1. Introduction
Since Madden and Julian (1972) found the eastward-propagating oscillations over the tropical Indo-Pacific region, many observational analyses have revealed a slowly eastward-propagating convective envelope characterized by planetary-scale circulation with a broad life span of 30–60 days (e.g., Lau and Chan 1986; Hendon and Salby 1994; Zhang 2005; Lau and Waliser 2005). Despite numerous studies about the Madden–Julian oscillation (MJO), some fundamental questions still remain to be answered, like what initiates an MJO and what determines the propagation and growth of MJOs. Our further progress hinges on a better understanding about multiscale interaction processes in MJO (Wheeler and Kiladis 1999; Mapes et al. 2006).
Understanding the MJO phenomenon is important for diagnosis and prediction of tropical weather and climate. For example, MJO is observed to virtually influence intraseasonal precipitation of the Asian and Australian monsoons (e.g., Sui and Lau 1992; Lawrence and Webster 2002). The cyclone geneses in the western Pacific, Indian Ocean, and Caribbean Sea basins are also modulated by the intraseasonal oscillation (e.g., Liebmann et al. 1994; Maloney and Hartmann 2000; Kim et al. 2008; Ching et al. 2010).
In a typical life cycle of the MJO, like primary and successive events classified by Matthews (2008), the convective envelop initiates from the Indian Ocean and propagates eastward to the Maritime Continent, where the MJO’s circulation weakens but reintensifies upon reaching the Pacific warm pool. Actually, many MJOs evolve differently from this typical life cycle: some behave more like a stationary dipole oscillation over the Indian Ocean and the warm pool.
The previous theories explaining the propagating mechanism can be separated into two sets of theories: the tropical wave dynamics and the moisture mode. In the set of wave dynamics, the eastward propagation of MJOs is first explained by Kelvin waves based on the forced wave dynamics with a wave–conditional instability of the second kind (wave-CISK)-type parameterization of convective heating (e.g., Lau and Peng 1987). The most unstable wave in such a simplified system is normally at a small wavelength, which is different from the observed planetary-scale circulation associated with MJOs. To remedy the scale selection problem, Wang (1988) and Wang and Li (1994) added friction-induced boundary layer convergence in the wave-CISK framework. In the wave-CISK framework, convection-reduced vertical stratification can slow down Kelvin wave speed (~15 m s−1), which is still faster than observed MJO (~5 m s−1). Since moisture is treated as a diagnostic variable in the wave-related theory, horizontal moisture advection is normally neglected and vertical moisture advection is crudely coupled with convective heating. As we will show in this study, these moistening processes are crucial for the MJO propagation.
In the second set of moisture-mode theories, the horizontal moisture advection and surface moisture flux that alter the moisture [therefore moist static energy (MSE)] tendency are regarded as the key processes of the MJO propagation (e.g., Neelin and Yu 1994; Hu and Randall 1994; Sobel and Maloney 2012, 2013). Some studies use column-integrated MSE or moist entropy because of the property of conservation, while most of the intraseasonal MSE variations are dominated by moisture variation with a relatively weak temperature variation. MJO simulation in some general circulation models can be improved by making deep convection sensitive to free-tropospheric moisture (e.g., Wang and Schlesinger 1999; Grabowski 2003). Recent studies performing moisture budgets and moist static energy budgets of the MJO show that cloud radiative feedback and horizontal moisture advection are important for the MJO’s growth and propagation (Maloney 2009; Hagos and Leung 2011; Andersen and Kuang 2012; Kim et al. 2014; Pritchard and Bretherton 2014; Hagos et al. 2014; Wang et al. 2015).
An important process recognized in both sets of theories is the premoistening ahead of the MJO convective phase as supported by contemporary observations revealing the prevalence of low and middle clouds during the suppressed phase of the MJO. Kikuchi and Takayabu (2004) used composite geostationary meteorology satellite data during TOGA COARE together with the upper-air soundings from the intensive observation period (IOP) of the experiment to obtain the vertical thermal structures and moisture structures. They showed lower-tropospheric moistening in the cloud-developing phase associated with the MJO, which is strongly modulated by shallow clouds and congestus cloud. Lau and Wu (2010) used 4 yr of Tropical Rainfall Measuring Mission data for MJO composite analyzing the top echo height and heating profile in different phases. They also showed the existence of shallow clouds and lower-tropospheric adiabatic heating during suppressed phase. An increase of moisture in the lower troposphere can destabilize the atmosphere for the following development of convection. Del Genio et al. (2012) analyzed CloudSat and Cloud–Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) data. Their composite of 10 MJO events revealed shallow and congestus clouds in advance of the peak deep clouds. Hsu and Li (2012) utilized 20-yr reanalysis data to show that the vertical advection, which results from boundary layer convergence, is the major moisture source of MJO evolution. Despite the studies cited above, the premoistening processes associated with the initiation and the propagation of the MJO are still unclear. In this study, we analyze scale-separated budget to quantify the dominant moistening process in different stages of the MJO life cycles. We also calculate apparent moisture sink (Q2) based on large-scale budgets (Yanai et al. 1973). The Q2 is an estimate of the collective effect of cloud moistening in a selected domain, which is large in the presence of deep convective system but weaker in the less convective environment, where shallow convection may moisten or dry the lower troposphere. The budget estimate requires data of higher quality and spatiotemporal resolution. The intensive sounding observations from the Dynamics of the MJO (DYNAMO)/Cooperative Indian Ocean Experiment on Intraseasonal Variability in Year 2011 (CINDY) field campaign provide an opportunity for us to perform a moisture budget analysis in this study.
In this study, we calculate a diagnostic moisture budget in the DYNAMO/CINDY sounding array in the Indian Ocean (IO). Instead of using reanalysis data, we use operational data from the European Centre for Medium-Range Weather Forecasts, which are assimilated with field observations and satellite data during the DYNAMO/CINDY IOP. In section 2, we describe the data and method utilized in this study. In section 3, the MJO evolutions from October to December 2011 are discussed. Section 4 presents the diagnostic moisture budget of the MJO. Section 5 presents concluding remarks.
2. Data and method
In this study, we use the following three datasets for the DYNAMO/CINDY period from 1 October to 31 December 2011.
a. Sounding observations
The DYNAMO/CINDY upper-air sounding network comprises two quadrilateral arrays, one north and one south of the equator (Fig. 1), with a total of six sounding sites, including three atolls, one island, and two scientific ships. Intensive observations were made at the six sounding sites from 1 October to 15 December 2011, except for a port call of the research vessel (R/V) Mirai in the south sounding array after 30 November and two port calls of the R/V Revelle from 31 October to 7 November and from 8 to 16 December [see Johnson and Ciesielski (2013) for complete description]. Thus, in this study, we only use the sounding data from 1 October to 30 November. In Fig. 1, the open and closed circles are intensive observation sites, which released from four to eight radiosondes per day. All radiosonde observations are mass weighted for every 50 hPa and time averaged for each day.
The 2-month (October and November) averaged SST (shading; °C) and 1000-hPa wind (vector; m s−1) in the DYNAMO/CINDY sounding array. The open and closed circles are intensive observation sounding sites that release four and eight radiosondes per day, respectively.
Citation: Journal of Climate 28, 8; 10.1175/JCLI-D-14-00416.1
b. Satellite data
The dataset of Tropical Rainfall Measuring Mission, version 7 (TRMMv7), including storm height (2A23) and rain rate (3B42), is used to describe cloud and rainfall evolutions in different phases of the MJO in DYNAMO. The horizontal resolution of TRMM level 2A data is 4 × 4 km2, with a temporal resolution of 12 h. The 3B42 data are on a 0.25° × 0.25° grid resolution and 3-h temporal resolution (Huffman et al. 2007). The sea surface temperature data are from the TRMM Microwave Imager (TMI) with spatial and temporal resolutions of 0.25° × 0.25° and 1 day, respectively. The interpolated daily outgoing longwave radiation (OLR) obtained from the National Oceanic and Atmospheric Administration (NOAA) polar-orbiting satellites (Liebmann and Smith 1996) is gridded at 2.5° × 2.5°.
c. ECMWF operational data
The budget analysis is computed using operational data from the European Centre for Medium-Range Weather Forecasts (ECMWF). About 95% of DYNAMO/CINDY sounding observations are transmitted to operational centers in real time, so the region of the sounding array is strongly influenced by in situ observations. Before doing budget analysis, we compare the ECMWF operational data and sounding data. The result (not shown) reveals that operational data capture synoptic-scale and intraseasonal-scale signal well. Ciesielski et al. (2014) carried out an assessment of moisture fields from the ECMWF operational products using quality-controlled upper-air sounding data. They found an overall good agreement, with the exception at upper levels, where the assimilated temperature values have a positive bias. The bias is not expected to affect our diagnostic result of lower-tropospheric budget. The 6-hourly operational data used in this study include zonal winds, meridional winds, vertical velocity, and specific humidity at 15 vertical layers from 1000 to 100 hPa and 0.25° × 0.25° latitude–longitude horizontal spatial resolutions.
d. Diagnostic moisture budget
e. Scale-separated moisture budget
The averaged power spectra of mean daily OLR (W m−2) in the 12 grid boxes of 5° × 5° over the central Indian Ocean (5°S–5°N, 60°–90°E, separated by every 5° × 5°) shown by vertical bars as a function of period. The red solid line is the red noise curve.
Citation: Journal of Climate 28, 8; 10.1175/JCLI-D-14-00416.1
To compare the time-filtered result with the theoretical waves structures, we also perform the time–space spectral analysis described by Wheeler and Kiladis (1999) to extract the MJO, equatorial Rossby waves, and equatorial Kelvin waves. The MJO is extracted by eastward-propagating components of wavenumbers 0–5 and periods of 30–90 days. The equatorial Kelvin waves are extracted by wavenumbers 0–5 and periods of 10–30 days, and the equatorial Rossby (ER) waves correspond to wavenumbers 1–10 and periods of 10–40 days.
3. The MJO evolution in the DYNAMO/CINDY period
a. Large-scale environment and rainfall evolution
Time averaged winds at 1000 hPa and sea surface temperature (SST) are shown in Fig. 1 for the IOP of the DYNAMO/CINDY (October and November). The SST shows a meridionally and zonally asymmetric pattern in the western and the southern Indian Ocean relative to the DYNAMO/CINDY sounding array. While most of the precipitation occurs in the region of SST higher than 28°C, the strongest rainfall is located in the region of large SST gradient, which is characteristic of substantial ITCZ rainfall (see Fig. 4 in Johnson and Ciesielski 2013). The near-surface winds are characterized by a pair of cyclonic flows to the north and south of the DYNAMO/CINDY sounding array. On the equator, the strong westerly is related to the two MJO events in October and November.
The overall evolution of the MJOs in the DYNAMO/CINDY is shown by the Hovmöller diagram of the 15–60-day bandpass-filtered OLR in Fig. 3 and unfiltered SST and precipitation in Fig. 4, all averaged within the 7.5°S–7.5°N latitude band. Also shown in Figs. 3 and 4 are space–time-filtered MJO from OLR and Kelvin waves from zonal wind at 850 hPa averaged over 7.5°S–7.5°N and ER waves from vorticity at 850 hPa averaged over 5°–15°N and 5°–15°S. From early October 2011 to January 2012, there are two strong MJO events (MJO1 and MJO2) and one weak event (MJO3). Both MJO1 and MJO2 appear to initiate at around 50°–60°E and propagate eastward. MJO1 starts over the Indian Ocean from a suppressed phase (anomalous high OLR) in early October with no prior propagating MJO event. The OLR anomaly then propagates eastward to initiate the following development of deep convective phase of MJO1 over the Indian Ocean in middle-to-late October. The MJO1 further leads to the initiation and propagation of MJO2 in November. The relevant mechanism for initiation and development is the focus of this study, which is discussed in section 4. The intraseasonal oscillation in December, on the other hand, is a less organized event, with an eastward-moving suppressed phase and a westward-moving convective phase (Fig. 3). The event is not discussed in this study. We find the convection center of both MJO1 and MJO2 located in the zonal wind convergence of Kelvin wave, but the major rainband shows more detailed features. In the MJO1 convective phase (15–30 October), precipitation exhibits 1–2-day oscillations emerging from the Maritime Continent near 120°E and propagating westward to central IO. The disturbance is related to the strong diurnal cycle in the suppressed phase of the MJO, especially over the Maritime Continent, as identified before in previous studies (e.g., Takayabu et al. 1996; Sui et al. 1997). This phenomenon is relatively weak and confined east of 90°E in MJO2. The MJO2 precipitation consists of two Kelvin wave–like rainbands (Fig. 4b) resulting from subtropical dry air intrusion (Kerns and Chen 2014). The SST in Fig. 4a exhibits multiscale variability that results from air–sea flux exchanges associated with the MJO evolution as well as ocean transports. The SST variability in October is not correlated with MJO1, likely more as a result of dominant ocean transports than the air–sea fluxes. However, the warming in the Indian Ocean in the first half of November is in phase with the suppressed phase of MJO2 and the cooling in the second half of November is in phase with the westerly Kelvin wave. These can be attributed to the enhanced solar heating in the suppressed phase and evaporation cooling associated with westerly wind burst and cloud albedo effect in the convective phase of MJO2, similar to the intraseasonal variability observed in the western Pacific (e.g., Sui and Lau 1992).
The Hovmöller diagram of 15–60-day bandpass-filtered OLR (shading; W m−2), space–time-filtered MJO from OLR (black contours; interval of 10 W m−2), Kelvin waves from zonal wind at 850 hPa (purple contours; interval of 0.5 m s−1) averaged over 7.5°S–7.5°N, and equatorial Rossby waves from vorticity at 850 hPa (green contours; interval of 3 × 10−6 s−1) averaged over 5°–15°N and 5°–15°S.
Citation: Journal of Climate 28, 8; 10.1175/JCLI-D-14-00416.1
Hovmöller diagrams of (a) sea surface temperature (°C) and (b) precipitation (mm h−1) averaged within 7.5°S–7.5°N. Contours superimposed in (a) as in Fig. 3.
Citation: Journal of Climate 28, 8; 10.1175/JCLI-D-14-00416.1
b. Evolution of moisture, vertical motion, and convective activity over the DYNAMO/CINDY sounding array
Figure 4 shows clearly the passage of MJOs through the DYNAMO/CINDY sounding array; we further examine the evolution of convective activities associated with MJOs within the sounding array. To help better define the evolution of MJOs at the sounding array, we first show the temporal evolution of 15–60-day-filtered OLR and specific humidity in Fig. 5. The filtered time–height distribution of specific humidity in Fig. 5a shows the dry–wet phases of the three MJO events with the maximum variability near 700 hPa. The moisture anomaly field below 700 hPa shows a distinct vertical tilt with time, showing a moistening boundary layer prior to the development of deep convective phase. This is an important feature that was noted previously by Sperber (2003) and Kiladis et al. (2005). In Fig. 5b we show the temporal evolution of 15–60-day-filtered OLR and vertically integrated [q]′ from the surface to 700 hPa to quantify the phase relation between clouds and lower-troposphere moisture in the MJO evolution. Based on the two variables, the evolution of MJO1 can be separated into the following four phases: suppressed phase (1–9 October, when OLR′ is positive and [q]′ is low); cloud-developing phase (10–19 October, when [q]′ grows to a maximum and OLR′ turns negative); convective phase (20–29 October, when OLR′ reaches minimum); and decaying phase (30 October–5 November, when OLR′ increases and [q]′ decreases to zero). In terms of the MJO index defined in Wheeler and Hendon (2004), the above four phases correspond to 567, 81, 2, and 34, respectively.
(a) Time–height structure of [q]′ (mm) and (b) time series of OLR′ (blue line) and 1000–700-hPa integrated specific humidity (green line). All variables are bandpass filtered at 15–60 days. The numbers shown above the x axis in (b) are the corresponding real-time multivariate MJO (RMM) index.
Citation: Journal of Climate 28, 8; 10.1175/JCLI-D-14-00416.1
Next, we examine in Fig. 6 the temporal evolution in quantities associated with convective activities including Q2, vertical velocity, frequency count of storm height, rain rate, and equivalent potential temperature observed in the DYNAMO/CINDY north sounding array. Here, Q2 and vertical velocity are derived from ECMWF operational analysis, equivalent potential temperature is derived from the DYNAMO/CINDY soundings, and the rest are from TRMM data, as discussed in section 2. Color shadings in Figs. 6a,b show time–height cross section of daily averaged Q2 (Fig. 6a) and normalized frequency time series of storm height (Fig. 6b). The vertical velocity is also shown in the two figures by contours. The vertically integrated frequency counts of storm heights in Fig. 6b give the cloud fraction (vertical bars in Fig. 6b). The most distinguished evolution from early October to late October is the dramatic increasing of cloud population. In the suppressed phase (1–9 October), large-scale downward motion confines the development of convection, with storm heights below 5 km and the domain cloud coverage less than 10%. The corresponding Q2 (Fig. 6a) in the MJO1 suppressed phase is significantly negative below 6 km, with the largest values exceeding −7 K day−1. Since negative Q2 can extend up to 600 hPa, where the boundary layer eddy activities cannot reach, the moistening implies reevaporation of shallow convection and congestus (Johnson et al. 1999). The shallow convection can efficiently moisten the low-level and midlevel troposphere and enhanced the convective instability that are required for the following development of deep convection.
Convective activities within the DYNAMO/CINDY north sounding array: (a) time–height cross section of 
Citation: Journal of Climate 28, 8; 10.1175/JCLI-D-14-00416.1
In the cloud-developing phase (10–19 October), we observe a gradual increase of cloud coverage and storm height. Some storms can exceed the freezing level by about 5 km, even to 10 km. The development of organized convection is associated with increasing domain-mean precipitation (Fig. 6c), weakening domain-mean downward motion, and changing Q2 in the low troposphere from negative to positive values. The above features indicate a gradual deepening of convection from nonprecipitating to precipitating clouds. In the convection active phase (21–27 October), the height of maximum Q2 matches well with the storm height distribution from TRMM observations, which is around 6 km, whereas the stronger convection can reach 12 km. The large positive values of Q2 indicate dominant condensation heating (exceeding 7 K day−1), as revealed by the intermittent strong precipitation (~1.5 mm h−1). The distribution of equivalent potential temperature (θe) shows a more neutral troposphere in the convective phase than that in the suppressed phase (Fig. 6d). In the decaying phase (30 October–5 November), the maximum storm height decreases to 8 km but the maximum vertical velocity and positive Q2 remain in upper troposphere, which is characteristic of prevailing stratiform clouds.
For MJO2, the evolution in the DYNAMO/CINDY sounding array is also divided into the four phases based on the same criteria as for MJO1: suppressed phase (6–12 November), cloud-developing phase (13–20 November), convective phase (21–25 November), and decaying phase (25 November–4 December). While there are some overall similarities between the two MJOs, some distinctive differences are noted. First, the duration of the four phases in MJO2 is generally shorter than that in MJO1, except the decaying phase. Second, the low-level Q2 in the suppressed phase of MJO2 (6–12 November) exhibits more positive Q2 values in the lower troposphere and gradually deepens to 600 hPa during the period from early to middle November (Fig. 6a). The positive Q2 here implies a drying (heating) effect by precipitating clouds. Contrary to recharging moisture by dominant nonprecipitating shallow convection in suppressed phase of MJO1, emerging precipitating convection discharges moisture in the suppressed phase of MJO2. This is supported by the relatively unstable atmosphere in the suppressed phase of MJO1 and a more neutral atmosphere in the suppressed phase of MJO2 (Fig. 6d), and the storm height in the suppressed phase of the MJO2 is significantly higher compared to the MJO2.
The overall evolution through the MJO life cycle discussed above is consistent with previous findings (Lau and Sui 1997; Kikuchi and Takayabu 2004; Benedict and Randall 2007; Mapes et al. 2006). Thus, the key to understand the mechanism of MJO propagation (local phase change) is to reveal specific processes causing lower-tropospheric moistening and drying. In the next section, we introduce the scale-separated moisture budget to quantify the dominant processes responsible for the MJO phase change.
4. Moisture budget
In this section, we present an analysis of vertically integrated moisture budget [Eq. (4)]. All budget terms are temporally filtered to retain the intraseasonal (15–60 day) band, areal averaged in the DYNAMO/CINDY north sounding array (0°–5°N, 73°–80°E) and vertically integrated in the lower troposphere from 1000 to 700 hPa (referred to here as the lower troposphere unless otherwise stated).
a. Projection of budget terms onto moisture anomaly and its tendency change
To determine the relative contribution of each budget term to the local moisture tendency change in the sounding array, we project time series of each vertically integrated budget term to the time series of tendency change term [∂q/∂t]′ in the period of the DYNAMO/CINDY intensive observation period (1 October–30 November). The projections shown in Fig. 7 are normalized to show fractional contribution of each moisture budget terms to [∂q/∂t]′. Figure 7a shows that both horizontal and vertical advection terms project to positive tendency change, while Q2 project negatively to moisture tendency change. The projection by each scale-separated advection terms are shown in Figs. 7b,c. The three leading terms, 
The projections of (a) [∂q/∂t]′, 
Citation: Journal of Climate 28, 8; 10.1175/JCLI-D-14-00416.1
The correlation coefficients between moisture budget terms and [∂q/∂t]′ in the DYNAMO/CINDY period. The correlations are calculated for two vertically integrated budgets: surface–700 hPa and 700–100 hPa. The correlation coefficients passing 95% significance of t tests are shown in boldface. The effective degree of freedom of bandpass-filtered data is 15 (Yan et al. 2004).
Time series of 1000–700-hPa integrated [q]′ (blue dashed line; mm), [∂q/∂t]′ (green dashed line; mm h−1), and respective moisture budget terms (red solid line; mm h−1): (a) 
Citation: Journal of Climate 28, 8; 10.1175/JCLI-D-14-00416.1
We also check the projection of each budget term onto [q]′, to evaluate their contribution to local moisture growing or decaying (Fig. 9). The figure illustrates that 
The projections of (a) 
Citation: Journal of Climate 28, 8; 10.1175/JCLI-D-14-00416.1
The correlation coefficients between moisture budget terms and [q]′ in the DYNAMO/CINDY period. The correlations are calculated for two vertically integrated budgets: surface–700 hPa and 700–100 hPa. The correlation coefficients passing 95% significance of t tests are shown in boldface.
b. Horizontal distribution of moisture budgets
To better understand the processes causing moisture changes, we examine horizontal distributions of dominant budget terms and associated variables in the four phases of MJO1 and MJO2 as identified in section 3b. Important features in the two MJOs are similar, so we only show results from selected phases in MJO1 here. Figure 10 shows the budget distribution averaged in a suppressed phase (1–9 October; Fig. 10, left) and a cloud-developing phase (10–19 October; Fig. 10, right). First, we show horizontal distribution of 
The horizontal distributions of (a) 
Citation: Journal of Climate 28, 8; 10.1175/JCLI-D-14-00416.1
We then show the term 
Similar to 
In Fig. 10e, we show the synoptic-scale advection term 
c. Moisture budget in the premoistening stage
Although projections of moisture budgets on the intraseasonal moisture anomaly and its derivative through the life cycles of MJO1 and MJO2 support the findings in previous studies that moisture advection is important for MJO propagation and growing/decaying, we note some differences in the premoistening processes from suppressed phase to cloud-developing phase between MJO1 and MJO2. In section 3b, we find large-scale downward motion is stronger in the suppressed phase of MJO1 than that in MJO2. The difference is primarily due to the low-frequency variability in background flow, as revealed in the time series of low-frequency vertical motion (
(a) Time series of 1000–700-hPa integrated 
Citation: Journal of Climate 28, 8; 10.1175/JCLI-D-14-00416.1
5. Conclusions and discussion
In this study, soundings, ECMWF operational data, and TRMM data are used to investigate the moistening processes in the lower troposphere (1000–700 hPa) for the October and November 2011 MJOs during the DYNAMO/CINDY IOP period. MJO1 starts over the Indian Ocean from a suppressed phase (high OLR anomaly) in early October with no prior propagating MJO event. On the other hand, MJO2 is a successive event following MJO1. The two MJOs are divided into four phases: suppressed, cloud developing, convective, and decaying, based on 15–60-day-filtered OLR and lower-tropospheric moisture.
By synthesizing vertical velocity, equivalent potential temperature, and convective activities, the two MJOs exhibit similar evolution except in the premoistening stage. In MJO1, the suppressed phase is characterized by large-scale downward motion, cloud population composed mostly of nonraining clouds with storm heights below 5 km and cloud coverage less than 10%, and negative Q2 indicating a collective effect of moistening by reevaporation in shallow convection and congestus. Different from that in MJO1, raining clouds and positive Q2 appear in the lower troposphere during the suppressed phase in MJO2, indicating a net drying and heating effect by convection. Overall, the MJOs evolve from a buildup of moisture in the lower troposphere that leads to enhanced potential instability and the consequent cloud-developing phase. A full development of deep convection neutralizes the troposphere and eventually terminates deep convection, leading to the decaying phase.
The dominant moistening processes responsible for the MJO evolution is investigated by a scale-separated moisture budget in the lower troposphere using decomposed wind and moisture fields in three frequency bands: low frequency (>60 days), MJO (15–60 days), and synoptic scale (<15 days). By projecting the budget terms onto the moisture tendency change term and checking their correlation over the whole DYNAMO/CINDY IOP, three broad-scale advection terms by the low-frequency and intraseasonal flow and moisture fields: namely, 
In addition to horizontal advection, the moistening of vertical advection by large-scale divergent flow and the counter drying by convection precipitation are also dominant processes for the MJO propagation. This is consistent with Hsu and Li (2012), who performed a diagnostic study to show that lower-tropospheric moisture convergence as the dominant moisture source for the MJO propagation. The two dominant terms, however, do not always show a clear phase relation with moisture tendency changes through the life cycle of the two MJOs. In fact, the premoistening in the lower troposphere ahead of the deep convection is observed only at the cloud-developing to convective phases of the two MJOs.
This study further reveals a diffusive effect of moisture advection by high-frequency disturbances 
Through the analysis of scale-separated moisture budget, we identify dominant moistening processes through life cycles or different phases of MJOs. Such an analysis can provide a better insight about the cause of MJO evolutions. The study also provides useful references for evaluating and understanding model simulations. However, we need to confirm the findings by analyzing more MJOs of different types in different seasons and regions. We also need to include a heat budget in addition to a moisture budget to understand the physical processes that determine the time and space scales of intraseasonal oscillations.
Acknowledgments
We thank Steve Williams of the National Center for Atmospheric Research (NCAR) Earth Observing Laboratory (EOL) for providing the ECMWF operational data. We also acknowledge Richard Johnson and Paul Ciesielski for sounding data support and Hung-Jui Yu for discussion of sounding data quality. We thank the NASA TRMM project for cloud and rainfall data. We acknowledge all participants of the DYNAMO/CINDY project who make the observation and analysis data available, especially Chi-Dong Zhang for his leadership in planning and carrying out the experiment. We also thank Po-Hsuing Lin, Wei-Ting Chen, and Chien-Ming Wu of the National Taiwan University for discussion of scientific issues. This research was fund by National Sciences Council in Taiwan. KCT and CHS were supported by the NSC under Grants NSC 102-2745-M-002-003-ASP.
REFERENCES
Andersen, J. A., and Z. Kuang, 2012: Moist static energy budget of MJO-like disturbances in the atmosphere of a zonally symmetric aquaplanet. J. Climate, 25, 2782–2804, doi:10.1175/JCLI-D-11-00168.1.
Benedict, J. J., and D. A. Randall, 2007: Observed characteristics of the MJO relative to maximum rainfall. J. Climate, 64, 2332–2354, doi: 10.1175/JAS3968.1.
Ching, L., C.-H. Sui, and M.-J. Yang, 2010: An analysis of multi-scale nature of tropical cyclone activities in June 2004: Climate background. J. Geophys. Res., 115, D24108, doi:10.1029/2010JD013803.
Ciesielski, P. E., and Coauthors, 2014: Quality-controlled upper-air sounding dataset for DYNAMO/CINDY/AMIE: Development and corrections. J. Atmos. Oceanic Technol., 31, 741–764, doi:10.1175/JTECH-D-13-00165.1.
Del Genio, A. D., Y. Chen, D. Kim, and M.-S. Yao, 2012: The MJO transition from shallow to deep convection in CloudSat/CALIPSO data and GISS GCM simulations. J. Climate, 25, 3755–3770, doi:10.1175/JCLI-D-11-00384.1.
Grabowski, W. W., 2003: MJO-like coherent structures: Sensitivity Simulations using the cloud-resolving convection parameterization (CRCP). J. Atmos. Sci., 60, 847–864, doi:10.1175/1520-0469(2003)060<0847:MLCSSS>2.0.CO;2.
Hagos, S., and L. R. Leung, 2011: Moist thermodynamics of the Madden–Julian oscillation in a cloud-resolving simulation. J. Climate, 24, 5571–5583, doi:10.1175/2011JCLI4212.1.
Hagos, S., Z. Feng, K. Landu, and C. N. Long, 2014: Advection, moistening, and shallow-to-deep convection transitions during the initiation and propagation of Madden-Julian oscillation. J. Adv. Model. Earth. Syst., 6, 938–949, doi:10.1002/2014MS000335.
Hendon, H. H., and M. L. Salby, 1994: The life cycle of the Madden–Julian oscillation. J. Atmos. Sci., 51, 2225–2237, doi:10.1175/1520-0469(1994)051<2225:TLCOTM>2.0.CO;2.
Hsu, P.-C., and T. Li, 2012: Role of the boundary layer moisture asymmetry in causing the eastward propagation of the Madden–Julian oscillation. J. Climate, 25, 4914–4931, doi:10.1175/JCLI-D-11-00310.1.
Hu, Q., and D. A. Randall, 1994: Low-frequency oscillations in radiative–convective systems. J. Atmos. Sci., 51, 1089–1099, doi:10.1175/1520-0469(1994)051<1089:LFOIRC>2.0.CO;2.
Huffman, G. J., R. F. Adler, D. T. Bolvin, G. Gu, E. J. Nelkin, K. P. Bowman, E. F. Stocker, and D. B. Wolff, 2007: The TRMM Multisatellite Precipitation Analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales. J. Hydrometeor., 8, 38–55, doi:10.1175/JHM560.1.
Johnson, R. H., and P. E. Ciesielski, 2013: Structure and properties of Madden–Julian oscillations deduced from DYNAMO sounding arrays. J. Atmos. Sci., 70, 3157–3179, doi:10.1175/JAS-D-13-065.1.
Johnson, R. H., T. M. Rickenbacg, S. A. Rutledge, R. E. Ciesielski, and W. H. Schubert, 1999: Trimodal characteristics of tropical convection. J. Climate, 12, 2397–2418, doi:10.1175/1520-0442(1999)012<2397:TCOTC>2.0.CO;2.
Kerns, B. W., and S. Chen, 2014: Equatorial dry air intrusion and related synoptic variability in MJO initiation during DYNAMO. Mon. Wea. Rev., 142, 1326–1343, doi:10.1175/MWR-D-13-00159.1.
Kikuchi, K., and Y. N. Takayabu, 2004: The development of organized convection associated with the MJO during TOGA COARE IOP: Trimodal characteristics. Geophys. Res. Lett., 31, L10101, doi:10.1029/2004GL019601.
Kiladis, G. N., K. H. Straub, and P. T. Haertel, 2005: Zonal and vertical structure of the Madden–Julian oscillation. J. Atmos. Sci., 62, 2790–2809, doi:10.1175/JAS3520.1.
Kim, D., J.-S. Kug, and A. H. Sobel, 2014: Propagating versus nonpropagating Madden–Julian oscillation events. J. Climate, 27, 111–125, doi:10.1175/JCLI-D-13-00084.1.
Kim, J.-H., C.-H. Ho, H.-S. Kim, C.-H. Sui, and S. K. Park, 2008: Systematic variation of summertime tropical cyclone activity in the western North Pacific in relation to the Madden–Julian oscillation. J. Climate, 21, 1171–1191, doi:10.1175/2007JCLI1493.1.
Lau, K.-M., and P. H. Chan, 1986: Aspects of the 40–50 day oscillation during the northern summer as inferred from outgoing longwave radiation. Mon. Wea. Rev., 114, 1354–1367, doi:10.1175/1520-0493(1986)114<1354:AOTDOD>2.0.CO;2.
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, 950–972, doi:10.1175/1520-0469(1987)044<0950:OOLFOI>2.0.CO;2.
Lau, K.-M., and C.-H. Sui, 1997: Mechanisms of short-term sea surface temperature regulation: Observations during TOGA COARE. J. Climate, 10, 465–472, doi:10.1175/1520-0442(1997)010<0465:MOSTSS>2.0.CO;2.
Lau, K.-M., and D. E. Waliser, 2005: Intraseasonal Variability in the Atmospheric-Ocean Climate System. Praxis, 436 pp.
Lau, K.-M., and H.-T. Wu, 2010: Characteristics of precipitation, cloud, and latent heating associated with the Madden–Julian oscillation. J. Climate, 23, 504–518, doi:10.1175/2009JCLI2920.1.
Lawrence, D., and P. J. Webster, 2002: The boreal summer intraseasonal oscillation and the South Asian monsoon. J. Atmos. Sci., 59, 1593–1606, doi:10.1175/1520-0469(2002)059<1593:TBSIOR>2.0.CO;2.
Liebmann, B., and C. A. Smith, 1996: Description of a complete (interpolated) outgoing longwave radiation dataset. Bull. Amer. Meteor. Soc., 77, 1275–1277.
Liebmann, B., H. Hendon, and J. Glick, 1994: The relationship between tropical cyclones of the western Pacific and Indian Oceans and the Madden–Julian oscillation. J. Meteor. Soc. Japan, 72, 401–411.
Madden, R. A., and P. R. Julian, 1972: Description of global-scale circulation cells in the tropics with a 40–50 day period. J. Climate, 29, 2665–2690, doi:10.1175/1520-0469(1972)029<1109:DOGSCC>2.0.CO;2.
Maloney, E. D., 2009: The moist static energy budget of a composite tropical intraseasonal oscillation in a climate model. J. Climate, 22, 711–729, doi:10.1175/2008JCLI2542.1.
Maloney, E. D., and D. L. Hartmann, 2000: Modulation of eastern North Pacific hurricanes by the Madden–Julian oscillation. J. Climate, 13, 1451–1460, doi:10.1175/1520-0442(2000)013<1451:MOENPH>2.0.CO;2.
Mapes, B. E., S. Tulich, J.-L. Lin, and P. Zuidema, 2006: The mesoscale convection life cycle: Building block or prototype for large-scale tropical waves. Dyn. Atmos. Oceans, 42, 3–29, doi:10.1016/j.dynatmoce.2006.03.003.
Matthews, A. J., 2008: Primary and successive events in the Madden-Julian oscillation. Quart. J. Roy. Meteor. Soc., 134, 439–453, doi:10.1002/qj.224.
Neelin, J. D., and J.-Y. Yu, 1994: Modes of tropical variability under convective adjustment and the Madden–Julian oscillation. Part I: Analytical results. J. Atmos. Sci., 51, 1876–1894, doi:10.1175/1520-0469(1994)051<1876:MOTVUC>2.0.CO;2.
Pritchard, M. S., and C. S. Bretherton, 2014: Causal evidence that rotational moisture advection is critical to the superparameterized Madden–Julian oscillation. J. Atmos. Sci., 71, 800–815, doi:10.1175/JAS-D-13-0119.1.
Sobel, A., and E. Maloney, 2012: An idealized semi-empirical framework for modeling the Madden–Julian oscillation. J. Atmos. Sci., 69, 1691–1705, doi:10.1175/JAS-D-11-0118.1.
Sobel, A., and E. Maloney, 2013: Moisture modes and the eastward propagation of the MJO. J. Atmos. Sci., 70, 187–192, doi:10.1175/JAS-D-12-0189.1.
Sperber, K. R., 2003: Propagation and the vertical structure of the Madden–Julian oscillation. Mon. Wea. Rev., 131, 3018–3037, doi:10.1175/1520-0493(2003)131<3018:PATVSO>2.0.CO;2.
Sui, C.-H., and K.-M. Lau, 1992: Multiscale phenomena in the tropical atmosphere over the western Pacific. Mon. Wea. Rev., 120, 407–430, doi:10.1175/1520-0493(1992)120<0407:MPITTA>2.0.CO;2.
Sui, C.-H., K.-M. Lau, Y. N. Takayabu, and D. A. Short, 1997: Diurnal variations in tropical oceanic cumulus convection during TOGA COARE. J. Atmos. Sci., 54, 639–655, doi:10.1175/1520-0469(1997)054<0639:DVITOC>2.0.CO;2.
Takayabu, Y. N., K.-M. Lau, and C.-H. Sui, 1996: Observation of a quasi-2-day wave during TOGA COARE. Mon. Wea. Rev., 124, 1892–1913, doi:10.1175/1520-0493(1996)124<1892:OOAQDW>2.0.CO;2.
Wang, B., 1988: Dynamics of tropical low-frequency waves: An analysis of moist Kelvin waves. J. Atmos. Sci., 45, 2051–2065, doi:10.1175/1520-0469(1988)045<2051:DOTLFW>2.0.CO;2.
Wang, B., and T. Li, 1994: Convective interaction with boundary dynamics in the development of a tropical intraseasonal system. J. Atmos. Sci., 51, 1386–1400, doi:10.1175/1520-0469(1994)051<1386:CIWBLD>2.0.CO;2.
Wang, S., A. Sobel, F. Zhang, Y. Sun, Y. Yue, and L. Zhou, 2015: Regional simulation of the October and November MJO events observed during the CINDY/DYNAMO field campaign at gray zone resolution. J. Climate, 28, 2097–2119, doi:10.1175/JCLI-D-14-00294.1.
Wang, W., and M. E. Schlesinger, 1999: The dependence on convection parameterization of the tropical intraseasonal oscillation simulated by the UIUC 11-layer atmospheric GCM. J. Climate, 12, 1423–145, doi:10.1175/1520-0442(1999)012<1423:TDOCPO>2.0.CO;2.
Wheeler, M., and G. N. Kiladis, 1999: Convectively coupled equatorial waves: Analysis of cloud and temperature in the wavenumber-frequency domain. J. Atmos. Sci., 56, 374–399, doi:10.1175/1520-0469(1999)056<0374:CCEWAO>2.0.CO;2.
Wheeler, M., and H. H. Hendon, 2004: An all-season real-time multivariate MJO index: Development of an index for monitoring and prediction. Mon. Wea. Rev., 132, 1917–1932, doi:10.1175/1520-0493(2004)132<1917:AARMMI>2.0.CO;2.
Yan, H.-M., M. Zhong, and Y.-Z. Zhu, 2004: Determination of the degree of freedom of digital filtered time series with an application to the correlation analysis between the length of day and the Southern Oscillation index. Chin. Astron. Astrophys., 28, 120–126, doi:10.1016/S0275-1062(04)90014-8.
Yanai, M., S. Esbensen, and J.-H. Chu, 1973: Determination of bulk properties of tropical cluster from large-scale heat and moisture budget. J. Atmos. Sci., 30, 611–627, doi:10.1175/1520-0469(1973)030<0611:DOBPOT>2.0.CO;2.
Zhang, C., 2005: Madden–Julian oscillation. Rev. Geophys., 43, RG2003, doi:10.1029/2004RG000158.
