Longitudinal localization of tropical intraseasonal variability


  • Jai Sukhatme

    Corresponding author
    1. Centre for Atmospheric and Oceanic Sciences, Indian Institute of Science, Bangalore, India
    • Centre for Atmospheric and Oceanic Sciences, Indian Institute of Science, Bangalore 560012, India.
    Search for more papers by this author


Intraseasonal time-scales play an important role in tropical variability. Two modes that contribute significantly to tropical intraseasonal variability (ISV) are the eastward-propagating Madden–Julian Oscillation (MJO), and westward-moving moist equatorial Rossby waves. This note reports on a correspondence between the longitudinal gradient of mean tropical precipitable water (PW), and the geographical regions of genesis, and convective activity, of both these large-scale tropical systems. Our finding is based on an analysis of PW from the MERRA reanalysis product. The data indicate that the mean tropical PW has a dominant wavenumber two (three) structure in longitude in the Northern (Southern) Hemisphere. Departures from a longitudinally homogeneous state are attributed to the influence of subtropical anticyclones, and are accentuated by monsoonal regions of both hemispheres. This mean structure results in a sharply localized longitudinal gradient of PW. Remarkably, regions with positive gradients (such as the Northern and Southern Hemisphere western Indian Ocean), i.e. they have larger PW to the east, are the very zones that are implicated in the formation, and show high levels of convective activity, of the eastward-moving MJO. On the other hand, regions with negative gradients (such as the Southern Hemisphere central Pacific) are the very regions where genesis, and maxima in variance, of westward-moving moist equatorial Rossby waves are known to occur. Apart from providing a first-order longitudinal footprint of the convective phase of these systems, this correspondence reinforces the role of the mean climatic state in tropical ISV. Copyright © 2012 Royal Meteorological Society

1. Introduction

Tropical variability is known to consist of a hierarchy of spatio-temporal scales (Takayabu, 1994; Wheeler and Kiladis, 1999). In particular, it has been observed that large-scale intraseasonal time-scales contribute significantly to this variability (Wheeler and Kiladis, 1999). The physical modes involved in tropical intraseasonal variability (ISV) are the Madden–Julian Oscillation (MJO), moist Kelvin waves, and moist equatorial Rossby waves (Wheeler and Kiladis, 1999). In many respects, the MJO stands on its own (Zhang, 2005), while the latter two are members of an extended family of convectively coupled equatorial waves (CCEWs; Kiladis et al., 2009). As moist Kelvin waves also influence shorter time-scales (and smaller length-scales) than the large-scale intraseasonal window, we choose not to include them in the present study. Instead we focus on the MJO and moist equatorial Rossby waves.

In addition to the central dynamical role of water vapour in these large-scale systems, a salient feature is the longitudinal localization of their signals in moisture and deep convection (Madden and Julian, 1971; Weickmann et al., 1985; Zhu and Wang, 1993; Hendon and Salby, 1994; Kayano and Kousky, 1999; Wheeler et al., 2000; Roundy and Frank, 2004). It is this geographical preference in genesis and convective activity exhibited by these two modes that is the subject of this note.

The MJO is an eastward-propagating, coherent, multiscale structure that contributes to tropical ISV in the 30–60 day window (Madden and Julian, 1971). Although some features of the MJO are felt throughout the Tropics, its moist or convective phase shows a clear longitudinal preference (Weickmann et al., 1985; Zhu and Wang, 1993; Madden and Julian, 1994; Hendon and Salby, 1994; Kayano and Kousky, 1999). In the Northern Hemisphere (NH) summer, genesis of MJO convective activity is observed to be north of the Equator in the western Indian Ocean (Madden and Julian, 1971). The moist phase has a broad peak that begins over the western Indian Ocean (often weakening over the Maritime Continent), and terminates in the western Pacific Ocean (Zhu and Wang, 1993; Zhang, 2005). In addition, a secondary site of heightened moist activity is noted over the eastern Pacific Ocean (Kayano and Kousky, 1999; Zhang and Dong, 2004; Roundy and Frank, 2004). During the Southern Hemisphere (SH) summer, moist MJO activity is mainly south of the Equator and its genesis occurs off the eastern African continent in the western Indian Ocean (Weickmann et al., 1985; Zhu and Wang, 1993; Hendon and Salby, 1994; Zhang and Dong, 2004). A second prominent peak in convective activity is observed east of the Australian continent in the Pacific Ocean (Hendon and Salby, 1994; Zhang and Dong, 2004; Zhang, 2005), and terminates in the South Pacific Convergence Zone (Zhu and Wang, 1993; Zhang and Dong, 2004)*

Northwestward-moving equatorial Rossby waves, which affect tropical variability on a quasi-biweekly time-scale, also show a clear longitudinal dependence in their convective phase (Kiladis et al., 2009). In particular, a quasi-biweekly mode is observed during the NH summer in the Indian and western Pacific Oceans (Murakami, 1976; Krishnamurti and Bhalme, 1976; Chen and Chen, 1993). This mode is a manifestation of large-scale moist equatorial Rossby waves that grow in the west Pacific (Hayashi and Sumi, 1986; Zhu and Wang, 1993; Wang and Xie, 1997; Roundy and Frank, 2004), and make their way northwestward during the South Asian monsoon (Chatterjee and Goswami, 2004; Kikuchi and Wang, 2009). Similar northwest-moving convective waves, which originate over continental equatorial and northern Africa, are observed in the West African monsoon during the NH summer season (Janicot et al., 2010). Equatorial Rossby waves have also been documented during the SH summer, and play a part in ISV of the Australian monsoon (Chatterjee and Goswami, 2004; Wheeler and McBride, 2005). In fact, the moist activity of equatorial Rossby waves peaks in the SH central Pacific Ocean, and a signal of these northwestward-moving systems is also seen in the NH west Pacific Ocean (Kiladis and Wheeler, 1995; Wheeler et al., 2000; Roundy and Frank, 2004).

The reason for this geographical preference in convective activity of the MJO and equatorial Rossby waves is not very clear. In the context of the MJO, plausible links, reviewed in Zhang (2005), include: correlations with prevailing sea surface temperatures, dependence on oceanic heat fluxes, the role of air–sea coupling, surface evaporation feedbacks, and interference with regional diurnal cycles. In addition, recent numerical work has drawn attention to the importance of the proper representation of the mean climatic state in connection with genesis of convection in MJOs (Inness et al., 2003). With regard to equatorial Rossby waves, Wang and Xie (1997) have alluded to the role of diminishing water vapour over the central Pacific, possibly resulting in a transition of the moist MJO into a dry Kelvin wave to the east, and moist Rossby waves to the west in an integrated framework for boreal intraseasonal oscillations. Further, the episodic influx of extratropical Rossby waves into the deep SH tropics has been tied to the observed high levels of westward-moving convective equatorial Rossby wave activity over the SH central Pacific Ocean (Matthews and Kiladis, 1999a,b).

In their convective phase, both the MJO and equatorial Rossby waves are moist dynamical systems. In particular, even though we lack a complete understanding of the MJO, explanations of the system require a sufficiently moist environment, and assign a prominent role to condensational heating as derived from the phase change of water vapour (Wang, 2005; Majda and Stechmann, 2009). Similarly, moist feedback is seen to be necessary for the proper scale selection of convective equatorial Rossby waves (Chatterjee and Goswami, 2004).

Exploiting this need for moisture, we examine precipitable water (PW) in the Tropics from the Modern Era Retrospective-analysis for Research and Applications (MERRA) product, for possible insight into the localization of convective activity in these eastward- and westward-moving systems. In particular, daily MERRA data at a spatial resolution of 0.5° latitude ×0.67° longitude spanning the decade 2001–2010 was used in this study. Details and downloading information can be found on the MERRA website, http://gmao.gsfc.nasa.gov/merra

2. Mean precipitable water in the Tropics

The spatial pattern of mean PW, as well as its average over the tropical region (30°N to 30°S), is shown in Figure 1. We see that the mean PW is composed of two distinct envelopes; PW in the first envelope increases from the western Indian Ocean up to the western Pacific, and then systematically decreases over the central and eastern Pacific. The second envelope of PW is of a smaller magnitude and spatial scale; this begins to rise over the east Pacific, through Central America and then decays over the eastern Atlantic and the African continent.

Figure 1.

Mean PW (mm) in the Tropics, as derived from MERRA data, (a) averaged over the period 2001–2010, and (b) from 30°N to 30°S (bold line). In (b) the thin line is an approximation to the mean PW generated by retaining the largest three wavenumbers. This figure is available in colour online at wileyonlinelibrary.com/journal/qj

Figure 2 shows the latitudinally averaged PW in the two hemispheres separately. In the NH (Figure 2(a)), we observe two regions of systematic decay. The first is from the central to eastern Pacific, and the second is from eastern Atlantic through to the North African continent. A similar decay is seen in the SH (Figure 2(b)), where, in addition to the central-eastern Atlantic and Pacific Oceans, we also observe a depletion in PW over the eastern Indian Ocean and western Australian continent. In both hemispheres, the regions of decreasing PW coincide with the descending eastward flanks of oceanic subtropical anticyclones (Chen et al., 2001; Rodwell and Hoskins, 2001; Miyasaka and Nakamura, 2005, 2010). In addition to the descent of tropical air, the import of dry air from the extratropics along the eastern flanks further reduces the PW in these longitudinal zones (Rodwell and Hoskins, 1996). Based on the number of subtropical subsidence zones in each hemisphere (denoted by arrows in Figure 2), we expect the mean PW in the NH (SH) to be dominated by a wavenumber two (three) structure. Indeed, as seen in Figure 2, the reconstructed longitudinal profile using two (three) wavenumbers captures the essential features of the PW in the two hemispheres.

Figure 2.

Mean PW (bold line), averaged over (a) 2001–2010, and (b) 30°N, 30°S to 0°. The thin curve represents an approximation to (a) the mean NH PW generated by retaining the largest two wavenumbers, and (b) the mean SH PW generated from the largest three wavenumbers. This figure is available in colour online at wileyonlinelibrary.com/journal/qj

Apart from the effects of subtropical highs, monsoonal regions increase the PW in a particular hemisphere's summer season. This is seen in Figure 3, which shows the difference in the mean PW during the summer and winter seasons (JFMA and JJAS respectively) of the two hemispheres. As anticipated, the summer PW in a given hemisphere is larger than the corresponding winter PW but, more interestingly, it is the longitudinal inhomogeneity in Figure 3 that confirms the impact of individual summer monsoons. In particular, for the NH, we note that the peaks in Figure 3 correspond to the Asian Monsoon (AsM), the North American Monsoon (NAM) and the North African Monsoon (NAfM), whereas for the SH, the peaks line up with the South African Monsoon (SAfM), the Australian Monsoon (AuM) and the South American Monsoon (SAmM).

Figure 3.

Seasonal differences in mean hemispheric PW. The dark line shows the NH mean PW (JJAS) minus mean PW (JFMA), and the lighter line the SH mean PW (JFMA) minus mean PW (JJAS). The horizontal lines are the mean differences in the two hemispheres. This figure is available in colour online at wileyonlinelibrary.com/journal/qj

Taken together, we see that the large-scale longitudinal structure of mean PW in the Tropics is modulated by the presence of subtropical high pressure zones. These anticyclones cause a systematic regional depletion of PW, resulting in a dominant wavenumber two (three) profile of mean PW in the NH (SH). This basic structure is accentuated by the seasonal increases in PW due to the various tropical monsoon systems. At a more fundamental level, as the monsoon systems play a role in the formation of subtropical highs (Chen et al., 2001; Rodwell and Hoskins, 2001), the two influences are not independent, rather they act in concert to shape the mean tropical PW. We also note that these processes that determine the longitudinal structure of PW are slower (seasonal) and, in a sense, external to the ISV under consideration.

3. Longitudinal gradient of mean tropical precipitable water

With this background, we turn to the longitudinal gradient of mean PW. A two-dimensional picture of the gradient field is presented in Figure 4. This field is constructed by retaining three (four) wavenumbers of the mean PW in the longitudinal (latitudinal) direction. Figures 4(a, b) show the contours of large positive and negative gradients respectively. As is evident, the gradient is highly localized in both hemispheres. Further, in both hemispheres, the regions of maximal eastward-propagating convective MJO activity (namely, the western Indian and Pacific Oceans and east Pacific in the NH, and the western Indian Ocean and the region east of the Australian continent in the SH) are clearly picked out by the positive gradient contours. Similarly, the zones of maximal westward-moving convective equatorial Rossby wave activity (namely, the east Atlantic and western continental Africa in the NH, and the central Pacific in the SH and NH) are demarcated by the negative gradient contours in Figure 4(b). Therefore, we have a remarkable correspondence between the longitudinal gradient of mean PW and the geographical preference in genesis and convective activity, of both the MJO and equatorial Rossby waves.

Figure 4.

Two-dimensional longitudinal gradient (mm/degree) of mean PW. (a) shows positive gradient contours, i.e. regions with larger PW to the east. (b) shows negative gradient contours. This figure is available in colour online at wileyonlinelibrary.com/journal/qj

In addition to their location, the tilt of contours in Figure 4 provides an idea of the meridional displacement of these two systems. Focussing on the SH, which records the largest convective MJO (Weickmann et al., 1985; Zhang and Dong, 2004) and equatorial Rossby wave activity (Wheeler et al., 2000; Roundy and Frank, 2004), we note that the Pacific Ocean has adjoining positive and negative gradient contours. In particular, following steepest descent, eastward propagation through the positive contours entails a southward drift. This is in accord with the record of MJO activity east of the Australian continent into the South Pacific Convergence Zone (Zhu and Wang, 1993). Similarly, as noted in observations of the central Pacific region (Roundy and Frank, 2004), steepest descent in a westward direction across negative contours naturally implies a significant northward component to convective equatorial Rossby wave propagation.

4. Discussion

Gathering these pieces together in Figure 5, we put forth a simple schematic picture of large-scale eastward- and westward-moving moist tropical systems. In essence, the principal longitudinal structure, or deviation from longitudinal homogeneity, of mean PW is determined by the joint influence of subtropical anticyclones and tropical monsoon systems. This structure consists of two (three) envelopes of PW in the NH (SH). One such (idealized) envelope is depicted in Figure 5. Each envelope supports a compact region of large positive (negative) longitudinal gradient of PW on its westward (eastward) side. As the PW envelope is fixed on intraseasonal time-scales, and the change in PW is maximal in regions of high gradients, any trigger (such as a pre-existing system) or perturbation is expected to be most effective at generating condensational heating in these localized high-gradient zones. Quite naturally, the response to this heating consists of a convective or moist (dry) eastward (westward) propagating system on the westward side of the envelope, and vice versa.

Figure 5.

Schematic of large-scale moist tropical systems and mean PW. The solid line shows an envelope of PW, and the dashed line its longitudinal gradient. The two vertical lines depict regions where the gradient attains a maximum. The central ellipse represents an area under the peak PW. The symbols dW, dE, mW and mE stand for dry westward, dry eastward, moist westward and moist eastward propagating systems. This figure is available in colour online at wileyonlinelibrary.com/journal/qj

Therefore, as shown in Figure 5, we expect the genesis, or rejuvenation, of moist eastward (westward) propagating systems in regions of high positive (negative) gradients, localized on the western (eastern) flanks of a PW envelope. These convective systems propagate through the envelope, and begin to decay after crossing the region of peak PW. Further, the region of peak PW represents a middle ground where one expects significant contribution to the variance of a convection proxy (such as outgoing long-wave radiation, OLR) from both eastward- and westward-propagating systems. This is exemplified by the NH western Pacific (around 120°E, Figure 2). Indeed, in addition to the MJO time-scale, the OLR in this region has a notable quasi-biweekly component attributed to the influence of westward-propagating convective Rossby waves (Zhu and Wang, 1993).

Regarding the genesis of eastward-moving convective systems on the westward flank of a PW envelope, we draw attention to the SH east Pacific, including the western coast of South America. As seen in Figure 4, this region supports a large positive PW gradient. In contrast to other zones with positive gradients, this is not a site of MJO genesis or enhanced MJO activity. Instead, it is a region associated with the robust generation of eastward-propagating moist Kelvin waves (Liebmann et al., 2009). This suggests that the gradient is not subtle enough to distinguish between the MJO and a moist Kelvin wave, both of which are eastward-moving systems. In fact, all it specifies is a geographical region of enhanced convective variability, and the associated direction of propagation.

In addition, as highlighted by Inness et al. (2003), the present correspondence also points to the importance of the mean climatic state (in this case, that of PW) in connection with tropical ISV.


The author thanks Professors V. Venugopal, D. Sengupta and R. Murtugudde for discussions. The comments of an anonymous reviewer helped to develop a better feel for the MJO and equatorial Rossby waves, and led to significant improvements in the manuscript.

  • *

    As highlighted by a reviewer, we caution the reader that in isolated cases, MJO convective activity can differ from this broad geographical prescription (e.g. Yu et al., 2011).

  • In the eastern hemisphere, the single (dual) peak in the NH (SH) is a stark feature of moist MJO activity (Zhang and Dong, 2004). As seen, the positive gradient contours pick out this structure with fair accuracy.

  • Why MJOs are not produced or rejuvenated in this region is a difficult question. Some possibilities are (i) the small scale of the PW envelope itself (Figure 1); (ii) The zonal propagation of existing MJOs (as candidates for rejuvenation) over the SH Pacific is likely to be inhibited by the large negative PW gradients in the central Pacific (Figure 4(b)), as well as the tilt of the positive PW gradient contours which indicate a movement towards the extratropics along the South Pacific Convergence Zone (Figure 4(a)); (iii) The landmass and topography immediately to the east of this zone could also possibly be an inhibiting factor in the development of large-scale convective systems; (iv) Also, as pointed out by a reviewer, retaining only three wavenumbers in the longitudinal direction makes the PW gradient appear fairly smooth in this region. Compared with other positive gradient regions, in reality the gradient here is highly localized on the west coast of South America.