Abstract

The evolution of total cloud cover and cloud types is composited across the Madden–Julian oscillation (MJO) using CloudSat data for June 2006–May 2010. Two approaches are used to define MJO phases: 1) the local phase is determined at each longitude and time from filtered outgoing longwave radiation, and 2) the global phase is defined using a popular real-time multivariate MJO (RMM) index, which assigns the tropics to an MJO phase each day.

In terms of local phase, CloudSat results show a familiar evolution of cloud type predominance: in the growing stages shallow clouds coexist with deep, intense, but narrow convective systems. Widespread cloud coverage and rainfall appear during the active phases, becoming more anvil dominated with time, and finally suppressed conditions return. Results are compared to the convectively coupled Kelvin wave, which has a similar life cycle to the MJO. Convection in the MJO tends to be modulated more by moisture variations compared to the Kelvin wave.

In terms of global phases, wide deep precipitating, anvil, cumulus congestus, and altocumulus types exhibit similar eastward propagation from the Indian Ocean to the central Pacific, while the narrow deep precipitating type only propagates to the Maritime Continent. These propagating types also show coherent Western Hemisphere signals. Generally, negative Western Hemisphere anomalies occur when anomalies are positive over the Indian Ocean.

In both approaches, sampling leads to pictorial renderings of actual clouds across MJO phases. These mosaics provide an objective representation of the cloud field that was unavailable before CloudSat and serve as a reminder to the complex nature of the MJO’s multiscale features.

1. Introduction

The Madden–Julian oscillation (MJO) is the dominant coherent mode of intraseasonal variability in the tropics (Madden and Julian 1972; Zhang 2005). At its most basic level, the MJO can be described as a multiscale envelope of organized convection and cloudiness that moves slowly eastward over the Indo-Pacific at approximately 5 m s−1. Motivation for studying the MJO comes from its profound influence on tropical weather over the Indo-Pacific, monsoon rains over the nearby continents of Asia and Australia, and tropical cyclones throughout the globe, as well as midlatitude teleconnections and influences on the evolution of the El Niño–Southern Oscillation (ENSO) [see reviews by Madden and Julian (1994) and Zhang (2005)]. Despite the MJO’s importance to society, complete understanding of the phenomenon remains elusive as illustrated by the failure of most global circulation models (GCMs) to accurately simulate the MJO (Zhang 2005, and references therein). Furthermore, a comprehensive theory explaining the MJO’s initiation, propagation, and time period is lacking [see Wang (2005) for a theoretical review].

Data from various sources have been used to describe the large-scale dynamical structure of the MJO. These data include model reanalysis products (e.g., Lin et al. 2004; Kiladis et al. 2005; Benedict and Randall 2007), radiosondes (e.g., Lin and Johnson 1996; Lin et al. 2004; Kiladis et al. 2005), and satellite sounders and precipitation retrievals (e.g., Myers and Waliser 2003; Tian et al. 2006; Benedict and Randall 2007; Morita et al. 2006). Results of these studies show that during the suppressed (dry) phases easterly winds predominate, with descending motion and anomalous dry conditions throughout the troposphere. As convection and ascending motion start to build, surface winds become westerly and low-level moistening occurs. During peak convective activity, low-level temperature anomalies are generally cool with warm anomalies through the majority of the upper troposphere and anomalous moisture in the middle troposphere. Following the wet phase, low-level drying ensues and builds upward and descending motion and easterlies return [see Fig. 3 of Benedict and Randall (2007) and Fig. 7 of Zhang (2005)].

With the above knowledge, schematics have been published to visually express convective activity and clouds through the MJO (Fig. 1). All three MJO schematics (Figs. 1a–c) generally show a progression of cloud types from shallow, to middle-topped, to deep clouds, followed by high-topped somewhat thick clouds, with an eventual return to shallow clouds. Interestingly, this progression of cloud types is similar to cloud evolution observed over the life cycle of mesoscale convective systems (MCSs) (Fig. 1d).

Fig. 1.

Schematics of clouds through MJO phases from (a) Lin and Johnson (1996), (b) Benedict and Randall (2007), and (c) Morita et al. (2006). (d) Cloud development through a mesoscale convective system from Zipser et al. (1981).

Fig. 1.

Schematics of clouds through MJO phases from (a) Lin and Johnson (1996), (b) Benedict and Randall (2007), and (c) Morita et al. (2006). (d) Cloud development through a mesoscale convective system from Zipser et al. (1981).

While schematics are helpful at summarizing results, they often lack objectivity. With the advent of CloudSat in spring 2006, direct observations of vertical cloud structure became possible (Stephens et al. 2002). The aim of this paper is to provide an objective view of clouds across the MJO, using CloudSat’s high sensitivity to cloud particles. Documenting cloud structure and its change across varying phases of the MJO will hopefully lead to a better understanding of convective organization within the MJO. Furthermore, we hope to understand the relative importance of different cloud types to different MJO phases.

The novelty of CloudSat is its ability to see midlevel and low-level clouds that are optically obscured by high clouds. Myers and Waliser (2003) and Tromeur and Rossow (2010) used the International Satellite Cloud Climatology Project (ISCCP) dataset to examine cloud structure across the MJO, but neither study was able to capture the vertical structure of clouds or the variation of cloud types in fine detail. Lau and Wu (2010) also did an analysis of cloud type variations across the MJO, but using Tropical Rainfall Measuring Mission (TRMM) data products and therefore focusing on precipitating cloud types. They used two approaches to define MJO phases, local and global, which is similar in spirit to our methods below. However, their analysis of cloud evolution for globally defined phases is limited to the western Pacific, whereas results in section 4 show tropics-wide variations in cloud types through MJO phases.

CloudSat helps fill the gaps of the above studies. For example, Masunaga et al. (2008) compared the vertical structure of reflectivity measured by CloudSat to simulated reflectivity during an MJO event from December 2006 to January 2007. Also, Jiang et al. (2011) used CloudSat to evaluate total cloud fraction and cloud fraction by cloud type during boreal summer intraseasonal variability events.

The next section discusses the data and methods used. Sections 3 and 4 present results of cloud variations across different phases of the MJO. Section 5 provides a discussion of the results, while section 6 summarizes the study.

2. Data and methods

a. CloudSat data

Four years (16 June 2006–15 May 2010) of CloudSat data were used in this study. A three-step process was implemented to study cloud variations across the MJO: 1) CloudSat hydrometeor echo objects were identified using version 5 of the Geometric Profile (2B-GEOPROF) CloudSat product; 2) each echo object was assigned a cloud type and MJO phase; and 3) composites of total cloudiness and of individual cloud types were made for each phase. The 2B-GEOPROF product contains 2D arrays of the radar reflectivity factor measured by the 94-GHz cloud-profiling radar (CPR), as well as a “cloud mask” value ranging between 0 and 40. Higher cloud mask values indicate an increased likelihood of hydrometeor detection (Mace 2004). The spatial resolution of the CPR is 2.5 km along track by 1.4 km across track and 240 m in the vertical.

In Riley and Mapes (2009, hereafter RM09) hydrometeor echoes were analyzed on two tiers: 1) as echo objects (EOs) and 2) as pixels comprising EOs. This study used the same approach. Briefly, EOs were defined as a contiguous region of cloud mask greater than or equal to 20 consisting of at least three pixels with their edges (not merely corners) touching. For each EO we save top and base height, width, and geographical information, as well as several other attributes that were not used in this study.

Following RM09, vertical temperature and moisture profiles were also available for each EO, constructed by averaging temperature and moisture at each altitude in the bounding box region of each EO from CloudSat’s European Centre for Medium-Range Weather Forecasts auxiliary product (ECMWF-AUX), which has been interpolated to the CloudSat grid on a pixel-by-pixel basis.

There are 10 979 780 EOs in the 4 yr of data used here. Because this paper focuses on clouds associated with the MJO, we are only concerned with EOs occurring in the tropics (15°S–15°N), of which there are 2 124 473.

A joint histogram of tropical (latitude < 15°) EO base height versus EO top height reveals natural separations in cloud types (RM09’s Fig. 1a, modified here as Fig. 2a). Low-based EOs lie along the left vertical axis, while layer-type clouds occur along the diagonal. By parsing the histogram into the indicated boxes, seven familiar cloud types are distinguished: 1) deep precipitating, 2) detached anvil (thick cirrus), 3) cirrus, 4) cumulus congestus, 5) alto clouds (including altostratus and altocumulus that are referred to as altocumulus throughout the text), and low clouds, divided into 6) stratocumulus (width > 10 pixels) and 7) cumulus (width ≤ 10 pixels). Each EO is assigned one of these cloud types. Random examples of each EO type are given in Fig. 2b. Our EOs differ from the CloudSat cloud classification product (Wang and Sassen 2007) in that an entire continuous object is assigned a single type. Also, our cirrus definition does not include very thin cirrus, as those clouds contain relatively small particles that go undetected by CloudSat’s millimeter wavelength (Marchand et al. 2008). CloudSat also misses some shallow clouds due to surface contamination (Sassen and Wang 2008).

Fig. 2.

(a) Distribution of horizontal cloud cover by EOs in the tropics (15°S–15°N) accounted for by clouds with tops and bases in the indicated bins for 16 Jun 2006–15 May 2010. Contour values are in units of 103 horizontal echo pixels per bin. Bin size is 240 m × 240 m. Lines and letters delineate EO types, roughly associated with cloud types: A, deep precipitation; B, detached anvil (thick cirrus); C, cirrus; D, cumulus congestus; E, altostratus and altocumulus; and F, low clouds, both stratocumulus (pixel width > 10) and cumulus (pixel width ≤ 10) (adapted from RM09) (b) Random examples of EO types. A subjective number of EOs were chosen to plot for each type, so there is no relation between the different elements of the figure. Type is indicated by transparent color background.

Fig. 2.

(a) Distribution of horizontal cloud cover by EOs in the tropics (15°S–15°N) accounted for by clouds with tops and bases in the indicated bins for 16 Jun 2006–15 May 2010. Contour values are in units of 103 horizontal echo pixels per bin. Bin size is 240 m × 240 m. Lines and letters delineate EO types, roughly associated with cloud types: A, deep precipitation; B, detached anvil (thick cirrus); C, cirrus; D, cumulus congestus; E, altostratus and altocumulus; and F, low clouds, both stratocumulus (pixel width > 10) and cumulus (pixel width ≤ 10) (adapted from RM09) (b) Random examples of EO types. A subjective number of EOs were chosen to plot for each type, so there is no relation between the different elements of the figure. Type is indicated by transparent color background.

b. MJO indices

Two approaches are used to define MJO events and phases. The first defines MJO phases locally relative to minima, maxima, and rate-of-change (i.e., slope) conditions in a space–time filtered outgoing longwave radiation (OLR) time series. Filtered OLR is used because it is a good proxy for tropical deep convection and is familiar in MJO studies (e.g., Nakazawa 1988, 1995; Lau and Chan 1985; Salby and Hendon 1994; Matthews 2008). The OLR data are twice daily on a 2.5° × 2.5° grid from 16 June 2006 to 15 May 2010. The MJO signal was obtained by space–time filtering raw OLR to isolate zonal wavenumbers 0–9 and periods between 30 and 96 days. We averaged the filtered OLR from 15°S to 15°N to obtain a longitude–time array (Fig. 4). This filtered OLR and its local time derivative were both standardized by dividing by their respective standard deviations within the entire array. For reference, the standard deviation of this filtered OLR is 5.82 W m−2.

From these standardized quantities, eight phases were defined based on regions in the scatterplot of Fig. 3a. A pure sine wave would be a unit circle on such a diagram, but because many frequencies are present the points fall along spiral forms. To isolate strong MJO events, we only consider locations and times when the amplitude (distance from the origin in Fig. 3a) is greater than 2 (colored numerals in Fig. 3a). Phase 1 (black numeral 1 symbols) represents the most suppressed phase of the MJO, with the highest filtered OLR values and a time derivative near zero. Opposite phase 1, phase 5 (green numeral 2 symbols) represents the most active phase of the MJO because it has the lowest filtered OLR values and a time derivative near zero. Phases 2–4 are the building phases and phases 6–8 are the decaying phases of the MJO’s convective activity. Perhaps a clearer rendering of the phases is in time–longitude space (Fig. 4). Each color band represents the same phase as the colored numbers in Fig. 3a. Phase 1 (black stripes) runs through maxima in the filtered OLR, phase 5 runs through minima, and phases 2–4 and 6–8 fill in the periods between the maxima and minima. We will refer to the phases defined this way as “pinwheel” phases, as the spiral form taken in Fig. 3a resembles a pinwheel.

Fig. 3.

(a) Pinwheel phases. Colors represent different pinwheel phases (see text for details). (b) WH04 RMM phases. Each colored diamond represents a day in our CloudSat dataset that qualified as part of an MJO event. There is no relation between the colors in the two panels.

Fig. 3.

(a) Pinwheel phases. Colors represent different pinwheel phases (see text for details). (b) WH04 RMM phases. Each colored diamond represents a day in our CloudSat dataset that qualified as part of an MJO event. There is no relation between the colors in the two panels.

Fig. 4.

Filtered MJO OLR in time–longitude space averaged over 15°S–15°N for 15 Jun 2006–15 May 2010. Contour interval is 10 W m−2 starting at 5 W m−2 (solid) and −5 W m−2 (dashed). Colored shading indicates MJO phase for MJO events. Horizontal bars indicate data gaps in CloudSat.

Fig. 4.

Filtered MJO OLR in time–longitude space averaged over 15°S–15°N for 15 Jun 2006–15 May 2010. Contour interval is 10 W m−2 starting at 5 W m−2 (solid) and −5 W m−2 (dashed). Colored shading indicates MJO phase for MJO events. Horizontal bars indicate data gaps in CloudSat.

Each EO in the latitudinal belt 15°S–15°N was assigned an MJO phase by looking up the phase value at its time and mean longitude in the longitude–time array of pinwheel phase. For EOs falling in longitude–time regions with standardized amplitude less than 2, a phase of −1 was assigned, and such EOs were excluded from further analysis.

The second approach defines eight tropics-wide MJO phases using Wheeler and Hendon’s (2004, hereafter WH04) Real-Time Multivariate MJO (RMM) index (http://cawcr.gov.au/staff/mwheeler/maproom/RMM/). These phases are a function of time only, based on a time series of the whole tropical belt’s resemblance to an MJO “mode” structure derived from empirical orthogonal function (EOF) analysis. The RMM1 and RMM2 indices measure the projection of daily 15°S–15°N-averaged longitude sections of OLR, 850-hPa zonal winds, and 200-hPa zonal winds onto the leading pair of EOFs of these same combined fields (with their annual and interannual variability removed) found from historical data. WH04 found that this spatial screening for large-scale structures in this combination of the three variables acts as an effective filter for intraseasonal frequencies associated with the MJO. The first EOF describes enhanced convection over the Maritime Continent, as well as low-level easterlies across the Pacific in conjunction with upper-level westerlies. The second EOF has enhanced convection over the Pacific Ocean and wind patterns, approximately in quadrature to those of the first EOF (Fig. 1 of WH04).

The standardized principal components (PCs) of the leading pair of EOFs are referred to as RMM1 and RMM2. Because RMM1 leads RMM2 by 10–15 days, the MJO appears as a point moving counterclockwise in their joint phase space (Fig. 3b). Labels in each wedge in the diagram roughly indicate the geographical location of enhanced convection associated with the MJO. Each diamond in Fig. 3b corresponds to a day within our CloudSat dataset. Colored diamonds are days on which (RMM12 + RMM22) > 1, qualifying them as part of an MJO event. Each EO is assigned a WH04 RMM phase based only on the day in which it occurred. Again, a phase flag of −1 was used to label EOs on days when the RMM-defined MJO was weak, and such EOs were excluded from further calculations.

3. Results as a function of pinwheel phase

a. EO statistics

Figure 5a shows total horizontal pixel coverage for each EO type by MJO pinwheel phase. Horizontal pixel cover is the number of cloudy (i.e., cloud mask ≥ 20) CloudSat profiles per EO, or simply the pixel width per EO. Figure 5b shows the same information, but normalized by total horizontal pixel cover in each phase bin. Deep precipitating (Dp) EOs have been subdivided into narrow (width < 200 km, red) and wide (width ≥ 200 km, orange) to test the conclusion of Morita et al. (2006) and Tromeur and Rossow (2010) that deep narrow (or isolated) convection is more prevalent during suppressed MJO conditions, while wide convection is more prevalent during active MJO conditions (Fig. 1c).

Fig. 5.

(a) Total horizontal pixel cover and (b) normalized horizontal pixel cover percentage for each EO type per MJO pinwheel phase. (c) Absolute difference between the phase with maximum (max) horizontal pixel cover and the phase with minimum (min) horizontal pixel cover for each EO type. The phase with the max horizontal pixels and min horizontal pixels per EO type is as follows: narrow deep precipitating (ndp) max phase is 4 and min phase is 7; wide deep precipitating (wdp) max phase is 5 and min phase is 1; anvil (an) max phase is 6 and min phase is 3; cirrus (ci) max phase is 6 and min phase is 1; cumulus congestus (cg) max phase is 5 and min phase is 8; altocumulus (ac) max phase is 5 and min phase is 2; stratocumulus (sc) max phase is 6 and min phase is 4; and cumulus (cu) max phase is 6 and min phase is 4. (d) The fractional increase from the minimum horizontal pixels to the maximum horizontal pixels for each EO type.

Fig. 5.

(a) Total horizontal pixel cover and (b) normalized horizontal pixel cover percentage for each EO type per MJO pinwheel phase. (c) Absolute difference between the phase with maximum (max) horizontal pixel cover and the phase with minimum (min) horizontal pixel cover for each EO type. The phase with the max horizontal pixels and min horizontal pixels per EO type is as follows: narrow deep precipitating (ndp) max phase is 4 and min phase is 7; wide deep precipitating (wdp) max phase is 5 and min phase is 1; anvil (an) max phase is 6 and min phase is 3; cirrus (ci) max phase is 6 and min phase is 1; cumulus congestus (cg) max phase is 5 and min phase is 8; altocumulus (ac) max phase is 5 and min phase is 2; stratocumulus (sc) max phase is 6 and min phase is 4; and cumulus (cu) max phase is 6 and min phase is 4. (d) The fractional increase from the minimum horizontal pixels to the maximum horizontal pixels for each EO type.

The most noticeable feature of Fig. 5a is the 1.5-fold increase of total EO coverage from suppressed to active phases. Although the wide deep precipitating contributes most to the EO coverage increase, all types of EOs are slightly more prevalent in the active phase. Figures 5c and 5d explicitly show this increase. Plotted in Fig. 5c is the absolute minimum to maximum change in horizontal pixel cover for each EO type. For example, the cirrus EO type has maximum cover in phase 6 (321 069 horizontal pixels) and minimum cover in phase 1 (208 977 horizontal pixels), giving a difference of 112 092 horizontal pixels (i.e., Fig. 5c, green bar). Since the wide deep precipitating type dominates the absolute change in horizontal pixel cover, Fig. 5d shows the fractional increase in horizontal pixel cover from minimum to maximum phase for each EO type. The wide deep precipitating, anvil, and altocumulus types increase about twofold or more, while the narrow deep precipitating and anvil types increase 1.5-fold, and cumulus congestus, stratocumulus, and cumulus increase less than 1.5-fold from their minimum to maximum horizontal pixel cover. These fractional changes indicate the degree to which the MJO modulates each EO type. In that regard, wide deep precipitating, anvil, and altocumulus types are modulated most, while cumulus congestus, stratocumulus, and cumulus are modulated least (i.e., fractional change < 1.5).

All EO types have maximum absolute horizontal pixel cover in phase 4, 5, or 6 (see Fig. 5 caption). The wide deep precipitating, cumulus congestus, and altocumulus EO types have their maximum horizontal pixel cover in phase 5 (coincident with the minimum in OLR), while the anvil, cirrus, stratocumulus, and cumulus EO type maxima trail the OLR minimum by one phase (phase 6) and the narrow deep precipitating type maximum leads the OLR minimum by one phase (phase 4). Each EO type’s minimum horizontal pixel cover occurs either before or after the active phases (phases 4–6), except for the shallow types (i.e., stratocumulus and cumulus). That anvil and cirrus maxima trail the wide deep precipitating maximum by one MJO phase is consistent with the well-established picture of convective cloud type evolution across various time and space scales: shallow to deep to high clouds (e.g., Zipser et al. 1981; Lin and Johnson 1996; Mapes et al. 2006; Morita et al. 2006; Benedict and Randall 2007; Kiladis et al. 2009).

The above results are mostly consistent with those of Lau and Wu (2010), who showed increased occurrence in all their precipitating and nonprecipitating cloud types during active phases, except for warm low clouds (their Figs. 6 and 7). Since their cloud types are defined by brightness temperature from TRMM’s visible and infrared scanner (VIRS) and TRMM’s Precipitation Radar (PR), their low clouds during active phases could have been obscured to the VIRS by the abundance of higher-level clouds during active phases and/or gone undetected by the TRMM PR threshold (dBZ ≥ 17; Kummerow et al. 1998) since the low clouds may not be raining or may be just drizzling.

In normalized terms (Fig. 5b), narrow deep precipitating EOs gradually increase their percentage contribution from phase 1 to 3 and then slowly decrease through phase 7, suggesting a somewhat different cloud behavior than in Morita et al. (2006) and Tromeur and Rossow (2010). Here the narrow deep convection is modulated according to building versus trailing phases compared to active versus suppressed phases. Anvil EOs have their greatest fractional contribution after the active phases, and their smallest during active phases. Shallow EO types (stratocumulus and cumulus) have their greatest fractional contribution during the suppressed phases (1, 2, 6, 7, and 8).

Because our EO analysis retains links to pixels, we can generate pictorial realizations of “actual” (i.e., CloudSat observed) clouds (EOs) across the MJO (e.g., Fig. 6). The aim of our pictorial realizations is to mimic schematics from previous works (e.g., see Fig. 1 and the references given there), but with objective methods and CloudSat-observed clouds. To make a pictorial realization, random samples are drawn from a random number generator from the set of MJO EOs and rendered as a set of pixels centered over the EO’s continuous pinwheel phase value (see the caption of Fig. 6 for how continuous phase relates to discrete phase). The conversion factor between EO width (in pixels) and width in this diagram (in phase units) is an arbitrary choice, adjusted manually for graphical clarity. Also for clarity, pictorial realizations sample only from EOs with amplitude greater than 3 (as opposed to a threshold of greater than 2 for all other statistics). The random selection and plotting of EOs is repeated until the desired amount of total horizontal echo coverage is reached. One final nod to clarity: wide Dp EOs are rendered in a gray-to-red reflectivity color scheme, while the pixels making up all other types are in a gray-to-blue scheme. The faintest discernible shade of red or blue occurs near 0 dBZ.

Fig. 6.

One example of a pictorial realization of CloudSat clouds across continuous pinwheel phases of the MJO. Each EO is plotted centered over its continuous phase, which is multiplied by 500 (the bottom x axis) to maximize viewing clarity. The top x axis corresponds to the actual EO’s continuous phase. Continuous phase relates to discrete phase as follows: EOs with 0.5 ≤ continuous phase < 1.5 are in discrete phase 1, 1.5 ≤ continuous phase < 2.5 are in discrete phase 2, etc., and continuous phase < 0.5 and continuous phase ≥ 7.5 are in discrete phase 8. Wide Dp EOs are gray to red, while all other EO types are gray to blue. Colored shading starts at dBZ > 0. The seed is the number inputted to the random number generator to recreate this example.

Fig. 6.

One example of a pictorial realization of CloudSat clouds across continuous pinwheel phases of the MJO. Each EO is plotted centered over its continuous phase, which is multiplied by 500 (the bottom x axis) to maximize viewing clarity. The top x axis corresponds to the actual EO’s continuous phase. Continuous phase relates to discrete phase as follows: EOs with 0.5 ≤ continuous phase < 1.5 are in discrete phase 1, 1.5 ≤ continuous phase < 2.5 are in discrete phase 2, etc., and continuous phase < 0.5 and continuous phase ≥ 7.5 are in discrete phase 8. Wide Dp EOs are gray to red, while all other EO types are gray to blue. Colored shading starts at dBZ > 0. The seed is the number inputted to the random number generator to recreate this example.

Figure 6 is one example of a pictorial realization (for others, see online at http://www.rsmas.miami.edu/users/eriley/Emily/JAS2011_Supplementary_Figs6.html). The example here highlights some of the statistics in the bar graphs (Fig. 5): There is more cloudiness during the active phases (phases 4–6), with the wide Dp EOs dominating these phases and completely absent in the most suppressed continuous phases (phases 0–2 and 7–8). Narrow Dp EOs are seen during both suppressed and active conditions (i.e., the tall blue towers near continuous phases 0.3, 1.7, 3.1, 4.8, 7.3, and 8). Shallow EOs are prevalent during all phases. The statistical variation of the other EO types is difficult to discern in this pictorial realization. However, in terms of the main features, Fig. 6 is quite comparable to Morita et al’s (2006) hand-drawn Fig. 14 (reproduced here as Fig. 1c) but has the virtue of objectivity.

b. Pixel-level statistics

The echoes in each pinwheel phase of the MJO can also be characterized with 2D histograms of pixel dBZ versus height [also known as contoured frequency by altitude diagrams (CFADs); Yuter and Houze 1995]. It is clearest to establish the mean or background cloudiness of MJO-affected regions, then examine anomalies relative to that. Figure 7a shows the normalized CFAD (NCFAD) for all EOs in all pinwheel phases of the MJO wherever pinwheel amplitude is greater than 2. The greatest number of pixels occurs around −25 dBZ between 11 and 14 km. Pixels are also common near the rainfall attenuation line, with high reflectivities below 5 km (white line in Fig. 7a). A broad minimum in the NCFAD occurs at reflectivities below 0 dBZ at midlevels (between 3 and 8 km). Figures 7b–i show each pinwheel phase’s NCFAD anomalies, overlaid by the background contours from Fig. 7a for reference.

Fig. 7.

(a) Normalized CFAD (2D histogram of pixel dBZ vs pixel height) for all eight pinwheel phases. White line indicates the approximate rainfall attenuation line. (b)–(i) Shading is the percentage difference between the normalized CFAD of the indicated phase and the normalized CFAD for all eight pinwheel phases [in (a)]. The black contours in each panel are the same as the black contours in (a).

Fig. 7.

(a) Normalized CFAD (2D histogram of pixel dBZ vs pixel height) for all eight pinwheel phases. White line indicates the approximate rainfall attenuation line. (b)–(i) Shading is the percentage difference between the normalized CFAD of the indicated phase and the normalized CFAD for all eight pinwheel phases [in (a)]. The black contours in each panel are the same as the black contours in (a).

During suppressed phases 1 and 2 (Figs. 7b,c), the most enhanced echoes (relative to the all-phase mean echo) have low-reflectivity values, both at low levels (z < 5 km) (the signature of nonprecipitating cumulus clouds) and mid- to upper levels (9 km < z < 13 km) (the signature of cirrus, as detailed in Fig. 3.21 of Riley 2009). Additionally, in phase 2 high-reflectivity upper-level enhancements emerge. By phase 3 (Fig. 7d), high reflectivities above the freezing level, suggesting intense updrafts, are well established, along with enhanced echo frequencies along the low-reflectivity and high-altitude margins of the background NCFAD. Low-level, high-reflectivity enhancements also appear in phase 3, indicating rainfall. In phase 4 (Fig. 7e), high-reflectivity enhancements increase, indicating deep convection and rainfall. Midlevel enhancements also appear in phase 4, as well as phase 5. In phase 5 (Fig. 7f), the highest reflectivities (dBZ > 10) above the freezing level diminish, suggesting less intense updrafts. The results during the active phases 4 and 5 are consistent with Fig. 4 of Masunaga et al.’s (2008) ,CloudSat CFAD differences from the wet and dry phases of the December 2006–January 2007 MJO event (where they defined wet and dry by rainfall measured from the TRMM PR). In phase 6 (Fig. 7g), rain-related enhancements are nearly gone. Midlevel enhancements remain over the breadth of the reflectivities, while upper-level enhancements below −10 dBZ shift down about 1 km. By phase 7 (Fig. 7h), rain-related enhancements disappear completely. What remains is a wedge of enhanced frequency from 5 to 12 km with moderate reflectivities along the base of the upper-level mode of the climatological distribution. These enhancements are a signature of tropical anvil NCFAD enhancements, as in Fig. 3.21 of Riley (2009). Finally, enhancements in phase 8 (Fig. 7i) are similar to those in phase 1, indicating a return to suppressed conditions.

Looking more closely at reflectivity enhancements, Fig. 8 shows the mean NCFAD and anomalous CFADs during phases 3 and 7 for just wide deep precipitating EOs. The most common pixels in the all-phase mean NCFAD are at low levels (z < 5 km) and high reflectivities (dBZ > 0) (Fig. 8a). During phase 3, the wide deep precipitating EOs have reflectivity enhancements at low levels and high reflectivities, as well as along the upper margins of the background NCFAD (Fig. 8b). Phase 7 contrasts almost perfectly with phase 3: positive enhancements are in a wedge from 5 to 14 km and −25 to 5 dBZ with negative anomalies elsewhere (Fig. 8c). This contrast suggests that wide deep precipitating EOs prior to the most active phase (i.e., phase 5) are more bottom heavy and may contain attached precipitating midlevel EOs, while after phase 5 the wide deep precipitating EOs contain more attached anvil.

Fig. 8.

(a)–(c) As in Figs. 7a, 7d, and 7h, respectively, except for wide deep precipitating (wd) EOs only. (d) Vertical normalized percentage profile of echo cover for wide deep precipitating (wdp; solid line) and narrow deep precipitating (ndp; dashed line) EOs. The total of each profile is 100%. The profiles are made by summing over the reflectivity dimension of the ndp and wdp NCFAD.

Fig. 8.

(a)–(c) As in Figs. 7a, 7d, and 7h, respectively, except for wide deep precipitating (wd) EOs only. (d) Vertical normalized percentage profile of echo cover for wide deep precipitating (wdp; solid line) and narrow deep precipitating (ndp; dashed line) EOs. The total of each profile is 100%. The profiles are made by summing over the reflectivity dimension of the ndp and wdp NCFAD.

Anomalous phase CFADs for narrow deep precipitating EOs are noisy and not shown. However, the all-phase mean NCFAD is useful to compare with the wide deep precipitating all-phase mean NCFAD. Summing each NCFAD over its reflectivity dimension yields the vertical normalized percent profile of echo cover (Fig. 8d). The total of each profile is 100%. The narrow deep precipitating profile (dashed line) is more bottom heavy than the wide deep precipitating profile (solid line), suggesting that the greater fractional contribution by narrow deep precipitating EOs during the building phases (phases 1–3) (Fig. 5b) is weighted more by bottom-heavy deep convective cloud systems.

Figure 9a shows total echo coverage across MJO pinwheel phases. The unnormalized CFAD for each MJO pinwheel phase was summed over the reflectivity dimension (as above) and divided by the total number of CloudSat-sampled profiles during that phase of the MJO to give echo coverage. Echo cover at all heights roughly doubles from suppressed (phases 1–3) to active (phases 4–6) conditions, while the height of the highest echoes rises about 1 km. These changes are consistent with Figs. 4 and 5, indicating the prevalence of wide intense deep convection in active phases. Each phase has two prominent echo cover peaks, one centered around 1.5 km and the other near 12 km. During active phases (4–6), the difference between the upper- and lower-level peaks is greatest. In suppressed conditions (phases 1–2, 8), the two peaks are relatively comparable. Figure 9a also shows that upper-level (10–13 km) cloudiness is slightly greater after active phases than before (i.e., phase 7 vs phase 3), again illustrating the prevalence of anvil-type EOs there, consistent with Figs. 5b and 7h,i.

Fig. 9.

MJO (a) total echo cover, (b) temperature anomalies (deviations from the all-phase mean), and (c) specific humidity anomalies for pinwheel phases. Solid (dashed) contours are positive (negative). The zero difference contour is thick.

Fig. 9.

MJO (a) total echo cover, (b) temperature anomalies (deviations from the all-phase mean), and (c) specific humidity anomalies for pinwheel phases. Solid (dashed) contours are positive (negative). The zero difference contour is thick.

c. ECMWF temperature and moisture

Figure 9b shows temperature anomalies during each pinwheel phase of the MJO. This figure was constructed by simple averaging of the T(z) profiles associated with all the EOs in each phase bin, then subtracting from each phase bin the average profile over all eight phase bins. In other words, the horizontal average of Figs. 9b and 9c at each altitude is zero by construction. The coherence of anomalies in adjacent phase bins indicates their robustness, and the main features here are consistent with previous studies (discussed below), so we have not made a careful estimation of statistical significance and will not attempt to attach interpretation to subtle features.

In the active phases (4–6), the lowest kilometer of the troposphere is cool. This simultaneous relationship apparently reflects the effects of rain evaporation and convective downdrafts (e.g., Lin and Johnson 1996; Kemball-Cook and Weare 2001), since sea surface temperature tends to lag convective activity (Sperber 2003). The 1–5-km layer has much weaker variations, lagging behind the planetary boundary layer (PBL) by perhaps one phase category. The 5–14-km layer varies coherently and clearly leads convection by a whole phase bin with opposite-signed fluctuations above that up to about 18 km, tilting forward (toward earlier phase) with height, consistent with upward propagating gravity waves excited by a moving heat source (Kiladis et al. 2001, 2005; Tian et al. 2006). Schwartz et al. (2008) found a similar eastward tilt (where eastward is comparable to our forward tilt) in water vapor and temperature anomalies from Aura’s Microwave Limb Sounder (MLS) and the Goddard Earth Observing System (GEOS) data assimilation system satellites, respectively. Virts and Wallace (2010) also detected an eastward tilt in Cloud–Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO)-detected tropical tropopause transition layer (TTL) cirrus. The range of variations is about 0.5 K for this unweighted composite of all fluctuations with pinwheel amplitude greater than two standard deviations (i.e., twice the 5.82 W m−2 standard deviation in filtered OLR).

The temperature structure of Fig. 9b is consistent with previous studies (e.g., Lin and Johnson 1996; Kiladis et al. 2005; Tian et al. 2006; Benedict and Randall 2007). In each of those studies, a cool–warm–cool vertical tripole structure was found during the active periods of the MJO. Before the active period both Kiladis et al. (2005) and Benedict and Randall (2007) showed anomalous low-level tropospheric warmth. After the active period these studies generally showed cooling through most of the troposphere.

Figure 9c shows specific humidity anomalies during each pinwheel phase of the MJO, constructed in the same manner as the temperature profiles. Generally, the troposphere is anomalously dry through all depths during the suppressed phases (i.e., phases 1 and 8) and moist during the active phases (i.e., phases 4 and 5), but there is systematic rearward tilt (toward later phases with height) to the anomalies. In phase 2, anomalously moist conditions appear in the lowest kilometer of the troposphere, while conditions remain dry aloft. The moisture anomaly deepens to 6 km at phase 3 and phases 4 and 5 are anomalously moist through the entire depth of the troposphere. Phases 6, 7, 8, and 1 are almost precisely the opposite of 2, 3, 4, and 5. Again the anomalies shown here are consistent with previous work (e.g., Kemball-Cook and Weare 2001; Sperber 2003; Myers and Waliser 2003; Kiladis et al. 2005; Tian et al. 2006; Benedict and Randall 2007). A comparison with Kelvin wave anomalies is done in section 5 (see Fig. 13) to help put these MJO anomalies in the context of other wave types.

4. Results: WH04 RMM phases

A limitation of the above analysis is that it does not convey how changes in the geographic distribution of cloudiness are associated with the MJO. This section therefore describes results using the more geographically discriminating RMM index of WH04. Figure 10 shows cloud cover in eighteen 20° longitude bins across the tropics during each of the eight RMM phases when amplitude is greater than one. The line plot in each panel shows the mean total horizontal pixel cover over all eight phases for each longitude bin. Here the background longitudinal structure pervades all RMM phases. Generally, cloud cover is higher over the Maritime Continent and western Pacific versus the eastern Pacific and Atlantic Oceans, although the extent to which this is true is modulated by MJO phase. Echo cover is greater over Central and South America (longitude bin 300°) versus the nearby eastern Pacific and Atlantic Oceans. Shallow clouds (i.e., stratocumulus plus cumulus EO types; black and purple) are more common in the eastern Pacific Ocean (longitude bins 220°–280°) than all other longitude bins. Deep precipitating EO types (both wide and narrow; orange and red) make their greatest percentage contribution to total cloud cover over the Maritime Continent, western Pacific Ocean, and South American longitude bins. One persistent feature through all RMM phases is the presence of the stratocumulus decks off the coast of South America (purple slab in longitude bin 280°, showing about 20%–25% coverage).

Fig. 10.

Horizontal pixel (hp) cover for each EO type in 20° longitude bins across the tropics during each of the eight WH04 RMM phases. The longitude values represent the center value of each 20° longitude bin. Total hp cover is the number of cloudy profiles divided by the total (both clear and cloudy) profiles. The mean total hp cover over all eight phases for each longitude bin is represented by the line plot in each panel. (bottom) A geographical map of the tropics is provided for reference.

Fig. 10.

Horizontal pixel (hp) cover for each EO type in 20° longitude bins across the tropics during each of the eight WH04 RMM phases. The longitude values represent the center value of each 20° longitude bin. Total hp cover is the number of cloudy profiles divided by the total (both clear and cloudy) profiles. The mean total hp cover over all eight phases for each longitude bin is represented by the line plot in each panel. (bottom) A geographical map of the tropics is provided for reference.

Perhaps more interesting than these cloud cover commonalities is the differences from phase to phase. Going from phase 2 to 8, the longitude of maximum cloud cover propagates from the Indian Ocean, over the Maritime Continent, into the western Pacific, and finally to South America. In phases 2 and 3, cloud cover is greatest over the eastern Indian Ocean (longitude bins 80° and 100°) and steadily increases over the Maritime Continent (longitude bin 120°). Phase 4 shows a broadening of the maximum cloud cover envelope from the Maritime Continent to the western Pacific (longitude bins 120°–160°). These longitude bins remain above average through phases 5 and 6. In phase 6, the maximum cloud cover envelope shifts farther eastward into the Pacific, extending from longitude bins 120° to 180°. Concurrent with this increase in cloud cover over the Maritime Continent and western Pacific is a decrease in cloud cover over the Indian Ocean (longitude bins 60° and 80°). Phase 6 is the cloudiest phase tropics-wide, with above-average cloud cover from the Maritime Continent eastward to South America (longitude bins 120°–340°, excluding longitude bin 280°). By phases 7 and 8, cloud cover over the Maritime Continent and western Pacific wanes. Phase 8 shows increased cloudiness over the eastern Atlantic, Africa, and the Indian Ocean (longitude bins 360° and 20°–100°).

Western Hemisphere cloudiness also appears to be modulated by the MJO, at least by this EOF-based definition. Cloud cover is largest over the eastern Atlantic and West Africa in phases 1, 2, and 8, when the cloud cover is relatively large over the Indian Ocean but small over the western Pacific compared to other phases. These results are consistent with previous studies that found a Western Hemisphere response to the MJO in temperature, wind, and precipitation (Salby and Hendon 1994; Bantzer and Wallace 1996). More recently, Martin and Schumacher (2011) found positive precipitation anomalies over the Caribbean during WH04 RMM phases 1 and 2 and negative anomalies during phases 5 and 6, consistent with our Fig. 10.

Another way to display the same information is as contour plots of cloud cover anomalies (relative to the all-phase mean) for each EO type separately (Fig. 11). The wide deep precipitating EO type anomalies account for much of the eastward propagation of the MJO (Fig. 11a; note that the color scale is different in Fig. 11a than in Figs. 11b–i). Positive anomalies start over Africa, the Indian Ocean, and Indonesia in phases 1–3, move over the western Pacific in phases 4–6, and then finally extend into the central Pacific in phases 7 and 8. In phases 7 and 8 the positive anomalies also sprawl the South American continent and Atlantic Ocean where they linger into phases 1 and 2. There is a noticeable break in the eastward propagating positive anomalies near 100° that is also present in the anvil, cirrus, cumulus congestus, and altocumulus types (Figs. 11c–f). This break or “jump” in cloud cover has been seen in previous observational work (e.g., Knutson and Weickmann 1987; Maloney and Hartmann 1998) and a modeling study by Newman et al. (2009) and is likely associated with topographical influences on convection over the Maritime Continent (Wu and Hsu 2009).

Fig. 11.

(a)–(h) Anomalous horizontal pixel cover for each EO type. The color scale for (a) goes from −10% to +10% while those for (b)–(h) range from −5% to +5%. In (a) [(b)–(h)] each contour level is 1% (0.5%) with positive anomalies shown as solid lines and negative anomalies as dashed lines. The thick solid line is the zero difference contour. The longitude values represent the middle value of individual 20° longitude bins. The diagonal line in each panel represents the propagation of the wide Dp EO type. (i) The 1–4-km temperature anomaly associated with all EO types. Contour interval is 0.1°C. (bottom) A map of the tropics from 15°S to 15°N is provided for reference.

Fig. 11.

(a)–(h) Anomalous horizontal pixel cover for each EO type. The color scale for (a) goes from −10% to +10% while those for (b)–(h) range from −5% to +5%. In (a) [(b)–(h)] each contour level is 1% (0.5%) with positive anomalies shown as solid lines and negative anomalies as dashed lines. The thick solid line is the zero difference contour. The longitude values represent the middle value of individual 20° longitude bins. The diagonal line in each panel represents the propagation of the wide Dp EO type. (i) The 1–4-km temperature anomaly associated with all EO types. Contour interval is 0.1°C. (bottom) A map of the tropics from 15°S to 15°N is provided for reference.

The narrow deep precipitating type also shows eastward propagating positive anomalies that lead the wide deep precipitating anomalies by 1–2 phases. Unlike the other EO types (except for cumulus), though, the positive anomalies only traverse the Indian Ocean and Maritime Continent, petering out over the western Pacific during phase 5. Also unique to the wide deep precipitating type (and cumulus type Fig. 11h), the positive anomalies sprawl the entire tropics when enhancements are positive over the Indian Ocean and Maritime Continent, although the pattern becomes weaker and noisier in the Western Hemisphere. A possible explanation for the tropics-wide positive enhancements is given below.

The anvil EO type anomaly pattern resembles the wide deep precipitating EO cloud cover anomaly pattern (cf. Figs. 11a,c), but shifted westward (or later in time) by mean tropical easterlies (or a long lifetime after convection, as seen in section 3). Unlike the wide deep precipitating type, the anvil type has a secondary band that propagates eastward from the central Pacific to South America. The Atlantic Ocean (~320°–360°) anomaly pattern is opposite the Indian Ocean and Maritime Continent with generally negative anomalies during phases 1–5 and positive ones during phases 6–8.

The cirrus EO type (Fig. 11d) shows eastward propagation of positive anomalies similar to the wide deep precipitating and anvil types. Positive cirrus cloud cover anomalies start over Africa and the Indian Ocean in phases 1 and 2, spread into the western and central Pacific Ocean during phases 3 and 4, and extend from the Maritime Continent to the eastern Pacific Ocean during phases 5 and 6. By phases 7 and 8, positive anomalies are mainly over the Western Hemisphere, in agreement with Virts and Wallace (2010), who found CALIPSO-detected TTL cirrus over South America and Africa in the late WH04 RMM phases. Comparison with CALIPSO results should be taken cautiously, though, as the cirrus Virts and Wallace (2010) discuss has base heights above 15 km, which go mainly undetected by CloudSat (see Fig. 2a).

The cumulus congestus EO type (Fig. 11e) cloud cover anomalies are similar to the wide deep precipitating and anvil types, although the congestus stops propagating eastward around phase 6, where the positive anomalies extend into the central Pacific. Similar to the anvil and cirrus types, there are generally negative anomalies over South America and the Atlantic in phases 1–5 and positive afterward. The altocumulus (ac) EO type cloud cover anomalies (Fig. 11f) propagate approximately in phase with the wide deep precipitating bands and have a Western Hemisphere signal similar to other EO types.

The stratocumulus have a weaker eastward propagating positive signal from the Indian to central Pacific Ocean that slightly lags the wide deep precipitating type (Fig. 11g). Also, anomalies over the eastern Pacific (240°–300°) and Atlantic (340°–360°) are positive during phases 2–6 and negative otherwise. The cumulus EO type exhibits a zonal anomaly structure (Fig. 11h) very much like the narrow deep precipitating type (Fig. 10b), suggesting that tropics-wide vertical instability for cumulus convection may be modulated by the MJO’s zonal mean temperature signal (Fig. 11i).

Figure 11i shows the 1–4-km temperature anomaly associated with all EO types across the RMM phases. The depth 1–4 km is chosen since Tulich and Mapes (2010) showed that convection was most sensitive to temperature and moisture variations below 4 km. During phases 1 and 2, temperatures tropics-wide are anomalously cool, in line with increased instability and anomalously positive cumulus convection (Figs. 11b,h). From phase 3 to 8, warm temperature anomalies extend from the Maritime Continent eastward, consistent with suppressed cumulus during these phases.

The information from Fig. 11 is used to make pictorial realizations of the MJO’s anomalous cloudiness across the tropics as a function of RMM phase. Akin to Fig. 6’s purpose, the aim now is to mimic Madden and Julian’s (1972) famous schematic (their Fig. 16, reproduced here as Fig. 12b), but again with objective methods and actual CloudSat observations of cloud structure. To make the pictorial realizations, EOs are randomly selected from RMM phase EOs and plotted centered over their true mean longitude (where longitude has been multiplied by 40 to maximize viewing clarity). Here the random selection is done as follows: wherever anomalous cloud cover in Fig. 11 is positive, random EOs are selected from the set of EOs belonging to that longitude bin, RMM phase, and EO type. The amount of cloudiness plotted per EO type is proportional to the anomaly strength in Fig. 11 (actually slightly more, since EOs are drawn until the desired amount of echo cover is reached and in practice exceeded).

Fig. 12.

(a) One example of a pictorial realization of EOs across the tropics during each WH04 RMM phase. Phases run downward from 1 to 8. Each stamp is centered over its true mean longitude × 40. The longitude scale (x axis) has been multiplied by 40 to maximize viewing clarity. (bottom) A map of the tropics from 15°S to 15°N is provided for reference to an EO’s true mean longitude. (b) Schematic of the tropics-wide cloud evolution and circulation associated with the MJO (Madden and Julian 1972).

Fig. 12.

(a) One example of a pictorial realization of EOs across the tropics during each WH04 RMM phase. Phases run downward from 1 to 8. Each stamp is centered over its true mean longitude × 40. The longitude scale (x axis) has been multiplied by 40 to maximize viewing clarity. (bottom) A map of the tropics from 15°S to 15°N is provided for reference to an EO’s true mean longitude. (b) Schematic of the tropics-wide cloud evolution and circulation associated with the MJO (Madden and Julian 1972).

Figure 12a is an example of one pictorial realization of clouds for RMM phases. With a different seed to the random number generator, another realization can be made. We have made several (see online at http://www.rsmas.miami.edu/users/eriley/Emily/JAS2011_Supplementary_Figs12.html) and selected one based on aesthetic criteria and consistency with the statistical results. As in Fig. 5, wide deep precipitating types are gray to red, while all other EO types are gray to blue, with color appearing at 0 dBZ. The eastward propagating band of wide deep precipitating is readily seen (red EOs), and the associated anvil type can be noticed. The tropics-wide enhancement of narrow deep precipitating types (tall blue columns) in phases 1–3 is also discernible. Other EO types are certainly present, but their small fractional contribution to total cloudiness makes them harder to evaluate in the pictorial realization.

Comparing this modern diagram to the Madden–Julian schematic of 1972 (Fig. 12b) is instructive. First, it must be acknowledged that they did a remarkable job of inferring cloudiness anomalies based only on spectral analysis of surface data. Their cumulus cartoon icons correspond best to our wide deep precipitating mesoscale convective systems, which carry the bulk of the MJO signal. The slightly lagging anvil clouds, related to the MJO’s tilted structure of moisture (Figs. 9b,c and 11) and heating (Lin et al. 2004), along with the zonal structure of shallow and narrow deep cumulus convection, are, however, more than their methods could have detected.

5. Discussion

In terms of pinwheel phases, the figures in this paper provide evidence for the validity of the schematics in Fig. 1. In those schematics, cloud structure and evolution were inferred from thermodynamic, dynamic, and precipitation data. Here, the ability of CloudSat to detect cloud particles augments previous work by providing a direct view of the vertical structure of cloud types across the MJO.

The likely importance of low clouds followed by mid-topped convection at preconditioning the environment prior to deep convection in the most active phase of the MJO has been highlighted in several studies (e.g., Kemball-Cook and Weare 2001; Kiladis et al. 2005; Tromeur and Rossow 2010; Lau and Wu 2010; Jiang et al. 2011). Figures 5 and 7 indicate that when MJO phases are defined locally (i.e., the pinwheel phases), shallow clouds (both stratocumulus and cumulus) have their biggest influence (i.e., largest fractional contribution) on cloud fields prior to MJO active phases, although total low-cloud amount is greater in the active phases. Perhaps in locally defined phases they are indeed critical to the low-level warming and moistening (seen in Fig. 9) ahead of peak MJO activity. However, mid-topped convection (i.e., cumulus congestus EO type) does not show much variation prior to active MJO phases as defined locally. Cumulus congestus fractional contribution peaks during phase 1 and then decreases through phase 5 (Fig. 5b), leaving their explicit role in moistening the environment prior to MJO active phases unclear from these results. However, some cumulus congestus cells are part of wide deep precipitating systems (Fig. 2b), which were shown to be more bottom heavy prior to the most active MJO phase (Figs. 8b,c), perhaps indicating more attached midlevel convection prior to versus after active phases.

Contrary to the low-level cloud signal in the pinwheel phases, the stratocumulus EO type positive anomalies trail the wide deep precipitating signal in the WH04 RMM phases (Fig. 11g). Why there is a difference between the role low-level clouds play in locally versus globally defined phases is unclear, but it is intriguing and perhaps warrants future study.

Changes in the type of deep convection across MJO events, as seen in Figs. 5 and 7, suggest that the organization of convection is being modulated by MJO phase. The MJO transitions from deep, narrow, less-organized convection in the first three phases to widespread, more-organized convection during active phases, to more anvil and stratiform in the latter phases, as indicated by fractional contribution of narrow deep precipitating, wide deep precipitating, and anvil EO types to total cloudiness in Fig. 5. High reflectivities above 0°C during phases 3 and 4 imply strong updrafts capable of lofting large particles, while phase 5 shows midlevel enhancements along with high-reflectivity low-level rainfall enhancements, suggesting more widespread, less intense rainfall (Fig. 7). Tromeur and Rossow (2010) offer an explanation for this change in convective organization: the large-scale wave moistens the lower troposphere, allowing a transition from smaller-scale, less organized convection to larger-scale, more-organized convection. The effects of this convective reorganization could then feed back on the larger scale (Moncrieff 2004). A change in convective scale may also indicate changes in updraft entrainment across the MJO; as Bacmeister and Stephens (2011) pointed out, “it seems intuitively reasonable that as the size of a convective cloud increases the bulk fractional entrainment of environmental air into the cloud will decrease” (p. 15). A method developed by Luo et al. (2010) to estimate buoyancy and entrainment from A-Train observations may help identify such entrainment changes across the MJO but is not used here.

To help put MJO cloud and thermodynamic variations in the context of other tropical waves, MJO pinwheel phase results are compared to convectively coupled Kelvin pinwheel phase analysis (Fig. 13). EOs are assigned a Kelvin pinwheel phase using the same method as the MJO pinwheel phases (section 2), except using filtered Kelvin OLR [i.e., the Kelvin box in Fig. 6b of Wheeler and Kiladis (1999)] between 10°S and 10°N. For strict comparison, Fig. 13 is based on MJO and Kelvin events with amplitudes between 2 and 3, which have a standard deviation of OLR of 10.0 and 11.6 W m−2, respectively. (MJO results are similar to Fig. 9 with amplitude greater than 2.)

Fig. 13.

(a)–(c) As in Fig. 9, but for the Kelvin wave with amplitudes between 2 and 3. (d) MJO and Kelvin wave difference in EO cover, where the MJO events selected for this plot had amplitudes between 2 and 3. Solid (dashed) contours are positive (negative). The zero difference contour is thick. See text for more details.

Fig. 13.

(a)–(c) As in Fig. 9, but for the Kelvin wave with amplitudes between 2 and 3. (d) MJO and Kelvin wave difference in EO cover, where the MJO events selected for this plot had amplitudes between 2 and 3. Solid (dashed) contours are positive (negative). The zero difference contour is thick. See text for more details.

The Kelvin wave echo cover structure (Fig. 13a) is similar to the MJO (Fig. 9a) in that both wave types have two prominent echo cover peaks near 1.5 and 12 km. Differences between the two waves are shown in Fig. 13d. In phase 1, the MJO has greater cloudiness from about 2 to 14 km. Cool colors in phases 2–5 show that the MJO has less cloudiness compared to the Kelvin wave. Warm colors during phases 1 and 6–8 show the converse cloud signal. Specifically, the lower- and upper-level echo cover peaks are cloudier during phases 6–8 for the MJO compared to the Kelvin wave.

The thermodynamic structure of the Kelvin wave also differs from that of the MJO. In Fig. 9, the MJO temperature anomaly structure tilts weakly backward (toward later phases) through the troposphere, while the moisture anomalies show a comparatively stronger backward tilt. By comparison, the Kelvin wave temperature structure is more tilted in the troposphere and its moisture anomalies are about 3 km lower than the MJO. Furthermore, Kelvin anomalies lead MJO anomalies by about one phase, with an exception of low-level (z ≤ 1 km) temperature anomalies, which show similar patterns in phases 1–6. Additionally, for a comparable temperature change across phases, the MJO has a greater change in moisture at low levels (z < 5 km) than the Kelvin wave (cf. Figs. 13b,c and 9b,c), in line with results of Mapes et al. (2006) and Tulich and Mapes (2010), who found greater moisture to temperature ratios for intraseasonal variations compared to shorter time scales (<2 days). In terms of absolute temperature differences, low-level temperature variations in the Kelvin wave are about 4 times greater than those in the MJO. Perhaps the time scale of these two wave types offers clues to the phase and height discrepancies in the moisture field. If the Kelvin phases are thought of as days and the MJO phases as pentads, then the MJO frequency resonates more with the residence time of water in the atmosphere (~9 days), giving more time to build moisture vertically in the MJO versus Kelvin wave.

The relative importance of moisture to temperature sheds light on underlying dynamical differences between the MJO and Kelvin waves. For the Kelvin wave, low-level temperature variations are relatively more important than moisture, suggesting that convection is more controlled through temperature effects on buoyancy. However, for the MJO, variations in moisture seem to have greater impacts on convection. Further evidence for this view is given in Yasunaga and Mapes (2011), which shows that the MJO modulates small TRMM echo objects, the convective fraction of rainfall, and precipitable water more than Kelvin waves of the same amplitude in rainfall. Their results are interpreted as indicating that convection in divergent wave types (e.g., Kelvin waves) is controlled more by convective inhibition (CIN), while rotational wave types (e.g., the MJO) have more impact through moisture. Raymond and Fuchs (2007, 2009) model results also support this idea. In Raymond and Fuchs (2007) two types of unstable modes are described, a slow-moving “moisture mode” and a fast-moving “gravity mode” or CIN-governed instability. They link the latter to the Kelvin wave and, in their 2009 paper, the former to the MJO.

Despite Kelvin wave and MJO differences that may imply different convective coupling mechanisms, their similarities indicate scale invariance to organization of cloud morphology (i.e., shallow to deep to stratiform). Overall, the MJO and Kelvin waves show similar cloud evolution (i.e., Kelvin EO type evolution is very similar to Fig. 5). More broadly, the evolution of predominant cloud types in the life cycle of the MJO is similar to all convectively coupled equatorial waves (CCEWs; Kiladis et al. 2009) and even the life cycle of individual MCSs (cf. Fig. 1d). The resemblance of the MCS life cycle to the MJO is perhaps explained as the aliasing of shallow, deep, and stratiform parts of the MCS onto larger time and space scales, as discussed in Mapes et al. (2006). Different proportions of shallow, deep, and stratiform clouds occur in individual MCSs embedded within different large-scale wave structures, so in a filtered sense there is an evolution from enhanced shallow convection to enhanced deep convection, followed by enhanced stratiform cloud on the large scale. Mapes et al. (2006) refer to this filtered view as the stretched building block hypothesis. In addition to the shallow to deep to stratiform convective evolution, results from both pinwheel and RMM phases highlight the importance of narrow deep convective evolution across MJO phases—specifically, narrow deep precipitating types are relatively more important to the MJO prior to widespread deep convection and therefore may play a part in premoistening the environment prior to active MJO phases.

6. Summary and conclusions

Using CloudSat observations we have documented the evolution of total cloud cover, cloud types, temperature, and moisture across the MJO. Broadly speaking, the evolution of predominant cloud types over the life cycle of the MJO is similar to the life cycle of individual MCSs (cf. Figs. 1a–c) and even more broadly to all CCEWs (Kiladis et al. 2009). Such similarity gives credence to the stretched building block hypothesis of Mapes et al. (2006).

Differences between the MJO and Kelvin wave offer clues to underlying dynamical differences in how the wave types are coupled to convection. For a given amplitude in OLR, the MJO affects low-level moisture more than the Kelvin wave, while the Kelvin wave shows larger low-level changes in temperature through the phases (Fig. 13). We speculate that this is because the MJO modulates moisture much more than the Kelvin wave, again for comparable temperature, echo cover, and OLR amplitudes. One interpretation, supported by the findings of Raymond and Fuchs (2007, 2009) and Yasunaga and Mapes (2011), is that Kelvin waves are controlled more by CIN, while the MJO is affected more by moisture variations. Future studies could refine and test this hypothesis through experiments with cloud models on moisture-limited versus inhibition-limited convection, and through a better understanding of moisture budget processes in the MJO.

In terms of the WH04 RMM phases, the wide deep precipitating, anvil, cirrus, cumulus congestus, altocumulus, and stratocumulus EO types showed eastward propagation from the Indian Ocean to the central Pacific. The anvil and stratocumulus band tend to lag the wide deep precipitating type. These propagating EO types also showed a coherent Western Hemisphere signal. Generally, negative anomalies occur over South America and the Atlantic Ocean during phases 1–5 or when anomalies are positive over the Indian Ocean and Maritime Continent, with an opposite signal in phases 6–8. The stratocumulus type EOs also showed coherent variations over the eastern Pacific, with positive anomalies during phases 1–4 and negative ones afterward.

The narrow deep precipitating EO type also showed eastward propagation, but only from the Indian Ocean to the Maritime Continent in phases 1–4. Narrow deep precipitating anomalies tend to lead the wide deep precipitating anomalies by about 1–2 phases (Fig. 11b). Along with the eastward propagating feature, positive anomalies tropics-wide were enhanced during phases 1–3 and suppressed during phases 4–8. A similar enhancement, suppression pattern was observed for the cumulus EO type. During phases 1–2 (6–8) the tropics is anomalously cool (warm) at low levels (Fig. 11i), suggesting that the tropics-wide vertical instability of cumulus convection is being modulated by the MJO’s zonal mean temperature signal.

The pictorial mosaics, Figs. 6 and 12, offer a novel reality check on how we interpret the multiscale nature of convectively coupled waves, in this case the MJO. Besides representing the statistics of cloud evolution across the MJO, they also preserve cloud system morphology differences across phases, as well as a richness of texture that remind us of the complexity of clouds across large-scale waves. The comparison of the CloudSat mosaic to the Madden and Julian’s (1972) schematic shows that their ability to infer cloud evolution across the tropics was remarkable, but incomplete: there is more to be learned with the detail and richness of CloudSat.

As with all detailed observations, there is a question of how to apply these results to improve models or theories that do not have such details. One avenue that CloudSat offers is a cloud water content product, which provides derived measurements of liquid and ice water content. A similar analysis, as was done here, could be done for the liquid water and ice water content variables and then compared to reanalysis data or model output, similar to what Jiang et al. (2011) did for boreal summer intraseasonal variability events.

Another approach is to use the multiscale modeling framework (MMF) or “superparameterization” as a clean comparison to detailed observations. In superparameterization, 2D cloud system–resolving models (CSRMs) are embedded within each GCM grid cell to serve as the convective parameterization (Grabowski 2001; Randall et al. 2003). Several studies (Grabowski 2003; Khairoutdinov et al. 2008; Thayer-Calder and Randall 2009; Benedict and Randall 2009) have shown the success of superparameterization in simulating a fairly realistic MJO, yet none of these studies examined cloud morphology per se. In the future we aim to do a similar analysis as in this current observational study, but applied to CSRM output from multiscale model runs of equatorial waves.

Acknowledgments

We thank Kazuaki Yasunaga for help in formulating the discussion section. George Kiladis kindly provided the filtered OLR data. Generous student travel funding for (and discussion with participants at) the Monsoon Intraseasonal Variability Modeling Workshop in South Korea is also appreciated. We thank CloudSat PIs Deborah Vane and Graeme Stephens for a subcontract that supported the first stages of this work. Comments from Duane Waliser and one anonymous reviewer greatly improved this work. This work was performed for the Jet Propulsion Laboratory, California Institute of Technology, sponsored by the National Aeronautics and Space Administration and supported by the National Science Foundation under Grant 0806553.

REFERENCES

REFERENCES
Bacmeister
,
J. T.
, and
G. L.
Stephens
,
2011
:
Spatial statistics of likely convective clouds in CloudSat data
.
J. Geophys. Res.
,
116
,
D04104
,
doi:10.1029/2010JD014444
.
Bantzer
,
C. H.
, and
J. M.
Wallace
,
1996
:
Intraseasonal variability in tropical mean temperature and precipitation and their relation to the tropical 40–50-day oscillation
.
J. Atmos. Sci.
,
53
,
3032
3045
.
Benedict
,
J. J.
, and
D. A.
Randall
,
2007
:
Observed characteristics of the MJO relative to maximum rainfall
.
J. Atmos. Sci.
,
64
,
2332
2354
.
Benedict
,
J. J.
, and
D. A.
Randall
,
2009
:
Structure of the Madden–Julian oscillation in the superparameterized CAM
.
J. Atmos. Sci.
,
66
,
3277
3296
.
Grabowski
,
W. W.
,
2001
:
Coupling cloud processes with the large-scale dynamics using the Cloud-Resolving Convection Parameterization (CRCP)
.
J. Atmos. Sci.
,
58
,
978
997
.
Grabowski
,
W. W.
,
2003
:
MJO-like coherent structures: Sensitivity simulations using the Cloud-Resolving Convection Parameterization (CRCP)
.
J. Atmos. Sci.
,
60
,
847
864
.
Jiang
,
X.
,
D.
Waliser
,
J.
Li
, and
C.
Woods
,
2011
:
Vertical cloud structures of the boreal summer intraseasonal variability based on CloudSat observations and ERA-Interim reanalysis
.
Climate Dyn.
,
36
,
2219
2232
,
doi:10.1007/s00382-010-0853-8
.
Kemball-Cook
,
S.
, and
B. C.
Weare
,
2001
:
The onset of convection in the Madden–Julian oscillation
.
J. Climate
,
14
,
780
793
.
Khairoutdinov
,
M.
,
C.
DeMott
, and
D.
Randall
,
2008
:
Evaluation of the simulated interannual and subseasonal variability in an AMIP-style simulation using the CSU Multiscale Modeling Framework
.
J. Climate
,
21
,
413
431
.
Kiladis
,
G. N.
,
K. H.
Straub
,
G. C.
Reid
, and
K. S.
Gage
,
2001
:
Aspects of interannual and intraseasonal variability of the tropopause and lower stratosphere
.
Quart. J. Roy. Meteor. Soc.
,
127
,
1961
1984
.
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
.
Kiladis
,
G. N.
,
M. C.
Wheeler
,
P. T.
Haertel
,
K. H.
Straub
, and
P. E.
Roundy
,
2009
:
Convectively coupled equatorial waves
.
Rev. Geophys.
,
47
,
RG2003
,
doi:10.1029/2008RG000266
.
Knutson
,
T. R.
, and
K. M.
Weickmann
,
1987
:
30–60-day atmospheric oscillations: Composite life cycles of convection and circulation anomalies
.
Mon. Wea. Rev.
,
115
,
1407
1436
.
Kummerow
,
C.
,
W.
Barnes
,
T.
Kozu
,
J.
Shiue
, and
J.
Simpson
,
1998
:
The Tropical Rainfall Measuring Mission (TRMM) sensor package
.
J. Atmos. Oceanic Technol.
,
15
,
809
817
.
Lau
,
K.-M.
, and
P. H.
Chan
,
1985
:
Aspects of the 40–50-day oscillation during the northern winter as inferred from outgoing longwave radiation
.
Mon. Wea. Rev.
,
113
,
1889
1909
.
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
.
Lin
,
J.
,
B.
Mapes
,
M.
Zhang
, and
M.
Newman
,
2004
:
Stratiform precipitation, vertical heating profiles, and the Madden–Julian oscillation
.
J. Atmos. Sci.
,
61
,
296
309
.
Lin
,
X.
, and
R. H.
Johnson
,
1996
:
Kinematic and thermodynamic characteristics of the flow over the western Pacific warm pool during TOGA COARE
.
J. Atmos. Sci.
,
53
,
695
715
.
Luo
,
Z. J.
,
G. Y.
Liu
, and
G. L.
Stephens
,
2010
:
Use of A-Train data to estimate convective buoyancy and entrainment rate
.
Geophys. Res. Lett.
,
37
,
L09804
,
doi:10.1029/2010GL042904
.
Mace
,
G. G.
,
2004
:
Level 2 GEOPROF product process description and interface control document
.
Cooperative Institute for Research in the Atmosphere, 43 pp. [Available online at http://www.cloudsat.cira.colostate.edu/ICD/2B-GEOPROF/2B-GEOPROF_PDICD_3.0.pdf.]
Madden
,
R. A.
, and
P. R.
Julian
,
1972
:
Description of global-scale circulation cells in tropics with a 40–50-day period
.
J. Atmos. Sci.
,
29
,
1109
1123
.
Madden
,
R. A.
, and
P. R.
Julian
,
1994
:
Observations of the 40–50-day tropical oscillation—A review
.
Mon. Wea. Rev.
,
122
,
814
837
.
Maloney
,
E. D.
, and
D. L.
Hartmann
,
1998
:
Frictional moisture convergence in a composite life cycle of the Madden–Julian oscillation
.
J. Climate
,
11
,
2387
2403
.
Mapes
,
B.
,
S.
Tulich
,
J.
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
.
Marchand
,
R. T.
,
G.
Mace
,
T.
Ackerman
, and
G. L.
Stephens
,
2008
:
Hydrometeor detection using CloudSat—An earth-orbiting 94-GHz cloud radar
.
J. Atmos. Oceanic Technol.
,
25
,
519
533
.
Martin
,
E. R.
, and
C.
Schumacher
,
2011
:
Modulation of Caribbean precipitation by the Madden–Julian oscillation
.
J. Climate
,
24
,
813
824
.
Masunaga
,
H.
,
M.
Satoh
, and
H.
Miura
,
2008
:
A joint satellite and global cloud-resolving model analysis of a Madden–Julian oscillation event: Model diagnosis
.
J. Geophys. Res.
,
113
,
D17210
,
doi:10.1029/2008JD009986
.
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
.
Moncrieff
,
M. W.
,
2004
:
Analytic representation of the large-scale organization of tropical convection
.
J. Atmos. Sci.
,
61
,
1521
1538
.
Morita
,
J.
,
Y. N.
Takayabu
,
S.
Shige
, and
Y.
Kodama
,
2006
:
Analysis of rainfall characteristics of the Madden–Julian oscillation using TRMM satellite data
.
Dyn. Atmos. Oceans
,
42
,
107
126
,
doi:10.1016/j.dynatmoce.2006.02.002
.
Myers
,
D. S.
, and
D. E.
Waliser
,
2003
:
Three-dimensional water vapor and cloud variations associated with the Madden–Julian oscillation during Northern Hemisphere winter
.
J. Climate
,
16
,
929
950
.
Nakazawa
,
T.
,
1988
:
Tropical super clusters within intraseasonal variations over the western Pacific
.
J. Meteor. Soc. Japan
,
66
,
823
839
.
Nakazawa
,
T.
,
1995
:
Intraseasonal oscillations during the TOGA-COARE IOP
.
J. Meteor. Soc. Japan
,
73
,
305
319
.
Newman
,
M.
,
P. D.
Sardeshmukh
, and
C.
Penland
,
2009
:
How important is air–sea coupling in ENSO and MJO evolution?
J. Climate
,
22
,
2958
2977
.
Randall
,
D.
,
M.
Khairoutdinov
,
A.
Arakawa
, and
W.
Grabowski
,
2003
:
Breaking the cloud parameterization deadlock
.
Bull. Amer. Meteor. Soc.
,
84
,
1547
1564
.
Raymond
,
D. J.
, and
Z.
Fuchs
,
2007
:
Convectively coupled gravity and moisture modes in a simple atmospheric model
.
Tellus
,
59A
,
627
640
,
doi:10.1111/j.1600-0870.2007.00268.x
.
Raymond
,
D. J.
, and
Z.
Fuchs
,
2009
:
Moisture modes and the Madden–Julian oscillation
.
J. Climate
,
22
,
3031
3046
.
Riley
,
E. M.
,
2009
:
A global survey of clouds by CloudSat
.
M.S. thesis, Division of Meteorology and Physical Oceanography, University of Miami, 134 pp
.
Riley
,
E. M.
, and
B. E.
Mapes
,
2009
:
Unexpected peak near −15°C in CloudSat echo top climatology
.
Geophys. Res. Lett.
,
36
,
L09819
,
doi:10.1029/2009GL037558
.
Salby
,
M. L.
, and
H. H.
Hendon
,
1994
:
Intraseasonal behavior of clouds, temperature, and motion in the tropics
.
J. Atmos. Sci.
,
51
,
2207
2224
.
Sassen
,
K.
, and
Z.
Wang
,
2008
:
Classifying clouds around the globe with the CloudSat radar: 1 year of results
.
Geophys. Res. Lett.
,
35
,
L04805
,
doi:10.1029/2007GL032591
.
Schwartz
,
M. J.
,
D. E.
Waliser
,
B.
Tian
,
D. L.
Wu
,
J. H.
Jiang
, and
W. G.
Read
,
2008
:
Characterization of MJO-related upper tropospheric hydrological processes using MLS
.
Geophys. Res. Lett.
,
35
,
L08812
,
doi:10.1029/2008GL033675
.
Sperber
,
K. R.
,
2003
:
Propagation and the vertical structure of the Madden–Julian oscillation
.
Mon. Wea. Rev.
,
131
,
3018
3037
.
Stephens
,
G. L.
, and
Coauthors
,
2002
:
The CloudSat Mission and the A-Train
.
Bull. Amer. Meteor. Soc.
,
83
,
1771
1790
.
Thayer-Calder
,
K.
, and
D. A.
Randall
,
2009
:
The role of convective moistening in the Madden–Julian oscillation
.
J. Atmos. Sci.
,
66
,
3297
3312
.
Tian
,
B.
,
D. E.
Waliser
,
E. J.
Fetzer
,
B. H.
Lambrigtsen
,
Y. L.
Yung
, and
B.
Wang
,
2006
:
Vertical moist thermodynamic structure and spatial-temporal evolution of the MJO in AIRS observations
.
J. Atmos. Sci.
,
63
,
2462
2485
.
Tromeur
,
E.
, and
W. B.
Rossow
,
2010
:
Interaction of tropical deep convection with the large-scale circulation in the MJO
.
J. Climate
,
23
,
1837
1853
.
Tulich
,
S. N.
, and
B. E.
Mapes
,
2010
:
Transient environmental sensitivities of explicitly simulated tropical convection
.
J. Atmos. Sci.
,
67
,
923
940
.
Virts
,
K. S.
, and
J. M.
Wallace
,
2010
:
Annual, interannual, and intraseasonal variability of tropical tropopause transition layer cirrus
.
J. Atmos. Sci.
,
67
,
3097
3112
.
Wang
,
B.
,
2005
:
Theory
.
Intraseasonal Variability of the Atmosphere–Ocean Climate System, W. K. M. Lau and D. E. Waliser, Eds., Praxis, 307–360
.
Wang
,
Z.
, and
K.
Sassen
,
2007
:
Level 2 cloud scenario classification product process description and interface control document
.
Cooperative Institute for Research in the Atmosphere, 50 pp. [Available online http://www.cloudsat.cira.colostate.edu/ICD/2B-CLDCLASS/2B-CLDCLASS_PDICD_5.0.pdf.]
Wheeler
,
M. C.
, and
G. N.
Kiladis
,
1999
:
Convectively coupled equatorial waves: Analysis of clouds and temperature in the wavenumber–frequency domain
.
J. Atmos. Sci.
,
56
,
374
399
.
Wheeler
,
M. C.
, 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
.
Wu
,
C.
, and
H.
Hsu
,
2009
:
Topographic influence on the MJO in the Maritime Continent
.
J. Climate
,
22
,
5433
5448
.
Yasunaga
,
K.
, and
B. E.
Mapes
,
2011
:
Differences between more divergent versus more rotational types of convectively coupled equatorial waves. Part II: Composite analysis based on space–time filtering
.
J. Atmos. Sci.
,
in press
.
Yuter
,
S. E.
, and
R. A.
Houze
,
1995
:
Three-dimensional kinematic and microphysical evolution of Florida cumulonimbus. Part II: Frequency distributions of vertical velocity, reflectivity, and differential reflectivity
.
Mon. Wea. Rev.
,
123
,
1941
1963
.
Zhang
,
C. D.
,
2005
:
Madden–Julian oscillation
.
Rev. Geophys.
,
43
,
RG2003
,
doi:10.1029/2004RG000158
.
Zipser
,
E. J.
,
R. J.
Meitín
, and
M. A.
LeMone
,
1981
:
Mesoscale motion fields associated with a slowly moving GATE convective band
.
J. Atmos. Sci.
,
38
,
1725
1750
.