Characteristics of deep convection measured by using the A train constellation

Transcription

1 Click Here for Full Article JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115,, doi: /2009jd013000, 2010 Characteristics of deep convection measured by using the A train constellation S. Iwasaki, 1 T. Shibata, 2 J. Nakamoto, 1 H. Okamoto, 3 H. Ishimoto, 4 and H. Kubota 5 Received 12 August 2009; revised 2 October 2009; accepted 28 October 2009; published 27 March [1] We show characteristics of a tropical deep convection observed in an experiment employing the A train constellation, the spaceborne imager Moderate Resolution Imaging Spectroradiometer (MODIS), the sounder Atmospheric Infrared Sounder (AIRS) advanced microwave sounding unit (AMSU), the cloud radar CloudSat, and the lidar Cloud Aerosol Lidar with Orthogonal Polarisation (CALIOP). CloudSat and CALIOP measured a vertical cross section of a deep convection at 1.1 km from its center, where the center is defined as the local minimum of the brightness temperature T B (11 mm) measured by using MODIS. This deep convection should be overshooting since its cloud top height measured by using CALIOP was 840 m higher than that of 380 K potential temperature as estimated by using AIRS AMSU data. The cloud morphology observed by using CALIOP indicates that deep convections raised the isentropic surface in the tropical tropopause layer and that there were downdrafts around the deep convection. The averaged mode radius of ice particles and ice water content (IWC) in the deep convection above 380 K are estimated as 23.0 ± 4.9 mm and 7.2 ± 8.0 mg/m 3, respectively, by the use of CloudSat and CALIOP data. The volume of the deep convection above a height of 380 K and the averaged IWC, of which the particle size is less than 20 mm, are estimated and the deep convection has the potential to hydrate the stratosphere with about t of water vapor. We also show deep convections above a height of 380 K are not rare phenomena over the tropical land and warm water pool. Citation: Iwasaki, S., T. Shibata, J. Nakamoto, H. Okamoto, H. Ishimoto, and H. Kubota (2010), Characteristics of deep convection measured by using the A train constellation, J. Geophys. Res., 115,, doi: /2009jd Introduction [2] Stratospheric water vapor plays an important role in chemical and radiative processes, such as ozone depletion [Kirk Davidoff et al., 1999], stratospheric temperature [Gettelman et al., 2004], and surface climate [Shindell, 2001]. Most of the water vapor in the lower stratosphere is transported from the tropical troposphere to higher latitudes in the stratosphere through the tropical tropopause layer (TTL) [Brewer, 1949]. [3] Overshooting refers to cloud intrusion through the level of neutral buoyancy above a deep convection. There are two hypotheses regarding the effects of overshooting in the stratosphere. The first hypothesis suggests that overshooting dehydrates the stratosphere [Danielsen, 1993; 1 Department of Earth and Ocean Sciences, National Defense Academy, Yokosuka, Japan. 2 Graduate School of Environmental Studies, Nagoya University, Nagoya, Japan. 3 Center for Atmospheric and Oceanic Studies, Tohoku University, Sendai, Japan. 4 Meteorological Research Institute, Tsukuba, Japan. 5 Research Institute for Global Change, Japan Agency for Marine Earth Science and Technology, Yokosuka, Japan. Copyright 2010 by the American Geophysical Union /10/2009JD Sherwood and Dessler, 2000, 2001]. Sherwood and Dessler [2000] demonstrated that air recently arrived in the TTL was dry, and these authors proposed a dehydration hypothesis in the stratosphere. In a following paper, Sherwood and Dessler [2001] explained the mixing drying process by simulation of convection and advection in the TTL. On the other hand, overshooting might rehydrate the stratosphere [Jensen et al., 2007; Corti et al., 2008], as proposed by the second hypothesis. Jensen et al. [2007] implemented a numerical simulation that showed the particle size was sufficiently small, less than 20 mm in radius, to allow a gradual fall and that the resulting overshoot hydrates the stratosphere. Corti et al. [2008] carried out airborne observations and showed that evaporation of ice particles hydrates the stratosphere. [4] One of the most unknown properties of an overshoot is the size of the ice particles in an overshooting cloud. That is, if the particle size is sufficiently large to sediment out rapidly, an overshoot tends to dehydrate the stratosphere by causing a cold air intrusion. If the particle size is sufficiently small for a gradual fall, however, an overshoot tends to hydrate the stratosphere. However, since in situ measurements at the top of a cumulonimbus cloud are too difficult to conduct, only a few studies have measured the microphysics of ice particles near the overshoot. Nielsen et al. [2007] 1of10

2 Table 1. List of Data Analyzed Spacecraft Payload Characteristic Product ID Description Aqua AIRS Infrared sounder AMSU Microwave sounder MODIS Visible infrared imager CloudSat CloudSat 94 GHz cloud radar AIRX2RET, level 2, version 5 AIRX2RET, level 2, version 5 Level 1B 2B GEOPROF Vertical profile of temperature. Horizontal and vertical resolutions are correspondingly 50 and 2 km. a Vertical profile of temperature. Horizontal and vertical resolutions are correspondingly 50 and 2 km. a Brightness temperature. Horizontal resolution is 1 km. Vertical profile of attenuated radar reflectivity factor. Horizontal and vertical resolution correspondingly km (along cross track), and 240 m. CALIPSO CALIOP 532 nm lidar Level 1B Vertical profile of attenuated backscattering coefficient. Horizontal and vertical resolution correspondingly km (along cross track), and 60 m. b a Resolutions underneath CloudSat and CALIOP in TTL. b Resolutions in TTL. Observation Time Lag 30 s 15 s performed measurements of ice particles in the tropical lower stratosphere by using balloon borne measuring devices and ground based lidar. They estimated the radii and number densities of the ice particles to be in the range of mm and (0.03 1) 10 6 m 3, respectively. However, their sampling point was 200 km away from the deep convection. The ice particles sampled in that study would have sublimated in the dry stratosphere and therefore their measurements may not represent the actual properties of ice particles associated with a typical overshoot. [5] There are also several studies that employ remote sensing for the study of particles associated with overshooting. Alcala and Dessler [2002] analyzed data from the spaceborne precipitation radar, tropical rainfall measuring mission (TRMM), and showed that 5% of deep convections are overshooting. They also estimated the mean radius of the ice particles as 138 mm; for this radius, the ice particles would be sedimented out rapidly, since their terminal velocity was 1 3 m/s. However, the above mentioned estimates of the ice particle size should not represent the typical particle size in an overshooting cloud, since the TRMM cannot measure particles smaller than about 100 mm due to weak Rayleigh scattering because of TRMM s transmitted wavelength of 21.7 mm. Thus, the TRMM is only suitable for the analysis of large particles. [6] To analyze small particles associated with overshooting, Dessler et al. [2006] used the spaceborne lidar, Geoscience Laser Altimeter System (GLAS), and demonstrated that 3% of thick clouds penetrated the TTL. On the basis of data collected by the spaceborne cloud radar, CloudSat, and the spaceborne imager, Moderate Resolution Imaging Spectroradiometer (MODIS), Luo et al. [2008] demonstrated the possibility of studying the life stages of an overshoot. Savtchenko [2009] statistically studied the relation between deep convections and water vapor at 300 hpa by use of A train data and showed that upper troposphere is humidified by deep convections. [7] The aim of this study is to determine if water vapor in a deep convection has the potential to hydrate the stratosphere. In this study, we use spaceborne multisensors installed on the A train constellation [Stephens et al., 2002]. The A train constellation consists of five satellites on nearly the same orbit. We analyze data obtained from three of the satellites: Aqua, CloudSat, and Cloud Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO). The time lag among the three satellites is within 1 min. [8] To determine the size of the cloud particles associated with overshooting, we apply the radar/lidar method [Okamoto et al., 2003] to analyze the measurements obtained by CloudSat and Cloud Aerosol Lidar with Orthogonal Polarisation (CALIOP). Since the transmission wavelength of CloudSat is 3.19 mm, it has the ability to measure cloud particles smaller than those measured by the TRMM. The CALIOP transmits 532 and 1064 nm laser pulses, and we use the 532 nm data to determine the cloud particle sizes. It should be noted that the radar/lidar method is the best algorithm for the study of cloud microphysics when employing a remote sensing approach [Heymsfield et al., 2008]. [9] Section 2 shows observational results of a deep convection and section 3 estimates the amount of water vapor in the deep convection above a height of 380 K potential temperature. Section 4 shows the estimation of number of deep convections in 1 year. We summarize the results in section Observation [10] We analyze data obtained by using five instruments: Atmospheric Infrared Sounder (AIRS), advanced microwave 2of10

3 CloudSat, dbz, below 27 dbz, respectively, are assumed to be noise. [12] Since the orbit of the A train satellites is sun synchronous and because they measure the same place every 16 days, we cannot measure the temporal transition of deep convections. Figure 1. The brightness temperature at a wavelength of 11 mm, T B (11 mm), measured with the MTSAT. We averaged from 7.5 S to 7.5 N and for 1 day. The cross denotes the position of the deep convection we analyze in this paper. sounding unit (AMSU), and MODIS, installed on Aqua; CloudSat, installed on CloudSat; and CALIOP, installed on CALIPSO (Table 1). All of these satellites measure almost simultaneously. Although on descend, as is the case here, footprints of CloudSat and CALIOP cross the equator about 200 km west from Aqua s ground track, collocation with MODIS, AIRS, and AMSU pixels is still possible because the latter three instruments have scanning functions. Although the Microwave Limb Sounder, MLS on Aura, part of A train, can measure water vapor in the TTL, its footprint was not collocated with CloudSat and CALIOP [e.g., Savtchenko et al., 2008] for the case in this study (2007); hence, we do not analyze MLS data for this particular case. In addition, since AIRS AMSU cannot retrieve water vapor in the TTL, we cannot directly estimate the amount of stratospheric water vapor in this study. [11] We analyze those data from AIRS and AMSU whose flags are of highest quality. The attenuated backscattering coefficient measured with CALIOP, b, below /m/str and the attenuated radar reflectivity factor measured by using 2.1. Multifunctional Transport Satellite Measurement [13] Figure 1 shows the time longitude section of the brightness temperature at a wavelength of 11 mm, T B (11 mm), averaged from 7.5 S to 7.5 N as measured by using the Multifunctional Transport Satellite (MTSAT), where MTSAT is the geostationary satellite and is not a member of the A train constellation. In Figure 1, cross denotes the position of the deep convection we analyze in this paper. Figure 1 shows the Madden Julian oscillation (MJO) moved to cross in the middle of January 2007 and the convections at cross were also classified as MJO in the wave numberfrequency spectrum analysis [Wheeler and Kiladis, 1999]. [14] The anomaly of the sea surface temperature in the Niño 3 region was greater than 0.5 C from September 2006 to January 2007, and the anomaly was 0.8 C in January 2007; therefore, January 2007 was in the El Niño condition. [15] Figure 21 of Ropelewski and Halpert [1987] showed that rainfall amounts in the central Pacific increase during El Niño due to eastward movement of convections induced by eastward expansion of warm pool and relocation of the upward location of the Walker circulation [e.g., Holton, 1992]. Therefore, the convections around cross became active by MJO and the Walker circulation affected by the El Niño condition CloudSat and CALIOP Measurements [16] Figure 2 shows the deep convection as measured by using CloudSat and CALIOP. There are two cloud clusters over the sea surface, marked A and B, in Figure 2a. The deep convection we analyze in this paper is the needle like cloud on cloud cluster A (the white downarrow). The deep convection appeared at 4.75 S, E at 1345 UT on 12 January (0145 LT on 13 January) 2007 and its cloud top was 18.2 km high. CloudSat measured a large radar reflectivity factor, greater than 15 dbz, at a height of less than 14 km. There are two layers of thin cirrus clouds whose geometrical depth is less than 400 m at a height of about 17 km (Figure 2b). One of the cirrus clouds extends 150 km southward starting from 4.5 S at a height of 17.3 km, while the other extends 180 km northward starting from 5.1 S at a height of 17.5 km. These cloud layers are known as subvisual cirrus clouds, or TTL clouds, and their characteristics are a small geometrical depth (<1 km), horizontal uniformity Figure 2. CloudSat and CALIOP measurements of the deep convection at 4.75 S, E at 1345 UT on 12 January The height latitude sections of (a) the attenuated radar reflectivity factor Z and (b) the attenuated backscattering coefficient b, of which the wavelength is 532 nm. The cloud top of the deep convection is 18.2 km. A and B denote cloud clusters. The deep convection we analyze in this paper is at the white downarrow, on top of A. (c and d) Same as Figures 1a and 1b, respectively, but a height from 10 to 20 km. The yellow lines denote heights of 380 K potential temperature obtained from the AIRS and AMSU data. The red lines denote a height of 380 K analyzed by the ECMWF. (e) The overlap of CloudSat and CALIOP measurements. The colored dots denote overlaps of the CloudSat and CALIOP data: red, blue, and green denote measurements by both CloudSat and CALIOP, only CALIOP, and only CloudSat, respectively. The solid line denotes a height of 380 K retrieved from AIRS and AMSU data. 3of10

4 Figure 2 4of10

5 (extending up to 2700 km), and a high occurrence frequency (>50% over the warm pool) [Winker and Trepte, 1998; Wang et al., 1996]. A TTL cloud is produced from at least four processes of large scale dynamics as mentioned by Fujiwara et al. [2009], who carried out three intensive shipborne observations with a lidar and 3 hourly radiosondes. [17] The two cloud layers should be continuous cloud layers on account of their horizontal homogeneity. However, both of the cloud layers were discrete around the deep convection (Figure 2b). Figure 1 of Jensen et al. [2007] demonstrated there were downdrafts on both sides of the overshoot. This implies that the TTL cloud layers may have been dissipated by the downdrafts of the stratospheric warmer air, and that discrete cloud layers were generated at the deep convection. [18] Figures 2c and 2d show magnified images of Figures 2a and 2b, respectively. The yellow lines denote interpolated heights of 380 K potential temperature obtained from the infrared and microwave sounder, AIRS and AMSU, where we interpolate them along the orbit by using the nearest neighbor interpolation. The height of 380 K is the climatological tropopause discussed by Holton et al. [1995]. The distance between the observation points of CloudSat CALIOP and the AIRS AMSU data is generally km and the farthest one is 57 km at 4.96 S, E. The red lines denote heights of 380 K calculated by the European Centre for Medium Range Weather Forecast (ECMWF) data. The difference of both heights is the largest at the cloud clusters and it is about 300 m. We utilize the potential temperature retrieved by the AIRS AMSU data in this paper because the ECMWF data should be used in large scale meteorological fields. Note that the nearest radiosonde was launched from Tarawa (1.35 N, E), Kiribati, at 1200 LT on 13 January However, the time difference and the distance between Tarawa and the deep convection were approximately 10 h and 900 km, respectively. Therefore, the data obtained by the radiosonde were not applicable for the determination of climatology fields. [19] The two TTL cloud layers are at their highest altitude around the deep convection and descend by about 500 m at their edges (Figure 2d). Since a TTL cloud is usually horizontally uniform and its terminal velocity is very small (about 3 cm/s [Iwasaki et al., 2004, Figure 3]), the variations in the height of the TTL cloud layers indicate an increase in the height of the isentropic surface in the vicinity of the deep convection, as simulated by Grosvenor et al. [2007]. Hence, a TTL cloud near a deep convection can be visualized as an isentropic surface in the first stage of generating gravity waves by a deep convection. [20] Figure 2d shows that the cloud top measured by using CALIOP was 840 m higher than a height of 380 K. Figure 10 of Tobin et al. [2006] shows that the root mean square error of temperature at a height of 100 hpa retrieved by AIRS AMSU is 2 K or less; hence, the induced error of a height of 380 K is about 150 m. The vertical resolution of CALIOP at a height of 100 hpa is 60 m (Table 1). Therefore, the height of 840 m over 380 K is significant; hence, the deep convection should be overshooting. Note that this is not exact proof of overshooting because the vertical distribution of the isentropic surface in an overshooting is complicated, as simulated in Figure 9 of Grosvenor et al. [2007], and the horizontal and vertical resolutions of AIRS AMSU data are not high enough to study the distribution of potential temperature in the overshoot. [21] Jensen et al. [2007] simulated overshooting and estimated that the lifetime of an overshoot was about 20 min. With this estimated lifetime and since the horizontal extension of the TTL clouds is 120 km or more from the deep convection, the two TTL cloud layers could not have been generated by the deep convection. This is because the extending speeds of the TTL clouds from the deep convection, which were 100 m/s ± = 120 km/20 min) or more, would be unrealistic. Therefore, these TTL cloud layers would be generated before the deep convection, and the deep convection penetrated these layers later. If the deep convection were to the top of a supercell, the lifetime would be much longer than that of a cell in a multicell storm [e.g., Atkinson, 1981, Houze, 1993], and therefore the above mentioned scenario would not be appropriate [e.g., Wang, 2003]. However, since a supercell is a rare phenomenon, there is a very low probability that this deep convection was the top of a supercell. [22] Figure 2e shows the overlap of the CALIOP and CloudSat data. Red, blue, and green denote measurements by both CALIOP and CloudSat, only CALIOP, and only CloudSat, respectively. Because the attenuation of the laser beam is significant, data below the deep convection could not be obtained by using CALIOP. On the other hand, CloudSat could not perform measurements for some clouds, for example, two layers of TTL clouds, because the cloud particles would be too small [e.g., of Iwasaki et al., 2004, Figure 2]. Note there are several green dots at cloud tops even though the attenuation of the laser beam is small and CALIOP received significant signals below the dots. There are two possible reasons for this. The first is that cloud particles are too large but the number density is too low. The second is the difference in horizontal resolution between CALIOP and CloudSat (Table 1). That is, if a cloud shape is not horizontally uniform, CloudSat has a chance to detect more clouds than CALIOP does at cloud top because the horizontal resolutions of CloudSat are coarser and CloudSat measures a wider area. However, because of insufficient data, we cannot distinguish which of the two causes are significant MODIS Measurement [23] Figure 3a shows a latitude longitude section of the brightness temperature at a wavelength of 11 mm T B (11 mm), band 31, measured by employing the MODIS, where we averaged T B (11 mm) in 10 km 10 km to be able to compare our results to those of Chen et al. [1996]. Chen et al. [1996] analyzed data of the Japanese Geosynchronous Meteorological Satellite (GMS) whose resolution is 10 km. They classified cloud clusters into four classes on the basis of an area of the precipitating core defined by its T B (11 mm) being less than 208 K [Chen et al., 1996, Figure 5]. Since the area of the precipitating core of the cloud cluster in which the deep convection is km 2, the cloud cluster is categorized as belonging to the smallest class of cloud clusters. Since small cloud clusters usually occur in 5of10

6 Figure 3. MODIS measurements of the deep convection. (a) Latitude longitude section of averaged T B (11 mm) whose spatial resolution is 10 km 10 km. The diagonal line denotes the orbit of CloudSat and CALIOP. (b) Magnification of Figure 1a around the deep convection, and its resolution is 1 km 1 km. Note that the color bars in Figures 1a and 1b are different. A and B are the same cloud clusters measured by using CloudSat (Figure 2a). The red denotes T B (11 mm) is greater than 220 K. The contours are drawn from 180 K to 220 K at 5 K increments. The orbit is divided into three colors by measurements of the deep convection (excluding the TTL clouds). White, gray, and black lines denote the places where CloudSat and CALIOP measured the deep convection, when only CALIOP measured the deep convection, and when neither of them measured deep convections above a height of 380 K, respectively. (c) Same as Figure 1b but destriped T B (6.7 mm) T B (11 mm) averaged in 3 3 pixels. The contours are drawn from 0 K to 5 K at 1 K intervals. tropical regions, deep convections of this scale could be a common phenomenon in the tropics as discussed in section 4. [24] Figure 3b shows the magnification of Figure 3a at a resolution of 1 km. A and B denote the same cloud clusters as shown in Figure 2a. The diameter of A is approximately 29 km and there are four deep convections whose coldest T B (11 mm) is less than about 190 K. The diagonal line denotes the orbit of CloudSat and CALIOP. Hence, these satellites measure one of four deep convections in A. The maturest T B (11 mm) of the deep convection on the orbit measured by using MODIS is K and the T B (11 mm) at the highest cloud top measured by using CALIOP and CloudSat is K, and the distance between both T B (11 mm) is 1.1 km. [25] Schmetz et al. [1997] simulated the difference between the brightness temperatures of the water vapor absorption band ( mm) and the infrared window ( mm) at an overshoot. Figure 4 of Schmetz et al. [1997] showed that the difference is positive when a cloud top is in the TTL. The difference is at a maximum when a deep convection reached the tropopause, and the difference decreases when the cloud top of a deep convection enters the stratosphere. [26] Since the MODIS data have striping noise in T B (6.7 mm), we destriped the data by applying the method of Weinreb et al. [1989]. Figure 3c shows the difference between destriped T B (6.7 mm) and T B (11 mm) measured with MODIS where we averaged the data in 3 3 pixels (approximately 3 3 km) because the noise was not completely removed by the algorithm. [27] The difference in the cloud clusters whose T B (11 mm) is less than 220 K is positive, and this implies the cloud top of the clusters reached into the TTL. The local minimums of T B (11 mm) in the center of cluster A have the local maximums of the difference of temperatures. On the other hand, the local minimum of T B (11 mm) in the east of A does not have the local maximum of the difference of temperatures. Therefore, these characteristics are not easy to explain since the difference of temperatures depends on stratospheric conditions, such as temperature and the amount of water vapor, as discussed by Schmetz et al. [1997]. Therefore, we do not discuss the difference between T B (6.7 mm) and T B (11 mm) in detail here. [28] Luo et al. [2008] combined CloudSat, MODIS, and ECMWF data and classified three life stages of an overshoot: growing, mature, and dissipating stages. For example, when the T B (11 mm) of an overshoot measured by using MODIS is greater than the temperature of the cold point tropopause (CPT) calculated by using the ECMWF data and its cloud top height is higher than that of CPT, they classified it as WH type, which indicates the dissipating stage since air at the cloud top would be mixed with the warm stratospheric air. They also showed statistically that the occurrence frequency of the WH type is the greatest in the three life stages. In our case, since the T B (11 mm) of the deep convection measured by using MODIS (191.1 K) is greater than the CPT temperature estimated by using the 6of10

7 where dn/dr and r denote a size distribution (m 4 ) and a particle radius (m), respectively. We assume that dn/dr is a lognormal distribution (equation (2)) with a standard deviation, s g, of 1.5, i.e., 2 3 dn dr ¼! N pffiffiffiffiffi 2 ln g r exp ln r ln r 2 4 g pffiffi 5; ð2þ 2 ln g where N denotes number density (m 3 ). The relationship of r eff and the mode radius, r g, in a lognormal distribution is h i 2 r g ¼ r eff exp 2:5 ln g : ð3þ Figure 4. Retrieval of (a) r eff and (b) IWC where s g is assumed to be 1.5. ECMWF data (186.6 K) and the cloud top height is higher than CPT height, the deep convection is classified as a WHtype and it would be in the dissipating stage. [30] Though r eff has less physical meaning than does r g, r eff is a useful parameter in remote sensing because r eff is more independent of s g and size distributions than is r g [e.g., Okamoto et al., 2003, Figure 1]. Two parameters, r g and N, are retrieved independently from two independent observation data, lidar data (geometrical optics) and radar data (Rayleigh scattering). Here, we show r eff, r g, and IWC as retrieved results. [31] We estimate the total hydration amount by following Jensen et al. [2007], who broadly evaluated that smaller particles (<20 mm in radius) in an overshoot would have the potential to rehydrate the stratosphere during 200 m sedimentation. [32] We averaged data of the deep convection above a height of 380 K. The retrieved average and standard deviation for r eff, r g, and IWC of the deep convection are 34.8 ± 7.4 mm, 23.0 ± 4.9 mm, and 7.2 ± 8.0 mg/m 3, respectively, where 1 mg/m 3 is approximately 10 ppmv. When s g is 1.7, r eff, r g, and IWC of the overshoot are 30.1 ± 6.9 mm, 14.9 ± 3.2 mm, and 5.4 ± 5.9 mg/m 3, respectively. It should be noted that the CloudSat data was assumed as noise level, 27 dbz, while CALIOP only measures cloud parameters (see blue dots in Figure 2e). Thus, results obtained from data observed by using only CALIOP would be overestimates. For example, although the mode radius of a TTL cloud is 4 30 mm [e.g., Iwasaki et al., 2007; Shibata et al., 2007], the obtained size of the TTL clouds is about mm. The retrieved r eff, r g, and IWC of the deep convection whose radar reflectivity is greater than the noise level (> 27 dbz) are 37.1 ± 4.0 mm, 24.6 ± 2.6 mm, and 15.3 ± 6.9 mg/m 3, respectively, when s g is 1.5. That is, r g and IWC in the deep convection are about 20 mm and 10 mg/m 3 at a rough estimation. [33] We estimate the volume, V, of the deep convection above a height of 380 K, by assuming that the height of the cloud top of the deep convection is represented by equation (4) (Figure 5): 3. Amount of Hydration [29] Figure 4 shows the height latitude section for the retrieved effective radius r eff and the ice water content (IWC) by use of the radar/lidar method, where equation (1) is the definition of r eff : Z r eff ¼ r 3 dn Z dr dr r 2 dn dr dr; ð1þ 7of10 H ¼ H 0 c 1 R c2 ; where H 0 (in km) is a parameter which denotes the height of the apex, and is defined as the coldest T B (11 mm) of the deep convection as measured by using MODIS. Here, R (in km) denotes the horizontal distance from the apex, c 1 and c 2 denote parameters, and H (in km) denotes the height of the cloud top of the deep convection at R. We then calculate H 0, c 1, and c 2 by fitting H with the cloud top height measured by ð4þ

8 IWC of only small particles (<20 mm). Thus, the deep convection has the potential to hydrate the stratosphere about t. Figure 5. A schematic diagram of the deep convection to estimate its volume. We assume that the height of the cloud top of the deep convection is described as equation (4). The apex is assumed at the coldest T B (11 mm) as measured by using MODIS and the bottom is a height of 380 K as estimated by using AIRS and AMSU. CloudSat and CALIOP measured one of the vertical cross sections (blue and red sections). using CALIOP. Here, H 0, c 1, and c 2 are 18.0 km, , and 3.1, respectively. V is calculated as 116 km 3 from equation (5). R at a height of 380 K is 8.7 km. We also estimate the horizontal cross section of CloudSat data at a height of 380 K by the same fitting method and it becomes 98 km 2, where we utilize the value in the next section V ¼ Z above height of 380 K R 2 dh: [34] The averaged IWC, the particle size of which is less than 20 mm in the deep convection above a height of 380 K, is calculated from equation (6) and the value becomes 0.56 mg/m 3 when s g is 1.5: * 20m Z + dn 4 IWC 20m ¼ dr 3 r3 dr 0 above height of 380 K ð5þ ; ð6þ where r denotes a density of water ice, kg/m 3. [35] Therefore, we simply estimate that the deep convection has the potential to hydrate the stratosphere up to 65 t (= 116 km mg/m 3 ) of water vapor. [36] Figure 6 shows the uncertainty of hiwc 20mm i above a height of 380 K induced by standard deviations of the lognormal distribution and assumed noise levels of CloudSat. Figure 6 shows that retrieved hiwc 20mm i would be tens of percents of uncertainty; e.g., the amount of water vapor becomes 76 t and 65 t when the noise levels of CloudSat and s g are assumed to be 40 dbz and 50 dbz and 1.5, respectively. When the noise levels of CloudSat and s g are 27 dbz and 1.7, respectively, it is 106 t. The assumed noise level is not significant for hiwc 20mm i. This is because smaller noise level reduces the number of larger particles and increases the number of smaller particles [e.g., Iwasaki et al., 2004, Figure 2] and because we calculated 4. Number of Deep Convections [37] CloudSat and AIRS AMSU data from September 2006 to August 2007 are analyzed to estimate the occurrence number of deep convections above a height of 380 K. Since CALIOP detects more clouds than CloudSat does in TTL and since shapes of most clouds measured by use of CALIOP are too complicated to carry out the radar/lidar method, we do not use CALIOP data for the statistical analysis. [38] Figure 7a shows the global map of deep convections above a height of 380 K in 1 year. Figure 7a shows that the occurrence frequency of the deep convections is greater over land and warm water pool as in Figure 2 of Alcala and Dessler [2002]. Figure 7b is the same as Figure 7a but the height of the cloud top is 0.8 km higher than the height of 380 K; that is, the same or deeper convections than this case study s. Table 2 shows the summary of the occurrence frequency. [39] We then estimate a number of deep convections above a height of 380 K in 1 year, m, by m hi T 1year h S 380K S Trop i ¼ A 380K A all ; ð7þ where hti and hs 380K i denote averaged lifetime and horizontal cross section of one deep convection whose echo top is higher than a height of 380 K. T 1year and S Trop denote time of 1 year and surface area of the Earth between 20 S and 20 N; that is, s and km 2, respectively. A all denotes area synergistically measured by the use of AIRS AMSU, CALIOP, and CloudSat in 1 year. A 380K is the same as A all but CloudSat detects deep convections above a height of 380 K. Table 2 shows A 380 K km /A all is about 10 times smaller than A 380 K /A all where A 380 K+0.8 km is the same as A 380 K but CloudSat detects deep convections 0.8 km above a height of 380 K; hence, the occurrence Figure 6. Retrieved uncertainty induced by the standard deviations of the lognormal distribution and the noise levels of CloudSat that we assumed. The Y axis denotes the averaged IWC whose particle size is less than 20 mm above a height of 380 K (see equation (6)). 8of10

9 Figure 7. (a) Global map of deep convection events whose cloud top measured by the use of CloudSat is higher than the height of 380 K. Brown and blue dots denote deep convections over land and ocean, respectively. (b) Same as Figure 7a but the height of the cloud top is 0.8 km higher than the height of 380 K. frequency of the deep convection whose cloud top is 0.8 km higher than a height of 380 K is about 10 times lower than those above a height of 380 K. [40] The value m is then estimated as by assuming hti and hs 380K i to be 20 min and km 2 (see section 3), respectively. That is, deep convections above a height of 380 K are not rare phenomena over the tropical land and warm water pool. 5. Summary [41] We carried out a case study of a deep convection in the tropics, 4.75 S, E, measured by using the A train satellites at 1345 UT on 12 January (0145 LT on 13 January) 2007 in order to evaluate the possibility that deep convection hydrates the stratosphere. [42] MODIS data showed that the cloud cluster of the deep convection belonged to the smallest class described by Chen et al. [1996]. There were four deep convections in the cloud cluster, whose diameter was approximately 29 km. The coldest T B (11 mm) in the deep convection was K and the coldest T B (11 mm) at the highest cloud top measured by using CloudSat and CALIOP was K. The distance between both of the coldest points was 1.1 km; CloudSat and CALIOP measured the vertical cross sections of the deep convections. [43] CALIOP data showed that two cirrus cloud layers, TTL clouds, had risen at the deep convection and were discrete around the deep convection. Therefore, a TTL cloud would be a good indicator of the generation of gravity waves and vertical wind near a deep convection in the TTL. The cloud top measured by using CALIOP was 840 m higher than a height of 380 K potential temperature determined by using AIRS and AMSU. Therefore, the deep convection should be an overshoot. [44] We applied the radar/lidar method to the CloudSat and CALIOP data and estimated the particle size and IWC of the deep convection. The averaged mode radius of ice particles and the IWC above a height of 380 K were 23.0 ± 4.9 mm and 7.2 ± 8.0 mg/m 3, respectively, assuming that the particle size distribution had a lognormal distribution and that the standard deviation was 1.5. When s g was 1.7, these values were 14.9 ± 3.2 mm and 5.4 ± 5.9 mg/m 3, respectively. [45] Because the volume of the deep convection above a height of 380 K was roughly estimated as 116 km 3 and because the averaged IWC which consisted of particles smaller than 20 mm in radius was 0.56 mg/m 3, the deep convection has the potential to hydrate the stratosphere by about t where the retrieval error induced by assumptions of the radar/lidar method would be tens of percent. [46] The occurrence number of deep convections above a height of 380 K was also estimated. The value becomes between 20 S and 20 N in 1 year, and the frequency over land is greater than that over ocean. Deep convections above a height of 380 K are not rare phenomena over the tropical land and warm water pool. Table 2. Occurrence Frequency of Deep Convection Total Day a Night a Ocean a Land a A 380 K /A all A 380 K+0.8 km /A all a Distinction of day/night and land/ocean is carried out by use of Day_Night_Flg (two categories; day and night) and Land_Water_Mask (eight categories) of CALIOP data. Data with the flags of shallow, moderate, and deep oceans are classified as Ocean and the others are classified as Land. 9of10

10 [47] Acknowledgments. Data from AIRS AMSU, MODIS, Cloud- Sat, and CALIOP were obtained from the NASA Goddard Earth Sciences Data and Information Services Center (GES DISC), the Level 1 and Atmosphere Archive and Distribution System (LAADS), the CloudSat Data Processing Center, and the Atmospheric Sciences Data Center (ASDC), respectively. The Niño 3 index was obtained from the Japan Meteorological Agency. The MTSAT data were provided by Kikuchi, Kochi University. The MODIS L1B infrared destriping software was provided by the International MODIS/AIRS Processing Package (IMAPP). This work was partly supported by the Ministry of Education, Culture, Sports, Science, and Technology of Japan through a grant in aid for Scientific Research ( ). References Alcala, C. M., and A. E. Dessler (2002), Observations of deep convection in the tropics using the Tropical Rainfall Measurement Mission (TRMM) precipitation radar, J. Geophys. Res., 107(D24), 4792, doi: / 2002JD Atkinson, B. W. (1981), Meso Scale Atmospheric Circulations, 495pp., Academic, London. Brewer, A. (1949), Evidence for a world circulation provided by the measurements of helium and water vapour distribution in the stratosphere, Q. J. R. Meteorol. Soc., 75, , doi: /qj Chen, S. S., R. A. Houze Jr., and B. E. Mapes (1996), Multiscale variability of deep convection in relation to large scale circulation in TOGA COARE, J. Atmos. Sci., 53, , doi: / (1996) 053<1380:MVODCI>2.0.CO;2. Corti, T., et al. (2008), Unprecedented evidence for deep convection hydrating the tropical stratosphere, Geophys. Res. Lett., 35, L10810, doi: /2008gl Danielsen, E. F. (1993), In situ evidence of rapid, vertical, irreversible transport of lower tropospheric air into the lower tropical stratosphere by convective cloud turrets and by large scale upwelling in tropical cyclones, J. Geophys. Res., 98(D5), , doi: /92jd Dessler, A. E., S. P. Palm, and J. D. Spinhirne (2006), Tropical cloud top height distributions revealed by the Ice, Cloud, and Land Elevation Satellite (ICESat)/Geoscience Laser Altimeter System (GLAS), J. Geophys. Res., 111, D12215, doi: /2005jd Fujiwara, M., et al. (2009), Cirrus observations in the tropical tropopause layer over the western Pacific, J. Geophys. Res., 114, D09304, doi: /2008jd Gettelman, A., P. M. de F. Forster, M. Fujiwara, Q. Fu, H. Vömel, L. K. Gohar, C. Johanson, and M. Ammerman (2004), Radiation balance of the tropical tropopause layer, J. Geophys. Res., 109, D07103, doi: /2003jd Grosvenor, D. P., T. W. Choularton, H. Coe, and G. Held (2007), A study of the effect of overshooting deep convection on the water content of the TTL and lower stratosphere from cloud resolving model simulations, Atmos. Chem. Phys., 7, Heymsfield, A., et al. (2008), Testing IWC retrieval methods using radar and ancillary measurements with in situ data, J. Appl. Meteorol. Climatol., 47, , doi: /2007jamc Holton,J.R.(1992),An Introduction to Dynamic Meteorology, 3rded., 511 pp., Academic, San Diego, Calif. Holton, J. R., P. H. Haynes, M. E. Mclntyre, A. R. Douglass, R. B. Rood, and L. Pfister (1995), Stratosphere troposphere exchange, Rev. Geophys., 33, , doi: /95rg Houze, R. A. (1993), Cloud Dynamics, 573 pp., Academic, San Diego, Calif. Iwasaki, S., et al. (2004), Subvisual cirrus cloud observations using a 1064 nm lidar, a 95 GHz cloud radar, and radiosondes in the warm pool region, Geophys. Res. Lett., 31, L09103, doi: /2003gl Iwasaki, S., et al. (2007), Characteristics of aerosol and cloud particle size distributions in the tropical tropopause layer measured with optical particle counter and lidar, Atmos. Chem. Phys., 7, Jensen, E. J., A. S. Ackerman, and J. A. Smith (2007), Can overshooting convection dehydrate the tropical tropopause layer?, J. Geophys. Res., 112, D11209, doi: /2006jd Kirk Davidoff, D. B., E. J. Hintsa, J. G. Anderson, and D. W. Keith (1999), The effect of climate change on ozone depletion through changes in stratospheric water vapour, Nature, 402, , doi: / Luo, Z., G. Liu, and G. L. Stephens (2008), CloudSat adding one new insight into tropical penetrating convection, Geophys. Res. Lett., 35, L19819, doi: /2008gl Nielsen, J. K., N. Larsen, F. Cairo, G. Di Donfrancesco, J. M. Rosen, G. Durry, G. Held, and J. P. Pommereau (2007), Solid particles in the tropical lowest stratosphere, Atmos. Chem. Phys., 7, Okamoto, H., S. Iwasaki, M. Yasui, H. Horie, H. Kuroiwa, and H. Kumagai (2003), An algorithm for retrieval of cloud microphysics using 95 GHz cloud radar and lidar, J. Geophys. Res., 108(D7), 4226, doi: / 2001JD Ropelewski, C. F., and M. S. Halpert (1987), Global and regional scale precipitation patterns associated with the El Niño/southern oscillation, Mon. Weather Rev., 115, , doi: / (1987) 115<1606:GARSPP>2.0.CO;2. Savtchenko, A. (2009), Deep convection and upper tropospheric humidity: A look from the A Train, Geophys. Res. Lett., 36, L06814, doi: / 2009GL Savtchenko, A., R. Kummerer, P. Smith, A. Gopalan, S. Kempler, and G. Leptoukh (2008), A train data depot: Bringing atmospheric measurements together, IEEE Trans. Geosci. Remote Sens., 46, , doi: /tgrs Schmetz, J., S. A. Tjemkes, M. Gube, and L. van de Berg (1997), Monitoring deep convection and convective overshooting with METEOSAT, Adv. Space Res., 19(3), , doi: /s (97) Sherwood, S., and A. E. Dessler (2000), On the control of stratospheric humidity, Geophys. Res. Lett., 27, , doi: /2000gl Sherwood, S., and A. E. Dessler (2001), A model for transport across the tropical tropopause, J. Atmos. Sci., 58, , doi: / (2001)058<0765:amftat>2.0.co;2. Shibata, T., H. Vömel, S. Hamdi, S. Kaloka, F. Hasebe, M. Fujiwara, and M. Shiotani (2007), Tropical cirrus clouds near cold point tropopause under ice supersaturated conditions observed by lidar and balloon borne cryogenic frost point hygrometer, J. Geophys. Res., 112, D03210, doi: /2006jd Shindell, D. T. (2001), Climate and ozone response to increased stratospheric water vapour, Geophys. Res. Lett., 28, , doi: /1999gl Stephens, G. L., et al. (2002), The CloudSat mission and the A Train: A new dimension of space based observations of clouds and precipitation, Bull. Am. Meteorol. Soc., 83, , doi: /bams Tobin, D. C., H. E. Revercomb, R. O. Knuteson, B. M. Lesht, L. L. Strow, S. E. Hannon, W. F. Feltz, L. A. Moy, E. J. Fetzer, and T. S. Cress (2006), Atmospheric Radiation Measurement site atmospheric state best estimates for atmospheric infrared sounder temperature and water vapor retrieval validation, J. Geophys. Res., 111, D09S14, doi: / 2005JD Wang, P. H., P. Minnis, M. P. McCormick, G. S. Kent, and K. M. Skeens (1996), A 6 year climatology of cloud occurrence frequency from Stratospheric Aerosol and Gas Experiment II observations ( ), J. Geophys. Res., 101(D23), 29,407 29,430, doi: / 96JD Wang, P. K. (2003), Moisture plumes above thunderstorm anvils and their contributions to cross tropopause transport of water vapor in midlatitudes, J. Geophys. Res., 108(D6), 4194, doi: /2002jd Weinreb, M. P., R. Xie, J. H. Lienesch, and D. S. Crosby (1989), Destriping GOES images by matching empirical distribution functions, Remote Sens. Environ., 29, , doi: / (89) Wheeler, M., and G. Kiladis (1999), Convectively coupled equatorial waves: Analysis of clouds and temperature in the wavenumber frequency domain, J. Atmos. Sci., 56, , doi: / (1999) 056<0374:CCEWAO>2.0.CO;2. Winker, D. M., and C. R. Trepte (1998), Laminar cirrus observed near the tropical tropopause by LITE, Geophys. Res. Lett., 25, , doi: /98gl H. Ishimoto, Meteorological Research Institute, 1 1 Nagamine, Tsukuba, Ibaraki, , Japan. S. Iwasaki and J. Nakamoto, Department of Earth and Ocean Sciences, National Defense Academy, Hashirimizu, Yokosuka, Kanagawa , Japan. (iwasaki@nda.ac.jp) H. Kubota, Research Institute for Global Change, Japan Agency for Marine Earth Science and Technology, 2 15 Natsushima cho, Yokosuka, Kanagawa, , Japan. H. Okamoto, Center for Atmospheric and Oceanic Studies, Tohoku University, Aoba, Aramaki, Aoba ku, Sendai , Japan. T. Shibata, Graduate School of Environmental Studies, Nagoya University, D2 1(510), Furo cho, Chikusa ku, Nagoya , Japan. 10 of 10