A climatology of tropical congestus using CloudSat


Corresponding author: C. Wall, Department of Atmospheric Sciences, University of Utah, 135 S 1460 E, Rm. 819 (WBB), Salt Lake City, UT 84112, USA. (christy.wall@utah.edu)


[1] Cumulus congestus clouds have long been identified as an important part of the spectrum of convective clouds in the tropics. These clouds—which range in size from growing cumulus to slightly smaller than a cumulonimbus—make important contributions to precipitation and latent heat fluxes in the tropics. Past studies have used numerical simulations and satellite observations to examine these clouds globally, although definitions of congestus vary between studies. In this study, congestus in the tropics are identified using contiguous cloud area with cloud tops between 5 and 8 km from 5 years of CloudSat reflectivity data. Alternatively, congestus clouds are defined by contiguous cloud areas with infrared brightness temperature ranging from 273 to 260 K and radar-detected surface rainfall from 14 years of Tropical Rainfall Measuring Mission (TRMM) data. Due to the resolution of CloudSat, the congestus identified using this method represent groups of congestus clouds rather than individual turrets. The regional, seasonal variations of the population of congestus are presented globally. Congestus clouds are found most frequently over the Amazon. There is a strong diurnal variation of congestus clouds over land with a peak in the early afternoon shown by TRMM. General differences are found between the properties of congestus over land and those over ocean, especially the shapes of groups of congestus over land and ocean. Ocean congestus clusters are more bell shaped, while land congestus clusters tend to have flatter sides and larger area above the freezing level. These differences have important implications in the proper representation of congestus in numerical models.

1 Introduction

[2] Many previous studies have examined the distribution of clouds in the tropics. The most studied cloud types have been shallow cumulus and cumulonimbus, or deep convection. Johnson et al. [1999] reminded us that there are three prominent cloud types in the tropics: shallow cumulus, cumulonimbus, and congestus. Congestus have not seen as much attention, perhaps due to the varied definitions of this cloud type. Congestus are considered to range from tall, skinny building cumulus to large clouds that extend above the freezing level but do not meet cumulonimbus criteria. The American Meteorological Society glossary defines congestus as “A strongly sprouting cumulus species with generally sharp outlines and, sometimes, with a great vertical development; it is characterized by its cauliflower or tower aspect, of large size” [Glickman, 2000]. This definition leaves considerable room for interpretation.

[3] Studies from past field campaigns show the impact of congestus on climate in the tropics. Congestus with tops between 4.5 and 9.5 km produced 57% of the precipitation occurring from convective clouds in the Tropical Ocean Global Atmosphere Coupled Ocean Atmosphere Response Experiment (TOGA COARE), and these same clouds contribute 28% of the total convective rainfall over the West Pacific (WPAC) warm pool [Johnson et al., 1999]. Stephens and Wood [2007] found that the typical mode of convection producing precipitation at the Manus Atmospheric Radiation Measurement (ARM) site is congestus-like convection underlying higher layers of cirrus clouds. Their study analyzes over 825,000 radar profiles in the tropics to find that the biggest change between different regimes of synoptically forced convection was the occurrence of different storm types, rather than cloud and precipitation structure. Storm class E from their study, with precipitation echo top heights (ETHs) between 4 and 6 km, is similar to the definition of congestus that we will use in this study.

[4] Congestus are also thought to play an important role in preconditioning the environment for deeper convection [Johnson et al., 1999], and this moistening may be very important in advancing the Madden–Julian Oscillation (MJO) from developing to mature stage [Kikuchi and Takayabu, 2004]. Waite and Khouider [2010] used a cloud-resolving model to show that detrainment of water vapor from congestus can moisten the lower troposphere. Riley et al. [2011] used CloudSat to classify congestus and did not see dramatic differences in congestus population during different stages of the MJO. Hohenegger and Stevens [2013] found that cumulus congestus transition to deep convection too quickly for congestus clouds to sufficiently moisten the atmosphere, and Takayabu et al. [2010] found that parts of the world, for example, the East Pacific (EPAC) between Hawaii and the Intertropical Convergence Zone (ITCZ), have significant quantities of rainfall from congestus but little rain from deeper convective features. The relationship between population of congestus and preconditioning for deeper convection remains unclear.

[5] Jensen and Del Genio [2006] examined the relationship between the environment and congestus at the ARM Nauru site and found that drying in the midtroposphere is more likely to be responsible for limiting congestus cloud top heights (CTHs) than stability of the freezing layer. Similar results were found by Takayabu et al. [2006] using rawinsonde data from research vessels in the western tropical Pacific. Brown and Zhang [1997] showed that dry air above the boundary layer can inhibit the growth of congestus and suppress deep convection using TOGA COARE soundings. Takemi et al. [2004] found that in the West Pacific, midlevel moisture is the biggest difference in environmental soundings when comparing instances where congestus were able to develop to days on which only shallow cumulus occur. Environment has an important impact on where congestus form, and differences in cloud properties could result from regional variations in humidity and temperature fields.

[6] These studies of congestus are necessarily limited in scope because they rely on data from field experiments or stationary ground-based instrumentation with data from limited locations or time periods. Other studies [Luo et al., 2009; Casey et al., 2011] use CloudSat, a satellite in the A-Train constellation that carries the cloud-profiling radar (CPR)—a nadir-pointing, 94 GHz cloud radar [Stephens et al., 2002]. The CloudSat CPR has a minimum detectable signal of −28 dBZ and is sensitive to cloud droplets and smaller rain drops. Luo et al. [2009] used a combination of CloudSat CPR, MODerate Resolution Imaging Sectroradiometer (MODIS), and European Centre for Medium-Range Weather Forecasts (ECMWF) analyses to determine whether or not congestus observed by a CloudSat overpass will continue to grow vertically (“transient convection”) or will cease growth (“terminal convection”). An analysis of data between 15°S and 15°N from 2007 shows that a significant fraction of tropical congestus will continue to grow. Another congestus study by Casey et al. [2011] used the CloudSat cloud classification product to group convective clouds into cloud features. They described the characteristics of cloud top and echo top height in oceanic congestus in the tropics from 2008. Around 31% of the convective features that they identified met the criteria for congestus clouds used by Luo et al. [2009].

[7] CloudSat allows us to investigate congestus globally. However, the increase in spatial coverage is offset by a lack of temporal resolution. Its sun-synchronous orbit results in two overpasses per day at 0130 and 1330 local time (LT). Because of this orbit, obvious difficulties arise when using CloudSat to examine any convective phenomenon. It is well known that there is a distinct diurnal cycle of convective clouds, which differs over land and ocean [Hall and Vonder Haar, 1999; Yang and Slingo, 2001; Liu and Zipser, 2008]. This diurnal cycle paired with the timing of the afternoon CloudSat overpass questions the representativeness of climatology of clouds generated by CloudSat [Liu et al., 2008a].

[8] The Tropical Rainfall Measuring Mission (TRMM) [Kummerow et al., 1998] satellite is not sun-synchronous and can observe the diurnal cycle of convection. TRMM carries the Visible and Infrared Scanner (VIRS) and precipitation radar (PR). The VIRS provides both visible and infrared images that are used to identify different types of clouds. Infrared images can provide cloud top height, while visible provides cloud optical depth. Infrared images can be used to identify different types of clouds [Rossow and Schiffer, 1991]. In addition to the horizontal cloud area indicated by the VIRS, the PR actively detects the vertical structure of precipitation with a minimum detectable signal of 18 dBZ, corresponding to light rain under these clouds. A combination of observations from these two satellites will be used to provide details about the global distribution and properties of congestus from different perspectives. With TRMM and CloudSat, we can observe the typical locations of congestus as well as cloud properties and how they vary over the globe.

[9] Using more than 4 years of CloudSat and 14 years of TRMM observations, this study addresses the following questions:

  1. [10] Where do congestus occur globally? What contribution do these congestus make to global cloud cover and precipitation?

  2. [11] How does the global occurrence of congestus change seasonally and during day and night?

  3. [12] What are the typical properties (size, height, thickness, and reflectivity profile) of congestus? Do these properties change over different regions, especially over land and ocean?

2 Data and Methodology

2.1 Defining Congestus Using CloudSat

[13] Many definitions of congestus have appeared in previous publications. Congestus are easy to identify visually. Johnson et al. [1999, Figures 7a and 7b] showed some examples of congestus clouds. When using radar data to identify congestus clouds, definitions vary by study. Jensen and Del Genio [2006] stated that congestus should have a cloud base near the lifting condensation level (LCL), heights near the freezing level, a lack of significant ice hydrometeors, and precipitation reaching the ground. A cloud top height restriction selects clouds with tops above the typical freezing level in the tropics (4.75 km), but not high enough to be dominated by ice processes. Jensen and Del Genio [2006, Figure 1] showed a pronounced peak in occurrence of clouds with CTH between 5 and 7 km. Luo et al. [2009] identified congestus in CloudSat data by requiring a CloudSat cloud top height (CTH) between 3 and 9 km, a continuous radar echo from CTH to the ground, 10 dBZ echo top height (ETH) within 2 km of the CloudSat CTH, 0 dBZ ETH within 1 km of the CloudSat CTH, and the Calipso CTH within 1 km of the CloudSat CTH. Casey et al. [2012] utilized a similar definition, requiring CTH between 3 and 9 km, cloud base within 1 km of the surface, and a definition of “cumulus” or “deep convection” in the CloudSat cloud classification product.

[14] In this study, the University of Utah CloudSat Database was used to identify congestus globally. This database spans 5 years from 2006 to 2011 and uses the level 2 geometrical profile product to identify clouds. Cloud features are identified by grouping contiguous pixels with reflectivity of at least −28 dBZ and a cloud mask greater than 20. The properties of each cloud feature, including maximum cloud top height, vertical profile of width, and maximum reflectivity in the cloud, are summarized. Figures 1a and 1b show two examples of congestus clouds over land and ocean, respectively, with contoured CloudSat reflectivity. One difficulty in defining congestus in this way is to separate adjacent congestus clouds adjoined by an area of low reflectivities (e.g., Figure 1c). Histograms of congestus area (see section 3) indicate that this type of feature occurs less frequently than the isolated congestus, as shown in Figures 1a and 1b. Section 3.1 discusses this issue in greater depth.

Figure 1.

(a)–(c) Examples of congestus identified by CloudSat over Africa, the West Pacific, and Arabian Sea. Reflectivity is contoured in color, with the dark black lines outlining −28 dBZ, which is the CloudSat minimum detectable signal. Titles above the figure show the date and time of the overpass. (d) An example of a TRMM VIRS cloud feature over Africa (denoted by “X”), with TB11 contoured in color. The black contour is 273 K. Dashed lines indicate the edges of the TRMM precipitation radar overpass. Yellow dashed lines indicate areas with near surface rainfall above 0.

[15] The first step of this study is to subset the congestus clouds from the whole cloud database. The method used to identify congestus is subjective and was derived through multiple iterations of selecting criteria and examining features. Our congestus are restricted to clouds with maximum cloud tops between 5 and 8 km, following the range of congestus cloud top shown by Jensen and Del Genio [2006], but with an expanded upper limit in order to capture the peak in cloud top occurrence between 7 and 8 km observed by Riley and Mapes [2009]. Cloud echo base is required less than 1.5 km above the terrain height to somewhat ensure surface precipitation. Further, we insist on congestus having a maximum reflectivity of at least −5 dBZ, a thickness of at least 4 km, and a horizontal (along-track) width of less than 30 CloudSat pixels (about 33 km). In order to compare CloudSat to TRMM observations, only congestus between 36°S and 36°N are examined. Some of the congestus selected using this method have overlying cirrus layers. Investigation of a subset of our congestus showed that these overlying layers typically had low reflectivities, and therefore, these cases were not excluded from our congestus population.

[16] It should be noted that the interpretation of the CloudSat echo base is ambiguous because the CPR sees both cloud and rain but cannot easily differentiate between the two. At some reflectivities, inferences can be made as to whether a cloud is precipitating, but for the remainder of this study, we use the term “cloud echo base” rather than “cloud base” in recognition of the problem of separating rain from cloud.

2.2 Defining Congestus Using TRMM

[17] As previously mentioned, CloudSat's sun-synchronous orbit may skew the climatology if there is a strong diurnal variation of clouds. In order to determine the impact of diurnal sampling, we use cloud features from the TRMM cloud and precipitation feature database [Liu and Zipser, 2008b]. In this database, cloud features are identified by grouping contiguous pixels of VIRS 11 µm brightness temperature (TB11) colder than 273 K. Then the size of the cloud, minimum TB11, and rainfall indicated by the TRMM PR inside cloud feature are summarized. To subset congestus, the cloud features with minimum infrared TB11 between 260 and 273 K and area less than 200 km2 are identified as candidate cloud features. These temperatures correspond to approximately 6.8–4.75 km based on climatological tropical soundings. Nonzero surface rainfall is required for a candidate cloud feature to be counted as a congestus cloud. This removes many samples with low reflectivity due to TRMM's minimum detectable signal of 18 dBZ. At the same time, this reinforces the definition as congestus of above average convective intensity, with relatively higher radar echoes. Features occurring in the winter hemispheres from 20° to 36° are omitted from the population in order to exclude the large number of cold winter clouds and cold surfaces over mountains that fit the TB11 criteria. Fourteen years (1998–2011) of TRMM data provide a large enough sample size to have a statistically robust data set for the entire diurnal cycle, so the VIRS congestus features will be used in the following sections to compare to CloudSat congestus to determine the effect of the CloudSat orbit on sampling.

[18] Figure 1d shows a plan view of one large area of cloud with infrared TB between 260 and 273 K and some isolated TRMM VIRS “congestus” clouds. The dark black lines show the boundaries of the features. By looking at the visual band from VIRS (not shown), it can be seen that the center cloud is a field of congestus, similar to what we see in Figure 1c but with a larger horizontal extent adjoined by many small congestus clouds. The resolution of the TRMM satellite is larger than that of the CPR, meaning that VIRS is only able to resolve fields of congestus such as this. Individual congestus are not resolvable.

3 Results

3.1 Global Distribution

[19] In total, 130,023 congestus clouds and 4,708,864 cloud features with infrared TB11 ranging from 260 K to 273 K are separately identified from 5 years (June 2006 to April 2011) of CloudSat and 14 years (1998–2011) of TRMM observations. Figure 2 shows the global distribution of CloudSat and TRMM VIRS congestus features. Figures 2a and 2b compare CloudSat and TRMM congestus, that occur within 30 min of the CloudSat overpass. These two figures are similar in many areas, lending confidence to our TRMM VIRS definition of congestus. Figure 2c also shows the relative frequency of TRMM VIRS congestus, but all times are included.

Figure 2.

The relative frequency (contoured in color) of (a) CloudSat congestus, (b) TRMM VIRS congestus occurring within 30 min of the local CloudSat overpass time, and (c) all TRMM VIRS congestus. Black boxes denote regions of interest.

[20] Including the entire diurnal cycle (Figure 2c) lessens the maxima seen over the Amazon and the West Pacific but enhances the East Pacific Intertropical Convergence Zone (ITCZ) and the northern Atlantic. TRMM VIRS does not capture the maxima in congestus over Africa well. This likely occurs because we insist that congestus must have surface rainfall when defining VIRS congestus. A lower fraction of congestus candidates (clouds with infrared TB11 at 260–273 K) have surface rainfall over Africa (5.1%) than over other regions, such as the Amazon (13.5%). The features with higher bases and no surface rainfall over Africa are excluded. The distribution of congestus over the Pacific and Indian Oceans corroborates the distribution found in Casey et al. [2012]. The difference between Figures 2b and 2c suggests a significant diurnal variation of congestus over the Amazon. The high fraction of congestus clouds over the Amazon in Figure 1a is partially due to the incomplete diurnal sampling of CloudSat. The magnitude of this maximum would be reduced if CloudSat observations were available for all hours.

[21] Five different regions are displayed with black boxes in Figure 2. These regions, the East Pacific, West Pacific (ocean only), Amazon, Africa, and Maritime Continent (land only), will be utilized in section 3 to look at differences in congestus over different regions. A variety of land and ocean regions were chosen, as important differences in convection have been shown to occur over land and ocean [Zipser and LeMone, 1980; Zipser and Lutz, 1994; Robinson et al., 2011].

[22] The chosen width criteria of less than around 33 km necessitate some assumptions about the congestus observed by CloudSat. It is impossible for an individual cumulus congestus to be 33 km in width. Individual congestus consist of multiple updrafts and growing areas as well as detrained cloud matter. Frequently, these clouds occur close together and may be connected near their bases. The CloudSat footprint of 1.1 km means that closely spaced congestus towers would be smeared together and would appear as a single cloud using both CloudSat (as in Figure 1c) and especially TRMM (see Figure 1d).

[23] The congestus that we observe with CloudSat can be conceptualized as a “cloud envelope,” in which individual congestus are closely spaced. Figure 3 shows the changes in global congestus distribution as the maximum width of congestus is reduced from around 33 km to around 11 km. Changing the width of congestus used in our definition does not change the global distribution of congestus, but it does greatly reduce our sample size. Thus, the 33 km maximum width restriction will be used for the remainder of the study.

Figure 3.

The occurrence (contoured in color) of (a) CloudSat congestus using a maximum width restriction of 33 km, (b) CloudSat congestus using a maximum width restriction of 22 km, and (c) CloudSat congestus using a maximum width restriction of 11 km. Readers are cautioned that the color scales are different for each panel.

3.2 Contribution of Congestus

[24] One of the most important questions regarding congestus clouds is how much they contribute to the cloud population. Johnson et al. [1999] found that while congestus did not produce as much rain as cumulonimbus, the rain from congestus still accounts for a large amount of the precipitation falling in the West Pacific warm pool. The contribution of our congestus to the total CloudSat cloud population is shown in Figure 3 and summarized for individual regions in Table 1. Figure 4a shows the percentage of all clouds that meet our definition of congestus, while Figure 4b shows the percentage of clouds with tops less than 8 km that are considered to be congestus in this study. The maximum over the Amazon is immediately apparent. In this region, a greater percentage of the cloud population is made up of congestus, and these congestus make up 14.5% of the cloud population with tops lower than 8 km. Central Africa and the Maritime Continent also have a higher proportion of congestus. Over the ocean ITCZ areas, generally 1%–5% of all clouds are congestus.

Table 1. Regional Contributions of Congestus to Area and Rain Volumea
 Percentage of All CloudSat Clouds That Are Congestus [%]Percentage of CloudSat Clouds With Top <8 km That Are Congestus [%]Mean Ratio of Area of CloudSat Congestus to Area of Tall CloudSat CloudsMean Percentage of Volumetric Rainfall Produced by TRMM VIRS Congestus [%]Percentage of VIRS Features With 263–270 K TB With Rain
  1. aTall clouds are clouds with maximum cloud top above 12 km. The rain fraction is calculated using the precipitation rate retrieved from TRMM radar reflectivity [Iguchi et al., 2000].
All land3.867.150.2550.7705.38
All ocean1.962.780.2621.0811.34
Maritime Continent6.6013.40.2130.99713.3
Figure 4.

Figures showing the percentage (indicated by color) of (a) all clouds observed by CloudSat that are congestus and (b) all clouds with tops less than 7 km observed by CloudSat that are congestus.

[25] Figure 5 shows the contribution of our congestus clouds to global cloud area and amount of rain. Figure 5a shows the ratio of the total coverage (along-track width) of congestus to the total coverage of clouds with maximum tops above 12 km. Areas with a small sample size of congestus are not shown. It is important to note that CloudSat only sees a swath through the cloud, so if the cloud was elliptically shaped and the satellite passed across the minor axis, the width may not reflect the true size of the cloud. In most cases, tall clouds should have a larger horizontal extent than the shorter congestus, which are restricted to smaller sizes by definition. Generally, areas dominated by warm rainfall [Schumacher and Houze, 2003; Liu and Zipser, 2009] and regions with large-scale subsidence, such as over the Southeast Pacific and South Atlantic oceans, have a larger ratio of congestus to deeper convection. Of areas with sizable populations of congestus, the Amazon region has a greater ratio of congestus to taller convection. Table 1 lists the values for each region specifically.

Figure 5.

Contributions of congestus to (a) tall cloud (cloud top higher than 12 km) area from CloudSat, (b) rain contribution from congestus identified using TRMM VIRS, and (c) the percentage of VIRS features with TB11 between 263 K and 270 K with measurable near-surface rainfall from the TRMM PR. Readers are cautioned that Figure 5a shows ratios, while Figures 5b and 5c show percentages.

[26] To evaluate the rainfall contribution from congestus, rainfall in each TRMM VIRS cloud feature is determined from TRMM precipitation radar reflectivity [Iguchi et al., 2000]. The fraction of near-surface rainfall from TRMM VIRS congestus is calculated and shown in Figure 5b. In general, groups of congestus produce a larger percentage of total rainfall over the oceans. These percentages of rainfall are far lower than those observed by Johnson et al. [1999] in TOGA COARE, but our precipitation data come from the TRMM PR, which is known to underestimate weak rainfall due to the radar's sensitivity and 5 km footprint [Berg et al., 2010; Lebsock and L'Ecuyer, 2011]. Figure 5b excludes small, individual congestus that cannot be observed by the PR and congestus with very light rainfall rates below the PR's minimum detectable signal. Over land, congestus produce a much smaller percentage of total rain, with the exception again being the Amazon region, where values approach those over the ocean. Values for individual regions can be seen in Table 1.

[27] Requiring surface rainfall in our definition of VIRS congestus likely leaves some congestus clouds out of our sample. Figure 5c shows the percentage of VIRS features with infrared TB11 between 263 and 270 K that have measurable surface rainfall. Ocean areas clearly have a higher fraction of congestus with rainfall, while Africa has a much smaller percentage of these cloud features with rainfall. This result is consistent with Casey et al. [2007], which showed that midlevel clouds in Africa are less likely to be raining than midlevel clouds over the Amazon. This is a reasonable explanation for the lack of VIRS congestus in Africa and likely results from a combination of higher cloud bases and more inefficient warm rain production. The Amazon region has a considerably higher percentage of VIRS features with precipitation.

[28] The Amazon region has the largest percentage of rain produced by congestus, as well as the greatest overall percentage of congestus compared to other areas. Congestus clearly make up a more important part of the spectrum of convective clouds in the Amazon. The West Pacific region and Africa are next in terms of the ratio of the area of congestus to tall clouds, as well as the percentage of low clouds that are congestus. The contributions of congestus in these regions are greater than those in other regions. Comparing the East Pacific to West Pacific also yields interesting results. The West Pacific has a considerably greater percentage of all clouds that are congestus, but the mean percentage of rainfall produced by congestus is similar in both regions. This suggests that the East Pacific has a larger population of nonraining or weakly raining clouds. Indeed, the East Pacific has a smaller percentage of VIRS features with TB11 between 263 K and 270 K producing rain (13.3%) than the West Pacific (16.6%).

3.3 Seasonal and Diurnal Distributions of Congestus

[29] Figure 6 shows the seasonal distribution of CloudSat congestus. Again, the maximum over the Amazon stands out. This region consistently has many more congestus than anywhere else in the world. The explanation for this is beyond the scope of the current study and will be explored in a subsequent paper. Globally, seasonal changes in the congestus population are more visible over the Amazon region and the West Pacific region. These regions along with Africa tend to have more congestus in the dry season (SON).

Figure 6.

Seasonal cycle of global distribution of population of CloudSat congestus. Colored contours show the number of features in (a) December–February, (b) March–May, (c) June–August, and (d) September–November.

[30] Figure 7a shows the diurnal cycle of VIRS congestus features over the selected regions. The black dashed lines denote the times of the CloudSat overpass. Significant differences are seen in the diurnal cycle of ocean and land congestus. Ocean congestus have a slight peak in occurrence at 0300 local time with a very small diurnal change. The diurnal cycle of land congestus has a much greater amplitude and peaks near 1330 local time, which coincides with the CloudSat overpass. Notice that the Amazon and the Maritime Continent regions peak slightly earlier, near 1245 local time, while Africa peaks later, between 1330 and 1500 local time.

Figure 7.

(a) Diurnal variation of population of TRMM VIRS congestus for different regions. The black dashed lines indicate the times of CloudSat overpass. (b) The difference in number of CloudSat congestus from day (1330 LT) to night (0130 LT), contoured in color. Red indicates more daytime congestus, while blue indicates more nighttime congestus. The peak over the Amazon is labeled with the peak occurrence.

[31] The different peaks in the diurnal cycle of occurrence match the results seen by Liu and Zipser [2008]. In their study, ocean precipitation features from the TRMM precipitation radar (PR) peak between 0200 and 0300. Likewise, all land features peak between 1400 and 1500, but the Congo region features peak in occurrence between 1500 and 1600. The times of congestus occurrence also match results seen in Liu and Zipser [2009], where rainfall from warm TRMM features peaks shortly after 1200 local time for land and between 0200 and 0300 local time over ocean.

[32] While CloudSat cannot see the entire diurnal cycle, it captures the afternoon maximum in congestus occurrence. The difference between the number of CloudSat congestus occurring during the day (1330 LT) and the number occurring at night (0130 LT) is shown in Figure 7b. Over land, more congestus occur during the day. Again, the peak over the Amazon is strong—over twice the number of congestus occur here compared to any other region. Areas to the west of the continents around the ITCZ (west of central America and central Africa) also have more congestus during the afternoon overpass. Many ocean areas have more congestus during the night, but the magnitude of the difference is not as great as that in regions that have more daytime congestus.

3.4 Properties of Congestus

[33] Different regions have different seasonal and diurnal cycles of congestus, but are properties of congestus similar for each region? Using CloudSat reflectivity to group features allows us to examine these cloud clusters on an individual basis, rather than sorting each column of pixels individually. Figure 8 compares four different CloudSat congestus properties over selected regions. Figure 8a shows the histogram of maximum reflectivity. This reflectivity is the highest return within the entire cloud—all vertical levels are included. Ocean congestus have a greater number of higher reflectivities, while the maximum reflectivity distribution of the land congestus tends to be more spread out. Some land congestus do have reflectivities in the 10–20 dBZ range, but values below 5 dBZ are more common over land than over ocean.

Figure 8.

Histograms of (a) maximum reflectivity within a CloudSat congestus, (b) maximum horizontal (along-track) width of CloudSat congestus, (c) CloudSat congestus maximum cloud top, and (d) CloudSat cloud thickness (depth between the maximum cloud top and the lowest cloud bottom).

[34] Figure 8b shows the histogram of maximum horizontal width of the cloud along the CloudSat track (the greatest horizontal extent of the cloud group at any level). Most regions have a peak in the 10–15 km range and have similar distributions. The East Pacific has a completely different distribution from any other region. Congestus groups in the East Pacific are wider overall than those occurring in the other regions. The difference in distributions is remarkable and could be caused by larger numbers of individual congestus being grouped more closely together. Cifelli et al. [2007] examined two small regions in the East Pacific and found that precipitation features in the easternmost regime had slightly larger mean and median sizes. They also found that features in this eastern Pacific regime were deeper than those in the regime farther to the west, but our CloudSat congestus data do not show a significant difference in the distributions of cloud tops among the East Pacific and other oceanic regions (Figure 8c).

[35] The distribution of cloud top heights is similarly shaped for many of the regions. Africa and the all land category are differently shaped, with the peak in the distribution at higher cloud tops. Africa has a peak at 6.25 km. Other land regions have discernible peaks, with the Amazon peaking near 5.5 km, the Maritime Continent near 5.75 km, and all land peaking near 6.0 km. Ocean regions have a slightly different shape to the distribution, with the maximum in occurrence at or just above 5.0 km and decreasing from there. This maximum is not necessarily representative of the population of ocean congestus because tops are required to reach 5 km, but ocean congestus tops in general are lower than land congestus tops. The Amazon region is a land region but has a peak around 0.5 km lower than mean land and nearly 1 km lower than Africa. This supports the results of Liu and Zipser [2009], who found that mean storm heights over the Amazon were around 1 km lower than other regions, and is consistent with the idea of the Amazon as a “green ocean” [Silva Dias et al., 2002].

[36] Cloud thickness is shown in Figure 8d. Cloud thickness is a function of both cloud echo base and cloud top. Again, CloudSat cannot differentiate between cloud droplets and rain drops, so cloud echo base represents the lowest detectable echo, which could be either cloud or rain. Land regions have higher mean echo bases (not shown), which when paired with the restrictions on cloud top height forces land congestus to be thinner than ocean congestus. Africa has the lowest peak in occurrence of cloud thickness, with a peak between 4.0 and 4.25 km. Mean land, Amazon, and the Maritime Continent are close behind, with peaks in thickness of around 4.5 km. The ocean regions (all ocean, EPAC, and WPAC) have thicker clouds, with their peaks occurring between 4.5 and 5.5 km.

3.5 Width of Congestus

[37] While the population of CloudSat congestus shows only small variations in the maximum horizontal extent of the cloud, surprising differences show up in the width of these cloud clusters at different levels. Figure 9a shows median profiles of the horizontal (along-track) width of congestus at different levels, as defined using the −20 dBZ reflectivity level. The East Pacific region again has larger congestus groups than any other region, but only below the freezing level (4.75 km in the tropics). At around 3 km, a reversal occurs in the width of the land and ocean congestus groups observed here. Below this level, the ocean regions have bigger features. Above this level, land features are larger. This reversal is featured most prominently in the top 25% and 10% (Figures 8b and 8c), where the largest land features have a much different shape than the largest ocean features. The difference in the relative occurrence of these profiles is shown in Figure 8d, in which it can be seen that land congestus are generally smaller than ocean congestus below 3–3.5 km, while ocean congestus are generally smaller than land congestus above this level.

Figure 9.

Profiles of (a) median, (b) top 25%, and (c) top 10% of horizontal (along-track) width of CloudSat congestus for different regions. Width is determined using the −20 dBZ reflectivity level. (d) The difference in relative frequency of land and ocean congestus at different altitudes. The median lines for land and ocean are red and blue, respectively.

[38] This difference in mean width of congestus clusters is shown globally in Figure 10. The top panel shows mean cloud width at 5 km, and the bottom panel shows mean cloud width at 2 km. At 5 km, features over land have larger mean widths, particularly in Africa. Ocean features at 5 km tend to have mean widths of 3–6 km, while some of the land widths reach nearly 15 km. At an altitude of 2 km, ocean features are clearly larger than land features in many areas, with ocean congestus reaching mean widths of 15 km, while over Africa, the mean width is between 6 and 10 km. These results point to congestus clusters over land and ocean having very different shapes. This idea will be discussed further in section 4.

Figure 10.

Mean width (defined using the −20 dBZ reflectivity level) at (a) 5 km and (b) 2 km in altitude for CloudSat congestus clusters. Colors indicate mean width in kilometers.

[39] The choice of width restrictions for our congestus means that we are looking at groups of congestus that are too closely space to be individually resolved. How does changing the width restriction affect the profiles of median congestus width over land and ocean? Figure 11 shows that as we restrict our congestus to narrower clouds, the mean widths decrease as expected, but the different shapes of land and ocean congestus remain. Land congestus are wider above 3 km, while ocean congestus are wider below this level.

Figure 11.

Profiles of median congestus width for groups of land and ocean congestus with different width restrictions. (a) Congestus less than 33 km across. (b) Congestus less than 22 km across. (c) Congestus less than 11 km across.

3.6 Profiles of Congestus Reflectivity

[40] The differences in cloud width with height inspire questions as to whether trends can be seen in maximum reflectivity with height. Profiles of maximum reflectivity with height can be seen in Figure 12. The ocean regions clearly have larger reflectivities below the freezing level. Above this level, changes between land and ocean are difficult to discern. Differences between the regions are also greatest for the median maximum reflectivity—once only the top 10% are considered, differences between the regions are smaller. If a congestus cloud was growing vertically, we would expect higher reflectivities closer to the cloud top. The top of the profile for the top 10% of maximum reflectivity may reflect the higher reflectivities seen in these growing congestus. Precipitating congestus should have higher reflectivities lower in the cloud as rain drops grow in size while falling through and accreting cloud droplets. Attenuation of these larger drops and the transition from the Raleigh scattering regime to Mie scattering for larger hydrometeors observed by CloudSat likely causes the decrease in reflectivity toward the surface in these profiles. Figure 12d shows the differences in contoured frequency by altitude figures for land and ocean. Ocean congestus clearly have a greater occurrence of higher maximum reflectivities.

Figure 12.

Profiles of (a) median, (b) top 25%, and (c) top 10% of maximum reflectivity for CloudSat congestus clusters in different regions. (d) The difference in contoured frequency between land and ocean profiles of maximum reflectivity. The land and ocean median lines are red and blue, respectively.

4 Discussion

[41] Congestus occur most frequently in the tropics, with some regions experiencing a larger number of these clouds than others. Some differences in congestus properties, including area, reflectivity profile, and levels of cloud top and bottom, are seen between these regions. The greatest differences can be attributed to disparities between land and ocean congestus.

[42] Figure 13 summarizes the properties of land and ocean congestus. This cartoon can be compared with the land and ocean congestus examples shown in Figures 1a and 1b. The shapes surrounding these cartoon clouds were derived from the mean width of our CloudSat land and ocean congestus clusters at each level (similar to Figure 8a) and represent a “cloud envelope” in which individual congestus are spaced too closely to be resolved. The mean tops and bottoms are indicated. These congestus within the cloud envelope observed by CloudSat not only consist of updrafts and growing cloud but also include detraining cloud material. Some congestus observed in our study may be terminal convection, which is no longer growing vertically. These mean values of cloud base agree with the results of Hahn et al. [2001], who found mean values for land and ocean congestus and cumulonimbus using surface observations. Our land and ocean congestus clusters have nearly the same mean width. The dramatic differences occur in cloud shape, with groups of land congestus being “muffin shaped,” while groups of ocean congestus are more “bell shaped.” Land congestus also have higher bases (which could result from higher lifting condensation levels and/or less rain beneath these congestus) and slightly higher tops overall than ocean congestus. Liu and Zipser [2009, Figure 7] showed cumulative distribution frequencies of warm PFs defined using the TRMM over land and ocean in the tropics. In their figure, land features are smaller than ocean features. Area in this case is defined by raining area, so their results match the differences we have observed with congestus—land congestus are smaller than ocean congestus below the freezing level.

Figure 13.

Theoretical schematic of (a) land and (b) ocean congestus, created from the mean width of CloudSat congestus over land and ocean at each altitude level. The mean cloud top height and cloud base height are indicated. Note that the hydrometeor type and placement are speculatively drawn.

[43] The locations and sizes of hydrometeors in this cartoon are purely speculative. Using only CloudSat, we have no definitive proof of quantities or sizes of water and ice particles, although higher reflectivities in oceanic congestus point to larger hydrometeors in those clouds. The slightly higher cloud tops on land congestus and the larger area suggest that perhaps there could be more ice hydrometeors in the top of land congestus. Rangno and Hobbs [2005] provide a comprehensive profile of the microphysics of ocean congestus in the West Pacific. A similar study of land congestus would be worthwhile. It should be noted that the CloudSat signal becomes attenuated in heavy rain, and this attenuation is observed in some of the CloudSat congestus population. A discussion of attenuation issues with millimeter-wave radars can be found in Stephens and Wood [2007].

[44] It is well known that important distinctions exist between deep convective clouds over land and ocean. Differences in heating profiles, updraft speeds, and number of cloud condensation nuclei (CCN) are all thought to play a role. Oceanic clouds generally occur in cleaner air [Twomey and Wojciechowski, 1969], and fewer CCN could result in a more efficient warm rain process and lead to larger raindrops. This is consistent with higher reflectivity in oceanic congestus at low levels. Less latent heat release and a weaker updraft above the freezing level in ocean congestus would lead to fewer ice particles. The higher cloud base observed over land, especially over Africa, could be playing a role in weaker warm rain processes and higher cloud tops. Lower mean cloud bases for ocean congestus also indicate that they produce more rain than land congestus. This can be seen with the VIRS congestus (Figure 5).

[45] Deep convective clouds over land have higher updraft speeds [Zipser and LeMone, 1980], which could force more condensate above the freezing level. If enough condensate is lofted above 0°C, the release of latent heat could cause additional lift and higher cloud tops [Zipser, 2003]. We have no way of knowing whether our congestus have reached their maximum height or will continue to grow in this manner.

[46] We did not investigate how many of these convective clouds over land and ocean continue to grow into larger, deeper cloud systems. Luo et al. [2009] found that 42% of their congestus over ocean and 36% of congestus over tropical land were transient and likely to continue to grow. The peak in the diurnal cycle of congestus observed in Figure 7 precedes the afternoon peak in deep convection. The possibility that some of these clouds will continue to grow into deeper convection will be examined in a future study.

[47] The presence of dry, warm air near the freezing level has been thought to control the amount of congestus that continue to grow into deeper convection [Malkus and Riehl, 1964; Johnson et al., 1999; Redelsperger et al., 2002]. Many of these ideas regarding land/ocean differences are purely speculative—we have not shown that these principles, most of which have been developed by studying deep convection, can be applied to congestus, but some of these processes could explain the differences between land and ocean congestus. A preliminary look at relative humidity profiles associated with our congestus showed that its effect on cloud parameters such as top and area is more complicated than the presence a dry layer and varies from region to region, so we leave the explanation for congestus occurrence relative to environmental parameters for another paper.

[48] Interactions between congestus and the environment will be critical to the study of the MJO. Waite and Khouider [2010] found that detrainment from congestus clouds can moisten the lower troposphere, while Kemball-Cook and Weare [2001] showed that the periodicity of the MJO may be controlled by the buildup and discharge of moist static energy in the lower atmosphere. The MJO is typically preceded by low-level convergence and moistening at low levels, followed by the development of shallow convection, and a gradual lofting of moisture by congestus [Kiladis et al., 2005]. The extent to which congestus precondition the environment for deeper clouds must be investigated.

[49] The next steps in the study of CloudSat congestus are to determine which clouds occur on days when congestus grow into deeper convective clouds and examine the environments in which this does and does not occur, with the goal of exploring the mechanisms by which these congestus are restricted to lower levels or allowed to grow into deeper clouds.

5 Conclusions

[50] Congestus clouds are an important part of the tropical climate. Despite its Sun-synchronous orbit, CloudSat is a viable tool for examining global congestus, as it is able to resolve both clouds and precipitation. Comparisons with TRMM show that it captures the afternoon maxima in congestus occurrence. CloudSat shows that congestus are found most frequently in the tropics, particularly along the ITCZ and make up anywhere from 0% to 12% of clouds. In places in South America, central Africa, and the Maritime Continent, congestus make up 18% of the population of low clouds. A significant maximum in the occurrence of congestus is observed in the Amazon.

[51] Important differences are observed in properties of congestus clouds in different parts of the world. Most of the regional differences in congestus can, at first glance, be attributed to differences in congestus over land and ocean. Oceanic congestus tend to have higher maximum reflectivities, lower bases, and lower cloud tops than land congestus. Congestus over the East Pacific region are larger than congestus in other (land or ocean) parts of the world. Africa has the largest number of higher congestus tops.

[52] Profiles of congestus width show that groups of land and ocean congestus observed by CloudSat have different shapes. The CloudSat-observed envelopes of ocean congestus are more bell shaped, with wider bases and sides that slant toward a skinny top. Land congestus are more muffin shaped, with sides that slope more slowly to a wider top. Many congestus over land are wider above the freezing level than near the ground. Figure 1a shows an example of a land congestus with a very wide top. The differences in the shapes of these cloud clusters are robust and are clearly seen within the means of land and ocean congestus. Differences in shapes of groups of congestus would be well related to different latent heating profiles and possibly differences in precipitation and radiation balances between congestus over land and ocean. These previously unobserved differences have large implications in the numerical modeling world, where the correct parameterization of tropical convection is critical in global climate models.

[53] How do these congestus fit into the diurnal and seasonal cycles of convection? More work is needed to determine the full impact of these clouds on precipitation, cloud cover, and radiation balances in the tropics. The mean environments of land and ocean congestus, including the presence of stable layers near the freezing level as well as midtropospheric moisture, need to be examined to see if either of these parameters can explain the population of global congestus. Further work on these topics is warranted in the future.


[54] CloudSat data were obtained from the CloudSat Data Processing Center (http://www.cloudsat.cira.colostate.edu/). This work was funded by NASA grant NNX11AG31G under the direction of Ramesh Kakar and NASA grant NNX08AK28G under the direction of Erich Stocker. Thanks also go to Erich Stocker and John Kwiatkowski and the rest of the Precipitation Processing System (PPS) team at NASA Goddard Space Flight Center, Greenbelt, MD, for data processing assistance. The authors would also like to thank Courtney Schumacher, Johnny Luo, and one anonymous reviewer.