The dominant source of infrasonic waves detected in Tahiti is associated with ocean swells generating permanent signals from 0.1 to several hertz. At thousands of kilometers from Tahiti, microbaroms are mainly produced by standing ocean waves near low-pressure systems in the south Pacific. Their monitoring over one year exhibits clear seasonal trends correlated with changes in the prevailing stratospheric wind direction. More locally, ocean waves propagating towards Tahiti also generate surf noise when impacting against the reefs. Analyzing the infrasonic recordings, the main source regions of surf signals are identified and their locations and seasonal variations are discussed.
 The infrasound station I24FR, Tahiti, is part of the global infrasound network of the International Monitoring System (IMS) for the compliance of the Comprehensive Nuclear-Test-Ban Treaty (CTBT). Such a network also provides an opportunity to monitor human activities and natural phenomena on a global scale [Hedlin et al., 2002]. The I24FR station consists of five microbarometers, 1 to 4 km apart, which can measure pressure fluctuations from 0.003 up to 27 Hz [Le Pichon et al., 2002]. Due to the geographic situation of I24FR, most of the detected infrasonic waves are produced by low-pressure systems in the Pacific Ocean [Willis et al., 2004].
 Past studies showed that microbaroms are related to severe weather in the ocean and the resulting high ocean surface waves [Daniels, 1962; Posmentier, 1967; Rind, 1980; Tabulevich, 1993]. A precise source mechanism describing the non-linear interaction of ocean waves with the atmosphere has recently been proposed [Arendt and Fritts, 2000; Garcés et al., 2002]. The temporal and spatial variations in the structure of the atmosphere substantially affect the propagation of infrasonic waves [Kulichkov et al., 2004]. Thus, observation of infrasound over long periods provides a comprehensive technique for monitoring parameters of the upper atmosphere, such as wind speeds. More specifically, the use of microbaroms to probe the upper atmosphere has already been pointed out in previous works [Donn and Rind, 1971; Rind, 1978].
 Surf-generated signals are also generated by ocean waves that propagate from major storm systems to the coastlines. The penetration of ocean waves into closed bays may produce this type of signal due to the compression and expansion of large volumes of air, or by breaking waves at a cliff or a reef [Garcés et al., 2003]. The volume of air suddenly released generates acoustic waves whose amplitude depends on the height and length of the waves. Continuous monitoring of coherent infrasonic waves at I24FR reveals the existence of such signals. An automatic location procedure has been developed in order to locate source regions of surf signals.
 Microbaroms and surf activity are permanent sources of infrasound and provide useful means of investigating seasonal changes. Furthermore, the huge number of detections associated with these natural sources makes it difficult to focus on transient signals, such as atmospheric explosions. The aim of this paper is to study and characterize these signals over a whole year. A better understanding of these natural phenomena is essential for routine identification of this class of infrasound.
2. Observations and Analysis
 The wave parameters of the infrasonic waves are calculated with the Progressive Multi-Channel Correlation method (PMCC) [Cansi, 1995]. The implementation of automatic processing of the signals recorded at I24FR clearly reveals the existence of microbaroms and surf activity. Figure 1 shows the results of typical recordings of microbaroms overlaying surf signals. As seen from the radar plot, the detected azimuths versus frequency indicate the existence of two distinct sources. Between 0.1 and 0.2 Hz, microbaroms are detected from 190 to 220°, and in the same period, surf signals with dominant frequencies of 1–7 Hz are observed in the opposite direction. Comparative noise spectra show the microbarom peak at 0.2 Hz and energy above 1 Hz associated with surf activity. Differences in amplitude measured between individual sensors are explained by attenuation and will be discussed at the end of this section. Figure 2 shows seasonal trends in the azimuth variations of microbaroms and surf signals between September 2002 and January 2004. Microbaroms are detected from 200–250° from March to October and from 70–170° from December to February. In the [1–4] Hz band, surf signals from 355–70° are mainly observed from October to March, while signals from 210–270° are detected permanently during the whole year.
 Noise due to breaking waves contributes significantly to the low-frequency ambient noise near shore [Wilson et al., 1985]. For surfers, the beaches of highest repute are Vairao and Teahupoo located a few kilometers to the southwest of the station, giving a good indicator of wave size and consequently, of the intensity of the acoustic signals they may generate. Wave heights between 2 m and 2.5 m are observed throughout most of the year. Since the maximum distance separating the main reefs and the station is nearly 20 km, the characterization of the wave fronts curvature can be used for locating the main source regions of surf signals. Arrival time measurements at each of the five array-elements allow us to apply classical inversion algorithms used in seismology [Geiger, 1910]. As recently demonstrated by Arnoult et al. , such methods proved to be very efficient for near-field location of infrasonic sources. The acoustic propagation path to the station is assumed to be rectilinear with a uniform sound speed calculated from temperature measurements. Figures 3c and 3d show the main source regions of surf signals obtained by the location procedure. The main zones of breaking waves are identified near reefs situated along the eastern coast of Tahiti with a dominant region found in the Taravao Bay. Most of the reefs around the peninsula are located behind the Tahiti peninsula volcanic cone, which is composed of steep cliffs. The propagation of waves is therefore not favored in those directions. Southwest surf is constant all over the year (Figure 2). The eastern and northern shores are exposed to strong swells during the austral summer. However, the volcano of the main island hides surf signals from the northern shore and may explain the only active area found along the closest part of the shoreline to the array. Figure 3a compares the number of surf detections per day to the NOAA Wavewatch III wave height predictive model [Tolman, 2002] at Papeete from February to May 2003. The number of detections is consistent with the wave height variations. The infrasound detectability at I24FR is sensitive to the high variability of the wind-generated noise. Consequently, the number of detections is lower during periods of high winds and may explain some discrepancies during large swells. These observations corroborate other recent studies in Hawai'i [Garcés et al., 2003] that clearly demonstrate the correlation between amplitude of surf signals and waves height. As shown by Figure 1b, differences in the amplitude of the surf signals are observed. These are explained by signal attenuation during propagation. Considering the measurements of signal amplitudes at each of the five array-elements and their known source locations, the wave's attenuation can be estimated (Figure 3b). When averaged over one year, a slope of ∼1.7 dB/km is obtained. This attenuation is consistent with statistical description of sound blast propagation over land already pointed out by Schomer . This attenuation may also have an alternate explanation. Since surf signals originate from a finite area or line of significant size compared to the distance to the station, the combination of multiple sources regions that are not coherent in amplitude and phase may also explain part of the differences in amplitude across the array.
3. Concluding Remarks
 Due to severe attenuation in the thermosphere, detection of distant sources of infrasound depends on the prevailing upper wind direction. Continuous monitoring of microbaroms from storm systems distributed all around the peri-Antarctic belt reveals seasonal inversion of the stratospheric wind. From October to May, stratospheric winds blow westwards in the southern hemisphere and eastwards the rest of the year which is consistent with the observed bearings of microbaroms. The weak number of detections from the north Pacific can be explained by the stratospheric winds inversion between the two hemispheres. Results from several IMS stations located in the southern hemisphere show that these variations are not specific to Tahiti. These findings will be discussed in a separate manuscript. The weak number of detections from the north Pacific can be explained by the stratospheric winds inversion between the two hemisphere and by a shorter correlation length for these more distant swells.
 Surf activity is also associated with low-pressure systems in the south Pacific throughout the year and major storms in the north Pacific during the austral summer. Ocean waves generated by these far-away swells propagate to the coast of Tahiti and break on its reefs. From the infrasonic measurements at one single array, the use of a simple inversion scheme can provide a precise location of nearby source regions. Those are located on emerged reefs, confirming that breaking waves are the main source mechanism of surf signals. Due to the higher frequency range, these signals are strongly attenuated over distance and cannot be used to monitor the upper atmosphere. However, the combination of an infrasound array with other wave height measurements may help to monitor ocean wave interactions near coastlines. Furthermore, in order to assess the performance of the infrasound network, it is necessary to study the broad suite of natural and local man-made events that deafen the array. Such natural phenomena offer an opportunity to develop and evaluate detection and location procedures for discrimination purpose with the aim of monitoring the CTBT. For all different geographical areas where infrasound stations are installed, a one-year learning process should help to build up experience and discriminate between environmental noise and events of interest.
 The authors gratefully thank G. Barruol of the University of Polynesia for his contribution to the processing and the analysis of the NOAA WW3 data. The authors are also grateful to Dr. Lars Ceranna for his interest in this study and for the helpful discussions we had during the completion of this work.