The Precipitating Cloud Population of the Madden-Julian Oscillation over the Indian and. West Pacific Oceans

Transcription

1 The Precipitating Cloud Population of the Madden-Julian Oscillation over the Indian and West Pacific Oceans Hannah C. Barnes and Robert A. Houze, Jr., Department of Atmospheric Sciences, University of Washington, Seattle, Washington, USA Corresponding author: H. C. Barnes Department of Atmospheric Sciences, University of Washington, Box , Seattle, WA Submitted to Journal of the Geophysical Research - Atmospheres November

2 2. Key Points TRMM PR characterizes MJO cloud population in central Indian Ocean and west Pacific. Stratiform regions vary the most by area but isolated, shallow echoes vary the most by number. During active stage, population related to large-scale wind shear and mid-upper moisture. 3. ABSTRACT The variability of the precipitating cloud population of the Madden-Julian Oscillation (MJO) is represented by statistics of echo features seen by the Tropical Rainfall Measuring Mission's Precipitation Radar over the central Indian Ocean (CIO) and West Pacific (WP). These echo features include isolated shallow echoes (ISE), deep convective cores (DCC), wide convective cores (WCC), and broad stratiform regions (BSR). ISEs are ever-present but most numerous during suppressed stages. BSRs dominate the variability in areal coverage with a strong maximum in active phases. DCCs and WCCs are more common and variable in number than BSRs. While both regions have similar magnitudes of variability, the active phase of the MJO is characterized in the CIO by a synchronous maximization of deep convective features and the WP is characterized by BSRs maximizing prior to WCCs. Reanalysis data show that ISEs are most numerous in a dry mid-troposphere with strong hpa (low-level) shear. Midtropospheric moisture increases before deeper convective features increase, possibly due to large-scale vertical motion, horizontal advection and/or convection at shorter timescales. Midtropospheric moisture then maximizes as the convective entities maximize and decreases as WCCs and BSRs decline. Active stage DCCs and WCCs preferentially occur in a moist midtroposphere with strong low-level shear, probably because acute shear favors enhanced downward momentum transport by downdrafts and more robust gust-front convective triggering. 2

3 BSRs maximize in a moist mid-troposphere with strong low- and upper- ( hpa) level shear that is not so strong that the stratiform region disconnects from its moisture source. 4. Index Terms AGU: 0399 General or miscellaneous, 9340 Indian Ocean, 9355 Pacific Ocean Author Key Words: Tropical Meteorology, Madden-Julian Oscillation, Tropical Convection, TRMM, vertical wind shear 5. Text 1. Introduction The Madden-Julian Oscillation (MJO) dominates intraseasonal (30-90 day) variability in the equatorial belt [Madden and Julian, 1971, 1972]. While upper-level circulation anomalies circumnavigate the globe, the MJO is most readily identifiable as a Kelvin-Rossby wave [Gill, 1980] that becomes convectively coupled in the central Indian Ocean (CIO), propagates into the Western Pacific Ocean (WP), and loses its convective coupling near the dateline. Even though the convective anomalies directly linked to the MJO are spatially limited to the equatorial belt in the eastern hemisphere, the MJO influences weather and climate over the globe [Zhang, 2005]. Given these teleconnections, an improved understanding of the MJO is expected to benefit global mid- and long-range forecasts. For a complete discussion of the MJO see Zhang [2005]. The convective coupling that occurs over the CIO and WP regions is not well understood, and a first step to improving this knowledge is to describe accurately the convective population and how it morphs into a population that includes deep cloud systems that couple with the largescale circulation of the MJO. The largest convective entities are mesoscale convective systems 3

4 (MCSs) [Houze, 2004], which are large, organized cloud systems whose precipitating regions cover at least 100 km in one horizontal direction and typically contain a population of deep convective elements along with relatively large stratiform rain areas. The importance of MCSs to the tropical cloud population was noted when Houze and Cheng [1977] and Cheng and Houze [1979] studied convection in the eastern tropical Atlantic and found that, while convective-scale precipitation radar echoes were far more numerous than mesoscale echoes, a little over 40% of the rainfall was associated with the relatively few mesoscale radar echoes. A global analysis by Yuan and Houze [2010] showed active MCSs to be associated with 56% of tropical precipitation. We might therefore expect the MJO to exhibit a higher frequency of MCSs in its convectively coupled active stage than its suppressed stage. This expectation is consistent with studies using outgoing longwave radiation [Mapes and Houze, 1993; Chen et al., 1996], radar data from the Tropical Ocean/Global Atmosphere Coupled Ocean-Atmosphere Response Experiment (TOGA- COARE) [DeMott and Rutledge, 1998; Yuter and Houze, 1998; Kingsmill and Houze, 1999a; Houze et al., 2000], A-Train satellite data [Del Genio et al.,2012; Riley et al., 2011; Yuan and Houze, 2012], and Tropical Rainfall Measuring Mission (TRMM) satellite products [Lin et al., 2004; Benedict and Randall, 2006; Lau and Wu, 2010]. Morita [2006] and Tromeur and Rossow [2010] found sporadic deep convection to be present during all stages of the MJO. Analyzing ship-borne radar data from the R/V Vickers during TOGA-COARE, DeMott and Rutledge [1998] found that the variability in the height of the 30 dbz reflectivity contour, which is used to represent deep intense convection, is small as the MJO passes through the WP. However, these intense convective cells were found to contribute proportionally less to the observed rainfall during active stages of the MJO when stratiform precipitation is more prevalent, consistent with the increased role of MCSs during the active phases. Lin et al. [2004] used five years of TRMM 4

5 Precipitation Radar (PR) data in the WP to show that 60% of the anomalous precipitation associated with the MJO comes from the stratiform portion of MSCs. This increased stratiform coverage during the active stage of the MJO is important since it is a signature of a more topheavy heating profile [Houze, 1982, 1989]. Since MCSs constitute the large end of the size spectrum of convective clouds, we focus here on how the cloud population of the MJO evolves in suppressed phases from having few MCSs to active phases hosting numerous MCSs. Despite the importance of MCSs, it should be kept in mind that these storms only account for about half of tropical precipitation. MCSs are not the only aspect of the MJO precipitating cloud population of interest. Observational and modeling studies have led some to hypothesize that the cloud population systematically transitions from shallow cumulus to cumulus congestus to MCSs as the MJO transitions from its convectively suppressed to convectively active stage as detrainment from convective clouds systematically increases the depth of the moist layer [e.g. Bladé and Hartmann, 1993; Benedict and Randall, 2007; Stephens et al., 2004]. While studies such as Del Genio et al. [2012] demonstrate that convection entities of all depths are present at all times, the authors also show that the relative frequency of deep convection increases as the MJO transitions into its convective stage. However, whether the shallower cumulus clouds are the proximate cause of the moistening and subsequent increasing convective depth remains a matter of debate since large-scale motions may also lead to changes in humidity, environmental wind shear, or large-scale ascent/subsidence. Further observational and modeling studies are needed to understand the interaction between shallow cumulus and the moisture field. However, whatever the answer, it will be important to understand how all elements of the convective cloud population (small, medium, and large) vary by phase of the MJO. We therefore we assess the 5

6 cloud population in this study by analyzing four echo entities that represent different stages of convective development. Analyses of synoptic-scale conditions within the MJO have shown that the variations in the precipitating cloud population coincide preferentially with certain large-scale atmospheric conditions. For example, Lin and Johnson [1996], Lin et al. [2004] and Chen et al. [1996] show that the precipitation maximum associated with the MJO in the WP occurs roughly 1-3 weeks prior to the westerly wind burst (WWB), 5-10 days before low-level ( hpa) shear and upper-level ( hpa) shear maximize, and 2-5 days before hpa relative humidity maximize. Numerical models continue to struggle to accurately represent the MJO. Zhang [2005] pointed out that even when models are able to accurately produce the propagation, speed, and scale of the MJO, the period and structure are often unrealistic. Kim et al. [2009] analyzed eight climate models using the guidelines set forth by the U.S. Climate Variability and Prediction (CLIVAR) MJO Working Group and found that no single model exceled in every diagnostic. While the models are more successful in representing large-scale circulation anomalies than convective anomalies, both the circulation and convective anomalies are often too weak. Significantly for this study, models often underrepresent the stratiform precipitation, which we will find from observations to exhibit the greatest phase-to-phase variability in terms of areal coverage. These previous authors suggest that model accuracy is dependent on the interaction between convection and moist physics. Haertel et al. [2008] found that the MJO is not only sensitive to the deep heating and circulation anomalies produced by deep convection and stratiform regions but that the shallow heating and upper level-cooling resulting from shallow cumulus and congestus is also important. Their results indicate that the relative amounts of 6

7 shallow clouds, deep convective elements, and stratiform precipitation are important in determining heating profiles associated with the clouds in the MJO. Consistent with this view, Zhang and Song [2009] found that a realistic MJO would not develop without shallow convection even if deep convection is accurately portrayed. Even though models are shown to be sensitive to the nature of the cloud population and its interaction with the large-scale atmospheric conditions, the exact relationship among these features is unknown. Thus, a first step to improving the model representation of the MJO is to provide detailed observations of all members of the cloud population and its concurrent large-scale environment in each MJO phase. Yuan and Houze [2012] used A-Train satellite data to analyze the variability of large, mature MCSs in the eastern Indian Ocean, Maritime Continent, and WP with phase of the MJO and related these changes to large-scale conditions using ERA-interim reanalysis. Connected- MCSs, which are very large systems consisting of at least two MCSs joined by precipitation and often having a thick cloud deck that efficiently produces stratiform precipitation, experience the greatest variability with phase of the MJO. While MODIS brightness temperature data indicate that the depth of convection varies little with phase of the MJO, brightness temperature data are incapable of accurately distinguishing between different types of deep convection. This limitation is resolved in the current study through the use of the TRMM Precipitation Radar (TRMM PR) [Kummerow et al., 1998], which successfully separates stratiform from convective rain, isolates the tallest and widest components of deep convection, and identifies small, shallow convection. The purpose of this study is to use the capability of the TRMM PR data to characterize the precipitating cloud population of the MJO via these echo types. Specifically we will: 7

8 Determine how the relative amounts of each type of convective entity seen by the TRMM PR change as a function of MJO phase. Compare the TRMM PR echo population properties in the Indian Ocean, where many MJOs become convectively coupled, with the West Pacific, where the MJO often moves through as a fully developed disturbance. Frame observed changes in the convective population in context of the large-scale relative humidity, circulation patterns, and vertical wind shear fields of the MJO using global reanalysis data coinciding with the TRMM PR data. 2. Data and methodology This study uses Version 6 of the TRMM Precipitation Radar data obtained over the central Indian Ocean (CIO) and West Pacific (WP) (Figure 1) during the months of October through February from The WP is further separated into the three sub-regions: the northwest (NWWP), northeast (NEWP), and southeast (SEWP) since the precipitating cloud population and large-scale environment markedly vary between these regions. These differences will be discussed in detail in the following sections. We have not included land areas in this analysis, so the southwest portion of the WP is excluded. Every echo feature is described by its intensity using 3D attenuation corrected reflectivity from the TRMM 2A-25 product [Iguchi, 2000] and separated into its echo rain type using the classification provided in the TRMM 2A-23 product [Awaka et al., 1997]. With a sensitivity of 17 dbz, the TRMM PR detects most of the rainfall in tropical oceanic regions and a considerable amount of the vertical structure of storms. Here we use the TRMM PR data to analyze four important components of the precipitating cloud population. The TRMM 2A-23 product divides the PR data into three categories: convective, stratiform, and 8

9 other [Awaka et al., 1997]. These categories are first used to separate each contiguous radar echo into its convective and stratiform components. Three subsets of convective echoes are identified in order to capture the fundamentally different characteristics of convective areas. One subset includes all shallow, isolated precipitating convective elements of the type analyzed previously by Schumacher and Houze [2003]. We refer to these entities as isolated shallow echoes (ISEs), and they represent precipitating clouds of shallow to moderate depth. To capture more intense convection, we consider two features defined by an echo threshold of 30 dbz. These entities are similar to those analyzed by Houze et al. [2007], Romatschke et al. [2010], and Romatschke and Houze [2010] used to investigate extreme convective systems over land. However, the reflectivity thresholds used in this study are lowered from 40 to 30 dbz to be consistent with differences between oceanic and land convection. Deep convective cores (DCCs) are identified as contiguous three-dimensional convective columns with radar echo 30 dbz reaching at least 8 km in altitude. Wide convective cores (WCCs) are defined as contiguous three-dimensional convective volumes with radar echo 30 dbz covering at least 800 km 2 at some altitude. DCCs and WCCs are two different indications of convective intensity whose populations often overlap but do not coincide. DCCs and WCCs are sometimes, but not always, contained within MCSs. A large region of convectively initiated stratiform precipitation, however, is a manifestation of an extremely well developed MCS (Houze 1993, 2004). To signify such a feature, a broad stratiform region (BSR) is defined as a contiguous stratiform echo covering at least 50,000 km 2. BSRs have no reflectivity threshold. In this study we take a census of the four above-defined echo object types (ISE, DCC, WCC, and BSR) as a means of determining the nature and variability of the precipitating cloud population associated with the MJO. 9

10 To relate the populations of these observed radar echo objects to the evolving large-scale environmental conditions comprised by the MJO, we analyze the large-scale circulation patterns, relative humidity, and vertical wind shear fields within which the clouds occur by using fourtimes-daily ERA-interim reanalysis data [Dee et al., 2011] at 1.5 x 1.5 degree resolution for the same geographic regions and time periods examined in the TRMM PR portion of this study. Tian [2010] compares the moist vertical profile of the ERA-interim data and AIRS data with respect to the MJO in the Indian Ocean and WP. While discrepancies between the global reanalysis and satellite data exist, including an underestimation of boundary layer moisture by the reanalysis, the ERA-interim reanalysis provides the most reasonable reanalysis representation of the largescale humidity and circulation fields. We use the Wheeler and Hendon [2004] index (WH index) to group the TRMM PR data and ERA-interim reanalysis data in relation to the structure of the MJO. For a complete discussion of the WH index see To ensure that composites represent robust MJO events, we consider only the months of October through February, which is the climatological peak MJO season [Madden, 1986], and only include days in which the amplitude of the WH index is greater than one. During the fourteen boreal winters in the TRMM PR dataset, approximately 63% of the days are characterized by a robust MJO. During each phase, a Monte-Carlo method is used to create twenty independent samples of 100 days to assess the significance of the variability of the precipitating cloud population using the student s t-statistic at the 99% confidence level. The mean of the twenty samples is almost identical to results found when all days in a given phase are used to construct the composites. For this reason, the ERA-interim composites are generated using all days within a given phase and these reanalysis composites do not significantly differ from those generated from the mean of the 10

11 twenty trials. It is important to note that since this study is based upon 14-year composites of the WH index and the duration of each phase in the WH index varies for individual MJO events, a given MJO may have stronger or more abrupt variability than the composites presented below. Evidence of this smoothing will be discussed in the conclusions using data collected during the Dynamics of the Madden-Julian Oscillation/Atmospheric Radiation Measurement (ARM) MJO Investigation Experiment (DYNAMO/AMIE) campaigns that took place from 1 October 2011 to 9 February 2012 in the central Indian Ocean. 3. Variations in the precipitating cloud population with phase and region Figure 2 from Wheeler and Hendon [2004] shows how the satellite-observed outgoing long-wave radiation (OLR) signal varies by phase of the MJO with the black vertical lines showing the geographic regions analyzed in the current study. The active phase, indicated by a minimum of OLR, propagates eastward such that it is centered over the CIO during phase 3 and over the WP in phases 5-6. While the eastward propagation of the convective envelope of the MJO is readily apparent in OLR, the composition of the cloud population (by size and type of cloud) is not well discriminated. In this section, TRMM PR data demonstrates how the composition of the precipitating cloud population varies from one phase to the next over the CIO and WP. Figures 3, 4, 5, and 6 show geographic maps of the distribution of the average frequency of ISE, DCC, WCC, and BSR echoes, respectively, for the mean of the twenty samples in the CIO and WP during phases 1, 3, 5, and 7. The frequency is reported as a percentage and represents how often the TRMM PR detects one of these echo entities as it passes over a given 0.5 x 0.5 grid box. Thus, this frequency is a measure of the areal coverage of each echo feature. Since this study focuses only on the oceanic precipitating cloud population, any grid box 11

12 containing land is blank in the maps. The color bar differs in each figure to most clearly depict the variability of each type of echo feature. The black lines in the WP maps show the boundaries of the NWWP, NEWP, and SEWP sub-regions. Each panel in Figure 7 summarizes the variability of one of the echo features by calculating the areal frequency of each entity in the same manner as in Figures 3, 4, 5, and 6. However, instead of calculating the frequency over individual 0.5 x 0.5 grid boxes and plotting the geographic distribution of echoes, Figure 7 calculates the frequency across the four domains shown in Figure 1 (CIO, NWWP, NEWP, and SEWP) and plots the frequency by phase. The blue lines represent frequency-phase series of the twenty samples, the black line is the mean of the samples, and the dashed red lines represent the 99% confidence interval based upon the twenty samples. The active stage in each geographic region is defined in this study to occur during the phase that has the maximum areal coverage of BSR echoes, which occurs during phases 2-3 in the CIO, phase 5 in the NWWP, phase 3-7 in the NEWP, and phase 6 in the SEWP. While comparison of the maps in Figure 3, 4, 5, and 6 reveals that BSR echoes experience the greatest variability in areal coverage with phase of the MJO in the CIO and WP, Figure 7 shows that the areal coverage of each echo entity significantly varies with phase in all geographic regions. ISEs are shown in Figure 3 to always be present in each geographic region. These echoes are especially common near 8 S in the CIO and 8 N in the WP. Schumacher and Houze [2003] showed that these bands of enhanced ISE frequency are associated with the poleward edges of the cloud band of the Intertropical Convergence Zone (ITCZ), which is climatologically located in the southern hemisphere in the CIO and on both sides of the equator in the WP. Inspection of Figure 3 reveals that ISEs are slightly less common during the active stage of the MJO, which 12

13 occurs during phase 3 in the CIO and phase 5 in the WP. Despite the visually subtle variability, Figures 7a, 7e, 7i, and 7m indicate that ISEs are significantly more frequent during the suppressed stage than the active stage in each geographic region. Figures 3 and 7 normalize the ISE coverage by the total area detected by the TRMM PR, which includes echo-covered and echo-free areas. If ISE coverage is normalized only by the area that is echo-free, ISE coverage continues to significantly vary with phase of the MJO (not shown). Thus, the variability of ISE coverage with phase of the MJO cannot purely be attributed to ISEs having less space to develop when deep and wide convective activity greatly increases during the active stage. Figure 3 indicates that the greatest reduction in ISEs in the CIO occurs within the ITCZ regions, which suggests the suppressed stage of the MJO is the active stage of the ITCZ. In the WP, however, Figure 3 shows that the greatest reduction occurs along the equator. While both regions experience a maximum in ISE coverage during the suppressed stage, the maximum in each geographic region occurs in different phases of the suppressed stage. Figure 7a shows that ISEs peak in the CIO during phase 5, which is two phases after the active stage and during the beginning of suppressed stage. However, ISEs in the WP are shown in Figures 7e, 7i, and 7m to peak during phase 3, which are 5-6 phases after the active stage and during the end of the suppressed stage. DCC echoes are much less frequent than ISEs. However, like ISEs, DCC echoes are always present and their variability is not easily seen by eye in Figure 4 but is shown to be statistically significant in Figure 7. Close examination of Figure 4 shows that in the CIO, DCC echoes occur somewhat more frequently north and south of the equator. Figure 7b indicates that DCC echoes in the CIO broadly peak during the active stage of the MJO in phases 1-3. However, this peak is not significantly greater than another peak occurring just prior to the active stage in 13

14 phase 7. Comparison of the panels in Figure 4 suggests that, overall, DCC echoes are more common in the WP than the CIO and are spread inhomogeneously across the WP. DCC echoes are somewhat more common over the warm-pool and within the South Pacific Convergence Zone (SPCZ), which are contained with the NWWP and SEWP, respectively. While Figure 7f and 7j indicate that DCC echoes in NWWP and NEWP, respectively, do not experience a statistically significant peak, Figure 7n shows that DCC echoes in the SEWP peak distinctly when the active phase concludes in phase 6. We attribute the consistent DCC frequency in the NWWP to the high instability that continually characterizes the Pacific warm-pool, which is centered in this region. The absence of a peak in DCC echoes during the active stage in the NEWP may be related to the tendency for the MJO in the WP to manifest most clearly in the region of the SPCZ where the water is warmer [Weickmann et al., 1985]. WCC echo variability with phase of the MJO in the CIO and WP is more visually identifiable in Figure 5 than the ISE and DCC echo variability in Figures 3 and 4, respectively. In the CIO, WCC echoes are homogeneously distributed throughout the region as they peak during the active stage in phase 3, which Figure 7c indicates is statistically significant. Furthermore, Figure 7c indicates that the increase in WCC echoes leading up to the active stage in the CIO is more gradual than their reduction immediately following the active stage. A distinct minimum is also evident in Figure 7c during phase 5; however, this minimum is not statistically significant and could be considered to extend broadly across phases 4-8. Comparing the panels of Figure 5 suggests that WCC echoes occur more frequently in the CIO than the WP. We attribute this tendency to the CIO consistently having stronger low-level and upper-level shear than the WP, which we will discuss in detail in subsequent sections. WCC echoes in the WP are shown in Figure 5 to continue to preferentially occur along a diagonal line from the warm-pool 14

15 in the NWWP to the SPCZ in the SEWP. Figure 7o indicates that the SEWP experiences a significant maximum in WCC echo coverage during phase 7, which is one phase after the BSR echo maximum and active stage. The increase and decrease in WCC coverage is more symmetric in the SEWP than is witnessed in the CIO in Figure 7c. WCC echoes broadly minimize in the SEWP during phases 2-5. The WCC echo peak in the NWWP is shown in Figure 7g to be less distinct as the frequency increases stepwise starting in phase 2 and maximizes during phase 6, which is one phase after the BSR maximum and active stage. However, the maximum in phase 6 is not statistically larger than a secondary peak during phase 8. WCC variability in the NEWP is distinctly different. While a single peak in WCC echo coverage occurs during the active stage of the MJO in the NWWP and SEWP, Figure 7k shows that WCC echoes in the NEWP maximize and minimize twice. After minimizing in phase 2 and maximizing in phase 4, WCC echoes in the NEWP gradually decline to a secondary minimum in phase 7 and increase to a secondary maximum in phase 8. None of these minima and maxima are statistically distinct. This dissimilar trend in WCC variability may be related to the tendency for the MJO to avoid the Pacific coldtongue and propagate into the SPCZ [Weickmann et al., 1985]. Figure 6 shows that BSR echoes experience the most visually apparent variability with phase of the MJO as they maximize in the active stage and minimize during the suppressed stage in each region. This trend is expected since BSR echoes are the largest of the echo entities that we use to represent the precipitating cloud population and their maximum designates the most frequent upscale growth of deep convection into organized MCSs. While Figure 7 indicates that the statistical uncertainty associated with BSR echoes in the active stage is comparable to the other echo entities, BSR echoes in the suppressed stage are associated with the smallest statistical uncertainty of all echo entities during any stage of the MJO. Thus, the variability of the 15

16 precipitating cloud population during the MJO is best characterized by a reduction in large, mesoscale systems during a relatively long suppressed stage. In the CIO, BSR echoes maximize during phases 2-3, which Figure 7d shows is highly statistically significant. Additionally, Figure 7d indicates that the variability in BSR echo coverage is asymmetric around the active stage as BSR echo coverage increases more slowly leading up to the active stage than it decreases after the active stage. The eastward propagation of the MJO is discernible in Figure 6. In the CIO, BSR echoes tend to occur in the western portion of CIO basin during phases 8-1 and in the central and eastern portions during phases 2-7. Even though BSR echoes continue to preferentially occur over the warm-pool and SPCZ, the gradual eastward propagation of the MJO is evident in the WP in Figure 6. Figures 7h, 7l, and 7p emphasize the inherent geographic differences in the WP. While the NWWP and SEWP have statistically significant maxima during phases 5 and 6, respectively, the NEWP has a broad peak in BSR echo coverage during phases 3-7. Additionally, BSR echoes in the SEWP vary asymmetrically around the active stage similar to the CIO but BSR echoes in the NWWP vary symmetrically around the active stage. In summary, inspection of Figure 7 indicates that the composition of the precipitating cloud population changes within each geographic region in terms of the different sizes and types of precipitating clouds as the MJO transitions from one phase to the next. Additionally, Figure 7 illustrates that the relationship between the population and MJO phase changes between the time that the MJO initiates in the CIO and propagates into and through the WP. Generally, the frequency of large stratiform regions decreases during the suppressed stage of the MJO in both geographic regions. In the CIO, ISEs are most common just after the active stage and DCC, WCC, and BSR echoes maximize synchronously, which suggests that the active stage is characterized by deep convection of all sizes and stages of development. In the WP, ISEs 16

17 maximize just prior to the active stage and DCC, WCC, and BSR echoes do not simultaneously maximize. With the exception of the NEWP, WCC echoes in the WP maximize one phase after BSR echoes peak, which suggests that throughout most of the WP the active stage is first characterized by extremely large MCSs followed by smaller MCSs. The NEWP has the least coherent variability with phase of the MJO, which may be associated with the tendency for the convective envelope of the MJO to manifest more strongly off the equator over the warmer water of the SPCZ region. The magnitude of the overall variability of each type of echo object (ISE, DCC, WCC, and BSR) is compared in Figure 8. Figures 8a-8d show the mean frequency (thick line) and 99% confidence intervals (thin lines) of ISEs, DCCs, WCCs, and BSRs in terms of areal coverage. BSR echoes dominate the variability in areal coverage of the precipitating cloud population in each geographic region and occur primarily in the active stage of the MJO. However, areal coverage is only one way of characterizing the variability of the precipitating cloud population and emphasizes the stratiform component of the convective population that is characterized by upper-level heating [Houze, 1982, 1989]. The variability in the number of echo entities is another important aspect of the precipitating cloud population, which emphasizes the importance of mass transport and heating in convective-scale elements. Figures 8e-8h show the mean (thick line) and 99% confidence intervals (thin line) of the number of DCCs, WCCs, and BSRs as a function of phase of the MJO. Figures 8i-8l show the mean and 99% confidence interval for the number of ISEs. ISEs are by far the most common and variable element of the precipitating cloud population by number in both geographic regions. Comparing only deeper convective features reveals that WCC echoes are most common and experience the greatest variability by number in the CIO. In the WP, however, DCC echoes are most common. Thus, while ISE, DCC, and WCC 17

18 echoes experience only subtle changes in areal coverage, these echo entities are highly variable in terms of their number. Haertel et al. [2008] showed that, in addition to deep heating and cooling from the deep convection and stratiform regions, the MJO is sensitive to heating by shallow convection. The following sections will investigate how each of the four components of the precipitating cloud population is related to the large-scale humidity and circulation fields. 4. Relationship between relative humidity and the precipitating cloud population Figure 9 shows the average vertical profile of relative humidity in each geographic region from the ERA-interim reanalysis. The solid black and blue lines represent the two phases leading up to the active stage, the solid green line represents the convectively active stage, and the solid red line is one phase after the active stage. The dashed lines represent the suppressed phases. Regardless of phase and geographic region, the average relative humidity in Figure 9 is consistently moist below 800 hpa i.e., the moist layer never goes away! This uniformity suggests that variations in the precipitating cloud population are unassociated with the largescale lower-tropospheric moisture field. The largest moisture variations in both geographic regions are found in the mid-upper troposphere from hpa and are characterized by a relative humidity maximum during the active stage and minimum during the suppressed stage. Figure 9 indicates that the greatest largescale mid-tropospheric relative humidity variability occurs over the CIO. The mid-tropospheric moisture in the CIO rapidly increases from phase 7 to 8, just after the secondary peak in DCC echoes and just prior to significant increases in WCC and BSR echo coverage (Figures 7b-7d). The source of moisture for this rapid increase in relative humidity is unclear but may include planetary wave induced vertical motion, horizontal advection, and/or accumulation from convective bursts occurring at shorter time-scales. Mid-tropospheric relative humidity continues 18

19 to increase slowly and maximize in phases 2-3, which corresponds in phase to the rapid increase and maximum areal coverage of DCC, WCC, and BSR echoes (Figure 7b-7d). Finally, midtropospheric relative humidity quickly declines as the frequencies of DCC, WCC, and BSR echoes significantly decrease in phase 4. Figures 7a and 8c indicate that ISEs are most common in phase 5; when the mid-troposphere in the CIO is relatively dry (Figure 9). Figure 9 indicates that the shape and variability of the relative humidity profile in the NWWP and NEWP is similar to the CIO. However, the profile in the SEWP is notably different since variability in the hpa layer is small. Despite this reduced variability in the SEWP, each of the WP regions displays a systematic association between the precipitating cloud population and mid-tropospheric moisture. For example, mid-tropospheric moisture in the NWWP is shown in Figure 9 to maximize in phase 5, which corresponds to the maximum in BSR echo coverage in Figure 7h. Even though Figure 7h shows that BSR echoes significantly decline in phase 6, Figure 9 indicates that mid-tropospheric moisture does not dramatically decrease until phase 7 when WCC echoes are shown in Figure 7g to significantly decline. Thus, mid-tropospheric drying in the NWWP is delayed by one phase relative to the CIO, where the mid-troposphere dramatically dries one phase after DCC, WCC, and BSR echoes synchronously maximize. Therefore, the NWWP suggests that WCC echoes can maintain a moist midtroposphere even if the frequency of BSR echoes significantly decline. This result is not surprising since both WCC and BSR echo objects are often associated with MCSs and the prevalence of WCC echoes probably indicates that numerous MCSs are present, but are smaller and not producing BSRs. The ability for WCC and DCC echoes to maintain a moist midtroposphere is also observed in the SEWP. Figure 9 shows that mid-troposphere moisture begins to decline in the SEWP in phase 8, which is shown in Figures 7n-7p to be two phases after BSRs 19

20 peak and one phase after DCC and WCC echoes peak. Similar to the CIO, Figure 9 and Figures 7e, 7i, and 7m indicate that ISEs in the WP are out phase with the mid-tropospheric relative humidity peak since these echoes occur most frequently when the mid-troposphere is dry. While it is unclear whether the large-scale mid-tropospheric relative humidity is causing or reacting to changes in the precipitating cloud population, based on these results it is evident that the large-scale mid-level moisture field varies systematically with the precipitating cloud population, particularly in terms of the presence of the deepest, widest, and most stratiform-rainproducing convection (DCCs, WCCs, and BSRs). 5. Relationship between large-scale winds and the precipitating cloud population Figure 10 shows maps of the average 1000 hpa wind direction (vectors) and zonal speed (shading) for phase 2, 4, 6, and 8 in the CIO and WP. The westerly wind burst (WWB) [Zhang, 2005] in the CIO propagates eastward just south of the equator and maximizes during phase 4, which Figure 7b-7d indicate is one phase after DCC, WCC, and BSR echoes maximize. This large-scale westerly feature is observed through the depth of the lower troposphere, up to the 500 hpa level (not shown). While the WWB is present in the WP, the westerly winds are much weaker, more confined in the meridional direction, and only discernible up to the 700 hpa level (not shown). Additionally, the WWB in the WP propagates southeastward along the SPCZ in the SEWP. Figure 10 indicates that the WWB peaks during phase 6 in the NWWP, which is one phase after BSR echoes maximum (Figure 7h) and similar to the CIO. However, unlike the CIO, the areal coverage of WCC echoes maximizes with the WWB peak in phase 6 as shown in Figure 7g. The SEWP is also characterized by lags of 1 and 2 phases between maxima in BSR and WCC echoes and the maximum WWB, respectively. Despite these differences, the peak WWB signals the conclusion of large, organized convection in both geographic regions, a fact 20

21 consistent with previous studies [e.g. Lau et al., 1989]. Figure 10 indicates that low-level westerly winds minimize during phase 8 in the CIO, which is near the minimum in BSR coverage and the end of the suppressed stage in Figure 7a. A similar relationship is observed during phase 3 throughout the WP. Strong easterlies dominate the upper-levels in the CIO during the active phase of the MJO (Figure 11a). These upper-level winds propagate eastward along to the equator and maximize in phase 4, which Figures 7b-7d show to be one phase after the DCC, WCC, and BSR echo maxima. Figure 11 indicates that the circulation field in the CIO has a sizable meridional component during phases 6-8, which corresponds to the suppressed stage of the MJO in that region. Easterly winds are also present throughout the WP. Similar to the CIO, these easterlies are shown in Figure 11 dominate the active stage and propagate eastward in the WP. However, the upper-level winds in the WP are much weaker than those observed in the CIO. Figure 11 shows that the areal coverage of strong easterly winds in the NEWP maximizes in phase 8 as BSR echoes are shown in Figure 7l to begin to decline. While upper-level easterly winds in the NWWP are also shown in Figure 11 to maximize in terms of areal coverage during phase 8, Figures 7f shows that this maximum is two phases after the peak of BSR echoes. This two-phase lag in peak 200 hpa winds with respect to the BSR echo maximum is also observed in the SEWP. While the lag between the 200 hpa easterly maximum and BSR peak varies between one and two phases, upper-level easterly winds consistently maximize shortly after the widest and deepest mesoscale echo entities peak in each geographic region. Motivated by the geographical differences in magnitude and propagation of the 1000 hpa and 200 hpa zonal winds with phase of the MJO, we investigate how the large-scale wind influences the precipitating cloud population in terms of the vertical shear of the horizontal winds. 21

22 6. Relationship between large-scale shear and the precipitating cloud population Figure 12 shows the average vertical profile of zonal winds for each geographic region using the same color scheme as Figure 9. The solid red line, which represents the profile of zonal winds one phase after BSR echoes peak during the active stage, is characterized by a maximum in westerly winds (or minimum in easterly winds) at approximately 750 hpa in each geographic region. This extremum is consistent with the WWB observed in Figure 10. Given that zonal winds below 750 hpa become more westerly (i.e. less easterly) with height and above 750 hpa become more easterly with height in each geographic region at this time, 750 hpa is used to define low-level ( hpa) and upper-level ( hpa) vertical wind shear. Maps of the magnitude (shading) and direction (vectors) of the average hpa shear for each phase is shown in Figure 13 for the CIO. Figure 13 shows that as the MJO transitions out of the suppressed stage and reaches its convectively active stage, the hpa shear increases. The simultaneous rise in mesoscale convective activity and low-level shear might be expected since MCS downdrafts transport mid-level winds to the surface, which helps to create the surface convergence, vertical motion, and convective initiation required to maintain MCSs [Houze, 1993, 2004]. As the active stage in the CIO concludes during phase 4 this complementary trend stops. Figure 13 indicates that large-scale low-level shear maximizes in phase 4 despite Figures 7b-7d showing a significant decrease in DCC, WCC, and BSR echo coverage. As will be discussed below, we attribute this sudden change to acute upper-level shear and/or a dry mid-troposphere. Figure 14 shows that the upper-level ( hpa) shear maximizes during phase 4 over the CIO, which is one phase after DCC, WCC, and BSR echoes maximize in areal coverage (Figure 7b-7d ). Stratiform regions within MCSs are sustained in part by the import of moisture 22

23 from the convective regions of the MCS [Houze et al., 1980; Houze, 2004]. While a moderate amount of upper-level shear strengthens the mid-level inflow and aids in mesoscale organization, excessively strong upper-level shear severs the stratiform region from its convective moisture source. Given that BSR echoes peak prior to the strongest upper-level shear, these results suggest that the variability in BSR echo coverage during the MJO is influenced by the degree of upperlevel shear. DCC and WCC echoes are likely less sensitive to upper-level shear since their development does not depend on moisture being transported from another part of the convective system that is relatively far away. Rather, the decline of DCC and WCC echoes in phase 4 is more likely related to the rapid drying of the mid-troposphere observed from phase 3 to 4 in Figure 9. The rapid decline in BSR frequency is also likely affected by this drying. While observations in the CIO suggests that the precipitating cloud population during the active stage of the MJO is systematically associated with the large-scale mid-tropospheric moisture, lowlevel shear, and upper-level shear, the WP provides the opportunity to investigate these relationships in greater detail since the DCC and WCC peaks during the active stage are temporally isolated from the BSR peak. Figures 15 and 16 show that both low-level and upper-level shear is weaker in the WP than in the CIO. However, despite differences in shear magnitude, the association between the precipitating cloud population during the active stage of the MJO and the large-scale atmospheric conditions throughout most of the WP is qualitatively consistent with the relationships observed during the active stage in the CIO. For example, during the active stage in the NWWP, WCC and BSR echoes maximize during phases 6 and 5, respectively (Figure 7g- 7h). At these times, low-level shear (Figure 15) is relatively strong and the mid-troposphere is very moist (Figure 9), which suggest that atmospheric conditions are favorable for WCC and 23

24 BSR echoes. Figure 16 indicates that upper-level shear in the NWWP maximizes during phase 6, which may account for the dramatic reduction in BSR echo frequency during this phase since the strong shear is likely separating the large stratiform regions from their convective moisture sources. Figure 9 indicates that mid-tropospheric relative humidity rapidly declines from phase 6 to 7, which corresponds to the statistically significant decrease in WCC echoes in Figure 7c. WCC and BSR echoes during the active stage of the MJO in the SEWP are shown in figures 7n-7p, 9, 15, and 16 to exhibit the same association with large-scale relative humidity, low-level shear, and upper-level shear as WCC and BSR echoes observed during the active stage in the NWWP. Additionally, the SEWP shows that DCC echoes observed during the active stage occur most frequently when the mid-troposphere is moist and the low-level shear is strong. DCC echoes are shown in Figure 7n to maximize in phase 7, which corresponds to the strongest lowlevel shear in Figure 15 and the highest mid-tropospheric relative humidity in Figure 9. This association between DCC echoes, mid-tropospheric moisture, and low-level shear is likely not witnessed in the NWWP due to the consistently high instability that characterizes the warm-pool and enables DCC echoes to occur at all times. While we have shown that DCC and WCC echo entities observed during the active stage of the MJO in the CIO and most of the WP appear to be coherently associated with the largescale mid-tropospheric relative humidity and low-level shear, these relationships are not always applicable to the suppressed stage of the MJO. For example, Figure 7b indicates that a secondary maximum in DCC echo coverage occurs in the CIO during phase 7, which is near the end of the suppressed stage. Figure 9 indicates that phase 7 corresponds to some of the lowest midtroposphere relative humidities in the CIO and Figure 13 shows that the low-level shear is relatively weak at this time. A similar secondary peak in WCC echoes occurs in the NWWP 24

25 during phase 8 (Figures 7g, 9, and 15). Apparently, isolated deep convective cores and smaller MCSs can occur in sufficiently unstable situations even if the large-scale shear and humidity fields are unfavorable. However, BSR echoes, the signatures of the largest MCSs, appear to have a consistent relationship with large-scale mid-tropospheric relative humidity, low-level shear, and upper-level shear during all phases of the MJO in each geographic region. I.e., fully developed MCSs require optimal conditions of environmental shear and humidity, and we conclude that active phases of the MJO are situations in which these favorable conditions are most often present. The association between the precipitating cloud population during the active stage and large-scale moisture and shear fields is weakest in the NEWP. For example, BSR echoes are shown in Figure 7l to broadly maximize in phases 3-7. Figure 16 indicates that upper-level shear remains moderately strong throughout this time, which suggests that the hpa shear is strong enough to foster mesoscale organization but not strong enough to destroy stratiform regions. However, Figures 9 and 15 indicate that mid-tropospheric moisture and low-level shear are highly variable throughout this time. Given that the MJO tends to manifest its convective coupling in the SPCZ at these longitudes [Weickmann et al., 1985], we should perhaps not be surprised that the relationship between the precipitating cloud population and large-scale shear and relative humidity is least distinct in this region. Mid-tropospheric moisture and low-level shear is also coherently associated with the ISE maximum in each geographic region. For example, over the CIO, ISEs maximize in phase 5, which corresponds to the suppressed stage of the MJO (Figure 7a). Figures 13 and 9 show that at the same time the low-level shear is relatively strong and the mid-troposphere is dry over the CIO. This pattern of strong hpa shear and dry mid-tropospheric conditions is observed 25

26 in each of the WP regions. ISEs do not have a consistent relationship with upper-level shear, which is not surprising since they are by definition shallow. In the CIO, the ISE maximum in phase 5 is characterized by strong hpa shear (Figure 14). However, ISEs in the WP maximize in phase 3 (Figures 7e, 7l, and 7m) when the upper-level shear minimizes (Figure 16). 7. Conclusions We have used four types of radar echo objects seen in the TRMM PR data to characterize the precipitating cloud population of the MJO. ISEs represent small, shallow precipitating clouds. DCC, WCC, and BSR echoes describe the deepest and widest convective and stratiform components of MCSs. Similar echo entities have been used to understand convective populations in the ITCZ [Schumacher and Houze, 2003], the Asian monsoon [Houze et al., 2007], and South America [Romatschke and Houze, 2010]. By examining these radar echo objects over the Indian Ocean and Western Pacific Ocean, we have shown that the areal coverage of each type of echo feature varies significantly with phase of the MJO. ISEs maximize during the suppressed stage while DCC, WCC, and BSR echoes maximize during the active stage in each geographic region. While these results are consistent with previous studies that used OLR and brightness temperature to show that the MJO varies the degree of mesoscale organization achieved by some elements of the cloud population [Chen et al., 1996; Houze et al., 2000; Benedict and Randall, 2007; Yuan and Houze, 2012], the TRMM PR enables us to determine how the relative amounts of different types of convective entities within the precipitating cloud population vary with phase of MJO, both as its convective coupling initiates over the CIO and later as it propagates through the WP. By using fourteen boreal winter seasons of TRMM PR data from , this study provides a robust climatology of the observed precipitating cloud population elements during each phase in the CIO and WP that can be used to verify numerical models and assess if 26

27 conditions observed during field campaigns such as TOGA-COARE, the Mirai Indian Ocean cruise for study of the MJO-convection Onset (MISMO), and the Dynamics of the Madden- Julian Oscillation/ARM MJO Investigation Experiment/Cooperative Indian Ocean experiment on the Intraseasonal Oscillation in the Year 2011 (DYNAMO/AMIE/CINDY-2011) are representative of climatological conditions. Regardless of geographic region and phase of the MJO, ISEs are found always to be present and are especially concentrated within the ITCZ portions of each oceanic region. In terms of number, ISEs are the most frequent component of the precipitating cloud population at all times. However, ISEs are also the most variable component of the precipitating cloud population in terms of number and tend to maximize during the suppressed stage of the MJO. BSR echoes dominate the variability of the precipitating cloud population in terms of areal coverage and are most prevalent during the active stage of the MJO. DCC and WCC echoes also maximize during the active stage and are more frequent and variable than BSR echoes in terms of number. While the variability of each echo entity has a similar magnitude across all geographic regions, subtle differences in the convective population between these regions exist. In the CIO, the spatial distribution of echoes is relatively homogeneous, WCC echoes are more common than DCC echoes, and ISEs peak shortly after the suppressed stage of the MJO. Also, a simultaneous peak in DCC, WCC, and BSR echoes characterizes the active stage of the MJO, which suggests that the precipitating cloud population of the active phase contains MCSs of all sizes and maturations in this region. In the WP, the geographic distribution of echoes and their relative amounts are inhomogeneous. DCC, WCC, and BSR echoes preferentially occur along a diagonal from the warm-pool in the NWWP to the SPCZ in the SEWP. Additionally, DCC echoes are more frequent than WCC echoes and ISEs peak shortly before the active stage of the 27

28 MJO. During the active stage in the NWWP and SEWP BSR echoes maximize one phase before WCC echoes, which suggests that extremely large MCSs are more common at the beginning of the active stage and smaller MCSs are more common as the active stage concludes. The NEWP displays the least systematic variability with phase of the MJO. Thus, the precipitating cloud population changes in a Lagrangian and Eulerian point of view with phase of the MJO. In this study, the large-scale atmospheric conditions are analyzed by compositing ERAinterim reanalysis data from the same fourteen boreal seasons that are used to characterize the precipitating cloud population from the TRMM PR data. We find that the composition of the precipitating cloud population during the initiation and propagation stages of the MJO exhibits a systematic relationship with large-scale atmospheric conditions similar to those found in TOGA- COARE rawinsonde data [Lin and Johnson, 1996]. Low-level westerlies and upper-level easterlies peak just after deep convective echo entities maximize, signaling the conclusion of organized convection in each geographic region. Additionally, mid-tropospheric relative humidity, low-level ( hpa) shear, and upper-level ( hpa) shear are found to vary systematically with the occurrence of ISEs, DCCs, WCCs, and BSR echoes. While the relationships summarized below appear to account consistently for the variability in ISEs and BSR echoes during all phases of the MJO, the relationships do not necessarily apply to the secondary DCC and WCC maxima observed during the suppressed stage in the CIO and NWWP, respectively. ISEs are profusely present in all phases of the MJO in both the CIO and WP. However, they maximize during the suppressed stage when the mid-troposphere is driest and the low-level shear is strongest. This maximum in the suppressed stage is not related to having more echo-free space available for ISE development. 28

29 DCC echoes in the CIO and SEWP occur most frequently during the active stage when the low-level ( hpa) shear is strong and the mid-troposphere is moist. A statistically significant peak in DCC does not occur in the NWWP, likely because the inherent instability of the warm-pool in that region allows DCC echoes to be continually present. There is no coherent relationship between DCC echoes, midtropospheric moisture, and low-level shear in the NEWP. WCC echoes are also most prevalent during the active stage when strong large-scale low-level ( hpa) shear occurs within a moist mid-troposphere. Winds aloft transported in MCS downdrafts to the surface create surface convergence and convective initiation. This implies that large-scale conditions foster gust front convergence in the vicinity of older convection within an environment with favorable moisture conditions. Thus, the initiation of new convection and greater WCC echo occurrence is expected. BSR echoes preferentially occur when a moderate amount of large-scale upper-level ( hpa) shear occurs within a moist mid-troposphere that has strong low-level shear. While strong low-level shear aids in convective initiation and moderately strong upper-level shear aids in mesoscale organization, we suggest that excessively strong upper-level shear disconnects the stratiform region from its convective moisture source and causes the frequency of BSR echoes to significantly decline. While strong upper-level shear has been found to favor stratiform convection in numerous studies [e.g. Saxen and Rutledge, 2000], the literature has not formally stated whether too much upper-level shear can be inimical to the mesoscale organization. If very strong upper-level shear is shown to be detrimental to MCSs, 29

30 the changing wind profile associated with a developing or propagating MJO could greatly influence the nature of the convective coupling since the heating feedback of large, organized mesoscale systems is notably more top heavy in nature [Houze, 1982]. Evidence in support of an upper-level shear threshold may be found in previous studies. For example, Lin and Johnson [1996] showed that the maxima in hpa shear lags the precipitation maximum by 5-10 days, which is consistent with the maximum hpa shear occurring one phase after the BSR maximum. The one exception to this consistency in the present study is the NEWP, which contains the WP cold-tongue and lacks a coherent association between upper-level shear and BSR echoes. These associations enable us to understand why DCC, WCC, and BSR echoes simultaneously maximize in the CIO but BSR echoes maximize one phase earlier than DCC and WCC echoes in most of the WP. While low-level and upper-level shear maximize one phase after the BSR peak in both regions, mid-tropospheric relative humidity dramatically decreases one phase after the BSR peak in the CIO and two phases after the BSR peak in the WP. Thus, DCC and WCC echoes in most of the WP appear to persist one phase longer than these echoes in the CIO since mid-tropospheric moisture remains elevated longer in the WP than the CIO. Uniformly moist conditions exist below 800 hpa regardless of the presence of ISEs or deep convective entities, phase of the MJO, and geographic region. Thus, similar to Yuan and Houze [2012], we find that the MJO chiefly modulates mid-tropospheric moisture. The CIO and most of the WP is characterized by an increase in mid-tropospheric moisture one phase prior to a statistically significant rise in DCC, WCC, and BSR echoes. While the source of this moisture is unclear, three possible mechanisms include vertical motion induced by planetary waves, 30

31 horizontal advection, and/or moisture accumulation from convective bursts that occur on a shorter timescale than the MJO. The moisture in the mid-troposphere is greatest as DCC, WCC, and BSR echoes maximize during the active stage. Substantial drying only commences once the frequency of WCC and BSR echoes significantly decline. While this study identifies systematic associations among the precipitating cloud population and the large-scale relative humidity, it is, of course, unclear if the precipitating cloud population is causing or reacting to large-scale conditions. Yuan and Houze [2012] also found that mid-tropospheric moisture varies in conjunction with large, mature MCSs and encountered the same attribution difficulty. Their study suggests that MCS in the active stage are initially a reaction to an increase in the large-scale mid-tropospheric moisture from either non-local advection or accumulation from previous convection. This conclusion is based on the observation that mid-tropospheric moisture is elevated during the active stage of the MJO regardless of the presence of deep convection. Yuan and Houze [2012] also suggested that after the initial MCS increase during the active stage, mesoscale convection acts as a positive feedback for mid-tropospheric moisture. The results of the current study are consistent with these two hypotheses. DCC, WCC, and BSR echoes appear at first to be a reaction to mid-tropospheric moisture since the mid-troposphere moistens prior to a statistically significant rise in DCC, WCC, and BSR echoes. Additionally, mid-tropospheric moisture continues to slowly rise as DCC, WCC, and BSR occurrence significantly increases, which supports the positive moistureconvection feedback. The present study expands the interpretations of Yuan and Houze [2012] by highlighting how the occurrence of deep convective entities in the MJO are related to the interplay of midlevel moistening and large-scale shear at both lower and upper levels. Specifically, we suggest that environmental lower- and upper-level shear influences the increase 31

32 or decrease of the mesoscale convective systems that produce BSRs, which dominate the areal variability of the precipitating cloud population and drive the heating profile toward a top-heavy configuration. While observational studies can suggest hypotheses concerning the association between the precipitating cloud population, large-scale relative humidity, and large-scale shear, modeling efforts will be necessary in order to decisively establish the physical connections linking the observations of radar echoes, wind shear, and humidity. The DYNAMO/AMIE/CINDY campaigns provide a unique opportunity to compare the composites of the precipitating cloud population and large-scale atmospheric conditions presented in the current study with three individual MJO events that occurred over the central Indian Ocean from 1 October 2011 to 15 January 2012 using the dual polarimetric SPolKa radar and rawinsonde data on Addu Atoll (0.6 S, 73.1 E). The variability in the precipitating cloud population observed during DYNAMO/AMIE is quite similar to the composites presented in the current study. Powell and Houze [in preparation, 2012] used SPolKa S-band radar data to show that the areal coverage of stratiform precipitation dominates the variability of the precipitating cloud population. Additionally, the pattern of variability by phase (see their Figure 4) is extremely similar to the composite presented in Figure 8a of the current study. Zuluaga and Houze [in preparation, 2012] characterized the mesoscale echoes observed by the SPolKa radar as DCCs, WCCs, and BSRs and composited eleven rain events over a period of 48 hours centered on the rain maximum. Each of these rain events occurred during the active stage. Their results suggested that the precipitating cloud population systematically transitions from DCCs to WCCs to BSRs and confirmed that young MCSs are characterized by DCCs and WCCs and mature MCSs are characterized by BSRs. Given that each phase of the MJO lasts on average

33 days, the simultaneous maximization of DCC, WCC, and BSR echoes observed in the current study is consistent with the results of Zuluaga and Houze [in preparation, 2012]. In terms of mid-tropospheric moisture and upper-level shear, however, the magnitude and rapidity of fluctuations is notably greater during DYNAMO/AMIE/CINDY campaign. For example, Powell and Houze [in preparation, 2012] used rawinsonde data to show that uniformly moist conditions existed below 925 hpa and relative humidity from hpa abruptly and rapidly increased just prior to the active stage and slowly decreased after the active stage. However, results from the current study indicate that uniformly moist conditions exist below 800 hpa and the increase in mid-tropospheric relative humidity between hpa is large but less abrupt and rapid. Thus, the overall pattern of large-scale variability is similar, but the magnitude and time-scale differ. This difference is likely attributable to the smoothing that results from compositing fourteen years of MJO events that each have slightly different time scales. Thus, while the magnitude and rapidity of variability observed in our 14-year composites may differ from individual MJO events, we are confident that the general pattern of variability in the precipitating cloud population and large-scale environment described in the current study qualitatively describes the variability observed during individual MJO events. 6. Acknowledgements The authors would like to thank Stacy Brodzik for initiating this study, Manuel Zuluaga for downloading and converting the reanalysis data, Scott Powell for helpful comments on the manuscript, and Beth Tully for improving the graphics. ERA-interim reanalysis data was provided by the European Centre for Medium-Range Weather Forecasts. This research is sponsored by the Department of Energy Grants DE-SC / ER and DE-SC , 33

34 the DYNAMO - National Science Foundation Grant AGS , and by the Precipitation Measuring Mission - National Aeronautics and Space Administration (PMM-NASA) Grant NNX10AH70. 34

35 7. References Awaka, J., T. Iguchi, H. Kumagai, and K. Okamoto (1997), Rain type classification algorithm for TRMM Precipitation Radar, paper presented at Proc Int. Geoscience and Remote Sensing Symp.(IGARSS 97) Remote Sensing: A Scientific Vision for Sustainable Development, Singapore, IEEE, 4(6), Benedict, J. J., and D. A. Randall (2007), Observed characteristics of the MJO relative to maximum rainfall, J. Atmos. Sci., 64(7), Bladé, I. and D. L. Hartmann, (1993), Tropical intraseasonal oscillations in a simple nonlinear model, J. Atmos. Sci., 50(17), Dee, D. P., S. M. Uppala, A. J. Simmons, P. Berrisford, P. Poli, S. Kobayashi, U. Andrae, M. A. Balmaseda, G. Balsamo, P. Bauer, P. Bechtold, A. C. M. Beljaars, L. van de Berg, J. Bidlot, N. Bormann, C. Delsol, R. Dragani, M. Fuentes, A. J. Geer, L. Haimberger, S. B. Healy, H. Hersbach, E. V. Holm, L. Isaksen, P. Kallberg, M. Kohler, M. Matricardi, A. P. McNally, B. M. Monge-Sanz, J. J. Morcrette, B. K. Park, C. Peubey, P. de Rosnay, C. Tavolato, J. N. Thepaut, and F. Vitart, (2011), The ERA-Interim reanalysis: Configuration and performance of the data assimilation system, Quart. J. Roy. Meteor. Soc., 137(656), Del Genio, A. D., Y. Chen, D. Kim, and MS Yao (2012), The MJO transition from shallow to deep convection in CloudSat/CALIPSO data and GISS GCM simulations, J. Clim., 25(11), DeMott, C. A., and S. A. Rutledge (1998), The vertical structure of TOGA COARE convection. Part I: Radar echo distributions, J. Atmos. Sci., 55(17),

36 Chen, S. S., R. A. Houze, Jr., and B. E. Mapes (1996), Multiscale variability of deep convection in relation to large-scale circulation in TOGA COARE, J. Atmos. Sci., 53(10), Cheng, C.-P., and R. A. Houze, Jr. (1979), The distribution of convective and mesoscale precipitation in GATE radar echo patterns, Mon. Wea. Rev., 107(10), Gill, A. E. (1980), Some simple solution for heating-induced tropical circulation, Quart. J. Roy. Meteor. Soc., 106(449), Haertel, P. T., G. N. Kiladis, A. Denno, and T. M. Rickenbach (2008), Vertical-mode decompositions of 2-day waves and the Madden-Julian Oscillation, J. Atmos. Sci., 65(3), Houze, R. A., Jr., and C. -P. Cheng (1977), Radar characteristics of tropical convection observed during GATE: Mean properties and trends over the summer season, Mon. Wea. Rev., 105(8), Houze, R. A., Jr., C. P. Cheng, C. A. Leary, and J. F. Gamache, (1980), Diagnosis of cloud mass and heat fluxes from radar and synoptic data, J. Atmos. Sci., 37(4), Houze, R. A., Jr. (1982), Cloud clusters and large-scale vertical motions in the tropics, J. Meteor. Soc. Japan, 60(1), Houze, R. A., Jr. (1989), Observed structure of mesoscale convective systems and implications for large-scale heating, Quart. J. Roy. Meteor. Soc., 115(487), Houze, R. A., Jr. (1993), Cloud Dynamics, 573pp., Academic Press, San Diego, Calif. 36

37 Houze, R. A., Jr., S. S. Chen, D. E. Kingsmill, Y. Serra, and S. E. Yuter, (2000), Convection over the Pacific warm pool in relation to the atmospheric Kelvin-Rossby wave, J. Atmos. Sci., 57(18), Houze, R. A., Jr. (2004), Mesoscale convective system,, Rev. Geophys., 2004, RG4003, doi: /2004rg Houze, R. A., Jr., D. C. Wilton, and B. F. Smull, (2007), Monsoon convection in the Himalayan region as seen by the TRMM Precipitation, Mon. Wea. Rev., 133(627), Iguchi, T., R. Meneghini, J. Awaka, T. Kozu, and K. Okamoto, (2000), Rain-profiling algorithm for TRMM Precipitation Radar data, Adv. Space Res., 25(5), Iguchi, T., T. Kozu, R. Meneghini, J. Awaka, and K. Okamoto, (2000), Rain-profiling algorithm for the TRMM Precipitation Radar, J. Appl. Meteor., 39(12), Kingsmill, D. E., and R. A. Houze, Jr., (1999a), Kinematic characteristics of air flowing into and out of precipitating convection over the west Pacific warm pool: An airborne Doppler radar survey, Quart. J. Roy. Meteor. Soc., 125(556), Kim, D., K. Sperber, W. Stern, D. Waliser, I. -S. Kang, E. Maloney, W. Wang, K. Weickmann, J. Benedict, M. Khairoutdinov, M. I. Lee, R. Neale, M. Suarez, K. Thayer-Calder, and G. Zhang, (2009), Application of MJO simulation diagnostics to climate models, J. Clim., 22(23), 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(3),

38 Lau, K. M., L. Peng, C. H. Sui, T. Nakazawa (1989), Dynamics of super cloud clusters, westerly wind bursts, day oscillation and ENSO: An unified view, J. Meteor. Soc. Japan, 67(2), Lau, K. M., and H. T., Wu, (2010), Characteristics of precipitation, cloud, and latent heating associated with the Madden-Julian Oscillation, J. Clim., 23(3), Lin, J., B. Mapes, M. Zhang, M. Newmann, (2004), Stratiform precipitation, vertical heating profiles, and the Madden Julian Oscillation, J. Atmos. Sci., 61(3), 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(22), Madden, R. A. (1986), Seasonal variations of the day oscillation in the tropics, J. Atmos. Sci., 43(24), Madden, R. A., and P. R. Julian, (1971), Detection of a day oscillation in the zonal wind in the tropical Pacific, J. Atmos. Sci., 28(8), Madden, R. A., and P. R. Julian, (1972), Description of global-scale circulation cells in the tropics with a day period, J. Atmos. Sci., 29(5), Mapes, B. E., and R. A. Houze Jr, (1993), Cloud clusters and superclusters over the oceanic warm pool, Mon. Wea. Rev., 121(5), 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, , doi: /j.dynatmoce

39 Saxen, T. R., and S. A. Rutledge (2000), Surface rainfall-cold cloud fractional coverage relationship in TOGA COARE: A function of vertical wind shear, Mon. Wea. Rev., 128(2), Schumacher, C., and R. A. Houze, Jr., (2003), The TRMM Precipitation Radar s view of shallow, isolated rain, J. Appl. Meteor., 42(10), Stephens, G. L., P. J. Webster, R. H. Johnson, R. Engelen, and T. L'Ecuyer, (2004), Observational evidence for the mutual regulation of the tropical hydrological cycle and tropical sea surface temperatures, J. Clim., 17(11), Riley, E. M., B. E. Mapes, and S. N. Tulich, (2011), Clouds associated with the Madden-Julian Oscillation: A new perspective from CloudSat, J. Atmos. Sci., 68(12), Romatschke, U., and R. A. Houze, Jr., (2010), Extreme summer convection in South America, J. Clim., 23(14), Romatschke, U., S. Medina, and R. A. Houze, Jr., (2010), Regional, seasonal, and diurnal variations of extreme convection in the South Asian region, J. Clim., 23(2), Tian, B., D. E. Waliser, E. J. Fetzer, and Y. L. Yung, (2010), Vertical moist thermodynamic structure of the Madden-Julian Oscillation in atmospheric infrared sounder retrievals: An update and a comparison to ECMWF interim re-analysis, Mon. Wea. Rev., 138(12), Tromeur, E. and W. B. Rossow, (2010), Interaction of tropical deep convection with the largescale circulation of the MJO, J. Clim., 23(7),

40 Weickmann, K. M., G. R. Lussky, and J. E. Kutzbach, (1985), Intraseasonal (30-60 day) fluctuations of outgoing longwave radiation and 250-mb streamfunction during northern winter, Mon. Wea. Rev., 113(6), 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(8), Yuan, J., and R. A. Houze, Jr., (2010), Global variability of mesoscale convective system anvil structure from A-train satellite data, J. Clim., 23(21), Yuan, J., and R. A. Houze, Jr., (2012), Deep convective systems observed by A-Train in the tropical Indo-Pacific region affected by the MJO, J. Atmos Sci., revised. Yuter, S E., and R. A. Houze Jr. (1998), The natural variability of precipitating clouds over the western Pacific warm pool, Q. J. R. Meteorol. Soc., 124(545), Zhang, C. (2005), Madden-Julian Oscillation, Rev. Geophys., 43, RG2003, doi: /2004rg Zhang, G. J., and X. Song, (2009), Interaction of deep and shallow convection is key to Madden- Julian Oscillation simulation, Geophys. Res. Lett., 36, 9708, doi: /2009gl Figure Captions Figure 1. Physical map of the geographic regions analyzed. The central Indian Ocean (CIO) extends from 10 S - 10 N, 60 E - 90 E. The northwest portion of the WP (NWWP) extends from 0-10 N, 140 E E and includes the warm-pool. The northeastern portion of the WP 40

41 (NEWP) extends from 0-10 N, 156 E E and includes the cold-tongue. The southeastern portion of the WP extends from 10 S - 0, 156 E E and includes the South Pacific Convergence Zone (SPCZ). Figure 2. DJF composite of OLR and 850 hpa wind anomalies during each phase of the MJO from 1979 to 2001 from Wheeler and Hendon (2004). Negative OLR anomalies are shaded, positive OLR anomalies are hashed, and wind anomalies significant at the 99% confidence level are in black. The total number of days in each phase and magnitude of the largest wind anomaly is in the lower right of each panel. Figure 3. Geographic distribution of the average ISE frequency over the CIO and WP during phases 1, 3, 5, and 7 for the twenty samples made using the Monte-Carlo method. The frequency is defined as the number of TRMM PR pixels that contain an ISE normalized by the total number of TRMM PR pixels detected, which includes both echo-covered and echo-free pixels. The frequency is calculated individually for each 0.5 x 0.5 grid box and plotted as a percentage. Any grid box containing land is excluded and left blank. The black lines in the WP show the boundaries of the NWWP, NEWP, and SEWP. Figure 4. Same as Fig. 3 except for DCCs. Figure 5. Same as Fig. 3 except for WCCs. Figure 6. Same as Fig. 3 except for BSRs. Figure 7. (a) Total ISE frequency in the CIO. The frequency is defined as the number of TRMM PR pixels that contain an ISE normalized by the total number of TRMM PR pixels, which includes both echo-covered and echo-free pixels. The frequency is calculated across the entire 41

42 CIO region for each phase and reported as a percent. The blue lines show the frequency-phase series for the twenty Monte-Carlo samples, the black line is the mean, and the dashed red lines are the upper and lower 99% confidence limits based on the student s t-statistic at the 99% confidence level. (b) Same as (a) except for DCCs in the CIO. (c) Same as (a) except for WCCs in the CIO. (d) Same as (a) except for BSRs in the CIO. (e-h) same as (a-b) except in the NWWP. (i)-(l) same as (a)-(b) except in the NEWP. (m-p) same as (a-b) except in the SEWP. Figure 8. (a-d) The average total frequency (thick line) and 99% confidence interval ISEs (red), DCCs (dark blue), WCCs (cyan), or BSRs (green) in terms of areal coverage in the CIO, NWWP, NEWP, and SEWP for the twenty samples given as a percent. (e-h) The average number (thick line) and 99% confidence interval (thin line) of DCCs (dark blue), WCCs (cyan), and BSRs (green) in the CIO, NWWP, NEWP, and SEWP for the twenty samples. (i-l) The average number (thick line) and 99% confidence interval of ISEs in the CIO, NWWP, NEWP, and SEWP for the twenty samples. Figure 9. The average relative humidity profile for each oceanic grid point in the ERA-interim reanalysis for the CIO, NWWP, NEWP, and SEWP. Dashed lines show the convectively suppressed phases for the respective region. Solid lines depict the convectively active phases with the solid green line representing the phase with the largest areal coverage of BSRs in each geographic region. The relative humidity is given as a percent. Figure 10. The average direction (vectors) and zonal speed (shading) of the 1000 hpa wind in the CIO and WP for all days during phases 2, 4, 6, and 8 in ms -1. Figure 11. Same as Fig. 10 expect for the average 200 hpa wind. Figure 12. Same as Fig. 9 expect for zonal wind in ms

43 Figure 13. The average magnitude (shading) and direction (vectors) of the hpa shear in the CIO for all days during each phase in ms -1. Figure 14. Same as Fig. 13 except for hpa shear. Figure 15. Same as Fig. 13 except located in the WP. The black lines show the boundaries of the NWWP, NEWP, and SEWP. Figure 16. Same as Fig. 13 except for hpa shear in the WP. The black lines show the boundaries of the NWWP, NEWP, and SEWP. 43

44

45