Infrasonic airwaves produced by active volcanoes provide valuable insight into the eruption dynamics. Because the infrasonic pressure field may be directly associated with the flux rate of gas released at a volcanic vent, infrasound also enhances the efficacy of volcanic hazard monitoring and continuous studies of conduit processes. Here we present new results from Erebus, Fuego, and Villarrica volcanoes highlighting uses of infrasound for constraining quantitative eruption parameters, such as eruption duration, source mechanism, and explosive gas flux.
 Understanding volcanic eruptions requires robust estimation of fundamental physical parameters such as the detailed history of gas flux. Seismograms are often ineffective for such analyses because they are typically complex superpositions of source processes and wave propagation phenomena in strongly scattering media [Chouet, 1996]. However, the acoustic airwaves produced by eruptions and radiated to distances of a few km tend to exhibit comparatively low atmospheric scattering/dissipation, and experience predictable (and frequently minor) echoing, site, or weather-dependent effects. For these reasons infrasound monitoring much more readily enables quantitative assessment of eruptive degassing. In basic monitoring situations, high-amplitude airwaves provide unequivocal evidence of eruptive degassing in progress. Such infrasound is also largely unaffected by cloud cover and does not rely on line-of-sight view of vents, as is the case with ancillary satellite or other visual/infrared observations. For these reasons, acoustic monitoring and quantitative analysis of infrasonic pressure waves is becoming increasingly well established at active volcanoes worldwide (Table 1).
Table 1. Summary of Some Volcanoes With Infrasound Studies
Maximum peak excess pressures in the near-infrasound bandwidth as cited in reference. For comparative purposes these pressures have been reduced here to 1 km, assuming an inverse pressure decrease with radial distance from vent.
 An illustrative example of the merits of infrasound can be shown at Fuego Volcano, Guatemala, where pyroclastic explosions occurred several times each hour during January 2003. At a monitoring site 2.6 km from the summit crater, explosion plumes were first visible about 1.5 s before the first emergent seismic arrivals and 8 s before the first impulsive infrasonic arrival (Figure 1). Though the seismicity associated with these events tended to be complex and drawn out in time, the acoustic records depicted an initial high-amplitude pressure pulse (compression/dilatation) followed by lower-amplitude ‘rumbling’ continuing for tens of seconds. The primary pulse of the Fuego infrasound often exceeded 40 Pa (∼126 dB) 2.6 km from the vent and was accompanied by an audible boom that sounded like distant thunder (<100 dB). As observed at other volcanoes [Vergniolle et al., 1996; Johnson, 2003], the Fuego infrasonic bandwidth appears to have several orders of magnitude greater spectral energy than the audible bandwidth.
 Volcanoes may efficiently generate long-wavelength (17 to 340 m) near-infrasound (20 to 1 Hz) because the vent source dimension, considered here to be the area of vigorous free surface degassing, is frequently much larger than typical sonic wavelengths. Though debate remains as to whether some volcanic infrasound can be generated internally (e.g., within a volcano's magma conduit [Buckingham and Garcés, 1996; Garcés and McNutt, 1997]), or due to vibrational modes of large intact gas bubbles [Vergniolle and Brandeis, 1994; Vergniolle et al., 1996], the majority of high-amplitude volcanic infrasound is explicable by the eruptive acceleration of compressed volatiles from vents [Reed, 1987; Firstov and Kravchenko, 1996; Yamasato, 1997; Ripepe and Gordeev, 1999; Gabrielson, 1998; Johnson, 2003]. Such infrasound may result from either a long-period acceleration of erupted gas at a compact vent or from an impulsive source distributed over a large region. In both cases, it is possible to model infrasound generation based upon the linear theory of sound [Lighthill, 1978; Dowling, 1998] where the acoustic wavefield is composed of a superposition of n volumetrically expanding monopole point sources. For a fixed source at the surface of a rigid half space radiating into a homogeneous atmosphere of velocity c, the excess pressure recorded at x (located distances ri from each source) is p(x, t) = , where each Qi is proportional to the time derivative of the associated mass flux. For a monopole point source, with negligible propagation effects, a first-order flux estimate is thus q(t) = 2πr∫p(t + r/c)dt [Firstov and Kravchenko, 1996; Johnson, 2003].
2. Volcano Infrasound Case Studies
 Volcano infrasound may be modeled by either a series of sources distributed over a diffuse vent region, or in special cases, as a single point source from which a gas or fluid volume is erupted. A point source model is suitable for single-bubble bursts at the intermediate viscosity (∼104 Pa-s) phonolitic lava lake at Mount Erebus, Antarctica [Dibble, 1994], where infrasonic wavelengths are large relative to source dimension, and the source history is simple. Erebus video reveals that intact bubbles with radii over 5 m emerge then burst from the surface of the lava lake, spewing volcanic bombs and ash, and releasing over-pressurized volatiles (primarily H2O and CO2) [Aster et al., 2003]. Corresponding infrasound waveforms have an N-wave shaped appearance, characteristic of a weak shock wave that is generated during bursting of an overpressurized (bubble) volume [Blackstone, 2000]. Erebus N-wave amplitudes observed from 1999 to present have ranged from 2 to 100 Pa at 1 km [Rowe et al., 2000], providing bubble gas mass estimates on the order of ∼103 kg for gas pressurized at a few atmospheres (Figure 2) [Aster et al., 2003, 2004]. Gas flux from these infrequent (<1 per day) events, indicates that only a small percentage of the estimated 106 kg degassing [Andres and Kasgnoc, 1998] is attributable to discrete explosions.
 Degassing during lava lake activity at Villarrica, Chile produces more continuous sustained infrasound than at Erebus (Figures 3a–3b). Like Erebus, Villarrica is also a fairly low-viscosity, open-vent system that at times hosts a lava lake at the tip of its conduit. However, following an N-wave onset similar to the Erebus bubble bursts, Villarrica infrasonic codas show continuing pressure oscillations lasting for 5–10 s with dominant energy ∼0.5 Hz that hint at a sustained sequence of bursting bubble slugs. Although absolute mass flux is not recoverable from these more complex waveforms, the infrasound has time-integrated power that is four orders of magnitude greater than at Erebus, with more than 103 discrete degassing events occurring daily during December 2002. If these degassing events are responsible for the bulk of Villarrica's 105–106 kg/day SO2 output [Witter et al., 2004], a typical degassing event releases 102–103 kg of SO2, with a correspondingly greater total volatile output (SO2, H2O, and CO2).
 Basaltic Vulcanian activity at Fuego, Guatemala in January 2003 was considerably more explosive than the Strombolian behavior described at both Erebus and Villarrica. Here, infrasound from explosions also began impulsively, reflecting an abrupt outward acceleration of gases, but the vigorous degassing continued for several tens of s (manifested by a relatively low-amplitude, tremor-like infrasonic coda), which helped to fuel energetic (>1500 m) ash-rich convective plumes. The dominant mechanism of these eruptions is likely the continued explosive foam disruption of small (≪10−3 m) pressurized vesicles, typical in more volatile-rich, explosive, silicic eruptions [Sparks et al., 1994; Mangan and Sisson, 2000]. Although Fuego's explosivity may be attributed in part to the magma's gas-rich nature (>106 kg/day flux since 2001 (W. Rose, written communication, 2004)), the observed activity and corresponding infrasound suggest that either the mafic magma behaves viscously or that the vent/conduit geometry is narrow. Both situations likely inhibit the formation of large individual bubbles. Fuego infrasound (envelope and frequency content) is reminiscent of discrete eruptions at Karymsky and Sangay [Johnson and Lees, 2000] and Volcán de Colima (N. Varley, unpublished data, 2003), but differs from eruptive activity at more viscous volcanoes. Very low-amplitude or indiscernible infrasound typically produced by more silicic volcanic systems like Unzen [Yamasato, 1997], Pichincha [Johnson et al., 2003], Monserrat (J. Neuberg, written communication, 2003), or Santiaguito [Johnson et al., 2004] may be attributed to dispersed source regions and/or to non-impulsive vent degassing. Non-impulsive surface gas release is likely for vesiculation that occurs along conduit margins at depth.
3. Current State of Volcano Infrasound Studies
 Volcano infrasound observations have been made with both single sensors and arrays. To date, infrasonic deployments have included volcanoes ranging from low-viscosity basaltic or phonolitic systems, to basaltic-andesite and andesitic stratovolcanoes, to highly viscous, silicic systems (Table 1). Infrasound arrays and networks, deployed as tight antennas or distributed at various azimuths around a volcano, show tremendous potential for enhanced event detection and localization [Yamasato, 1997; Ripepe and Marchetti, 2002; Garcés et al., 2003; Johnson et al., 2003; Johnson, 2004]. The low velocity of sound (∼343 m/s at STP) facilitates precise localization (<few m) of acoustic sources and allows for the tracking of evolving source locations with comparable resolution [Yamasato, 1997]. Cross-correlation techniques have been used at multi-vent systems such as Stromboli [Ripepe and Marchetti, 2002; Johnson, 2004] and Pu'u ‘O’o, Kilauea [Garcés et al., 2003] to monitor as many as 6 distinct vents. Integrating infrasound with other geophysical measurements, such as seismic, thermal, gas flux, and video, shows tremendous potential for improved understanding of fluid transport and conduit processes [Aster et al., 2004; Ripepe et al., 2001, 2002]. Comparison of radiated acoustic, seismic, and thermal power [Johnson et al., 2004] promises new insights for eruption source dynamics such as melt properties [Garcés et al., 1998; Hagerty et al., 2000], volatile fragmentation depth, and other source and near-source conditions (J. B. Johnson and R. C. Aster, Relative partitioning of acoustic and seismic energy during Strombolian eruptions, submitted to Journal of Volcanology and Geothermal Research, 2004).
 We acknowledge B. McIntosh, L. Clabaugh, R. Wolf, A. Harris, E. Calder, W. Rose, and the IRIS PASSCAL Instrument Center at New Mexico Tech for field assistance. M. Garcés and W. Rose provided helpful comments. This work supported through NSF grants OPP-9814291, OPP-0116577, OPP-0229305, and EAR-0106349.