1. Introduction
Convectively coupled Kelvin waves are large-scale, eastward-propagating tropical convective disturbances whose spectral characteristics in the deep cloudiness field lie along the theoretical dispersion curves for dry Kelvin waves of shallow equivalent depth (Takayabu 1994; Wheeler and Kiladis 1999, hereafter WK99; Wheeler et al. 2000, hereafter WKW00; Straub and Kiladis 2002, hereafter SK02). Deep convection in these waves propagates eastward at approximately 17 m s–1, and is accompanied by large-scale kinematic and thermodynamic perturbations in the troposphere and lower stratosphere that are broadly consistent with the theoretical internal Kelvin wave response to an eastward-propagating upper tropospheric heat source (Holton 1972; WKW00). Convectively coupled Kelvin waves are thus thought to arise from the interaction between a dry Kelvin mode and moist convective processes. The observed phase speed of convectively coupled Kelvin waves of 17 m s–1 is slower than that of dry Kelvin waves observed in the atmosphere, which propagate eastward at 40–50 m s–1 (Andrews et al. 1987; Milliff and Madden 1996), and is faster than that of the Madden–Julian oscillation (MJO), in which convection propagates eastward at 5–10 m s–1 (Madden and Julian 1971, 1994).
Previous studies of convectively coupled Kelvin waves have illustrated the observed horizontal and vertical structures of these waves in their developed stage (Takayabu 1994; WKW00; SK02), but have not addressed the mechanisms by which Kelvin waves might be initially excited. In this paper, we argue that a substantial fraction of convectively coupled Kelvin waves are excited through a distinct forcing from the extratropics. Extratropical forcing has also been suggested to play an important role in initiating lower-frequency MJO events (Liebmann and Hartmann 1984; Hsu et al. 1990; Meehl et al. 1996; Lin et al. 2000; Straus and Lindzen 2000).
Convectively coupled Kelvin waves are observed most frequently in the central Pacific intertropical convergence zone (ITCZ) during austral winter [June–July–August (JJA)]. During this season, the subtropical jet in the Southern Hemisphere attains its largest amplitude and northernmost position in the vicinity of Australia (see Fig. 1), and Rossby waves excited in the jet propagate eastward and equatorward from this region, as shown by Chang (1999) and discussed in more detail in section 4. In the current study, we present observations that link these equatorward-propagating Southern Hemisphere subtropical jet disturbances with the initiation of tropical Kelvin wave activity in the Pacific ITCZ, and suggest a possible dynamical mechanism for this interaction.
This paper is organized as follows. A literature review is presented in section 2. In section 3, the datasets and methodology used in this study are detailed. A climatology of the parameters relevant to Kelvin wave activity during austral winter is presented in section 4. Sections 5 and 6 discuss the typical horizontal and vertical structures of the circulation anomalies associated with a convectively coupled Kelvin wave in the tropical Pacific. In section 7, the longer timescale relationship between Kelvin wave activity in the Pacific and the variability of the Southern Hemisphere subtropical jet is described. In section 8, a case study is presented that illustrates the relationship between the extratropical circulation and Kelvin wave activity during two contrasting months. Finally, a summary and conclusions are given in section 9.
2. Background
A number of previous observational and modeling studies have explored the relationship between the extratropical circulation and the initiation of tropical convection and equatorially trapped waves. Observations suggest that tropical convection can be forced from the extratropics via pressure surges from strong midlatitude systems in the winter hemisphere (Lau 1982; Love 1985a,b; Hsu et al. 1990; Meehl et al. 1996; Compo et al. 1999), or by an upper tropospheric trough propagating equatorward through the Tropics during boreal winter, in a region of equatorial upper tropospheric mean westerlies (Liebmann and Hartmann 1984; Kiladis and Weickmann 1992, 1997; Kiladis 1998; Liebmann et al. 1999). Williams (1981) and Lau (1982) present case studies of northeasterly cold surges off the Asian continent during boreal winter that are followed by the initiation of equatorially symmetric, eastward-propagating convective anomalies near Borneo. Both authors suggest that these convective anomalies are related to Kelvin waves. Kiladis (1998) presents observations of lower tropospheric equatorially trapped Rossby wave structures initiated following a deep convective outbreak in the eastern Pacific, which itself is initiated by an equatorward-propagating extratropical Rossby wave train. Magaña and Yanai (1995) document a correlation between Southern Hemisphere westward-propagating extratropical circulation anomalies and the initiation of equatorially trapped mixed Rossby–gravity waves in the central Pacific during boreal summer and fall. In summary, observations suggest that extratropical circulation features may be responsible for forcing certain large-scale regions of deep tropical convection, and may also play a role in initiating equatorial waves.
Theoretical and modeling studies have also addressed the topic of the extratropical forcing of equatorially trapped waves and tropical convection. For example, Mak (1969) demonstrates that stochastic forcing at the latitudinal boundaries of a two-layer model can excite tropical waves with mixed Rossby–gravity and equatorial Rossby mode characteristics. Lamb (1973) presents solutions to a more complex model, which show that lateral forcing from the extratropics can excite equatorially trapped waves whose structures are then modified by the presence of condensational heating. Lim and Chang (1981) illustrate in a shallow water model that extratropical mass forcing analogous to a lower tropospheric pressure surge excites a spectrum of equatorially trapped waves including Kelvin, mixed Rossby–gravity, and n = 1 equatorial Rossby waves. Matthews and Kiladis (2000) excite equatorward-propagating Rossby waves in the eastern Pacific in a baroclinic model by perturbing the climatological boreal winter east Asian jet, and show that lower tropospheric n = 1 equatorial Rossby wave structures can be excited through the addition of diabatic forcing. Lin et al. (2000) simulated convectively coupled equatorial waves with dispersion characteristics similar to both Kelvin waves and the MJO using the quasi-equilibrium tropical circulation model of Neelin and Zeng (2000). These waves were significantly suppressed when extratropical disturbances were not allowed to excite low latitude waves, suggesting that a substantial portion of at least eastward-propagating equatorial wave activity can be excited by lateral forcing.
Two additional studies that are particularly relevant to the present work are those of Zhang (1993) and Hoskins and Yang (2000). Both studies find that a propagating extratropical vorticity source can force a spectrum of equatorially trapped waves in the Tropics, including Kelvin waves, even in regions of mean tropical easterlies. Previously, it had been thought that such tropical–extratropical interaction would be suppressed in regions of tropical easterlies, based on the Rossby wave propagation theories of Charney (1963) and Webster and Holton (1982). These theories predict that Rossby waves become evanescent at their critical line, where their zonal phase speed equals that of the background zonal wind. However, Zhang (1993) demonstrates that Kelvin wave amplitudes should actually be larger in the presence of equatorial easterlies, for eastward-propagating forcing with periods greater than 6 days. Similarly, Hoskins and Yang (2000) show that the strongest Kelvin wave response to eastward-propagating forcing occurs in tropical easterlies, for low zonal wavenumbers. In general, the equatorial Kelvin wave response is shown to maximize when the Doppler-shifted forcing frequency approaches the modal eigenvalue, that is, when the local forcing speed matches the theoretical Kelvin wave phase speed for a given basic-state flow. These results allow for the possibility of the remote forcing of Kelvin waves from the extratropics even when the equatorial winds are easterly, as is the case in the tropical Pacific during austral winter.
The question then arises as to which Kelvin wave phase speeds can be forced from the extratropics in the earth's atmosphere. Observations show that dry Kelvin waves exist in the Tropics over a wide range of frequencies and phase speeds (see Salby and Garcia 1987), but that moist (or convectively coupled) Kelvin waves are constrained to Doppler-shifted phase speeds between 15–20 m s–1 (Takayabu 1994; WK99). If, then, convectively coupled Kelvin waves do represent an intrinsic mode of the moist atmosphere, it appears that they could potentially be forced by extratropical vorticity perturbations that also propagate eastward at 15–20 m s–1. We thus hypothesize that extratropical vorticity forcing in the form of eastward-propagating perturbations in the subtropical jet might be able to force equatorial Kelvin waves, even in the presence of equatorial easterlies. The remainder of this study seeks to show that this is indeed the case in the Pacific during austral winter.
3. Data and methodology
Two primary datasets are utilized in this study. The National Oceanic and Atmospheric Administration (NOAA) outgoing longwave radiation (OLR) data are used to represent large-scale deep tropical convection, and European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis fields are used to represent the atmospheric circulation. Both datasets are available globally, on a regular 2.5° horizontal grid, and are averaged to daily time resolution. The reanalysis data are available on 16 pressure levels in the vertical, from 1000 to 10 hPa, and extend from 1979 to 1993. The OLR data extend from 1979 to 2001 (Liebmann and Smith 1996).
The relationship between Kelvin wave convection and the tropical and extratropical circulation is determined through a linear regression technique, in a similar manner to many previous studies of this type (Kiladis and Weickmann 1992, 1997; Kiladis 1998; WKW00; SK02). The majority of the results in the present study are obtained by linearly regressing daily ECMWF reanalysis fields (horizontal and vertical winds, temperature, and geopotential height) and total OLR against an index of Kelvin wave OLR, for the 15 austral winters (JJA) from 1979 to 1993, as a function of temporal lag. The Kelvin wave OLR index is constructed by filtering the total OLR data in wavenumber–frequency space such that the output includes only variability in the region surrounding the climatological Kelvin wave spectral peak (see WKW00 for more details). The filter spans periods of 2.5 to 17 days and wavenumbers 1–14, specifically isolating disturbances with eastward phase speeds from 8 to 30 m s–1 (see WK99, WKW00, or SK02 for an illustration). A reverse transform in space and time then results in a filtered dataset that includes only variability on these preferred Kelvin wave time- and space scales. The Kelvin wave index used in this study is the daily value of the Kelvin wave–filtered OLR at the point of its climatological variance maximum in austral winter, which is at 7.5°N, 172.5°W (see Fig. 1). The linear regression results based on this index are then scaled to a −40 W m–2 anomaly in OLR at the basepoint on day 0, a typical value for a strong Kelvin wave event. The statistical significance of these results is calculated based on a local two-sided significance test, which takes into account the correlation coefficients and a reduced number of degrees of freedom based on the decorrelation timescale, as in Livezey and Chen (1983). Results are considered significant at the 95% level or greater.
To assess the robustness of the relationship between convection and circulation determined from the Kelvin wave–filtered OLR, regressions were also calculated based on several other indices of tropical convection and the extratropical circulation. These results show strong similarities with those based on the Kelvin wave OLR, and are discussed in section 5a. In addition, we have calculated composite disturbances based on strong Kelvin wave events. The composite results are also very similar to those calculated in the regressions, based on both positive and negative OLR perturbations. This suggests a strong linearity in the observed Kelvin wave dynamical fields, and supports the use of the linear regression technique.
In section 7, 200-hPa National Centers for Environmental Prediction (NCEP)–National Center for Atmospheric Research (NCAR) reanalysis data are used in place of the ECMWF reanalysis data in constructing composites of monthly averaged circulation parameters related to Kelvin wave activity in the Pacific, because of the longer data record, which extends from 1979 to 2001. The analysis presented in section 7 has been carried out using the overlapping periods of 1979–93 in both the ECMWF and NCEP–NCAR datasets, and the results are nearly identical, as are the case study results presented in section 8. Finally, monthly averaged Reynolds sea surface temperature (SST) data are utilized in section 8. These data were obtained from the NOAA–Cooperative Institute for Research in Environmental Sciences (CIRES) Climate Diagnostics Center on a 1° grid, and interpolated to a 2.5° grid to match the resolution of the OLR and reanalysis data.
4. Climatology
The total OLR field in Fig. 1 (dark shading) is dominated by convection in the Asian monsoon region. Low OLR extends across the Pacific ITCZ, from the maritime continent to the coast of Central America, along 5°–10°N. The distribution of subseasonal convective variance closely follows the distribution of OLR, with greater subseasonal variance in regions of lower OLR (not shown). In the central Pacific, between the two centers of intense convective activity in the Asian and North American monsoon regions, lies the maximum in Kelvin wave activity, as represented by the variance of the Kelvin wave–filtered OLR (dark contours). The maximum variance lies to the north of the equator, just to the east of the dateline, with high variance extending along the latitude of maximum total convection (minimum OLR). Note that the maximum in Kelvin wave activity is collocated with a relative minimum in total convection along the ITCZ. A similar overall relationship emerges between Kelvin wave OLR variance and a variety of parameters representing overall convection and the large-scale moist environment in the Pacific, including total OLR variance, SST, and boundary layer equivalent potential temperature (not shown). These relationships suggest that while tropical Kelvin wave activity is modulated by the large-scale lower boundary conditions (that is, activity exists only in regions satisfying certain threshold conditions on low-level moisture, SST, etc.), other factors are involved in determining the distribution of this activity within the ITCZ.
Figure 1 also shows the 200-hPa zonal wind field (light contours) during austral winter. The extratropical wind field is dominated by the Southern Hemisphere subtropical jet, which maximizes to the east of Australia along 30°S. Extending along the jet core is a distinct maximum in submonthly meridional wind variance (light shading), which stretches from Australia eastward to approximately 120°W. The meridional wind variance is chosen as a proxy for extratropical Rossby wave activity. A similar distribution of activity is also found using streamfunction or vorticity variance. The preferred direction of energy propagation of these submontly jet perturbations can be estimated by the direction of the E-vectors, which point northeastward. It thus appears that during austral winter, submonthly Rossby wave perturbations in the subtropical jet tend to propagate eastward and equatorward in the vicinity of Australia. The equatorward propagation of individual Rossby wave trains should theoretically be facilitated by the existence of westerlies at relatively low latitudes in the central Pacific, allowing Rossby waves to propagate quite close to the equator before they are influenced by their critical line (Webster and Holton 1982). Note that the maximum Kelvin wave variance is located close to the longitude at which the subtropical westerlies extend farthest toward the equator.
Observations by Chang (1999) show that the dominant wave propagation pattern in this region during austral winter is consistent with the above conclusions. A wave train of zonal wavenumber 6 propagates eastward over the south Indian Ocean at approximately 8–12 m s–1. When this wave train reaches Australia, it splits into two branches. The stronger northern branch propagates northeastward, toward the equator, and the weaker southern branch continues to propagate eastward along 50°–60°S. This preferred wave pattern will be shown in section 5 to be associated with the initiation of convectively coupled Kelvin waves in the central Pacific.
5. Horizontal structure
In this section, results are presented that link equatorward-propagating transient wave activity over Australia to the initiation of a convectively coupled Kelvin waves in the Pacific. The results are based on linear regressions of ECMWF reanalysis data against the Kelvin-filtered OLR index at the basepoint 7.5°N, 172.5°W, for the 15 austral winters from 1979 to 1993, as described in section 3.
a. 200 hPa
Figure 2 shows the regressed values of OLR (shading) and 200-hPa streamfunction (contours) and winds (vectors, plotted only where significant at the 95% level or greater), for days −9, −5, 0, and +3. On day −9 (Fig. 2a), a highly statistically significant wave train in the Southern Hemisphere stretches eastward from South America to the east of Australia, near the date line. This wave packet propagates eastward in time, with new circulation centers forming to the east of existing centers, due to the eastward dispersion of energy. The wave train splits into two branches to the east of Australia, as seen on day −5 in Fig. 2b, with the stronger northern branch first propagating equatorward and then eastward along 20°S, and the weaker southern branch propagating eastward along approximately 50°S. This wave train pattern represents the preferred dispersion path for energy in the Southern Hemisphere subtropical jet during austral winter (Chang 1999), as discussed in section 4. The positive tilt of the anomalies in the northern wave train, from northwest to southeast, implies an equatorward transport of wave energy and a poleward transport of westerly momentum. The structure, tilt, and spatial scale (zonal wavenumber 6) of the circulation anomalies on day −5 are reminiscent of the LC1 “backward-tilted” baroclinic wave life cycle of Thorncroft et al. (1993, their Fig. 7), suggesting that these disturbances may represent the most unstable baroclinic mode of the subtropical jet. Straus and Lindzen (2000) have suggested that subtropical jet instability is also important in forcing MJO events.
Low (high) OLR, indicating enhanced (suppressed) deep convection in the Tropics (or upper-level cirrus in the subtropics) and represented by the dark (light) shading, forms to the northwest of the high (low) pressure center in the northern branch of the wave packet by day −5 (Fig. 2b), and is linked with poleward (equatorward) 200-hPa flow to its south in the Southern Hemisphere subtropics. The horizontal scale of the tropical OLR anomalies is similar to the scale of the subtropical circulation anomalies, with a zonal wavelength of approximately 60°. Anomalous 200-hPa divergence (convergence) exists in the poleward (equatorward) subtropical flow as the wave train propagates equatorward (not shown), as might be expected based on quasigeostrophic Rossby wave vorticity arguments. The upper tropospheric divergence fields and their associated vertical motion fields spread northward and eastward as the wave train approaches the equator. Figure 3 illustrates the 400-hPa vertical motion field on day −5 (shading), along with the OLR (hatching) and 200-hPa streamfunction (contours) and winds (vectors). Low (high) OLR on the equator is collocated with 400-hPa upward (downward) motion, which has propagated into the Tropics from the subtropics. This structure is consistent with the modeling results of Hoskins and Yang (2000; see their Figs. 9 and 10), in which an eastward-propagating extratropical wave train is shown to excite Kelvin wave–like vertical motion anomalies in the Tropics. Since the observed extratropical wave train exists for several days before the tropical and subtropical Pacific convective anomalies appear, it is suggested that the circulation forces the vertical motion and cloudiness signals at this time.
In the following days, the tropical and subtropical OLR anomalies propagate eastward with the subtropical circulation centers at phase speeds between 15–20 m s–1. The tropical OLR anomalies propagate somewhat faster than the subtropical circulation anomalies, such that by day 0 (Fig. 2c), low (high) OLR is located almost directly to the north of the subtropical high (low) pressure cell. It is unclear at this developed stage the extent to which the circulation is forcing the convection, or the convection is forcing the circulation, since upper tropospheric divergence associated with deep convection would be expected to spin up anticyclonic circulations in the subtropics, with the stronger circulation in the winter hemisphere, as observed. However, the faster phase speed of the convection, coupled with the divergent wind anomalies emanating from the center of low OLR, suggests that a substantial portion of the upper tropospheric circulation anomalies at this time is due to the deep convection itself.
The deepest convection on day 0, as represented by the negative OLR anomalies in Fig. 2c, is centered to the north of the equator at 7.5°N (as determined by the choice of basepoint). Convection continues to maximize in the Northern Hemisphere as it propagates eastward into the eastern Pacific at positive lags (Fig. 2d). This off-equatorial shift of convection is believed to be a consequence of the asymmetric SST distribution in the eastern Pacific, where cold SSTs lie along the equator and warm SSTs exist farther to the north (SK02).
Several additional regressions were calculated to assess the robustness of the above results. First, 30-day high-pass-filtered OLR was used as the independent variable instead of the Kelvin wave–filtered OLR. The results from this regression are similar to those shown in Fig. 2, demonstrating that eastward-propagating Kelvin wave OLR anomalies represent the dominant submonthly convective variability in the central Pacific ITCZ during austral winter. Regressions were also calculated based on Kelvin-filtered OLR at a basepoint on the equator and 200-hPa Kelvin-filtered zonal wind on the equator, and results are also similar to those shown in Fig. 2. A regression was then calculated to determine whether an eastward-propagating tropical OLR signal could be detected when the independent variable is based purely on an extratropical dynamical field. The 200-hPa meridional wind was used as the independent variable, which was filtered in the same space–time region as the Kelvin wave OLR. This isolates disturbances with eastward phase speeds between 8 and 30 m s–1. The basepoint for this regression was the point of the maximum JJA filtered meridional wind variance, at 27.5°S, 127.5°E. The results using this circulation index are similar to those using tropical Kelvin wave OLR, with an upper tropospheric wave train propagating eastward and equatorward over Australia (Fig. 4). The tropical OLR signal is weaker, but still retains the same phase relationship with the circulation anomalies as in the original Kelvin OLR regression (cf. with Fig. 2b). These results clearly demonstrate that Kelvin waves are associated with perturbations in the subtropical jet, and suggest that Kelvin waves might be initiated as a response to the eastward-propagating forcing of the subtropical Rossby wave dynamical fields.
Based on the regressed fields shown in Fig. 2 and those discussed in the preceding paragraph, we hypothesize that enhanced (suppressed) convection, as represented by low (high) OLR, is forced at least partially as a response to the upper tropospheric divergence (convergence) field and upward (downward) motion induced by the equatorward-propagating Rossby wave train. Of course, the initiation of deep tropical convection depends not only on large-scale vertical motion, but also on the convergence of warm, humid air in the boundary layer. This topic is addressed in sections 5c and 6.
b. 850 hPa
Figure 5 shows the 850-hPa circulation anomalies on day −5. Only the southern branch of the 200-hPa wave train shown in Fig. 2b extends into the lower troposphere, with anomalies at 850 hPa centered along approximately 40°S. As the upper tropospheric wave train splits into two branches prior to day −5, its equatorward-propagating branch largely decouples from the lower troposphere, while its poleward branch retains a baroclinic structure, with circulation anomalies displaying poleward and westward tilts with height. The northern wave train becomes progressively shallower in the vertical as it approaches the equator, due to the influence of the lower tropospheric equatorial easterlies, in a similar manner to the equatorward-propagating Rossby waves in the eastern Pacific studied by Kiladis (1998).
A strong cyclone is centered to the south of Australia near 140°E on day −5, directly to the south of the emerging low OLR anomaly on the equator. It will be shown in the following section that this cyclone affects the low-level tropical wind and pressure fields so as to promote convergence and the initiation of deep convection to its north. Throughout the evolution of the tropical Kelvin wave fields, the southern branch of the wave train remains remarkably coherent, both in the vertical and horizontally with respect to the Kelvin wave OLR anomaly.
c. 1000 hPa
The large-scale extratropical circulation features associated with the growth and decay of the tropical Kelvin wave OLR signal are similar between 850 hPa and 1000 hPa. An interesting feature of the 1000-hPa maps, however, is the existence of Kelvin wave–like height and temperature anomalies in the Tropics, which appear to originate as a response to the extratropical baroclinic waves, and which precede the development of the initial convective anomalies.
Figure 6 shows the 1000-hPa geopotential height field (Fig. 6a) and temperature field (Fig. 6b) on day −5, corresponding to the 200- and 850-hPa maps shown in Figs. 2b and 5. To the east of the high OLR anomaly (single hatching, representing suppressed deep convection) is a region of high heights and low temperatures at 1000 hPa, which has a fairly symmetric distribution with respect to the equator, suggesting a Kelvin wave–like structure. Similar structures are seen in the extratropically forced modeling results of Hoskins and Yang (2000; their Fig. 11b). The observed height and temperature anomalies in Fig. 6 reached the Tropics on day −8 (not shown), having propagated northward from a high pressure system centered to the southeast of Australia, in a surgelike manner similar to that documented by Love (1985a) and analogous to the surges often seen propagating southeastward from Asia in boreal winter [Compo et al. (1999) and references therein]. The surge propagates northward at a phase speed of at least 25 m s–1, and is constrained to the lowest 2 km of the atmosphere. Regressions using radiosonde data from Darwin and Alice Springs, Australia, for the years 1979–2000 confirm the existence of the surge, its shallowness, and its timing relative to the initiation of deep convection in the Tropics (not shown). The subtropical high responsible for this pressure surge has since propagated eastward to 180° by day −5, and remains connected to the eastward-propagating height anomaly along the equator.
Between the tropical high OLR anomaly at approximately 165°E and the developing low OLR anomaly to its west on day −5 is a similar height and temperature signal of the opposite sign. Low heights and warm temperatures are centered on the equator, symmetric with respect to the equator, and are linked to extratropical perturbations of the same sign associated with the developing low pressure system centered at 40°S, 145°E. These tropical height and temperature anomalies appear first on day −5, and then propagate eastward with the tropical convection and extratropical circulation. Note that high (low) 1000-hPa geopotential heights lead high (low) OLR by a quarter wavelength, a relationship consistent with linear Kelvin wave theory if high (low) OLR is considered a proxy for divergence (convergence), as might be expected for the lower troposphere.
The equatorial pressure gradient on day −5 between 150°E and 170°W is followed by the appearance of easterly winds flowing from high to low heights on day −4 between 10°S and 10°N (not shown). The rapid development of tropical wind anomalies in response to an extratropically forced equatorial pressure gradient is also documented by Chu (1988). These easterlies produce a region of low-level zonal wind convergence to the east of the developing low OLR anomaly, which converges moisture and increases the convective available potential energy (CAPE) in this region (not shown). This relationship, of low-level convergence leading low OLR, is retained throughout the evolution of the wave, as shown in Fig. 7, a longitude–time diagram of the regressed OLR (shading) and 1000-hPa convergence (contours), averaged from 5°S–10°N. Note that the maximum 1000-hPa convergence leads the minimum OLR by approximately 15° longitude, or less than a quarter wavelength, throughout the wave's life cycle. The regressed 1000-hPa specific humidity, CAPE, and 700-hPa upward motion fields are all in phase with the 1000-hPa convergence (not shown), and thus also lead the minimum OLR by less than a quarter wavelength. The low OLR anomaly is itself in phase with 400-hPa upward motion. From these relationships we conclude that the deepest convection develops only after a period of lower tropospheric convergence and upward motion, which gradually moistens the lower troposphere over a time span of approximately one day. This idea is consistent with results from a case study of a developed Kelvin wave in the eastern Pacific by SK02, which show that shallow convection moistens the lower troposphere (below 700 hPa) for approximately 24 h prior to the lowest OLR anomaly.
Based on the results in this section, it appears that the initial convective anomaly associated with the Kelvin wave is forced by the upper tropospheric upward motion anomaly that propagates equatorward from the extratropics. The lower tropospheric fields then converge boundary layer moisture to the east of this initial OLR anomaly, creating a region favorable for subsequent deep convection. The tropical convective portion of the wave propagates eastward with the subtropical circulation anomalies, but presumably also provides its own internal dynamical and moisture feedbacks that modulate the influence of the extratropical anomalies.
6. Vertical structure
In this section, the vertical structures of the extratropical and tropical circulations associated with convectively coupled Kelvin wave OLR signals are examined. The focus in the current study is primarily on the circulations coincident with the development of the tropical OLR anomalies. The developed wave circulations associated with a convectively coupled Kelvin wave in the Indian Ocean region are discussed in WKW00, and a future study will more thoroughly describe the evolution of the developed wave circulations in the Pacific ITCZ.
Figure 8a is a longitude–height cross section along 20°S of the regressed meridional wind (contours), vertical motion (shading), and zonal–vertical circulation (vectors) on day −5. The northern branch of the subtropical Rossby wave train (see Fig. 2b) is clearly visible in Fig. 8a, as evidenced by the vertically coherent positive and negative meridional wind anomalies (contours), which alternate in sign in the horizontal and maximize at 200 hPa. Vertical motion anomalies (shading) are in phase with the meridional wind maxima, with upward (downward) motion below 200 hPa in the regions of northerly (southerly) flow. Opposite-signed anomalies occur above 200 hPa. Temperature anomalies along 20°S (not shown) are consistent with the expected structure of upper tropospheric potential vorticity anomalies, with warm (cold) air above and cold (warm) air below the centers of cyclonic (anticyclonic) vorticity.
The vertical motion anomalies centered at 130° and 160°E in Fig. 8a extend northward to the equator in the upper troposphere, as shown in Fig. 3 and as also seen in latitude–height cross sections along these longitudes (not shown). The circulations in the equatorial plane are shown in Fig. 8b, a longitude–height cross section along the equator showing the regressed temperature (contours), vertical motion (shading), and zonal–vertical circulation (vectors) on day −5. Vertical motion anomalies are centered at 135° and 160°E, directly to the north of those pictured in Fig. 8a, with maxima centered at approximately 300 hPa. OLR anomalies (top plot) are consistent with the vertical motion anomalies, with low (high) OLR corresponding to regions of upper tropospheric upward (downward) motion. A zonal–vertical circulation cell is present between 120°E and 180°. Note also that the upward motion field extends eastward below 600 hPa, such that upward motion in the lower troposphere is positioned approximately 15° to the east of that in the upper troposphere. This lower tropospheric upward motion anomaly is horizontally collocated with the maximum 1000-hPa convergence on day −5 (not shown). This relationship suggests that lower tropospheric upward motion is forced to the east of the developing OLR anomaly by the convergence of surface easterlies, which themselves develop in response to the geopotential height fields forced from the extratropics, as discussed in section 5c.
Also shown in Fig. 8b is the regressed temperature field on day −5 (contours). The temperature signals are weak throughout the troposphere at this time, and their relationship to the vertical motion fields no longer resembles the upper tropospheric potential vorticity anomaly structures at 20°S, where temperature and vertical motion anomalies are in quadrature. Instead, upward (downward) motion in the Tropics is collocated with a very weak warm (cold) anomaly in the upper troposphere and a cold (warm) anomaly in the lower troposphere, suggesting the presence of diabatic processes in the Tropics and a transitioning from a barotropic to a baroclinic structure as the wave moves equatorward (Wang and Xie 1996). The weak tropical temperature anomalies also suggest that upper tropospheric cold advection is not a primary factor in the initiation of deep convection in the Kelvin wave. These results contrast with those of Kiladis (1998), who documented strong upper tropospheric cold advection prior to the initiation of deep convection in eastern Pacific equatorward-propagating Rossby waves during boreal winter. Instead, convection in the present study appears to be initiated in regions of upper tropospheric upward motion and low-level moisture convergence forced from the extratropics.
As the coupled convective and dynamical fields of the Kelvin wave propagate eastward, the vertical motion anomalies strengthen and extend throughout the depth of the troposphere. Figure 8c is a longitude–height cross section of temperature (contours), vertical motion (shading), and the zonal–vertical circulation (vectors) along 7.5°N on day 0, when the Kelvin wave OLR has reached its minimum value at the basepoint 7.5°N, 172.5°W (see Figs. 2c and 5). The vertical motion field retains a westward tilt with height below 600 hPa, such that surface convergence continues to precede the lowest OLR anomaly by approximately 15°. In the region of negative OLR anomalies, between 180° and 160°W, the troposphere is warm above 500 hPa and cold below, and the tropopause is cold (near 100 hPa). This temperature structure is consistent with other studies of convectively coupled tropical waves (Reed and Recker 1971; Dunkerton 1993; Takayabu and Nitta 1993; Haertel and Johnson 1998; WKW00), and suggests the presence of a tropospheric heat source that includes both deep convective heating and a second baroclinic mode heating over cooling structure, which arises from stratiform precipitation. A case study of a convectively coupled Kelvin wave in the eastern Pacific by SK02 shows that both deep convective and stratiform precipitation signals do exist within a developed wave, with stratiform precipitation dominating after an initial convective period.
In the region of positive OLR anomalies in Fig. 8c, where deep convection is suppressed, the temperature signal is opposite in sign to that associated with enhanced deep convection. The upper troposphere is cold, and the lower troposphere and tropopause regions are warm. The temperature response of the atmosphere to the convectively coupled Kelvin wave heating fields thus appears to be quite linear, as also documented in composite fields (not shown). In the stratosphere, above 100 hPa, the temperature anomalies tilt eastward with height, as expected for an upward-propagating dry Kelvin wave forced from below (Andrews et al. 1987). The vertical wavelength of these temperature anomalies is approximately 6 km, which is consistent with the expected structure of an upward-propagating dry Kelvin wave of equivalent depth 40 m.
7. Longer timescale relationships
In this section, the relationship between jet variability and Kelvin wave variance on longer timescales is examined through the calculation of composite, or averaged fields. Monthly averages are used as a simple means of measuring the variability of tropical convection and the extratropical circulation on timescales longer than one typical Kelvin wave event, which lasts approximately 10 days from initiation to dissipation.
A time series of monthly averaged Kelvin wave variance in a 5° × 5° box around the Kelvin wave OLR basepoint (5°–10°N, 170°–175°W) is shown in Fig. 9, for the period 1979–2000. Note the highly skewed variability over this 22-yr period, with a number of months standing out as containing extremely high Kelvin wave variance. A majority of these months falls within the JJA period: for example, 13 of the 20 months with highest variance occur during JJA. A composite for these high variance events is calculated by averaging months within the JJA season that have a variance greater than one standard deviation above the JJA climatological mean (above 301 W2 m–4). There are 11 such months. A composite approach is chosen in this context instead of a linear regression approach because, as can be seen in Fig. 9, there is no corresponding set of months with anomalously low variance to provide the necessary linearity for such a technique. Instead, a composite of low variance events is calculated based on the 11 months during the JJA season with lowest variance. The composites are constrained to include months within the JJA period to remove any effects due to the seasonal cycle.
Shown in Figs. 10a and 10b are the high and low variance composites. Each plot includes anomalous Kelvin wave variance (contours), 200-hPa Kelvin wave–filtered meridional wind variance (shading), and 200-hPa Kelvin wave–filtered E-vectors. The Kelvin wave–filtered meridional wind variance and E-vectors were created by first filtering the daily 200-hPa zonal and meridional wind fields to the same wavenumber–frequency region as the Kelvin wave–filtered OLR, as described in section 3. This constrains the filtered disturbances to be eastward-propagating with phase speeds of 8–30 m s–1. This Kelvin wave–filtered variance represents between 25% and 50% of the 30-day high-pass-filtered variance, depending on location, in the region of the subtropical jet (as shown in Fig. 1). The wind fields were filtered in this manner to limit them to the same spatial scales and phase speeds as the tropical Kelvin wave OLR signals, since variability on these time-and space scales is dominant in the regressions. Monthly averaged fields were then calculated from the daily data. Finally, monthly anomalies were calculated by subtracting the average annual cycle at each grid point from the raw monthly data.
When monthly averaged Kelvin wave variance is anomalously high in the 5° × 5° box in the central Pacific, it is also anomalously high across the Pacific ITCZ, from 140°E to 80°W (Fig. 10a). Meridional wind variance at 200 hPa is anomalously high over Australia and to its east, in the same region as the climatological subtropical jet maximum in austral winter (see Fig. 1). This result illustrates that enhanced eastward-propagating wave activity within the subtropical jet is associated with enhanced Kelvin wave activity in the Pacific. These relationships are also observed in individual months (not shown). The composite E-vector anomalies give an indication of the change in the direction of energy propagation of the disturbances from the mean seasonal values (which are similar in direction to the 30-day high-pass-filtered E-vectors shown in Fig. 1, but smaller in magnitude due to the additional constraints of the Kelvin wave filtering). In the region of enhanced meridonal wind variance, the E-vectors primarily point eastward, signaling an enhanced eastward propagation of disturbance energy in the Kelvin-filtered band.
On the other hand, when monthly averaged Kelvin wave variance is anomalously low in the central Pacific, it is also anomalously low across the Pacific ITCZ (Fig. 10b). At the same time, meridional wind variance is decreased over Australia and to its west, signaling a decrease in eastward-propagating wave activity in the subtropical jet entering the region from the Indian Ocean. The anomalous E-vectors in this region of low meridional wind variance point westward, signaling that the disturbances that are present during these months have a group velocity with a smaller eastward component than the more typical disturbances in the jet.
These results, coupled with the regression results presented in Fig. 2, suggest that Kelvin wave activity in the central Pacific is sensitive to the amplitude and character of wave activity within the subtropical jet during austral winter. When disturbances in the jet are strong and move eastward with zonal phase speeds between 8–30 m s–1, Kelvin wave activity in the Pacific is enhanced. A decrease in eastward-propagating wave activity in the subtropical jet is associated with reduced Kelvin wave activity in the tropical Pacific.
One might argue that the enhanced subtropical wind variance in the high variance composite (Fig. 10a) can be accounted for by circulations forced by the Kelvin wave OLR perturbations in the Tropics, instead of the reverse, as is suggested above. While it is certainly the case that enhanced Kelvin wave activity in the Tropics should lead to enhanced circulation variability in the subtropics, the regressions shown in Fig. 2 confirm that the circulation anomalies occur prior to the initiation of the convective anomalies. In addition, the negative composite shows a decrease in wave activity to the west of the Kelvin wave OLR variance minimum, in the region in which the equatorward-propagating Rossby wave packets originate. This, again coupled with the regressions shown in Fig. 2, suggests that a decrease in extratropical wave activity over Australia leads to a decrease in Kelvin wave activity in the Pacific.
Plots of individual months (not shown) suggest that the magnitude of Kelvin wave activity in the central Pacific depends not only on the wave activity in the subtropical jet, but, as might be expected, also on the underlying thermal and moisture boundary conditions in the Pacific. When SSTs in the central Pacific are anomalously low, Kelvin wave activity is suppressed even when the subtropical jet is very active. On the other hand, Kelvin wave activity can be suppressed even when the thermal boundary conditions are favorable, as will be shown in the following section. These relationships suggest that an enhancement of Kelvin wave activity in the central Pacific depends on both warm SSTs and an influx of extratropical wave energy.
8. Case study: July 1986 versus July 1987
The conclusions reached in the previous section can be illustrated even more dramatically by considering a case study. The months of July 1986 and July 1987 provide an interesting juxtaposition, and will be examined in this section.
In Fig. 11, monthly averaged maps of total SST (hatching), total Kelvin wave OLR variance (dark contours), anomalous 200-hPa Kelvin wave–filtered meridional wind variance (shading), and total E-vectors are shown for (a) July 1986 and (b) July 1987. SST in the western and central Pacific is quite warm during both months, consistent with the fact that both months fall within the 1986/87 warm El Niño–Southern Oscillation (ENSO) event. The warm event was in its initial stages in July 1986 and near its peak in July 1987, as measured by Niño-3.4 SST anomaly values of +0.33 and +1.78, respectively (see Trenberth 1997). Between 160°E and 140°W, SSTs are significantly warmer in July 1987 than in July 1986, with a distinct maximum in SST just south of the equator near the dateline in July 1987. Total OLR variance follows this SST pattern, with higher variance in the central Pacific during July 1987 (not shown). Based on these SST and OLR distributions alone, one might predict that Kelvin waves should be more active in the Pacific during July 1987 than during July 1986.
However, as shown by the dark contours in Figs. 11a and 11b, Kelvin wave activity is unusually intense during July 1986, and very weak during July 1987. The peak Kelvin wave variance in July 1986 exceeds 400 W2 m–4, while in July 1987, values maximize at 100 W2 m–4, a difference of a factor of 4. Consistent with the hypothesis that extratropical wave activity is associated with the initiation of Kelvin waves, the anomalous 200-hPa meridional wind variance distribution also shows substantial differences between July 1986 and July 1987. During July 1986 (Fig. 11a), meridional wind variance is anomalously strong eastward of 120°E, most notably in the region to the northeast of Australia, in the climatologically preferred region for equatorward-propagating wave trains in austral winter. Total E-vectors point northeastward toward the Tropics in this region. The 200-hPa jet core lies over southeastern Australia during July 1986, centered at approximately 150°E (not shown).
Contrast this situation with that in July 1987 (Fig. 11b). The meridional wind variance is anomalously weak between the subtropical Indian Ocean and eastern Pacific, signifying a decrease in eastward-propagating subtropical wave activity in the jet region. The total E-vectors in this region point to the northwest. The jet core during July 1987 is located much farther to the east than during July 1986, centered at approximately 160°W (not shown), which is consistent with the eastward displacement of the local Hadley circulation during a warm ENSO event.
Daily maps of submonthly circulation anomalies for the months of July 1986 and July 1987 illustrate two quite different flow regimes (not shown), as might be expected based on the difference in the 200-hPa meridional wind variance maps. During July 1986, wave packets flow freely to the east, often propagating northeastward over Australia as in Fig. 2. On the other hand, during July 1987, a blocking episode over Australia from 9 to 15 July disrupts the eastward propagation of waves. The difference between these two cases suggests the important role of fluctuations in the subtropical circulation on Kelvin wave variability in the Tropics.
9. Summary and conclusions
A linear regression technique is used to determine the preferred circulation and convection patterns associated with the initiation and development of a convectively coupled Kelvin wave in the Pacific ITCZ during austral winter. A baroclinically developing wave packet in the Southern Hemisphere subtropical jet propagates eastward and equatorward over Australia, with a phase speed of approximately 15 m s–1. The upper tropospheric divergence–convergence pair associated with the subtropical ridge and trough induces vertical motion anomalies that spread toward the Tropics and acquire equatorially trapped, Kelvin wave–like characteristics. The zonal phase speed and wavelength of the developing Kelvin wave anomalies are similar to those in the extratropical perturbations, suggesting that the extratropical circulation anomalies are of the necessary spatial scale and frequency to excite a convectively coupled Kelvin wave. As the upper tropospheric vertical motion anomalies develop, a lower tropospheric pressure surge excited by the baroclinically developing extratropical wave train also produces Kelvin-like temperature and height anomalies in the equatorial region. An easterly trade surge forms as mass flows from high to low pressure at the surface, causing moisture convergence and upward motion in the lower troposphere ahead of the developing low OLR anomaly. The initiation of deep convection in the Kelvin wave appears to be due to a combination of the upper tropospheric vertical motion forced by the upper tropospheric Rossby wave train, and lower tropospheric moisture convergence and upward motion due to the extratropically generated pressure surge.
Once a convectively coupled Kelvin wave is established, it appears that it can be self-sustaining; that is, in many individual cases, eastward-propagating convective anomalies persist even after the initial subtropical circulation anomalies dissipate. This observation is consistent with the results of Lamb (1973), who suggested that equatorial waves may be initially excited by extratropical forcing, but that condensational heating provides the energy for the disturbances to continue propagating once they are excited. In addition, recent primitive equation model simulations suggest that an eastward-propagating, Kelvin wave–like thermal forcing can induce a low-level convergence signal that propagates eastward out ahead of the forcing, further promoting the eastward propagation of convection and the Kelvin wave structure as a whole (M. Wheeler 2002, personal communication). Furthermore, Haertel and Johnson (2000) show in a dry linear model that a moving thermal forcing creates a stronger gravity mode response in its direction of motion. Thus an eastward-propagating Kelvin wave thermal forcing should affect the boundary layer fields more strongly to its east, and through convergence and upward motion it may precondition the atmosphere for deep convection in this region. Mapes (2000) and Majda and Shefter (2001) suggest that convectively coupled Kelvin waves can be viewed as a result of “stratiform instability,” which requires only two vertical modes: a deep convective heating and a stratiform heating over cooling. These studies suggest that convectively coupled Kelvin waves, once initiated, may provide the necessary internal feedbacks to maintain their convective–dynamical structure as they propagate eastward.
On timescales longer than one individual wave, Kelvin wave variability in the Pacific is modulated by the strength of eastward-propagating extratropical wave activity in the subtropical jet. Increased Kelvin wave activity in the Pacific is associated with enhanced eastward-propagating wave activity in the jet, which typically propagates with a more equatorward group velocity than typical jet perturbations. Suppressed Kelvin wave activity, on the other hand, is associated with suppressed jet activity, and a weaker equatorward group velocity.
Two contrasting months during the 1986/87 warm ENSO event are compared, illustrating that Kelvin wave activity can be more strongly modulated by the strength of eastward-propagating subtropical wave activity than by changes in local SST. During July 1986, SSTs in the Pacific were marginally warm and extratropical wave activity was quite strong. Kelvin wave activity was anomalously strong during this time period. On the other hand, during July 1987, SSTs in the Pacific were very warm, but extratropical wave activity was weak. Kelvin wave activity during this period was suppressed. These two months provide an interesting comparison and support our theory that extratropical wave activity may be a precursor to Kelvin wave activity in the Pacific. An interesting sidenote to this case study is that, while July 1987 was largely devoid of Kelvin wave activity in the Pacific, a strong convectively coupled mixed Rossby–gravity wave packet was observed in this region during this time (Dickinson and Molinari 2002). This observation suggests that the equatorial modes into which tropical convection is organized may be dependent on the space–time characteristics of wave energy impinging on the tropical atmosphere from the extratropics.
Acknowledgments
The OLR, ECMWF reanalysis, NCEP–NCAR reanalysis, and SST data used in this study were obtained from the NOAA–CIRES Climate Diagnostics Center. Thanks to Adrian Matthews, David Randall, Wayne Schubert, Matthew Wheeler, and two anonymous reviewers for providing helpful comments on earlier versions of this paper. This work was supported by the Pan American Climate Studies Program of the NOAA Office of Global Programs under Project GC98-627.
REFERENCES
Andrews, D. G., J. R. Holton, and C. B. Leovy, 1987: Middle Atmospheric Dynamics. International Geophysics Series, Vol. 40, Academic Press, 489 pp.
Chang, E. K. M., 1999: Characteristics of wave packets in the upper troposphere. Part II: Seasonal and hemispheric variations. J. Atmos. Sci., 56 , 1729–1747.
Charney, J. G., 1963: A note on large-scale motions in the Tropics. J. Atmos. Sci., 20 , 607–609.
Chu, P-S., 1988: Extratropical forcing and the burst of equatorial westerlies in the western Pacific: A synoptic study. J. Meteor. Soc. Japan, 66 , 549–564.
Compo, G. P., G. N. Kiladis, and P. J. Webster, 1999: The horizontal and vertical structure of east Asian winter monsoon pressure surges. Quart. J. Roy. Meteor. Soc., 125 , 29–54.
Dickinson, M., and J. Molinari, 2002: Mixed Rossby–gravity waves and western Pacific tropical cyclogenesis. Part I: Synoptic evolution. J. Atmos. Sci., 59 , 2183–2196.
Dunkerton, T. J., 1993: Observation of 3–6 day meridional wind oscillations over the tropical Pacific, 1973–1992: Vertical structure and interannual variability. J. Atmos. Sci., 50 , 3292–3307.
Haertel, P. T., and R. H. Johnson, 1998: Two-day disturbances in the equatorial western Pacific. Quart. J. Roy. Meteor. Soc., 124 , 615–636.
Haertel, P. T., and R. H. Johnson, 2000: The linear dynamics of squall line mesohighs and wake lows. J. Atmos. Sci., 57 , 93–107.
Holton, J. R., 1972: Waves in the equatorial stratosphere generated by tropospheric heat sources. J. Atmos. Sci., 29 , 368–375.
Hoskins, B. J., and G-Y. Yang, 2000: The equatorial response to higher-latitude forcing. J. Atmos. Sci., 57 , 1197–1213.
Hoskins, B. J., I. N. James, and G. H. White, 1983: The shape, propagation and mean-flow interaction of large-scale weather systems. J. Atmos. Sci., 40 , 1595–1612.
Hsu, H-H., B. J. Hoskins, and F-F. Jin, 1990: The 1985/86 intraseasonal oscillation and the role of the extratropics. J. Atmos. Sci., 47 , 823–839.
Kiladis, G. N., 1998: Observations of Rossby waves linked to convection over the eastern tropical Pacific. J. Atmos. Sci., 55 , 321–339.
Kiladis, G. N., and K. M. Weickmann, 1992: Extratropical forcing of tropical Pacific convection during northern winter. Mon. Wea. Rev., 120 , 1924–1938.
Kiladis, G. N., and K. M. Weickmann, 1997: Horizontal structure and seasonality of large-scale circulations associated with submonthly tropical convection. Mon. Wea. Rev., 125 , 1997–2013.
Lamb, V. R., 1973: The response of the tropical atmosphere to middle latitude forcing. Ph.D. thesis, University of California, Los Angeles, 151 pp.
Lau, K-M., 1982: Equatorial response to northeasterly cold surges as inferred from satellite cloud imagery. Mon. Wea. Rev., 110 , 1306–1313.
Liebmann, B., and D. L. Hartmann, 1984: An observational study of tropical–midlatitude interaction on intraseasonal time scales during winter. J. Atmos. Sci., 41 , 3333–3350.
Liebmann, B., and C. A. Smith, 1996: Description of a complete (interpolated) outgoing longwave radiation dataset. Bull. Amer. Meteor. Soc., 77 , 1275–1277.
Liebmann, B., G. N. Kiladis, J. A. Marengo, T. Ambrizzi, and J. D. Glick, 1999: Submonthly convective variability over South America and the South Atlantic convergence zone. J. Climate, 12 , 1877–1891.
Lim, H., and C-P. Chang, 1981: A theory for midlatitude forcing of tropical motions during winter monsoons. J. Atmos. Sci., 38 , 2377–2392.
Lin, J. W-B., J. D. Neelin, and N. Zeng, 2000: Maintenance of tropical intraseasonal variability: Impact of evaporation–wind feedback and midlatitude storms. J. Atmos. Sci., 57 , 2793–2823.
Livezey, R. E., and W. Y. Chen, 1983: Statistical field significance and its determination by Monte Carlo techniques. Mon. Wea. Rev., 111 , 46–59.
Love, G., 1985a: Cross-equatorial influence of winter hemisphere subtropical cold surges. Mon. Wea. Rev., 113 , 1487–1498.
Love, G., 1985b: Cross-equatorial interactions during tropical cyclogenesis. Mon. Wea. Rev., 113 , 1499–1509.
Madden, R. A., and P. R. Julian, 1971: Detection of a 40–50 day tropical oscillation in the zonal wind in the tropical Pacific. J. Atmos. Sci., 28 , 702–708.
Madden, R. A., and P. R. Julian, 1994: Observations of the 40–50-day tropical oscillation—A review. Mon. Wea. Rev., 122 , 814–837.
Magaña, V., and M. Yanai, 1995: Mixed Rossby–gravity waves triggered by lateral forcing. J. Atmos. Sci., 52 , 1473–1486.
Majda, A. J., and M. G. Shefter, 2001: Models for stratiform instability and convectively coupled waves. J. Atmos. Sci., 58 , 1567–1584.
Mak, M-K., 1969: Laterally driven stochastic motions in the Tropics. J. Atmos. Sci., 26 , 41–64.
Mapes, B. E., 2000: Convective inhibition, subgrid-scale triggering energy, and stratiform instability in a toy tropical wave model. J. Atmos. Sci., 57 , 1515–1535.
Matthews, A. J., and G. N. Kiladis, 2000: A model of Rossby waves linked to submonthly convection over the eastern tropical Pacific. J. Atmos. Sci., 57 , 3785–3798.
Meehl, G. A., G. N. Kiladis, K. M. Weickmann, M. Wheeler, D. S. Gutzler, and G. P. Compo, 1996: Modulation of equatorial subseasonal convective episodes by tropical–extratropical interaction in the Indian and Pacific Ocean regions. J. Geophys. Res., 101 , 15033–15049.
Milliff, R. F., and R. A. Madden, 1996: The existence and vertical structure of fast, eastward-moving disturbances in the equatorial troposphere. J. Atmos. Sci., 53 , 586–597.
Neelin, J. D., and N. Zeng, 2000: A quasi-equilibrium tropical circulation model—Formulation. J. Atmos. Sci., 57 , 1741–1766.
Reed, R. J., and E. E. Recker, 1971: Structure and properties of synoptic-scale wave disturbances in the equatorial western Pacific. J. Atmos. Sci., 28 , 1117–1133.
Salby, M. L., and R. R. Garcia, 1987: Transient response to localized episodic heating in the Tropics. Part I: Excitation and short-time near-field behavior. J. Atmos. Sci., 44 , 458–498.
Straub, K. H., and G. N. Kiladis, 2002: Observations of a convectively coupled Kelvin wave in the eastern Pacific ITCZ. J. Atmos. Sci., 59 , 30–53.
Straus, D. M., and R. S. Lindzen, 2000: Planetary-scale baroclinic instability and the MJO. J. Atmos. Sci., 57 , 3609–3626.
Takayabu, Y. N., 1994: Large-scale cloud disturbances associated with equatorial waves. Part I: Spectral features of the cloud disturbances. J. Meteor. Soc. Japan, 72 , 433–448.
Takayabu, Y. N., and T. Nitta, 1993: 3–5 day-period disturbances coupled with convection over the tropical Pacific Ocean. J. Meteor. Soc. Japan, 71 , 221–246.
Thorncroft, C. D., B. J. Hoskins, and M. E. McIntyre, 1993: Two paradigms of baroclinic-wave life-cycle behavior. Quart. J. Roy. Meteor. Soc., 119 , 17–55.
Trenberth, K. E., 1997: The definition of El Niño. Bull. Amer. Meteor. Soc., 78 , 2771–2777.
Wang, B., and X. Xie, 1996: Low-frequency equatorial waves in vertically sheared zonal flow. Part I: Stable waves. J. Atmos. Sci., 53 , 449–467.
Webster, P. J., and J. R. Holton, 1982: Cross-equatorial response to middle-latitude forcing in a zonally varying basic state. J. Atmos. Sci., 39 , 722–733.
Wheeler, M., and G. N. Kiladis, 1999: Convectively coupled equatorial waves: Analysis of clouds and temperature in the wavenumber–frequency domain. J. Atmos. Sci., 56 , 374–399.
Wheeler, M., G. N. Kiladis, and P. J. Webster, 2000: Large-scale dynamical fields associated with convectively coupled equatorial waves. J. Atmos. Sci., 57 , 613–640.
Williams, M., 1981: Interhemispheric interaction during Winter MONEX. Proc. Int. Conf. on Early Results of FGGE and Large-Scale Aspects of Its Monsoon Experiments, Vol. 10, Tallahassee, FL, WMO, 12–16.
Zhang, C., 1993: Laterally forced equatorial perturbations in a linear model. Part II: Mobile forcing. J. Atmos. Sci., 50 , 807–821.
Austral winter climatological fields: total OLR (dark shading), Kelvin wave–filtered OLR variance (dark contours), 200-hPa zonal wind (light contours), 30-day high-pass-filtered 200-hPa meridional wind variance (light shading), and 30-day high-pass-filtered E-vectors, based on the years 1979–93. OLR is shaded at 200, 220, and 240 W m–2, and is set to zero outside the range 20°S–30°N for plotting purposes. Kelvin OLR variance is contoured from 120 to 210 W2 m–4 by 30 W2 m–4. Zonal wind is contoured every 10 m s–1. Meridional wind variance is shaded at 160 and 200 m2 s–2. Longest E-vectors are 150 m2 s–2; vectors are not plotted below 10 m2 s–2
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Austral winter climatological fields: total OLR (dark shading), Kelvin wave–filtered OLR variance (dark contours), 200-hPa zonal wind (light contours), 30-day high-pass-filtered 200-hPa meridional wind variance (light shading), and 30-day high-pass-filtered E-vectors, based on the years 1979–93. OLR is shaded at 200, 220, and 240 W m–2, and is set to zero outside the range 20°S–30°N for plotting purposes. Kelvin OLR variance is contoured from 120 to 210 W2 m–4 by 30 W2 m–4. Zonal wind is contoured every 10 m s–1. Meridional wind variance is shaded at 160 and 200 m2 s–2. Longest E-vectors are 150 m2 s–2; vectors are not plotted below 10 m2 s–2
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Austral winter climatological fields: total OLR (dark shading), Kelvin wave–filtered OLR variance (dark contours), 200-hPa zonal wind (light contours), 30-day high-pass-filtered 200-hPa meridional wind variance (light shading), and 30-day high-pass-filtered E-vectors, based on the years 1979–93. OLR is shaded at 200, 220, and 240 W m–2, and is set to zero outside the range 20°S–30°N for plotting purposes. Kelvin OLR variance is contoured from 120 to 210 W2 m–4 by 30 W2 m–4. Zonal wind is contoured every 10 m s–1. Meridional wind variance is shaded at 160 and 200 m2 s–2. Longest E-vectors are 150 m2 s–2; vectors are not plotted below 10 m2 s–2
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Regressed values of OLR (shading) and 200-hPa streamfunction (contours) and winds (vectors), based on a −40 W m–2 anomaly in Kelvin wave–filtered OLR at the basepoint on day 0, for (a) day −9, (b) day −5, (c) day 0, and (d) day +3. OLR is shaded at ±6 and 15 W m–2; dark shading represents negative OLR anomalies. Streamfunction contour interval is 7.5 × 105 m2 s–1; the zero contour has been omitted. The longest wind vectors correspond to a 10 m s–1 wind, and are plotted only where either the u or υ component is significant at the 95% level or greater
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Regressed values of OLR (shading) and 200-hPa streamfunction (contours) and winds (vectors), based on a −40 W m–2 anomaly in Kelvin wave–filtered OLR at the basepoint on day 0, for (a) day −9, (b) day −5, (c) day 0, and (d) day +3. OLR is shaded at ±6 and 15 W m–2; dark shading represents negative OLR anomalies. Streamfunction contour interval is 7.5 × 105 m2 s–1; the zero contour has been omitted. The longest wind vectors correspond to a 10 m s–1 wind, and are plotted only where either the u or υ component is significant at the 95% level or greater
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Regressed values of OLR (shading) and 200-hPa streamfunction (contours) and winds (vectors), based on a −40 W m–2 anomaly in Kelvin wave–filtered OLR at the basepoint on day 0, for (a) day −9, (b) day −5, (c) day 0, and (d) day +3. OLR is shaded at ±6 and 15 W m–2; dark shading represents negative OLR anomalies. Streamfunction contour interval is 7.5 × 105 m2 s–1; the zero contour has been omitted. The longest wind vectors correspond to a 10 m s–1 wind, and are plotted only where either the u or υ component is significant at the 95% level or greater
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
As in Fig. 2b except that OLR is represented by hatching (cross hatching negative, single hatching positive), and shading represents 400-hPa vertical motion (dark positive; shading at 0.2 cm s–1)
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
As in Fig. 2b except that OLR is represented by hatching (cross hatching negative, single hatching positive), and shading represents 400-hPa vertical motion (dark positive; shading at 0.2 cm s–1)
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
As in Fig. 2b except that OLR is represented by hatching (cross hatching negative, single hatching positive), and shading represents 400-hPa vertical motion (dark positive; shading at 0.2 cm s–1)
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Regressed OLR (shading, dark negative, at ±5 and 9 W m–2) and 200-hPa streamfunction (contours, interval 15 m2 s–1) and winds (vectors, longest vector 20 m s–1) on day +2, based on a +20 m s–1 anomaly in Kelvin-filtered 200-hPa meridional wind at the basepoint 27.5°S, 127.5°E, on day 0
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Regressed OLR (shading, dark negative, at ±5 and 9 W m–2) and 200-hPa streamfunction (contours, interval 15 m2 s–1) and winds (vectors, longest vector 20 m s–1) on day +2, based on a +20 m s–1 anomaly in Kelvin-filtered 200-hPa meridional wind at the basepoint 27.5°S, 127.5°E, on day 0
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Regressed OLR (shading, dark negative, at ±5 and 9 W m–2) and 200-hPa streamfunction (contours, interval 15 m2 s–1) and winds (vectors, longest vector 20 m s–1) on day +2, based on a +20 m s–1 anomaly in Kelvin-filtered 200-hPa meridional wind at the basepoint 27.5°S, 127.5°E, on day 0
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
As in Fig. 2b except at 850 hPa. Streamfunction contour interval is 4.0 × 105 m2 s–1, and longest vectors correspond to a 5 m s–1 wind
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
As in Fig. 2b except at 850 hPa. Streamfunction contour interval is 4.0 × 105 m2 s–1, and longest vectors correspond to a 5 m s–1 wind
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
As in Fig. 2b except at 850 hPa. Streamfunction contour interval is 4.0 × 105 m2 s–1, and longest vectors correspond to a 5 m s–1 wind
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
As in Fig. 3 except at 1000 hPa for (a) geopotential height and (b) temperature on day −5. Geopotential height contour interval is 5 m from 5 to 20 m, then 40 m; temperature contour interval is 0.05 K from 0.05 to 0.1 K, then 0.2 K. Longest vectors correspond to a 5 m s–1 wind
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
As in Fig. 3 except at 1000 hPa for (a) geopotential height and (b) temperature on day −5. Geopotential height contour interval is 5 m from 5 to 20 m, then 40 m; temperature contour interval is 0.05 K from 0.05 to 0.1 K, then 0.2 K. Longest vectors correspond to a 5 m s–1 wind
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
As in Fig. 3 except at 1000 hPa for (a) geopotential height and (b) temperature on day −5. Geopotential height contour interval is 5 m from 5 to 20 m, then 40 m; temperature contour interval is 0.05 K from 0.05 to 0.1 K, then 0.2 K. Longest vectors correspond to a 5 m s–1 wind
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Longitude–time diagram of regressed OLR (shading) and 1000-hPa convergence (contours), averaged from 5°S–10°N, from day −11 to day 8. OLR is shaded at intervals of ±5 W m–2, with dark shading representing negative anomalies. Contour interval for 1000-hPa convergence is 2.5 × 10–7 s–1, with solid contours representing convergence. The zero contour has been omitted
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Longitude–time diagram of regressed OLR (shading) and 1000-hPa convergence (contours), averaged from 5°S–10°N, from day −11 to day 8. OLR is shaded at intervals of ±5 W m–2, with dark shading representing negative anomalies. Contour interval for 1000-hPa convergence is 2.5 × 10–7 s–1, with solid contours representing convergence. The zero contour has been omitted
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Longitude–time diagram of regressed OLR (shading) and 1000-hPa convergence (contours), averaged from 5°S–10°N, from day −11 to day 8. OLR is shaded at intervals of ±5 W m–2, with dark shading representing negative anomalies. Contour interval for 1000-hPa convergence is 2.5 × 10–7 s–1, with solid contours representing convergence. The zero contour has been omitted
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Longitude–height cross sections of regressed (a) meridional wind (contours) and vertical motion (shading) along 20°S (lower plot), and OLR along equator (upper plot) on day −5; (b) temperature (contours), vertical motion (shading), and OLR along equator on day −5; and (c) temperature (contours), vertical motion (shading), and OLR along 7.5°N on day 0. Contour interval in (a) is 0.5 m s–1, and in (b) and (c) is 0.1 K, and all zero contours have been omitted. Regressed vertical motion is shaded at ±0.1 and ±0.3 cm s–1 in (a) and (b), and at +0.1, +0.6, −0.1, and −0.3 cm s–1 in (c). Zonal–vertical circulation is shown by vectors, where the vertical component has been multiplied by 700 to account for the small aspect ratio of the plot. The longest vectors represent winds of (a) 5.0. (b) 3.5, and (c) 9.0 m s–1
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Longitude–height cross sections of regressed (a) meridional wind (contours) and vertical motion (shading) along 20°S (lower plot), and OLR along equator (upper plot) on day −5; (b) temperature (contours), vertical motion (shading), and OLR along equator on day −5; and (c) temperature (contours), vertical motion (shading), and OLR along 7.5°N on day 0. Contour interval in (a) is 0.5 m s–1, and in (b) and (c) is 0.1 K, and all zero contours have been omitted. Regressed vertical motion is shaded at ±0.1 and ±0.3 cm s–1 in (a) and (b), and at +0.1, +0.6, −0.1, and −0.3 cm s–1 in (c). Zonal–vertical circulation is shown by vectors, where the vertical component has been multiplied by 700 to account for the small aspect ratio of the plot. The longest vectors represent winds of (a) 5.0. (b) 3.5, and (c) 9.0 m s–1
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Longitude–height cross sections of regressed (a) meridional wind (contours) and vertical motion (shading) along 20°S (lower plot), and OLR along equator (upper plot) on day −5; (b) temperature (contours), vertical motion (shading), and OLR along equator on day −5; and (c) temperature (contours), vertical motion (shading), and OLR along 7.5°N on day 0. Contour interval in (a) is 0.5 m s–1, and in (b) and (c) is 0.1 K, and all zero contours have been omitted. Regressed vertical motion is shaded at ±0.1 and ±0.3 cm s–1 in (a) and (b), and at +0.1, +0.6, −0.1, and −0.3 cm s–1 in (c). Zonal–vertical circulation is shown by vectors, where the vertical component has been multiplied by 700 to account for the small aspect ratio of the plot. The longest vectors represent winds of (a) 5.0. (b) 3.5, and (c) 9.0 m s–1
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Time series of monthly averaged Kelvin wave variance (in W2 m–4) at 7.5°N, 172.5°W, from Jan 1979 to Dec 2000
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Time series of monthly averaged Kelvin wave variance (in W2 m–4) at 7.5°N, 172.5°W, from Jan 1979 to Dec 2000
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Time series of monthly averaged Kelvin wave variance (in W2 m–4) at 7.5°N, 172.5°W, from Jan 1979 to Dec 2000
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Monthly averaged Kelvin wave–filtered OLR variance (contours), 200-hPa Kelvin-filtered meridional wind variance anomalies (shading), and total Kelvin-filtered 200-hPa E-vectors for (a) high and (b) low variance composites, based on the 11 months during the JJA season with highest or lowest monthly averaged Kelvin-filtered OLR variance in a 5° × 5° box centered at 7.5°N, 172.5°W. Kelvin OLR variance is contoured from 25 W2 m–4 by 50 W2 m–4. Meridional wind variance is shaded at ±3 and 10 m2 s–2; dark shading represents positive values. The longest E-vectors are 40 m2 s–2; vectors are not plotted below 2 m2 s–2
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Monthly averaged Kelvin wave–filtered OLR variance (contours), 200-hPa Kelvin-filtered meridional wind variance anomalies (shading), and total Kelvin-filtered 200-hPa E-vectors for (a) high and (b) low variance composites, based on the 11 months during the JJA season with highest or lowest monthly averaged Kelvin-filtered OLR variance in a 5° × 5° box centered at 7.5°N, 172.5°W. Kelvin OLR variance is contoured from 25 W2 m–4 by 50 W2 m–4. Meridional wind variance is shaded at ±3 and 10 m2 s–2; dark shading represents positive values. The longest E-vectors are 40 m2 s–2; vectors are not plotted below 2 m2 s–2
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Monthly averaged Kelvin wave–filtered OLR variance (contours), 200-hPa Kelvin-filtered meridional wind variance anomalies (shading), and total Kelvin-filtered 200-hPa E-vectors for (a) high and (b) low variance composites, based on the 11 months during the JJA season with highest or lowest monthly averaged Kelvin-filtered OLR variance in a 5° × 5° box centered at 7.5°N, 172.5°W. Kelvin OLR variance is contoured from 25 W2 m–4 by 50 W2 m–4. Meridional wind variance is shaded at ±3 and 10 m2 s–2; dark shading represents positive values. The longest E-vectors are 40 m2 s–2; vectors are not plotted below 2 m2 s–2
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Monthly averaged total SST (hatching and light contours), Kelvin wave–filtered OLR variance (dark contours), Kelvin wave–filtered 200-hPa meridional wind variance anomalies (shading), and Kelvin wave–filtered 200-hPa total E-vectors, for (a) Jul 1986 and (b) Jul 1987. SST is contoured at 28°, 29°, and 29.5°C, with single hatching at 29°C and cross hatching at 29.5°C. Kelvin OLR variance is contoured from 150 W2 m–4 by 100 W2 m–4. Meridional wind variance anomalies are shaded at ±5 m2 s–2; dark shading denotes positive anomalies. Longest E-vectors are 150 m2 s–2, and are not plotted below 5 m2 s–2
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Monthly averaged total SST (hatching and light contours), Kelvin wave–filtered OLR variance (dark contours), Kelvin wave–filtered 200-hPa meridional wind variance anomalies (shading), and Kelvin wave–filtered 200-hPa total E-vectors, for (a) Jul 1986 and (b) Jul 1987. SST is contoured at 28°, 29°, and 29.5°C, with single hatching at 29°C and cross hatching at 29.5°C. Kelvin OLR variance is contoured from 150 W2 m–4 by 100 W2 m–4. Meridional wind variance anomalies are shaded at ±5 m2 s–2; dark shading denotes positive anomalies. Longest E-vectors are 150 m2 s–2, and are not plotted below 5 m2 s–2
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
Monthly averaged total SST (hatching and light contours), Kelvin wave–filtered OLR variance (dark contours), Kelvin wave–filtered 200-hPa meridional wind variance anomalies (shading), and Kelvin wave–filtered 200-hPa total E-vectors, for (a) Jul 1986 and (b) Jul 1987. SST is contoured at 28°, 29°, and 29.5°C, with single hatching at 29°C and cross hatching at 29.5°C. Kelvin OLR variance is contoured from 150 W2 m–4 by 100 W2 m–4. Meridional wind variance anomalies are shaded at ±5 m2 s–2; dark shading denotes positive anomalies. Longest E-vectors are 150 m2 s–2, and are not plotted below 5 m2 s–2
Citation: Journal of the Atmospheric Sciences 60, 3; 10.1175/1520-0469(2003)060<0526:EFOCCK>2.0.CO;2
The group velocity vector subtends twice the angle from the zonal direction that the E-vector does (Hoskins et al. 1983).
