Using the polarization diversity lidar data collected during March 1998 to April 2001, the characteristics of the tropical cirrus clouds observed over Gadanki (13.5°N; 79.2°E) are presented in this paper. Out of 210 nights of observations during this period, cirrus clouds were observed over the lidar site on 170 nights. The cloud mean height is found to be in the range of 8–17 km with peak occurrence at 13–14 km, just below the tropopause. The cloud thickness has values ranging 0.6–4.2 km with maximum occurrence at 0.9–1.2 km. The scattering ratio and the linear depolarization ratio are in the ranges of 1.14–36 and 0.01–0.75 with peak occurrences seen at 1.14–2 and 0.01–0.05, respectively. The optical depth can be as high as 2, but values less than 0.1 account for most (>80%) of the clouds. The clouds with low optical depth occur most frequently during fall equinox and high optical depth during summer. The formations of thin and thick cirrus clouds are seen to be closely related to the minimum tropospheric temperature and cumulonimbus outflows, respectively.
 Cirrus clouds in the upper troposphere and the lower stratosphere have attracted much attention recently because of their impact on the radiation budget [Liou, 1986; Ackerman et al., 1988; Boehm et al., 1999; Dowling and Radke, 1990; Prabhakara et al., 1993]. They affect the earth's climate by reflecting incoming sunlight and regulating heat loss from the earth's surface [Liou, 1986; Ramanathan and Collins, 1991; Arking and Zaskin, 1994]. Due to the differences in their microphysical and optical properties, they cause a warming effect at tropics and a cooling effect at midlatitudes [Liou, 1986; Nicolas et al., 1997]. Cirrus clouds are further thought to be the platform for the genesis of heterogeneous reactions that play an important role on the ozone budget in lower stratosphere/upper troposphere [Borrmann et al., 1996; Roumeau et al., 2000]. Accordingly, a study on such clouds in tropics assumes significance, as vertical transport of air from troposphere to stratosphere is believed to occur mostly in this latitude region [Brewer, 1949]. The observations on cirrus clouds in the tropical region, however, are very few, notwithstanding the fact that formation of cirrus in this region is more prevalent because of very cold environment [Jensen et al., 1996; Platt et al., 1998]. Although satellite observations like that of Stratospheric Aerosol and Gas Experiment-II (SAGE-II) provide a global picture on cirrus [Wang et al., 1994, 1996], a more detailed study requires high temporal and spatial resolution data to supplement them. Ground-based lidars have widely been used to study high temporal and spatial structure of cirrus at mid and high latitudes (see Dowling and Radke  for an overview).
 Large number of experiments were conducted to investigate the importance of cirrus clouds in the radiation budget using various techniques employing lidars and satellites [Liou, 1986; Prabhakara et al., 1988; Ramanathan et al., 1989; Sassen et al., 1989; Rossow and Schiffer, 1991; Ansmann et al., 1992; Wang et al., 1996; Dessler, 1998; Roumeau et al., 2000; Hartmann et al., 2001; Goldfarb et al., 2001]. Based on the results from these observations, and using optical depth (τc) as a differentiator, the cirrus clouds can be classified into two categories, namely, optically thin (τc < 0.03) and thick (τc ≥ 0.03). The thin cirrus has not only less optical depth but also less spatial thickness than the thick cirrus and it has been widely observed at different locations. Initially, the observations of thin cirrus were started at Kwajalein using the lidar measurements [Uthe and Russell, 1977]. They found the presence of these clouds, having an average lifetime of the order of few days, in almost all the seasons. Generally, the horizontal extents of the cirrus clouds were found to be in the range of 100 km to 1000 km. Dowling and Radke  reviewed the existing observations on thin cirrus clouds and concluded that such clouds have the thickness of the order of 1.5 km and usually occur at an altitude of about 9 km. Climatological study on the subvisible cirrus clouds over midlatitude using 3 years of data from Observatory Haute Province lidar (44°N, 6°E) by Goldfarb et al.  showed frequent occurrence of thin subvisible cirrus near to the tropopause height with maximum frequency of occurrence during fall equinox. On the other hand, the existing reports [Heymsfield and McFarquahar, 1996; Jensen et al., 1996; Platt et al., 1998; Sassen et al., 1998] suggest that thick cirrus clouds have lesser life-time (∼3–4 hours), occur mostly around 2 km below the tropopause, and have varying height and thickness as a function of time. These clouds, appearing as falling streaks, form at higher heights with large thickness and descend to lower heights with decreasing thickness and eventually disappear. Boehm et al.  have made a model study on the maintenance of tropical cirrus cloud and found that the regions of enhanced ice water content were associated with upward motion and reduced ice water content with downward motion.
Prabhakara et al. , using Nimbus 4 Infrared Interferometer Spectrometer (IRIS) measurements, studied the global distribution of thin cirrus. They found that such clouds are observed mostly in the tropics. Heymsfield and McFarquahar  and Roumeau et al.  further confirmed the prevalence of thin cirrus over tropics. Using SAGE II observations, Wang et al.  studied the climatological aspects of thin cirrus systems over the equator and showed that these clouds have optical depths less than 0.02, have 45% occurrence frequency and occur at an altitude of about 15 km. The observations from the First International Satellite Cloud Climatology Project Region Experiment (FIRE) and Central Equatorial Pacific Experiment (CEPEX) further provided a detailed study on cirrus [Jensen et al., 1996; Winker and Trepte, 1998]. These results, however, are mostly confined to southern part of western Pacific Ocean and information on cirrus clouds in northern tropical region is still scarce. Though the satellite observations provide valuable information on cirrus clouds, they have limitations on spectral, temporal and spatial sampling characteristics. The lidar technique has been successfully used to measure the scattering and depolarization characteristics of the high altitude clouds with good time and height resolutions.
 In this paper, the characteristics of cirrus clouds using lidar observations at a tropical station, Gadanki (13.5°N; 79.2°E), are presented. The characteristics are presented in terms of occurrence height, thickness, optical depth, scattering ratio and linear depolarization ratio. The study also includes the statistical variations of the above parameters based on 210 nights of observations made during March 1998–April 2001.
2. System Description
 A state-of-the art lidar system has recently been established at the National MST Radar Facility (NMRF), Gadanki (13.5°N, 79.2°E), India under a joint project between Department of Space, India and Communication Research Laboratory (CRL), Ministry of Posts and Telecommunications, Japan. The lidar uses an Nd-YAG Laser operating at its second harmonic of 532 nm with energy of 550 mJ and a pulse repetition frequency of 20 Hz. It comprises of two independent receivers, viz., Rayleigh receiver and Mie receiver. The Rayleigh receiver is used to measure Rayleigh backscatter from which temperature profiles can be derived over the height range of 30–80 km. The Mie receiver is used to measure the Mie scattering from aerosols to derive aerosol concentration and to study the cloud characteristics.
 The Mie receiver employs a compact Schmidt-Cassegraine type telescope with an effective diameter of 35 cm. A narrow band interference filter centered at 532 nm with Full Width at Half Maximum (FWHM) of 1.13 nm is used for cutting down the background light. A polarized beam splitter is employed for splitting the beam into co- and cross-polarized components with two channels designated as P and S, using identical photo-multiplier tubes (PMTs) and pulse discriminators. The outputs of the pulse discriminators are connected to a PC-based photon counting data acquisition system which records 5000 laser-shot averaged photon count profiles for both channels as one frame with a time resolution of 250 s and range resolution of 300 m. More details on NMRF-CRL lidar can be found in Raghunath et al.  and Bhavanikumar et al. .
3. Data Analysis
3.1. Lidar Data
 The basic lidar data are in the form of photon count profiles with a time resolution of 250 s and a height resolution of 300 m. Each profile represents a time integration that corresponds to 5000 laser shots. The backscattered signal, expressed in terms of scattering ratio S(z), is given by Fernald :
where βm(z) and βa(z) are the molecular and aerosol backscattering coefficients.
 The scattering ratio computation involves separating the aerosol and molecular contributions in the backscattered signal. It is accomplished by the normalization of the photon count with molecular density at a specified height (18 km) taken from a model atmosphere (CIRA-86). The scattering ratio profiles as computed above from the co-polarized channel (P) are employed for the purpose of studying the cloud characteristics and the other polarized channel (S) is used for the computation of linear depolarization ratio.
 The linear depolarization ratio (LDR) is expressed as the ratio of the scattering ratios of the cross-polarized (S⟂) and the co-polarized (Sll) components of the backscattered signal and is denoted by δ,
where δa (=0.014) is the LDR of the air particles.
 The opacity of the cloud is related to optical depth, defined as:
where α (z) is the extinction coefficient.
 The cirrus clouds are identified using the following criteria. The value of the scattering ratio should exceed 1.14 and should persist for at least three consecutive range bins. The data used for the present study are limited to nighttime, the daytime observations being not possible due to a high level of background photon count associated with the daylight.
3.2. Radiosonde Data
 Since there is no co-located radiosonde observation facility at Gadanki, the temperature measurements used for the study were taken from the radiosonde flight conducted by India Meteorological Department (IMD), Chennai (80.2°E, 13.1°N; 120 km southwest to Gadanki). Usually radiosondes are launched at Chennai twice a day, once at 0530 LT (00 GMT) and again at 1730 LT (12 GMT). For the background atmospheric temperature at the cloud heights, we have used 0530 LT measurements, since the time of this launch is closer to the lidar observation period. Here, we have considered the cold point tropopause for identifying the tropopause height.
 The basic lidar data are in the form of photon count profiles with a height resolution of 300 m and a time resolution of 250 s. The photon count profiles, showing exponential decay with height in the absence of a cloud, are reduced to scattering ratio profiles and are used to detect clouds and determine their spatial and temporal characteristics. The presence of a cloud results in a sharp enhancement in the scattering ratio making its detection quite unambiguous. Figure 1a, which shows the altitude profile of scattering ratio, illustrates the presence of a cloud on the night of 08–09 May 1998. In this figure, the scattering ratio is found to increase sharply to a value as high as 6 between 13 and 15 km. The thickness of the cloud is of the order of 2 km.
 The dual polarization (co- and cross) measurements of the backscattered signal can be used to determine the relative concentrations of water and ice in the clouds. Since the ice particles are different in their orientation, shape and size, they cause corresponding changes in the depolarization value [Sun et al., 1989; Sassen et al., 1990; Sassen, 1995; Sassen and Benson, 2000; Jensen et al., 1996]. The measured values of scattering ratio coupled with the depolarization are conventionally used to identify the composition of the cloud. The low value of scattering ratio with high depolarization indicates that the clouds are composed of anisotropic particles (like ice). On the other hand, high scattering ratio and low depolarization value identifies the clouds composed of isotropic particles. Lidar observations of cirrus clouds have regularly shown large depolarization, indicating the presence of ice in their composition. For the example shown in Figure 1b, the maximum depolarization ratio is found to be 15% (0.15), indicating that the cloud has significant component of ice. The formations of ice nucleation at those heights have been extensively reported [Platt et al., 1989; Sun et al., 1989; Sassen et al., 1990; Sassen, 1995; Jensen et al., 1996; Sassen and Benson, 2000].
5. Results and Discussion
5.1. Height-Time Structures of Cirrus Clouds
5.1.1. Optically Thin Cirrus Clouds
Figures 2a and 2b show height-time contour plots of scattering ratio, representing two typical events of stratiform clouds observed on the nights of 18 March 1998 and 13 January 1999. The observations show the occurrence of thin cirrus clouds over the height range of about 11–12 km on 18 March 1998 and 14–15 km on 13 January 1999. The thickness of cloud in both the cases is of the order of ∼1 km. This type of cirrus is called laminar cirrus and has the characteristics of long persistence and small optical depth (τc < 0.03) [Dowling and Radke, 1990; Prabhakara et al., 1988, 1993; Winker and Trepte, 1998]. The generative cause of these clouds at those height ranges is attributed to the ice nucleation [Sassen et al., 1990; Sassen, 1995; Prabhakara et al., 1988, 1993; Macke, 1993; Goldfarb et al., 2001]. The background temperature is found to be ∼220 K at 12 km on 18 March 1998 and ∼206 K at 15 km on 13 January 1999.
 Using the data collected from March 1998–April 2001 an attempt has been made to study the morphology of cirrus clouds over Gadanki (13.5°N; 79.2°E), a northern tropical station. Figures 3a–3e present the occurrence distributions for the cloud parameters of mean height, thickness, scattering ratio, depolarization ratio and optical depth. The statistical results obtained from 170 cloud events can be summarized as follows:
5.2.1. Occurrence Height
Figure 3a presents the distribution of the cirrus cloud mean height. The occurrence height of the cirrus clouds is confined to the height range of 8–17 km. The cirrus is generally found to occur somewhat close to the tropopause, with the maximum occurrence in the range of 13–14 km. The cold point tropopause obtained from the IMD data is used to study the occurrence height of the cirrus in relation to the tropopause level (Figure 4). It is found that the occurrence of thin cirrus (τc < 0.03) is mostly close to the tropopause height [Dowling and Radke, 1990; Prabhakara et al., 1988, 1993; Winker and Trepte, 1998; Boehm et al., 1999]. In a few cases, the occurrence height of cirrus was found to be above the tropopause level. At midlatitudes, on the other hand, the height of occurrence of thin cirrus clouds is mostly between 8.5 and 11.5 km [Sassen et al., 1990; Ansmann et al., 1992; Jensen et al., 1996; Heymsfield and McFarquahar, 1996; Goldfarb et al., 2001]. This seems to be due to the fact that tropopause occurs at higher heights in tropics (16–17 km) than at midlatitudes (10–12 km).
 The thickness of the cloud is found to be in the range ∼0.6–4.2 km (Figure 3b). The distribution shows maximum occurrence of cloud thickness in the range of 0.9–1.2 km. The clouds having low values of thickness (≲1 km) are also generally optically thin and those having optical depth less than 0.03 are referred to as subvisible cirrus [Jensen et al., 1996; Goldfarb et al., 2001]. Due to the entrainment and detrainment of air at the boundaries of the cloud, its thickness may decrease and sometimes the cloud may also decay [Platt et al., 1989; Sassen et al., 1989; Boehm et al., 1999].
5.2.3. Scattering Ratio
 The distribution for the mean scattering ratio, representing the average over the cloud extent, is as shown in Figure 3c with two major peaks, one at 0–2 and the other at 34–36. The large values of scattering ratio suggest the presence of scatterers of large scattering cross-section which may be due to homogeneous nucleation. The thin cirrus clouds, however, are composed of ice (nonhomogeneous) and are characterized by lower values of scattering ratio. The study based on the zonally averaged interactive chemistry-radiation-dynamical model also shows the scattering ratios to be in the range of 3–15 at 603 nm [Rosenfield et al., 1998]. The scattering ratio is mainly dependent on the size of the particles composed in the cloud. A modeling study on the maintenance of tropical cirrus revealed that the top of the cloud layer is composed of a large number of small crystals with increasing size and the crystal number density decreases with decreasing altitude due to differential sedimentation [Boehm et al., 1999].
5.2.4. Linear Depolarization Ratio (LDR)
Sassen  classified the cloud composition based on the computed LDR values as shown in Table 1. The LDR values obtained are distributed over the range of 1% to 75%, with peak occurrences over the range of 1–5% (Figure 3d). The wide variations in the LDR values are due to changes in the cloud composition resulting from the corresponding nucleation processes. The large values of LDR are believed to be due to the presence of corona producing ice crystals [Sassen et al., 1998]. In addition, breaking of Kelvin-Helmholtz wave at a boundary of the cloud can result in large values of LDR (0.6–0.7) in the crests of Kelvin waves [Boehm and Verlinde, 2000]. Usually, the thick cirrus clouds have large scattering ratio with low LDR value at the cloud base, suggesting the presence of reduced ice water content [Platt et al., 1989; Boehm et al., 1999]. Boehm et al.  found that the regions of enhanced ice water content were associated with upward motion, where as the regions of reduced ice water content were associated with downward motion. The measured temperature and LDR show an anticorrelation, i.e., if the temperature is low, the value of LDR is high and vice versa. This is to be expected since the ice nucleation would be more at lower temperatures.
Table 1. Classification of Cloud Composition
Composition of Cloud
supercooled (like ice)
complex ice crystals or aggregates
moderate to heavy rimed ice particles
5.2.5. Optical Depth
 Optical depth is a crucial parameter concerning the radiation and scattering processes of the cloud [Platt et al., 1989]. The measured values of optical depth are found to extend over a range of 0.001 to 3.0 with peak occurrence confined to a narrow range of 0.001–0.01 as shown in Figure 3e. The distribution shows that optically thin cirrus (τc < 0.03) account for nearly 70% of the clouds, while optically thick clouds (τc > 1.0) make up for less than 10%. The variability in optical depth depends on the nature of composition and the thickness of the cloud [Platt et al., 1989; Sassen et al., 1989; Sassen and Benson, 2000; Jensen et al., 1996].
5.2.6. Seasonal Variation and Life Period
 The cirrus clouds were found to occur on ∼80% of the observational nights which is due to the favorable conditions for their formation close to tropopause level over tropics. The percentage of occurrences of cirrus during spring, winter, summer and fall are 71.43, 74.7, 70.6, and 94.6, respectively. The cirrus clouds observed during equinoctial periods are generally thin with low values of optical depth (τc < 0.1) as has been observed also over midlatitude [Goldfarb et al., 2001]. The clouds having high optical depth with values approaching as high as 2–3 were observed mostly during summer when the convective activity over tropics is most intense. Cumulonimbus turrets, which are very common during this period in tropics, dissipate and leave their anvil at higher heights. These anvils, rich in water vapor content, metamorphoses in optically thick cirrus [Sassen et al., 1998; Goldfarb et al., 2001]. Knollenberg et al.  determined the ice number densities in the cloud composition and found that they are greater than 10 cm−3 and ice water content of up to ∼0.07 gm−3 in tropical anvils. Heymsfield and McFarquahar  also measured mean ice water content of 0.007 gm−2 in the lower-altitude topical cirrus. Owing to strong turbulence, the thick cirrus clouds have a short lifetime and decay within ∼3–4 hours [Jensen et al., 1996; Sassen et al., 1998], whereas the thin cirrus have lifetimes extending from a few hours to a few days [Boehm and Verlinde, 2000; Winker and Trepte, 1998; Jensen et al., 1996]. Data collected on 45 consecutive nights show that the thin cirrus clouds have persistence of 3–4 days, in contrast to the thick cirrus, which persists not more than 3–4 hours. Ackerman et al.  have proposed an explanation for the persistence of tropical cirrus for long time and it is based on radiative destabilization leading to cloud circulation which acts to maintain ice in the cloud layer. The nonavailability of daytime data and the passage of low-level clouds hampering the observations made it difficult to study the time-history of the cirrus clouds. Statistical studies on cirrus clouds have shown that the thin cirrus is observed more frequently during nighttime than during daytime [Dowling and Radke, 1990; Boehm et al., 1999].
5.2.7. Background Temperature
 The minimum tropopause temperature (∼190–210 K) occurs over tropics, favoring frequent occurrence of cirrus at these latitudes. This suggests a dominant presence of ice in the cirrus clouds over tropics. The cirrus clouds occurring much below the tropopause have greater temperatures and are accordingly of different composition compared to that occurring close to the tropopause level. The presence of water vapor in the thick cirrus is due to the cumulonimbus anvils that bring water from lower heights [Knollenberg et al., 1993; Heymsfield and McFarquahar, 1996; Jensen et al., 1996] as well as due to radiative forcing [Boehm et al., 1999]. The model-based calculation shows that the required background temperature for the formation of cirrus at tropics is in the range of ∼200–216 K [Rosenfield et al., 1998].
 The lidar data collected over Gadanki (13.5°N; 79.2°E) for 210 nights were used to characterize the tropical cirrus through a statistical study. The data revealed the occurrence of cirrus clouds on most of the observational nights (∼80%). The maximum occurrence (∼95%) was found to be during fall equinox, the clouds being generally of stratiform kind with low values of optical depth (τc < 0.1). The clouds having high optical depth with values ranging as high as 2–3 were observed mostly during summer when the convective activity is most intense over the tropics. The formations of the above mentioned thin and thick cirrus clouds seem to be related to the processes of ice nucleation close to tropopause and cumulonimbus outflows, respectively.
 The National MST Radar facility (NMRF) is operated by the Department of Space (DOS) with partial support from the Council of Scientific and Industrial Research (CSIR), Government of India. The authors would like to acknowledge with thanks the numerous contributions by their colleagues at the NMRF, India, and the Communications Research Laboratory (CRL), Japan, in establishing and operating the Lidar facility under an Indo-Japanese collaboration program.