7.1. Latitudinally Trapped Waves
 Inertia-gravity waves can have periods shorter than 2 days and zonal wavelengths of about 2,000 km as studied in this paper. However, the short-period disturbances are not likely due to inertia-gravity waves because the intrinsic wave period of the short-period disturbances (e.g., about 30 h in August) is obviously longer than the inertial period (i.e., about 14 h at 60°S) that is the upper limit of gravity wave period. Moreover, gravity wave components in the reanalysis data are usually suppressed by the normal-mode initialization [Gibson et al., 1997].
 Higher harmonics of larger-scale waves such as planetary waves is another candidate for the short-period disturbances. However, this is also unlikely because the ground-based phase velocity of larger-scale waves is not greater than 10 m s−1 throughout the year (Figure 2), while the short-period disturbances have much faster phase velocities.
 The Charney-Drazin theorem [Charney and Drazin, 1961] provides a simple condition on the phase speed c in order that the wave propagates upward in quasi-geostrophic flow on a beta plane:
where is a zonal-mean zonal wind, c ≡ β [k2 + l2 + f2/(4N2H2)]−1 is the critical zonal-mean wind, k and l are the zonal and meridional wave numbers, f is the constant Coriolis parameter, N is the Brünt-Väisälä frequency, and H is the pressure-scale height. The short-period disturbances have an eastward ground-based phase velocity of about 40 m s−1 at 60°S at θ = 540 K in winter. The critical level (i.e., = c) for the short-period disturbances is certainly located between the tropopause and the 540-K isentropic surface. Thus, the short-period disturbances are not due to the waves propagating from the troposphere.
 Spatial scales of filaments generated by the planetary wave breaking around the polar-night jet during winter and early spring can be as small as the short-period disturbances. The filaments are usually thought to be passively advected by the large-scale flow. While the short-period disturbances in April, May, and November have a nearly zero intrinsic phase velocity, they have a non-zero intrinsic phase velocity (i.e., not just advected by the large-scale flow) from June through October near the core of polar-night jet. Thus, the short-period disturbances in April, May, and November may be due to the filaments passively advected by the large-scale flow, although the short-period disturbances from June through October cannot be due to those passively advected filaments.
 The most probable candidate for the short-period disturbances from June through October is trapped waves. The latitudinally evanescent structure with the maximum amplitude near the core of polar-night jet suggests that the short-period waves are latitudinally trapped in the large latitudinal PV gradients there, similarly to the medium-scale waves trapped around the midlatitude tropopause. In case of the medium-scale waves, the region with large PV gradients is confined both latitudinally and vertically, and the medium-scale waves are trapped both latitudinally and vertically near the midlatitude tropopause. On the other hand, the maximum of PV gradients around the polar-night jet is attributed to the local maximum of second-order latitudinal derivative of mean zonal wind, so that the large latitudinal PV gradients are not vertically confined. Thus, the short-period waves are merely latitudinally trapped around the polar-night jet.
 A direct observation of those trapped waves is very difficult because it needs high temporal and/or spatial resolution enough to detect those short-period small-scale disturbances. The satellite data [cf. Gibson et al., 1997] and radiosonde data [cf. Yoshiki and Sato, 2000] used in the data assimilation process for ECMWF reanalysis are too sparse to resolve those trapped waves. This fact means that the trapped waves observed in the reanalysis data may be an artifact of the data assimilation process. However, it is natural that there are waves trapped around the maximum of latitudinal PV gradients [cf. Sato et al., 1998; Rivest et al., 1992]. Moreover, the resolution of observational data used in the data assimilation process is enough for portraying the structure of polar vortices in both hemispheres. Thus, if the model used for data assimilation appropriately describes atmospheric dynamics, the trapped waves observed in the model must also exist in the real atmosphere.
7.2. Source of Trapped Waves
 In spite of the fact that the wave propagation is considered easier in the Southern Hemisphere because of larger PV gradients, the short-period disturbances are more active in the Northern Hemisphere where the planetary waves are also more active. Moreover, the barotropic and baroclinic instability are unlikely, because the horizontal and vertical shear of zonal-mean flow is smaller in the Northern Hemisphere. These facts suggest that the generation of short-period disturbances in the Northern Hemisphere is related to the planetary wave activity there.
 Previous multi-layer modeling studies have pointed out that the filaments produced by the planetary wave breaking have a nearly barotropic structure [Dritschel and Saravanan, 1994; Waugh and Dritschel, 1999; Polvani and Saravanan, 2000]. Using the contour advection and reverse domain filling methods, Schoeberl and Newman  showed that the filaments forming in the edge region of polar vortex have a fairly barotropic structure often over the whole stratosphere. When the filaments are trapped in the region of large latitudinal PV gradients, they may have a non-zero intrinsic phase velocity as shown in the previous section.