Abstract

Ground-based high temporal and vertical resolution datasets from observations during 2002–07 at the Atmospheric Radiation Measurement (ARM) tropical western Pacific (TWP) site on Manus Island are used to examine the characteristics of clouds and rainfall associated with the active phase of the Madden–Julian oscillation (MJO) passing over Manus. A composite MJO event at Manus is developed based on the NOAA MJO index 4 and precipitation using 13 events. The cloud characteristics associated with the active phase of the MJO at Manus show a two-phase structure as the wave passes over Manus. During the development phase, congestus plays an important role, and the enhanced convection is located between surface westerly and easterly wind anomalies (type-I structure). During the mature phase, deep convection is the dominant cloud type, and the enhanced convection is collocated with the westerly wind anomalies (type-II structure). Consistent with this two-phase structure, the heavy rainfall frequency also shows a two-peak structure during the MJO disturbance, while light rainfall does not show a clear relation to the intraseasonal disturbance associated with the MJO. In addition, a positive relationship between the precipitation rate and precipitable water vapor exists at Manus, and the atmospheric column is less moist after the passing of the MJO convection center than before.

1. Introduction

The Madden–Julian oscillation (MJO) (Madden and Julian 1972, 1994) is a dominant mode of intraseasonal (30–90 days) variability in the tropical atmosphere. Its characteristics and structure have been well documented by previous studies (e.g., Kiladis et al. 2005; Lau and Waliser 2005; Zhang 2005; Wu et al. 2007; Jiang et al. 2009; Zhang et al. 2010). The organization and evolution of tropical convection is a major component of the MJO and generally includes several eastward-moving super cloud clusters with several westward-moving cloud clusters within each super cluster (e.g., Nakazawa 1988). The MJO-related convection starts from the Indian Ocean, propagates through the western Pacific Ocean, and decays in the central Pacific with an eastward speed of about 5 m s−1 (Weickmann et al. 1985; Knutson et al. 1986). The convective aspects of the MJO, observed in fields such as outgoing longwave radiation (OLR), surface precipitation, and latent heat flux, generally end around the date line while the atmospheric circulation (wave activity) associated with the MJO, observed in fields such as the 200-hPa velocity potential and zonal wind and the 850-hPa zonal wind, can continuously propagate farther east (e.g., Maloney and Hartmann 2001; Sperber 2004; Zhang and Dong 2004).

In general, the MJO displays a zonally asymmetric structure, which corresponds to the eastward propagation of the MJO convective center: east of the deep convection associated with the MJO, low-level moisture convergence and upward motion anomalies favor the development of new convection; while west of the MJO deep convection, low-level divergence, downward motion, and dry anomalies discourage the development of convection (e.g., Zhang 2005; Kiladis et al. 2005). Four theoretical models of the MJO, showing proposed relationships between deep convection and zonal wind anomalies, are summarized in Zhang and Anderson (2003) and Zhang (2005). In the type-I model, deep convection with enhanced precipitation and reduced OLR is located between the surface westerly and easterly anomalies. In the type-II model, the deep convection system is collocated with the prevailing westerly anomalies, while the easterly anomalies occur to the east (ahead of) the deep convection center. The type-III and -IV models have not been observed, and will not be discussed further. The type-I model is more often observed in the Indian Ocean while the type-II model often occurs over the western Pacific (e.g., Zhang and McPhaden 2000). However, a single MJO event may in fact display different structures (type I, II, or in between) during different stages of the MJO life cycle.

Theories have been proposed to explain the physical mechanism of the MJO: wave–conditional instability of the second kind (CISK; Lau and Peng 1987), wind–evaporation feedback (Emanuel 1987; Neelin et al. 1987), frictional moisture convergence (e.g., Wang 1988), and the discharge–recharge mechanism (Bladé and Hartmann 1993; Maloney and Hartmann 1998). Although none of these theories has been generally accepted to explain all aspects of the MJO, each one of them contributes to part of the MJO understanding. Hu and Randall (1994, 1995) suggested that the MJO is an atmospheric response to forcing sources and that nonlinear interactions among radiation, cumulus convection, and surface moisture fluxes are responsible for the low-frequency oscillation of the forcing sources. Bladé and Hartmann (1993) introduced a discharge–recharge mechanism, which proposes that the period of the low-frequency oscillation of convective heating is determined by the discharge time of convective stabilization together with the recharge time of moist static instability. Follow-on studies about this theory suggest a gradual building up of low-level warming and moistening cloud processes (e.g., shallow cumulus and cumulus congestus) that pave the way for the subsequent MJO deep convection (e.g., Benedict and Randall 2007).

The tropics contain a wide distribution of convective cloud types. Several authors have examined the possible role of the distribution of cloud types including shallow cumulus (tops near the trade inversion) and cumulus congestus (moderately deep convective clouds with tops around 5–6 km or 0°C) in the modulation of tropical convection and the MJO. Kikuchi and Takayabu (2004) propose five stages for the MJO convection using data from Tropical Ocean and Global Atmosphere Coupled Ocean–Atmosphere Response Experiment (TOGA COARE). Their developing stage (3–4 days) is related to the melting level and is dominated by the congestus. They suggest that the moistening process between the melting level and trade inversion is the most influential process for further convective development. Based on analysis of the Atmospheric Infrared Sounder (AIRS) datasets, Tian et al. (2006) suggest that the low-level moisture preconditioning in front of the MJO deep convection may be caused by congestus and shallow cumulus. Benedict and Randall (2007) further discuss this gradual increase of lower-tropospheric heat and moisture by the congestus and shallow cumulus ahead of the MJO deep convection based on the Tropical Rainfall Measuring Mission (TRMM), Global Precipitation Climatology Project (GPCP) and 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) datasets. Chen and Del Genio (2009) show an increase of congestus before the MJO convection center and reduction afterward. In addition, congestus and shallow cumulus dominate before the MJO convection center; from lag −10 to −5 days, the relative frequency of occurrence (RFO) for congestus is the largest.

The objective of this paper is to use high-resolution radar data to further explore the atmospheric and cloud characteristics when the convectively disturbed phase of the MJO passes over a fixed site. A thorough understanding of the MJO process is needed to fully evaluate the proposed mechanisms for the MJO and to pave the way for better MJO simulations. We examine the cloud variation associated with the propagation of the MJO past a fixed location using high-resolution ground-based datasets from the Atmospheric Radiation Measurement (ARM) tropical western Pacific (TWP) site on Manus Island, Papua New Guinea. Previous time series analysis by Wang et al. (2011) indicates that Manus Island experiences a strong MJO signal in shortwave cloud radiative forcing and fractional sky cover. The long time series of observational data allows us to combine multiple MJO events over 6 winter seasons to develop a composite analysis of the active phase of the MJO as it passes over the Manus site, while the high resolution of the data allows examination of cloud processes on a daily level, rather than the pentad scale often used. Section 2 describes the data and methods used to build the MJO composite. Section 3 shows the results of the MJO composite analysis, with a focus on cloud characteristics, and section 4 presents a summary and discussion of the results.

2. Data and methods

a. ARM TWP datasets

ARM TWP ground-based datasets from 2002 to 2007 at Manus Island, Papua New Guinea (2.060°S, 147.425°E), are used in this paper to create a composite view associated with the MJO disturbance (Table 1). Variables presented include cloud frequency, maximum precipitation, specific humidity, temperature, and zonal wind. The cloud frequency data is based on a combined product from the Millimeter Wavelength Cloud Radar (MMCR) and micropulse lidar (MPL). Clutter-screened reflectivity from the 35-GHz MMCR (at 90 m, 10-s resolution) and attenuated backscatter from the 532-nm MPL (30 m, 30-s resolution) are averaged to a common temporal (120 s) and vertical (~30 m) grid. A cloud is identified from the radar as any point with reflectivity >−50 dBZ, which may also include precipitation. A cloud is identified from the lidar using the algorithms of Comstock and Sassen (2001) and Wang and Sassen (2002). Grid points containing rain or drizzle are considered “cloud” in all of the following analysis. Cloudy layers are defined as four or more vertically adjacent cloud bins (≥120 m); two cloudy layers separated by three or fewer clear bins are merged into a single layer.

Table 1.

ARM TWP Manus Island datasets.

ARM TWP Manus Island datasets.
ARM TWP Manus Island datasets.

Precipitation rate is measured at the site by an optical rain gauge (ORG). The ORG measures at 1-s resolution and reports minimum, mean, and maximum precipitation over a 60-s interval. Here, we use the maximum precipitation rate within a 2-min interval (aligned with the cloud dataset above). Because of the uncertainty of the ORG, precipitation rate values less than 0.1 mm h−1 are considered to be zero.

Vertical profiles of relative humidity, temperature, and wind are measured by the balloon-borne sounding system (sonde) and are interpolated to 125-m resolution between 0 and 20 km. For sondes that do not reach 20 km, grid points above the sonde-top height are filled with missing values. As a basic quality control check, if the temperature at the surface is less than 10°C or the temperature value at 5 km is larger than 10°C, the sonde is not used. Typically, two balloons per day are released during the study period, at approximately 0000 and 1200 UTC. Specific humidity is calculated based on the relative humidity and temperature, and the mean annual cycle (based on 2002–07) is removed from the specific humidity, temperature, and wind fields.

b. Other data

To examine the large-scale spatial structure of convection associated with the MJO near Manus, we also use daily National Oceanic and Atmospheric Administration (NOAA) interpolated OLR from January 2002 to December 2007 (Liebmann and Smith 1996) at 2.5° latitude × 2.5° longitude resolution.

c. MJO events

To identify the strongest local convection associated with the MJO disturbance when it passes Manus, we first use the NOAA MJO index 4 to identify the periods with strong MJO signals in the general region of Manus. We then assign the local largest daily precipitation rate during this period as the peak of the local MJO event to construct a composite MJO event at Manus. The NOAA MJO indices are obtained online (at http://www.cpc.noaa.gov/products/precip/CWlink/daily_mjo_index/pentad.shtml). As described on this website, these indices are calculated based on an extended empirical orthogonal function (EEOF) analysis (e.g., Weare and Nasstrom 1982) of the tropical 200-hPa velocity potential anomalies between 30°S and 30°N during neutral or weak El Niño–Southern Oscillation (ENSO) winters from 1979 to 2000. Pentad values of the 200-hPa velocity potential anomalies are then regressed onto the 10 time-lagged patterns of the first EEOF to create ten MJO indices. Each index is then normalized by its standard deviation over the 1979–2000 period. Large negative values of index 4 indicate MJO-enhanced convection centered near 140°E (near Manus Island).

We use pentad index 4 values less than −1 between November and April to identify when the active phase of the MJO is near Manus Island (Table 2). Since this is a pentad index, and also indicates the peak MJO activity over a large spatial region, we use the largest Manus daily precipitation rate within that pentad to determine the associated peak day for each MJO event. Each MJO event is defined to include the 25 days before and after the MJO peak convection (51 days per event). Time series of the daily-averaged vertical profiles of specific humidity, temperature, wind, and cloud frequency are calculated for each MJO event.

Table 2.

19 MJO cases during November–April 2002–07. Boldface indicates the MJO events with at least 40% of cloud-frequency data available.

19 MJO cases during November–April 2002–07. Boldface indicates the MJO events with at least 40% of cloud-frequency data available.
19 MJO cases during November–April 2002–07. Boldface indicates the MJO events with at least 40% of cloud-frequency data available.

d. Composite process

To develop a picture of the key structures in the thermodynamic, wind, and cloud fields associated with the propagation of the active phase of the MJO over Manus, we average the 51-day time series for each MJO event discussed above into a composite time series. Also, we remove the events that have a significant amount of missing data based on the daily datasets. Figure 1 illustrates the percentage of available data for cloud frequency, specific humidity at 5 km, and precipitation rate for each event. Clearly, the cloud frequency data are the limiting factor. We require each event to have at least 40% of the cloud frequency data available, resulting in 13 events (shown in boldface in Table 2) out of the original 19 events. For each variable (cloud frequency, specific humidity, OLR, etc.), all of the daily data available during these 13 events are averaged to produce composite fields associated with the MJO disturbance over Manus Island. Although we only have 13 events using the NOAA MJO index, the basic features of the composite are robust and are not sensitive to small variations in the number of events (not shown). The significance of the composite cloud fields is discussed further in section 3e.

Fig. 1.

Percentage of available radar data per event at Manus Island for (a) cloud frequency, (b) specific humidity at 5 km, and (c) available days per event for precipitation rate. Nineteen events (x axis) are presented based on the NOAA MJO index 4 and local precipitation. Each event includes 51 days.

Fig. 1.

Percentage of available radar data per event at Manus Island for (a) cloud frequency, (b) specific humidity at 5 km, and (c) available days per event for precipitation rate. Nineteen events (x axis) are presented based on the NOAA MJO index 4 and local precipitation. Each event includes 51 days.

3. Composite analysis of MJO

In this section, the composite analysis based on the 13 events described above is examined to show the characteristics of the convective and thermodynamic fields before, during, and after the passage of the mature phase of the MJO over Manus Island. First, the general structure of the active MJO is examined using the OLR, cloud frequency, zonal wind, specific humidity, and temperature anomalies. Then, the total cloud frequency is broken down by cloud type to examine the contribution of different cloud types associated with the MJO disturbance. The relationship between rainfall and clouds and water vapor during the passage of the MJO active phase is shown. Finally, use of a bandpass filter to construct the composite thermodynamic fields and an examination of the significance of the composite cloud fields relative to a randomly constructed composite are discussed.

a. Composite MJO event

To indicate the large-scale pattern of convective cloudiness associated with the propagation of the MJO over Manus Island, the time–space variation of the composite OLR based on 13 events is shown in Fig. 2 from lag −25 to +20 days with a 5-day interval. A minimum OLR band is generally located over the eastern Indian Ocean and western Pacific, which represents the strong convection center of the active MJO and is consistent with previous studies (e.g., Deng and Wu 2010). The MJO associated strong convection, initially observed in the Indian Ocean near the equator before lag −10 days, propagates eastward to the western Pacific from lag −5 to +15 days. The OLR shows the minimum value at Manus Island (blue cross in Fig. 2) around lag 0 days, which suggests that the use of the NOAA index 4 and precipitation to build the MJO event composite at Manus is appropriate.

Fig. 2.

Spatial pattern of NOAA OLR (W m−2) for the mean composite MJO event based on 13 MJO events. The blue cross indicates Manus Island.

Fig. 2.

Spatial pattern of NOAA OLR (W m−2) for the mean composite MJO event based on 13 MJO events. The blue cross indicates Manus Island.

The composite cloud frequency, composite zonal wind, specific humidity, and temperature anomalies at Manus Island associated with the 13 MJO events over Manus are presented in Fig. 3. Because of data limitations (e.g., missing values), it is difficult to define a mean for the cloud frequency, so only the zonal wind, specific humidity, and temperature fields have the mean removed. The significance of the composite cloud frequency and the effect of bandpass filtering the other variables are discussed in section 3e.

Fig. 3.

Composite analysis of (a) cloud frequency (%), (b) specific humidity (g kg−1), (c) zonal wind (m s−1), and (d) temperature (°C). The mean annual cycle has been removed from (b)–(d).

Fig. 3.

Composite analysis of (a) cloud frequency (%), (b) specific humidity (g kg−1), (c) zonal wind (m s−1), and (d) temperature (°C). The mean annual cycle has been removed from (b)–(d).

The composite cloud frequency (Fig. 3a) reveals the percentage of time that a cloud was detected at each altitude by the MMCR or MPL. During the passage of the MJO disturbance, large variability in the cloud frequency exists in the upper troposphere. The cirrus clouds appear to be organized primarily on a shorter time scale than the intraseasonal scale, as there are numerous peaks in cirrus frequency throughout the composite period. Previous analysis has shown that much of the cirrus observed at Manus Island is not formed locally, but is associated with convection over the Maritime Continent (Mather 2005). The cirrus time scale may also be influenced by equatorial Kelvin wave dynamical process in the tropical tropopause layer over the western Pacific (e.g., Fujiwara et al. 2009), although this is more likely to impact cirrus above 14 km.

The strongest peak in column cloud frequency, with a high frequency of cloud existing at all altitudes, is located around lag 0 days. This peak is associated with the MJO deep convection disturbance and illustrates that the combination of the large-scale MJO index and the local precipitation used in this study is relevant for the local-scale cloudiness observed at Manus Island. Another peak in column cloud frequency is located from lag −6 to −4 days, and owing to the lower cloud-top heights, is associated with congestus clouds (also see further discussion of cloud types in section 3b). Between these two peaks, there is less convective cloud but a similarly high frequency of midlevel cloud (around lag −3 days) that is likely associated with the detrainment from the preceding convective clouds. After the deep convection peak, there is a high frequency of both midlevel and high clouds, again likely associated with detrainment from the convective clouds.

Analysis of the zonal wind anomalies (Fig. 3c) shows that in the lower troposphere, easterly anomalies are located before the MJO convection center passes over Manus, while robust westerly anomalies appear during and after it. In the upper troposphere, the direction of wind anomalies is opposite to that in the lower troposphere, which indicates a baroclinic wind anomalous structure and suggests a lower-troposphere convergence and upper-troposphere divergence anomalous structure when coupled with the eastward-propagating MJO convection center (e.g., Kiladis et al. 2005). In addition, the prevailing surface westerly anomaly at Manus is collocated with the MJO deep convection center around lag 0 days, which agrees with the type-II MJO structure in Zhang and Anderson (2003) and Zhang (2005).

Corresponding to the two peaks of column cloud frequency, specific humidity (Fig. 3b) also shows two moistened columns. One column is around from lag −2 to 0 days and is collocated with the MJO-related deep convection center. The other column is around lag −6 days, and is associated with the congestus peak. In addition, the lower-troposphere moisture anomaly around lag −10 days in front of the two peaks suggests preconditioning of moist static energy (MSE; specific humidity is the dominant term) before the active phase of the MJO passes over Manus and is consistent with previous results that the building up of MSE or moistening before the onset of strong MJO-related deep convection appears necessary (e.g., Sobel et al. 2004; Peters and Neelin 2006; Maloney 2009). Between these two peak moist columns, positive midtroposphere moisture anomalies appear with slightly dry anomalies occurring in the lower troposphere around lag −4 days, corresponding to the relative minimum in cloud amount between 0 and 4 km. The positive moisture anomalies in the midtroposphere may be caused by the upward transportation of moisture by convective clouds (e.g., through congestus-associated detrainment, downdrafts, and/or evaporation of precipitation) or caused by advection of moisture into the region. This moister midtroposphere favors the development of future deep convection because an air parcel can keep its positive buoyancy longer in a wet environment than in a dry environment (e.g., Kikuchi and Takayabu 2004). A strong moisture anomaly also exists around lag +6 days in the midtroposphere after the deep convection peak, which is caused by decayed deep convection. The composite specific humidity shows that the atmosphere is wetter in front of the MJO deep convection (~lag 0 days) than behind it, which agrees with previous results (e.g., Kiladis et al. 2005).

The final composite variable based on the 13 events is the temperature anomaly (Fig. 3d). At Manus Island, the temperature anomalies show large variability around 2 km (lower troposphere), 12 km (upper troposphere), and 18 km (tropopause), which matches well with previous observational studies in the western Pacific (e.g., Lin et al. 2005). Near the tropopause, the amplitude of the temperature anomalies is the largest relative to all other levels, and the tropopause temperature anomalies are out of phase with the upper-troposphere temperature anomalies (e.g., Lin et al. 2005). In addition, these temperature anomalies agree with the eastward tilt with height structure at the tropopause found in previous studies [e.g., Fig. 8b in Weare (2010)] and suggest the importance of Kelvin wave dynamics for the MJO. In the upper troposphere, the positive temperature anomalies are present before and during the MJO deep convection over Manus (e.g., Kiladis et al. 2005; Zhang 2005). Note that cold temperature anomalies after the MJO deep convection center passes Manus illustrate a more stabilized atmosphere, and agree with the suppressed period behind the MJO-related deep convection in previous results (e.g., Benedict and Randall 2007).

The variation of cloud frequency, zonal wind, specific humidity, and temperature anomalies was first examined separately using the composite height–time–lag plots. To further understand how those variables vary together with respect to the time lag, 850-hPa zonal wind anomalies, vertically averaged (column) cloud frequency, vertically averaged (column) specific humidity, and temperature anomalies are presented in Fig. 4 along with the composite precipitation rate. All variables except specific humidity show their largest peaks around lag 0 days corresponding to the strong deep convection associated with the MJO disturbance. The secondary peak in front of the deep convection is between lag −10 days and lag −4 days, corresponding to the congestus clouds column around lag −6 days in Fig. 3a. Based on these results and previous studies (e.g., Kikuchi and Takayabu 2004), we define two key phases associated with the passage of the active MJO: the development phase (from lag −10 to −4 days) and the mature phase (from lag −4 to +4 days). During the development phase, the peaks in cloud frequency, moisture, and precipitation rate exist in front of the westerly wind anomaly peak (Figs. 3 and 4), which agree with classic MJO structure (e.g., Madden and Julian 1972) and type I in Zhang and Anderson (2003). During the mature phase, cloud frequency, moisture, precipitation rate, and temperature peaks are in phase with the westerly wind anomaly peak, and fit the type-II MJO structure, which is often observed over the tropical western Pacific. These results suggest that type-I and -II MJO structures exist at Manus Island for different phases, with type-I structures occurring in the development phase, before the MJO deep convection center reaches Manus, and type-II structures occurring in the mature phase, when the deep convection center is over Manus.

Fig. 4.

Composite analysis for precipitation rate (green; mm h−1), 850-hPa zonal wind anomalies (orange, dashed; m s−1) and vertically averaged cloud frequency (black; 10%), specific humidity anomalies (blue; 0.1 g kg−1), and temperature anomalies (red; 0.1°C). All variables have been scaled to fit on the same axis. The vertical blue lines show the development phase (from lag −10 to −4 days) and mature phase (from lag −4 to +4 days).

Fig. 4.

Composite analysis for precipitation rate (green; mm h−1), 850-hPa zonal wind anomalies (orange, dashed; m s−1) and vertically averaged cloud frequency (black; 10%), specific humidity anomalies (blue; 0.1 g kg−1), and temperature anomalies (red; 0.1°C). All variables have been scaled to fit on the same axis. The vertical blue lines show the development phase (from lag −10 to −4 days) and mature phase (from lag −4 to +4 days).

b. Cloud types associated with the composite MJO event

To further illustrate the cloud characteristics associated with the MJO propagation past Manus, we define cloud types using the top, base, and thickness of each cloud layer. The definitions of different cloud types are based on previous studies (e.g., Johnson et al. 1999; Luo and Rossow 2004; Riley and Mapes 2009). A schematic plot of the types (shallow clouds, congestus, altocumulus/stratus, deep convection (includes stratiform rainfall), cirrostratus, and cirrus) is shown in Fig. 5. We note that these cloud type definitions are for convenience, and do not correspond exactly to meteorological cloud types; however, they represent distinct differences in the cloud structures associated with the passage of the active MJO. The composite analysis of the frequency of different cloud types before, during, and after the MJO disturbance is presented in Fig. 6, and the corresponding number of cloud layers per day (hereafter cloud number) for the first phase (period before the development phase; from lag −25 to −11 days), development phase, mature phase, and fourth phase (period after the mature phase; from lag +5 to +25 days) are summarized in Table 3. The basic feature of the results presented is not sensitive to small (1 km) changes in the cloud-base/top-height choices (not shown).

Fig. 5.

Schematic for cloud type definitions based on the cloud top, base, and thickness. To allow thin layers near freezing level, altocumulus may have bases <4 km, if tops are >4 km and thicknesses are <1.5 km. Similarly, cirrus clouds may have bases <8 km, if tops are >8 km and thicknesses are <1.5 km.

Fig. 5.

Schematic for cloud type definitions based on the cloud top, base, and thickness. To allow thin layers near freezing level, altocumulus may have bases <4 km, if tops are >4 km and thicknesses are <1.5 km. Similarly, cirrus clouds may have bases <8 km, if tops are >8 km and thicknesses are <1.5 km.

Fig. 6.

Composite analysis of cloud number for shallow clouds (blue), congestus (green), deep convection (red), altocumulus and altostratus (orange, dashed), cirrostratus (blue, dashed), and cirrus (red, dashed). A 3-day running mean has been applied on all variables. The vertical blue lines show the development and mature phases.

Fig. 6.

Composite analysis of cloud number for shallow clouds (blue), congestus (green), deep convection (red), altocumulus and altostratus (orange, dashed), cirrostratus (blue, dashed), and cirrus (red, dashed). A 3-day running mean has been applied on all variables. The vertical blue lines show the development and mature phases.

Table 3.

Cloud number of different cloud types. Here, ⇒ means the cloud number increases with time from phase to phase; corresponding “%” represents the percentage of cloud number increase divided by the cloud number of start phase.

Cloud number of different cloud types. Here, ⇒ means the cloud number increases with time from phase to phase; corresponding “%” represents the percentage of cloud number increase divided by the cloud number of start phase.
Cloud number of different cloud types. Here, ⇒ means the cloud number increases with time from phase to phase; corresponding “%” represents the percentage of cloud number increase divided by the cloud number of start phase.

Cirrus and shallow clouds are described first as they are not as sensitive to the intraseasonal disturbance at Manus Island as other types of clouds. Cirrus clouds are the most prevalent type of clouds at Manus. The cirrus number during the MJO events is large at around 624 per day (mean cirrus in Table 3), but the change in cirrus number with phase during the passage of the MJO disturbance is small. Over the entire composite, the maximum change between phases is just 3%. This small variability suggests that cirrus clouds are not strongly controlled by the intraseasonal disturbance at Manus Island and cirrus frequency is likely more related to synoptic variability. Shallow clouds also show little variability. Although they increase slightly from the first phase through the mature phase, the maximum variability between phases is only 9%, much less than the other cloud types discussed below.

All other types of clouds (congestus, deep convection, altocumulus/stratus, and cirrostratus) show comparably strong variability associated with the intraseasonal disturbance over Manus, with increase in cloud frequency from the first phase through the mature phase, and then a decrease in frequency after the mature phase. During the development phase, the peak in congestus (around lag −6 days) followed by the peak in altocumulus/stratus (lag −5 days) explains 24% of the total cloud number in the development phase. The explained percentage of total cloud number by deep convection and anvil in the development phase is much smaller (9%), although the anvil shows the largest increase in variability. This indicates that the congestus and altocumulus/stratus play important roles in the development phase.

From the development phase to the mature phase, the cloud number for all four types of clouds increases. The increases in congestus and associated altocumulus/stratus (30% and 13%, respectively) are smaller than the increases in deep convection (131%) and associated cirrostratus (28%). This large increase in deep convection in the mature phase suggests that the deep convection dominates the mature phase. From the mature phase to the fourth phase, the frequency of all cloud types decreases, especially deep convection (−54%). This decrease corresponds to the convection-suppressed period behind the MJO deep convection. Note that there is a congestus peak around lag +7 days during the fourth phase in Fig. 6 that, corresponding to the moisture anomalies and cloud frequency in Fig. 3, is caused by the decayed deep convection.

c. Rainfall

To further examine the relationship between rainfall and different cloud types before, during, and after the MJO active phase, the rainfall data are separated based on the intensity. Heavy rainfall is defined as a precipitation rate larger than 1 mm h−1 (which is the 80th percentile of rain intensity during the 2002–07 period) and light rainfall as precipitation rate between 0.1 and 1 mm h−1. There is a visible peak in the total rainfall frequency associated with the MJO deep convection around lag 0 days (Fig. 7a), while there appear to be separate peaks in the heavy rainfall frequency (Fig. 7b) during the development (lag −5 days) and mature phases (lag 0 days) of the MJO. The light rainfall frequency signal is less clear (Fig. 7c). There is no visible peak in the light rainfall frequency during the development phase, and a pretty small peak in the mature phase around lag +1 day. The total and heavy rainfall frequency features more clearly show the intraseasonal disturbance than the light rainfall frequency at Manus site.

Fig. 7.

Composite analysis for (a) total, (b) heavy, and (c) light rainfall percentage with precipitation rate larger than 0.1 mm h−1, larger than 1 mm h−1, and between 1 and 0.1 mm h−1, respectively. The lines show the development and mature phases.

Fig. 7.

Composite analysis for (a) total, (b) heavy, and (c) light rainfall percentage with precipitation rate larger than 0.1 mm h−1, larger than 1 mm h−1, and between 1 and 0.1 mm h−1, respectively. The lines show the development and mature phases.

The relationships between rainfall frequency and clouds are examined to determine which cloud types contribute most to the total, heavy, and light rainfall. Although the correlation coefficient values are not very high, a larger positive correlation exists between the total rainfall frequency and the deep convection (0.52) relative to the coefficients between the light rainfall and all cloud types. The heavy rainfall frequency depicts similar features as the total rainfall but with larger correlation coefficients (0.73 for deep convection and 0.50 for congestus), which agrees with previous studies that the convective component modulated by MJO dominates the heavy rainfall and is the major contributor to the total rainfall (e.g., Jiang et al. 2009). Figure 8 displays this high correlation coefficient (or in phase relationship) between the time series of the heavy rainfall and deep convection and congestus frequency during the Manus MJO composite. The generally stronger rainfall peaks appear during the intense convective phases (the development and mature phases) relative to other phases. It is interesting to note that the contribution of congestus to the heavy rainfall frequency is comparable to that for the deep convection, especially during the development phase.

Fig. 8.

Composite analysis for deep convection (red; number), congestus (green; number), and heavy rainfall frequency (blue, dashed; %; using right axis; precipitation rate >1 mm h−1). The dashed vertical lines show the development and mature phases.

Fig. 8.

Composite analysis for deep convection (red; number), congestus (green; number), and heavy rainfall frequency (blue, dashed; %; using right axis; precipitation rate >1 mm h−1). The dashed vertical lines show the development and mature phases.

The relationship between rainfall and precipitable water vapor is also examined to illustrate rainfall-related features before, during, and after the MJO disturbance at Manus. Figure 9 shows a composite of daily rainfall as a function of precipitable water vapor (from a two-channel microwave radiometer). For this composite, only 12 events are used because of a lack of water vapor data for the 15th event (20070105 in Table 2). Each daily value used here is the mean of the available 60-s gridded data for the 24-h period. The water vapor data are screened for precipitation and gaps less than 12 h are filled by interpolation (e.g., Holloway and Neelin 2010). Following Thayer-Calder and Randall (2009), the data from each MJO event are first smoothed by a 5-day running mean, then the data are composited across MJO events, and finally a 13-day running mean is applied to the composite life cycle (K. Thayer-Calder 2011, personal communication). Consistent with previous studies (e.g., Thayer-Calder and Randall 2009), there is a positive relationship between the precipitation rate and precipitable water vapor associated with the MJO disturbance. The increase of precipitable water vapor and precipitation rate in the development phase (from lag −10 to −4 days) is associated with the build-up of the congestus and midlevel clouds, and in the first part of the mature phase (from lag −4 to 0 days) is associated with the MJO deep convection. In addition, the observed column at Manus is less moist after the passing of the MJO convection center than before, which is similar to that seen in the TOGA COARE composite, but is opposite to the ERA-40/TRMM and climate model analysis (Fig. 11 in Thayer-Calder and Randall 2009). We note that the TOGA COARE composite was constructed from one event, observed at six different sites while the Manus composite is constructed from 12 events observed at a single site. The difference between the ground-based composites from TOGA COARE and Manus and the large-scale composites from the ERA-40/TRMM and the climate model may be related to the spatial scales of the water vapor/rainfall data used in the ground-based datasets compared to those in the other analyses.

Fig. 9.

Daily-averaged rainfall as a function of daily-averaged precipitable water vapor for each day during composite MJO life cycle (from lag −25 to +25 days) based on 12 events at ARM TWP Manus Island site. A 13-day running mean is performed.

Fig. 9.

Daily-averaged rainfall as a function of daily-averaged precipitable water vapor for each day during composite MJO life cycle (from lag −25 to +25 days) based on 12 events at ARM TWP Manus Island site. A 13-day running mean is performed.

d. Bandpass-filtered composite

Many studies of the MJO use 20–100-day bandpass filtering to isolate the intraseasonal signal (e.g., Deng and Wu 2011). The above results are based on unfiltered time series because of the data limitations of the cloud frequency measurements (Fig. 1). To examine the effect of bandpass filtering on the noncloud variables, we apply a 20–100-day bandpass filter on the moisture, zonal wind, and temperature before the composite process. The bandpass-filtered composite (Fig. 10a) shows that the moistening of the boundary layer appears around lag −10 days and the largest positive moisture anomalous peak is around lag 0 days at 6 km. These results are similar to those in Fig. 3b, but the two-phase moisture structure is not as clear. However, if we average the moisture anomalies for the lower troposphere (0–4 km) and the midtroposphere (4–8 km) separately (Fig. 10b), we clearly see that the lower-troposphere moisture peaks earlier (around lag −6 days) than the midtroposphere moisture (around lag 0 days), which corresponds to the two-phase structure in the unfiltered Fig. 3b. The filtered wind anomalies (Fig. 10c) show similar baroclinic wind anomalous structure and the easterly–westerly anomaly sequence in the lower troposphere associated with the MJO disturbance to the unfiltered data in Fig. 3c. The filtered temperature Fig. 10d also presents similar features as the unfiltered Fig. 3d.

Fig. 10.

Similar to Figs. 3b–d, but the 20–100-day bandpass filter has been used. Composite analysis of (a) specific humidity (g kg−1), (c) zonal wind (m s−1), and (d) temperature (°C). The 0–4 km (black) and 4–8 km (green) average lines of composite specific humidity (g kg−1) in (a) are presented in (b).

Fig. 10.

Similar to Figs. 3b–d, but the 20–100-day bandpass filter has been used. Composite analysis of (a) specific humidity (g kg−1), (c) zonal wind (m s−1), and (d) temperature (°C). The 0–4 km (black) and 4–8 km (green) average lines of composite specific humidity (g kg−1) in (a) are presented in (b).

In general, the composite results from the filtered time series are similar to those from the unfiltered data. However, when the bandpass filter is used, some high frequency signals (e.g., synoptic systems) are removed. The importance of coupling between synoptic scales (from hours to days) and intraseasonal scales in understanding the multiscale convection associated with the MJO signal (e.g., Nakazawa 1988) is widely recognized, thus it is useful to also include synoptic scales in analyses of the MJO.

e. Significance of the composite

In the above analysis, we assume that the variation of the cloud fields is associated with the passage of the active MJO over Manus. To examine this assumption, we use a resampling or bootstrap analysis (e.g., Wilks 1995) to test the significance of the composite time height structure centered on the time of peak MJO convection at Manus. In this method, we build up a collection of artificial time–height composites of the same length as our MJO events, and test whether structures of the MJO composite are significantly different from a composite created from random time series.

First, we separate our dataset into individual “events,” each containing 51 contiguous days of data (as in our original MJO event selection in section 2c). Because of the somewhat limited size of the dataset, we allow each case to overlap from 5 to 45 days with a 5-day interval. As only those cases that have at least 40% of cloud frequency data available can be used, we have a total of 303 cases. We then randomly select 13 cases at a time to build a composite. We build 1000 different composites and then compare the time and height autocorrelations between the original MJO composite (Fig. 3a) and these random composites to examine whether the original picture shows a more coherent time–height structure for the multiscale convective system associated with the MJO disturbance than the random composites.

Figure 11a is the time autocorrelation coefficient of the original composite MJO cloud frequency (Fig. 3a) at each height. At lag +1 day, there is high correlation (>0.5) of cloud frequency below 12 km, with much lower correlations above 12 km. As the lag increases, the correlation decreases, but the higher correlations are consistent for the midtropospheric levels (4–8 km). In addition, around 12 km, the coefficients decrease quickly with the increased lag, which is consistent with the discussion in section 3a that the cirrus clouds vary on shorter time scales than some of the other cloud types. Figure 11b shows the percentile of the correlation coefficient in Fig. 11a compared with that for the 1000 random composite pictures. Large values (e.g., 90th percentile) mean that the time autocorrelation of the cloud frequency in the original MJO composite is stronger than that in most (e.g., 90%) of the 1000 random composites. The highest percentile is centered between lag +1 and +8 days for heights less than 12 km, which suggests that the structure of the cloud system below 12 km, based on the precipitation and MJO index is a robust feature when compared with the 1000 random composites. Similarly, the percentile value of the height autocorrelation coefficient compared with the 1000 random composites at each lag days (not shown) describes the cloud vertical variation associated with the passage of the MJO disturbance over Manus is most significant at a lag time from −5 to +10 days with peaks around lag days −5, 0, +5, and +7.

Fig. 11.

Time autocorrelation coefficients (a) at different heights for the cloud frequency in Fig. 3a with different time lags. Percentile value (b) of the autocorrelation coefficient in (a) compared with the other 1000 random autocorrelation coefficients.

Fig. 11.

Time autocorrelation coefficients (a) at different heights for the cloud frequency in Fig. 3a with different time lags. Percentile value (b) of the autocorrelation coefficient in (a) compared with the other 1000 random autocorrelation coefficients.

The time and height autocorrelation comparisons illustrate that the process of creating the MJO cloud composite based on the precipitation rate and NOAA MJO index does produce a robust cloud signal, relative to similar but randomly created composites.

4. Summary and discussion

In this study, ground-based high temporal (~120 s) and vertical (~30 m) resolution datasets from several years of observations at the Atmospheric Radiation Measurement tropical western Pacific site of Manus Island are used to conduct a Eulerian study on the characteristics of the convective clouds associated with the MJO disturbance. We examine the thermodynamic, kinematic, and cloud fields before, during, and after the passage of the active phase of the MJO over Manus Island. The high-resolution datasets at Manus Island provide insight into the vertical structure of the multiscale convective systems associated with the MJO disturbance at a range of temporal scales. The local daily precipitation rate and NOAA MJO index 4 are used to identify the MJO events for the composite analysis. A summary of the composite MJO event is presented in this section.

a. Summary

The ARM observations at Manus Island indicate that the cloud characteristics associated with the passage of the MJO over a fixed location has a two-phase structure consisting of a development phase and a mature phase. During the development phase (from lag −10 to −4 days), the observations resemble the classic MJO structure (type I), suggesting the convection is located between the surface westerly and easterly anomalies. The related congestus peak before the MJO deep convection is associated with a positive moisture anomaly in the lower and middle troposphere. The midtroposphere moisture anomaly, along with a high frequency of midlevel nonconvective clouds (e.g., detrained altocumulus/stratus), remains until the deep convection peak. During the mature phase (from lag −4 to 0 days), the deep convection is the dominant cloud type and the composite around lag 0 days is consistent with the type-II MJO-related structure, with the surface westerly anomalies’ peak collocated with the convection. After the mature phase, the midtroposphere remains moist for several days. The frequency of congestus and deep convection decrease rapidly after the mature phase while altocumulus/altostratus and cirrostratus frequencies decrease more slowly.

Consistent with the two-phase structure for the cloud characteristics, the heavy rainfall frequency also shows a two-peak structure before and during the active phase of the MJO. The congestus plays an important role during the first heavy rainfall peak, while the deep convection dominates the second peak. The light rainfall does not show a clear relation to the MJO disturbance over Manus Island. A positive relationship between the precipitation rate and precipitable water vapor exists at Manus, and the atmospheric column is less moist after the passing of the MJO convection center than before, consistent with composite results from multiple observational sites during TOGA COARE.

b. Discussion

Although we show the gradual moistening of the atmosphere (e.g., Figs. 3b and 10a) before the MJO-associated deep convection that appears in previous studies, we do not see a clearly coupled gradual variation from shallow to deep convection (e.g., Fig. 3a) at Manus. Instead, we see frequent shallow cloudiness (e.g., Fig. 6), driven by time scales shorter than the MJO. Between 8 and 4 days before the MJO convective peak at Manus, deeper convection develops, but, likely inhibited by the dry midtroposphere, it does not extend beyond 8 km. In our results, the transition from shallow to deep convection is not a gradual process similar to the moistening process but appears to be a relatively fast transition. We note that (unlike the moisture field) the cloud frequency field is a total field rather than an anomaly field; however, we do not believe this significantly impacts the sharpness of the transition. It is possible that this fast transition is related to the multiscale features of convective processes associated with the MJO, including coupling between the large-scale flow and synoptic-scale disturbances. We also suggest that some of the features of this process in previous studies may have been masked by the use of pentad rather than daily data or by averaging over larger areas.

On the other hand, in Fig. 3b of Kikuchi and Takayabu (2004), the moist layer is limited to below the melting level during the development stage and then suddenly deepens up to 10 km during the mature stage. A similar feature can be seen in our Fig. 10a, which is the bandpass-filtered moisture anomaly. However, in our Fig. 3b, which includes high-frequency variability, the moisture vertical structure is similar during both the development and mature phases. A possible explanation for this difference is the impact of high-frequency signals within the MJO envelope, or synoptic convective activity embedded in the MJO low-frequency variability, on the upper-tropospheric moisture. To further understand the role of synoptic upscaling on the MJO, it is useful to include smaller (e.g., synoptic) scales in future analyses of the MJO with available high-resolution datasets.

These results are based on a composite MJO event from 7 years of ARM TWP Manus Island observational datasets. Further analysis is needed to check the generality of the cloud characteristics at different locations. A similar set of radar observations is available from October 2011 to February 2012 from a site on Addu Atoll in the Indian Ocean (0°37′S, 73°6′E) as part of the ARM MJO Investigation Experiment (AMIE) and Dynamics of the MJO (DYNAMO) field campaigns, and it will be interesting to see whether the same structure and cloud characteristics are seen during the initiation phase of MJO in the Indian Ocean as are seen in the more mature phase observed at Manus Island. Additionally, the large-scale sounding array associated with that campaign should allow investigation of the question whether the midtropospheric moisture anomaly develops locally or by advection.

Acknowledgments

We thank Dr. Jennifer Comstock for producing the radar/lidar datasets used in this analysis and Dr. Jason Hou for valuable suggestions regarding the composite analysis. Comments on the manuscript by Dr. William Gustafson, Dr. Samson Hagos, and two anonymous reviewers are greatly appreciated. Finally, we thank the ARM TWP operations team for their continued efforts to produce high quality data from the ARM TWP sites. This work was supported by the Atmospheric System Research (ASR) Program in the U.S. Department of Energy’s Office of Biological and Environmental Research. Pacific Northwest National Laboratory is operated by Battelle for the U.S. Department of Energy under Contract DE-AC06-76RLO1830.

REFERENCES

REFERENCES
Benedict
,
J. J.
, and
D. A.
Randall
,
2007
:
Observed characteristics of the MJO relative to maximum rainfall
.
J. Atmos. Sci.
,
64
,
2332
2354
.
Bladé
,
I.
, and
D. L.
Hartmann
,
1993
:
Tropical intraseasonal oscillations in a simple nonlinear model
.
J. Atmos. Sci.
,
50
,
2922
2939
.
Chen
,
Y.
, and
A. D.
Del Genio
,
2009
:
Evaluation of tropical cloud regimes in observations and a general circulation model
.
Climate Dyn.
,
32
,
355
369
,
doi:10.1007/s00382-008-0386-6
.
Comstock
,
J. M.
, and
K.
Sassen
,
2001
:
Retrieval of cirrus cloud radiative and backscattering properties using combined lidar and infrared radiometer (LIRAD) measurements
.
J. Atmos. Oceanic Technol.
,
18
,
1658
1673
.
Deng
,
L.
, and
X.
Wu
,
2010
:
Effects of convective processes on GCM simulations of the Madden–Julian oscillation
.
J. Climate
,
23
,
352
377
.
Deng
,
L.
, and
X.
Wu
,
2011
:
Physical mechanisms for the maintenance of GCM-simulated Madden–Julian oscillation over the Indian Ocean and Pacific
.
J. Climate
,
24
,
2469
2482
.
Emanuel
,
K. A.
,
1987
:
An air–sea interaction model of intraseasonal oscillations in the tropics
.
J. Atmos. Sci.
,
44
,
2324
2340
.
Fujiwara
,
M.
, and
Coauthors
,
2009
:
Cirrus observations in the tropical tropopause layer over the western Pacific
.
J. Geophys. Res.
,
114
,
D09304
,
doi:10.1029/2008JD011040
.
Holloway
,
C. E.
, and
J. D.
Neelin
,
2010
:
Temporal relations of column water vapor and tropical precipitation
.
J. Atmos. Sci.
,
67
,
1091
1105
.
Hu
,
Q.
, and
D. A.
Randall
,
1994
:
Low-frequency oscillations in radiative–convective systems
.
J. Atmos. Sci.
,
51
,
1089
1099
.
Hu
,
Q.
, and
D. A.
Randall
,
1995
:
Low-frequency oscillations in radiative–convective systems. Part II: An idealized model
.
J. Atmos. Sci.
,
52
,
478
490
.
Jiang
,
X.
, and
Coauthors
,
2009
:
Vertical heating structures associated with the MJO as characterized by TRMM estimates, ECMWF reanalyses, and forecasts: A case study during 1998/99 winter
.
J. Climate
,
22
,
6001
6020
.
Johnson
,
R. H.
,
T. M.
Rickenbach
,
S. A.
Rutledge
,
P. E.
Ciesielski
, and
W. H.
Schubert
,
1999
:
Trimodal characteristics of tropical convection
.
J. Climate
,
12
,
2397
2418
.
Kikuchi
,
K.
, and
Y. N.
Takayabu
,
2004
:
The development of organized convection associated with the MJO during TOGA COARE IOP: Trimodal characteristics
.
Geophys. Res. Lett.
,
31
, L10101,
doi:10.1029/2004GL019601
.
Kiladis
,
G. N.
,
K. H.
Straub
, and
P. T.
Haertel
,
2005
:
Zonal and vertical structure of the Madden–Julian oscillation
.
J. Atmos. Sci.
,
62
,
2790
2809
.
Knutson
,
R. R.
,
K. M.
Weickmann
, and
J. E.
Kutzbach
,
1986
:
Global-scale intraseasonal oscillations of outgoing longwave radiation and 250 mb zonal wind during Northern Hemisphere summer
.
Mon. Wea. Rev.
,
114
,
605
623
.
Lau
,
K.-M.
, and
L.
Peng
,
1987
:
Origin of low-frequency (intraseasonal) oscillations in the tropical atmosphere. Part I: Basic theory
.
J. Atmos. Sci.
,
44
,
950
972
.
Lau
,
K.-M.
, and
D. E.
Waliser
, Eds.,
2005
: Intraseasonal Variability of the Atmosphere-Ocean Climate System. Springer, 474 pp.
Liebmann
,
B.
, and
C. A.
Smith
,
1996
:
Description of a complete (interpolated) outgoing longwave radiation dataset
.
Bull. Amer. Meteor. Soc.
,
77
,
1275
1277
.
Lin
,
J.
,
M.
Zhang
, and
B.
Mapes
,
2005
:
Zonal momentum budget of the Madden–Julian oscillation: The source and strength of equivalent linear damping
.
J. Atmos. Sci.
,
62
,
2172
2188
.
Luo
,
Z.
, and
W. B.
Rossow
,
2004
:
Characterizing tropical cirrus life cycle, evolution, and interaction with upper-tropospheric water vapor using Lagrangian trajectory analysis of satellite observations
.
J. Climate
,
17
,
4541
4563
.
Madden
,
R. A.
, and
P. R.
Julian
,
1972
:
Description of global-scale circulation cells in the 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.
,
2009
:
The moist static energy budget of a composite tropical intraseasonal oscillation in a climate model
.
J. Climate
,
22
,
711
729
.
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
.
Maloney
,
E. D.
, and
D. L.
Hartmann
,
2001
:
The sensitivity of intraseasonal variability in the NCAR CCM3 to changes in convective parameterization
.
J. Climate
,
14
,
2015
2034
.
Mather
,
J. H.
,
2005
:
Seasonal variability in clouds and radiation at the Manus ARM site
.
J. Climate
,
18
,
2417
2428
.
Nakazawa
,
T.
,
1988
:
Tropical super clusters within intraseasonal variation over the western Pacific
.
J. Meteor. Soc. Japan
,
66
,
823
839
.
Neelin
,
J. D.
,
I. M.
Held
, and
K. H.
Cook
,
1987
:
Evaporation–wind feedback and low-frequency variability in the tropical atmosphere
.
J. Atmos. Sci.
,
44
,
2341
2348
.
Peters
,
O.
, and
J. D.
Neelin
,
2006
:
Critical phenomena in atmospheric precipitation
.
Nat. Phys.
,
2
,
393
396
.
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
.
Sperber
,
K. R.
,
2004
:
Madden–Julian variability in NCAR CAM2.0 and CCSM2.0
.
Climate Dyn.
,
23
(
3–4
),
259
278
.
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
.
Wang
,
B.
,
1988
:
Dynamics of tropical low-frequency waves: An analysis of the moist Kelvin wave
.
J. Atmos. Sci.
,
45
,
2051
2065
.
Wang
,
Y.
,
C. N.
Long
,
J. H.
Mather
, and
X.
Liu
,
2011
:
Convective signals from surface measurements at ARM tropical western Pacific site: Manus
.
Climate Dyn.
,
36
(
3–4
),
431
449
,
doi:10.1007/s00382-009-0736-z
.
Wang
,
Z.
, and
K.
Sassen
,
2002
:
Cirrus cloud microphysical property retrieval using lidar and radar measurements. Part I: Algorithm description and comparison with in situ data
.
J. Appl. Meteor.
,
41
,
218
229
.
Weare
,
B. C.
,
2010
:
Extended Eliassen–Palm fluxes associated with the Madden–Julian oscillation in the stratosphere
.
J. Geophys. Res.
,
115
,
D24103
,
doi:10.1029/2010JD014390
.
Weare
,
B. C.
, and
J. S.
Nasstrom
,
1982
:
Examples of extended empirical orthogonal function analyses
.
Mon. Wea. Rev.
,
110
,
481
485
.
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
,
941
961
.
Wilks
,
D. S.
,
1995
: Statistical Methods in the Atmospheric Sciences. Academic Press, 648 pp.
Wu
,
X.
,
L.
Deng
,
X.
Song
,
G.
Vettoretti
,
W. R.
Peltier
, and
G. J.
Zhang
,
2007
:
Impact of a modified convective scheme on the Madden–Julian oscillation and El Niño–Southern Oscillation in a coupled climate model
.
Geophys. Res. Lett.
,
34
,
L16823
,
doi:10.1029/2007GL030637
.
Zhang
,
C.
,
2005
: Madden–Julian oscillation. Rev. Geophys.,43, RG2003, doi:10.1029/2004RG000158.
Zhang
,
C.
, and
M. J.
McPhaden
,
2000
:
Intraseasonal surface cooling in the equatorial western Pacific
.
J. Climate
,
13
,
2261
2276
.
Zhang
,
C.
, and
S. P.
Anderson
,
2003
:
Sensitivity of intraseasonal perturbations in SST to the structure of the MJO
.
J. Atmos. Sci.
,
60
,
2196
2207
.
Zhang
,
C.
, and
M.
Dong
,
2004
:
Seasonality in the Madden–Julian oscillation
.
J. Climate
,
17
,
3169
3180
.
Zhang
,
C.
, and
Coauthors
,
2010
:
MJO signals in latent heating: Results from TRMM retrievals
.
J. Atmos. Sci.
,
67
,
3488
3508
.