelated to the rotation of the blade but caused by the obstacle-flow in-teraction between the wind flow and the large mast, which creates periodical releas-es of eddies, sources of infrasound emission. It is at the moment of the interac-tion with the fixed obstacle and only then, that another tonal component of the event ‘wind-turbine in production’ is created. Finally, the information carried by the elec-tric signal of the microphone(s) that capture these infrasonic waves, can include three different elements in its signature. The first two are the images of the two ‘true infrasounds’, one produced by the periodic air compression between a blade

Infrasonic signature presenting a marked frequential element with a variation between approximately 3 to 4 Hz. The analysis of this trace shows that the angle of incidence coincides with a dam at some 20 km, which suggests that it is infrasonic radiation caused by low frequency vibrations from the deck of the dam due to strains from turbines, water or wind. On the upper sonogram, the signal is mixed with stronger para-sitic sounds, which are due to gusts of wind. The lower sonogram is the result of a filter which is used to reduce the parasitic disturbance.

Infrasonic signature (spectrogram*) of the Concorde NY-London in supersonic descent over northern Scotland, captured from the Swiss Alps at a distance of about 900 km. The travel time of the wave is about an hour and the infrasonic event lasts about 15 minutes, which reflects the path of the infrasonic rays in the different layers of the atmosphere (curve of the sound rays influenced by wind profiles and temperature). *A spectrogram is a visual representation of the spectrum of frequencies of a sig-nal as it varies over time. A common format is a graph with two geometric dimensions: one axis represents time, and the other axis represents frequency; a third dimension indicating the amplitude of a particular frequency at a particular time is represented by the intensity or colour of each point in the image. [Source: Wikipedia, the free encyclopaedia]


7/31

and the mast of the wind-turbine, the other by periodical releases of eddies (and even a third caused by the same releases on the blades themselves). A third element that has not been mentioned until now, is the "parasitic infrasonic signal" consisting of the air friction on the acoustic inlets of the microphone(s) (under the assumption that the observation system can also be locally exposed to the same windy episode as the one which causes the movement of the wind-turbine). This last point is often important in practice since the majority of the infrasonic acoustic surveillance applications are often exposed to wind, breezes or even only weak air flows like thermal effects, since they are often deployed outside.

In many other situations, acoustic signatures can include elements coming from superposed sources or phenomena (distinct or linked, near or far, etc.). It is essential to have access to a method that separates the sources, to be able to distinguish these different superposed elements in a single signal.

For example, when monitoring avalanche activity in a medium or high altitude mountain range, including for example important infrastructure, construction sites or flight paths and roads, etc., experience has shown how difficult it is to distinguish an avalanche event from a wind gust, the passing of an airliner (in verticality with the observation system) or a heavy train in a tunnel, or yet an occasional movement of a construction engine, or yet other echoing phenomena, or eventually sometimes a much further removed event, up until tens or hundreds or thousands of kilometres away. By contrast, source separation is much easier with other types of events with very different signatures, like for example closer or much shorter events (transitory or pulse signals) like for example a rock or ice falling or a mining explosion.

The example hereinafter shows a series of 5 explosions, which are immediately visualised like pulse events and whose acoustic power is distributed over all the infrasonic bandwidths. They appear as vertical lines in the spectrogram where time is in abscissa, the frequencies in ordinate with the colour getting warmer as the infrasonic amplitude intensifies. On the contrary, the passage of a helicopter, that produces a practically constant frequency sound (as has been mentioned for the blade-rotation cycles) translates as horizontal lines, since the observed frequency variation only corresponds to the Doppler-effect. In other words, the frequency diminishes when the helicopter moves away and increases when it approaches. Regarding these acoustic events, the avalanche is neither a pulse event (short) nor an event likely to produce a pure sound (tonal). It typically materialises as an oblong stain which lasts from a few to several dozens of seconds, its infrasonic spectrum can extend to several Hz up until the low audible and even audible, in case of humid snow avalanches.

These two images represent the rec-orded sounds by a microphone placed under the compact snow cover of about 1m (old groomed snow), during a series of avalanches, artificially triggered by detonation of explosive charges from a helicopter (5 charges detonated on one mountainside in a little less than 2 minutes, Blasting SAREM/Armand Dussex, Anzère, Swit-zerland, January 1995). The vertical lines are clearly visible, corresponding with the 5 consecutive explosions. In the lower graph, the horizontal lines can clearly be distinguished, showing the additional presence of the helicopter, and four characteristic stains showing four successfully triggered avalanches (by the first detonation, the third, the fourth and the fifth).


8/31

Acoustics & Seismics

Aquatic environments aside, infrasounds in the air and seismic signals in the ground can mainly be distin-guished by their propagation environment and speed. Infrasound has acoustic longitudinal waves with the speed of sound, whereas seismic signals have longitudinal and transversal waves with 10 times faster speeds. Nevertheless, for a given event and with two recorders: one seismic (geophone) and the other acoustic (microphone), placed more or less close to each other, the two types of signals are often very simi-lar. As we will see later on, there is even a reciprocity because an infrasound can print a seismic signature on the surface of the ground, just like a vibration of the ground can locally entail an infrasonic emission.

In the opposite example, a geophone and a microphone are placed next to each other at about a hundred metres from an isolated road. They both record the passage of a 20 ton truck. The acoustic signature (up-per image) is of a higher fre-quency whereas the seismic signature (lower image) is precocious (NB: this is invisible at this scale) as well as longer: the acoustic reply transmits the direct aerial emission of the vehicle (engine, gear, roll-ing noise, vibrations of the truck bed) whereas the seismic reply translates the global trembling of the ground as a result of the mass-effect of the contact of the wheels with the road.

This resemblance between seismic and infrasonic signa-tures of an event exists as a general rule for all superficial sources at proximate distances. The former systematically contain more noises from further removed ele-ments, whereas the latter contain local wind conditions.

In the second example hereinafter, recorded with the same sensors (in the same locations) as the previous example, the recorded event is a gas release detonation from a fixed installation for the artificial triggering of avalanches. The spectrograms of the acoustic infrasonic signal (see top graph on the next page) and the seismic signal (at the bottom) show the characteristic signature of a pulse event, i.e. a vertical line. This shot being positive, it is equally easy to distinguish the characteristic signature of the artificially triggered avalanche following the explosion, which manifests itself as an oblong stain translating wide frequency content, without any pronounced tonal elements, and of a certain duration.

Simultaneous, compared recording, lasting 5 minutes, between an infrasonic micro-phone (upper image) and a geophone (lower image), of a passing heavy vehicle at low speed on an isolated mountain road.

Fixed remote avalanche control system: example of a remote-controlled gas exploder (type GAZEX

TM.) The exploder

is remotely triggered for preventive explosions to purge the local slope.


9/31

A more detailed inspection of these signatures as well as the measuring situation and the propagation con-ditions respectively in the air and in the ground (cf. diagram) allows to highlight the fact that the opposite seismic signature is not the image of the arrival of the seismic waves which could have been propagated over the ground between the location of the event and the point of measurement of the vibration of the ground, but the image produced by the passage of the infrasonic wave, and so rather the seismic reflection thereof. In other terms, in this particular case, the infrasonic sensor logically captures the infrasounds pro-duced by the avalanche directly, whereas - more surprisingly - the seismic sensor captures the local response from the ground at the passage of this infrasound.

This is explained by the fact that since the triggered avalanche was a small powdery avalanche for a so-called anticipated purge of the relevant slope (and not a high-energy avalanche), powder snow was suffi-cient to produce a well-detectable infrasound at a distance of 2 km, even in the presence of a little wind (gusts visible on the upper graph). Its mass however, was sufficient to induce a noticeable trembling of the ground. At the back of the seismic signature, one notices another line with a weak amplitude which corre-sponds to a tonal source caused by a nearby ski-lift (whose weight resounds on the ground), while this emission is acoustically undetectable because the ski-lift’s size is not important enough for an effective acoustic radiation in very low frequencies.

Infrasonic acoustic signal noticed at a preventive positive detona-tion at d1 = 1800 m distance as the crow flies, i.e. having triggered a slide.

Seismic signal noticed at the same detonation, the seismic sensor (geophone) is located at 70 m from the infrasonic sensor and at a distance of d2 = 1870 m as the crow flies from the detonator, or d3 = 2100 m according to shortest ground line.

A calculation establishing respectively the travel times according to the three propagation paths (aerial exploder – microphone and aerial exploder – geophone, at the speed of sound in the air; and structure-borne exploder-geophone at a seismic speed) proves that the signal recorded by the geophone is the reaction of the ground on the passage of the aerial infra-sonic wave.


10/31

Sound Recording Qualities

In an outdoor environment, the most important quality of acoustic sensing is the largest possible sensitivity to coherent waves of interest and the weakest sensitivity to perturbations due to wind on the acoustic in-lets of the sensor itself. As already pointed out, even the lowest wind speeds expose the acoustic inlets of the microphones to an aerodynamic excitation, and this is all the more parasitic with lower frequencies. A recording possessing these qualities innately provides the connected acquisition unit with the best signal-to-noise-ratio (SNR), i.e. potentially the largest capacity of source separation and recognition in the subse-quent levels of signal treatment and intelligence. In this sense, there is an analogy with image recognition, where, the clearer an image is, the easier the recognition is.

The sought-after wind immunity must first be achieved on the smallest scale. That is at the level of the microphonic transducer(s), followed at a larger scale by the incorporation of the transduc-er(s) in the body or the head of the microphone (often referred to, by extension, as "the micro-phone"). And at last, when a passive external protection device is included, as well as its basic electronic equipment (power supply, amplification, filtering, symmetrisation, cabling, etc.), this whole integrated assembly forms eventually what is called "the sensor".

Under this macroscopic view of the sensor, the op-timization of the sound recording’s engineering often lies in the fine adjustment of all those aerau-

lic-electro-acoustic characteristics (compliances, resistances and mechanic, acoustic, aeraulic and electric inductances). Seen from the outside of a sensor and while in-cluding most of the time a supplementary external protective envelope against the elements (UV, ozone, hydro and litho meteorites, frost, ice, lightning, etc.) and other nuisances (birds, rodents, etc.), numerous kinds of solutions can be considered and combined, ranging from a simple wire-meshed shield to more important structures (like pipes, fences, etc.). The acoustic inlet of a sensor can for example consist of one or several pierced hoses with a length of several metres, placed in a star-shape in order to produce the effect of a spatial filter. In another example the microphone is placed in the centre of a several metres high octagonal wind fence. In a third exam-ple the microphone has been made extremely sensitive by deploying giant membranes (reciprocally used speakers) and wind protection has been dealt with by burying them in the ground, under the snow cover.

Extensive testing and comparing of different solutions have led to the conclusion that each and every imaginable integration and implementation reaches a certain performance limit which is almost identical in terms of sensitivity to infrasounds and wind immunity. Eventually the following four principal contingencies guide the defi-nition of the implantation:

1. Searching a site with the least possible exposition to dominant winds (concavity of the relief, forest, orientation, etc.)

2. Obtaining a power supply solution for the sensors and the con-nected system, keeping in mind that the application must almost always be able to record signals for a long period of time and often without access to a permanent power supply.

Buried sound-recording

Sound recording under plant-cover

Sound recording under a protective windshield


11/31

Infrasonic wave captured by 4 microphones, which are about 30 m apart

3. Acquiring long-term resistance to weather conditions, since besides occasional applications, infrasonic sensing most often lasts for longer pe-riods (years or decades) in hostile conditions (temperature, wind, radiation, precipitation, frost, sand, etc.) or likely to create various accu-mulations on the sensors (snow cover, sand) or their acoustic inlets to (debris, dead leaves, ice, etc.).

4. The possibility of construction or the set-up of fixating structures for the sensors on the terrain, since sensing applications have to be able to op-erate on various terrains.

In summary, the main qualities that are expected from an infrasonic sound recording are, in order of im-portance: the sensor’s maximum global immunity to wind, maximum acoustic sensitivity, maximum resistance to external conditions and power consumption.

For the principal quality of immunity to wind, our research has so far identified one solution as the one yielding the most benefits. This solution combines two techniques, one passive technique consisting of a microphone-head equipped with a tranquilisation chamber (design of the sensor itself) and a second tech-nique, of signal treatment this time, which is called SCOH3 (‘sonogram of cubic coherence’) originally suggested by Bédard1, it applies to the simultaneous use of the signals of several microphones. This tech-nique is therefore not applicable with one single microphone and as a consequence, before even speaking of beamforming in the sense of directional acoustics (determination of the direction of the propagation of an acoustic wave) this technique justifies the need of a sound recording with at least two sensors. In sum-mary, the biggest benefit in signal-to-noise is obtained on the one hand by the optimization of the sensor itself and on the other hand by the exploitation of the assembly of the signals of several sensors.

The effect of ‘whitening of the parasitic components due to wind in an infrasonic signal’ can be explained as follows: given the fact that the signal-of-interest that we are looking to capture with the greatest possible sensitivity is, by definition, an infrasound which propagates (i.e. an acoustic wave), whereas parasitic noises from which we seek im-munity are the effects of local turbulent frictions on each sensor (by situating themselves in the same bandwidths as those of the infrasound-of-interest), the information that is carried by the elec-tric signals from each sensor will be a superposition of these com-ponents.

Yet, and provided that the sensors are not too far apart considering the dimension of the pattern of the wave (the length of the wave of interest, for example approximately 34m at 10Hz), but far enough apart considering the dimension of the phenomena that produce parasitic noises (for example, 10 times the size of a sensor, or typi-cally 1m for a sensor of 10cm), in the electric signal in which the two components are superposed, the component of the wave is barely deformed on each of the signals from the four sensors, while the component of the noises is rather different on each of the signals of the four sensors because it are caused by a local turbulence at each sensor.

1 A. J. Bedard, G. E. Greene, J. Intrieri, and R. Rodriguez, “On the feasibility and value of detecting and characterizing avalanches

remotely by monitoring radiated sub-audible atmospheric sound at long distances,” presented at A Multidisciplinary Approach to Snow Engineering, Santa Barbara, CA., 1988, E. F. Conference, 267-275; Hoplinger, 1983

Test of a wooden windshield, Bonneval-sur-Arc, France, 2003


12/31

It is this fundamentally different physical phenomenology of the two components (one local at each sensor, the other propagating with very little deformation on the four sensors) which allows them to be effectively separated by basic signal treatment functions. Thus, the signal of the wave of the noisy signal is extracted, operating the functions of antennae equipment, then identification and finally recognition, in parallel (since the two things are independent and do not pursue the same goal), as we will see in the next chapter.

Elimination of the noise caused by the wind with the SCOH3 technique (red areas in the upper graph): the system automatically eliminates all the uncor-related noises, so that only the red area produced by an infrasonic activity subsists. Here the natural avalanche of La Met on February 18th 2000, at 11h58. At the top, all the other high amplitude contributions are perturba-tions from the wind on the acoustic inlets


13/31

Beamforming, identification and classification

In summary, beamforming refers to the techniques used to determine the direction of a wave, and the ensuing possibilities to locate the source or determinate a trajectory by triangulation between several directions. Just like the electro-acoustics of microphones, beamforming is a separate field of applied acoustics. In this document we will only explain those aspects that are to be taken into account for the choice of typology of a means of infra-sonic recording or surveillance.

Since beamforming determines in which direction the source of an observed wave can be found, or, more precisely, its direction of arrival (DOA), seen from the array of sensors, we speak about vectorised acoustic signals (acous-tic pressure, direction) rather than scalar signals (acoustic pressure). This point is capital when it comes to recognizing the potential origin of a wave (source classification), since its directional characteristics (for example evolu-tion of an elevation and an azimuth every n seconds) establish an indicator of the apparent trajectory of the source. This trajectory very often provides information about the eligibility of a certain source-type at the origin of the

observed wave rather than a certain other source-type. The perceived trajectory of an aeroplane, a mete-orite or an avalanche as well as the distance from the source to the system, produce very different trajectory signatures indeed.

This is where beamforming and recognition (i.e. automatic classification) are so very complementary for medium or long range detection of events.

The capacity to attribute a label corresponding to a more or less precise description of the detected event (for example ‘aircraft’, ‘airplane’, ‘airplane of this type’, ‘airplane this manoeuvre’ or yet ‘rock slide’, ‘ava-lanche’ ‘powder avalanche’, ‘artificially triggered powder avalanche’, etc.) to signals picked up by the sensors, is the complementary result of an acoustic signature on the one side and the analysis of the acoustic trajectory on the other side, whichever algorithmic analysis is used, since all these techniques are henceforth brought together under the designation artificial intelligence.

These considerations finally lead to the following contours of an optimal acoustic sensing solution. Firstly, for a sound pick-up with maximum immunity to wind, which allows a determination of the propagation direction of the wave over the observation system, the system should be composed of several sensors

which are separated from each other by a greater distance than those of the local aeraulic turbu-lences and smaller than the biggest wavelength (several me-tres to several dozens of metres). In practice, such a system would dispose of at least four sensors, or the minimal number to allow the unambiguous determination of the propagation direction of a wave (azimuth and elevation of the wave by the determination of offset be-tween six pairs of sensors set up like that).

Microphonic tetrahedron anten-na with 1 m long ribs, used for the localization of aircraft trajec-tories.

Global principle of an acoustic beamformer: the signals captured by spatially distanced microphones are weighted and time-shifted in order to obtain focali-zation of the antenna in a preferential direction.


14/31

Secondly, the system must allow the local registration or transmission to a distant recorder of the raw signals or failing that, a reduced version of the raw signals, like segments or an envelope for example.

On this infrasonic recording, lasting 6 minutes, it is shown that in a whole central part, the incidence direction of the captured infrasonic wave is very well determined and varying monotonously, whereas before and after no stable incidence direction is detect-ed. This type of behaviour indicates the occurrence of an infrasonic event in the sense that for a certain period of time, the energy of the signal is sufficiently coherent to allow the determination of a solidly determined directiond’une direction.

Jan. 24th, 2005: 7 visible events within a 10 minutes observation window: helicopter flight, Lama or Alouette turbine type (17.5 Hz), with visible Doppler frequency shift effect indicating non stationary flight sections; second helicopter, stationary; series of explosions at 15h14m35s, azimuth 239°, 15h16m28s, azimuth 264°, other unmapped explosions at 15h12m10s, 15h14m12s, 15h15m09s; avalanche at 15h16m36s, duration about 1min46s, potentially in sector 5 with extension down to the road.


15/31

Triangulation, range

When the infrasonic sensing requires a more precise localization than a simple indication of the direction, the solution with two associated distant systems allows, when triangulating the directions produced by each of the two systems, not only for the extrapolation of the apparent localization of the infrasound at the point where two directions cross each other. It also takes into account the propagation durations between the considered points (retro-propagation taking into account the differences in travel time for a source), by means of temporal readjustments if necessary. For example, when an aircraft moves at a great speed, close to two stations, without being equidistant to them, the triangula-tion cannot be executed precisely by com-comparing the two received directions at a given moment, since the directions, sim-ultaneously received by the two stations, do not correspond to one and the same emission, but to two different emission positions (the source has moved between each of the two emissions). In another situation, for an infrasound whose origin is situated at several hundreds of kilometres from two listening stations, which are in turn several hundreds of metres removed from each other, the directions produced by each of these two stations would be practically parallel, rendering the extrapolation of the localization based on these two directions particularly imprecise. We are thus able to prognosticate the infrasonic event direction "doubly" well but the prognosis of the distance of the event (at which dis-tance the origin of the event is situated) would comprise an important incertitude. If, on the contrary, the two stations had been placed a couple of dozens of kilometres apart from one another, the localization would have been much more precise. Inversely, a good angular discrimination and consequently a good localization can be obtained for two stations which are close to each other and equally close to a source of infrasound.

These basic considerations belong of the more global field of infrasonic goniometry (goniometry = meas-urement of angles) and the propagation of sound at long dis-tance. This includes the notions of de sound beam curves in the atmos-phere (especially gradients of temperature and effects of the wind), overlap between a wave and its reflections, loss by geometrical absorption and in the air, and the effects of obstacles and reliefs. With-out going into the details of these questions here – extensive literature is available on these issues - but giv-en the fact that these aspects depend in the first place on the con-nection between the wavelengths and sounds of interest and the dis-

In this example, the intersection (triangulation) of the obtained directions by respectively 2 infrasonic surveillance stations at a distance of 600m, allow for the direct localization of the position of the source of the infra-sonic wave emission (here a helicopter).

In this other example of acoustic localization by beamforming, the emission source of the sound wave (shockwave of a transitory explosion signal) is ob-tained by a deformable antenna whose acoustic inlets are about a hundred metres apart.


16/31

tances of interest, and keeping in mind that that we restrict ourselves in this document to sound frequen-cies in the range of 1 to 40 Hz, the most essential consideration to be taken into account when defining a sensing solution is to find a way to immediately place its application in a range of interest. In reality an in-frasonic sensing system has no range since it is only passive (no emission). Nevertheless, one can define the range of a system as the radius, outside of which one can exclude the elements that come from it, for ex-ample the scale of a mountainous relief or the perimeter of an industrial facility, up until a wider zonal, territorial or geographical perimeter.

To get an idea, one can say that ranges from several hundreds to several thousands of kilometres – for very powerful events – are realistic in the biggest wavelengths of our frequency range of interest (1 Hz, or ap-proximately 340 m at 20°C), and that these ranges are reduced to just several dozens to about a hundred metres at the other extremity of our frequency range of interest (40 Hz, or approximately 8.5 m at 20 °C).

Thus, a sensing system 1Hz-40Hz could for example detect the infrasonic signature of a part of a flight of a supersonic aircraft at a distance of thousands of kilometres (for example: detection by an Alpine ski-resort of an infrasonic emission of the supersonic-subsonic transition of the commercial flight New York – Paris), but only in favourable detection conditions in the atmosphere. It would however be incapable to detect the infrasonic signature of a powder avalanche, even of an important size (several millions of m3), situated at only a dozen or several dozens of kilometres. In the first example, the circumstances that are most in fa-vour of detection are typically summer-like atmospheric conditions whose temperature gradients are in favour of long-distance southward propagation of the components close to 1 Hz. In the second example, the atmospheric effects occur not very often but the conditions of direct view or reflections and masking by the surrounding reliefs are predominant.


17/31

Optimal Basic Typology for Local Sensing

In all the cases where local sensing is the first choice, i.e. that one is inter-ested in the surveillance of infrasonic events in a radius of typi-cally several kilometres around a central point (natural gravitational hazards for example), or along a line of interest (communication line or circumference of a site), the most generic solution which is neverthe-less particularly effective, based on our experience, can be described as follows. It is a system formed by four to eight sensors placed on the ter-rain in a three-legged star-form, i.e. one central sensor and three pe-ripheral sensors placed at 120° at a distance of about thirty metres from the central microphone, plus eventually three other intermediate sensors (the three external sensors provide the basic precision of the antenna and the three optional intermediate sensors provide an in-creased capacity to lift ambiguities). This system is, in comparison to a station that consists of only one

sensor, first of all, far more immune to parasitic perturbations caused by wind (limited immunity, like explained previ-ously, very quickly becomes a major limitation to the sensing), and secondly able to provide the information of direc-tion of the provenance of the infrasound. As already pointed out, this last point is a considerable advantage in the context of local detection, since the characteristics of the site of interest are generally well known and charted, as the word ‘local’ suggests. Most of the time, it is possible to link an effect to a cause, which makes the classification system more and more reliable in its prognoses.

As far as the precision of the localization is concerned, the question often arises which precision can realis-tically be expected from an infrasonic goniometer: in general, one can say that the precision of the determination of the direction of an incidental wave by an acoustic antenna depends above all on the size of the antenna – that is to say on the distances between the sensors – compared to the dimension of the incidental wave, as well as the biggest extension of the antenna seen from the direction from which the wave arrives.

Two sensors that are for example very far apart but aligned in the same direction as the propagation of the wave, will be seen by the wave as very close to one another, even coincidental (extension of zero). On the contrary, if the two sensors are placed transversally to the arrival direction of the wave, their extension will be equal to the distance that separates them. As this effect translates itself on the horizontal plan (azimuth) as well as on the vertical plan (elevation), ideally the four basic sensors of the antenna have to present the largest possible extension, seen from all directions. The ideal geometry consists of a regular tetrahedron

In the case of the application of infrasonic surveillance on the avalanche activity, we are particularly interested in monitoring the corridor close to a slope over-hanging a segment of an exposed transport route.

In general, a local surveillance system of a site requires the implantation of several sensors linked to a central unit with a power supply and if possible, a wired or wireless communication line.


18/31

with four summits. In the solution that we recommend, the three peripheral sensors, each about thirty metres separated from a central sensor, form precisely the triangular base of a tetrahedron. That being said, the matter left to deal with in order acquire this ideal geometry, is the height of the central sensor. If we require an equally accurate precision in elevation as in azimuth, it becomes indispensable to place the central sensor at a height of around 20 to 30 metres. This represents a real practical challenge. Numerous trials have shown that even if we manage to overcome this obstacle (for example by placing the central microphone on top of a mast), the gain in elevation precision is counterbalanced by a strong degradation of the global immunity of the antenna due to an overexposure to wind of the central sensor. Because of the global mean characteristic of the wind in relation to the ground (speed profile logarithmically increasing with the elevation), and because of the fact that the acoustic band-width of interest of 1 to 40 Hz is typically, in microphonic recording, a bandwidth which is very sensitive to air-flow turbulences, the dif-ference in parasitic noise between a sensor placed at a height of about twenty metres, as compared to other sensors placed close to the ground, is typically 20 or even 30 dB.

Fortunately, in applications of local sensing, besides particular applications like for example the precise localization of a target, a high preci-sion in elevation is eventually not very useful. For, on a site whose characteristics are well-known, from the moment a direction is precise-ly indicated, one becomes aware that in order to be able to understand the origin of the source of the infrasound, this direction is more discriminating than its elevation. Yet the preci-sion in azimuth, obtained with a horizontal extension in triangle like previously described, is typically of a degree, even with a temporal resolution of integration of just a couple of seconds. Such azimuthal precision (from several hundreds of metres to several kilometres), enables the very precise localization of an event on a relief or an installation. This basic typology can equally evolve according to the need, in order to obtain increasingly precise locali-

zation and detection prognoses. In our experience, a basic system of four sensors deserves two types of evolution when better per-formances in terms of azimuthal localization are necessary.

The first evolution concerns the basic antenna itself and consists in increasing the number of sensors from four to eight, set up in the form of two layered stars, a big and an a small one, in such a way that the big star provides the precision, whereas the small one takes away any direction ambiguities (ghost direction2).

2 It is worth noting that this geometry can equally be realised with seven sensors while keeping one and the same sensor in the

centre, but since most purchasing systems function with an even number of tracks, the eighth sensor can usefully be placed in the centre of the stars as well, but higher off the ground, in order to slightly increase the vertical extension of the antenna without however risking that the central sensor is systematically overexposed to the wind. When speaking about a function of ambiguity reduction, this directly concerns the calculation or signal treatment methods that need to be implemented in the calculation of the beamformer. They consist in determining, between to signals, the value of their time offset: no matter which technique is imple-mented, there is always a need for readjustment or offset calculation and this requires calculations of so to say inter-correlation, i.e. a function of which the maximum allows to determine a time offset. Yet, with real and non-ideal signals, the maximum of this function, calculated according to a certain integration duration, is not always entirely determined: several eligible maximums (of very close values) can occur in a certain window; we then speak of ambiguity because, according to the maximum chosen by the

In this example, a beamformer typology with 8 microphones, 2 cen-tral, 3 on an external radius of maximum extension (ex. R=30m), 3 intermediaries on a half-radius (R=15m), allows, seen from the central unit to form three sub-arrays of 4 microphones each, the one with the biggest extension providing the precision of the antenna and the one with the smallest extension providing an augmented discrimination capacity (elimination of ambiguous localizations)


19/31

The second evolution of the system consists of doubling not just the number of sensors in the system but the system itself, i.e. to set up two systems (of four or eight sensors each), thus creating the possibility of a real triangulation. In certain cases, the application situation (requested range and precision, favourable or unfavourable alignment of reliefs or installations, etc.) could require several systems. We then speak of a network of substations, each substation producing a first information independent from the others. The aggregation of these independent data then provides additional data, until the obtention for example of an integral trajectory monitoring of and infrasonic event.

In summary, it can be schematized as follows. For applications of local infrasonic goniometry, it evolves from the simplest geometry – which would be the station with four sensors in a tetrahedron constellation (that is the Arfang-SR2-4 system) – to the most precise with eight sensors in a flattened tetrahedron

(Arfang-SR2-8), then subsequently, in order to obtain a first triangulation, to a double station (2 Arfang-SR2-4 or -8), until n stations according to the size of the site and the necessary surveillance ability. From a topographical point of view, the scope of the surveillance site, brought back to the functional range of the Arfang-S2R system, requires a comparative evaluation of two solutions. The first solution consists in the implantation of one central system, maximizing the detection performance in the corresponding axis of the site, with progressive lateral degradation. This is all the more important since the criteria of distance, direct view and angular discrimination of the corridors are less and less valid. The second solution consists of two systems, with two possible varieties:

Symmetrical implantation of the two systems, either with a cover (i.e. lining) in the centre, or opting for the lateral sectors.

Implantation according to a so-called principal system, which typically corresponds to the most extend-ed side, or the one presenting most pronounced surveillance interest, and a secondary back-up system, for the rest of the considered site (typically for slopes with double orientation).

algorithm, the direction of the eventually prognosticated wave could be very different. By increasing the number of sensors of the system from four to eight and by ensuring that the distances between the sensors are distributed over two scales, one obtains combinations of trajectories between all the sensors that allow to obtain sharper and more robust prognoses of the incidence direction.

Competitive implantation principles for sites with great extension

Axi

s 1

Variant with cover

Site contour (extension n km)

Axi

s 2

Syst

em a

xe

Lateral zone(s) with light cover (1 side or 2 sides)

A) CENTRAL SYSTEM Axi

s 1

Lateral variant

Axi

s 2

B) DOUBLE SYSTEM, ON SINGLE ORIENTED SLOPE

Bac

k-u

p a

xis

Mai

n a

xis

B) DOUBLE SYSTEM, ON DOUBLE ORIENTED SLOPE


20/31

Evaluation of a Detection Performance

An automatic detection system is naturally expected to provide a detection called "true positive" when a sought-after event really occurred, and no detection (so a "negative" detection) when no event oc-curred. In other words, to "detect everything that occurs without ever giving a false alarm". A definition of the perfor-mance of an automatic detection system, like for example facial recognition, can therefore be defined by the so-called con-fusion matrix formed by the following four indicators: a percentage of true positives (TP), true negatives (TN), false positives (FP) and false negatives (FN). An ideal performance would be 100% true positives and true negatives and 0% false positives and false negatives. From our experience with infrasonic detection, we can state that the main difficulty when trying to establish these scores is to obtain observations and prove their veracity.

The following graph illustrates the result taken from a deterministic classification of a 452 day-period of Arfang-S2R surveillance (2012) of an avalanche activity of a site. The classification in this example was real-ized according to 5 criteria:


21/31

1. Criterion of global SCOH3 coherence of the event

2. Criterion of frequential composition

3. Criterion of the duration of the event

4. Criterion of the azimuthal direction of the event

5. Criterion of azimuthal direction variation of the event

For each event, the values of each of the criteria were determined and compared to acceptance markers or thresholds. In order for an event to be prognosticated avalanche, it had to meet all 5 criteria. These thresh-olds had initially been set by default. After a first winter of self-learning, they were fine-tuned in order to achieve optimal settings. The result showed that of the 44228 infrasonic events recorded during the obser-vation period of 452 days, 97.7% of events had been eliminated because of criterion 1, then 65.8%, 77.1% and 6.9% because of criteria 2, 3, 4 and 5 respectively. This eventually led to the retention of 137 events classified as avalanches, which represented a global rejection rate of 99.7%

These scores regularly progress with the evolution and implementation of artificial intelligence techniques, however in terms of infrasonic avalanche detection, we have to bear in mind that the nature of physics is such that the basic typical scores presented above, i.e. on a given observation site, can typically reach high or very high event ratios between a detectable event (i.e. presence of an propagating infrasound) and an event classified as an avalanche.


22/31

Technical Considerations

Two practical questions frequently occur, depending on the situation in which an infrasonic surveillance station must operate and even in the case of a temporary implantation. The first question concerns power supply, the second the possibilities of data transmission and consequently the necessary additional energy to charge this device if it is wireless.

These two questions are almost always linked to each other from the moment there is no infrastructure available nearby, which is often the case in natural environments. They then become the most critical point in the choice of operating method of an installation. This often leads to two possible operating methods for an infrasonic surveillance station:

Operating method, type datalogger/recorder only for applications where no communication system is available in close proximity. The complete, local recording of raw signals is then necessary, typically 40 GB/year/station to stock full raw signals of 1-40 Hz to channels in 16 bits. There should not be any constraints linked to time and technolo-gy, given the current possibilities in storage memory.

Operating method, type data streamer for applications where a wired or wireless means of communication is available. Ideally, the most comfortable option is integral streaming of the raw signals to a distant reception server, on which the signals can be archived and from which an automated service can pro-duce all necessary treatments and supply all means of visualization, alert, management, etc.

From those two options, the most appropriate architecture is the one where the untreated information that is produced by the system is a real-time data stream, which is either directly recorded by the system in the form of raw signal files or forwarded as a stream to a peripheral communication unit of any type (ex-ample: an ADSL or 3G gateway, typical speed 6 Kbits/s).

In the first case, the architecture of the hard- and software resembles a pure recorder. Its 5 watts, continu-ously available power, produced for example by a battery + solar panels can meet almost any of the needs of testing or experimental research. The main inconveniences of the implementation of this type of installa-tion are firstly, the need to deploy solar panels and regulate and charge the corresponding batteries and secondly, the fact that the exploitation of the data requires visits to the system to remove the recordings.

In the second case the architecture of the hard- and software resembles a streamer and the electric power supply may need to be increased for a continuous 3G emission, i.e. with approximately 2 additional watts, or a total required power-supply of about 7W.

Note: These two operating methods, streamer or recorder respective-ly, are at the root of the suffix S2R (two modes streamer & recorder) in the name of the Arfang-S2R system. For most applications, the streamer method is the most appropriate, since it provides its users with comfort and the informative power of live access to all data of the system. In certain cases, however, generally for temporary scien-tific applications or in environments that are far removed from any infrastructure, the recorder method is the only possible method.


23/31

History and Latest Developments in Avalanche Surveillance

Knowledge about avalanche activity on the slopes on the scale of a sector or a mountain range is important for the forecast and general management of risks in mountainous regions. For example, knowledge about the avalanche activity in certain corridors in given snow and weather situations on a site, im-proves the forecast of upcoming events in similar nivo-meteorological situation in those same corridors.

We know that this information, whether it concerns natural or artificially triggered avalanches, can be compromised, or even

impossible to obtain by visual observation only, for lack of visibility due to bad weather or darkness, masking reliefs, etc.

In the 1990s the Laboratoire d’Electro-magnétisme et d’Acoustique of the Ecole Poly-technique Fédérale de Lausanne (LEMA-EPFL) in Switzerland, with the support of Service des Dangers Naturels du Canton du Valais, looked into the feasibility of local surveillance of the avalanche activity by means of infrasonic sens-ing3.

This research was preceded by research from an organisation for exploratory research in Switzerland in the larger area of high-infrasonic sensing 1 Hz – 40 Hz, which had confirmed, as already consid-ered by research by the National Oceanic and Atmospheric Administration (NOAA)4, that avalanches, by the masses that are set in motion, induce a trem-bling of the air and the ground, which then propagates in the form of a seismic wave and an acoustic wave.

3 "Acoustic detection system for operational forecasting", International Snow and Science Workshop, Banff, Canada, 1996; "Snow

avalanches: automatic acoustic detection for operational forecasting", Forum Acousticum 1996, EAA Convention, Antwerpen, Bel-gium; "Détection acoustique des avalanches", Symposium International Sciences et Montagnes de Chamonix, ANENA, 1995; "Microphones pour la détection d'infrasons en haute montagne et par conditions hivernales", Journée d'étude sur les Trans-ducteurs et capteurs en milieu hostile, Groupe Electroacoustique de la Société Francaise d'acoustique, Paris, 1995; "Annals of Glaciology 26 1998 Infrasonic monitoring of snow avalanche activity: what do we know and where do we go from here?", V. Adam, V. Chritin, M. Rossi, E. Van Lancker. 4 A. J. Bedard, G. E. Greene, J. Intrieri, and R. Rodriguez, “On the feasibility and value of detecting and characterizing avalanches

remotely by monitoring radiated sub-audible atmospheric sound at long distances,” presented at A Multidisciplinary Approach to Snow Engineering, Santa Barbara, CA., 1988, E. F. Conference, 267-275; Hoplinger, 1983.


24/31

A first experimental campaign, undertaken during the winter of 1994-1995 by LEMA-EPFL on the Anzère site, in the centre of the Valais region in Switzerland, allowed for the detection and localization of 22 artifi-cially triggered avalanches in a radius of between 0.3 and 5 km, with measured acoustic levels of 40 to 90 dBSPL in the frequency range of 1 to 10 Hz. The four used microphones, of an electrostatic type, were placed on top of masts in a protective windshield. The tops of the four masts were separated 50 m from one another and formed an ele-vated horizontal level in relation to the ground. At a sound speed of about 340m/s, it takes 15 hundredths of a second for an acoustic wave to propagate from one microphone to the other. This equipment allowed the confirmation of the emission of infrasounds by the avalanches, and the demon-stration of the possibility to localize the active passes at a distance. However, this first disposition had the inconven-ience that masts were required, it was also hindered by strong winds which can induce important infrasonic noise.

For avalanches typically situated at a kilometric distance, the infrasonic intensity typically varies between 30 and 90 dBSPL in the 1 to 10 Hz bandwidth. The weakest recorded levels correspond to avalanches that barely emerge from the infra-sonic noise. The duration of the infrasonic emission in seconds goes from several seconds to 1 minute. The maxi-mum number of occurrences corresponds to durations from 15 to 25 seconds. The longest emissions have been observed until 2 minutes. As far as artificially triggered avalanches are concerned, the time that separates the explosion from the infrasonic emission is in general a dozen of seconds. Steep slopes, proximity and direct view favour the precocious emission of infrasounds.

As a general rule, the avalanches in the same corridor present similar spectral characteristics. Frequencies inferior to 2 Hz are only present during very large avalanches, frequencies superior to 6.3 Hz are recorded close to the avalanche (distance < 1.5 km) in direct view and concerning the bandwidths with higher energy they range from 1 to 6.3 Hz for large, far removed avalanches (> 1.5 km) and from 2 to 12.5 Hz for small, close avalanches (< 1.5 km).

There generally is a close correlation between the size of the avalanche and the acoustic energy it produc-es. The size of avalanches is a difficult parameter to estimate but globally, they can be classified in three categories: small, intermediate and large.

Small avalanches

Large amplitude avalanches


25/31

On the one hand, the width of the avalanche varies and, on the other hand the height of the snow cover carried into the flow area is far below the height of the fracture. However, the snow cover quantifies the size of the avalanche well enough to deduce the parameters influencing the infrasonic level from it.

The comparison of the levels per corridor (previous page) shows that this level increases with the size of the avalanche, identically for each corridor. The study of the topographical data for each corridor allow for an improvement of the interpretation. At a constant volumetric index, the observed infrasonic level on the goniometer increases when the corridor is nearby and in direct view. As far as the duration is concerned, it provides information about the temporal continuity of the event and facilitates the decision.

The straight lines of equal speed corre-spond to the average speed of the avalanche if it would have covered the distance in the same time as the dura-tion of the emission. The duration of the emission is however always inferior to the real duration of the avalanche (contrary to the seismic emission). The following lessons can be learned: the results, for all corridors, follow the same pattern. The upper limit of the duration of the emission increases when the length of the avalanche in-creases and corresponds to the real duration of the avalanche. The lower limit increases when the length of the avalanche increases. As for the emis-sion level, a specific behaviour can be observed for each corridor. When the length of the monitored corridor is known, we can estimate the minimum and maximum duration of the emis-sions. Long emission durations, corresponding to very low speeds, prove that another emission can occur after the front of the avalanche has reached the deposition zone. This is the case, for example, when important masses of snow that form the ava-lanche’s body, still jump the over the rock barriers, whereas the front is al-ready on the ground at the end of its path.

In another case of experimental appli-cation, we evaluated the detection capacities of purging avalanches by a fixed gas installation as well as by hand-held or helicoptered solid charg-es, with 437 recorded events between November 11th 2012 and April 9th 2013, 190 of them spread in fixed di-rections and 302 gathered in 7 main directions. Here, a basic Arfang-S2R system with four microphones proved its ability to identify and localize the

Graphic extract of a database of 437 infrasonic events, classified as ava-lanches on the Arfang-S2R station in La Giettaz (France), winters 2011 to 2013


26/31

periods with an increased number of artificially triggered avalanches within a perimeter including a surveil-lance zone with a range of 4 km.

Another example of use, of which the performances are extremely well-mastered, is the case hereinafter in which a variant of the Arfang-S2R was implemented in a so-called ARTEX form which consisted in placing the sensors on the terrain, not in a star-shape but on the fall-line under a fixed installation. With this im-plantation, the possibility to localize is removed, it allows however for a very sensitive proximate detection of the result of the triggering at the moment of each detonation from the installation.

Example of an infrasonic acoustic signature collected by a linear variant of the Arfang-S2R system, in which the four sensors are placed downhill from a fixed installation for the artificial triggering of avalanches at respectively 10, 20, 40 and 80 m from the mouth of the installation. The first signature (above) includes a trace of reverberation of the detonation in the tube whereas the third signature gives evidence of the movement of the snow cover, which manifests itself the clearest on the sensor which is the most downhill of the system.


27/31

A more macroscopic information which can be drawn from the knowledge about a site coming from local, long-term, infrasonic surveillance is the notion of the increase of the number of events. The following ex-ample presents the statistics of the number of avalanche events and the shots for artificial avalanche triggering during a 5 month winter period in the ski resort of La Giettaz, France, 2012-2013, as well as a map identifying the directions of 563 shots noticed during that same period (in white unique simple shots, in blue multiple shots gathered under a same azimuthal value with a tolerance of ±2.5°).

La Giettaz, France, Winter 2012-2013


28/31

In the following graph (Taesch-Zermatt, Switzerland, 2006-2007) the representation of the increase in numbers is compared to a measurement of the referential snow-height on the ground on the relevant slope, as well as the average speed per hour gathered at the same referential location.

Finally, the two following figures present, on an extract of the existing Arfang-S2R database, an example of the fine-tuning of the classification, according to the latest artificial intelligence technology which is actually being implemented in the service.

Taesch-Zermatt, Switzerland, 2006-2007


29/31

***

Arfang-S2R’s raw signals database contains so far approximately 15 years of continuous re-cordings, on 7 main pilot sites spread over the Swiss and French Alps, plus a site in the Himalaya. It totals approximately 200'000 infrasonic events, several thousands of which are de-tections of detonations confirmed by observation, several thousands of avalanches, about a hundred of which are confirmed by a direct observation and several hundred confirmed by an a posteriori observation.

The authors wish to thank the key persons and institutions who have lent their support and ex-pertise to 25 years of intensive experimental field-research and development.


30/31

Bibliography

[1] Bolognesi R., Buser O., Good W., (1994). Local avalanche forecasting in Switzerland: strategy and tools. International Snow Science Workshop, Snowbird, UT, USA.

[2] Chritin V., Rossi M., (1995). Détection acoustique des avalanches, Site La Sionne - Anzère, Valais, Suisse. Symposium International Sciences et Montagnes, Chamonix.

[3] Hejda D., (1995). Caractérisation acoustique des avalanches. Mémoire de diplôme, LEMA-EPF-Lausanne.

[4] Rossi M., Chritin V., (1995). Microphones pour la détection d'infrasons en montagne. Journée d'étude sur les transducteurs en milieu hostile, Groupe électroacoustique de la Société Française d'Acoustique, Paris.

[5] Chritin V., Rossi M., (1995) Actes de la Journée d'étude sur les Transducteurs et capteurs en milieu hostile, Groupe Electroacoustique de la Société Française d'Acoustique, « Microphones pour la dé-tection d'infrasons en haute montagne et par conditions hivernales ».

[6] Chritin V., Rossi M., (1995) Actes du Symposium International Sciences et Montagnes de Chamonix, ANENA, « Détection acoustique des avalanches ».

[7] Chritin V., (1996) Actes du colloque mensuel IFENA, Suisse, « Emission acoustique des avalanches... ».

[8] Chritin V., Rossi M., Bolognesi R., Forum Acusticum, Antwerpen, Belgium, April 1-4, 1996, Proc. Acta Acustica, Supplement 1, pp S173, Janvier / Février 1996, « Avalanches : détection acoustique au-tomatique pour la prévision opérationelle ».

[9] Bolognesi R., Chritin V., Rossi M., Proc. of International Snow and Science workshop, pp 149-153, Banff, Canada, 1996, « Acoustic Detection System For Operational Avalanche Forecasting ».

[10] Robert A., (1996). Caractérisation acoustique des avalanches. Mémoire de diplôme, LEMA-EPF-Lausanne.

[11] Chritin V., Bolognesi R., revue NXLOG 1996, « L'acoustique appliquée au service de la prévision ré-gionale des risques d'avalanches ».

[12] Adam V., Chritin V., Rossi M., Van Lancker E., Proc. International Society of Glaciology, pp 324-328, Symposium on Snow and Avalanches, Chamonix, 1997, « Infrasonic Monitoring Of Snow Avalanches: What Do We Know, Where Do We Go From Here? ».

[13] Adam V., Chritin V., Rossi M., Van Lancker E., (1997). Aerial acoustics emission from snow ava-lanches: what do we know, what don’t we know, where do we go from there? International Symposium on Snow and Avalanches, Chamonix.

[14] Chritin V., Adam V., Rossi M., Bolognesi R., Actes de la Table Ronde Européenne Avalanche Control Saas-Fée’97, 1997, « Acoustic Detection System For operational Avalanche Forecasting ».

[15] Van Lancker E., Rossi M., SSA, DAGA’98, Zurich, Switzerland, pp 142-143, 1998, « Acoustic Goniome-try Antennas and Algorithms ».

[16] Chritin V., Rossi M., Revue ANENA Neige et Avalanches N° 78, pp 2-7, Juin 1997, « A l'écoute... des avalanches ».

[17] Van Lancker E., Proc. of the 5th French Congress on Acoustic, Presses Polytechniques et Universi-taires Romandes, Lausanne, Switzerland, vol. 1, pp 603-606, 2000, « Goniométrie acoustique et estimation des différences de temps de propagation ».

[18] Van Lancker E., Chritin V., Revue ANENA Neige, et Avalanches N° 96, pp. 10-14, 2001, « Des ava-lanches sous bonne surveillance.


31/31

[19] Van Lancker E., Thèse EPFL N°2408, Presses Polytechniques Romandes, 2001, « Acoustic Goniometry, a spatio-temporal approach »

[20] Van Lancker E., Chritin V., Revue Neve e Valanghe AINEVA, Neve 51/8, 2003, « Valanghe sotto con-trollo ».

[21] PROCEEDINGS OF SPIE, SPIEDigitalLibrary.org/conference-proceedings-of-spie, SPIE International Society for Optics and Photonics, SECURITY + DEFENCE Conference – Edinburgh, United Kingdom, 26-29 Sept. 2016. Target and Background Signatures Transient acoustic detection for hostile fire indica-tion for helicopters, Chritin, Vincent, Van Lancker, Eric, Wellig, Peter, Ott, Beat, Perseguers, Sébastien, et al.

[22] PROCEEDINGS OF SPIE, SPIEDigitalLibrary.org/conference-proceedings-of-spie, SPIE International Society for Optics and Photonics, SPIE Optics + Photonics 2020 Digital Forum – 24 Aug. - 4 Sept. 2020. Target and Background Signatures High infrasonic goniometry applied to the detection of a helicop-ter in a high activity environment » V. Chritin, E. Van Lancker, Peter Wellig, Beat Ott.

Applied acoustic goniometry

Technical Addendum

Documents