On June 23, 2001, a strong earthquake measuring Mw 8.4 occurred along the coast of south-central Peru. Coherent infrasonic waves were detected over a period of one hour by the IS08 infrasound station in Bolivia. Analysis of the ground-coupled air waves shows that the rupture propagated from the northwestern to the southeastern part of the fault with a rupture velocity of 3.3 km/s. The azimuth variation of the infrasonic waves is attributed to a distribution of secondary sources along the highest mountain ranges, which excite infrasonic waves that are diffracted to the ground. The predominant source of infrasound is likely distributed along the Andean Cordillera. Using the azimuth and arrival time determination, the horizontal scale size of the distant source regions of infrasonic waves is reconstructed over distances greater than 400 km.
 Sources such as atmospheric or buried explosions and shallow earthquakes are known to produce infrasonic pressure waves. Infrasonic waves from the sudden strong vertical ground displacements have been detected during the Alaskan earthquake of March 27, 1964, at distances of thousands kilometers from the origin [Bolt, 1964, Mikumo, 1968]. Donn and Posmentier  related the observed pressure measurements to the vertical radiation of air waves above the seismic waves. Cook  reported on pressure waves radiated locally by seismic waves during the Montana earthquake on August 18, 1959. Two distinct mechanisms of infrasonic signal generation during large earthquakes can be identified:
- (1)Ground-coupled air waves generated at the station by the vertical displacement of the seismic waves - these may involve both wave propagation and pressure changes associated with ground motion.
- (2)Infrasonic signals generated in the atmosphere by violent ground motion in the epicenter region [Mikumo, 1968], and associated to pressure waves reradiated when the seismic surface waves travel through region a of high mountains [Young and Greene, 1982].
 On June 23, 2001, at 20:33:13 UTC, a strong earthquake measuring Mw 8.4 (NEIC) occurred along the coast of south-central Peru. The earthquake origin (16.15°S, 73.40°W, focal depth ∼30 km) was centered along the Peruvian coast about 600 km southeast of Lima and 110 km northwest of Camana (Figure 1). The Pacific Tsunami Warning Center reported a moderate tsunami struck the Peruvian coast.
 The purpose of this paper is to present a thorough analysis of the infrasonic waves generated by this earthquake. Infrasonic waves have been detected at the IS08 station (Bolivia - La Paz, 16.26°S, 68.45°W) that is part of the global infrasonic network of the International Monitoring System (IMS). The detection at short distances (about 500 km from the epicenter) allows a precise determination of the pressure wave parameters of both local and distant generation of infrasonic waves.
2. Observations and Wave Parameters
 The IS08 station consists of four microbarometers, 1.5 to 3 km apart. The wave parameters of the infrasonic waves are calculated with the Progressive Multi-Channel Correlation method (PMCC). This method, originally designed for seismic arrays, proved to be very efficient for infrasonic data and is well adapted for analyzing low-amplitude coherent waves within non-coherent noise [Cansi, 1995; Le Pichon et al., 2002]. With a sampling rate of 20 Hz, the expected numerical resolution at 0.5 Hz is of the order of 0.5° for the azimuth and 5 m/s for the apparent horizontal phase velocity.
2.1. Seismic Coupled Air Waves
 The infrasonic field locally generated may be due to an integrated effect of the sound radiated by seismic waves over a large area [Cook, 1971]. The measured pressure changes may also depend on the sensor response to accelerations and ground level elevations independently of any pressure changes occurring in the atmosphere [Bedard, 1971]. The conversion from seismic waves to the sound pressure has been already observed on microbarometers or microphones at teleseismic distances [Cook and Young, 1962] or at regional distances [Takahashi et al., 1994]. Due to the coupling at the earth-air interface, the horizontal phase velocity, or trace velocity, of the ground-coupled air waves and the seismic waves are identical. In the present study, the trace velocity and azimuth of the regional waves recorded on the infrasound station are determined. The distance between IS08 and the hypocenter is equal to 530 km. The distribution of the aftershocks and the source inversion given by Kikuchi and Yamanaka  shows that this distance is approximately twice the length of the fault activated during the earthquake (Figure 1). This kind of record is unique and allows us to give some details concerning the dynamic of the rupture.
Figure 2 shows the recorded signal at one microbarometer of IS08 with the theoretical arrival times of the regional waves. The S and Rayleigh waves are observed with a better coupling in the atmosphere compared to the P waves, as it was already observed by Cook and Young . The calculated trace velocities correspond to the seismic waves propagating at regional distances. The azimuths vary from 290° to 240° while the trace velocity decreases from 7.8 km/s for refracted Pn waves to 2.5 km/s for the Rayleigh waves during a time interval of 225 s. This calculation shows that the source moved from the northern to the southern part of the fault. Since the duration of the source is greater than the difference in propagation time between the different regional waves, a mixture of the different waves is observed. In order to simplify the interpretation, the arrivals of the Pn waves are selected. Figure 2b shows that the duration of the Pn arrivals is around 40 ± 5 s with a small resurgence around 70 s, while the azimuth decreases from 290° to 240°. This variation indicates a southward rupture propagation along the fault towards the aftershock epicenters. The rupture propagation is simulated along a line for three different angles (130 ± 10°) covering the repartition of the aftershocks. At each step of the rupture, the azimuth and the arrival time of the Pn waves are calculated at IS08. The best fit of the Pn arrival is obtained with a rupture velocity of 3.3 ± 0.3 km/s (Figures 2b and 2c). Considering an aftershock distribution of 300 km along the fault, a source duration of 90 ± 10 s is obtained. This result is consistent with the source duration estimated at 107 s by Kikuchi and Yamanaka .
2.2. Distant Generation of Infrasound
Figure 3 shows the time variations of the azimuth and the horizontal trace velocity of the infrasonic waves detected after the Mw 8.4 earthquake. The distant propagation of infrasonic waves (referred as groups 1 to 4 on Figure 3) are characterized by a trace velocity of 340–370 m/s. Four distinct arrivals are detected from 20:39 to 23:39 and related to one of the following aftershocks (Table 1):
- Group 1 is detected between 20:39 and 21:20 with an azimuth increasing from 238 to 272°. The main period is 2 s and the amplitude increases from 1.6 Pa at 20:40 to 5.2 Pa peak-to-peak at 20:53. Using a typical apparent speed of propagation of thermospheric phases, for a propagation range of 530 km (section 3), the infrasonic waves generated from the epicenter of the main earthquake are expected at 21:05 with an azimuth of 269°.
- Group 2 is detected between 21:20 and 21:28. The wave train is consistent with infrasonic waves generated by the mb 5.8 aftershock. This group is not generated from the epicenter area since the measured azimuths (277–280°) are greater than the expected theoretical values (∼253°).
- Group 3, detected between 21:50 and 21:52 with an azimuth ranging from 252 to 264°, is associated to the mb 6.3 aftershock. Figure 3 also reveals coherent waves of 10 to 20 s period from the southwest (in blue). These longer period waves may be Mountain Associated Waves generated by turbulent flows over mountain ranges [Larson et al., 1971]. Such waves are frequently observed from IS08 in the direction of the Southern Andean Cordillera.
- Group 4, detected from 23:13 to 23:39 with an azimuth of 268–279°, is related to the seismic ground-coupled waves of the mb 5.9 aftershock. At 23:10:57, the seismic ground-coupled waves are also detected with a trace velocity ranging from 3 to 8 km/s.
Table 1. Source Parameters of Aftershocks of the Peru Earthquake of June 23, 2001 at 20:33:13 UTC (NEIC, mb > 5)
Origin time (UT)
 The observed azimuth variations and the expansion of the signal duration can be explained by: (1) a radiation area along the fault rupture, (2) the increase of the effective source regions when the seismic surface waves travel from the fault rupture through region of high mountains.
3. Infrasound Propagation and Secondary Sources Location
 The source location of the coherent wave trains measured from 20:39 to 21:28 is now considered. The input parameters of the location procedure include the azimuths and arrival times of groups 1 and 2, the origin times of the Mw 8.4 earthquake and the mb 5.8 aftershock. The propagation model is based on a velocity of 3.1 km/s for seismic surface waves propagating from the epicenter area, and sound and wind speed profiles in the atmosphere.
 Ray-tracing simulations are performed in order to evaluate the effective propagation speed, or celerity of infrasonic propagation, which depends on the ray path and the propagation range. Ray paths are computed for waves propagating in a stratified atmosphere under the influence of winds [Garcés et al., 1998]. The wind and temperature profiles used in the simulations are obtained from the MSISE-90 and HWM-93 empirical reference models for the atmospheric conditions of June 23, 2001 at 21 h UTC. The simulations are performed for air waves launched along the Andean Cordillera [Young and Green, 1982], at altitudes lower than 5 km in the 60° to 110° quadrant.
 For propagation ranges greater than 400 km, the dominant predicted ray paths are thermospheric returns with ray parameter values consistent with the measured trace velocity (355–365 m/s, Figure 4). Depending on the source location and the azimuth, the effective celerity ranges from 270 to 290 m/s. From 20:40 to 20:53, the amplitude of the signals filtered in the 0.05–1 Hz frequency band increases by a factor of 3.2. This variation is interpreted as the result of the amplitude absorption through the shadow zone whose effect is stronger at short propagation ranges, i.e. at the beginning of the detection. We assume that the coherent arrivals detected before 20:50 (propagation ranges lower than 400 km) are related to the ground reception of stratospheric elevated ducted waves with boundaries between 2 km and 43 km. Such trapped waves can be observed when the source is located above the ground [Weber and Donn, 1982]. The stratospheric ducted waves celerity of these scattered or diffracted to the ground is 300–305 m/s. The measured trace velocity (330–340 m/s) is consistent with the mean theoretical value (335 m/s). Figure 4 shows the reconstructed horizontal scale size of the distant source regions using both thermospheric and stratospheric paths. In order to take into account uncertainties due to the measurements, atmospheric profiles and propagation model used, the accuracy of the azimuth and the effective celerity are estimated at 10° and 30 m/s respectively.
 The strong earthquake of June 23, 2001 that occurred along the coast of south-central Peru produced large infrasonic waves that were measured by the IS08 IMS station. The calculated trace velocity range from several kilometers per second to the sound velocity. These values are characteristic of seismic waves and distant generation of infrasonic waves, respectively.
 Although the aperture of the IS08 array is designed for the reception of sound waves, the PMCC analysis yields some useful estimates of the seismic wave field. The azimuth and the velocity of all the regional seismic waves are determined. The azimuth variation for the Pn waves indicates that the rupture propagated from the northwestern to the southeastern part of the fault. This observation is consistent with the spatial distribution of the aftershock epicenters located to the southwest of the mainshock hypocenter. The rupture velocity can also be estimated to 3.3 ± 0.3 km/s, which is the mean crustal shear wave velocity in this region. The source duration of 90 ± 10 s is close to the teleseismic source inversion.
 The observed azimuth variations and the expansion of the infrasonic signal duration suggest that regions far from the epicenter are excited and act as secondary sources. In order to reconstruct the horizontal scale size of the distant source regions, ray-tracing simulations have been performed to identify several modes of propagation through the atmosphere. The predominant lower thermosphere source of infrasound is attributed to radiation of pressure waves during seismic reverberation of the Andean Cordillera. These infrasonic waves are observed with a detection threshold close to mb5.8. A distribution of the radiating zone, near 100 km width and 400 km long, has been reconstructed.
 It is likely that these distant source regions can be located only for larger earthquakes with favorable setting within a region of high mountains. Infrasonic observation of such secondary sources may be a value for better understanding of the remote effects of earthquakes, particularly where there is an absence of surface motion instrumentation. It is hoped that a study combining the infrasonic measurements with seismic source characteristics will allow a better understanding of the physical mechanism for infrasonic wave generation.