Topographic Effects on the Eastward Propagation and Initiation of the Madden–Julian Oscillation

Huang-Hsiung Hsu Department of Atmospheric Sciences, National Taiwan University, Taipei, Taiwan

Search for other papers by Huang-Hsiung Hsu in
Current site
Google Scholar
PubMed
Close
and
Ming-Ying Lee Department of Atmospheric Sciences, National Taiwan University, Taipei, Taiwan

Search for other papers by Ming-Ying Lee in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

This study investigates the relationship between deep convection (and heating anomaly) in the Madden–Julian oscillation (MJO) and the tropical topography. The eastward propagation of the deep heating anomalies is confined to two regions: the Indian Ocean and the western Pacific warm pool. Superimposed on the eastward propagation is a series of quasi-stationary deep heating anomalies that occur sequentially and discretely downstream in a leapfrog manner in the central Indian Ocean, the Maritime Continent, tropical South America, and tropical Africa.

The deep heating anomaly, usually preceded by near-surface moisture convergence and shallow heating anomalies, tends to occur on the windward side of the tropical topography in these regions (except the central Indian Ocean) under the prevailing surface easterly anomaly of the MJO. It is suggested that the lifting and frictional effects of the tropical topography and landmass induce the near-surface moisture convergence anomaly, which in turn triggers the deep heating anomaly. Subsequently, the old heating anomaly located to the west of the tropical topography weakens and the new heating anomaly east of the topography develops because of the eastward shift in the major moisture convergence center to the east of the mountains. Therefore, the deep heating anomaly shifts eastward from one region to another. The equatorial Kelvin wave, which is forced by the tropical heating anomaly and propagates quickly across the ocean basins in the lower troposphere, plays an important role by helping to strengthen the easterly anomaly and lowering the surface pressure.

This process is proposed to further our understanding of the shift in the deep convection from the Indian Ocean to the western Pacific, the reappearance of the deep convection in tropical South America, and the initiation of the MJO in the western Indian Ocean. It is suggested that the fast eastward propagation and the slow development of quasi-stationary convection together determine the quasi-periodicity of the MJO.

Corresponding author address: Dr. Huang-Hsiung Hsu, Department of Atmospheric Sciences, National Taiwan University, No. 61, Lane 144, Section 4, Keelung Road, Taipei, Taiwan. Email: hsu@atmos1.as.ntu.edu.tw

Abstract

This study investigates the relationship between deep convection (and heating anomaly) in the Madden–Julian oscillation (MJO) and the tropical topography. The eastward propagation of the deep heating anomalies is confined to two regions: the Indian Ocean and the western Pacific warm pool. Superimposed on the eastward propagation is a series of quasi-stationary deep heating anomalies that occur sequentially and discretely downstream in a leapfrog manner in the central Indian Ocean, the Maritime Continent, tropical South America, and tropical Africa.

The deep heating anomaly, usually preceded by near-surface moisture convergence and shallow heating anomalies, tends to occur on the windward side of the tropical topography in these regions (except the central Indian Ocean) under the prevailing surface easterly anomaly of the MJO. It is suggested that the lifting and frictional effects of the tropical topography and landmass induce the near-surface moisture convergence anomaly, which in turn triggers the deep heating anomaly. Subsequently, the old heating anomaly located to the west of the tropical topography weakens and the new heating anomaly east of the topography develops because of the eastward shift in the major moisture convergence center to the east of the mountains. Therefore, the deep heating anomaly shifts eastward from one region to another. The equatorial Kelvin wave, which is forced by the tropical heating anomaly and propagates quickly across the ocean basins in the lower troposphere, plays an important role by helping to strengthen the easterly anomaly and lowering the surface pressure.

This process is proposed to further our understanding of the shift in the deep convection from the Indian Ocean to the western Pacific, the reappearance of the deep convection in tropical South America, and the initiation of the MJO in the western Indian Ocean. It is suggested that the fast eastward propagation and the slow development of quasi-stationary convection together determine the quasi-periodicity of the MJO.

Corresponding author address: Dr. Huang-Hsiung Hsu, Department of Atmospheric Sciences, National Taiwan University, No. 61, Lane 144, Section 4, Keelung Road, Taipei, Taiwan. Email: hsu@atmos1.as.ntu.edu.tw

1. Introduction

Eastward propagation is one of the most fascinating features of the Madden–Julian oscillation (MJO), which is most prominent in the Indian Ocean and the western Pacific. An eastward-propagating MJO often originates in the western Indian Ocean. Although numerous studies have explored the nature of the MJO, the reason for its origin in the western Indian Ocean is not fully understood. Hsu et al. (1990) suggested a triggering effect by extratropical perturbations on the tropical convection based on a case study. Kiladis and Weickmann (1992), however, found that the triggering effect might be more effective in the eastern Pacific where the equatorial westerly exists in the upper troposphere and allows the extratropical Rossby wave to propagate into the Tropics (Hsu and Lin 1992). Lau and Peng (1987) and Chang and Lim (1988) proposed that the eastward-propagating Kelvin wave might trigger the deep convection when it propagates into the moist and warm Indian Ocean from the west. The subsequent coupling between the circulation and tropical convection results in the eastward-propagating MJO in the Indian Ocean and the western Pacific. However, how this deep convection is trigged has not been studied in detail. Conversely, Hu and Randall (1994) and Bladé and Hartmann (1993) suggested that a self-adjustment in the tropical atmosphere could produce this intraseasonal oscillation between active and inactive convection regimes without Kelvin wave initiation. This mechanism, however, cannot explain the origin of the MJO in the western Indian Ocean or in other regions where the background state is also suitable for the development of deep convection.

Several mechanisms, mostly based on the Kelvin–Rossby wave conditional instability of the second kind (wave-CISK), have been proposed to explain the eastward propagation of the MJO. A summary of these proposed mechanisms can be found in Seo and Kim (2003) and Hsu et al. (2004). However, the eastward propagation is by no means a smooth propagation at a constant speed. It has long been noticed that the convection associated with the MJO tends to flare in the Indian Ocean and western Pacific warm pool. During enhancement in these areas, the MJO tends to be stationary for certain periods before continuing its eastward movement (e.g., Khalsa and Steiner 1988; Wang and Rui 1990a). Hsu et al. (1990) and Weickmann and Khalsa (1990) reported a sudden shift in the deep convection from the Indian Ocean to the western Pacific. The regions where the MJO convection amplifies and becomes stationary are the regions of high sea surface temperature (SST) and moisture content. Several studies (e.g., Salby et al. 1994; Wang and Xie 1998) suggested that the positive feedback between the convection and large-scale circulation results in MJO amplification in these areas. Zhang and Hendon (1997) concluded that there is little evidence of standing oscillation based on the space–time spectral analysis and suggested that the amplification in the equatorial Indian and western Pacific Oceans yields the impression of a standing oscillation occurring. However, the spectral analysis they applied might project local signals to the global wavenumber domain and have the standing signals diffused.

The geographic dependence discussed above suggests that the eastward propagation driven by the wave-CISK mechanism might be affected by the characteristics of the underlying surface conditions. The frictional effect was proposed by several studies (e.g., Wang and Rui 1990b; Hendon and Salby 1994) to result in moisture convergence to the east of the deep convection, which in turn contributes to the eastward propagation. The strength of this frictional convergence is likely to depend on the underlying surface characteristics. For example, the frictional effect might be more significant in the mountainous land area than in the open sea. In that case, one might wonder whether the complex land–sea contrast and topography in the Maritime Continent (Fig. 1) has an effect on the shift in convection from the Indian Ocean to the western Pacific. The possible effect of the topography and land–sea contrast on the propagation nature of the MJO is an interesting and yet hardly explored subject.

Previous studies (e.g., Knutson and Weickmann 1987; Hsu 1996) found that the velocity potential often reveals the global-scale feature propagating around the globe along the equator. Conversely, the OLR reveals local convection propagating eastward mainly in the Indian Ocean and western Pacific. It has been noted that the occurrence of a deep convection in tropical South America lags its counterpart in the western Pacific by certain periods. This lagged relationship explains the discrepancy between the OLR and velocity potential anomalies. The former is a proxy variable representing the local deep convection in the Tropics. The latter can be obtained from the divergence field through the inverse Laplacian operator. It therefore represents a greatly smoothed picture of the divergence field. When the deep convection is weakening in the western Pacific and the deep convection is developing in tropical South America, the computed velocity potential is likely to have a maximum in the eastern Pacific. Further development of the deep convection in tropical South America and the dissipation of the deep convection in the western Pacific results in the farther eastward propagation of the velocity potential. This effect can be seen clearly when the velocity potential and outgoing longwave radiation (OLR) anomalies were plotted in the same figure, as demonstrated by Hsu (1996). While this explanation explains the discrepancy in visualizing the OLR and velocity potential, it does not explain the reappearance of the deep convection in tropical South America, which carries the MJO signal farther eastward.

The equatorial Kelvin wave was proposed to explain the eastward propagation of the velocity potential beyond the date line. The wave forced by the deep convection in the western Pacific and sustained by the boundary forcing propagates through the drier and cooler eastern Pacific and triggers convection in tropical South America (e.g., Sui and Lau 1989). However, as in the origin problem, more studies are still needed to understand how the convection in tropical South America is triggered.

This study examines the evolution of the MJO and finds that the stationary component of the tropical convection associated with the MJO tends to occur nearby the mountainous regions (Fig. 1) beside the high SST regions. This observation could shed light on the three problems discussed above: the origin of the MJO in the western Indian Ocean, the shift in convection from the Indian Ocean to the western Pacific, and the reappearance of deep convection in tropical America. Section 2 describes the data and analysis procedure. The evolution and geographic dependence of the deep convection and near-surface moisture divergence associated with the MJO are examined in section 3. A detailed vertical structure of the MJO at different stages and its relationship with the topography are explored in section 4. Possible effects of the tropical topography on the MJO are discussed in section 5.

2. Data and analysis procedure

The data used in this study includes 1) the European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA; Gibson et al. 1997), and 2) the interpolated OLR (Liebmann and Smith 1996). The ERA and OLR data covers 14 winters (defined as the period from November to April of the next year) from 1979/80 to 1992/93 on a 2.5° × 2.5° grid. To isolate the intraseasonal signals, all data were filtered through a 30–60-day bandpass Butterworth filter (Hamming 1989; Kaylor 1977). Zonal means were then removed to emphasize the eddy components of the circulation. To show the large-scale structure clearly, OLR, divergence, vertical velocity, and diabatic heating were spatially smoothed using a T24 spectral smoothing technique proposed by Sardeshmukh and Hoskins (1984). One might suspect that the spatial smoothing could smear out the detailed structure associated with the MJO. The statistics presented below were double checked by comparing them to those derived from the nonsmoothed fields. It was found that, although some of the detailed structures might be smoothed out, the smoothing procedure did not remove evidence that would lead to different conclusions. Smoothing also helped isolate the relevant large-scale features embedded in the MJO.

Regression analysis was applied to various variables to delineate the temporal evolution and the spatial structure of the MJO. Statistical significance tests were applied to all regression maps. The t score of each regression coefficient was computed. The null hypothesis of no relationship was rejected when the computed t score is larger than the t value at 95% in the Student’s t distribution. Only those results rejecting the null hypothesis are plotted in most of the following figures.

3. General characteristics

This section reviews the MJO characteristics related to the three problems previously raised. As discussed in the introduction, the MJO amplitude exhibits a strong geographic dependence. This can be seen in Fig. 2a, which presents the geographic distribution of the bandpass-filtered OLR variance. A large variance is found in the Indian Ocean and the western Pacific with the maximum variance located near the equator around 90°, 125°, and 165°E. Also shown in Fig. 2a is the climatological mean SST. The large OLR variance tends to be located near the high SST areas. In the equatorial eastern Pacific, where the SST is relatively low, the variance is small. This collocation demonstrates the close relationship between the MJO and the SST. The variance is also large over the land area in eastern South America and southeastern Africa where the deep convection is climatologically active.

Figure 2b presents the propagation tendency of the bandpass-filtered OLR at each point. The vectors, which are drawn based on the 5-day lagged one-point correlation maps for each grid point, represent the distance and direction of an OLR anomaly propagating from the base point to the location where the lagged correlation is the largest 5 days later. Although a large OLR variance associated with the MJO is observed in the Tropics at almost all longitudes except the eastern Pacific, the eastward propagation occurs mainly in two regions: the eastern Indian Ocean (e.g., from 60° to 120°E) and the western Pacific warm pool (from 135°E to 180°). The Maritime Continent where the eastward propagation is weak separates these two regions. This suggests that the deep convection originated from the Indian Ocean does not continuously propagate eastward through the Maritime Continent. Instead, it seems to trigger a new convection in the east, which propagates farther eastward to the date line. This view is consistent with the finding of Seo and Kim (2003), which suggested the initiation of a new convection in the western Pacific. Beyond the date line, the eastward-propagation tendency weakens drastically. Interestingly, the large variance in tropical South America and Africa are not associated with the eastward propagation of the OLR anomaly. There is even westward propagation of the OLR anomaly in the equatorial eastern Pacific and the tropical Atlantic. The propagation tendency of the MJO apparently exhibits stronger geographic dependence than the variance distribution. This contrast suggests that the 30–60-day OLR variability in the Tropics is not always associated with the eastward propagation of the deep convection.

Nevertheless, the regions of large OLR variance are all under the influence of the global MJO, which is best depicted by the velocity potential, as demonstrated in Knutson and Weickmann (1987). Because this study focuses on the MJO that exhibits a strong tendency to propagate beyond the date line and even globally, the velocity potential seems to be an appropriate variable for identifying this phenomenon. Empirical orthogonal function (EOF) analysis was applied to the 200-hPa velocity potential between 40°N and 40°S. The first two EOFs explain 36.5% and 28.4% of the total variance. Both EOFs exhibit a zonal wavenumber-1 structure and fluctuate in time with the first principal component leading the second one by about a quarter of a cycle. These two EOFs, which explain about 65% of the variance, form a tropical pattern propagating eastward around the globe. Because the EOF result resembles the pattern shown in many previous studies (e.g., Knutson and Weickmann 1987), the spatial structure and temporal variation in the first two EOFs are not shown here for the sake of brevity. The first principal component was chosen as the index for further analysis described below. Similar results can be obtained if the second principal component were chosen as the index.

Figure 3 shows the regression coefficients of the OLR and 925-hPa moisture divergence regressed upon the first principal component. Only the equatorial strip from 15°N to 20°S is presented from days −22 to 22 for every 2 days. This figure is arranged in such a manner as to mimic the Hovmöller diagram, and at the same time, retain the spatial structure. Due to the filtering procedure, the evolution beyond day 22 can be inferred from day −22 onward. Figure 3 can be viewed as a loop of the MJO evolution.

At first glance at Fig. 3, one generally notices the eastward propagation of the overall OLR anomaly from the western Indian Ocean to the date line in about 30 days. However, each regional center tends to move at a much slower speed or even becomes stationary. The centers located near 95° and 125°E are nearly stationary from the onset to the dissipation. For example, the center in the Indian Ocean moves eastward by only about 10° longitudes from days −14 to 4. The center near 125°E hardly moves from days −4 to 10. The center in the western Pacific warm pool exhibits the strongest tendency to propagate eastward and moves from 160°E to about 160°W from days 10 to 22. While the western two centers exhibit the stationary tendency, the overall anomalous convection activity develops systematically eastward at a much greater speed. Moreover, the three centers described above appear one after another in sequence in a manner such that the western center starts dissipating when the eastern center starts developing. The development in the Indian Ocean and the western Pacific between days −4 and 2 is similar to the shift in convection from the eastern Indian Ocean to the western Pacific reported by Hsu et al. (1990) and Weickmann and Khalsa (1990). These characteristics suggest that the eastward development of the overall convection activity is a combination of the eastward propagation associated with the tropical wave propagation and the sequential downstream development at specific geographic locations.

The occurrence of the near-surface moisture divergence anomaly tends to precede the OLR anomaly by 8–10 days. For example, the moisture convergence in the Indian Ocean is already evident at day −22, while the OLR anomaly does not appear until day −14. The moisture convergence in the Maritime Continent is observed at day −12, while the OLR anomaly does not develop until day −2. The existence of the near-surface moisture convergence in these regions appears to be a precondition for the later deep convection development. The stationary component also exists in the 925-hPa moisture divergence field. This feature is particularly evident near Borneo (e.g., 125°E) and New Guinea (e.g., 150°E). The moisture divergence anomalies in these two regions hardly move during the active periods (e.g., from days −12 to 8). It is interesting to note that these two regions are located near the mountainous islands in the Maritime Continent (Fig. 1). This suggests that the topography and the land–sea contrast in the Maritime Continent might contribute to the observed stationary components.

During the course of eastward propagation in the western Pacific warm pool (e.g., beyond 150°E), the near-surface moisture divergence anomaly is located to the east of the OLR anomaly. This phase relationship between the moisture convergence and the deep convection is consistent with the frictional wave-CISK mechanism and the observational results presented in previous studies (e.g., Salby et al. 1994; Maloney and Hartmann 1998; Mathews 2000). There is indication of the continuing eastward propagation of moisture convergence beyond the date line. However, the negative OLR anomaly stalls near the date line. Apparently, the frictional wave-CISK mechanism ceases to function over the cooler ocean surface in the equatorial eastern Pacific.

Although the deep convection ceases propagating eastward in the eastern Pacific, a new deep convection in tropical South America begins developing at day 10 and continues strengthening afterward. This time-lagged relationship between the deep convection in the western Pacific and tropical South America results in the continuously eastward propagation in the velocity potential field as discussed in the introduction. Like its counterpart in the Indian Ocean and the western Pacific, the OLR anomaly in tropical South America is also preceded by the near-surface moisture divergence anomaly. For example, an anomalous moisture convergence appears on the eastern side of the Andes (Fig. 3) as early as day −8, long before the deep convection initiation. The OLR anomaly is in fact positive at this time, indicating inactive convection. The anomalous convergence zone shifts eastward to northern Brazil at day 10 when the deep convection starts developing there. During the evolution, the near-surface moisture convergence is always located to the east of the deep convection. To understand the lagged teleconnection between the convection activity in the western Pacific and tropical South America, it would be essential to explore how the near-surface moisture divergence anomaly is induced in situ in tropical South America when there is no clear moisture divergence anomaly signal propagating eastward from the western Pacific to the region.

As discussed in the introduction, the eastward-propagating MJO tends to originate in the western Indian Ocean. This can be seen clearly in Fig. 3. For example, the negative OLR anomaly in the central Indian Ocean appearing at day −14 is preceded by the near-surface moisture convergence, which can be traced farther west to the region off the eastern African coast. There is no indication of moisture convergence propagating from the Atlantic into the region before the onset of the moisture convergence off the eastern African coast. This near-surface moisture convergence, which seems to occur in situ, could be the key to understanding the initiation of the eastward-propagating MJO in the western Indian Ocean. Interestingly, this moisture convergence occurs to the east of the eastern African highland (Fig. 1) where the topographic lifting effect might contribute to the initiation of the near-surface convergence.

The above results suggest that the three problems raised in the introduction might have to do with the topography and land–sea contrast (Fig. 1). This relationship is explored in detail in the following section.

4. Horizontal and vertical structures

Horizontal and vertical structures associated with the MJO, shown in Fig. 3, are presented in this section to delineate the possible relationship with the topography and land–sea contrast. Figure 4 shows a cross section of the circulation and diabatic heating averaged from 10°S to 5°N for every 5 days from days −20 to 20. The diabatic heating is defined as the residue from the thermodynamics equation. The terrain height averaged over the land areas within the latitudinal band is also plotted. The lower portion in each plot is stretched purposely to emphasize the features in the lower troposphere. The corresponding horizontal distribution of the 1000-hPa height and wind anomalies are shown in Fig. 5. The features presented in Figs. 4 and 5 complete about a cycle of the time evolution from days −20 to 20. The plots for days −20 and −15 can be viewed as the plots for days 25 and 30, and so on.

The sequential downstream development of the OLR anomalies described in the preceding section can be seen even more clearly in Fig. 4. The sequence starts with the development of the positive diabatic heating anomaly in the Indian Ocean at day −15. Two new anomalies appear to the east of mountainous Borneo and New Guinea, respectively, at day 0 when the anomaly in the Indian Ocean dies out. Note that the western anomaly leads the eastern anomaly by about 5 days. The anomaly located to the east of New Guinea moves eastward in the following period and dissipates before reaching the date line at day 15. At the same time, deep heating anomalies appear in both tropical South America and western Africa. While these two heating anomalies dissipate 15 days later (e.g., day −10), a positive heating anomaly appears in the Indian Ocean. This upstream dissipation and downstream development sequence clearly reveals that the deep heating anomalies develop sequentially and discretely downstream in a leapfrog manner.

a. Shift in convection from the Indian Ocean to the western Pacific

The development of each deep heating anomaly is always preceded by a near-surface moisture convergence (i.e., negative anomaly). At days −10 and −5 when the deep heating anomaly in the Indian Ocean is still strong and the ones in the Maritime Continent have not yet developed, two centers of near-surface moisture convergence and shallow heating are evident east of Borneo and New Guinea, respectively. Note that the moisture convergence in the Maritime Continent is well separated from its counterpart in the Indian Ocean, which is located right under the deep heating anomaly.

In contrast to the diabatic heating and moisture convergence, the wind pattern in the lower troposphere exhibits continuous eastward propagation. The prevailing easterly anomaly propagating eastward from the western Indian Ocean reaches the Maritime Continent at day −15 and the date line at day −10. This feature can also be clearly seen in the 1000-hPa height and wind anomalies shown in Fig. 5. A tongue of negative height and easterly anomaly extends quickly across the equatorial Indian Ocean from days −20 to −15. The corresponding height anomalies in the west are, however, off the equator. The height and wind anomalies exhibit characteristics similar to the theoretic Rossby–Kelvin wave packet (Matsuno 1966; Webster 1972; Gill 1980). However, the situation is more complicated than the idealized case, especially when both heating and cooling exist simultaneously. For example, the easterly anomaly tends to lead the height anomaly. The eastern end of this easterly anomaly is located near the deep cooling anomaly in the Maritime Continent (e.g., day −15) and is likely to be part of the Rossby wave response to the tropical diabatic cooling. This explanation is consistent with Seo and Kim (2003) in which they showed that the equatorial wind anomaly is the combination of the Kelvin and Rossby wave responses to a heating–cooling dipole in the equator.

Interestingly, the easterly anomaly does not end right beneath the deep heating. Instead, it extends to the eastern edge of the deep heating anomaly (day −10 in Fig. 4). A careful examination of Fig. 5 from days −20 to −10 reveals that the easterly anomaly tends to be located between the negative and positive height anomalies. This suggests that the westward pressure gradient maintains the easterly anomaly and explains why the easterly anomaly extends farther eastward than the deep cooling.

The moisture convergence in the Maritime Continent is not well developed until day −10 when the deep cooling propagates away from the Maritime Continent into the western Pacific. At this time, the easterly anomaly prevails across the mountainous islands in the Maritime Continent, and two centers of anomalous moisture convergence and shallow heating appear on the windward side of the mountain ranges in Borneo and New Guinea. It is likely that the lifting and frictional effects of mountainous Borneo and New Guinea induce the near-surface moisture convergence in the Maritime Continent. The other possibility might be the frictional convergence associated with the equatorial Kelvin wave as proposed in previous studies. However, there is little evidence showing that the eastward-propagating moisture convergence is associated with the negative 1000-hPa height anomaly near the equatorial Indian Ocean, for example, days −20 to −10 in Fig. 4. Instead, the moisture convergence develops in situ in the Maritime Continent. On the other hand, the arrival of this Kelvin wave–type perturbation around day −15 might enhance the easterly anomaly and lower the surface pressure in the mountainous Maritime Continent and indirectly strengthen the near-surface moisture convergence.

After the initiation of the deep heating anomaly in the Maritime Continent (e.g., day 0), the westerly anomaly originally penetrating to the central Indian Ocean (e.g., at day −5) extends farther eastward to the Maritime Continent, probably induced by the strengthening moisture convergence in the Maritime Continent. Consequently, the moisture convergence in the Indian Ocean weakens and the convergence in the Maritime Continent strengthens. The major center of the moisture convergence therefore shifts eastward. This eastward shift might lead to a sudden dissipation of the deep heating anomaly in the Indian Ocean between days −5 and 0. The moisture convergence in the Maritime Continent starts weakening at day 5 when the deep heating anomaly moves eastward into the western Pacific warm pool and the easterly anomaly is replaced by the westerly anomaly. Consequently, the heating anomalies in the Maritime Continent dissipate quickly afterward. The moisture convergence in the lower troposphere continues propagating to the eastern Pacific, while the heating anomaly dissipates near the date line.

b. Deep convection development in tropical America

Another dramatic eastward-propagation event in the negative 1000-hPa height and easterly anomalies occurs between days −5 and 0 during the development of a deep heating anomaly in the Maritime Continent (Figs. 5d and 5e). The leading edge of the height anomaly travels across the Pacific in about 5 days, which is equivalent to the phase speed of a dry Kelvin wave, as will be shown later. This result is consistent with the finding by Milliff and Madden (1996), in which they found fast-propagating signals in surface pressure that took only 4 days to propagate from Kanton Island (2.8°S, 171.7°W) to Balboa (9.0°N, 79.6°W). The anomalous circulation discussed above exhibits characteristics similar to the Rossby–Kelvin wave packet. There is no indication of eastward-propagating moisture convergence associated with the negative height anomaly (e.g., Figs. 3 and 4).

This eastward propagation might have to do with a fully developed deep heating anomaly in the Maritime Continent, the dissipation of the diabatic cooling anomaly in the central Pacific, and the strengthening diabatic cooling anomaly in tropical South America. The dramatic changes in the first two heating anomalies favors the Kelvin wave–like feature to propagate quickly across the eastern Pacific. The diabatic cooling anomaly in tropical South America favors the development of the easterly winds associated with the Rossby wave–like feature forced by tropical forcing. Both effects favor the quick development of the easterly anomaly across the Pacific basin. Interestingly, a region of near-surface moisture convergence appears on the windward (eastern) side of the Andes where the easterly anomaly prevails (Figs. 3 and 4e). This anomalous moisture convergence moves slowly eastward to eastern South America where the deep heating anomaly appears at day 20 (Figs. 3 and 4i). This deep heating anomaly and moisture convergence exist for more than 10 days. It weakens dramatically 15 days later (e.g., at day −10, Fig. 5c) when the westerly anomaly spreads quickly across the eastern Pacific to South America and the moisture convergence is replaced by divergence.

c. Deep convection development in the western Indian Ocean

Near-surface moisture convergence and a shallow heating anomaly are evident off the eastern African coast 20 days (e.g., day 15, Fig. 4h) before the onset of the deep heating anomaly in the central Indian Ocean (e.g., day −10, Fig. 4c). This moisture convergence and shallow heating anomaly moves eastward to the eastern Indian Ocean where the deep heating anomaly flares at day −10 over the warm SST. It is interesting to note that the moisture convergence and shallow heating also occur in the prevailing easterly anomaly. This situation is similar to that observed in the Maritime Continent and tropical South America.

It is suggested that the moisture convergence and shallow heating might be induced by the lifting effect of the western African highland. In that case, the initiation and maintenance of the easterly anomaly is crucial. The first appearance of the easterly anomaly coincides with the development of the deep cooling anomaly in the central Indian Ocean at day 15. This easterly anomaly is associated with a pair of off-equator positive height anomalies in the western Indian Ocean, which resemble a Rossby wave (e.g., days 10 and 15 in Fig. 5).

The heating anomaly in western Africa starts developing after the onset of the easterly anomaly in the western Indian Ocean at day 15. This development coincides with the arrival of the negative height anomaly in the lower troposphere (e.g., day 15, Fig. 5h) and the in situ development of the near-surface moisture convergence (e.g., day 12, Fig. 3) in western Africa. This deep heating anomaly in western Africa might be able to force a Kelvin wave–like response crossing the Indian Ocean between days −20 and −15 (Fig. 5) and, therefore, enhance the easterly anomaly in the eastern Indian Ocean and the Maritime Continent.

d. Equatorial Kelvin wave and topography

Several features presented above exhibit the characteristics of equatorial Kelvin waves. Figure 6, which presents a cross section of the height anomaly averaged from 10°N to 10°S, reveals the vertical structure of these features. The height anomaly, which tends to change signs in the middle troposphere, exhibits a vertical structure similar to the first internal mode vertical structure found in theoretical studies (e.g., Gill 1980; Wang and Chen 1989) and reported in empirical studies (e.g., Mathews 2000). The sign reversal is most evident in the regions where the deep heating or cooling anomalies are located, for example, eastern Africa, the central Indian Ocean, the Maritime Continent, the western Pacific, and tropical South America. This indicates that the sign reversal can largely be attributed to the tropical atmospheric response to the deep heating or cooling anomalies.

Three spells of eastward propagation are observed in the lower troposphere. These occur mostly in the open oceans, namely, the Indian Ocean, the Pacific, and the Atlantic. Conversely, the eastward propagation is much weaker in the upper troposphere where the perturbation tends to be stationary. The most persistent propagation in the upper troposphere occurs from the eastern Indian Ocean to the western Pacific. However, both the speed and zonal scale are much smaller than the counterparts are in the lower troposphere. A comparison between the evolution of the wind anomaly in the upper and lower troposphere in Fig. 4 reveals a similar contrast between the lower and upper troposphere. This contrast is probably due to interruption by the extratropical waves, which are more likely to propagate into the Tropics in the upper troposphere (e.g., Hsu and Lin 1992; Hsu 1996; Kiladis and Weickmann 1992; Mathews 2000).

These propagating features in the lower troposphere move at such a fast speed that they often cross the oceans in less than 5 days. For example, the leading edge of the negative height anomaly in the western Pacific at day −5 propagates from 165°E to 120°W in about 5 days (e.g., days −5 and 0 in Figs. 5 and 6). The propagation speed is about 22 m s−1, which is equivalent to the phase speed of a dry Kelvin wave. Similar features also cross the Atlantic (days 10 to 15) and the Indian Ocean (days −20 to −15) in less than 5 days. Between the three eastward-propagation events there is always a period of little eastward propagation, for example, days −15 to −5 in the Maritime Continent, days 15 to −20 in Africa, and days 5 to 10 in tropical South America. These periods of stalling coincide with the development of new deep convection in the tropical mountainous areas. A new event of eastward propagation is observed once the deep convection is fully developed and induces a Kelvin wave–like perturbation moving eastward mostly in the lower troposphere. This “on and off” situation indicates that it is not the same Kelvin wave traveling around the globe. Instead, a new wave tends to be generated by the newly developed heating near the downstream topography when the old one stalls and dissipates. Milliff and Madden (1996) and Seo and Kim (2003) also reported the initiation of a fast-propagating Kelvin wave in the Maritime Continent.

The eastward propagation is often interrupted by the topography. The negative 1000-hPa height anomaly, originated in the western Pacific, propagates across the Pacific and stalls near the Andes between days 0 and 10, while the easterly anomaly continuously moves eastward to the Atlantic at day 10. This discontinuity in the height field was also noted in previous studies (e.g., Mathews 2000; Seo and Kim 2003). Interestingly, during this period, the negative height anomalies continue propagating northward and southward along the western coasts of North and South America (Fig. 7 and days 5 and 10 in Fig. 5). Simultaneously, a narrow negative height anomaly propagates northward along the Andes as shown in Fig. 7. The propagation of this feature stalls at the northern tip of South American where the Andes ends. The common feature for these three coastlines is the coastal mountain ranges (Fig. 1). The propagation occurs with the terrain at the right and left in the Northern and Southern Hemisphere, respectively. These features exhibit characteristics similar to those of the coastal Kelvin wave (Gill 1977).

5. Summary and discussion

This study investigated the relationship between the deep convection (and heating anomaly) in the MJO and the topography. The eastward propagation of the deep heating anomalies is confined to two regions: the Indian Ocean and the western Pacific warm pool. Superimposed on the eastward propagation is a series of quasi-stationary deep heating anomalies in the central Indian Ocean, the Maritime Continent, tropical South America, and tropical Africa. This study found that the major deep heating anomalies tend to be quasi-stationary and do not propagate eastward associated with the equatorial waves. Instead, they occur sequentially and discretely downstream in a leapfrog manner. The sequential downstream development of these deep heating anomalies carries the MJO signals continuously eastward. This explains why the OLR anomaly tends to stall at the date line while the velocity potential continues propagating eastward, as discussed in the introduction.

In these four regions, the near-surface moisture convergence and shallow heating anomalies always occur preceding the onset of the deep heating anomalies. Interestingly, in all four regions, except the central Indian Ocean, the convergence and shallow heating anomalies occur on the windward side of mountain ranges under the prevailing easterly anomaly. It is proposed that the lifting and frictional effects of the topography induce the near-surface moisture convergence and shallow heating anomalies, which tend to occur initially underneath the cooling anomaly and anomalous subsidence. Upon the dissipation of the cooling and subsidence anomaly, the deep heating anomaly flares.

The existence of the easterly anomaly is one of the important preconditions for the development of the deep heating anomaly. The easterly anomaly is often located to the east of the deep heating anomaly and to the west of the deep cooling anomaly, as shown above and observed in previous studies (e.g., Seo and Kim 2003). According to the theoretical studies (e.g.,Webster 1972; Gill 1980), an equatorial heating anomaly can force the near-surface easterly anomaly (i.e., the Kelvin wave) to the east while an equatorial cooling can force the near-surface easterly anomaly (i.e., the Rossby wave) to the west. Since the deep heating and cooling anomalies tend to appear in pairs, a strong easterly anomaly can be maintained between a heating–cooling pair with the heating anomaly in the west and the cooling anomaly in the east. If the easterly anomaly prevails over the mountain ranges located between the heating–cooling pair, the near-surface moisture convergence and shallow heating anomaly would occur on the eastern (and windward) side of the mountain ranges and precondition the lower troposphere for the later development of the new deep heating anomaly. The deep heating anomaly originally located in the west would dissipate quickly with the maximum near-surface convergence located to its east on the windward side of the mountain ranges. This results in the eastward shift in the deep convection from one region to another. A schematic diagram shown in Fig. 8 illustrates this hypothesized process, which is proposed to explain the observed shift in the deep convection from the Indian Ocean to the Maritime Continent (Hsu et al. 1990; Weickmann and Khalsa 1990), and the shift from the date line to tropical South America.

The same argument can also help explain the tendency for the eastward-propagating MJO initiated in the western Indian Ocean. Before the initiation, the heating anomalies in tropical South America and Africa and the cooling anomaly in the Indian Ocean form a heating–cooling pair accompanied by the easterly anomaly prevailing in tropical Africa and the western Indian Ocean. The moisture convergence and shallow heating anomalies, which are induced on the windward side of the eastern African highland, might initiate the development of the deep heating anomaly. The heating anomaly subsequently propagates eastward along with the Kelvin–Rossby wave packet and strengthens in the central and eastern Indian Ocean when entering the warmer SST region (Fig. 1) as suggested in theoretical studies (e.g., Salby et al. 1994; Wang and Xie 1998).

The equatorial Kelvin waves excited by the deep heating anomalies seem to play an important role in the sequential downstream MJO development. The Kelvin wave signals, which are mostly evident in the lower troposphere, propagate across the ocean basins in just a few days to help establish the easterly anomaly and reduce the pressure in the lower troposphere near the downstream mountain ranges. Although the Kelvin waves tend to stall on the western side of the mountain ranges (e.g., the Andes), the new heating anomaly subsequently developed on the eastern side of the mountain ranges might trigger a new Kelvin wave that carries the MJO signal continuously eastward.

The results presented here suggest that the propagation of the MJO is not a continuous eastward propagation around the globe. Instead, it is a combination of three fast eastward-propagating spells across three ocean basins and the quasi-stationary features occurring near the tropical mountainous land areas. The eastward propagation, which is most evident in circulation and height fields, tends to stall near these mountainous areas where a new convection develops and induces the next spell of eastward propagation. The sequential downstream development of these fast propagation and slow quasi-stationary features results in the eastward propagation of the MJO around the globe. It is suggestive that the fast eastward propagation and the slow development of quasi-stationary convection together determine the quasi-periodicity of the MJO.

Theoretical studies found that the MJO signal is amplified through the conversion of eddy potential energy and slows down because of the reduced static stability in the high moisture content and SST regions. While the theoretical results can be used to explain the MJO amplification and slow propagation in the warm and moist areas, they cannot explain the existence of the stationary features appearing in specific regions where the topography and landmass exist. The collocation of these stationary features and the topography suggests that the topographic forcing and the land–sea contrast effect might be a plausible mechanism worthy of consideration. Theoretical and numerical studies are definitely needed to evaluate the plausibility of this conjecture.

Acknowledgments

The authors appreciate the valuable comments of two anonymous reviewers. This study was supported by the National Science Council, Taiwan, under Grants NSC 92-2111-M-002-016-AP4 and NSC 93-2111-M-002-004-AP4.

REFERENCES

  • Bladé, I., and D. L. Hartmann, 1993: Tropical intraseasonal oscillation in a simple nonlinear model. J. Atmos. Sci., 50 , 29222939.

  • Chang, C-P., and H. Lim, 1988: Kelvin wave-CISK: A possible mechanism for the 30–50-day oscillations. J. Atmos. Sci., 45 , 17091719.

    • Search Google Scholar
    • Export Citation
  • Gibson, J. K., P. Kallberg, S. Uppala, A. Hernandez, A. Nomura, and E. Serrano, 1997: ERA description. ECMWF Re-analysis Project Report Series 1, European Centre for Medium-Range Weather Forecasts, Reading, United Kingdom, 72 pp.

  • Gill, A. E., 1977: Coastally trapped waves in the atmosphere. Quart. J. Roy. Meteor. Soc., 103 , 431440.

  • Gill, A. E., 1980: Some simple solutions for heat-induced tropical circulation. Quart. J. Roy. Meteor. Soc., 106 , 447462.

  • Hamming, R. W., 1989: Digital Filters. Prentice-Hall International, 284 pp.

  • Hendon, H. H., and M. L. Salby, 1994: The life cycle of the Madden–Julian oscillation. J. Atmos. Sci., 51 , 22252237.

  • Hsu, H. H., 1996: Global view of the intraseasonal oscillation during northern winter. J. Climate, 9 , 23862406.

  • Hsu, H. H., and S-H. Lin, 1992: Global telconnections in the 250-hPa stream function field during the Northern Hemisphere winter. Mon. Wea. Rev., 120 , 11691190.

    • Search Google Scholar
    • Export Citation
  • 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 , 823839.

    • Search Google Scholar
    • Export Citation
  • Hsu, H. H., C-H. Weng, and C-H. Wu, 2004: Contrasting characteristics between the northward and eastward propagation of the intraseasonal oscillation during the boreal summer. J. Climate, 17 , 727743.

    • Search Google Scholar
    • Export Citation
  • Hu, Q., and D. R. Randall, 1994: Low-frequency oscillations in radiative–convective systems. J. Atmos. Sci., 51 , 307319.

  • Kaylor, R. E., 1977: Filtering and decimation of digital time series. Institute of Physical Science Technology Tech. Note BN 850, University of Maryland, College Park, 42 pp.

  • Khalsa, S. J. S., and E. J. Steiner, 1988: A TOVS dataset for study of the tropical atmosphere. J. Appl. Meteor., 27 , 851862.

  • Kiladis, T. R., and K. M. Weickmann, 1992: Extratropical forcing of tropical Pacific convection during northern winter. Mon. Wea. Rev., 120 , 19241938.

    • Search Google Scholar
    • Export Citation
  • Knutson, T. R., and K. M. Weickmann, 1987: The 30–60-day atmospheric oscillations: Composite life cycles of convection and circulation anomalies. Mon. Wea. Rev., 115 , 14071436.

    • Search Google Scholar
    • Export Citation
  • Lau, K. M., and L. Peng, 1987: Origin of low-frequency (intraseasonal) oscillations in the tropical atmosphere. Part I: Basic theory. J. Atmos. Sci., 44 , 950972.

    • Search Google Scholar
    • Export Citation
  • Liebmann, B., and C. A. Smith, 1996: Description of a complete (interpolated) outgoing longwave radiation dataset. Bull. Amer. Meteor. Soc., 77 , 12751277.

    • Search Google Scholar
    • Export Citation
  • Maloney, E. D., and D. L. Hartmann, 1998: Frictional moisture convergence in a composite life cycle of the Madden–Julian oscillation. J. Climate, 11 , 23872403.

    • Search Google Scholar
    • Export Citation
  • Mathews, A. J., 2000: Propagation mechanism for Madden–Julian oscillation. Quart. J. Roy. Meteor. Soc., 126 , 26372651.

  • Matsuno, T., 1966: Quasi-geostrophic motions in the equatorial area. J. Meteor. Soc. Japan, 44 , 2542.

  • 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 , 586597.

    • Search Google Scholar
    • Export Citation
  • Salby, M. L., R. Garcia, and H. H. Hendon, 1994: Planetary-scale circulation in the presence of climatological and wave-induced heating. J. Atmos. Sci., 51 , 23442367.

    • Search Google Scholar
    • Export Citation
  • Sardeshmukh, P. D., and B. J. Hoskins, 1984: Spatial smoothing on the sphere. Mon. Wea. Rev., 112 , 25242529.

  • Seo, K. H., and K. Y. Kim, 2003: Propagation and initiation mechanisms of the Madden–Julian oscillation. J. Geophys. Res., 108 , D13. 43844405.

    • Search Google Scholar
    • Export Citation
  • Sui, C. H., and K. M. Lau, 1989: Origin of low-frequency (intraseasonal) oscillations in the tropical atmosphere. Part II: Structure and propagation of mobile wave-CISK modes and their modification by lower boundary forcings. J. Atmos. Sci., 46 , 3756.

    • Search Google Scholar
    • Export Citation
  • Wang, B., and J. Chen, 1989: On the zonal-scale selection and vertical structure of equatorial intraseasonal waves. Quart. J. Roy. Meteor. Soc., 115 , 13011323.

    • Search Google Scholar
    • Export Citation
  • Wang, B., and H. Rui, 1990a: Synoptic climatology of transient tropical intraseasonal convection anomalies: 1975–1985. Meteor. Atmos. Phys., 44 , 4361.

    • Search Google Scholar
    • Export Citation
  • Wang, B., and H. Rui, 1990b: Dynamics of the coupled moist Kelvin–Rossby wave on an equatorial β–plane. J. Atmos. Sci., 47 , 397413.

    • Search Google Scholar
    • Export Citation
  • Wang, B., and X. Xie, 1998: Coupled modes of the warm pool climate system. Part I: The role of air–sea interaction in maintaining Madden–Julian oscillation. J. Climate, 11 , 21162135.

    • Search Google Scholar
    • Export Citation
  • Webster, P., 1972: Response of the tropical atmosphere to local, steady forcing. Mon. Wea. Rev., 100 , 517540.

  • Weickmann, K. M., and S. J. S. Khalsa, 1990: The shift of convection from the Indian Ocean to the western Pacific Ocean during a 30–60 day oscillation. Mon. Wea. Rev., 118 , 964978.

    • Search Google Scholar
    • Export Citation
  • Zhang, C., and H. H. Hendon, 1997: Propagating and standing components of the intraseasonal oscillation in tropical convection. J. Atmos. Sci., 54 , 741752.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Terrain height in tropical Africa, the Maritime Continent, and tropical South America. The plotted contours are the same as the shading shown in the figure (units in m).

Citation: Journal of Climate 18, 6; 10.1175/JCLI-3292.1

Fig. 2.
Fig. 2.

(a) The 30–60-day filtered OLR variance plotted on top of the climatologically mean SST. Variances larger than 100 W2 m−4 are plotted in shading for every 100 W2 m−4. The four contours for the SST are 27°, 28.5°, 29°, and 29.5°C with the temperature increasing toward the center. (b) The propagation tendency of the OLR anomaly derived from the 5-day lagged one-point correlation maps. The vector represents the speed and direction of an OLR anomaly propagating from the base point to the region where the lagged correlation coefficient is maximum 5 days later. The length of the reference vector shown in the lower-right corner is equivalent to 5 m s−1. The contours represent those lagged correlation coefficients larger than 0.75. Contour interval is 0.05.

Citation: Journal of Climate 18, 6; 10.1175/JCLI-3292.1

Fig. 3.
Fig. 3.

Hovmöller diagram of the lagged regression coefficients between the first principal component of the 200-hPa velocity potential and the OLR and 925-hPa moisture divergence anomalies by pasting together the equatorial strip plots from days −22 to 22 for every 2 days. The OLR and moisture divergence anomalies are shown in colors and contours, respectively. The contour intervals for the OLR and moisture divergence are 2 W m−2 and 10−9 g Kg−1 s−1, respectively. Dashed and solid lines denote negative and positive moisture divergence anomalies, respectively. Regression coefficients have been multiplied by one std dev of the principal component.

Citation: Journal of Climate 18, 6; 10.1175/JCLI-3292.1

Fig. 4.
Fig. 4.

Zonal cross sections of the lagged regression coefficients between the first principal component of the 200-hPa velocity potential and the diabatic heating (heavy contours), zonal wind, pressure velocity (multiplied by 100), and moisture divergence (shaded and light contours) averaged from 10°S to 5°N at days (a) −20, (b) −15, (c) −10, (d) −5, (e) 0, (f) 5, (g) 10, (h) 15, and (i) 20. Contour intervals are 1 × 10−6 K s−1 and 10−9 g Kg−1 s−1 for the diabatic heating and moisture divergence, respectively. Length of the reference arrow is equivalent to 2 m s−1 or 0.02 hPa s−1. Solid and dashed lines indicate positive and negative values, respectively. Regression coefficients have been multiplied by one std dev of the principal component and only those that are significant at the 0.05 level are plotted. The lower portion of each plot is stretched linearly to emphasize the near-surface features.

Citation: Journal of Climate 18, 6; 10.1175/JCLI-3292.1

Fig. 5.
Fig. 5.

Lagged regression coefficients between the first principal component of the 200-hPa velocity potential and the 1000-hPa height and wind anomalies at days (a) −20, (b) −15, (c) −10, (d) −5, (e) 0, (f) 5, (g) 10, (h) 15, and (i) 20. Height anomalies larger than 1 m and less than −1 m are contoured in solid (dark) and dashed (light) lines (shading), respectively, and the contour interval is 1 m. Unit vector equivalent to 1 m s−1 is shown in the lower-right corner. Regression coefficients have been multiplied by one std dev of the principal component and only those that are significant at the 0.05 level are plotted.

Citation: Journal of Climate 18, 6; 10.1175/JCLI-3292.1

Fig. 6.
Fig. 6.

Zonal cross sections of the lagged regression coefficients between the first principal component of the 200-hPa velocity potential and the height anomaly averaged from 10°S to 10°N at days (a) −20, (b) −15, (c) −10, (d) −5, (e) 0, (f) 5, (g) 10, (h) 15, and (i) 20. Contour interval is 1 m. Solid and dashed lines indicate positive and negative values, respectively. Regression coefficients have been multiplied by one std dev of the first principal component. Shading denotes the area where the anomalies are significant at the 0.05 level.

Citation: Journal of Climate 18, 6; 10.1175/JCLI-3292.1

Fig. 7.
Fig. 7.

Lagged regression coefficients between the first principal component of the 200-hPa velocity potential and the 1000-hPa height near the American continent at days (a) 0, (b) 5, and (c) 10. Height anomalies larger than 1 m and less than −1 m are contoured in solid and dashed lines, respectively. The contour interval is 1 m. Regression coefficients have been multiplied by one std dev of the principal component and only those that are significant at the 0.05 level are plotted.

Citation: Journal of Climate 18, 6; 10.1175/JCLI-3292.1

Fig. 8.
Fig. 8.

Schematic diagram illustrating the topographic and land–sea contrast effect on the eastward shift in the deep convection associated with the MJO along the equator. This diagram only represents the two-dimensional vertical structure of the three-dimensional MJO along the equator. Therefore, the sketch shown here does not necessarily form a closed circulation in certain regions. Arrows indicate the direction of flow in the upper and lower troposphere and subsidence. The cloudlike symbols denote the enhanced convection. Dashed arrows represent the flows that exhibit a general direction but are not necessarily continuous. Question marks denote the flows that do not show consistent direction.

Citation: Journal of Climate 18, 6; 10.1175/JCLI-3292.1

Save
  • Bladé, I., and D. L. Hartmann, 1993: Tropical intraseasonal oscillation in a simple nonlinear model. J. Atmos. Sci., 50 , 29222939.

  • Chang, C-P., and H. Lim, 1988: Kelvin wave-CISK: A possible mechanism for the 30–50-day oscillations. J. Atmos. Sci., 45 , 17091719.

    • Search Google Scholar
    • Export Citation
  • Gibson, J. K., P. Kallberg, S. Uppala, A. Hernandez, A. Nomura, and E. Serrano, 1997: ERA description. ECMWF Re-analysis Project Report Series 1, European Centre for Medium-Range Weather Forecasts, Reading, United Kingdom, 72 pp.

  • Gill, A. E., 1977: Coastally trapped waves in the atmosphere. Quart. J. Roy. Meteor. Soc., 103 , 431440.

  • Gill, A. E., 1980: Some simple solutions for heat-induced tropical circulation. Quart. J. Roy. Meteor. Soc., 106 , 447462.

  • Hamming, R. W., 1989: Digital Filters. Prentice-Hall International, 284 pp.

  • Hendon, H. H., and M. L. Salby, 1994: The life cycle of the Madden–Julian oscillation. J. Atmos. Sci., 51 , 22252237.

  • Hsu, H. H., 1996: Global view of the intraseasonal oscillation during northern winter. J. Climate, 9 , 23862406.

  • Hsu, H. H., and S-H. Lin, 1992: Global telconnections in the 250-hPa stream function field during the Northern Hemisphere winter. Mon. Wea. Rev., 120 , 11691190.

    • Search Google Scholar
    • Export Citation
  • 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 , 823839.

    • Search Google Scholar
    • Export Citation
  • Hsu, H. H., C-H. Weng, and C-H. Wu, 2004: Contrasting characteristics between the northward and eastward propagation of the intraseasonal oscillation during the boreal summer. J. Climate, 17 , 727743.

    • Search Google Scholar
    • Export Citation
  • Hu, Q., and D. R. Randall, 1994: Low-frequency oscillations in radiative–convective systems. J. Atmos. Sci., 51 , 307319.

  • Kaylor, R. E., 1977: Filtering and decimation of digital time series. Institute of Physical Science Technology Tech. Note BN 850, University of Maryland, College Park, 42 pp.

  • Khalsa, S. J. S., and E. J. Steiner, 1988: A TOVS dataset for study of the tropical atmosphere. J. Appl. Meteor., 27 , 851862.

  • Kiladis, T. R., and K. M. Weickmann, 1992: Extratropical forcing of tropical Pacific convection during northern winter. Mon. Wea. Rev., 120 , 19241938.

    • Search Google Scholar
    • Export Citation
  • Knutson, T. R., and K. M. Weickmann, 1987: The 30–60-day atmospheric oscillations: Composite life cycles of convection and circulation anomalies. Mon. Wea. Rev., 115 , 14071436.

    • Search Google Scholar
    • Export Citation
  • Lau, K. M., and L. Peng, 1987: Origin of low-frequency (intraseasonal) oscillations in the tropical atmosphere. Part I: Basic theory. J. Atmos. Sci., 44 , 950972.

    • Search Google Scholar
    • Export Citation
  • Liebmann, B., and C. A. Smith, 1996: Description of a complete (interpolated) outgoing longwave radiation dataset. Bull. Amer. Meteor. Soc., 77 , 12751277.

    • Search Google Scholar
    • Export Citation
  • Maloney, E. D., and D. L. Hartmann, 1998: Frictional moisture convergence in a composite life cycle of the Madden–Julian oscillation. J. Climate, 11 , 23872403.

    • Search Google Scholar
    • Export Citation
  • Mathews, A. J., 2000: Propagation mechanism for Madden–Julian oscillation. Quart. J. Roy. Meteor. Soc., 126 , 26372651.

  • Matsuno, T., 1966: Quasi-geostrophic motions in the equatorial area. J. Meteor. Soc. Japan, 44 , 2542.

  • 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 , 586597.

    • Search Google Scholar
    • Export Citation
  • Salby, M. L., R. Garcia, and H. H. Hendon, 1994: Planetary-scale circulation in the presence of climatological and wave-induced heating. J. Atmos. Sci., 51 , 23442367.

    • Search Google Scholar
    • Export Citation
  • Sardeshmukh, P. D., and B. J. Hoskins, 1984: Spatial smoothing on the sphere. Mon. Wea. Rev., 112 , 25242529.

  • Seo, K. H., and K. Y. Kim, 2003: Propagation and initiation mechanisms of the Madden–Julian oscillation. J. Geophys. Res., 108 , D13. 43844405.

    • Search Google Scholar
    • Export Citation
  • Sui, C. H., and K. M. Lau, 1989: Origin of low-frequency (intraseasonal) oscillations in the tropical atmosphere. Part II: Structure and propagation of mobile wave-CISK modes and their modification by lower boundary forcings. J. Atmos. Sci., 46 , 3756.

    • Search Google Scholar
    • Export Citation
  • Wang, B., and J. Chen, 1989: On the zonal-scale selection and vertical structure of equatorial intraseasonal waves. Quart. J. Roy. Meteor. Soc., 115 , 13011323.

    • Search Google Scholar
    • Export Citation
  • Wang, B., and H. Rui, 1990a: Synoptic climatology of transient tropical intraseasonal convection anomalies: 1975–1985. Meteor. Atmos. Phys., 44 , 4361.

    • Search Google Scholar
    • Export Citation
  • Wang, B., and H. Rui, 1990b: Dynamics of the coupled moist Kelvin–Rossby wave on an equatorial β–plane. J. Atmos. Sci., 47 , 397413.

    • Search Google Scholar
    • Export Citation
  • Wang, B., and X. Xie, 1998: Coupled modes of the warm pool climate system. Part I: The role of air–sea interaction in maintaining Madden–Julian oscillation. J. Climate, 11 , 21162135.

    • Search Google Scholar
    • Export Citation
  • Webster, P., 1972: Response of the tropical atmosphere to local, steady forcing. Mon. Wea. Rev., 100 , 517540.

  • Weickmann, K. M., and S. J. S. Khalsa, 1990: The shift of convection from the Indian Ocean to the western Pacific Ocean during a 30–60 day oscillation. Mon. Wea. Rev., 118 , 964978.

    • Search Google Scholar
    • Export Citation
  • Zhang, C., and H. H. Hendon, 1997: Propagating and standing components of the intraseasonal oscillation in tropical convection. J. Atmos. Sci., 54 , 741752.

    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    Terrain height in tropical Africa, the Maritime Continent, and tropical South America. The plotted contours are the same as the shading shown in the figure (units in m).

  • Fig. 2.

    (a) The 30–60-day filtered OLR variance plotted on top of the climatologically mean SST. Variances larger than 100 W2 m−4 are plotted in shading for every 100 W2 m−4. The four contours for the SST are 27°, 28.5°, 29°, and 29.5°C with the temperature increasing toward the center. (b) The propagation tendency of the OLR anomaly derived from the 5-day lagged one-point correlation maps. The vector represents the speed and direction of an OLR anomaly propagating from the base point to the region where the lagged correlation coefficient is maximum 5 days later. The length of the reference vector shown in the lower-right corner is equivalent to 5 m s−1. The contours represent those lagged correlation coefficients larger than 0.75. Contour interval is 0.05.

  • Fig. 3.

    Hovmöller diagram of the lagged regression coefficients between the first principal component of the 200-hPa velocity potential and the OLR and 925-hPa moisture divergence anomalies by pasting together the equatorial strip plots from days −22 to 22 for every 2 days. The OLR and moisture divergence anomalies are shown in colors and contours, respectively. The contour intervals for the OLR and moisture divergence are 2 W m−2 and 10−9 g Kg−1 s−1, respectively. Dashed and solid lines denote negative and positive moisture divergence anomalies, respectively. Regression coefficients have been multiplied by one std dev of the principal component.

  • Fig. 4.

    Zonal cross sections of the lagged regression coefficients between the first principal component of the 200-hPa velocity potential and the diabatic heating (heavy contours), zonal wind, pressure velocity (multiplied by 100), and moisture divergence (shaded and light contours) averaged from 10°S to 5°N at days (a) −20, (b) −15, (c) −10, (d) −5, (e) 0, (f) 5, (g) 10, (h) 15, and (i) 20. Contour intervals are 1 × 10−6 K s−1 and 10−9 g Kg−1 s−1 for the diabatic heating and moisture divergence, respectively. Length of the reference arrow is equivalent to 2 m s−1 or 0.02 hPa s−1. Solid and dashed lines indicate positive and negative values, respectively. Regression coefficients have been multiplied by one std dev of the principal component and only those that are significant at the 0.05 level are plotted. The lower portion of each plot is stretched linearly to emphasize the near-surface features.

  • Fig. 5.

    Lagged regression coefficients between the first principal component of the 200-hPa velocity potential and the 1000-hPa height and wind anomalies at days (a) −20, (b) −15, (c) −10, (d) −5, (e) 0, (f) 5, (g) 10, (h) 15, and (i) 20. Height anomalies larger than 1 m and less than −1 m are contoured in solid (dark) and dashed (light) lines (shading), respectively, and the contour interval is 1 m. Unit vector equivalent to 1 m s−1 is shown in the lower-right corner. Regression coefficients have been multiplied by one std dev of the principal component and only those that are significant at the 0.05 level are plotted.

  • Fig. 6.

    Zonal cross sections of the lagged regression coefficients between the first principal component of the 200-hPa velocity potential and the height anomaly averaged from 10°S to 10°N at days (a) −20, (b) −15, (c) −10, (d) −5, (e) 0, (f) 5, (g) 10, (h) 15, and (i) 20. Contour interval is 1 m. Solid and dashed lines indicate positive and negative values, respectively. Regression coefficients have been multiplied by one std dev of the first principal component. Shading denotes the area where the anomalies are significant at the 0.05 level.

  • Fig. 7.

    Lagged regression coefficients between the first principal component of the 200-hPa velocity potential and the 1000-hPa height near the American continent at days (a) 0, (b) 5, and (c) 10. Height anomalies larger than 1 m and less than −1 m are contoured in solid and dashed lines, respectively. The contour interval is 1 m. Regression coefficients have been multiplied by one std dev of the principal component and only those that are significant at the 0.05 level are plotted.

  • Fig. 8.

    Schematic diagram illustrating the topographic and land–sea contrast effect on the eastward shift in the deep convection associated with the MJO along the equator. This diagram only represents the two-dimensional vertical structure of the three-dimensional MJO along the equator. Therefore, the sketch shown here does not necessarily form a closed circulation in certain regions. Arrows indicate the direction of flow in the upper and lower troposphere and subsidence. The cloudlike symbols denote the enhanced convection. Dashed arrows represent the flows that exhibit a general direction but are not necessarily continuous. Question marks denote the flows that do not show consistent direction.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 487 92 8
PDF Downloads 200 44 5