Continuous monitoring of microbaroms generated by storm systems distributed all around the peri-Antarctic belt reveals similar trends in the measurements. Stratospheric winds in the southern hemisphere blow eastward from June to November and westward the rest of the year, which is consistent with the observed bearings of microbaroms for Austral stations. For these stations, the weak number of detections originating from the northern hemisphere oceans could be explained by the stratospheric wind inversion in the equator region. In the northern hemisphere, the same azimuthal variations are observed showing an approximately 6 month delay. Globally and on large temporal scales of weeks to months, microbarom azimuths are rather constant whereas wider ranges are expected if we consider the extent of the ocean swells distribution. Such wider ranges can be identified in the life span of a typical swell with timescales of days [Garcés et al., 2004]. This observation confirms a strong directionality induced by the prevailing zonal wind. For these measurements, the arrival azimuths roughly coincide with the stratospheric wind direction at the latitude of the ocean swells. Rays launched eastward from the ACC during the Austral wintertime and westward in summer, reach all IMS stations with predicted arrival azimuths close to the observed ones.
 The seasonal variation in the number of detections follow the variation of the stratospheric wind strength along the source-receiver path, while some daily variability can either be related to short timescale variability of the atmosphere along the raypaths, or explained by changes in the amount of ocean swell energy. At ∼10 km altitude, the wind speed is mainly governed by large storm systems. Because of its variability, tropospheric waveguides generally do not persist for over long propagation ranges. Considering the strong temperature gradient above 90 km, thermospheric paths are always predicted. However, due to the low particle density and dissipation in the upper atmosphere, thermospheric returns are strongly attenuated. For propagation range of 1000 km and for a frequency of 0.25 Hz, rays refracting at 100–120 km are attenuated by ∼50–100 dB [Bass and Sutherland, 2004]. At shorter distances (less than ∼2000 km), previous studies demonstrated that microbaroms from energetic swells refracted back to the ground at the thermosphere may be observed [Rind, 1978; Garcés et al., 2004]. At larger distances, thermospheric returns are unlikely. Thus we assume that the observed signals propagate efficiently for thousands of kilometers in the stratospheric duct, which is consistent with the low trace velocity values generally observed.
 Figure 4a presents the seasonal variations of the maximum amplitude of signals from the South Pacific detected in Bolivia, along with fluctuations in the NRL-G2S wind-corrected sound speed in the 30 to 49 km range. These arrivals are prominent from May to November. Microbarom detections start in April with positive increasing values (and disappear in November with negative values) of the effective sound speed above ∼35 km. The maximum amplitude coincides with the highest wind speed which is greater than 50 m/s between 47 and 49 km in July. Downward refraction occurs when the effective sound speed averaged along the raypaths exceeds the sound speed at the ground level. As a result, downwind propagation with increasing wind speed in the stratosphere decreases the refracting height. Figures 4b and 4c compare the annual variations of the observed signal amplitude at I08BO and I26DE to the results of ray tracing simulations. To interpret the observations, we consider the dissipation of acoustic energy for a frequency of 0.25 Hz. The upper boundaries of the stratospheric ducts reach as high as 45 km in April and November during the seasonal reversal in the stratospheric general circulation, and as low as 30 km during the downwind season. The simulation results show that the acoustic signal suffers a negligible amplitude decrease below 45 km, while rays refracting at higher altitudes are subject to increased attenuation. These simulations are in agreement with the observed seasonal trend in the amplitude for both stations. From winter to summer, the decrease in amplitude can partly be explained by acoustic attenuation in the stratosphere, where the additional absorption due to the difference in travel paths between 30 and 45 km is slightly lower than 10−3 dB/km. Earlier researches dealing with atmospheric tidal circulation in the thermosphere pointed out semidiurnal fluctuations of the amplitude of infrasound signals [Rind and Donn, 1975; Le Pichon et al., 2005]. Such fluctuations, typical of long-duration infrasound propagating through the lower thermosphere, were not clearly observed in these measurements. Figure 4d presents, on a semilogarithmic scale, the measured amplitude versus the stratospheric wind speed. Using a standard least squares procedure, the following linear amplitude-scale relationship is derived:
where P is the zero-to-peak amplitude (in Pa), R is an approximate distance between the source and the receiver (in km), n = 0.3 ± 0.1 is the distance-scaling component, k = 0.0096 s/m is the wind effect normalization parameter estimated with a unit standard deviation of ∼10%, and Vs is the component of the stratospheric wind velocity (in m/s) in the direction of propagation. The scattering in our scaling relation may either be related to stochastic variations in the atmosphere not predicted by the atmospheric models used, or changes in the amount of ocean swell energy. Blanc et al.  proposed for k a value of 0.0116 s/m derived from a combination of atmospheric nuclear and chemical explosion data covering ranges of 400–7000 km. Mutschlecner et al.  adopted a value of 0.016 s/m from observations of nuclear tests carried out at the Nevada test site for ranges of 100–300 km. Stevens et al.  provided an evaluation of various scaling laws derived from nuclear tests with k = 0.019 s/m. These discrepancies may be explained by differences in the frequency content of the signals and the dependence of amplitude upon yield and propagation ranges. In order to investigate small-scale variations of the amplitude, (1) an improved knowledge of the swell distribution over the oceans, (2) a more precise radiation modeling of microbarom signals, and (3) a more realistic long-range propagation modeling are needed. Willis  used global ocean wave spectra provided by the National Oceanic and Atmospheric Administration's (NOAA's) Wavewatch3 (WW3) model [Toldman, 2002], and developed a detailed source pressure formulation base upon interactions of surface wind and ocean waves. Such works provide a basis for a better quantification of the relationship between infrasonic observables and atmospheric specification problems.