Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
 The sudden ejection of material during an explosive eruption generates a broad spectrum of pressure oscillations, from infrasonic to gravity waves. An infrasonic array, installed at 3.5 km from the Soufriere Hills Volcano has successfully detected and located, in real-time, the infrasound generated by several pyroclastic flows (PF) estimating mean flow speeds of 30–75 m/s. On July 29 and December 3, 2008, two differential pressure transducers, co-located with the array, recorded ultra long-period (ULP) oscillations at frequencies of 0.97 and 3.5 mHz, typical of atmospheric gravity waves, associated with explosive eruptions. The observation of gravity waves in the near-field (<6 km) at frequencies as low as about 1 mHz is unprecedented during volcanic eruptions.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Infrasonic waves produced by volcanic eruptions, at short source-receiver distance (<10 km), exhibit low atmospheric scattering/dissipation and experience predictable echoing, site, or weather-dependent effects. Infrasound is also largely unaffected by cloud cover and does not rely on line-of-sight view of eruptive vents. Arrays of infrasonic sensors, deployed as small antennas or distributed at various azimuths around a volcano, show tremendous potential for enhanced event detection and localization [e.g., Garces et al., 2008]. Acoustic monitoring is becoming increasingly used by the World Meteorological Organization and by the International Civil Aviation Organization as the appropriate monitoring tool to promptly detect volcanic eruptions world wide.
 At short distances from the source, the almost constant velocity of sound (∼343 m/s at STP) makes precise localization (within a few meters accuracy) possible and allows to rack evolving sources with comparable resolution [e.g., Ripepe et al., 2007].
 Infrasound, often associated with volumetric sources rapidly expanding into the atmosphere, provides valuable insights into eruption dynamics and the state of volcanic activity. On volcanoes characterized by lava dome growth, infrasound may also be generated by non-explosive moving sources related to dome instability (e.g., pyroclastic flows) and can efficiently be used to estimate the speed and direction of propagation of these density currents [Yamasato, 1997].
 The 1995-ongoing eruption at the Soufrière Hills Volcano (SHV), Montserrat (Figure 1a), has been characterized by cyclic growth and collapse of a summit lava dome interspersed by vulcanian explosions and accompanied by lahars, pyroclastic, and debris flows. Violent explosions and dome collapses represent a constant threat to the local population on Montserrat. Over the past 15 years, the Montserrat Volcano Observatory (MVO) has progressively established a multi-parameter monitoring network that includes seismic, ground deformation, and gas measurements. In this paper we describe the installation of a permanent small-aperture infrasonic array and demonstrate its use to detect and track PF propagation. We also present unprecedented near-field observations of rare gravity wave phenomena associated with explosive activity at SHV.
2. Instrumentation and Methods
 In March 2008 the monitoring system on Montserrat was enhanced by the installation of a permanent, small-aperture infrasonic array within the framework of a collaborative effort between MVO and the University of Florence, Italy. The array, located at a distance of about 3500 m from the summit of SHV, at the St. George's Hill site (SGH), has an aperture of 200 m and a “triangle” geometry that consists of 3 satellite stations each at 100 m distance from a central receiver station (Figure 1b) and equipped with microphones sensitive 5mV/Pa in the 0.1–20 Hz band. All the elements of the array are linked to the receiver station by fiber optic cables. A 8-channel digitizer records acoustic and pressure signals (16-bit), ambient temperature (8-bit ±1°C), and battery voltage (8-bit). The central station also includes two Honeywell differential pressure sensors (±250 Pa and ±2500 Pa, respectively, in the 0.01–20 Hz). All the data collected by the array are sent via radio modem link to the MVO where they are archived and processed in real-time.
 Coherent infrasound propagating across the array is located using a grid-search for the best multichannel semblance once the signals from each array element have been delayed for the theoretical node-to-receiver travel time. The output is the propagation backazimuth for both planar and spherical wavefronts with spatial resolution of one degree [Ripepe and Marchetti, 2002].
3. Infrasound Generated by Pyroclastic Flows
 Following a 1-year period of relative quiescence, during May 2008 SHV entered a renewed phase of activity characterized by numerous small explosions. Some of these events generated ash plumes reaching up to several thousands meters above sea level, and most were accompanied by moderate-to-large size PF predominantly in the western sector of the edifice [Cole et al., 1998].
 During May, July, and December 2008, the SGH array recorded clear signals with durations of 400–1500 s (Figure 1c) associated with large vulcanian explosions and the propagation of energetic pyroclastic fronts. Coherent infrasound across the array (Figure 2d and 3b) generally started with an impulsive onset and reached peak amplitudes of 40–80 Pa during the most vigorous phase of the event. The infrasound recorded at the SGH microphones is relatively broadband, ranging from 0.4 to 7 Hz. The waveforms recorded on July 29 and December 3, 2008 exhibit remarkably similar frequency spectra (Figure 1d), peaked at 0.9 Hz. In contrast the smaller event of May 29 is characterized by a marked peak at 2 Hz. According to preliminary field observations and geological estimates of the intensity and size of the flows, we speculate that the frequency of the dominant spectral peak may scale inversely to the size of the flow and reflects the propagation dynamics and grain composition of the PF. This observation, if confirmed by additional evidence, would have important implications for real-time hazard assessment as it suggests that infrasound could be effectively used to evaluate size of PF.
3.1. Tracking the Propagation of Pyroclastic Fronts
 The largest pyroclastic flow detected and located by the SGH array, occurred at 03:38:13 UT on July 29, 2008. We estimated a backazimuth of 117°N pointing to a source in the direction of the SHV lava dome (Figure 1b). The sustained acoustic waveform lasting 197 s (from 03:38:13 to 03:41:30 UTC in Figure 2d), exhibits elevated coherency across the array and multi-channel semblance greater than 0.6 (Figure 3c). In spite of the severe weather conditions, the acoustic records maintained high signal-to-noise ratio for ∼226 s (Figure 3c). This interval coincides with the time of maximum infrasonic amplitude (∼30 Pa) and with the most intense and vigorous phase of the PF. Assuming a linear amplitude decay of the signal with distance from the source, the maximum acoustic pressure scaled at 1 m from the dome would yield a value of 10 kPa. After 03:41:30 UT, the multi-channel semblance drops to a value of about 0.4 (Figure 2e) which is still representative of coherent infrasound propagation across the array [Ripepe and Marchetti, 2002]; this observation suggests that the PF was still active although sensibly less energetic. The flow terminated at 03:46:40 UTC, ∼500 seconds after its onset, when the semblance returned to pre-explosion values of ∼0.2.
 During the most energetic infrasonic phase, the array detected several changes in the backazimuth (Figure 3d). The first shift between 117°N (i.e., the dome direction) and 132°N occurred in ∼24 seconds. This variation suggests that the PF was moving away from the dome (117°N) towards the South, in the direction of Plymouth (132°N). Assuming a source-receiver distance of ∼3500 m, the 15° variation of the infrasonic backazimuth corresponds to a minimum linear distance of 855 m. This distance was covered by the PF in 24 s, thus, at a speed of 35.6 m/s, (about 128 Km/h). The change in backazimuth direction from the dome towards SSE was observed four more times during the high-amplitude infrasonic phase, suggesting a pulsating nature of the flow. The mean velocity of the flow's propagation front varied between 30 m/s and 75 m/s, in good agreement with theoretical modelling of non-explosive gravitational PFs at SHV [Wadge et al., 1998; Hooper and Mattioli, 2001] and with other estimates for the velocity of PFs at Unzen volcano [Yamasato, 1997] and Mount St. Helens [McEwen and Malin, 1989].
3.2. Seismo-acoustic Time Delay Constraints
 The July 29 event was accompanied by an emergent seismic signal (Figure 2a) starting as early as 03:32:30 UTC, 343 seconds before the infrasonic onset (Figure 2c). The broadband frequency content (1–10 Hz) of this signal is consistent with previous measurements of PF activity [e.g., De Angelis et al., 2007]. Considering the source-receiver distance of 3500 m, only a time delay of ∼10 seconds can be ascribed to the difference in propagation velocity between seismic and acoustic phases. The larger delay between the seismic and infrasonic onsets suggests that the PF was initiated by a source lasting ∼343 s. This source shows a sharp and clear onset in the displacement seismogram (Figure 2b) at 03:32:30 UT that is not accompanied by infrasonic waves, thus, indicating that it may have been confined within the dome. We suggest that the increasing seismic amplitudes may reflect progressive failure of the shallower part of the dome leading to a minor collapse, and consequent explosion. These observations demonstrate how the integrated seismo-acoustic approach may have important consequences on our understanding of the dome collapse process as well as on the development and application of early-warning systems for PF and volcanic explosions.
4. ULP Acoustic and Gravity Waves
 During the July 28 and December 3 2008 eruptions, the two differential pressure transducers, installed at the SGH site, recorded ultra long-period (ULP) acoustic waves with a period of ∼280 s (Figure 4a), corresponding to a frequency of about 3.5 mHz. During both eruptions, the ULP signal started with a compressive onset followed by a larger decompression. The initial compressive phase for the July and December events had peak amplitudes of ∼25 Pa and ∼80 Pa, respectively. These values correspond to excess pressures of ∼0.1 and ∼0.3 MPa, respectively, at 1 m from the source. The ULP initial compressive phase moving away from the vent likely represents the early expansion of an eruption cloud in the atmosphere caused by the vent opening blast. The outflow of volcanic material soon evolved into a sustained column capable of entraining air and heating the sorrounding atmosphere during its ascent. This process may explain the large decompressive phase and suggests that the plume was fed, for about 170–220 s, by air flowing towards the dome into the high temperature column towering above the vent.
 Following this initial ULP phase at ∼3.5 mHz, the pressure signal evolved into much longer period (∼14 and 17 minutes) oscillations with frequencies of 1.15 mHz and 0.97 mHz, for the July and December 2008 events respectively (Figure 4a). These wavelets propagated for more than 1 hour after the end of the main eruptive phase (Figure 4a) and their frequency content (Figure 4c) falls within that of atmospheric gravity waves (GW). Our observations are also confirmed by microbarometric data collected within the framework of the CALIPSO project, at ∼5200 m from the dome and other locations across Montserrat (J. Gottsmann, personal communication, 2010).
 Observations of GW were previously reported at microbarographic stations in Japan associated with the eruption of Mt. Pinatubo in 1991 [Tahira et al., 1996]. Pressure perturbations with periods of ∼300 seconds were also recorded as acoustic-gravity waves for their near-sound speed of ∼320 m/s during large Plinian explosion at El Chichon in 1982 [Mauk, 1983] and Mount St. Helens in 1980 [Donn and Balachandran, 1981]. The SHV atmospheric gravity waves are, however, unprecedented as associated with smaller scale lava dome explosions and recorded as close as 3.5 km from the source. Further, the gravity waves at SHV have much longer period than any previous example.
 In a stable, inviscid, and isothermal atmosphere, density stratification provides the conditions for the generation and propagation of GW which transfer momentum and energy through atmospheric regions by displacement of the atmospheric particles. For volcanic eruptions, their generation can be ascribed to the interplay between the motion of air parcels, thermally or mechanically displaced during the explosive phase of an eruption, and the force of gravity that acts to restore equilibrium [Harkrider, 1964]. The restoring buoyancy force acting on a vertically displaced particle in any incompressible, continuously stratified fluid medium (e.g., the atmosphere) is characterized by the Brunt-Vaisala buoyancy frequency N defined by
where g is gravity, ρo a constant reference density, and ρbk is background atmospheric density in conditions of hydrostatic equilibrium. N is, in general, dependent on the height z although, for simplicity, it is often assumed to be constant (typical values in the troposphere, 0.01–0.02 Hz). At wavelengths shorter than the atmospheric scale height, (about 6–8 km) internal GW are clearly distinguished from acoustic waves by their transverse velocity field and much slower propagation (10–30 m/s). From the equation of the vertical velocity [Kundu and Cohen, 2008] a well known dispersion relation for internal GW is inferred:
where θ is the angle between the wavenumber vector and the horizontal direction. Equation (2) suggests that the frequency, ω, of the internal GW, lies between 0 and N. This reveals the significance of the buoyancy frequency as the maximum possible frequency of internal GW in a stratified fluid. Kanamori et al.  estimated typical values of 1/N of about 300 s notably lower than the values observed at SHV. Further, they proposed models of point mass and energy injection sources for the generation of untrapped internal GW. These models predict GW frequencies ≤1 mHz at distances ≥50 km from a source at the height of 10 km.
 Many discussions on the propagation of internal GW during volcanic eruptions have been confined to waves on uniform background states (i.e., constant N). However, the most dramatic GW will be found when N is non uniform, in which case wave trapping, constructive interference, and consequent amplification may occur. The 0.97 and 1.15 mHz modes recorded on Montserrat, may represent the propagation of energy trapped in the lower atmosphere [Gossard and Hookes, 1975]. The existence of an upper atmospheric boundary may effectively duct GW, preventing vertical propagation of the wave energy, hence, its dissipation. At the same time a ducting region represents an effective medium for horizontal propagation of the GW. Whilst our speculation on the mechanisms of GW at SHV are consistent with preliminary results from other independent studies (S. I. Sacks, personal communication, 2010), full modeling of their source processes and propagation effects is beyond the aim of this manuscript.
 The observations presented in this paper represent clear evidence of the infrasonic signature of pyroclastic flows and vulcanian eruptions recorded in the near-field (<5 km). Infrasound complement other geophysical measurements, particularly seismic, and assist with their interpretation by yielding information on the nature of lava dome failure mechanisms, direction of propagation and speed of PF. Likewise, moderate-to-large Vulcanian explosions can be monitored providing key insights on their timing and mechanisms. The SGH infrasonic array on Montserrat has remarkably proved its ability to monitor an active lava dome volcano detecting explosive eruptions and tracking PF in real-time.
 The unprecedented observation of ∼1 mHz GW, traveling at speeds as low as few tens of meters per second and capable of transporting energy over long distances with little or no attenuation, can have important implications on volcano monitoring operations. A large displacement of fluid volume is required to initiate them. The data gathered at SHV during July and December 2008, suggest that the intensity of GW may be related to the explosivity of eruptions. The amplitude of the GW recorded at SHV was lower for the smaller eruption of July 29 than for the large event of December 3. We believe that continued observations of GW may confirm these preliminary findings, thus, have future application in assessing the relative magnitude of volcanic explosions. The infrasonic array at the MVO provided useful data that will help to constrain theoretical models aimed to unravel the complex dynamics of pyroclastic flow propagation and large volcanic explosions.
 We thank MVO for technical support in the field. This work has been realized under the financial support of Dipartimento della Protezione Civile, Italy.