a. 200 mb

In this section we examine the horizontal structure of the synoptic-scale circulation at 200 and 850 mb associated with ITCZ convection during northern winter. Although these circulations are isolated using OLR as a predictor, it will be shown that these modes are quite typical of the type of equatorward-propagating wave activity discussed, for example, by Hsu and Lin (1992) and Tomas and Webster (1994) using circulation parameters as a predictor. The presentation here is based on regression relationships obtained using 6–30-day filtered OLR averaged over the region 5°–15°N, 150°–140°W. As will be shown below, this is near the longitude of maximum upper-level wave activity in the eastern tropical Pacific during DJF, and the results are representative of the entire sector between about 170° to 120°W.

Figure 1 shows the sequence of 200-mb streamfunction and wind anomalies scaled to a −40 W m⁻² OLR deviation averaged over the base region, which is located within the shaded region on day 0 in Fig. 1c. In these plots, wind vectors are shown only at those points where the correlation coefficients imply statistical significance at better than the 5% level. The theoretical group propagation of a stationary Rossby wave on a sphere in an atmosphere with constant angular velocity is along a great circle (e.g., Karoly and Hoskins 1982). For later reference, a portion of a great circle route is shown in Fig. 1, having a northern turning point at 30°N, 160°E and crossing the equator at 110°W. Also shown is a meridionally oriented line along 145°W, for use in vertical cross sections below.

Four days prior to the peak in ITCZ convection (day −4, Fig. 1a), statistically significant circulations are confined to the western North Pacific and eastern Asia, with an anticyclonic anomaly over southern China tied to enhanced westerlies near the mean position of the Asian jet entrance region. Near the date line there is a trough south of and a ridge north of the Asian jet axis,associated with anomalously weak jet flow. Opposite anomalies over the eastern North Pacific complete a quadrupole, with the northern branch leading into a weak wave-train pattern over North America and into the North Atlantic. Evidence presented by Meehl et al. (1996) shows that this pattern, and subsequent developments in Fig. 1, are often preceded by equatorial convection in the eastern Indian Ocean at about day −6.

Two days later (Fig. 1b), the jet streak over China has advanced into the jet core, and the trough to its east has shifted eastward and developed a northeast–southwest tilt consistent with the shear deformation zone of positive ∂u/∂y along the southern margin of the jet. The ridge to its southeast has also developed a stronger positive tilt, while the northern wave train moves more slowly eastward.

By day 0 (Fig. 1c) the downstream portions of both wave trains have strengthened significantly, while the OLR signal peaks in the ITCZ, ahead of the upper trough located near Hawaii. Group propagation toward the southeast is evident, with clear dispersion apparent as downstream features to the south of the equator are strengthening successively while those upstream weaken. Eastward dispersion along the wave train over North America is also evident. By day +2 (Fig. 1d) the OLR signal has moved eastward and northward, and a substantial circulation has developed just south of the equator near 100°W. Between day −2 and day +4 (Fig. 1e), it is also evident that the pattern has the opposite phase, due to the propagation of features with respect to the earth’s surface. This yields a mean period of around 12 days for the disturbance, in agreement with the period derived by Tomas and Webster (1994) and with the spectral results referred to in the previous section.

It is of interest to compare the observed phase and group velocity of the wave train with those for a nondivergent barotropic Rossby wave on a beta plane. Figure 2 is a time–longitude diagram of the streamfunction perturbations along the great circle line in Fig. 1 from day −12 to day +12, within the longitude range 160°E–80°W. This line lies close to the path of the apparent phase and group propagation of the features in Fig. 1, although deviations from this path are observed and tobe expected (see below). A phase speed line of 5 m s⁻¹ is shown, which approximates the zonal phase speed of individual troughs and ridges within the wave train around day 0, although disturbances within the jet on the left side of the diagram have much faster speeds. Also shown is a 24 m s⁻¹ line that gives an estimate of the rate of zonal energy dispersion within the wave train, as shown by the successive development of perturbations farther east over time. This estimate is seen to be most valid for the wave train around day 0, and the next (opposite) phase of the disturbance from days 5 through 10, when the wave energy is crossing the equator.

The approximate spatial scale of the waves can also be estimated from Figs. 1 and 2. From these pictures the wave along the great circle has a zonal wavelength of around 50° longitude, giving a zonal planetary wavenumber of approximately k = 7. The meridional wavenumber l at the equator can then be estimated by noting that the angle α that the great circle makes with the longitude–latitude (x–y) coordinate system is

View Expanded

Since α = −30° at the equator for the great circle shown in Fig. 1, this gives an estimate of around l = −4 for the meridional planetary wavenumber at this latitude.The negative wavenumber is indicative of the positive tilt of the waves in this region.

We will compare the theoretical phase and group velocity with respect to the earth’s surface to that of a nonstationary, nondivergent barotropic Rossby wave (e.g., Yang and Hoskins 1996). The dispersion relationship for such a Rossby wave with respect to the earth’s surface is given by

View Expanded

with the zonal phase speed c of

View Expanded

and group velocities u_g and υ_g in the zonal and meridional directions, respectively, given by

View Expanded

where ν is the frequency and u is the value of the background zonal wind speed. Setting u to zero for the moment, and using the values of β, k, and l at the equator in order to estimate the intrinsic phase and group velocity of such a wave yields c = −14.3 m s⁻¹, u_g = 7.3 m s⁻¹, and υ_g = −12.3 m s⁻¹. However, the waves are embedded in a strong background basic state that Doppler shifts these phase and group speeds. Referring to maps of the mean DJF 200-mb flow over the region (e.g., Fig. 1a of Kiladis and Weickmann 1997), we see that along the tropical portion of the ray path in Fig. 1 the zonal flow is nearly constant at about 20 m s⁻¹. Taking u = 20 m s⁻¹ then gives ν = 6.3 × 10⁻⁶ s⁻¹, corresponding to a period of 11.6 days, which agrees well with the mean period of the waves as described above. The zonal phase and group velocity with respect to the ground is then c = 5.7 m s⁻¹ and u_g = 27.3 m s⁻¹, values that match the observed speeds in Fig. 2 reasonably well, considering the gross assumptions made.

If the group velocity were indeed along the great circle route, the observed meridional group speed at the equator would simply be υ_g = u_g tanα = −13.8 m s⁻¹, a value that is again not far from that predicted. However, in reality, the ray path for a nonstationary barotropic Rossby wave will deviate from a great circle due to changes in both zonal and meridional advection and refraction by the zonal and meridional gradients in the flow (e.g., Karoly 1983). The meridional flow in the domain of the wave train is certainly not negligible, and moreover, as we shall show, the disturbances are associated with strong divergence in the region of the ITCZ. Nevertheless, initial value simulations in a barotropic model with a realistic basic state, to be presented elsewhere, do confirm that the observed upper-level waves indeed behave similarly to that predicted by simple Rossby wave theory.

b. 850 mb

Figure 3 shows the evolution of the 850-mb circulation obtained from the identical OLR predictor used to produce Fig. 1. As at 200 mb, the sequence suggests a substantial extratropical influence on ITCZ convection, although the signal is not one of wave energy propagation into the Tropics.

At day −4 in Fig. 3a a large anticyclone is present at 850 mb, which lies beneath the 200-mb anticyclone and subtropical trough at the date line in Fig. 1a. This anticyclone already extends far into the Tropics and is linked to anomalous equatorial easterly flow just to the east of the date line. The collocation of the subtropical portion of this feature with an upper trough is indicative of a negative temperature perturbation in the midtroposphere, discussed in more detail below. To the west is a cyclone over Japan, which shows a pronounced poleward tilt with height. Over North America and the Atlantic, 850-mb signals are for the most part in phase with those at upper levels, although some westward tilt with height is evident, particularly in the feature located near the east coast of North America.

An amplification of the 850-mb anticyclones over the mid-Pacific and North America is seen by day −2, resulting in an increase of the trade wind flow over a broad region of the east Pacific (Fig. 3b). The cyclone off Japan in Fig. 3b moves rapidly northeastward in the Pacific storm track by day 0 (Fig. 3c), at the same time that the Pacific anticyclone surges equatorward and further strengthens the trade wind flow into the region of developing convection. To the west of the convection, a low-level cyclone has formed, with equatorial westerly anomalies coupled to a weak cyclonic feature to the south of the equator. This southern cyclone strengthens significantly by day +2 (Fig. 3d) and propagates westward along with the equatorial westerlies over the next couple of days (Fig. 3e).

Throughout the sequence of Figs. 1 and 3 the evolution of the extratropical features are characteristic of baroclinic wave development, where westward and poleward tilts with height are evident near Japan and the southeastern coast of North America, becoming more equivalent barotropic as the disturbances move northeastward into the storm tracks over the Pacific and Atlantic. The low-latitude vertical structure is more complicated, with like-signed wind perturbations in some regions, such as along the equator near 160°W on day 0, as opposed to very weak perturbations at low levels beneath strong signals at 200 mb over the east Pacific.

The pattern of westward-propagating paired cyclonic anomalies along the equator at 850 mb in Fig. 3 is characteristic of the n = 1 equatorial Rossby waves isolated in the 6–30-day band by Kiladis and Wheeler (1995). The equivalent barotropic vertical structure, with like-signed wind anomalies through a deep layer of the troposphere near the equator and convection in the poleward flow to the east, is also consistent with our observations of those modes. One interpretation of this analysis is that low-level equatorial Rossby waves may be initiated by upper-level wave activity from the extratropics, as has been suggested for the generation of mixed–Rossby gravity waves (e.g., Zhang and Webster 1992; Magaña and Yanai 1995). However, the role of ITCZ convection as an energy source is still an open question, and resolution of these issues will likely require additional theoretical and modelling work, as discussed further in section 6.

c. Mass circulation

The OLR signal in Figs. 1 and 3 is accompanied by a substantial mass circulation at 200 mb, as shown in Fig. 4. In Fig. 4a a region of perturbation divergence is almost perfectly collocated with the negative OLR signal, with a wave train of divergence anomalies in quadrature with the streamfunction anomalies of Fig. 1. This relationship gives us confidence that at least the qualitative pattern, if not the exact magnitude, of the divergence field is well-captured by the NCEP reanalysis data, since OLR is virtually independent of the analysis. The divergent wind (Fig. 4b) shows rather large- scale upper-level outflow away from the convective region, which is stronger toward the winter hemisphere but also crosses the equator into the eastern South Pacific convergence zone (SPCZ) region. A sharply defined line of convergence is found at the junction between the tropical outflow and divergent outflow from the storm track cyclone farther northwest. The strength of the tropical divergent circulation is substantially larger than that observed for the mean circulation in this region (e.g., Mo and Rasmusson 1993); thus these transient events are likely important for the vorticity balance of this region.

At 850 mb, convergence into the convective region yields a pattern qualitatively opposite to that observed in Figs. 4a and 4b (not shown). This implies upward motion in the negative OLR region, which must then peak somewhere in the midtroposphere, a structure which is verified in section 4.

d. Isentropic potential vorticity

In the case study examined by Kiladis and Weickmann (1992b), an upper-level trough at low latitudes was shown to be accompanied by a strong signal of equatorward propagating high isentropic potential vorticity (IPV) air on the 350-K potential temperature (θ)surface (see Hoskins et al. 1985). A negative OLR anomaly in the ITCZ was located ahead of this IPV anomaly, in a region of strong positive PV advection, where upward motion should be maximized.

Figure 4c maps the regressed horizontal distribution of IPV on the 350-K surface at day 0 corresponding to the streamfunction pattern in Fig. 1c. This θ surface lies close to the 200-mb pressure level at all latitudes. The IPV distribution was computed by adding the regressed wind and temperature fields using the OLR predictor to those corresponding to the DJF mean at the appropriate pressure levels. These values of PV were then interpolated to the 350-K surface and mapped in “PV units” (=1 × 10⁶ m² s⁻¹ K kg⁻¹).

A positively tilted “trough” of high PV air is present immediately northwest of the OLR anomaly in Fig. 4c. Lagged maps (not shown) depict this signal originating as a small ripple propagating eastward within the strong PV gradient along the Asian jet a few days prior to day 0, followed by a rapid amplification and equatorward propagation from the jet exit region into the Tropics. This evolution matches that of the case studied by Kiladis and Weickmann (1992b) quite closely, although in individual events the PV ridge will typically be muchmore narrow and sharper than the feature in Fig. 4c, due to the “smearing” from the noise inherent in the regressions. It is also notable that the negative streamfunction perturbation near the equator, 100°W in Fig. 1c, is associated with another trough of positive PV crossing into the Southern Hemisphere in Fig. 4c.

In lagged maps of IPV perturbations (not shown) the positive IPV anomaly present to the northwest of the OLR anomaly weakens considerably by day +2 as it moves eastward. This weakening can occur through the advection and consequent flux divergence (or “rearrangement”) of PV along θ surfaces by barotropic processes, or by diabatic effects that transport mass across θ surfaces and effectively “dilute” the PV (Haynes and McIntyre 1987). The former is unlikely to be the dominant process here, since the PV appears not so much“spread out” over time but disappears very rapidly more or less in situ. Most likely the diabatic heating associated with the convection is primarily responsible for this evolution; however, quantification of this effect through a PV budget analyses, and modeling, will be necessary to confirm this.

a. Potential vorticity

Tomas and Webster (1994) examined the vertical structure of equatorward propagating disturbances in theeastern Pacific during northern winter by correlating PV at 200 mb to that at standard pressure levels from 1000 to 100 mb in the vertical. Their analysis showed that the waves propagated freely across the equator in the upper-level westerlies, but at lower levels only weak signals penetrated into the trade winds. Thus, the vertical extent of the waves was seen to become more shallow as lower latitudes were approached. It was speculated that the zero zonal wind line acted as a critical line in the vertical, such that the wave activity was precluded from propagating from the upper westerlies into the low-level easterlies.

Here we take a somewhat different approach by using OLR as the independent variable and examining the actual regressed values of PV on pressure surfaces. Vertical cross sections of the evolution of PV on the 14 standard pressure levels were produced in the zonal and meridional directions, as well as along the great circle route shown in Fig. 1. Since the regression analysis allows the relative amplitude of a given parameter at each level to be preserved, a detailed view of the typical vertical distribution of PV associated with ITCZ convection can be depicted.

Figure 5a shows the day 0 distribution of PV perturbations on pressure surfaces in the zonal direction along 10°N. Since lower tropospheric PV is so small near the equator, the contour interval has been reducedto 0.02 PV units for values less than 0.14 in this figure, and 0.10 otherwise. The largest PV anomalies at this latitude are in the lower stratosphere, with the location of the OLR signal, at 145°W, in between the largest tropospheric PV anomalies. There is a pronounced westward tilt with height of the stratospheric signals, and weaker but opposite tilts below about 100 mb.

The total PV on day 0 and day +2 in Figs. 5b and 5c is obtained by adding the DJF mean state for the study period to the perturbations. The waves are represented by vertical undulations in the pressure levelof PV, which progress eastward over time. In Fig. 5b two marked downward excursions of the contours represent the anomalously high PV in the mid and lower troposphere to the west and east of the convection, as was also seen in Fig. 4c. Also notable is the 0.2 contour of PV below 700 mb centered at the date line, which represents the cyclonic circulation at 850 mb in Fig. 3c. While this lower signal is initially vertically aligned with the midtropospheric anomaly in previous lags (not shown), over time it drifts westward and becomes detached from the eastward-moving upper anomaly.

Figure 6 shows the evolution, from day −2 to day +2, of PV anomalies on pressure surfaces in the meridional plane along 145°W (marked by the vertical line in Figs. 1 and 3), which passes through the region of maximum OLR perturbation. These sections are similar to those along the great circle (not shown). Pronounced poleward tilts of the waves are seen in the stratosphere, and as in the zonal direction there are weaker but opposite tilts below 100 mb. Another feature evident in Fig. 6, besides the equatorward progression of the anomalies, is their apparent vertical propagation into thestratosphere, consistent with their poleward and westward tilts with height above the 100-mb level.

One way to understand the apparent upward propagation of the disturbances in a qualitative sense is by considering the vertical orientation of the PV field in the meridional direction. Figure 7a shows the climatological distribution of total PV along the meridional cross section. The PV contours in these plots (as in the zonal mean) are quasi-horizontal poleward of 30°N and increasingly sloped to the south, to become nearly vertical at the equator. In the midlatitudes the tropopause generally lies around the 1.5-PV unit level. Clearly at low latitudes, however, this contour must cut across the tropopause, which is at around 100 mb at the equator. The strong, vertically sloping PV gradient can act as a wave guide to sufficiently large perturbations along it (Hoskins et al. 1985), leading to the upward propagation of wave activity. The effect of the day 0 PV perturbation on the total field is shown in Fig. 7b, where the downward excursion of contours is evident at 20°N. The“bulges” associated with the wave activity progress equatorward and upward along the strong gradient over time (not shown), with the waves subtly altering the vertical orientation of the PV distribution near the equator, as can be seen by comparing Figs. 7a and 7b.

b. Temperature

A deep layer of cold air advection accompanies the upper PV anomaly and associated trough, which was also a prominent feature of the case study in Kiladis and Weickmann (1992b). A meridional section (Fig. 7c) shows the largest negative temperature perturbation exceeding 2 K; it peaks at 250 mb poleward of 30°N and extends downward into the subtropics and Tropics, crossing the equator. There is a warm anomaly at the surface and in the stratosphere in the Tropics. The net effect of this temperature distribution is to make the lower troposphere less stable at low latitudes, as shown by the change in ∂T/∂p, which has a local maximum near the 700-mb level at about 15°N (not shown).

c. Vertical motion

The combination of dynamical forcing by positive PV advection at upper levels and decreased stability at low levels due to temperature advection is to induce upward motion along the ITCZ. Figure 8a shows the mean DJF 1979/80–1994/95 vertical distribution of ω (=dp/dt) along the 145°W reference line. Also shown is a representation of the mass circulation in the meridional plane, obtained by combining the divergent meridional wind and ω, and scaling the arrows such that a 1 m s⁻¹ meridional divergent wind is equal to a 5 × 10⁻² Pa s⁻¹ vertical motion. The ITCZ is well defined in the reanalysis data, with mean upward motion between 2.5°N and 10°N, peaking at 7.5°N. Mean tropospheric subsidence at the equator is in agreement with profiler observations at Christmas Island, 12° of longitude farther west (Gage et al. 1991). The SPCZ is also seen as a broad region of negative ω to the south of the equator, as is the storm track poleward of 40°N. Most of the subtropical subsidence in the Northern Hemisphere subtropics is confined to the lowest levels at this longitude, as opposed to zonal mean conditions (not shown), where it is maximized at around 400 mb.

Figure 8b shows the sum of the mean state from Fig. 8a and the perturbation ω field, scaled once again to a−40 W m⁻² OLR anomaly. Upward motion is greatly enhanced from the ITCZ northward to beyond 20°N. The wave forcing concentrates a vertical velocity maximum at about the 300-mb level at 20°N, with a secondary maximum at the latitude of the ITCZ itself. Interestingly, there is still downward motion below 700 mb centered on 15°N. There is also little change in the mass circulation away from the northern subtropics, although the perturbation field in Fig. 8c shows that aweak region of subsidence completes an anomalous local Hadley circulation extending out to about 40°N. The magnitude of this horizontal divergent circulation (see also Fig. 4b) is certainly large enough to produce a substantial vorticity tendency at upper levels of the subtropics. The quantification of this tendency and its effect on the evolution of the subtropical westerlies and associated cloud band will be explored in a future study.

d. Momentum and heat transports

A view of the of the meridional transport of westerly momentum in the region of the wave train is shown in Fig. 9a. This was obtained by calculating the daily product u′υ′ at each grid point and pressure level from day −12 to day +12 from the regressed fields. These products were then averaged over the day −12 to day +12 cycle and also over the longitude range 160°E to 100°W, which encompasses the domain of the wave train in the zonal direction. This 25-day averaging procedure covers two “cycles” of the mean period of the wave and is meant to give a qualitative picture of the vertical distribution of poleward momentum flux associated with the disturbances. Since this calculation includes only products between separately filtered and regressed 6–30-day u′ and υ′, and not cross products that would beassociated with interactions between this band and other frequencies, it comprises only a portion of the impact of these transients on the momentum budget, although calculations of these other quantities (not shown) show negligible contributions.

As expected from the positive tilt of the wave train in Fig. 1, there is a poleward flux of westerly momentum extending from just south of the equator to 50°N associated with the cycle, which peaks at around the 200- mb level. The primary maximum of poleward flux at 20°N suggests an acceleration of the westerlies at around 30°N, and a flux divergence and deceleration of the westerlies in the deep Tropics. Thus the propagation of this type of wave activity into the westerly duct acts as a brake on the equatorial westerlies themselves.

The calculation of υ′T′ by the same method (Fig. 9b) shows two peaks in the poleward flux of heat at 40°N in the lower and upper troposphere, with positive values extending to just south of the equator below 850 mb and in the stratosphere. Despite the cold advection present near the convective region on day 0, the net effect of the wave activity results in equatorward (poleward) movement of warm (cold) air in the mid- to upper levels of the subtropics. This countergradient heat flux is typical of the climatological transient eddy statistics at this time of year (e.g., Newell et al. 1974) and is also consistent with the vertical tilts of the eddies at low latitudes in Fig. 5 (e.g., Hoskins et al. 1983), where upward (downward) group propagation is associated with a poleward (equatorward) heat flux in the stratosphere (troposphere).

a. Mean conditions

The term υ′² − u′² is a measure of the mean anisotropy or shape of the waves. For example, if υ′² is consistently larger than u′², the waves are preferentially elongated in the meridional direction and the E vector points eastward, as is the case for high-frequency (<10 day) intraseasonal fluctuations at the Asian jet exit region (Hoskins et al. 1983; Wallace and Lau 1985). They-component −u′υ′ is minus the time–mean northward flux of westerly momentum associated with the perturbations. Together the two components approximate the preferred direction of the group velocity of the waves using suitable approximations, including the assumption of quasigeostrophy.

Kiladis and Feldstein (1994) showed that the east Pacific Rossby wave activity propagating into the Tropics dominated the less than 14-day E vector field in that region for DJF 1979/80–1989/90 NCEP operational analyses at 200 mb. Here we perform a similar calculation for 6–30-day transients using the 1979/80 through 1994/95 NCEP reanalysis data.

Another useful diagnostic for representing the mean background state in which the transients are embedded involves the calculation of the stationary Rossby wavenumber:

View Expanded

where

View Expanded

is the meridional gradient of absolute vorticity associated with the basic state. Here, K_s is the total wavenumber at which a barotropic Rossby wave is stationary at a particular location in a given background zonal flow. Karoly and Hoskins (1982) and Hoskins and Ambrizzi (1993) show that the distribution of K_s can be used to infer the location of critical lines and wave guides for stationary Rossby waves, again subject to some approximations. Although the waves studied here are clearly not stationary, Yang and Hoskins (1996) show that a modified K for transient waves can still give useful information on the dependence of nonstationary eddy activity on the background state. Here we will use K_s to give a qualitative picture of the effects of the basic- state flow within the westerly duct on the equatorward wave propagation.

Figure 10a shows the 6–30-day mean 200-mb DJF E vector field over the Pacific domain for 1979/80–1994/95, along with a calculation of K_s for the same period. The imprint of the equatorward-propagating wave activity is evident as a large region of southeastward- pointing E vectors that extends from the Asian jet exit region near the dateline into central North America. These eddies reach farthest into the Tropics from about the dateline to around 120°W, matching the longitude range of maximum correlation between OLR and upper level wave activity (Kiladis and Weickmann 1992b, 1997). A small region of zonally oriented eddies in the Asian jet core gives way to meridionally elongated signatures to the east and south. The transients again become more zonally elongated at about 10°N, where the E vectors begin to point southwestward. There is a notable lack of E vector amplitude near the critical lineregion of the tropical west Pacific, as shown by the thick contour where K_s approaches infinity (U = 0). Consistent with refractive index considerations, the E vectors tend to point across the gradient and toward regions of higher K_s, such as toward the ITCZ south of Hawaii, and avoid regions of local K_s minima, as near 10°N in the easternmost Pacific and south of the Asian jet.

Hoskins et al. (1983) show that a convergence of E vectors implies that the transient eddies are acting to decelerate the background zonal wind. This occurs along the northern edge of the central Pacific ITCZ and eastward into southern North America in Fig. 10a. Kiladis and Feldstein (1994) calculated the DJF zonal momentum budget over the eastern tropical Pacific at 200 mb from NCEP operational analyses and found that the transient eddy terms were indeed crucial to the budget and acted to decelerate the westerlies over that region. This budget was recalculated using 1979–95 reanalysis data, and again a good balance was obtained, with the residuals in most locations less than 20% of any one term in the budget (not shown). Filtering reveals that most of this transient eddy deceleration of the zonal wind occurs on timescales of less than 30 days.

b. E vector signal of wave activity associated with ITCZ convection

The 200-mb E vector signature of the 6–30-day wave activity is mapped in Fig. 11, calculated in the same way as the covariance terms in Fig. 9 over the 25-day cycle of lagged regressions. The mean E vectors of the wave activity have a signature similar to the time mean DJF picture in Fig. 10a, with zonally elongated eddies in the Asian jet, meridionally elongated and southeastward-propagating transients to the south of the jet exit region, and E vectors turning southwestward at around 10°N. The poleward flux of westerly momentum, which is maximized at this level in Fig. 9b, and the convergence of E to the north of the ITCZ, are also reflected in the mean picture. One notable difference is that the transients we have isolated as being associated with convection at 145°W remain zonally elongated past the dateline, as can be seen in Fig. 1. Of course, these disturbances contribute to only a portion of the climatological transient eddy effects, although it appears that they are the dominant features leading to the climatological E-vector picture over the study region in Fig. 10a.

c. Interannual variability of wave activity

An example of the interannual variability in the east Pacific wave activity can be obtained by referring to Figs. 10b and 10c, which show the 6–30-day 200-mb E vectors and K_s for two individual DJF seasons, 1982/83 and 1988/89. The time–mean zonal wind contributing to Fig. 10b differs substantially from the mean picture in Fig. 10a due to the extreme warm event conditions associated with unusually high sea surface temperature (SST) and eastward and equatorward expansion of convection over the eastern equatorial Pacific during that period (e.g., Gill and Rasmusson 1983). The eastward shift of the critical line compared to climatology is evident during that season, which is directly related to the establishment of upper-level easterlies and the breakdown of the westerly duct as a response to the anomalous convection. The E-vector signal shows a marked reduction in its normal amplitude to the south of 20°N and a northward shift in the latitude where the vectors turn southeastward and the eddies become zonally elongated. This reduction in low-latitude Rossby wave propagation is in agreement with the observation of McGuirk et al. (1987) and Iskenderian (1995) that, despite the extremely active convection along the equator during that season, the 1982/83 winter had little cloud band activity. Thus the equatorial ITCZ convection during warm events behaves more like the usual warm pool convection over the west Pacific, which is forced by high SST rather than lateral forcing from waves originating in the extratropics.

As a contrast, Fig. 10c shows the same calculations for the DJF 1988/89 cold event season. In this case SST was much below normal over a large portion of the tropical east Pacific, and the upper-level westerlies were stronger than normal in the westerly duct. As a result the critical line is west of its mean position and the wave activity is also shifted south and west, with higher amplitude when compared to climatology. A large region of low OLR was observed around 15°–20°N, 140°–120°W, well north of the mean ITCZ location (Arkin 1989), despite the fact that the SST was below normal throughout this region during the period. This is where there is strong anomalous convergence of E vectors in Fig. 10c, suggesting that increased wave activity into the westerly duct was responsible for forcing this northward shift in the ITCZ along with enhanced cloud band activity.