Influence of Madden-Julian Oscillation on water budget transported by the Somali low-level jet and the associated Indian summer monsoon rainfall


  • Paulina Ordonez,

    Corresponding author
    1. Departamento de Sistemas Físicos, Quimicos y Naturales, Universidad Pablo de Olavide, Sevilla, Spain
    2. Departamento de Engenharias, Universidade de Tras-os-Montes e Alto Douro, Vila Real, Portugal
    • Corresponding author: P. Ordonez Perez, Universidade de Trás-os-Montes e Alto Douro, Departamento de Engenharias, Quinta de Prados, Apartado 1013, PT-5001-801 Vila Real, Portugal. (

    Search for more papers by this author
  • Pedro Ribera,

    1. Departamento de Sistemas Físicos, Quimicos y Naturales, Universidad Pablo de Olavide, Sevilla, Spain
    Search for more papers by this author
  • David Gallego,

    1. Departamento de Sistemas Físicos, Quimicos y Naturales, Universidad Pablo de Olavide, Sevilla, Spain
    Search for more papers by this author
  • Cristina Pena-Ortiz

    1. Departamento de Sistemas Físicos, Quimicos y Naturales, Universidad Pablo de Olavide, Sevilla, Spain
    Search for more papers by this author


[1] Recent studies suggest that there is a strong linkage between the moisture uptake over the equatorial area of the Somali low level jet (SLLJ) and the rainfall variability over most of continental India. Additionally, the Madden-Julian Oscillation (MJO) strongly modulates the intraseasonal variability of the Indian summer monsoon rainfall, since the northward propagation of the boreal summer MJO is closely associated with the active and break phases of monsoon rainfall. But a question remains open: is there a relationship between the moisture transported by the SLLJ and the MJO evolution? In this paper, a Lagrangian approach is used to track the evaporation minus precipitation (E − P) evolution along trajectories of particles initially situated over the equatorial region of SLLJ. The impact of the MJO on the water budget transport of the SLLJ is examined by making composites of the obtained (E-P) fields for the different MJO phases. The spatial structures of the boreal summer intraseasonal oscillation are revealed in our results, which strongly suggest that the main responsible for the rainfall variability associated to the MJO in these regions are the changes in the moisture advected by the SLLJ. In order to assess the MJO-SLLJ interaction, an analysis of the total-column mass and the total-column specific humidity transported by the SLLJ during the MJO life cycle is performed. While a systematic difference between air mass advected to India during active and break phases of MJO is not detected, changes in the moisture of particles are found, with wet (dry) anomalies over enhanced (suppressed) convection region. This result implicitly leads to assume air-sea interaction processes.

1. Introduction

[2] The Indian summer monsoon (ISM) is a distinctive component of the Asian weather and climate, which brings enormous changes in the hydrological cycle [Gadgil, 2003; Cane, 2010]. Local agriculture and water resources can be greatly affected by the timing and the intensity of the summer monsoon that brings, during some months, most of the total annual rainfall. The most characteristic feature of the intraseasonal variability of the precipitation are prolonged periods of heavy rainfall over the Indian subcontinent (active monsoon conditions) and break spells that are distinguished by cessation/reduction in rainfall over central and western India, with increase over foothills of Himalayas, and northeast India [Krishnamurthy and Shukla, 2000; Joseph et al., 2009; Rajeevan et al., 2010].

[3] The ISM is characterized by a strong south to north cross-equatorial low level jet stream over the western Indian Ocean [Findlater, 1969a, 1969b; Krishnamurthi et al., 1976]. The so-called Findlater Jet or Somali low level jet (SLLJ) originates around the east of Madagascar and reaches mainland Africa near 5°S. After flowing north along the eastern coast of the African continent, it gets into the Arabian Sea at about 10°N. Winds continue across the Arabian Sea and flow over India, bringing with them the monsoon rains. Numerous studies have appeared in the literature addressing the relationship between the changes in the lower-tropospheric monsoon circulation and the intraseasonal variability of rainfall [e.g., Webster et al., 1998; Ramesh Kumar et al., 1999; Krishnan et al., 2000; Goswami et al., 2003; Bhatla et al., 2004]. Different characteristics of the SLLJ associated with the active-break cycle of the ISM rainfall have also been discussed [Joseph and Sijikumar, 2004; Camberlin et al., 2010; Roja Raman et al., 2011]. However, although rainfall is directly related to the moisture supply, only a few works have dealt with the water vapor transported by the SLLJ and its relationship with the ISM precipitation. Recently, Ordóñez et al. [2012] used a Lagrangian dispersion model to perform a dynamical analysis of the water transport toward India founding that the main responsible of the precipitation variability over most of continental India was the water uptake over the equatorial area.

[4] On the other hand, the Madden-Julian Oscillation (MJO) [Madden and Julian, 1971, 1972, 1994] has been identified as a dominant mode of the intraseasonal variability of the ISM rainfall [Hoyos and Webster, 2007; Pai et al., 2011]. The MJO can be characterized by large-scale convective anomalies that develop over the tropical Indian Ocean in a time scale ranging between 30 and 60 days [e.g., Qi et al., 2008; Joseph et al., 2010; Moron et al., 2012] and propagates eastward, equatorially trapped, during winter; but it is characterized by a prominent northward propagation over the south Asian monsoon region during the boreal summer [Yasunari, 1979; Sikka and Gadgil, 1980; Lau and Chan, 1986; Wang and Rui, 1990, among others]. This northward propagation is closely associated with the active and break phases of the ISM rainfall [Lawrence and Webster, 2002], and although it has received intensive attention during past decades [Jiang and Li, 2005], its cause is still an open question [Abhik et al., 2013]. A MJO event is initiated by a slow buildup of moisture in the boundary layer over the Indian Ocean [Kemball-Cook and Weare, 2001; Majda and Stechmann, 2009], vertical transport of low level moisture through shallow convection acts to precondition the atmosphere for deep convection, which, after initiation, is sustained for multiple days [Benedict and Randall, 2007; Jiang et al., 2009]. The MJO propagation involves turbulent fluxes at the air-sea interfaces in the tropical Indian Ocean [Sobel et al., 2010] and a significant modulation of sea surface temperature (SST) [Woolnough et al., 2000; Klingaman et al., 2008; Benedict and Randall, 2011]. Dynamical details pertaining particularly to the initiation mechanism and moisture cycling are unclear [Berkelhammer et al., 2012]. While significant progress has been made in theoretical understanding of this oscillation, a unified theory for the MJO system remains elusive (see Zhang [2005] and Wang [2012] for reviews). General circulation models typically fail to produce a coherent MJO [Berkelhammer et al., 2012]. Accurately simulating the MJO remains a formidable challenge [Lin et al., 2006; Waliser et al., 2009].

[5] Two of the main ideas behind the study are presented in this paper. First, as seen in previous paragraphs, the SLLJ is known to be the main source of moisture that contributes to precipitation events over India. Second, the MJO strongly modulates the ISM rainfall. However, the relationship between moisture transported by the SLLJ and the evolution of the MJO had not been investigated. The main objective of this paper is to examine the impact of the MJO propagation on the water budget transport of the SLLJ and the associated variation in the rainfall over India by using a Lagrangian model. An analysis of the total mass and the specific humidity transported by the SLLJ during different phases of the MJO life cycle is used to assess the MJO-SLLJ interaction.

2. Data and Methods

2.1. Lagrangian Diagnostic of E-P

[6] The method developed by Stohl and James [2004, 2005] is employed in the present study to implement a dynamical analysis of the water vapor transport by using the Lagrangian particle dispersion model FLEXPART [Stohl et al., 1998]. The FLEXPART model is supported by a large number of peer-reviewed publications, and it is well validated and of proven robustness [Gimeno et al., 2013]. A brief outline of the model is presented here, further details of the model can be found in the FLEXPART technical note [Stohl et al., 2005]. At the model start, the atmosphere is homogeneously divided into a large number of air parcels (particles), each representing a fraction of the total atmospheric mass. Then, the particles are allowed to move freely (forwards or backwards in time) with the observed wind, maintaining their mass constant. FLEXPART is driven with meteorological analysis data from the European Centre for Medium-Range Weather Forecasts (ECMWF) [Uppala et al., 2005]. They have a temporal resolution of 6 h (0000, 0600, 1200, and 1800 UTC) and a horizontal resolution of 1° × 1° on 60 levels to calculate the grid-scale advection. Values of specific humidity q are interpolated to the particle positions from the ECMWF analysis grid; both of these (particle positions and q values) are also stored at regular intervals.

[7] The diagnostic method tracks the changes in specific humidity (q) along the path of air parcels, which predominantly reflect the effects of precipitation (p) and evaporation (e) processes,

display math(1)

where m represents the particle's mass. By adding (e-p) for all the particles in the atmospheric column over a given area “A”, the net surface freshwater flux (E-P) is obtained by

display math(2)

where E and P are the evaporation and precipitation rate per unit of area and K is the number of particles residing over the area A. Thus, it is also possible to track (E-P) from any specific region backward or forward in time (see Stohl and James [2004, 2005] for more details). The areas where E – P > 0 have gained moisture in total column average (so the area acts as a moisture source for the atmosphere). On the other hand, there is a predominance of precipitation over evaporation in those areas where E – P < 0 (moisture sink).

[8] FLEXPART was run for June, July, and August (JJA) of a 5 year period (2000–2004).

[9] Although the ISM season typically comprises from June to September, JJA are the most interesting months in our study. During September the relevance of moisture supply from SLLJ declines notably and the recycling process becomes more relevant over most of peninsular India [Ordóñez et al., 2012]. A maximum travel time of 10 days was considered. This is estimated to be the average time that water vapor resides in the atmosphere [Numaguti, 1999]. In this paper, we trace E-P forwards in time from air parcels located within the SLLJ humidity source region for the ISM, thus tracking the fresh water flux transported toward India by this jet. In succession, math formula will designate the total-column moisture budget value for the day “n” (n varying from the first to tenth day) after the particles start the movement over the SLLJ. The addition of math formula to math formula will be indicated as math formula. In this sense, math formula is a measure of the average gain or loss of moisture of the atmosphere between 1 to n days after the air particles left the SLLJ. It must be pointed out that while math formula is not a direct measure of precipitation (it only accounts for the net changes of water vapor of the selected air particles in a given interval), it is profoundly related.

[10] To initialize the algorithm, it is necessary to define the source region of air particles representing the area of water vapor recharge that will be advected by the SLLJ. The extent of this area was recently analyzed by Ordóñez et al. [2012]. Figure 1a shows the selected domain for the present study. It should be noted that the SLLJ, when completely developed, extends to the Indian coast [Ordóñez et al., 2012], but for this study, the northern extreme of the SLLJ domain has not been considered for two reasons: first, Ordóñez et al. [2012] found that the main moisture change affecting the development and intensity of the ISM rainfall were located around SLLJ equatorial portion; second, a source area too large would imply merging air particles traveling to India from 1 (essentially those particles next to the continental coast) up to 7 or more days before reaching the continent (those particles arriving from the southern extreme of the SLLJ). Since the MJO phases are on the daily scale, “mixing” air particles arriving to India in such a wide temporal window would not allow a clear characterization of the water transport variability associated to the MJO. The area selected in Figure 1a limits the temporal range of the particles arriving to India to an average window of 2–6 days, thus allowing the study for a relatively shorter time range.

Figure 1.

(a) Moisture source region defined. (b–h) Seasonal average (JJA) value of number of particles by grid point of 1° × 1° for 1, 2, 3, 4, 5, 6, 10 days forward in time.

[11] Figure 1 helps to understand the dynamics of the tracked particles. The average number of particles over the selected SLLJ domain is around 16,500 at any given day. Figures 1b–1h can be interpreted in terms of the average travel time of the air particles starting over the selected SLLJ. At day 1, much air remains very close to the starting area. The bulk of air particles reach India between days 2 (Figure 1c) and 6 (Figure 1g), in consequence in the next sections, the net freshwater flux from day 2 to 6 math formula will be taken as the basic measure of water vapor transport toward India through the SLLJ. It should be noted that a significant number of particles which at day 0 were over the SLLJ area have also spread into other directions (light gray tones in Figures 1c–1h). By day 10, the particle distribution is rather uniform, but there is still evidence of a greater concentration from the southern Bay of Bengal toward the Indochina Peninsula (Figure 1h).

2.2. Analysis of E-P as a Function of the MJO Phase

[12] The MJO if often defined by the observed eastward migration of the main convection centers over the equatorial Indian Ocean. This oscillation is generally broken up into eight “phases” [Madden and Julian, 1972]. A strong Phase 1 is characterized by the presence of a large negative outgoing longwave radiation (OLR) anomaly around 60°E, while Phase 2 involves convection centered near 80°E and so on. Each progressive phase marks an increasingly eastward location of the convective center. For a typical MJO event, the phases are separated in time by approximately 5–10 days [Jones and Carvalho, 2011].

[13] In the present study, the characterization of the MJO phase has been determined with the MJO index developed by Wheeler and Hendon [2004]. This MJO index is computed as the principal component time series of the two leading empirical orthogonal functions (EOFs) of the combined daily mean fields of 850 hPa and 200 hPa zonal winds and satellite observed OLR averaged over the tropics (15°S-15°N). Projection of the daily observed data onto the multiple-variable EOFs, with the annual cycle and the low-frequency variability associated with ENSO removed, results in two principal component time series that enhances the intraseasonal MJO variability. The pair of principal component series that form the index are called the Real-time Multivariate MJO series 1 (RMM1) and 2 (RMM2). This MJO index has a great advantage for estimating the strength of a MJO event. In this direction, Wheeler and Hendon [2004] suggested a 2-D Phase—Space defined by RMM1 and RMM2. When the amplitude of the event ( math formula) is ≥ 1, the complete MJO event can be categorized as “strong.” In this work, this approach has been adopted. To compute the math formula anomalies representative of the net water vapor changes for each of the MJO phases, the last day of the interval (day 6) was matched with every strong MJO day for the summer monsoon season (JJA). Table 1 shows the number of days used in the analysis of each phase. This method produces an anomaly map for each stage of the MJO that highlights how the (E-P) general characteristics evolve over south Asian monsoon region during a MJO life cycle.

Table 1. Number of Days Under Various MJO Phases Used for Compositing
MJO PhaseNumber of Days of Strong MJO Category

[14]  math formula cannot be interpreted as a measure of precipitation. Moreover, as Figure 1 evidences, the tracked air particles are only part of all the atmospheric particles involved in the generation of the monsoon precipitation (hopefully, the most significant ones). Consequently, it is not expected that the spatial pattern of math formula exactly matches the precipitation climatology over the study area. Nevertheless, if both the period between 2 and 6 days and the SLLJ source region are representative of a significant part of the monsoon precipitation, the spatial patterns of math formula for each of the eight MJO phases and the concurrent precipitation distribution should bear a strong resemblance. To test this, composites of Indian rainfall seasonal anomalies were computed. The Global Precipitation Climatology Project (GPCP) dataset (version 1DD) is used, it provides a global coverage for the 2000–2004 period, on a 1° × 1° grid resolution and a daily basis [Huffman et al., 2001]. This database combines the precipitation information available from several sources; both satellite estimates and gauge observations into a final merged product [Huffman et al., 2007; Vila et al., 2009]. The composite of active and break cycles from GPCP data within the Indian continent were very similar to that obtained from station data [Goswami, 2012].

[15] Then, in this work we investigate the mechanism behind the impact of the MJO life cycle on the SLLJ moisture transport that results in different spatial patterns of math formula. We looked for changes in both the number of particles aimed to the study area and the value of the specific humidity transported by these particles:

[16] 1. According to equation (2), variations in (E-P) in a given area can be caused by a change in the number of particles (K) that contribute to (E-P) in this area. To test this, we traced math formula (n varying from the first to tenth day) forward in time from the air parcels located over the SLLJ, being math formula the total number of particles over each grid cell. Composites of math formula seasonal anomalies for the MJO lifecycle were also computed in order to compare with the spatial patterns of math formula previously obtained.

[17] 2. Another possibility to explain variations of (E-P) patterns over an area are the changes in the (e-p) values of individual particles within this area (equation (2)). Since the mass of the particles remains constant in the model, (e-p) of a particle depends on the specific humidity q carried by the particle (equation (1)). To check this, we trace the total-column specific humidity content math formula from 1 to 10 days after the particles start the movement over the SLLJ. Analogously, composites of math formula seasonal anomalies along different phases of MJO were computed for comparing with the water budgets math formula during the MJO lifecycle.

[18] Finally, significance of composite differences is tested by the bootstrap method [Efron and Tibshirani, 1993], where the original time series is permuted in a nonparametric bootstrap method to evaluate the statistical significance [Wei et al. 2012; Gimeno et al., 2013]. Considering phase 1 as an example, we randomly selected 56 days (see Table 1) within the five monsoon seasons and calculated their difference with the long-term seasonal mean. This process was repeated 1000 times (this is a reasonable number of repetitions, according to Efron and Tibshirani [1993]). To be considered significant, the absolute value of the composite of differences needs to be higher than 90% of the 1000 randomly selected differences.

3. Results

3.1. Relation of (E-P) With the Changes Associated to the MJO

[19] Previous studies [e.g., Annamalai and Sperber, 2005; Hoyos and Webster, 2007; Yang et al., 2008; Pai et al., 2011; Suhas et al., 2013] characterize the MJO evolution during the boreal summer based on deviations in the rainfall or the OLR (which have often been used as proxy for precipitation in areas of deep tropical convection [Wheeler and Weickmann, 2001]). Figure 2 shows the JJA composite of math formula anomalies stratified by the MJO phase. When these composites of fresh water flux anomalies and large-scale anomalous patterns of convection from previous studies are compared (see for example, the Figure 3 in Pai et al. [2011], where an excellent climatology is provided), they both depict a similar structure.

Figure 2.

Composites of (E-P)2–6 during the eight strong MJO phases for the period of JJA (2000–2004). The dotted areas indicate where the differences are significant at 90% confidence level, according to a bootstrap test.

Figure 3.

Composites of daily rainfall during the eight strong MJO phases for the period of JJA (2000–2004). The dotted areas indicate where the differences are significant at 90% confidence level, according to a bootstrap test.

[20] During phase 1, initial stage of the MJO, above normal convection appears throughout the equatorial Indian Ocean (active convection) and convection anomalies of opposite signs are seen over the Inter-Tropical Convergence Zone (ITCZ) (suppressed convection), forming a dipole pattern. In the same way, Figure 2a shows negative fresh water flux (E-P) anomalies over the SLLJ region; therefore, precipitation dominates over evaporation in this area. Positive (E-P) anomalies are seen over the monsoon through region and southeastward over the Bay of Bengal, indicating that in this area the air masses are already gaining moisture compared with the expected seasonal mean.

[21] In the next two phases (phases 2 and 3), the above normal convection over the Indian Ocean gradually builds up and expands eastward, with the zone of maximum convection oriented southeast to northwest, extending from the equator to about 15°N. The suppressed convection over the ITCZ seen during phase 1 shrinks and moves northeastward, but new suppressed convection anomalies appear over equatorial Indian Ocean. For (E-P), during phase 2 (Figure 2b), negative anomalies increase over the Arabian Sea reaching the southern peninsular India. Positive (E-P) anomalies over central India and the Bay of Bengal move northeastward. The equatorial portion of the SLLJ begins to display above normal uptake of moisture. During phase 3 (Figure 3c), the zone of maximum negative (E-P) anomalies are nearly the same as that during phase 2 except for a northeastern shift by about 3–4° over Bay of Bengal and an increase in its magnitude. Cross-equatorial SLLJ reveals its maximum moisture supply.

[22] By phase 4, once the convection hits the maritime continent, it moves off the equatorial region into both hemispheres, being the northern branch much stronger than the southern one. The region of positive rainfall anomalies of the north branch extends from Arabian Sea to the Bay of Bengal, strengthening the eastern part of the monsoon trough over India. Meanwhile, suppressed convection anomalies are seen along most of equatorial Indian Ocean. Accordingly, Figure 2d shows that particles over the northern Arabian Sea and the southern Bay of Bengal are characterized by (E-P) < 0. Some air masses over central India also present an anomalous loss of moisture. In contrast, (E-P) anomalies are positive for air masses arriving to the northern India and Bangladesh as well as for some air masses over equatorial Indian Ocean, indicating that in these regions the air masses coming from SLLJ are already gaining moisture compared with the seasonal average.

[23] In phase 5, the convection dominates the Bay of Bengal and central India, forming a northwest-southeast rainband which gradually moves northeastward during phase 6. It leaves the equatorial Indian Ocean controlled by suppressed convection. Therefore, the dipole in the convective pattern observed during the first three phases of MJO is now reversed. Figure 2e shows that during phase 5 central India displays (E-P) <0 anomaly values, with positive anomalies on either sides (south and north) of this region. Figure 2f illustrates that this entire area of anomalous negative water budget also develops a northward shift during phase 6. In the south (over the equatorial region), a change in the (E-P) anomaly pattern is visible. The situation initiated during phase 1 is also clearly reversed in phase 6 (see Figures 2a and 2f).

[24] During the final two phases of the MJO (phases 7 and 8), the suppressed convection region over the Indian Ocean expands northward and eastward as the monsoon trough becomes weaker. It is known that during these two stages, an anomalous anticyclonic circulation is observed in the lower troposphere over the India subcontinent, resulting in the appearance of northeasterlies over Arabian Sea that weakens the cross-equatorial flow [Pai et al., 2011]. By phase 7 (Figure 2g), positive freshwater flux anomalies observed over the equatorial region in the previous phase, shift to the north cutting off the northwest-southeast (E-P) < 0 band. Only a small area over the northeastern Bay of Bengal shows (E-P) > 0 anomalies. This area is displaced to the Indochinese Peninsula by phase 8 (Figure 2h). Meanwhile, new negative freshwater flux anomalies emerge off the coast of the eastern Africa initiating the onset of the next event.

[25] These results show that anomalous moisture source and sink regions, identified from particles in transit from SLLJ, provide a good representation of large-scale intraseasonal variability in rainfall/OLR fields. It is seen that anomalies of rainfall over the Indian Ocean sector are originated in the interaction of MJO and equatorial area of SLLJ. Therefore, it seems clear that the MJO modulate convection in south Asian monsoon sector through its influence on SLLJ, with different moisture advection in different MJO phases.

3.2. Comparison of Rainfall and (E-P) Over India

[26] In this section, the agreement between intraseasonal variability of ISM rainfall for the study period (2000–2004) and the math formula field over India is assessed. Figure 3 represents the rainfall anomaly (JJA) for the eight different phases of the MJO for the 2000–2004 period. Figure 4 shows a “zoom” of the corresponding math formula values (presented in Figure 2) over India. The color scale of Figure 4 has also been selected to ease the comparison.

Figure 4.

Zoom of (E-P)2–6 of Figure 2 over the same area of Figure 3 during the eight strong MJO phases for the period of JJA (2000–2004). The color scale has been modified to facilitate the comparison of Figures 3 and 4. The dotted areas indicate where the differences are significant at 90% confidence level, according to a bootstrap test.

[27] The general concordance of patterns represented in Figures 3 and 4 is remarkable considering the different methodologies involved. However, some differences in distributions can be identified. The anomaly patterns of the phase 1 (Figures 3a and 4a) resemble the rainfall pattern associated with break monsoon conditions [Rajeevan et al., 2006, 2010; Goswami, 2012] or the absence of low-pressure systems [Krishnamurthy and Ajayamohan, 2010]. During this phase, the main difference between Figures 3.1 and 4.1 is found in northwestern India, which shows opposite sign anomalies. During phase 2, the behavior of rainfall anomalies (Figure 3b) and water budget anomalies (Figure 4b) are again quite similar. In both cases, anomalies present in phase 1 weaken over the monsoon trough zone and strengthen over southern and northeastern India. Again, the northwestern extreme area of India shows an opposite behavior. During phase 3, the comparison of Figures 3c and 4c suggests that our methodology tends to underestimate the moisture loss of the atmosphere over areas of central India and the Arabian Sea, which are characterized by a positive precipitation anomaly (Figure 3c) but show a positive value of E-P (Figure 4c). Figure 3d shows that positive rainfall anomalies spread northward during phase 4 while negative anomalies are only prominent in the western coast. The corresponding (E-P) pattern is similar in the southern half but presents a noticeable difference over the northern sector (Figure 4d).

[28] Phases 5 (Figures 3e and 4e) and 6 (Figures 3f and 4f) are in good agreement. The positive (negative) rainfall (E-P) anomalies extend over most central India in phase 5, showing a northward shift during phase 6 across the monsoon trough zone. Weak negative (positive) rainfall (E-P) anomalies are observed over the rest of the country. The rainfall anomaly pattern of Phase 6 is opposite to that found for Phase 1, and it resembles the rainfall anomaly pattern associated with the active monsoon conditions [Rajeevan et al., 2006, 2010; Goswami, 2012] or the presence of low-pressure systems [Krishnamurthy and Ajayamohan, 2010]. During the final phases (7 and 8), with decreased large-scale convection, negative rainfall anomalies are seen over the Indian subcontinent (Figure 3g and 3h), being the positive precipitation anomalies restricted to some parts of the northeastern India during phase 7. Accordingly, Figures 4g and 4h show that the evaporation dominates over precipitation over most of India.

[29] From the comparison of Figures 3 and 4, it is clear that the computed E-P fields, representing water vapor transport from the SLLJ, are quite similar to the average precipitation over and southward of the typical area of the monsoon trough, suggesting that in these regions, the MJO modulates the rainfall by a direct influence on the SLLJ water transport. In general, (E-P) anomaly field in the northern regions does not provide such a good representation of rainfall, suggesting that northern India rainfall could not be directly modulated by the MJO activity or by its influence over the water fluxes transported by the SLLJ. In this regard, a few studies have addressed the intraseasonal variability of rainfall/convection over northern India. Hartmann and Michelsen [1989] reported that MJO-scale periodicity predominated over most parts of India south of 23°N, whereas spectral peaks around submonthly timescales were limited to northern India where the MJO-scale signal did not dominate. More recently, Fujinami et al. [2011] suggested that submonthly variation in timescale of 10–20 days or quasi-biweekly timescales [Chatterjee and Goswami, 2004] is more common than MJO-scale variation over and around Bangladesh. The modes of these two timescales differ in their spatial and temporal structures; whereas the MJO is associated with northward propagation of convection from the equatorial Indian Ocean, the submonthly scale mode is a westward propagating feature from the western Pacific [e.g., Chen and Chen, 1993; Annamalai and Slingo, 2001; Hoyos and Webster, 2007], thus having a different moisture origin.

3.3. Analysis of Moisture Advection Associated to the SLLJ

[30] The Lagrangian approach adopted in this work allows the estimation of both the number of particles advected (total mass of air) and the total atmospheric specific humidity in the atmospheric column. The analysis of changes in the total number of particles advected during a MJO life cycle math formula (figure not shown) revealed that it does not increase over India during active monsoon phases. In contrast, changes in the moisture of these particles math formula are found, with moist (dry) anomalies over enhanced (suppressed) convection region.

[31] Figure 5 illustrates the average temporal sequence of math formula during the successive phases of the MJO life cycle (2000–2004 JJA average). By phase 1, positive specific humidity (wet) anomalies are observed over the SLLJ in the west Indian Ocean. From phase 2 to 3, the wet anomalies advance progressively eastward along the equatorial Indian Ocean (as do the active convection). Meanwhile, north Arabic Sea, Indian Peninsula and Bay of Bengal are dominated by negative specific humidity (dry) anomalies (as well as suppressed convection). In phases 4 and 5, the dry anomalies weaken and move northward, while wet anomalies expand toward the north as well. From phase 6 to 8, these wet anomalies advance over India toward the northeast, while an anomalously dry atmosphere is found over the Arabian Sea and the south of India. Similar patterns were found by previous studies [Fu et al., 2006; Yang et al., 2008; Tian et al., 2010], which documented 3-D structure and evolution of specific humidity during boreal summer MJO, both with conventional global reanalysis and current available satellite data. They also indicated that in the region of strongly enhanced (suppressed) convection; the moist (dry) anomalies usually occupy the whole troposphere, but with differences in the altitude of the maximum specific humidity center.

Figure 5.

Composites of (qT)2–6 during the eight strong MJO phases for the period of JJA (2000–2004). The dotted areas indicate where the differences are significant at 90% confidence level.

[32] From these results, it seems reasonable to infer that the intraseasonal rainfall variability at MJO scale over India is not modulated by the quantity of particles transported by the SLLJ (i.e., the total advected mass of air), but mainly by the water content of those particles. Therefore, air-sea coupling appears as the essential process in the MJO influence over the monsoon rainfall over India.

4. Summary and Concluding Remarks

[33] There are two ideas widely accepted concerning the ISM rainfall variability. First, it is known that the MJO strongly modulates its intraseasonal signal [e.g., Pai et al., 2011]. Second, the SLLJ is the main moisture source of the ISM rainfall [e.g., Ordóñez et al., 2012]. However, the relationship between the water vapor transported by the SLLJ and the MJO evolution has not yet been directly addressed. In this work, a Lagrangian particle dispersion model has been used to analyze the water transported by the SLLJ toward India in terms both of the air mass advected and changes in its moisture content as a function of the MJO phase.

[34] We have studied the average conditions over a 5 year period (2000–2004) using available data. While short, this period has been used in previous studies [Nieto et al., 2007, 2008; Durán-Quesada et al., 2010; Gimeno et al., 2010], and it has been considered as a good representation of the mean climatic conditions since it does not include extreme events of the major modes of climate variability, such as El Niño-Southern Oscillation (ENSO) or the North Atlantic Oscillation (NAO). In fact, this period has been used by some authors to identify and quantify the moisture sources in different climatic regions such as the Sahel [Nieto et al., 2006], Iceland [Nieto et al., 2007], Orinoco Basin [Nieto et al., 2008], central America [Durán-Quesada et al., 2010], or the Iberian Peninsula [Gimeno et al., 2010]. However, a particular analysis was made for the 2002 ISM season due to its exceptional weakness [Fasullo, 2005; Chaudhari et al., 2010]. To assess the robustness of the method, we also compared (E-P) patterns with the precipitation climatology for the same period and the results matched quite well.

[35] Despite the length of the study period, the modulation of the MJO in precipitation over India is in good accordance with previous studies. Thus, during MJO phases 1 and 2, a below average rainfall band is observed along the monsoon trough (break monsoon type rainfall over India). Subsequently, a gradual northward shift of the positive rainfall anomaly band from the southern Indian Peninsula toward the north India is observed. During phases 5 and 6, the above normal rainfall band is observed along the monsoon trough (active monsoon type rainfall distribution). During the final phases (7 and 8), a general decrease in rainfall is observed over most of the country.

[36] The spatial (E-P) pattern computed for the particles previously over the SLLJ which arrive to India between 2 and 6 days later, are quite similar to that of the observed precipitation during the entire MJO cycle over the northern Indian Ocean and southern India. This result strongly suggests that the main responsible for the rainfall variability associated to the MJO in these regions are the changes in the moisture transported by the SLLJ. On the other hand, (E-P) over northern India does not provide a good representation of rainfall. Thus, these regions do not seem to be so affected by this modulation of the water vapor transported by the SLLJ. This last result is in good agreement with the observations made by Hartmann and Michelsen [1989].

[37] The analysis of the total mass and the specific humidity transported by the SLLJ during MJO life cycle does not reveal a systematic difference between air mass advected to India during active and break phases of MJO. In contrast, moist (dry) anomalies over enhanced (suppressed) convection region are found. We conclude that MJO impact is not reflected as much in the quantity of air masses advected to India by the SLLJ but in the moisture content of these air masses. Therefore, air-sea coupling seems to be an essential mechanism in the MJO influence over ISM rainfall.

[38] The origin of the northward propagation of the MJO during the boreal summer is one of the most striking features of the South Asian monsoon and many important contributions have addressed the underlying mechanisms [Abhik et al., 2013]. Our results may shed light over the difficult understanding of this origin. While different mechanisms have been proposed (e.g., Rossby wave emanation from an eastward-propagating equatorial Kelvin-Rossby wave packet [Wang and Xie, 1997; Lawrence and Webster, 2002], air-sea interaction [Kemball-Cook and Wang, 2001; Fu et al., 2003; Yang et al., 2008], a moisture-convection feedback mechanism [Jiang et al., 2004], easterly vertical shear mechanism [Jiang et al., 2004; Drbohlav and Wang 2005; Wang et al., 2005] or vertical troposphere diabatic heating [Xavier et al., 2007; Jiang et al., 2011, Wong et al., 2011]), our results further emphasize that the MJO signal propagates northward by the specific humidity advection by the meridional wind. A pronounced meridional gradient of specific humidity is found during the MJO life cycle, being the quantity of specific humidity transported northward modulated by air-sea interaction.

[39] Finally, the authors would like to remark that the anomalies in the intraseasonal distribution of the ISM rainfall can be important, and they have an enormous socioeconomic impact for India. Therefore, the prediction of the monsoon rainfall is imperative for India, and definitely, the predictability of the Indian monsoon may depend on the understanding of the mechanisms involved in its variability. The understanding of the sources and the transport of the atmospheric moisture is essential to the interpretation of the mechanism originating rainfall variability. Moreover, our results could be directly applicable to the generation of a MJO predictor based in the observation of the moisture transported by the SLLJ.


[40] We thank Andreas Stohl for providing the trajectory data. This work is a contribution to the project TRODIM (CGL2007-65891-C05-04) and to the research group RNM-356 belonging to the Plan Andaluz de Investigacion Desarrollo e Innovacion.