Locating explosions, volcanoes, and more with infrasound

The way length scales are described varies widely among acousticians. In the infrasound community, "global" implies ranges greater than 1000 km, "regional" implies ranges of 100–1000 km, and "local" implies ranges less than 100 km. Scales less than 10 km have recently been labeled "tactical," a term borrowed from the military. One expanding area of research for our group is the localization of tactical sources.

Consider a pair of omnidirectional microphones and a distant source, all at ground level. Provided the speed of soundc is known, measurement of the time delay between the sound arrivals at each microphone gives direction in the ground plane θ from which the sound arrived, up to a twofold ambiguity. If c is unknown, the ambiguity increases because different combinations of θ and c yield the identical time delays. Humans, admittedly, use only two ears to locate sources of sound, but we mitigate that design flaw by using perceptual clues and the fleshy parts of our ears to extract directional frequency content.

A third microphone not collinear with the first two but still on the ground would allow for an unambiguous determination of θ and the sound speed. With a fourth microphone not on the ground, one can deduce θ, the magnitude of the sound velocity, and the elevation angle of a source not at ground level.

Due to the long wavelengths associated with infrasound, the microphones typically need to be separated by kilometers. Building kilometer-tall, stable towers to house microphones is difficult, so in practice, arrays are placed on the ground. At kilometer scales, although an array is not strictly planar, it may be assumed to be planar for real-world applications. Even with planar arrays, field workers often deploy more than three sensors. After all, the infrasound environment is a noisy one, and the additional sensors enhance the estimates of θ and c.

Directional information is important, but so too is a knowledge of how far a source is from an array. The most commonly used technique is illustrated in panel a of the figure; a group of three non-collinear arrays enables triangulation of the source location based on individual arrival angles. The figure does not address a well-understood complication that must be taken into account when applying the triangulation technique: The widely separated arrays lie on Earth's curved surface, not on a plane.

As in individual arrays, redundancy is desirable in a system based on triangulation. Consider, for example, CTBT arrays, which are separated by thousands of kilometers. High winds at one station may drop the signal-to-noise ratio to an unacceptable level and render the station useless. Or winds aloft and atmospheric temperature gradients may duct the signal away from a particular station. Our field experience measuring low-frequency infrasound has shown that at tactical ranges, worries associated with ducting and topography can often be ignored. Near-field localization, it seems, is well suited to triangulation.

When our group began analyzing signals from our first CTBT array in Windless Bight, Antarctica, we recognized that many of those signals came from Mount Erebus, a volcano 28 km from the detector. Volcanologists are interested in such signals, which provide information about the character of eruptions. About halfway between our array and Mount Erebus is another constant source of infrasound: Vee Cliffs, where continental ice spills onto the Ross Ice Shelf. For glaciologists, signals originating from the cliffs can indicate rates of ice flow and fracture. We could not always separate the volcano and cliff signals, nor did we have the resources to place additional arrays on the ice. Therefore, we had to come up with a way to estimate range using a single array. Some techniques, often found in the seismology literature, allow an inference of range from wavefront curvature. Unfortunately, the noise present in our infrasound data was enough to defeat those methods.

One of us (Arnoult) did his doctoral work in astrophysics before studying infrasound. Drawing on that background, he considered a planar array in spacetime. As shown in panel b of the figure, the array's sensors, stationary on the ground, move vertically in time on parallel world lines. An acoustic point source radiates a spherical wavefront that traces out a four-dimensional sound cone in spacetime. Arrival times give the intersections of the sensor world lines with the sound cone. Enough of those intersections allow for a determination of the sound cone; its vertex is the source location.

A necessary condition for the location to be uniquely determined is that a planar array have four sensors, no three of which are collinear. Given enough sensors, the spacetime-cone technique will allow a fully 3D localization, even when a constant wind blows, in which case the sound cone is oblique. In actual operation, we deploy at least six sensors roughly in a plane and ignore the effects of local winds.

When a source is sufficiently far from a single detector, time delays will no longer provide enough information and the assumptions underlying the spacetime technique begin to break down; triangulation becomes the method of choice. Otherwise, both methods are viable. To see how they compare, the other of us (Szuberla) ran a numerical simulation. To our surprise, the spacetime technique gave an order-of-magnitude improvement in localization precision. A month later, as luck would have it, a visiting program manager expressed interest in getting more precise acoustic locations of small explosions detonated at distances ranging from 100 m to 10 km; here, "small" means equivalent to less than 200 kg of TNT. An impromptu laptop presentation of the simulation results quickly led to funding for a field test (the online version of this Quick Study includes photographs) in which we verified the order-of-magnitude precision improvement under realistic conditions with wind and clutter (coherent signals unrelated to what we were trying to localize).

Since that first test, spacetime localization has pinpointed both explosions and continuous, single-frequency signals. In one instance, Arnoult noticed a long-lasting signal interrupted for a few minutes by various popping sounds. With the help of his GPS, he drove out to the coordinates of the source and found the smoldering ruins of a house—happily, no one was living in it at the time. Fire-induced turbulence, it seems, gave rise to the main infrasound signal, and the pops were caused by exploding propane cylinders.

Most of the current interest in infrasound research was spurred by the CTBT. Indeed, current research in near-field infrasound localization is primarily aimed at defense-related studies. But the clandestine atmospheric tests that the CTBT infrasound arrays are designed to detect do not appear to be a threat in the current geopolitical climate—the four tests since 1996 have all been underground. In their PHYSICS TODAY article of more than a decade ago, Alfred Bedard and Thomas Georges perceptively stressed the promise of non-treaty-related infrasound research. Nowadays, scientists and engineers are increasingly interested in geophysical problems such as localizing active vents on volcanoes or civil problems involving persistent industry-generated infrasound.

As befits its connection to the CTBT, infrasound research has historically addressed acoustic phenomena at ranges exceeding thousands of kilometers, and that long-range work is still underway. Will long-range research sustain the current interest in infrasound, or will shorter-range applications come to the fore? We await future studies with interest.

Locating explosions,volcanoes, and more with infrasound.

Curt Szuberla is an assistant professor of physics and Ken Arnoult is a research scientist, both at the Geophysical Institute, University of Alaska Fairbanks.