Global distribution of convection penetrating the tropical tropopause

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 110,, doi: /2005jd006063, 2005 Global distribution of convection penetrating the tropical tropopause Chuntao Liu and Edward J. Zipser Department of Meteorology, University of Utah, Salt Lake City, Utah, USA Received 12 April 2005; revised 23 August 2005; accepted 28 September 2005; published 3 December [1] Tropical deep convection with overshooting tops is identified by defining five different reference heights using a 5-year TRMM database. The common properties of these extreme convective systems are examined from a global perspective. It is found that 1.3% of tropical convection systems reach 14 km and 0.1% of them may even penetrate the 380 K potential temperature level. Overshooting convection is more frequent over land than over water, especially over central Africa, Indonesia and South America. The seasonal, diurnal and geodistribution patterns of overshooting deep convection show very little sensitivity to the definition of the reference level. The global distribution of overshooting area, volume and precipitating ice mass shows that central Africa makes a disproportionately large contribution to overshooting convection. A semiannual cycle of total overshooting area, volume and precipitating ice mass is found. Citation: Liu, C., and E. J. Zipser (2005), Global distribution of convection penetrating the tropical tropopause, J. Geophys. Res., 110,, doi: /2005jd Introduction [2] Tropical cumulonimbus clouds have long been recognized not only to be essential to the global energy balance by transporting moist static energy to upper troposphere [Riehl and Malkus, 1958], but also by playing an important role on global circulation and mass exchange between troposphere and stratosphere [Yulaeva et al., 1994; Rosenlof, 1995]. Some recent theoretical studies [Sherwood and Dessler, 2000; Dessler, 2002; Salby et al., 2003] and a modeling study [Jensen and Pfister, 2004] suggest that these clouds may have a major impact on the Tropical Tropopause Layer (TTL) water vapor budget. [3] It is now well known that statistically, tropical cumulonimbus over oceans are generally weak, if strength is equated with vertical velocity [LeMone and Zipser, 1980; Jorgensen et al., 1985; Jorgensen and LeMone, 1989; Lucas et al., 1994; Weietal., 1998]. It is precisely because these storms are weak, and therefore generally safe to penetrate with research aircraft, that direct measurements are abundant in tropical oceanic cloud systems. It is widely accepted that their continental counterparts are stronger, but most direct comparisons are statistical, and are based upon indirect inferences. The global databases with longest period of record are from satellites, and those most commonly used are outgoing visible and long-wave radiation (OLR). Examples are numerous, e.g., Mapes and Houze [1993], Liu et al. [1995], Hall and Vonder Haar [1999], Roca and Ramanathan [2000], Massie et al. [2002], and Gettelman et al. [2002]. However, information on cloud top temperatures, without corresponding knowledge of the vertical structure of deep convection, leaves many questions unanswered. For example, infrared satellite data show that deep convection is more Copyright 2005 by the American Geophysical Union /05/2005JD006063$09.00 frequent over the western tropical Pacific than over continents, but not whether these presumably weaker convective clouds have a larger impact on the tropical tropopause layer than the less frequent but more intense convection over tropical continents. That is the main question addressed by this study. [4] The goals of the Tropical Rainfall Measuring Mission (TRMM) program were outlined by Simpson et al. [1988] and remain largely valid today. The TRMM satellite measurements began in December 1997 and continue as of this writing. The instrument suite is described by Kummerow and Barnes [1998]. These measurements are not only invaluable for global precipitation estimates [Adler et al., 2000; Hou et al., 2001; Schumacher and Houze, 2003; Hirose and Nakamura, 2004], but also have provided a unique opportunity to study the properties of the precipitating cloud systems [Nesbitt et al., 2000; Peterson and Rutledge, 2001; Toracinta et al., 2002; Cecil et al., 2002, 2005; Nesbitt and Zipser, 2003]. Furthermore, the high vertical resolution of the TRMM Precipitation Radar (PR) permits quantification of the heights reached by deep convection in the tropics. Alcala and Dessler [2002] described the distribution of deep convection with radar tops above 14 km over the entire tropics using 4 months of TRMM data. Their study analyzed the overshooting (radar echo above the reference height, i.e., 14 km) properties on the basis of the TRMM PR pixel level measurements. [5] There are a number of motivations for this study. As described above, the TRMM database can quantify the properties of cloud systems and their distributions throughout the global tropics. One of those properties is the intensity of deep convection. The magnitude of overshooting into the stratosphere is obviously related to the magnitude of the updraft velocity near the tropopause, and thus may be a good proxy for convective intensity. If global atmospheric models (and their cumulus parameterizations) 1of12

2 can simulate the geographical distribution of intense convection, they will have met an important challenge. This paper aims at providing such a data set. [6] Our second objective is to use a larger database than has been applied to these issues before to quantify the distribution and properties of convective systems that contain overshooting tops. In this study, we address the following questions: [7] 1. Is overshooting most common over the ocean, especially over the West Pacific? [8] 2. What are the patterns of diurnal and seasonal variation of the overshooting area, volume and ice mass? [9] 3. What are the sizes and the convective intensities of the convection with overshooting tops? How does their intensity compare over land and over ocean? [10] 4. What are the differences between the conclusions of overshooting studies using visible and infrared radiance, and those using PR measurements of radar reflectivity profiles? [11] In the following sections, the size, overshooting area, volume and ice mass, geodistribution, diurnal and seasonal variations of convection with overshooting tops are quantitatively investigated with 5 years of TRMM data using the Utah TRMM precipitation feature database. The convective intensity proxies of overshooting convection are defined, and compared with existing convective system definitions by Nesbitt et al. [2000]. Then, the results from this study and the results using overshooting as defined with visible and infrared radiance are specifically compared. 2. Data and Methodology [12] Since the TRMM satellite was launched in December 1997, more than 7 years of observations have been obtained. However, during 2001, the satellite altitude was increased to lengthen the mission. During the orbit boost, the quality of radar data is questionable. So in this study, we use the data from January 1998 to November 2000 and from December 2001 to December Five years of TRMM precipitation radar (PR) measured reflectivity are grouped into precipitation features (PFs) with adjacent pixels using the method described by Nesbitt et al. [2000]. To minimize noise, the minimum near surface 4 adjacent pixels greater than 20 dbz (about 80 km 2 ) were required for every PF. As the result, more than 5 million PFs were found in 20 N 20 S tropical region. Then the environment sounding for each one of these PFs are obtained by interpolation from , 6 hour interval NCEP reanalysis data described by Kistler et al. [2001]. After recording the general information of PF size (total pixel numbers), mean rain rate, convective rain contribution etc., the convective intensity proxies for each PF, such as minimum 85 GHz Polarization Corrected Temperature (PCT) [Spencer et al., 1989] estimated from TRMM Microwave Imager observed radiance, the maximum heights of PR 20 and 40 dbz, maximum PR reflectivity at 6 and 9 km and flash counts observed by TRMM Lightning Imaging Sensor (LIS) for each PF are analyzed. Additional information on the Utah PF database and its uses is given by Nesbitt et al. [2000], Toracinta et al. [2002], Cecil et al. [2002, 2005], and Nesbitt and Zipser [2003]. [13] To find the PFs with overshooting tops, a definition of reference height is required. In addition to the 14 km used by Alcala and Dessler [2002], four different reference heights are defined and used here: level of NCEP reanalysis tropopause (hereinafter referred to as Z trop ), level of potential temperature q equal to 380 K (hereinafter referred to as Z 380K ) calculated from NCEP sounding, Level of Neutral Buoyancy (LNB) calculated using NCEP sounding and surface equivalent potential temperature q e (hereinafter referred to as LNB sfc ) and Level of Neutral Buoyancy calculated using q e at 925 and 1000 mb, whichever is greater (hereinafter referred to as LNB 925&1000 ). Then the Overshooting PFs (OPFs) are found from 5 million PFs when maximum heights of PR 20 dbz of PFs are above the reference heights respectively. It is important to note that in this study, because only clouds containing large ice particles are likely to attain 20 dbz reflectivity, OPFs defined with PR data are representing the area of large ice particles lifted above the reference heights by strong updrafts in convective cores. Anvil clouds with small ice particles are not detectable by the PR and are not included. This will be discussed further in section 3.8. [14] The overshooting area of each OPF is calculated by multiplying the number of overshooting pixels with more than 20 dbz at the reference height by the pixel size, km 2 before the satellite orbit change and km 2 after the orbit boost. Here the pixel size is the effective pixel size. After the boost, the TRMM PR scan swath width increased about 15%, but scanning rate did not change. Therefore actual pixels overlap with each other mainly in the along track direction and the effective pixel size only increased 14%. The results shown in this study were checked with those before and after the orbit boost, and no significant difference was found. [15] In addition to overshooting area, the overshooting volume and precipitating ice mass for each OPF are calculated as well. The overshooting volume of each OPF is calculated by summing all pixels above the reference heights with reflectivity >20 dbz and multiplying the unit volume of each pixel (4.48 km 3 before the satellite orbit change and 5.09 km 3 after the orbit boost). The overshooting mass is calculated by integrating the precipitating ice mass of pixels above the reference level. The precipitating ice mass of each pixel is estimated using the Z-M ice relation described by Carey and Rutledge [2000] in equation (1). M ice ¼ 1000pðr ice ÞN 3=7 0 ð5: =720 Z ice Þ 4=7 UnitVolume ð1þ where N 0 = m 4, r ice = 917 kgm 3,Z ice is the PR reflectivity and ice mass M ice is in unit (g). [16] Population, mean overshooting distance, and overshooting area of the identified OPFs are listed in Table 1. Of the 5 definitions for OPF, it is not surprising that the mean Z 380K is the highest at 16.8 km. Whereas 1.38% of PFs have a 20 dbz signal reaching 14 km, only 0.1% of PFs were found reaching the level of 380 K. On average, 8% of OPF raining area detected by the PR has overshooting above 14 km from these OPFs, which is only 0.023% of the total PR sample area. When higher reference heights are 2of12

3 Table 1. Population, Overshooting Distance, and Area of Overshooting Precipitation Features (OPFs) Identified With Respect to Five Definitions of Reference Heights Reference Heights 14 km LNB sfc LNB 925&1000 Z trop Z 380K OPFs population, number Population percentage, % Mean reference height, km Mean Z 20dBZ, km Mean overshooting distance, km Mean OPFs overshooting area, km Mean overshooting area/opf raining area, % Total overshooting area/total raining area, % Total overshooting area/total sample area, % used, the PFs tend to have larger overshooting area and smaller raining area indicating that as the definition becomes restricted to the very tallest and strongest storms, they may be at a very early stage of their life cycle, or isolated severe convective storms that may not ever evolve into mesoscale convective systems (MCSs). Comparing with 0.5% overshooting area in the tropics reported by Gettelman et al. [2002], only 0.008% of the total sample area has 20 dbz echo above the NCEP tropopause. The difference can be attributed to the difference between the Figure 1. Location of identified overshooting PFs at 20 N 20 S using different reference heights. The PFs with overshooting distance greater than 2 km are in blue, and those greater than 3 km are in red. 3of12

4 Figure 2. Seasonal variation of number density of OPFs. Number density is defined as total number of PFs > 14 km in each 5 5 bin divided by TRMM 3A25 total pixel numbers to remove sampling bias. Units are parts per thousand. The locations of PFs with more than 40% overshooting area are marked with a cross. The locations of PFs with more than 10 5 km 2 horizontal area are marked with a diamond. height of the 20 dbz surface and the height of the infrared radiometric top of the convective clouds, or the large cloud area of anvils, and will be discussed further in section Results and Discussion [17] With five reference heights, 5 year TRMM observed OPFs were identified in five groups. In this section, the global distribution, seasonal, diurnal variation and the properties of these OPFs will be discussed Global Distribution of Population of OPFs [18] Figure 1 demonstrates the locations of PFs penetrating different reference heights. The similar geolocation distribution pattern exists in all five groups: OPFs were found mainly over the West Pacific, Central Africa, South America, the Intertropical Convergence Zone (ITCZ) and the South Pacific Convergence Zone (SPCZ). The overshooting tops above 17 km (red dots in the first panel of Figure 1) happened mainly over land. About the same number of PFs above 14 km were found over land and ocean. However, more land PFs were capable of penetrating the higher reference heights (Z 380K and Z tropopause ). There is less overshoot activity in the ITCZ over the East Pacific and Atlantic. Two regions were found with more frequent and higher overshoot activities: Congo basin and Panama. Over the extreme West Pacific, more accurately the Maritime Continent region, OPFs were most frequent over land rather than the oceans Seasonal and Diurnal Variation of Population of OPFs [19] In order to show the seasonal variation of penetrative convection, the population of identified OPFs between 20 N and 20 S are counted in 5 5 boxes for different seasons. Then the total numbers of OPFs in each box are normalized with TRMM 3A25 total pixel numbers to remove the sampling bias. After weighting by the number of total OPFs, the number density distribution is obtained. Because the distribution is very similar for PFs identified with different reference heights, we only show the PFs with overshoot above 14 km here in Figure 2. It is clear in Figure 2 that the OPF population is driven by the large-scale seasonal migration of the ITCZ, and the Indian, South American and African monsoons. The highest population of overshooting happens over the Congo basin, Amazon and Indonesia region during all seasons, over Panama in northern summer and over Darwin in Southern summer. [20] Consistent with the deep convection diurnal cycle analysis by Alcala and Dessler [2002], there is a strong diurnal cycle of OPF over land, maximizing in the afternoon 4of12

5 Figure 3. Diurnal variation of population of 20 N 20 S OPFs identified with five reference heights over land (solid) and ocean (shaded). (Figure 3). Similar afternoon maxima are found over South America and Africa independently. Over ocean, the diurnal cycle is weak and more nocturnal OPFs were found. The overshoot population has a similar diurnal cycle pattern as that of tropical convection described in the literature [Soden, 2000; Nesbitt and Zipser, 2003]. Even with big variations in the populations (Table 1), the same diurnal cycle pattern is found for the OPFs identified with different reference heights Size and Overshooting Area Percentage of OPFs [21] Since overshooting convection may play an important role on the troposphere-stratosphere mass exchange, information of overshooting area and size of OPFs is tabulated. The joint histogram of overshooting area and horizontal size is generated in Figure 4. From Table 1, only 0.19% of the cloudy area is found above the tropical tropopause. However, this does not mean that every convective system has a small overshooting area. As shown in Figure 4, there are PFs with overshooting area above the tropopause up to 5000 km 2. Most land OPFs penetrating Z trop have horizontal size about 1000 km 2 and 20% overshooting area. Most oceanic OPFs with tops higher than Z trop have a relatively larger size and smaller overshooting area than those over land. This leads to the larger overshooting area percentage for land OPFs. The OPFs with extremely high overshooting area percentage, for example, with >40% (cross marks on Figure 2), were only found over Africa, South America during all seasons and over Darwin during southern summer. On the other hand, there are many oceanic OPFs with huge size (>10 5 km 2, diamonds on Figure 2) found over the west Pacific, which rarely happened over land. A similar pattern was found for OPFs from other reference heights and is not shown here Convective Intensity Proxies of OPFs [22] To have an overshooting top reach higher heights, we expect the convection to be stronger. With larger overshooting distance, OPFs have lower minimum 85 GHz PCT and more flash counts (Figure 5). Most OPFs with 2 km overshooting distance have more than 100 flash counts and less than 150 K minimum 85 GHz PCT. As listed in Table 2, when we use a higher reference height, the population of OPFs decreases, but the intensity of OPFs increase, with higher maximum heights of 20 dbz (Z 20dBZ ) and 40 dbz (Z 40dBZ ), lower minimum PCT at 37 GHz and 85 GHz, and higher flash counts. These proxies also suggest that overshooting convection over land is more intense than over ocean, with higher reflectivity, stronger ice scattering signature in microwave frequencies, and more flashes. For the OPFs reaching 14 km, there are twice the number of OPFs having flashes over land than those having flashes over ocean. Compared to the existing definition of MCSs by using the TRMM PR and TMI measurements [Nesbitt et al., 2000], the OPFs are more intense with smaller size. They Figure 4. Joint histogram of area and overshooting area of OPFs (20 N 20 S) with cloud top above Z trop over land (solid contours) and over ocean (shaded with dash contours). The 10% (dot line), 20% (dash line) and 40% (solid line) area percentage line are overlapped. 5of12

6 Figure 5. Scatterplot of 20 N 20 S OPF maximum overshooting distance above LNB sfc versus minimum 85 GHz Tb (PCT). Plot includes all OPFs in TRMM s PR swath for 5 years. Symbols correspond to the flash counts in the OPFs, according to the color scale on the right. Solid lines are contours of frequency distribution, contours are 10, 100 OPFs for the bin sizes with 5 K (Tb) and 0.1 km (overshooting distance). are more likely to be storm systems at an early stage with strong convective cores. As the updraft weakens and large particles descend, the apparent strength of these OPFs (as indicated by the TRMM proxy variables) would decrease, the size of these OPFs would often increase, and eventually some of them would develop into MCSs covering large areas with convective intensity varying case by case Global Distribution of Tropical Overshooting Area, Volume and Precipitating Ice Mass [23] To show the global distribution of overshooting area above each reference height, the overshooting area from each OPF is accumulated in 5 5 boxes. The count of TRMM 3A25 total PR pixels is used to remove the sampling bias. After weighting by the total overshooting area, the global distribution of overshooting is obtained and shown in Figure 6. The global distribution of overshooting area from OPFs with different reference heights shows a very similar pattern. The overshooting is not randomly distributed, but concentrates in specific regions, including the Congo Basin, Panama, Colombia, Indonesia, Malaysia and Northern Australia. Congo Basin and Panama are the most active regions. The global distribution of tropical overshooting volume and precipitating ice mass are generated using similar methods. The distributions of volume and precipitating ice mass demonstrate an almost identical pattern as that of Figure 6 (not shown). [24] In order to demonstrate the contribution of overshooting from different parts of the tropics, the latitude belt from 20 N to 20 S is divided into eight equal area regions as shown in Figure 6e. The total contributions of OPF numbers and overshooting area, volume and precipitating ice mass over each region, as well as over land and ocean, are calculated and listed in Table 3. Between 20 N and 20 S, only 23% of the area is land, 77% ocean. However, more than half of OPFs above 14 km and their overshooting area, volume and precipitating ice mass were found over land. This land domination is stronger when higher reference heights are used. About 73% of the overshooting area and volume above the 380 K level are over land. The contribution of overshooting area, volume and precipitating ice mass from 0 to 45 E (including Africa) is the largest among the eight regions. The contribution from Africa alone is greater than the total contributions from Central and East Pacific ( W), Atlantic (45 W 0 ) and Indian (45 90 E) Oceans combined. South America and the West Pacific also contribute significant overshooting areas, although the contribution from the west Pacific is mainly from the land areas of the Maritime continent (Figure 6) Seasonal Variation of Tropical Overshooting Area, Volume and Precipitating Ice Mass [25] The global distributions of tropical overshooting area at different seasons are analyzed using the same method and Table 2. Convective Intensity Proxies of OPFs, MCSs and PFs With Flashes Over Land and Ocean Population, number Area, Z 20dBZ, km 2 km Z 40dBZ, km PCT 85, K PCT 37, K OPFs With Flashes, % Ocean OPFs 14 km LNB sfc LNB 925& Z trop Z 380K MCSs PFs with flashes Land OPFs 14 km LNB sfc LNB 925& Z trop Z 380K MCSs PFs with flashes Flashes, number 6of12

7 Figure 6. Global distribution of overshooting area contribution from OPFs above different reference heights: (a) 14 km, (b) Z 380,(c)Z trop, (d) LNB 925&1000, and (e) LNB surface. The overshooting area in each 5 5 bin has been divided by TRMM 3A25 total pixel number to remove the sampling bias. Units are parts per thousand. shown in Figure 7. Consistent with the OPF population density distribution in Figure 2, overshooting area over Africa has a large contribution in all seasons. There are large contributions from South America in southern spring and summer, the local Panama region in northern summer, and the local Darwin region in southern summer. These results indicate that the Darwin area s much-studied thunderstorms [e.g., Keenan et al., 2000] include some more extreme storms than the Maritime Continent as a whole. To demonstrate the seasonal variation of total overshooting area, volume and precipitating ice mass, these overshooting properties are integrated with 2 latitude intervals for each month. After removing the sampling bias with TRMM 3A25 total pixels, the percentage of overshooting contribution at each 2 latitude region in each month is calculated by weighting the total amount of overshooting area, volume and precipitating ice mass respectively. As shown in Figure 8, there are two main overshooting seasons in Northern tropics, May, and August October. These two overshooting seasons are mainly driven by Africa (Figure 7). Note that there is a semiannual cycle of overshooting area, volume and precipitating ice mass (Figure 8). A similar pattern is found for OPFs defined with other reference heights (not shown) Comparison to Overshooting Defined by IR Temperature [26] As shown in section 2, the overshooting area from OPFs defined with the PR reflectivity is much smaller than the overshooting area defined with the IR temperature colder than tropopause reported by Gettelman et al. [2002]. They found a large area of cloud top temperature colder than the tropopause over the West Pacific Ocean. However, large OPFs overshooting area was found over land in this study. The difference may have two explanations. First, the 20 dbz top is typically lower than the IR detected cloud top. Therefore some overshooting events detected by cold IR temperature are missed by the OPF 7of12

8 Table 3. Global Distribution of Tropical OPF Population and Overshooting Area, Volume and Precipitating Ice Mass Using Different Reference Heights a Reference Heights W W W 45 W E E E E Land Ocean Number 14 km LNB sfc LNB 925& Z trop Z 380K Area 14 km LNB sfc LNB 925& Z trop Z 380K Volume 14 km LNB sfc LNB 925& Z trop Z 380K Mass 14 km LNB sfc LNB 925& Z trop Z 380K a Distribution is in units %. Figure 7. Seasonal distribution of overshooting area from OPFs above 14 km. The overshooting area in each 5 5 bin has been divided by TRMM 3A25 total pixel number to remove the sampling bias. Units are parts per thousand. 8of12

9 Figure 8. Seasonal and latitudinal distribution of contribution of overshooting area, volume and precipitating ice mass from the OPFs above 14 km (in units %). The contribution contour is generated with 1 month and 2 intervals. definitions. Second, large areas of anvil clouds no longer including any active convection with low IR temperature could have lower reflectivity and may not be included in the OPFs in this study. [27] To investigate how many events might be missed by the OPF definition, the minimum IR temperature (calculated from TRMM VIRS measured radiance at 10.8 mm channel) for each PF in 2003 is compared with the NCEP reanalysis tropopause temperature T trop. 1.4% of PFs are found with minimum IR temperature < T trop. Their locations are shown in Figure 9. In contrast with the land domination of OPFs shown in Figure 1, a large number of PFs with cold IR tops is found over the ocean. To demonstrate the global number density of these PFs, the populations of these PFs as well as the OPFs defined with Z trop in selected regions (Figure 9) are listed in Table 4. The population of cold IR top PFs over ocean is greater than that over land. However, normalizing populations by area suggests that those cold IR top PFs over Africa, South America and Indonesia happen more frequently than those over the West Pacific Ocean, East Pacific Ocean and South Pacific convergence Zone (SPCZ). The total number of PFs with 20 dbz reaching 14 km was found 9of12

10 Figure 9. Locations of PFs with minimum IR temperature < T trop in Also shown are the geolocation of selected regions with frequent cold IR top, including West Pacific Ocean, East Pacific Ocean, SPCZ, Africa, South America and Indonesia (the shaded area in the dash line box). The populations of these PFs in the regions are listed in Table 3. to be 7 times less than the number of PFs with IR temperature < T trop. However, they are 10 times less frequent over ocean but only 4 times less frequent over land. However, they are more convectively intense than the PFs with cold IR top. Over land, convection is stronger, so larger ice particles are transported to higher altitudes and cause the higher reflectivity and lower PCTs. As a result, the infrared tops are closer to the 20 dbz tops over land. Over ocean, the convection is weaker than over land, only small particles are able to be transported to high altitudes. Thus, even though the infrared tops may be very cold, the 20 dbz top could be at a much lower level Implications of Observed Overshooting Properties [28] It must be admitted candidly that some of the specific calculations in this paper have error bars that may be substantial and of unknown magnitude. For example, the NCEP reanalysis data used herein have relatively coarse temporal, horizontal, and vertical resolution, so it is not clear how accurately the tropopause height or level of neutral buoyancy have been estimated. Confidence in the main features of the results comes only after comparison between the five largely independent estimates, and with results of previous studies, demonstrates no significant differences. The higher frequency of intense convection over continents compared with that over oceans, and the relative dominance of Africa versus South America or the Maritime Continent are some of those results. [29] The overshooting area, volume and mass shown in this study only apply directly to precipitating ice particles with 20 dbz echoes above the reference heights in convection. Most likely these particles would fall out of the TTL in a short time [Alcala and Dessler, 2002]. A major question that we cannot answer at this time is the relationship between what we can measure with the TRMM PR, and what we cannot measure. What is the relationship between the precipitating ice mass, and the mass of small ice particles which accompany the larger particles in intense convective overshooting towers? These small ice particles fall slowly and may remain aloft for a long time, and may evaporate at unknown altitudes within the TTL. To relate the overshooting air and ice mass to the troposphere and stratosphere mass exchange, it is necessary to know how much air and ice mass would remain in the TTL. However, the mixing and detrainment of air and ice mass from the overshooting cloud tops to the environment in the TTL is a complex process and still under investigation [e.g., Sherwood and Dessler, 2003]. This limits our ability to relate our results to the troposphere-stratosphere transport problem. [30] However, it may be reasonable to speculate that there is a positive relationship between the TRMM-detected Table 4. Comparison of Populations and Convective Intensity Proxies of PFs With Minimum IR Temperature < T trop and OPFs Defined With Z trop at Selected Regions (Shown in Figure 9) in 2003 WP Ocean EP Ocean SPCZ Africa South America Indonesia Land 20 S 20 N Ocean 20 S 20 N Land Minimum T IR <T trop Population, number Normalized by area, 10 3 #km Z 20dBZ, km Z 40dBZ, km PCT 85, K PCT 37, K Mean flashes, number OPFs Defined With Z trop Population, number Normalized by area, 10 3 #km Z 20dBZ, km Z 40dBZ, km PCT 85, K PCT 37, K Mean flashes, number Population ratio of 12

11 overshooting area and the amount of air and mass of small ice particles injected into TTL by intense convection. If so, the global distribution of overshooting area found in this study may represent the distribution of air colder and dryer than the environment TTL. In that case, the results in this study may lead to a contradiction. Some studies suggest that the deep convection has an important impact on the water vapor budget [Sherwood and Dessler, 2000]. However, the deep convection overshooting area has a semiannual cycle, not the well known annual cycle of water vapor in the low stratosphere [McCormick et al., 1993], suggesting that convective overshooting by itself cannot explain dehydration of air entering the stratosphere. Thus the clarification of the relationship between the mass of large ice particles and the mass of small ice particles in the overshooting tops, and the rate of mixing and detraining of air and ice mass from overshooting tops to the TTL environment demands further study. 4. Summary [31] Five groups of tropical deep convection with overshooting tops are identified with five different reference heights using 5-year TRMM data. The common properties of these extreme convective systems are examined from a global perspective. It is found that 1% of deep convection systems reach 14 km and 0.1% of them may even penetrate the 380 K potential temperature level. Convection with large overshooting distance tends to have more flashes and strong ice scattering signature in the 85 GHz channel. A high population of overshooting deep convection over Africa is found during all the seasons. Stronger overshooting convection is found over land, especially central Africa, South America and the large islands of the Indonesian region. Overshooting area and the relative percentage for convective systems over land tend to be larger than that over ocean. There is a large seasonal variation of overshooting over South America. The largest contribution of overshooting area, volume and precipitating ice mass is found from convection over central Africa. There is a semiannual cycle in the total overshooting area, volume and precipitating ice mass over the tropics as a whole. The seasonal, diurnal and geodistribution patterns show a good consistency among the five groups of overshooting deep convection, differently defined. There are more PFs with IR temperature colder than the tropopause than those above the tropopause height from the TRMM radar, the difference being greater over ocean than over land. [32] Acknowledgments. This work was supported by the NASA TRMM office under grant NAG The data were processed by the TRMM Science Data and Information System (TSDIS). We appreciate the comments of A. Del Genio and two anonymous reviewers that led to significant improvements in the manuscript. References Adler, R. F., G. J. Huffman, D. T. Bolvin, S. Curtis, and E. J. Nelkin (2000), Tropical rainfall distributions determined using TRMM combined with other satellite and rain gauge information, J. Appl. Meteorol., 39, Alcala, C. M., and A. E. Dessler (2002), Observations of deep convection in the tropics using the tropical rainfall measuring (TRMM) precipitation radar, J. Geophys. Res., 107(D24), 4792, doi: / 2002JD Carey, L. D., and S. A. Rutledge (2000), The relationship between precipitation and lightning in tropical island convection: A c-band polarimetric radar study, Mon. Weather Rev., 128, Cecil, D. J., E. J. Zipser, and S. W. Nesbitt (2002), Reflectivity, ice scattering and lightning characteristics of hurricane eyewalls and ranbands. part I: Quantitative description, Mon. Weather Rev., 130, Cecil, D. J., S. J. Goodman, D. J. Boccippio, E. J. Zipser, and S. W. Nesbitt (2005), Three years of TRMM precipitation features. part I: Radar, radiometric and lightning characteristics, Mon. Weather Rev., 133, Dessler, A. E. (2002), The effect of deep, tropical convection on the tropical tropopause layer, J. Geophys. Res., 107(D3), 4033, doi: / 2001JD Gettelman, A., M. L. Salby, and F. Sassi (2002), Distribution and influence of convection in the tropical tropopause region, J. Geophys. Res., 107(D10), 4080, doi: /2001jd Hall, T. J., and T. H. Vonder Haar (1999), The diurnal cycle of west Pacific deep convection and its relation to the special and temporal variations of tropical MCSs, J. Atmos. Sci., 56, Hirose, M., and K. Nakamura (2004), Spatiotemporal variation of the vertical gradient of rainfall rate observed by the TRMM precipitation radar, J. Clim., 17, Hou, A. Y., S. Q. Zhang, A. M. da Silva, W. S. Olson, C. D. Kummerow, and J. Simpson (2001), Improving global analysis and short range forecast using rainfall and moisture observations derived from TRMM and SSM/I passive microwave sensors, Bull. Am. Meteorol. Soc., 82, Jensen, E. J., and L. Pfister (2004), Transport and freeze-drying in the tropical tropopause layer, J. Geophys. Res., 109, D02207, doi: / 2003JD Jorgensen, D. P., and M. A. LeMone (1989), Vertical velocity characteristics of oceanic convection, J. Atmos. Sci., 46, Jorgensen, D. P., E. J. Zipser, and M. A. LeMone (1985), Vertical motion in intense hurricanes, J. Atmos. Sci., 42, Keenan, T., et al. (2000), The maritime continent thunderstorm experiment (MCTEX): Overview and some results, Bull. Am. Meteorol. Soc., 81, Kistler, R., et al. (2001), The NCEP-NCAR 50-year reanalysis: Monthly means CD ROM and documentation, Bull. Am. Meteorol. Soc., 82, Kummerow, C., and W. Barnes (1998), The tropical rainfall measuring mission (TRMM) sensor package, J. Atmos. Oceanic Technol., 15, LeMone, M. A., and E. J. Zipser (1980), Cumulonimbus vertical velocity events in GATE. part I: Diameter, intensity, and mass flux, J. Atmos. Sci., 37, Liu, G., J. A. Curry, and R.-S. Sheu (1995), Classification of clouds over the western equatorial Pacific Ocean using combined infrared and microwave satellite data, J. Geophys. Res., 100, 13,811 13,826. Lucas, C., M. A. LeMone, and E. J. Zipser (1994), Vertical velocity in oceanic convection off tropical Australia, J. Atmos. Sci., 51, Mapes, B., and R. Houze (1993), Cloud clusters and super clusters over the oceanic warm pool, Mon. Weather Rev., 121, Massie, S., A. Gettelman, W. Randel, and D. Baumgardner (2002), Distribution of tropical cirus in relation to convection, J. Geophys. Res., 107(D21), 4591, doi: /2001jd McCormick, M. P., E. W. Chiou, L. R. McMaster, W. P. Chu, J. C. Larsen, D. Rind, and S. Oltmans (1993), Annual variation of water vapor in the stratosphere and upper troposphere observed by the Stratospheric Aerosol and Gas Experiment II, J. Geophys. Res., 98, Nesbitt, S. W., and E. J. Zipser (2003), The diurnal cycle of rainfall and convective intensity according to three years of TRMM measurements, J. Clim., 16, Nesbitt, S. W., E. J. Zipser, and D. J. Cecil (2000), A census of precipitation features in the tropics using TRMM: Radar, ice scattering, and lightning observations, J. Clim., 13, Peterson, W. A., and S. A. Rutledge (2001), Regional variability in tropical convection: Observations from TRMM, J. Clim., 14, Riehl, H., and J. S. Malkus (1958), On the heat balance in the equatorial trough zone, Geophysica, 6, Roca, R., and V. Ramanathan (2000), Scale dependence of monsoonal convective systems over the Indian Ocean, J. Clim., 13, Rosenlof, K. H. (1995), The seasonal cycle of the residual mean meridional circulation in the stratosphere, J. Geophys. Res., 100, Salby,M.,F.Sassi,P.Callaghan,W.Read,andH.Pumphrey(2003), Fluctuations of cloud, humidity, and thermal structure near the tropical tropopause, J. Clim., 16, Schumacher, C., and R. A. Houze Jr. (2003), Stratiform rain in the tropics as seen by the TRMM precipitation radar, J. Clim., 16, Sherwood, S. C., and A. E. Dessler (2000), On the control of stratospheric humidity, Geophys. Res. Lett., 27, of 12

12 Sherwood, S. C., and A. E. Dessler (2003), Convective mixing near the tropical tropopause: Insights from seasonal variations, J. Atmos. Sci., 60, Simpson, J., R. F. Adler, and G. R. North (1988), A proposed tropical rainfall measuring mission (TRMM) satellite, Bull. Am. Meteorol. Soc., 69, Soden, B. J. (2000), The diurnal cycle of convection, clouds, and water vapor in the tropical upper troposphere, Geophys. Res. Lett., 27, Spencer, R. W., H. G. Goodman, and R. E. Hood (1989), Precipitation retrieval over land and ocean with the SSM/I: Identification and characteristics of the scattering signal, J. Atmos. Oceanic Technol., 6, Toracinta, E. R., D. J. Cecil, E. J. Zipser, and S. W. Nesbitt (2002), Radar, passive, microwave, and lightning characteristics of precipitation systems in the tropics, Mon. Weather Rev., 130, Wei, D., A. M. Blyth, and D. J. Raymond (1998), Buoyancy of convective clouds in TOGA COARE, J. Atmos. Sci., 55, Yulaeva, E., J. R. Holton, and J. M. Wallace (1994), On the cause of the annual cycle in the tropical lower stratospheric temperature, J. Atmos. Sci., 51, C. Liu and E. J. Zipser, Department of Meteorology, University of Utah, 135 S 1460 E Rm. 819, Salt Lake City, UT , USA. (liuct@met.utah.edu) 12 of 12