Page 12 - January 2006

barometers that measure the absolute or differential pres- sure variations in a frequency band of roughly 0.02 to 10 Hz. These instruments, acting as extremely sensitive micro- phones, measure the minute pressure fluctuations associat- ed with acoustic energy propagating through the atmos- phere. The pressure fluctuations reach the microbarometers through spatial filters that are designed to suppress wind noise by spatial averaging. Most wind noise is incoherent across distances of several meters, while infrasonic signals can be coherent over distances in excess of 100 meters. The wind filter averages pressure fluctuations over an area larg- er than the coherence limit of noise but smaller than the coherence limit of signals. In this manner, the effects of wind-induced noise are minimized while the signal is pre- served.
A common form of wind filter is the pipe rosette—a modern version of the linear pipe array wind filters first used in the 1950s to make long distance recordings of atmospher-
12
ic nuclear tests . A pipe rosette consists of plastic or metal
pipes arranged in a hub-and-spoke configuration (Fig. 2). The outer ends of the pipes are open to the atmosphere (through screens that resemble shower heads) and at the hub, all the pipes are connected together in a summing manifold. The hubs of multiple rosettes are then connected via pipes to a central summing manifold to create a single large rosette. The microbarometer is connected to this central manifold, so it effectively averages the pressure fluctuations over the entire circular surface area covered by the rosette (anywhere from 18 m to 70 m in diameter).
Fig. 2. Partially installed infrasound site in Warramunga, Australia. The pipe rosette will be buried so that only the intakes (white “shower heads”) will be above ground and the four rosettes will be connected to the microbarometer located in the cement vault. (Photo by D. Christie).
Data from the IMS infrasound network are recorded on low noise 24-bit digitizers at sample rates of 10 to 40 sam- ples per second. Data packets from a central recording com- puter are then transmitted in near real-time to national and international data centers using satellite communications systems. Local meteorological conditions and state-of- health data channels are transmitted along with the infra- sound data.
A typical infrasound station actually consists of an array of microbarometers providing improved signal detection and processing capability over that of a single sensor. There are typically four to eight individual microbarometers compris- ing an array, and each is connected to its own wind-filtering pipe rosette. The spacing between microbarometers is any-
where from tens of meters to one or two kilometers. Infrasound arrays are typically used to observe signals from sources hundreds to thousands of kilometers away from the station. Detection of weak signals from distant sources requires a sophisticated approach to data processing. The sig- nals from the array of microbarometers are processed togeth- er, using a variety of standard techniques, such as beam- forming and frequency-wavenumber analysis that are designed to enhance weak signals and provide improved directional response. Most infrasound signal detection algo- rithms require a combination of signal power and correlation of the signal across array elements to identify a signal. Most signal detectors use strategies to avoid detector “capture” wherein strong signals from a single direction are detected while simultaneous, but weaker, signals from another direc- tion are missed. This is particularly critical in infrasound processing, since signals from noise or clutter sources can be quasi-continuous, coherent and larger than signals from sources of interest. Infrasound investigators have used a vari- ety of detection algorithms. These range from correlation detection to the F-statistic estimator3 and the Progressive
4 Multi-Channel Correlation (PMCC) method .
Once signals are detected, additional processing is used to locate and characterize the source. A frequency-wavenum- ber (f-k) analysis of the array elements can be used to deter- mine the azimuth of incoming signals (the bearing from the station to the signal source). Source location processing uti- lizes the triangulation of bearings and arrival times of signals observed at two or more stations. In some cases, the infra- sound observations are combined with seismic observations of the same source to generate a fused event location.
Source characterization is a very active area of study since the rapidly expanding global infrasound network is providing many new observations. Sources are characterized through the analysis of signal features such as duration, peak amplitude, and frequency content.
Sources of infrasound in nature
A broad suite of natural phenomena in Earth’s surface and atmosphere produce infrasound. While the sources we review in this article are natural, infrasound can also be gen- erated by anthropogenic (man-made) sources including large chemical or nuclear explosions, rockets and aircraft.
Earthquakes
Earthquakes may produce infrasound in various ways.
Acoustic-gravity waves from strong vertical ground dis-
placements can propagate thousands of kilometers from
5
their source . Under certain earthquake source conditions,
and when the shear speed of the ground is commensurate
with the atmospheric sound speed, ground-coupled air-
6 waves propagating at acoustic velocities may be produced .
Infrasonic waves may also be radiated by topography when
7 seismic surface waves travel through mountainous regions .
8
Le Pichon et al. performed the first acoustic estimates of
fault rupture speeds from the June 23, 2000 Arequipe earth-
9
quake. Olson et al. also estimated fault rupture dynamics
from infrasound signals from the magnitude 7.9 Denali Fault earthquake of November 3, 2002. Analysis of the
10 Acoustics Today, January 2006