Characteristics of tropical thin cirrus clouds deduced from joint CloudSat and CALIPSO observations



[1] The joint detection characteristics of both the CloudSat radar and Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) lidar are used to study tropical thin cirrus observed between 20°N and 20°S. The thin ice cloud category (TIC-1) of cirrus consists of those clouds detected by the lidar but not the radar whereas the TIC-2 cirrus category consists of clouds detected by both sensors. Tropical TIC-1 cirrus clouds between 20°N and 20°S are high, are optically thin, and have an approximate cloud cover in the defined region of 30%. Almost a third of this occurrence is in the form of single layers of cloudiness without any clouds below. These TIC-1 clouds also exhibit a marked seasonal variation, especially away from the equator, consistent with the shifts in annual cycle of convection with latitude. Lidar-based estimates of optical depth, uncorrected for multiple scattering, suggest that the TIC-1 optical depths range between 0.02 and 0.3. The ice water path of TIC-1 clouds is also estimated to be between 0.5 and 4 g m−2. The radiative properties of the TIC-1 clouds are also deduced from CloudSat flux data products at the top, at the bottom, and within the atmosphere. The influence of these clouds on the instantaneous reflected solar fluxes is determined to be less than 2 W m−2. The effects of TIC-1 clouds on the instantaneous outgoing longwave fluxes are estimated to be ∼20 W m−2, and the impact of these TIC-1 clouds on the tropics-wide average of the infrared heating is ∼4 W m−2.

1. Introduction

[2] Clouds affect the radiative balance of the Earth's atmosphere through competing greenhouse and albedo effects. These effects are of basic importance to our understanding of the Earth's climate system [Stephens, 2005]. It has been realized for some time that the greenhouse effect of high thin cirrus, in particular, exceeds their albedo effect and thus increases or decreases of thin cirrus in global warming experiments could enhance or reduce projected warmings [Stephens and Webster, 1981; Stephens et al., 1990]. It is also speculated that high tropical cirrus clouds, either by their greenhouse effect or from their influence on solar radiation, regulate the tropical atmosphere and sea surface temperature [Ramanathan and Collins, 1991; Lindzen et al., 2001]. In addition to regulating the radiative heating of the tropical atmosphere and surface, thin cirrus clouds are also deemed important in adjusting the tropospheric/stratospheric transport and to stratospheric dehydration [Corti et al., 2006; Hartmann et al., 2001].

[3] Thin cirrus is thought to be a ubiquitous feature of the tropical atmosphere [Dessler and Yang, 2003] but our ability to quantify the radiative effects of these thin clouds has proven difficult. They are composed of particles of complex shapes that complicate the scattering and absorption of radiation and models of radiative transfer through these clouds [Lo et al., 2006; Liou, 1972; Stephens, 1980]. These clouds are also difficult to observe with in situ probes given their general remoteness and the high altitude at which they form. Consequently, satellite observations of thin cirrus have emerged over the years as valuable tools to study these clouds. Infrared radiances from the TIROS-N Observational Vertical Sounders (TOVS) have been used to examine the seasonal variability of thin cirrus properties [Stubenrauch et al., 2006] and trends in thin clouds [Wylie and Menzel, 1999] but primarily for clouds with optical depths greater than about 0.2. The solar occultation measurements of the Stratospheric Aerosol and Gas Experiment (SAGE-II) has provided vertical layer structure of the thinnest of these cirrus clouds (with extinctions less than 0.03/km [e.g., Wang et al., 2001]) and the limb observations of cirrus clouds by the Microwave Limb Sounder (MLS) add valuable information about the ice water contents of the thin cirrus [Wu et al., 2008].

[4] The new satellite capabilities provided by CloudSat and CALIPSO complement these earlier satellite observations by providing a unique opportunity to document the distribution of thin cirrus clouds and their vertical structure [e.g., Sassen and Wang, 2008] in much greater detail. These new observations also offer the opportunity to characterize their radiative and microphysical properties. The purpose of this paper is to describe the optical and microphysical properties of tropical thin cirrus gleaned from joint CloudSat and CALIPSO observations and to provide an estimate of their contribution to the solar and infrared radiative fluxes and atmospheric absorption. It has been known for sometime that the potential importance of these clouds on the radiative budget of the atmosphere is grossly affected by the vertical structure of clouds and consequently the tropics wide impact of thin cirrus on the radiation budget of the tropics was not known prior to the advent of these new A-Train observations. Stephens et al. [2002], for example, hypothesized that the tropics-wide effects of the thinnest cirrus clouds on the radiative budget of the atmosphere is most likely small in comparison to the effects of other thicker upper tropospheric cloud layers due in part to the assumption that these thin clouds more often occur in the presence of other clouds in the column of atmosphere. This hypothesis is tested in this study using the information about cloud vertical structure and cloud optical properties from A-Train observations.

[5] After briefly describing data sources, a definition of two types of Thin Ice Clouds (TIC-1, TIC-2) are introduced in section 3 on the basis of these combined observations. In section 4 we analyze 2 years of data and find that TIC-1 clouds (observed only by the lidar) occupy about 30% of the tropics as defined between 20°N/S and that about 1/3 of these clouds exist as isolated layers with no clouds below. The TIC-1 clouds appear to shift with the seasonal migration of the ITCZ. The optical and microphysical properties of both TIC-1 and TIC-2 are discussed in section 5 are obtained from 3 months of data based largely on the optical depths retrieved using the CALIPSO lidar backscatter profiles and the matched ice microphysical products derived from 3 months of CloudSat radar data. Section 6 quantifies the associated radiative effects of thin ice clouds for this same 3 months and it is shown that the radiative effects on the tropical atmosphere as a whole is small although the effects of thin clouds locally on the outgoing longwave radiation is about 20 W m−2. A summary of the research and conclusions reached are presented in section 7.

2. Data Description

2.1. CloudSat Radar and CALIPSO Lidar Observations

[6] The CloudSat and the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) missions were jointly launched on 28 April 2006 on a Delta-2 rocket. Together the two satellites form the A-Train with previously launched satellites Aura, Aqua, and PARASOL. By flying CloudSat in tight formation with CALIPSO, the two satellites are able to observe the same scene about 15 s apart. Thus the observing system of A-Train satellites provides the opportunity to combine different types of observations for improving our understanding of the properties of clouds from space. This study focuses on the joint use of CloudSat and CALIPSO observations that are combined to determine the properties of thin cirrus clouds.

[7] The CloudSat Cloud Profiling Radar (CPR) is a 94-GHz nadir looking radar. At this wavelength it is possible to penetrate through ice clouds with negligible attenuation. The along track resolution of the CPR is 1.8 km with a vertical sampling of 240 m [Stephens et al., 2008]. CALIPSO's lidar CALIOP, (Cloud-Aerosol Lidar with Orthogonal Polarization, [Winker et al., 2004, 2007]) is used to detect thin clouds and aerosols, but thicker, optically deep clouds attenuate at the lidar. CALIOP produces two linearly polarized pulses at 532 nm and 1064 nm. The lidar information is averaged in such a way that the spatial resolution of the lidar varies with height from a horizontal resolution of 5 km above 30 km down to 333 m below 8 km. The vertical resolution varies from 300 m above 30 km to 30 m below 8 km [Winker et al., 2004].

2.2. Data Description

[8] The cloud mask information in CloudSat's 2B-GEOPROF product [Mace et al., 2007] and the lidar cloud fraction contained in the CloudSat 2B-GEOPROF-LIDAR product [Mace et al., 2009] are used to classify the cirrus as described in the following section. The latter product in particular exploits the advantages offered by the combination of the lidar and radar. Optical depths of thin cirrus are derived from lidar backscatter profiles using a transmission method similar to Lo et al. [2006], Comstock and Sassen [2001] and also Young [1995]. This method is applied to the level 1 CAL_LID_L1-Prov-V1-10 data and the results produced are similar to the optical depth data recently released by the CALIPSO science team but is not used in this study. MODIS radiances provided in the CloudSat MODIS-AUX product that is specifically matched to the CloudSat footprint, are also used. Two versions of the water and ice product 2B-CWC-RO are used and the 2B-FLXHR and 2B-FLXHR-LIDAR data sources [e.g., L'Ecuyer et al., 2008] are also used to characterize the radiative properties of thin cirrus clouds.

3. Classification of Thin Ice Clouds

[9] Two types of cirrus clouds are defined. Thin ice cloud type 1 (TIC-1) are those clouds observed only by the lidar according to the lidar cloud fraction exceeding 50% in a single radar volume as given in the 2B-GEOPROF-LIDAR product [see Mace et al., 2009]. TIC-2 clouds on the other hand, are those clouds observed jointly by the lidar and the radar. Only part of the cloud is required to be detected by both sensors to be considered as TIC-2. To minimize cases of obvious lidar attenuation, TIC-2 clouds are only counted when no more than 2 radar only bins are observed at the bottom of the cloud.

[10] An example of TIC-1 and TIC-2 categorization of thin clouds observed along a portion of an orbit is illustrated in Figure 1. Figure 1 (top) shows each pixel classified by cloud type according to whether the cloud was observed by the lidar, radar, or both sensors. Figure 1 (bottom) shows how the thin cloud classification is applied to this portion of the orbit.

Figure 1.

Cloud types as seen by CloudSat's CPR radar and CALIPSO's CALIOP lidar. (top) Cloud type defined as clouds observed by lidar only (gray circles), radar only (black squares), and both lidar and radar (light gray hatched). (bottom) Classification of TIC-1 (gray circles), TIC-2 (light gray hatched), or neither (black squares).

4. Distribution of Tropical TICs

4.1. Spatial Structures

[11] Figures 2a, 2b, and 2c respectively present the layer thickness, cloud top height and cloud base height characteristics of the TIC-1 clouds observed by the A-Train sensors for 2 years of data (July 2006 to June 2008). This composite information offers gross inferences about the vertical structure of TIC-1 clouds. These clouds are typically 1 to 2 km thick with cloud top heights ranging between 13 and 18 km with a mode in the distribution of cloud top heights occurring at 16 km.

Figure 2.

Occurrence of TIC-1 (a) cloud thickness, (b) cloud top height, and (c) cloud base height. Means of the data are 1.6 km, 14.8 km, and 13.2 km, respectively.

[12] Analysis of the statistics of the occurrence of all TIC-1 observed over this 2 year period confined between 20°N/S are obtained from a combination of both day and night 2B-GEOPROF-lidar data. For this period, we find that 30% of the defined tropical region contains TIC-1 cloudiness. A third of this occurrence (i.e., 33%) is in the form of single layers of cloudiness without any clouds below. Slightly less that 50% of the time TIC-1 cloud are detected, they occur with clouds that are located in a single lower layer below.

[13] The mean geographic distribution of occurrence of TIC-1 is presented in Figure 3 again on the basis of the 2 year period analyzed. The maps of occurrence were constructed by finding a percentage of CloudSat pixels in each 2 by 2° grid box that contained a TIC-1 cirrus cloud. The distributions of TIC-1 cirrus shown in Figure 3 also resembles the results of Bréon et al. [2005] who used GLAS lidar data to map clouds with optical depths less than 0.2. Throughout much of the Western Pacific and Indian Ocean, the occurrence of TIC-1 cirrus clouds approaches 50%. These areas of high occurrence correspond to climatological regions of deep convection as typified by low values of OLR (not shown).

Figure 3.

The percentage of CloudSat pixels that have a type 1 cirrus cloud present in each grid box.

[14] TIC-1 clouds cover large areas and occur frequently in regions where deep convection forms and most often occur with optically thicker clouds below them. As noted above, 2/3 of the time TIC-1 clouds are detected other clouds are also present within the profile. Since the profile of cloudiness through the entire column exerts an important influence on the radiative heating effects of thin cirrus, it is important to characterize the properties of TIC-1 clouds in relation to other clouds below. TIC-1 cirrus that occur above sufficiently high and deep clouds, for example, produce little radiative heating [Hartmann et al., 2001] in contrast to the radiative heating by these thin cirrus layers when lofted sufficiently high above any clouds below.

[15] Since the radiative heating by high clouds is determined in part by the temperature contrast between the upper level TIC cloud and any clouds below [Stephens, 1980], we present the following analysis in terms of the temperature difference ΔT between the base of TIC-1 clouds and the tops of the clouds below the TIC-1 layer in question. Using CloudSat's ECMWF-AUX temperatures, a cloud base temperature was determined for the last cloudy bin of TIC-1 cirrus as was the temperature for the top cloudy bin of nearest cloud below the TIC-1. Temperature differences are presented in degrees Kelvin and hereafter referred to as the cloud layer temperature difference (e.g., Figure 4a). Two distinct modes in the distribution of the ΔT emerge. The first corresponds to cloud layers with tops about 10° warmer than the base temperature of the TIC-1 and a second broad peak occurs at approximately ΔT = 75° warmer corresponding to warm boundary layer clouds lying well below the TIC-1 cirrus. To classify this second cloud type further, the joint distribution of the thickness of these lower clouds and the temperature difference is presented in Figure 4b. This analysis reveals that the clouds below TIC-1 cirrus at ΔT = 75° are boundary layer clouds that range between 0.5 and 2 km in thickness. The peak in frequency of clouds just 10° warmer than TIC-1 cloud base corresponds to a second optically thicker cirrus layer with characteristic thickness of approximately 1 km.

Figure 4.

TIC-1 clouds are characterized by where the cloud below occurs. The temperature difference is the difference between the cloud base temperature of TIC-1 cirrus and the cloud top temperature of the cloud directly below the TIC-1. (a) A pdf of cloud occurrence based on cloud temperature difference. (b) The occurrence of TIC-1 cirrus by cloud temperature difference and the thickness of the second cloud based on 2 years of observation.

4.2. Seasonal Distributions

[16] Tropical TIC-1 cirrus undergoes a distinct seasonal cycle evident in the approximately 2 years of CloudSat and CALIPSO data shown in Figures 5 and 6. Figure 5 shows the occurrences averaged over the entire region of study as well as for each hemisphere between the equator and 20°N/S and Figure 6 illustrates the seasonal cycle for clouds within selected latitude bands. The seasonal variation of the Northern Hemisphere is smaller than the Southern Hemisphere tropical averages and peaks around August with a minimum in DJF. The Southern Hemispheric variation ranges from a minimum of 20% in the JJA season and maximum near 40% in DJF. The seasonal cycles also appear more distinct the further away from the equator the clouds form as illustrated by the top two plots of Figure 6 that show high occurrence of TIC-1 clouds for both the Northern and Southern Hemisphere averages between 15–20° and 10–15° in the respective summer hemispheres.

Figure 5.

Monthly averages of frequency of occurrence of TIC-1 cirrus clouds from July 2006 to May 2008. Plotted are Northern Hemisphere tropical averages (20°N to equator) as the dashed line, Southern Hemisphere tropical averages (equator to 20°S) as the dotted line, and a tropical average (20°N to 20°S) as the black solid line.

Figure 6.

Monthly occurrence of TIC-1 cirrus as a zonal average in 5° latitude bands between July 2006 and May 2008. The dotted line represents Northern Hemisphere zonal averages, and the dashed line represents the Southern Hemisphere zonal averages in each latitude range.

[17] It is of some interest to contemplate the processes that might govern the seasonal characteristic of TIC-1 cirrus shown. The existence of any broad-scale uplift is one mechanism of sustenance and any seasonal variation of such uplift might determine the character of the observed seasonal cycle. The “tape recorder” effect [Mote et al., 1996] describes the production of large-scale upward advection of water vapor entering the tropical stratosphere and the marking of the water vapor that enters there. The tape speed, or large-scale upward advection speed, has a substantial annual variation with a maximum in water vapor anomalies occurring in August–September and a minimum in February–March [Pumphrey et al., 2000]. The expected variation of cirrus by this mechanism is barely consistent with the observed seasonal cycle of the Northern Hemisphere TIC-1 clouds and inconsistent with the tropics-wide variation of the clouds (black solid line of Figure 5).

[18] Another obvious mechanism that provides some control on TIC-1 and TIC-2 clouds is the detrainment of ice from deep convection. The observed seasonal variations of TIC-1 clouds appears to synchronize with the seasonal shifts of the ITCZ and its convection hinting at the likely important role of deep convection on the observed seasonal cycle of TICs. We cannot determine from this study whether the main controlling mechanism of the TIC-1 cloud observed is due to large-scale uplift as might be related to the tape recorder effect or from deep convection and related detrainment processes. We remark, however, that the results presented here are consistent with McFarquhar et al. [2000] who found that thin cirrus that we interpret to be of the TIC-1 type are more often associated with deep convection, whereas the even thinner subvisible cirrus most likely form because of synoptic-scale uplift.

5. Optical and Microphysical Properties of Isolated TIC-1 and TIC-2 Clouds

[19] The properties reported in this section are derived for isolated cloud layers with a cloud base lying above 7 km and no intervening midlevel clouds between 3 km and 7 km. The reason the analysis is limited to these isolated cloud layers is to minimize complicating effects of clouds below on the radiative signatures observed from above. The radiative effects of these clouds are generalized to include all cloud profiles in section 6.

5.1. Optical Properties of Isolated TICs

[20] The optical depths of TIC-1 and TIC-2 are obtained using the lidar transmission method [e.g., Lo et al., 2006; Comstock and Sassen, 2001]. First, the nearest along-track CALIPSO footprint was matched to each CloudSat footprint. The next nearest 15 footprints were then averaged to create a 5 km average profile with much reduced noise in the average backscatter profile compared to a single profile. Even with this averaging, the optical depth analysis is limited to nighttime only as daytime profiles were still deemed too noisy for this application. ECMWF heights, pressures, and temperatures are also used and interpolated to the heights of the lidar in order to create a reference profile of Rayleigh backscatter. The Rayleigh backscatter is expressed as

equation image

where σm is Rayleigh backscatter cross section. Above the cloud layer this derived Rayleigh backscatter closely matches the measured backscatter as illustrated in the example of Figure 7. Below the cloud, given a clear column of sufficient depth, the measured backscatter is less than the calculated Rayleigh backscatter because of extinction by the intervening cloud layer. The measured backscatter in this region has the form

equation image

where the ratio of the measured backscatter below the cloud to the Rayleigh backscatter determined for the cloud free profile is the ratio of the coefficients, AM/AR. This ratio is simply a measure of the two-way cloud transmission, namely,

equation image

where τcld is the optical depth of the cloud and η is an empirical factor that crudely accounts for possible multiple scattering effects (η < 1 [e.g., Platt, 1981]). In the example shown in Figure 7, a thick cloud layer is observed between approximately 9 and 12 km. Both the measured backscatter and the Rayleigh backscatter above and below this cloud layer are displayed along with dashed fit lines. The two-way transmission and optical depth derived from (3) are 0.1 and 1.1, respectively, for this example with η = 1.

Figure 7.

An example of a backscatter profile used to calculate optical depth. The peak in the measured backscatter corresponds to a cloud from approximately 9 km to 13 km. A fit line to the measured data is represented by the dotted line near zero. The smooth line with a dotted line on top is the calculated Rayleigh backscatter and the associated fit line.

[21] The major sources of uncertainty in the calculation of the lidar backscatter derived optical depths include (1) the backscatter measurement errors due to calibration uncertainties, (2) multiple scattering effects, (3) measurement noise errors, (4) errors in the formulation of the Rayleigh scattering primarily due to errors in temperatures, and (5) the actual errors arising from fitting the data to equation (2). Errors in the formal representation of Rayleigh backscatter profiles from temperature errors are small, typically less then 2%, [Stephens et al., 2002] compared to other sources. However, when fitting the measured backscatter to (2), errors can be substantial, primarily because of noisy nature of the data especially below the cloud. Measurement noise errors are included in the errors associated with fitting equation (2) to the backscatter data. Measurement errors in the backscatter profile from noise are reduced to below 1% with averaging and the total calibration error is under 2% (Z. Liu et al., Validation of CALIPSO Lidar (CALIOP) calibration, paper presented at 23rd International Laser Radar Conference, Indeco, Inc., Nara, Japan, 2006). Combining these errors we estimate the optical depth error as

equation image

Assuming a minimum detectible optical depth of τ = 0.01 (D. M. Winker, personal communication, 2008), the total error in the calculation of optical depth for TIC-1 and TIC-2 clouds is approximately 11% which we deem optimistic.

[22] Providing an error estimate due to uncertain effects of multiple is difficult. Multiple scattering is complex and ignoring its effects entirely (i.e., η = 1) will lead to a bias in estimated optical depth that is expected to vary with optical depth and other factors. The degree of bias should increase with increasing optical depth and it likely to be small for optical depths significantly less than unity [e.g., Mitrescu et al., 2005]. However, Winker [2003] suggests that this bias may be as large as 25% to 30% (i.e., η = 0.7–0.75) and independent of optical depth. There is much uncertainty in accounting for multiple scattering effects in spaceborne lidar observations and much more research is needed. The simple approach inherent to (3) above is itself questionable [e.g., Mitrescu et al., 2005] and as a consequence we choose to set η = 1 and hereafter refer to our estimated optical depths as uncorrected realizing that the optical depths quoted hereafter may be underestimated by as much as 25%.

[23] Optical depths for TIC-1 clouds were derived using the method described above. The histogram of the uncorrected optical depth for all isolated layers of tropical TIC-1 clouds collected over the June, July, and August (JJA) season of 2006 between 20°N and 20°S is shown in Figure 8. The peak of the histogram occurs at values of 0.07, the mean corresponds to an optical depth of 0.1 with a standard deviation of 0.1 with optical depths below 0.3 lying within 2 standard deviations of the mean. These clouds are classified as thin cirrus according to Sassen and Cho [1992]. The histogram of optical depths presented for the JJA season does not change when derived for other seasons (not shown).

Figure 8.

Optical depths for TIC-1 derived from CALIPSO backscatter data for the months of June, July, and August 2006. Data only include nighttime profiles.

5.2. Microphysics of Isolated TIC-2

[24] The uncorrected optical depths were also derived for many of the classified isolated TIC-2 cirrus clouds that were clearly not completely attenuated, as evident by a detectable Rayleigh return and other returns below the cloud. These optical depths could then be matched to the CloudSat 2B-CWC-RO ice water path product [Austin et al., 2009] thus providing a novel way of relating the optical properties to the ice water paths using completely independent observations. PDFs of ice water content, ice water path, and effective radius derived from the 2B-CWC-RO product for these TIC-2 cirrus are shown in Figure 9. The ice water contents of these matched TIC-2 clouds range between 2 to 5 mg m−3 ice water paths from 1 to 6 g m−2, and effective radius range between approximately 40 to 60 μm.

Figure 9.

TIC-2 microphysical properties from 2B-CWC-RO. (top) Ice effective radius, (middle) ice water content, and (bottom) ice water path.

[25] The expression [Stephens, 1978]

equation image

where ρ is the density of ice provides the expectation of a relation between the optical depth and ice water path W and thus a way of checking the consistency between the radar observations and the lidar observations of these isolated TIC-2 clouds. This consistency is examined in Figure 10 which shows a remarkable relation between TIC-2 ice water path (IWP) (from either using the R03 and R04 versions of 2B-CWC-RO) and the lidar-based optical depths. The slope of this relationship defines re ∼ 85 μm which is larger than the 2B-CWC-RO range (Figure 9) but falls within this range if the 25% multiple scattering bias of Winker [2003] is assumed (i.e., η = 0.75).

Figure 10.

Optical depth versus ice water path for TIC 2 cirrus clouds. The asterisks represent R03 data with the dashed fit line, and the diamonds are R04 data with the solid fit line. Fit lines for the two data sets have a slope of 56 for R04 and 54 for R03. These slope values are a measure of particle effective radius. The radius associated with each data set is 40.7 (R04) and 39.3 (R03).

[26] By design, there is no CloudSat IWC or ice water path information available for the TIC-1 clouds. In order to provide some idea of the ice water path of these clouds, we simply assume a 40 μm particle size as being characteristic of TIC-1 clouds. This is presumably an over estimate of the expected particle sizes of such thin cirrus [McFarquhar et al., 2000] however there is no reliable information on the particles sizes of these clouds available for this study. This simple conversion produces the results shown in Figure 11 and indicates that the IWP of TIC-1 clouds peaks slightly below 2.0 g m−2 with a mean value of 2.8 g m−2.

Figure 11.

Estimated TIC-1 cirrus clouds ice water path from two sources. The solid line with diamonds is the IWP distribution inferred from the optical depth information assuming the effective radius used is 40 μm as inferred from the TIC-2 cirrus cloud data of Figure 5. The second distribution (solid line) is based on the flux emissivities. The stars represent an IWP derived from optical information using a 20 μm effective radius for reference.

6. Influence of TIC-1 on the Earth's Radiative Budget

[27] The influence of TIC-1 cirrus on the top of atmosphere (TOA), atmosphere, and bottom of atmosphere (BOA) radiation budgets are analyzed using the 2B-FLXHR and 2B-FLXHR-LIDAR products [L'Ecuyer et al., 2008]. As described above, the only difference between the 2B-FLXHR-LIDAR and 2B-FLXHR products is the addition of those clouds observed by the lidar that are included in the former product. Thus the 2B-FLXHR-LIDAR products include not only the TIC-1 cirrus clouds in question but also some amount of low clouds [L'Ecuyer et al., 2008] that are also undetected by the CloudSat radar. The differences reflect the additional contribution primarily by these two cloud types to the radiative fluxes deduced at the top, bottom and within the atmosphere.

[28] The cloud radiative effects (CRE), defined as the difference between all sky and clear sky fluxes (and expressed in W m−2), are each derived from 2B-FLXHR-LIDAR and 2B-FLXHR products. The quantities presented here are the differences in these two CREs and the sense of this difference is such that it is positive downward.

6.1. Radiative Effects of Isolated TIC-1 Clouds

[29] The cases in which only TIC-1 clouds occur in the column (above 7 km) were identified in the data and the results are presented in the form of TOA and BOA cloud radiative effect differences (these cases are listed as TIC-1 in Table 1). These cases are also referred to as isolated cirrus and two forms of these CRE differences are presented. The first are instantaneous flux differences produced for a column of atmosphere when only TIC-1 clouds occur in that column. The second quantity is the flux difference obtained after averaging over the tropics and thus includes both clear and cloudy columns and weights the cloud coverage into these flux estimates. These latter values represent the effects of isolated TIC-1 clouds on the total radiative budget of the tropics.

Table 1. Calculated Average Cloud Radiative Effect of TIC-1 Cirrus From the Difference of 2B-FLXHR-LIDAR and 2B-FLXHRa
 LWLW Tropics AverageSWSW Tropics AverageNETNET Tropics Average
  • a

    Five different cases are reported: isolated TIC-1 cirrus (TIC-1), TIC-1 above high clouds (high), TIC-1 above midlevel clouds (MID), TIC-1 above low clouds (LOW), and TIC-1 in all circumstances (ALL). The net effect along with the longwave and shortwave components are shown for the top of the atmosphere (TOA), the surface, and the atmospheric column. The average CRE of each cloud type on the entire tropics (20°N/S) is also given.


[30] The average TOA and BOA shortwave cloud radiative effects due to isolated TIC-1 clouds are determined to be −1.4 W m−2 and −1.6 W m−2 respectively. That is, on the basis of the optical properties inferred from the measurements we calculate 1.4 W m−2 more solar radiation is reflected when isolated TIC-1 clouds are present and 1.6 W m−2 less solar radiation reaches the surface than under clear sky conditions. In most cases, the change in solar radiation at the TOA is mirrored by a corresponding change at the BOA. By contrast, the average difference in the TOA longwave CRE of isolated TIC-1 clouds is 20.9 W m−2. Since positive is defined in the downward direction, this value indicates a decrease in outgoing longwave radiation due to the presence of these clouds. The BOA radiative longwave effect difference however is only 0.5 W m−2 where the differences in emission by TIC-1 clouds are almost entirely muted at the surface by the large overburdens of water vapor in the tropics. Thus we conclude that the average effect of isolated TIC-1 clouds on the atmosphere is overwhelmingly to instantaneously heat the atmosphere by 20.6 W m−2 primarily through the enhanced greenhouse effect of these clouds. When averaged throughout the tropics, isolated TIC-1 cirrus heat the atmosphere by 1.7 W m−2.

6.2. Radiative Effects of TIC-1 Clouds in Combination With Other Clouds

[31] Table 1 also shows the CRE differences due to TIC-1 clouds that are observed above 7 km in combination with clouds below. The average difference in the shortwave CRE of all clouds with all clouds present is −15.5 W m−2 (Table 1) and the longwave effect at the TOA is 25.8 W m−2. The shortwave and longwave BOA flux differences are −17.3 W m−2 and 11.9 W m−2 respectively resulting in a net surface flux difference of −5.4 W m−2. The difference between these values (notably the increased reflected solar, decreased transmitted solar and increased longwave flux to the surface) compared to the isolated TIC-1 clouds is primarily a result of the increased low cloudiness present in the 2B-FLXHR-LIDAR product in contrast to the 2B-FLXHR product. The addition of these low clouds and the increased cooling they represent slightly offsets the net atmospheric heating by TIC-1 reducing the instantaneous net radiative effect from 20.6 W m−2 to 15.7 W m−2. However, the tropics-wide average accounts for the frequency of occurrence, and as noted the isolated TIC-1 have a tropics-wide effect of 1.7 W m−2, whereas all nonradar cirrus combined with all clouds have a 3.9 W m−2 radiative heating effect on the tropics. This total 3.9 W m−2 heating arises mostly from the combination of longwave heating by isolated TIC-1 clouds (1.7 W m−2) and the thicker cirrus below them (1 W m−2).

[32] Figure 12 expands on the results of Table 1 and shows the CRE difference more specifically as a function of the temperature difference between the base of the TIC-1 and the top of the cloud layer below. The longwave and shortwave CRE of TIC-1 cirrus with no other clouds present is represented by the solid line in Figure 12 and the value this line corresponds to is the same respective values given in Table 1. Similarly, the dashed line is the average effect of all “nonradar” clouds with all other specific cloud conditions (also give in Table 1). The symbols are the contribution of all nonradar clouds as a function of the temperature difference between the base of the TIC-1 (located above 7 km) of the immediate underlying cloud below. Similarly Figure 13 is the net CRE difference for all nonradar clouds as a function of this same temperature difference. The large increases in the reflected shortwave radiation at the TOA and surface occurring evident between 60 and 100° temperature difference supports the assertion that these effects are attributed to an increase in low level lidar clouds and these clods tend to be most prevalent when low clouds are also determined to exist in the vicinity of these lidar clouds. Since the TOA and surface shortwave effects are very similar in magnitude, the atmospheric contribution is determined to be very small, ranging between 0 and 5 W m−2. The longwave effects differ considerably between the top of the atmosphere and surface creating an atmospheric effect that largely determines the net atmospheric radiative effect as already noted. Both in the longwave and net, nonradar clouds with high and midlevel clouds add a radiative heating effect between 15 and 20 W m−2 (heating) which then shifts to a cooling between 0 and −15 W m−2 for these clouds over lower clouds approximately 80–100° warmer than the TIC-1 clouds.

Figure 12.

The average cloud radiative effect of nonradar clouds cirrus displayed as a change in the cloud radiative effect between 2B-FLXHR-LIDAR minus 2B-FLXHR. The average cloud radiative effect change is plotted as a function of the temperature difference between the TIC-1 cloud identified above 7 km and the next cloud below. The average cloud radiative effect of only TIC-1 cirrus in an atmospheric column is represented by the solid line. The dotted line is the average cloud radiative effect of all nonradar clouds with other clouds in the column.

Figure 13.

The average change in net cloud radiative effect with the addition of all nonradar clouds to the column. This is found as the difference in 2B-FLXHR-LIDAR minus 2B-FLXHR cloud radiative effects at the (top) top of the atmosphere, (middle) surface, and (bottom) atmospheric column. The average cloud radiative effect change is plotted as a function of the temperature difference between the TIC-1 cloud identified above 7 km and the next cloud below. This change represents the cloud radiative effect of these nonradar clouds.

6.3. Infrared Emissivities

[33] Another way to consider these longwave effects is in terms of the emissivities inferred from the flux products. These emissivities provide a different way of evaluating the IWP information presented in Figure 11. The emissivities of isolated TIC-1 cirrus clouds are deduced using

equation image

where the TOA fluxes for both the observed cloudy column and an equivalent clear-sky column are available in the 2B-FLXHR-LIDAR product and midcloud temperatures are found using CloudSat's ECMWF-AUX data. The emissivity results are presented in Figure 14 for all isolated cases of TIC-1 cirrus for the June, July, August season of 2006. The mean emissivity of the distribution is 0.13 which is approximately equivalent to a visible optical depth of 0.17 that is larger than the uncorrected lidar derived optical depths (Figure 8). (The flux emissivity follows as ɛ = 1 − image where the 1.66 is the diffusivity factor and the factor of 1/2 is the approximate conversion of the visible optical depth to infrared absorption optical depth. A value of τvis ∼ 0.1 leads to ɛ ∼ 0.08.) These TIC-1 emissivities are also applied to the relationship [Stephens, 1978]

equation image

to derive the IWP that are also shown in Figure 11 (solid line). In comparison to the optical depth and effective radius derived IWP, the emissivity-derived values are smaller with an average value of 1.9 g m−2 and a peak in the distribution at 0.8 g m−2. For reference, the IWP inferred from optical data assuming an effective radius of 20 μm is also shown in Figure 11.

Figure 14.

The distribution of the emissivity of TIC-1 cirrus calculated from the 2B-FLXHR-LIDAR fluxes. The average value is 0.13 and a standard deviation of 0.1.

7. Summary and Conclusion

[34] Observations collected by CloudSat and CALIPSO are combined to study tropical thin cirrus clouds. Previous studies that have used satellite lidar data have been severely limited in time (3 days) and mostly refer to occurrence properties [Dessler and Yang, 2003]. Other studies that utilize infrared radiance data are limited in their detection of thin cirrus with optical depths above 0.2 [e.g., Wylie and Menzel, 1999; Prabhakara et al., 1993; Stubenrauch et al., 2006] and thus miss the thinnest cloud layers. In this study we introduce a new classification of thin cirrus based on the joint detection characteristics of both the CloudSat radar and CALIPSO lidar. We introduce TIC-1 cirrus as the cirrus detected by the lidar but not the radar and TIC-2 cirrus as those high clouds detected by both sensors.

[35] Most of the focus of this study is on the radiative effects of TIC-1 clouds between 20°N and 20°S. We combine this classification with other properties provided in the 2B-GEOPROF, 2B-GEOPROF-LIDAR, the 2B-CWC-RO and the 2B-FLXHR and 2B-FLXHR-LIDAR products to determine the optical and microphysical properties of TIC-1 and TIC-2 clouds and their radiative effects. The main findings of this paper are:

[36] 1. Tropical TIC-1 cirrus clouds are high and optically thin. We find these cloud layers are typically a kilometer thick or deeper and have a characteristic optical depth of approximately 0.1 and greater. They reside just below the tropical tropopause with tops ranging between 13 and 18 km.

[37] 2. Approximately 30% of the tropical region between 20°N/S contains TIC-1 cloudiness. Almost a third of this occurrence is in the form of single layers of cloudiness without any clouds below. Slightly less that 50% of the time TIC-1 clouds are detected, they occur with clouds that are located in a single lower layer below, typically either a second cirrus layer or boundary level clouds.

[38] 3. A clear seasonal cycle exists in the occurrence of the tropical TIC-1 cirrus observed by the A-Train. The seasonal variation of the Northern Hemisphere (Figure 5) is smaller than the Southern Hemisphere tropical averages where the former variation peaks around August and has a minimum in DJF. The seasonal cycle become more marked in regions further away from the equator (Figure 6) and appears to be associated with the seasonal shifts of tropical convection, moving with the location of the ITCZ.

[39] 4. On the basis of lidar transmission estimates of optical depth, we infer the TIC-1 optical depths range between 0.02 to 0.3 (Figure 8) with a mean value of 0.1. These optical depths are not corrected for multiple scattering and may be underestimated by as much as 25% because of uncertain effects of multiple scattering. The mean value of optical depth, even when uncorrected for multiple scattering, is consistent with the flux emissivity estimates derived from the flux data analyzed in this study. The ice water path (IWP) of TIC-1 clouds is also estimated to be between 0.5 and 4 g m−2 assuming a 40 μm sized particle. This range in IWP is likely to be an overestimate given that the sizes of TIC-1 cloud particles are expected to be smaller than the 40 μm assumed.

[40] 5. On the basis of the 2B-CWC-RO product of CloudSat, the TIC-2 cirrus clouds, detected by both the lidar and the radar, exhibit ice water contents that range between 2 to 5 mg m−3, ice water paths ranging between 1 to 6 g m−2, and effective radius from 40 to 60 μm (Figure 9). The latter particle sizes reasonably match the particle sizes derived from the lidar optical depths of TIC-2 clouds and the 2B-CWC-RO IWPs inferred from Figure 10 only with a multiple correction.

[41] 6. The radiative properties of the TIC-1 clouds were deduced from the difference between the 2B-FLXHR and 2B-FLXHR-LIDAR products. The influence of isolated TIC-1 clouds on the instantaneous shortwave fluxes calculated at the TOA and BOA are small (less than 2 W m−2). The influence on instantaneous longwave fluxes at the surface is also small (less than 1 W m−2) however the effects of TIC-1 clouds on the instantaneous outgoing longwave radiation is significant estimated to be approximately 21 W m−2. TIC-1 clouds are a source of infrared heating of the tropical atmosphere of ∼20.6 W m−2 for an instantaneous average value or ∼4 W m−2 for the tropics-wide average. This heating primarily arises from the longwave heating by isolated TIC-1 clouds and TIC-1 clouds in association with a thicker cirrus layer below.


[42] This work is supported by in part by NASA grants NNX07AR11G and NNX06AB63G.