a. ISO convection and circulation anomalies

The regressed OLR (shading) and 850-hPa circulation anomalies (streamfunction contours, vector winds) during one cycle of the boreal summer ISO are shown in Fig. 1, for days −15, −5, +5, and +15. On day −15 (Fig. 1a), low OLR, representing enhanced convection, extends over a large portion of the equatorial Indian Ocean, while a band of suppressed convection stretches southeastward from India toward the date line in the Northern Hemisphere. The band of suppressed convection had moved northward and eastward prior to day −15 as the new ISO active convective phase intensified in the Indian Ocean. Associated with the suppressed convective phase on day −15 is an elongated 850-hPa anticyclonic circulation anomaly in the Northern Hemisphere. Easterly wind anomalies are collocated with a large portion of the suppressed convective region. This anticyclonic circulation anomaly represents a strengthening and westward shift of the climatological North Pacific subtropical high. A period displaying these convection and circulation features is often referred to as a break monsoon period (Krishnamurti and Subrahmanyam 1982).

The enhanced ISO convection moves northward and expands eastward by day −5 (Fig. 1b), while the suppressed phase dissipates to the north of the Philippines. Easterly anomalies remain present to the north of the enhanced convective region, but are now associated with the developing trough over southern India. As the enhanced convective region continues to move northward and eastward, the trough strengthens such that strong 850-hPa westerly anomalies are collocated with the convection on day +5 (Fig. 1c). This pattern represents an active monsoon period, with a strong monsoon trough and a weaker North Pacific subtropical high. Finally, by day +15, the enhanced phase of the ISO is located to the north of the Philippines, while the suppressed phase has also moved northward and is located over the northern Indian Ocean.

The temporal and spatial evolution of OLR and circulation in Fig. 1 is consistent with earlier studies such as Krishnamurti and Subrahmanyam (1982), Knutson and Weickmann (1987), and Lawrence and Webster (2002). The ISO convection and circulation fields are quite linear in their evolution, such that the maps on days −15 and +5 display opposite-signed anomalies, as do the maps on days −5 and +15. From these figures the period of the ISO convective anomalies can be estimated to be approximately 40 days, also consistent with previous studies.

b. Mixed Rossby gravity waves–TD-type disturbances

Westward-propagating synoptic-scale disturbances in the western Pacific have been studied extensively by Reed and Recker (1971), Hendon and Liebmann (1991), Lau and Lau (1992), Takayabu and Nitta (1993), Sobel and Bretherton (1999), Dickinson and Molinari (2002), and many others. The broad spectrum of westward-propagating disturbances in this region is often categorized into three wave types: easterly waves, mixed Rossby–gravity waves, and tropical depression-type disturbances. The distinctions between these wave types are not necessarily clear, however, such that observed disturbances may display characteristics of more than one type, or represent a transition between types.

Previous studies agree that large-scale, westward-propagating disturbances often behave like theoretical MRG waves in the central Pacific (near the date line), with enhanced (suppressed) convection in the region of lower tropospheric off-equatorial convergence (divergence) associated with cross-equatorial flow. Observations of MRG waves coupled to convection show that they generally have zonal wavelengths on the order of 6000–8000 km and westward phase speeds of 20–25 m s⁻¹ (Liebmann and Hendon 1990; Hendon and Liebmann 1991; Dunkerton and Baldwin 1995; WKW00). As these waves propagate westward into the western Pacific, however, they tend to lose their equatorial antisymmetry in convection and their symmetry in the meridional wind field, and evolve into off-equatorial vortical structures whose centers fill with convection (Takayabu and Nitta 1993; Dickinson and Molinari 2002). These disturbances are generally referred to as tropical depression–type disturbances, as they often serve as the precursors to tropical cyclones in the western Pacific (Lau and Lau 1992; Takayabu and Nitta 1993; Sobel and Bretherton 1999; Dickinson and Molinari 2002).

In this study, MRG and TD-type disturbances are identified using the MRG OLR index of WKW00, which is based on the climatological OLR spectral peak along the MRG dispersion curves. Unlike WKW00, however, in the present study both the antisymmetric and symmetric components of OLR are allowed to pass through the MRG filter. With this change, the index is now able to represent both the antisymmetric MRG waves in the central Pacific and the off-equatorial TD-type disturbances in the western Pacific. To illustrate the difference between these two wave types, Fig. 2a shows the regressed OLR (shading) and 850-hPa circulation fields (streamfunction and vector winds) based on the MRG–TD index in the central Pacific, at 5°N, 175°E, on day 0. Note the antisymmetric OLR pattern, with low (high) OLR in the Northern (Southern) Hemisphere in the region of lower-tropospheric convergence (divergence). The streamfunction anomalies are centered on the equator, as expected. The zonal wavelength is approximately 70°, and the circulation anomalies move westward at 20–25 m s⁻¹. The fields obtained using the MRG OLR index in Fig. 2a are very similar to those obtained by Liebmann and Hendon (1990) using 850-hPa meridional wind as a predictor, by Hendon and Liebmann (1991) using 4–5-day filtered antisymmetric OLR fluctuations, and by Takayabu and Nitta (1993) using 2.5–10-day filtered geostationary meteorological satellite infrared (IR) temperature perturbations.

Figure 2b shows a similar plot, but for the regressed OLR and circulation fields based on the MRG–TD index in the western Pacific, at 10°N, 135°E. The circulation anomalies are now centered off of the equator and propagate northwestward. The OLR anomalies are in phase with the circulation anomalies, with low (high) OLR associated with anomalous cyclonic (anticyclonic) 850-hPa circulations. The scale of the circulation and convection anomalies is smaller than in Fig. 2a, with a wavelength of approximately 35°, and the anomalies propagate westward much more slowly, at 5–10 m s⁻¹. These results are consistent with those obtained for TD-type disturbances by Lau and Lau (1992) and Sobel and Bretherton (1999) using 850-hPa vorticity as a predictor, and by Takayabu and Nitta (1993) using 2.5–10-day filtered GMS IR data.

Figure 3 shows the evolution of the MRG–TD wave activity field (as represented by the variance of the filtered MRG–TD dataset) associated with the low-frequency ISO convection and circulation fields, on days −15, −5, +5, and +15. The cross-hatched areas represent the enhanced ISO convective anomalies (identical to the dark shading in Fig. 1), and the dark contours without shading represent the suppressed ISO convective anomalies. The dark (light) shading represents enhanced (suppressed) MRG–TD activity. The solid (dashed) light contours represent 850-hPa westerly (easterly) anomalies.

On day −15 (Fig. 3a), weakly enhanced MRG–TD activity is collocated with the enhanced low-frequency ISO convection in the equatorial Indian Ocean. MRG–TD activity is strongly suppressed in the region of ISO suppressed convection, particularly in the South China Sea from 5°–15°N, in the region of 850-hPa easterly anomalies. As the enhanced ISO convection moves northward and eastward in time, the region of enhanced MRG–TD activity also moves northeastward, as can be seen on day −5 (Fig. 3b). The suppressed MRG–TD activity expands eastward along the northwest–southeast-tilted axis of suppressed ISO convection, remaining collocated with 850-hPa easterly anomalies. By day +5 (Fig. 3c), the regions of enhanced MRG–TD activity, ISO convection, and 850-hPa westerlies have all continued to move northeastward and intensify, while the suppressed MRG–TD activity, suppressed ISO convection, and easterly anomalies have largely dissipated. The MRG–TD activity spreads eastward by day +15 (Fig. 3d), along with the enhanced ISO convection and 850-hPa westerlies. The MRG–TD wave activity fields show a remarkably coherent relationship with the ISO convection and 850-hPa zonal wind fields. Relationships are not as coherent in the upper troposphere, at 200 hPa.

Aiyyer and Molinari (2002) suggest that the convergent lower-tropospheric wind field of the ISO allows equatorial MRG waves to transition into smaller scale, off-equatorial TD disturbances, based on idealized shallow water model simulations. Observations do show a positive correlation between enhanced MRG–TD activity and 850-hPa ISO convergence anomalies (not shown); however, the spatial correlation appears to be stronger between enhanced MRG–TD activity and 850-hPa westerly anomalies.

There has been little theoretical work to date on the relationship between convectively coupled (moist) equatorial waves and the sign and magnitude of the background wind field. Zhang and Webster (1989) demonstrate that dry MRG waves are less meridionally trapped in westerlies than in easterlies, suggesting that westerly winds may be conducive to the formation of higher amplitude, more latitudinally extensive MRG waves. In addition, Zhang (1993) shows that extratropically forced MRG waves have larger amplitudes in basic-state westerlies when the period of the forcing is greater than 4 days. Since the period of the observed convectively coupled MRG–TD waves in this study varies between 4 and 6 days from the central Pacific to the western Pacific, and the level of MRG–TD activity has been shown to vary strongly as a function of the 850-hPa zonal wind, it is possible that the amplitude of MRG–TD activity may simply be a function of the low-frequency “background” ISO winds. However, as will be shown in section 4, the convection associated with the enhanced MRG–TD activity may compose a substantial fraction of the low-frequency ISO signal, and as such, may itself be partially responsible for accelerating the westerlies and deepening the monsoon trough. A feedback mechanism may exist that allows the enhanced MRG–TD activity to provide the conditions necessary for the continued development of both the ISO and the wave activity itself.

c. Kelvin waves

Convectively coupled Kelvin waves in the Indian Ocean and western Pacific regions have been studied by WKW00, SK02, and SK03. SK02 identified convectively coupled Kelvin waves as the “super cloud clusters” studied by Nakazawa (1988) and Takayabu and Murakami (1991). Convectively coupled Kelvin waves are convective disturbances of zonal scale 2000–4000 km and eastward phase speed 15–20 m s⁻¹, that are coupled to equatorially trapped, Kelvinlike dynamical fields in the tropical troposphere and stratosphere. Nakazawa (1988) and SK02 show that these eastward-propagating convective envelopes are often composed of smaller-scale, westward-propagating “cloud clusters.”

The evolution of Kelvin wave activity with respect to the ISO is shown in Fig. 4, which illustrates the regressed Kelvin wave variance (shading), ISO OLR (hatching, as in Fig. 3), and 200-hPa circulation (streamfunction contours, vector winds). The 200-hPa circulation is illustrated in the Kelvin wave case due to its stronger relationship with Kelvin wave activity. On day −15 (Fig. 4a), weakly enhanced Kelvin wave activity is collocated with the developing low-frequency ISO OLR anomaly in the equatorial Indian Ocean, in a similar manner to the MRG–TD activity illustrated in Fig. 3a. Thus, as the ISO convective anomaly intensifies in the Indian Ocean region on day −15, both eastward-propagating Kelvin waves and westward-propagating MRG–TD waves may be contributing to this lower-frequency convective anomaly. At the same time, suppressed Kelvin wave activity stretches across the Pacific in the Northern Hemisphere, far removed from the suppressed ISO OLR signal in the northern Indian Ocean and western Pacific.

As the ISO convective region moves northward and expands eastward by day −5 (Fig. 4b), the enhanced Kelvin wave activity moves rapidly eastward into the western and central Pacific. Enhanced Kelvin wave activity is no longer collocated with the enhanced ISO OLR signal, as is the MRG–TD activity in Fig. 3b. The enhanced Kelvin wave activity lies in a region of 200-hPa anomalous westerlies (and 850-hPa easterlies), to the north of a zonally elongated subtropical trough in the Southern Hemisphere. By day +5 (Fig. 4c), the enhanced Kelvin wave activity has expanded even farther eastward, stretching across the entire Pacific basin. Finally, by day +15 (Fig. 4d), as the enhanced ISO convection dissipates to the north of the Philippines, the enhanced Kelvin wave activity also decreases over the Pacific.

These results indicate that the low-frequency ISO OLR signal may initially consist of both MRG–TD and Kelvin wave activity in the Indian Ocean region, but that as the ISO convection moves northward, it no longer represents an enhancement of Kelvin wave (or super cluster) activity. Hendon and Liebmann (1994) also found that ISO convective envelopes contained very little eastward-propagating super cluster variance. Instead, it is shown here that enhanced Kelvin wave activity is found well to the east of the ISO OLR convection, maximizing in the central Pacific. This relationship suggests that Kelvin waves may be initiated by the active convection within the ISO envelope, and then subsequently propagate eastward across the Pacific. Although dry Kelvin waves have theoretically been shown to be unaffected by vertical wind shear (Wang and Xie 1996), it is possible that convectively coupled Kelvin waves preferentially form (or are more readily sustained after formation) in the enhanced low-level basic-state easterlies or weakened upper-level basic-state easterlies in the Pacific that are associated with the large-scale dynamical response to ISO convection.

As an alternative hypothesis to explain the enhancement of Kelvin wave variance to the east of the ISO convective anomaly, SK03 show that the initiation of convectively coupled Kelvin waves in the central Pacific during boreal summer is associated with an equatorward-propagating Rossby wave train excited in the Southern Hemisphere subtropical jet. If the circulation fields associated with the low-frequency ISO convection cause low-frequency changes in the characteristics of wave activity in the jet, it is possible that these changes could subsequently cause low-frequency changes in tropical Kelvin wave activity. A related mechanism for the boreal winter ISO is proposed in Matthews and Kiladis (1999), in which ISO convection forces low-frequency changes in both the Northern Hemisphere Asian jet and the downstream basic state, which then affects the supply and evolution of equatorward-propagating Rossby waves in the eastern Pacific.

The Southern Hemisphere subtropical jet is located along approximately 30°S during JJA, and maximizes to the east of Australia. Rossby waves excited in the jet often propagate northeastward over Australia, then eastward along approximately 20°S (Chang 1999; SK03). Therefore, low-frequency changes in the circulation along 30°S represent fluctuations in the strength and location of the subtropical jet, and may be correlated with changes in wave activity within the jet. The variance of <30-day filtered 200-hPa meridional wind is used here as a measure of Rossby wave activity in the jet, as in SK03.

On day −15 (Fig. 5a), a subtropical low pressure anomaly is located over Australia, which developed in association with the anomalous upper-tropospheric inflow to (i.e., weaker than normal outflow from) the suppressed convective region in the western Pacific. Easterly anomalies extend along 35°S from the central Indian Ocean to eastern Australia. Since the climatological subtropical jet core is located over eastern Australia, the easterly anomalies to the west of 140°E represent a weakening of the entrance region of the jet. Wave activity along the jet is weaker than normal, as evidenced by the negative anomalies in submonthly meridional wind variance over Australia and to its east. Kelvin wave activity in the Pacific is substantially diminished at this time, as shown in Fig. 4a.

As the suppressed ISO convective region dissipates and the enhanced convective region expands, the cyclonic circulation anomaly in the Southern Hemisphere expands eastward (see Fig. 4b), consistent with a rotational response to the strengthening equatorial westerlies forced by the upper-tropospheric outflow from the ISO convection. At this time, the anomalous easterlies along 30°S to the east of 120°E weaken the climatological exit region of the subtropical jet, and wave activity begins to propagate northward and southward of the climatological jet location (not shown).

By day +5 (Fig. 5b), the entrance region of the subtropical jet is strengthened by the ISO-influenced westerly anomalies associated with the anticyclonic outflow from the enhanced ISO convection in the Northern Hemisphere. Anomalously high Rossby wave activity is found along the climatological jet axis at this time, extending along the anomalous southwesterly flow toward the Tropics. Consistent with the analysis in SK03, anomalous E-vectors point northeastward at this time (not shown), signifying the enhanced equatorward propagation of Rossby wave activity over Australia. These equatorward-propagating Rossby wave perturbations may then excite convectively coupled Kelvin waves in the Pacific, as illustrated in SK03, and produce the enhanced Kelvin wave variance signal seen in the Tropics of the Northern Hemisphere in Fig. 4c.

The results presented in sections 3b and 3c have been confirmed using several other methodologies. The MRG–TD results are consistent with regressions based on 850-hPa 30–70-day filtered zonal wind in the northern Indian Ocean region, and the Kelvin wave results are consistent with regressions based on the 200-hPa 30–70-day filtered zonal wind in both the Southern Hemisphere subtropical jet region and the equatorial Pacific. Both sets of results can also be duplicated through a calculation of composite fields during strong ISO events.

The results presented in this section suggest two possible mechanisms for the enhancement of Kelvin wave activity to the east of an active ISO region. First, a region of active low-frequency ISO convection may directly excite shorter-timescale Kelvin waves that subsequently propagate eastward across the Pacific. Second, active ISO convection may indirectly cause an enhancement in Kelvin wave activity through its induced low-frequency extratropical response, which triggers an increase in higher-frequency wave activity in the subtropical jet, consequently exciting tropical Kelvin waves in the Pacific via equatorward-propagating Rossby waves. Both mechanisms are supported by the data, and it is difficult at present to determine which is the more likely candidate.

a. Westward-propagating modes

It is clear from Fig. 6 that the eastward-propagating active ISO convective signal during August 1987 is composed of many shorter timescale, smaller spatial scale convective events, many of which propagate westward with time. For example, a strong MRG–TD wave packet composes a substantial portion of the enhanced ISO convective phase. Horizontal maps of the convection and circulation fields during the enhanced ISO phase are shown in Fig. 7, which illustrates the total OLR (shading), 30–96-day filtered 850-hPa zonal wind (light contours) and wind vectors, and ISO-filtered OLR (dark contour surrounds negative OLR anomalies less than −10 W m⁻²), every 5 days from 13 August 1987 to 7 September 1987. Locations of named tropical storms are indicated. It will be shown in this section that these western Pacific tropical cyclones and their associated circulation anomalies are the precursors to the MRG–TD disturbances indicated in Fig. 6, and also provide much of the convective variability within the active ISO envelope.

On 13 August (Fig. 7a), the active ISO envelope of convection stretches in a northwest–southeast direction from the northwestern Indian Ocean to the equatorial region to the east of the Maritime Continent. The ISO filtered OLR on 13 August is quite similar to the regressed OLR on day −5 in Fig. 3b. A region of 30–96-day filtered 850-hPa westerly perturbations exists in the western Indian Ocean, just to the north of the equator, also as seen in the regressed fields on day −5 in Fig. 3b. A large region of deep convection is located over southeastern India, which appears to have developed in situ and does not represent a westward-propagating mode. In the western Pacific, two tropical cyclones, Betty and Cary, have formed on the northern edge of the ISO envelope. Betty is located within a well-developed, large-scale 850-hPa cyclonic circulation, while Cary is located on the northwestern side of an 850-hPa anticyclone (not shown). Five days later, on 18 August (Fig. 7b), the remnants of Supertyphoon (ST) Betty have crossed southeast Asia, remaining coherent as the convection and circulation propagate westward. Typhoon (TY) Cary continues to strengthen in the western Pacific. The OLR and <10-day filtered 850-hPa circulation anomalies associated with these disturbances on 18 August are shown in Fig. 8. A wave train of cyclonic and anticyclonic vortices arcs southeastward from ex-ST Betty, through TY Cary, toward the equator, in a similar manner to the MRG–TD regressed fields in Fig. 2b. The scale of the circulation anomalies increases toward the east, and the circulations become more equatorially centered, suggesting that Betty and Cary may have formed in association with an MRG-to-TD-type transition. In addition, the region of low OLR at 10°N, 165°E in Fig. 8, located in the southeasterly flow of the MRG-like disturbance centered at 160°E, eventually develops into ST Dinah as it propagates northwestward. A similar evolution of MRG waves to TD disturbances during early July 1987 is illustrated by Dickinson and Molinari (2002).

Once ex-ST Betty reaches the southeast Asian landmass, it accelerates westward and projects onto the MRG–TD OLR band, as seen in Fig. 6 by the westward-propagating features enclosed by the MRG–TD filtered OLR contours. A similar evolution is observed for Typhoon (TY) Cary. On 23 August (Fig. 7c), TY Cary is still analyzed as a tropical cyclone, even though it is located over land, due to its coherence as a tropical stormlike structure. Saha et al. (1981) document numerous instances of western Pacific tropical cyclones propagating westward across southeast Asia, many regenerating into monsoon depressions in the Bay of Bengal.

The relationship between convection and the lower-tropospheric circulation during the development and propagation of Betty, Cary, and Dinah is shown in Fig. 9a, which illustrates the evolution of the OLR anomaly field (shading, total OLR minus the mean and first three annual harmonics) and the <10 day filtered 850-hPa meridional wind (contours), averaged from 5°–20°N, from 8 to 28 August. Individual convective disturbances propagate westward, and are flanked by northerly (southerly) anomalies to their west (east), representing the wave train of lower-tropospheric cyclonic and anticyclonic circulations that propagate westward with the OLR. An eastward dispersion of energy can be seen in the wind field, which may account for the continued development of subsequent tropical cyclones to the east of their previous locations. It is suggested that this eastward dispersion of energy from an existing wave train produces new 850-hPa circulations to its east, which then couple with convection to produce new tropical storms.

For comparison purposes, Fig. 9b shows the regressed OLR (shading) and 850-hPa meridional wind (contours) associated with a +20 W m⁻² perturbation in the MRG–TD filtered OLR at the basepoint 10°N, 95°E, also averaged from 5° to 20°N. In a similar manner to the case study data in Fig. 8a, the regressed OLR and meridional wind fields are in quadrature, with northerlies (southerlies) to the west (east) of the negative OLR anomalies. A similar eastward dispersion of energy is seen in both the OLR and meridional wind fields. The strong resemblance between the case study fields in Fig. 9a and the regressed fields in Fig. 9b suggests that the evolution of these westward-propagating modes may be predictable, and validates the use of the filtered fields in the statistical analysis in section 3.

Returning to Fig. 7, note that throughout the period shown in Fig. 9, the 30–96-day filtered 850-hPa westerlies accelerate and extend eastward in the region to the south of the enhanced convection, as the individual tropical storms and MRG–TD disturbances propagate westward. This evolution suggests that the higher-frequency westerlies associated with the individual storms project onto the lower-frequency ISO filtered band, and extend farther eastward in time due to the eastward dispersion of MRG–TD energy and continued tropical storm development. Furthermore, the slow eastward phase speed of the ISO may be related to the slow eastward group velocity of an MRG–TD wave packet: both are approximately 5 m s⁻¹ (see Fig. 12a in WKW00 for observations of the eastward group velocity of MRG waves).

On 23 August (Fig. 7c), westerly anomalies extend to 120°E, and begin to develop near 140°E. Two more tropical cyclones, Dinah and Ed, have formed in the ISO active convective envelope by this time, to the south of a region of low-frequency easterlies along 10°–20°N. The convection and circulation anomalies on 23 August are quite similar to the regressed fields on day +5 in Fig. 3c. It is not surprising that tropical cyclones continue to develop in the large region of enhanced cyclonic zonal wind shear along 10°N in the western Pacific: climatologically, 42% of tropical disturbances in the western Pacific form in this type of monsoon shear line environment (Ritchie and Holland 1999). On 28 August (Fig. 7d), Dinah and Ed are still visible in the northern edge of the active ISO envelope, and westerly anomalies extend into the western Pacific.

By 2 September (Fig. 7e), anomalous westerlies extend nearly to the date line (compare with the regressed fields on day +15 in Fig. 3d), and the low-frequency ISO OLR signal consists of three developing tropical storms: Freda, Gerald, and Holly, which again line up roughly along the shear line in the monsoon trough. On 7 September (Fig. 7f), these three tropical storms extend across the western Pacific from 120° to 170°E, and are flanked to their south by strong westerly anomalies. The development of these storms does not appear to correspond to a continued eastward dispersion of disturbance energy, however; instead, these storms all form at approximately the same time in a manner consistent with the breakdown of a monsoon trough via barotropic instability (Ferreira and Schubert 1997). It is suggested that the eastward movement of low-level westerly anomalies, initiated by the eastward dispersion of disturbances in the MRG–TD band, is responsible for the development of the strong monsoon trough circulation and its later breakdown through barotropic instability.

The spatial and temporal development of multiple tropical cyclones and MRG–TD disturbances in the context of the low-frequency ISO convection and westerly wind anomalies during this case study is intriguing. The evolution of these fields suggests the possibility of a complex interaction between the development of high-frequency tropical disturbances, their net effect on the low-frequency convection and circulation, and the resulting effects of the low-frequency circulation back onto the development of the high-frequency disturbances. If the cyclonic circulation anomalies associated with the tropical cyclones and MRG–TD disturbances are consistently stronger than the anticyclonic anomalies, as might be expected due to the effects of latent heating, a net westerly (easterly) acceleration could be produced to the south (north) of the disturbances. This effect could be responsible for at least a part of the low-frequency westerly acceleration observed to the south of the convective anomalies in the case study. As mentioned, these westerly anomalies may then provide an environment more conducive to the continued development of tropical storms.

In summary, the low-frequency ISO signal during August–September 1987 consists of seven tropical storms in the western Pacific, two of which continue to propagate westward and project onto MRG–TD disturbances over southeast Asia and India. An eastward dispersion of energy appears to provide the low-level cyclonic circulations necessary for the continued development of several tropical storms (Betty, Cary, and Dinah). An eastward propagation of low-frequency westerly anomalies may also influence the development of storms along the axis of the monsoon trough. As demonstrated statistically in section 3b, the low-frequency ISO signal is collocated with an enhancement of activity in the MRG–TD band and strong 850-hPa westerly anomalies.

b. Eastward-propagating modes

The statistical results presented in section 3c suggest that Kelvin wave activity tends to be enhanced in the Pacific during the active convective phase of the ISO. As shown in Fig. 6, two Kelvin wave events propagate eastward across the Pacific at the end of the active ISO phase, during late August and early September 1987. Another Kelvin wave exists just prior to the first strong event in mid-August, but as its convective signal is centered near the equator, it does not appear in Fig. 6 due to the effects of averaging over Northern Hemisphere latitudes. No significant eastward-propagating convective events are observed during the suppressed phase of the ISO in July and early August.

Figure 10 shows the averaged ISO filtered OLR (hatching), total Kelvin wave variance (dark contours), 30–96-day filtered 200-hPa winds (streamfunction contours and vector winds), and 200-hPa <30-day filtered meridional wind variance anomalies (shading) for the suppressed (Fig. 10a) and active (Fig. 10b) periods of the ISO during July–August 1987. The suppressed and active periods were each chosen to be 20 days in length, or one-half of the typical 40-day ISO cycle. The date range for each period was determined by matching the daily maps of ISO filtered OLR with the regressed fields in Fig. 4, such that the averaged period corresponds with the period between day −10 and day +10 in the regression. The suppressed period extends from 20 July to 8 August, and the active period extends from 11 to 30 August. The 200-hPa meridional wind variance anomalies were calculated by subtracting the 20-day average for each period from the 20 July–30 August average. Admittedly, the sample periods are quite short and are somewhat subjectively defined; however, Fig. 10 is intended only to illustrate that the statistical relationships presented in Fig. 4 hold even in an individual case study.

The suppressed ISO phase in Fig. 10a contains very little Kelvin wave variance compared to the active ISO phase in Fig. 10b. The 30–96-day filtered 200-hPa wind anomalies are almost exactly opposite in sign between the two periods, with an anomalous anticyclonic (cyclonic) circulation over Australia and an elongated cyclonic (anticyclonic) circulation extending across the Pacific during the active (suppressed) ISO phase. Rossby wave activity over Australia is enhanced (suppressed) during the active (suppressed) ISO period, and activity over the Pacific is located anomalously north (south) of its period mean. These relationships suggest that during the active phase of the ISO, wave activity propagates northeastward over Australia, and then eastward at a more northern latitude than in the suppressed ISO phase, providing the subtropical circulation anomalies shown to be associated with Kelvin wave initiation in SK03. Specifically, daily maps of OLR and 200-hPa circulation show that on 18 August, a Rossby wave train propagates northeastward over Australia and is associated with the initiation of a Kelvin wave event in the central Pacific (not shown; however, the northwest–southeast tilted band of convection associated with the equatorward-propagating wave train can be seen to the northeast of Australia in Fig. 7b).

The results from the case study during July–September 1987 provide evidence that the statistical results presented in section 3c can be seen even in individual events. Kelvin waves were shown to be more frequent during the active phase of the ISO than during the suppressed phase. This change may have been due to low-frequency fluctuations in the subtropical jet, which change the propagation path of subtropical Rossby wave disturbances, or perhaps to an overall increase in the convective activity in the western Pacific, which would allow Kelvin waves to initiate in that region and subsequently propagate eastward.