Journal of Geophysical Research: Atmospheres

Spectral width of premonsoon and monsoon clouds over Indo-Gangetic valley


Corresponding author: T. V. Prabha, Indian Institute of Tropical Meteorology, Dr. Homi Bhabha Road, Pashan, Pune 411008, India. (;


[1] The combined effect of humidity and aerosol on cloud droplet spectral width (σ) in continental monsoon clouds is a topic of significant relevance for precipitation and radiation budgets over monsoon regions. The droplet spectral width in polluted, dry premonsoon conditions and moist monsoon conditions observed near the Himalayan Foothills region during Cloud Aerosol Interaction and Precipitation Enhancement EXperiment (CAIPEEX) is the focus of this study. Here σis small in premonsoon clouds developing from dry boundary layers. This is attributed to numerous aerosol particles and the absence/suppression of collision-coalescence during premonsoon. For polluted and dry premonsoon clouds,σ is constant with height. In contrast to premonsoon clouds, σ in monsoon clouds increases with height irrespective of whether they are polluted or clean. The mean radius of polluted monsoon clouds is half that of clean monsoon clouds. In monsoon clouds, both mean radius and σ decreased with total cloud droplet number concentration (CDNC). The spectral widths of premonsoon clouds were independent of total droplet number concentrations, but both σ and mean radius decreased with small droplet (diameter < 20 μm) number concentrations in the diluted part of the cloud. Observational evidence is provided for the formation of large droplets in the adiabatic regions of monsoon clouds. The number concentration of small droplets is found to decrease in the diluted cloud volumes that may be characterized by various spectral widths or mean droplet radii.

1. Introduction

[2] Aerosol particles (AP) in the atmosphere from both natural and anthropogenic emissions influence the Earth-Atmosphere system through direct and indirect effects. The latter modifies clouds, precipitation and radiation budgets [Ramanathan et al., 2001; Nakajima et al., 2001; Kim et al., 2003; Zhang et al., 2004; Grabowski, 2006; Rosenfeld et al., 2008; Khain 2009]. It is well known that an increase in aerosol loading results in increase in cloud condensation nuclei (CCN) concentrations. This further leads to increases in cloud droplet number concentration (CDNC) and cloud albedo [Twomey, 1977]. The effective radius is related to the mean volume radius and the σ of the cloud droplet spectra [Martin et al., 1994; Liu and Daum, 2000], which is related to cloud radiative properties [Liu and Daum, 2002]. Droplet spectral width (the standard deviation of droplet size distribution) depends on many parameters such as aerosol chemical composition, aerosol size distributions, vertical velocity, entrainment and other parameters that determine droplet size distribution [Khain et al., 2000; Yum and Hudson, 2005; Liu et al., 2006a; Peng et al., 2007].

[3] Cloud droplet size distributions exhibit a variety of spectral shapes [Martin et al., 1994; Hudson and Yum 1997; Liu and Daum 2000; Miles et al., 2000]. The addition of anthropogenic aerosol to marine air masses enhances not only the CDNC, but also the spectral dispersion. The increased spectral dispersion has been described to offset the cooling due to the Twomey effect by as much as 10% to 80% [Liu and Daum, 2002; Pandithurai et al., 2012].

[4] According to theoretical predictions, σ should decrease with height (or with mean radius) in adiabatic clouds. Various studies [Hudson and Yum, 1997; Miles et al., 2000] illustrate that theoretical predictions give much smaller droplet spectral widths than observations [Politovich, 1993; Martin et al., 1994; Hudson and Yum, 1997]. For a given flight, local droplet concentration varied considerably during the Second Aerosol Characterization Experiment (ACE2). The width of cloud droplet spectra was typically in the range of 1 to 2 μm [Pawlowska et al., 2006]. The spectral width did not vary systematically between pristine and polluted clouds and a small difference between near-adiabatic and diluted cloud regions was noted in that study. The width of the droplet size distribution and the relative dispersion were similar to that of stratocumulus clouds in the lowest few hundred meters during RICO (Rain In Cumulus over the Ocean) [Arabas et al., 2009], but they were significantly larger in the upper parts of the clouds. Miles et al. [2000] gives a comprehensive list of observations encompassing marine and continental cases. However, there are no studies of dispersion in tropical continental deep convective clouds and especially for highly polluted premonsoon and monsoon conditions.

[5] In an attempt to better understand cloud-aerosol interactions, a major experiment named Cloud Aerosol Interaction and Precipitation Enhancement EXperiment (CAIPEEX) is underway in India. Observations of convective clouds, aerosol and CCN were carried out during this experiment, mainly over continental regions using an instrumented aircraft (SOAR Piper Cheyenne II, seeAxisa et al. [2005]). The observations during Phase I of the experiment were taken during May–September 2009. A detailed description of the above project, science plan and instruments mounted on the aircraft can be found at∼caipeex and in Kulkarni et al. [2012]. Premonsoon (March, April and May) and monsoon (June, July, August, and September) clouds develop in different thermodynamic and aerosol conditions. Premonsoon clouds develop in dry and polluted conditions; so that cloud base height exceeds 2 km. Cloud base height for monsoon clouds is well below 2 km as they develop in very humid environments. Though the concentration of aerosol particles (AP) and CCN during monsoon periods is quite high, it is still significantly lower than concentrations during premonsoon periods [Prabha et al., 2011]. Aerosol indirect effect and what fraction of it can be offset by droplet dispersion is quantitatively estimated for warm continental cumulus over the Indian sub-continent using CAIPEEX phase 1 data sets [Pandithurai et al., 2012].

[6] In this study, vertical variations of spectral width in premonsoon and monsoon clouds developing under different thermodynamic and aerosol environments over the Indo-Gangetic Plains (IGP) (21°45′ to 31°N latitudes and 74°15′ to 91°30′E longitudes) are discussed. During the premonsoon, clouds are super-continental microphysically, which means that the cloud droplets are very small and prevent any warm rain. During the CAIPEEX campaign, observations were performed at heights up to 7 km inside deep convective clouds over the foothills of the Himalayas. Such cloud microphysical observations of deep convective clouds over the Indian subcontinent and especially over the highly polluted IGP are carried out for the first time. The remainder of this article is organized as follows. Insection 2, data and methods are discussed; detailed descriptions of the cases considered in the present study are described in section 3. Results and their brief discussion are presented in Section 4. Section 5 summarizes the findings.

2. Data and Methods

[7] Data for the current study are from CAIPEEX (Phase-I conducted in 2009). Five research flights, each characterized by different aerosol and moisture conditions were analyzed for premonsoon (May 24; PRE1, May 28; PRE2) clouds over Pathankot (32.28°N, 75.65°E) and monsoon convective clouds (August 23; MON1, August 24; MON2, August 25; MON3) over Bareilly (28.22°N, 79.27°E). The observations over Pathankot were located close to the Himalayan slopes and the Bareilly observations were in the main Gangetic Valley.

[8] The cloud droplet probe (CDP) of Droplet Measurement Technologies (DMT) was used for the measurement of cloud droplet size distributions (DSD) in 30 bins between 2 and 50 μm. Liquid water content (LWC), droplet effective radius, CDNC were also derived from the CDP. Mean radius ( inline image), spectral width/standard deviation of DSD (σ), and relative dispersion (ε =  inline image/σ) were derived from CDP DSD. The standard deviation of these parameters at 100 m height intervals refers to spatial variation (aircraft may make several point measurements at any specific level). Temperature, relative humidity, and winds were measured by the Aircraft Integrated Meteorological Measurement System (AIMMS). CCN concentration (NCCN using DMT CCN counter) and aerosol concentration (using Passive Cavity Aerosol Spectrometer Probe; PCASP at 0.1–3 μm) were also measured. Aerosol data is considered only outside of clouds (CDNC < 10 cm−3) and are averaged for every 100 m level. Cloud microphysics data (from the CDP probe) used for the current study is at 1 Hz sampling frequency, i.e., over approximately 100 m of horizontal distance. Cloud microphysics data during ascent and descent through single deep convective clouds are used for the present analysis. In-cloud data is defined by CDNC > 10 cm−3. In the CCN measurements, set supersaturations (SS) of 0.2%, 0.4% and 0.6% are used. Each of these SS corresponds to a specific temperature gradient in the CCN column. During SS cycles with SS = 0.2%, 0.4% and 0.6%, the temperature of the column is changed to adjust for the new SS setting. We remove CCN data when the difference between the top of the column and the base of the column temperature changes by more than 0.1°C per second. This criterion makes sure that the CCN data that we report corresponds to droplets formed at a SS close to that set by the instrument. Little deviation from the set SS was noted and errors on the SS spectrum estimates are less than 5%.

3. Description of Cases

[9] CAIPEEX was conducted during the premonsoon, monsoon and the transition period between them. Figure 1a shows averaged CDNC, plotted against water vapor mixing ratio (r) in the mixed layer (the layer between 100 m below cloud base and 100 m above the ground), and aerosol number concentration (Na) in the mixed layer (ML). Each point in the figure corresponds to a CAIPEEX cloud observation and the flight tracks as indicted in Figure 1b. Three regimes based on the available water vapor in the boundary layer and progression of monsoon are identified (shown with ellipses in Figure 1a). This showed a very dry regime primarily in the premonsoon, a moderately wet transition and wet regime during the monsoon. The premonsoon, CDNC was higher than during the monsoon. The premonsoon regime was characterized by dry conditions and highly polluted conditions (Na > 1000 cm−3). Within the monsoon regime, the CDNC increases with increase in r. A few coastal cloud observations (indicated as 0707 and 0705) during monsoon are also shown in the figure. It may be noted that a few continental observations in the monsoon and premonsoon regime show relatively high droplet number concentrations. These observations were conducted in the Ganges valley (Figure 1b) during the premonsoon (May 24: PRE1 and May 28: PRE2) and active monsoon (Aug 23: MON1, Aug 24: MON2, Aug 25: MON3) periods. The high CDNC monsoon cases are associated with very moist and polluted conditions.

Figure 1.

(a) Variation of CDNC with boundary layer water vapor for cloud observations during CAIPEEX Phase I. Labels are marked with date of observation (mmdd; month and date) and grouped (shown with ellipses) to premonsoon, transition and monsoon period. Rectangular boxes show the observations used in this study, with two extreme conditions of water vapor and high pollution. The Nais shown with color and cloud averaged effective radius with size of the symbols as described in the legend. (b) The flight tracks over the Indian region and HYSPLIT back trajectories (at 3, 4 and 6 km levels) over the longitude-height cross section (c) at Pathankot and (d) over Bareilly. Latitude corresponding to trajectory is given in color map.

[10] The large-scale dynamical conditions under which these clouds developed are investigated with Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model ( 24 h trajectories based on the Global Data Assimilation System (GDAS) data products. It may be noted that HYSPLIT is based on GDAS meteorology input and is used as a tracer model. However, the representation of aerosol and chemistry effects is not part of this analysis. Figures 1c and 1dshow longitude-height cross-sections of 24 h back trajectories that are reaching over Pathankot and Bareily, respectively, at three vertical levels (3 km, 4 km and 6 km indicated as A, B, C in theFigures 1c and 1d). Color map indicates latitude of the trajectory location. In the PRE cases, high level back trajectories originated from northern regions that are source areas of dust and biomass burning. Six km level trajectories originated at higher levels and subsidence is noted over the location of aircraft observations. Four km level trajectories originated in the boundary layer, but trajectories again descended. The large-scale dynamical effects as illustrated by trajectories are not favorable for cloud formation in the PRE cases, while trajectories showed oscillatory nature, possibly attributed to the gravity waves and upslope flows due to diurnal heating. In the MON cases, subsidence is reduced and especially for MON3, the trajectories begin at very low level in the southeastern region and rise to higher elevations. In such a scenario, it is likely that boundary layer aerosols are transported to elevated layers in cloud.

[11] Aerosol vertical profiles from the PCASP onboard the CAIPEEX aircraft for five cases are shown in Figure 2a. This figure shows that Na is very high during premonsoon (exceeding 2000 cm−3) throughout the lowest 4 km. Cloud microphysical observations on PRE1 and PRE2 were carried out during premonsoon under extremely polluted conditions over Pathankot. Elevated pollution layers reaching 6 km are characteristic of these observations. Reasons for such elevated pollution layers are discussed in Prabha et al. [2012a]. The vertical profile of r derived from AIMMS probe (outside cloud) is shown in Figure 2b. The premonsoon cases showed less than half the boundary layer water vapor content compared to monsoon conditions. Additionally, the upper layer r in the premonsoon conditions was also much less. This means that clouds developed in the very polluted and dry ambient conditions in this region. Boundary layer (BL) Na for these two cases were 2500 cm−3 and 2700 cm−3, respectively (Table 1). Average relative humidity (in the boundary layer) was about 44% and 36%. Average r was about 7.2 gkg−1 and 4.7 gkg−1, respectively.

Figure 2.

Vertical profile of (a) Na and (b) water vapor.

Table 1. Boundary Layer Characteristics of Different Flights During CAIPEEX-2009 and Thermodynamic Characteristics of Different Flights From Radiosonde Observations
May 24 PRE1May 28 PRE2Aug. 23 MON1Aug. 24 MON2Aug. 25 MON3
Mixing Ratio (g/kg)7.2284.6919.3522.5720.25
Relative Humidity (%)44.2636.6872.4382.2567.59
Aerosol Concentration (cm−3)253827291689.542752764
NCCN (cm−3) at 0.4% SS4312685562980091718
Constant C (cm−3) from CCN spectrum (valid over the range 0.1 to 0.6% SS)13091526314543150644668
Slope of CCN spectrum (k)0.1450.610.540.580.57
Sounding CAPE (Joule kg−1)1407188331423101958
Precipitable Water (cm)22455
LCL height (hPa)679663854900845
Temperature (LCL) (°C)67222419
Cloud base height (m) From sounding, From aircraft3252, 34033438, 45001419, 1180988, 7351505, 650
TRMM (3B42.007) Rainfall (mm) within an area of 1° × 1° during the flight0.1061.80.00.5080.0

[12] Monsoon cases were cleaner and high Na is noticed primarily below 2 km. It is to be noted that Na decreased drastically above the boundary layer as the monsoon progressed. Average mixed layer relative humidity for these three cases were 72%, 82%, 68%, respectively, and corresponding water vapor mixing ratios were 19 gkg−1, 22 gkg−1, 20 gkg−1. For MON1 and MON2, clouds developed above a comparatively high polluted boundary layer. Average boundary layer Na for these two cases were 1689 cm−3 and 2752 cm−3, respectively. For MON3, average relative humidity in the boundary layer was 68%, mixing ratio was 20 gkg−1 and Na was 764 cm−3. Compared to the premonsoon cases with very dry boundary layer, the monsoon cases can be described as having a moist boundary layer and relatively low Na, depending on scavenging by rain.

[13] Thermodynamic characteristics of the cloud observations (Table 1) considered in this study were derived from radiosonde observations taken prior to the aircraft flights (at 0600 UTC; 1130 IST). The Convective Available Potential Energy (CAPE) for PRE1 and PRE2 calculated from the radiosonde observations were 1407 Joules kg−1 and 1883 Joules kg−1, respectively, and integrated precipitable water was 2 cm. Lifting condensation level (LCL) temperatures were 6°C and 7°C at pressure levels 679 hPa and 663 hPa for PRE1 and PRE2, respectively. Cloud base heights on PRE1 and PRE2 from aircraft observations were at the altitude of 3403 m, and 4500 m, respectively. Actual time (1130 LST) and location of radiosonde observations were 25–30 km away from the slopes compared to aircraft observations, which makes observed cloud base height different from that derived from radiosonde observations (3252 m and 3438 m).

[14] Radiosonde observations for MON1, MON2 and MON3 from Bareilly showed signatures of monsoon conditions. CAPE for these cases was 3142 Joules kg−1, 3101 Joules kg−1 and 958 Joules kg−1, respectively. Integrated precipitable water for these three cases was 4 cm, 5 cm and 5 cm, respectively. There was no change in the precipitable water for MON2 and MON3 and continuous advection of water vapor was present over the region. LCL temperatures were 22°C, 24°C and 19°C at pressure levels 854 hPa, 900 hPa and 845 hPa for MON1, MON2, and MON3, respectively. Cloud base heights for these cases are located at altitudes of 1180, 735, and 650 m, respectively.

4. Results

4.1. CCN Activation Spectra and Vertical Velocities

[15] NCCN measured in the sub cloud layer for four cases with the SS ranging from 0.1 to 0.6 SS are presented in Figure 3. The CCN spectra was found in the SS range of 0.2–0.6. ‘CCN observations’ are the NCCNmeasured in the subcloud layer at various SS. Note that in the log-log plot, curves with different values of k are straight lines. The interesting result is that the formula with k = constant leads to a good approximation of activity spectrum within the range 0.1–0.6% of SS without any tendency of the curve to saturate (i.e., k = const, and does not tend to zero to the end of the supersaturation range). This indicates that the aerosol spectrum contains small CCN. Fine aerosol particles can become CCN and activate at higher SS [Pruppacher and Klett, 1997]. The CCN spectrum (NCCN = CSk) shows similar C values for the PRE2, MON1 and MON2 cases. The PRE2 case has a slightly higher slope of the CCN spectrum (k) than the MON cases. The NCCNfor MON3 is one-third of the other three cases. All monsoon cases show k within a narrow range (0.54 to 0.57). High C and lower k lead to spectral broadening in single cloud parcels and internal mixing may cause spectral broadening for higher k as illustrated byHudson and Yum [1997]. Oscillations (alternating updraft and downdrafts) were observed in CAIPEEX monsoon cloud observations as well [Prabha et al., 2012b], but was not found to contribute to spectral broadening in PRE clouds.

Figure 3.

CCN spectra derived from the subcloud CCN cycles for respective cloud samples.

[16] The frequency distribution of vertical velocity observations below (in the sub cloud layer) and inside cloud is presented in Figure 4a. It may be noted that there were strong updrafts and downdrafts in the premonsoon case (especially PRE1).It may also be noted that trajectory analysis also showed large scale subsidence in the PRE cases, with more displacement of trajectories in the vertical. PRE cases showed more updrafts. PRE cases are also closer to the topographic barriers than the MON cases and may also be influenced by orographically forced localized circulation patterns with embedded strong vertical motions. CAPE is estimated from radiosonde observations prior to the flight (indicated in Table 1). The calculated value of maximum updraft using a relationship with CAPE ( inline image [Dutton, 1976] is much higher than the aircraft observations of updraft velocities. It is to be noted that the formula inline image can be used only for simplified estimation of maximum updrafts. In MON1, MON2 CAPE is two times larger, so were Wmax. An interesting question is whether high fluctuations of updrafts in PRE cases are caused by aerosol effects, specifically convective invigoration? Indeed, in the PRE we have larger amount of small droplets that may continue growing without sedimentation. The resulting latent heat may lead to increased updrafts in PRE. So, in spite of lower CAPE, and large scale subsidence as indicated by trajectories (Figure 1c), Wmax in PRE is higher because of the possibility of aerosol-induced convective invigoration.Figure 4a also show existence of strong downdrafts in convective clouds, but these downdrafts do not reach lower levels inside the cloud. This may indicate the possibility of internal oscillations in the cloud, which might lead to internal mixing and spectral broadening in polluted clouds as indicated by Hudson and Yum [1997].The proximity to the Himalayan range and thus orographic forcing can affect the intensity of convective scale updrafts and downdrafts. Vertical velocity fluctuations are also introduced as larger scale (>10 km wavelength) gravity wave effects in the presence of clouds [Prabha et al., 2012a]. It may also be noted that below cloud base vertical velocities in the PRE cases showed stronger updrafts and downdrafts.

Figure 4.

(a) Frequency distribution of in-cloud (thick lines) subcloud (symbols) vertical velocity and (b) averaged drop spectrum distribution (DSD) for 1000 m above cloud base for all five cases.

[17] Higher CCN and k in PRE1 along with strong updrafts may indicate a greater spectral broadening as found by [Hudson and Yum, 1997; Miles et al., 2000]. According to the parcel theory, increase in W should lead to decrease in droplet size in updrafts due to more activated droplets competing for condensate. As we see from Figure 4a, W at cloud base in PRE1 is much higher than in MON. So, in spite of similar NCCN, CDNC is higher and DSD is narrower in PRE (Figure 4b). In this case we see a combined dynamical and microphysical effect on DSD. Table 2 shows the average inline image, CDNC, ε, σ, updraft and downdraft at the cloud base (at 300 m above cloud base). Table 3shows these data for the non-precipitating part (using the criteria of effective radius <12 μm 300 m above cloud base). The CDNC at the cloud base of MON cases are less than half of that in the PRE cases. The inline image is slightly higher for MON cases. Here σ is nearly the same for all cases at the cloud base. However, in the nonprecipitating part of the cloud (Table 3) σ in MON cases are almost double that of the PRE cases, which may be due to rapid diffusional growth (like in a marine case where CDNC is lower but droplet spectra is wider). We observe high updrafts and weak downdrafts in PRE1 and MON3, while strong updrafts and downdrafts in PRE2, MON1 and MON2 in the nonprecipitating region. These three cases have high C values and are all polluted cases.

Table 2. Averaged Cloud Parameters for Cloud Base +300 m Abovea
Parameter24 May PRE128 May PRE223 August MON124 August MON225 August MON3
  • a

    CDNC is cloud droplet concentration, inline image is mean radius, σ is spectral width of cloud droplet spectra and ε is relative dispersion, mean updraft (w+) and downdraft (w−) with spatial spread.

CDNC (cm−3)639.9 ± 425.9853.79 ± 756.38297.48 ± 268269.79 ± 236.57258.08 ± 86.49
inline image (μm)3.24 ± 0.522.94 ± 0.73.6 ± 0.974.34 ± 1.014.09 ± 1.25
σ (μm)1.41 ± 0.21.5 ± 0.161.48 ± 0.201.61 ± 0.231.42 ±0.22
ϵ0.43 ± 0.040.52 ± 0.070.42 ± 0.050.38 ± 0.060.36 ± 0.07
w+ (ms−1)2.58 ± 2.162.10 ± 1.372.07 ± 1.491.21 ± 0.663.42 ± 2.17
w− (ms−1)−1.9 ± 1.33−0.58 ± 0.49−0.68 ± 0.46−0.29 ± 0.2−0.68 ± 0.37
Table 3. Averaged Cloud Parameters for Non-precipitating Part of Cloudsa
Parameter24 May PRE128 May PRE223 August MON124 August MON225 August MON3
  • a

    CDNC is cloud droplet concentration, inline image is mean radius, σ is spectral width of cloud droplet spectra and εis relative dispersion, mean updraft (w+) and downdraft (w−) with spatial spread. Data presented are averaged within the height of 300 m above cloud bases in non-precipitating part (i.e., inline image < 12 μm and T > 0°C).

CDNC (cm−3)481.7 ± 268.8476.98 ± 300.63258.13 ± 181.02326.16 ± 210.33117.85 ± 86.49
inline image (μm)5.06 ± 0 0.954.94 ± 0.886.37 ± 0.896.60 ± 1.5110.57 ± 1.0
σ (μm)1.53 ± 0.181.41 ± 0.142.49 ± 0.552.68 ± 0.793.23 ± 0.51
ϵ0.31 ± 0.0560.29 ± 0.0670.38 ± 0.050.4 ± 0.0560.30 ± 0.047
w+ (ms−1)2.29 ± 2.54.69 ± 2.563.63 ± 2.452.34 ± 1.922.96 ±1.78
w− (ms−1)−0.71 ± 0.46−3.8 ± 3.1−3.37 ± 3.01−2.79 ± 3.1−0.69 ± 0.49

[18] Average inline image (Table 3) for clean monsoon case (e.g., MON3) is about 11 μm to about 5 μm for polluted premonsoon (e.g., PRE1, PRE2) and 6.5 μm for polluted monsoon cases (e.g., MON1, MON2). In the case of polluted clouds, competition between large numbers of CCN for a given amount of water vapor produces droplets of smaller sizes. The CDNC is highest for the PRE cases and is only 118 cm−3 for MON3. The corresponding inline image and σ increased for this case with low CDNC. But in the MON1 and MON2, σ is higher than PRE2, and droplets have a higher inline image. CDNC is greater than in the MON3 clean monsoon case. These observations suggest that in spite of the presence of pollution aerosol, the drop spectrum broadens at a lower height in the MON cases. The increase in inline image also means that the mass in the smallest droplet diameter bins decreases and in the largest droplet diameter bins increases. This is illustrated with averaging of DSDs. The DSD averaged for regions within cloud 1 km above cloud base is presented in Figure 4b for comparison. It may be noted that PRE cases show narrow DSD with only one mode. Meanwhile all three monsoon cases show multiple modes in the DSD. The monsoon cases show an increase in concentration of large droplets with a decrease in pollution aerosol and an increase in moisture. The bimodal nature of DSD showed clear indications of small mode concentration peak below 20 μm. Such bimodal spectra were simulated in the presence of high relative humidity in the boundary layer by Segal et al. [2003]. Prabha et al. [2011] investigated details of bimodal spectra in monsoon clouds during CAIPEEX. It was reported that the small mode is associated with nucleation of small CCN which are not activated at cloud base and activate at some distance above cloud base where supersaturation in strong updrafts exceeds that at cloud base. CDNC increases with altitude in regions with strong updrafts. Small droplets were also found in downdrafts (at sub saturated regions) that are caused by partial droplet evaporation. Several theoretical studies have illustrated bimodal DSD due to various reasons. In the last section of this manuscript we investigate how the presence of small droplets (diameter < 20 μm) may change the inline image and σ.

4.2. Vertical Distribution of Cloud Microphysical Parameters

[19] The vertical variation of inline image (Figure 5a), σ (Figure 5b), ε (Figure 5c) droplet number concentration (Figure 5d) and LWC (Figure 5e) for all five cases are presented for heights above cloud base. inline image shows a slight variation with height, for PRE1 and PRE2 while for monsoon clouds there is large (2–12 μm) variation with height. σ is constant (≈1.5) with height in the premonsoon cases. In the monsoon cases, σ increases with height and reaches 5 μm at height above 5 km. The σ was similar in all monsoon cases irrespective of polluted or clean condition. Differences among the monsoon cases are mainly in the inline image with MON3 (clean case) showing higher inline image than MON1 and MON2 (polluted) at higher elevations (1 km above cloud base). Another important observation is the spatial variation (indicated by error bars calculated as the standard deviation of inline image and σ within 100 m vertical distance). In premonsoon cases, the standard deviation representing the spatial variation of inline imageand σ is less and it appears that the inline image and σ does not vary with height. If we use σ for a comparison between the cases, it is reasonable to use a constant value of 1.5 for premonsoon cases. For monsoon cases, σ increased with height. However, the relative dispersion in the polluted monsoon (MON1, MON2) increased with height. This is due to the fact that in the clean monsoon case (MON3), inline image was considerably higher than in polluted monsoon cases.

Figure 5.

Variation of (a) droplet inline image, (b) σ, (c) relative dispersion ε = σ/r, (d) droplet number concentration, and (e) LWC corresponding to five cloud samples.

[20] It is to be noted that for premonsoon conditions, the available water vapor is low, CCN are high and LWC is low (<0.5 g−3 for premonsoon compared to >1 g−3 for monsoon cases). It is also to be noted that premonsoon clouds are only 2 km deep compared to 7 km deep monsoon clouds. This difference is attributed to the amount of water vapor present. Saturation is reached at a higher altitude during dry conditions, compared to the moist monsoon conditions, which results in drastic differences in the lifting condensation level (LCL) (see Table 1). It may be noted that inline image in PRE cases were very low and σ did not change with height. This shows that there is very little droplet growth with height in these clouds as they are constantly entrained and diluted (as shown in the next section). In Prabha et al. [2011], premonsoon polluted clouds from the peninsular Indian region were analyzed and showed high LWC and droplet growth with height. This seems to be due to the presence of small droplets that do not precipitate, i.e., remaining in the cloud for further condensational growth. This feature was not seen in the PRE clouds analyzed in the present study. The subsidence noted for PRE cases also indicates that the large scale synoptic conditions did not favor cloud development.

[21] As these droplets grow by condensation, entrainment of dry non cloudy air (ambient conditions are drier for premonsoon; Figure 2b) from the surroundings causes depletion of LWC, reduction of CDNC [Warner, 1973], changing DSD, and a reduction in inline image from homogeneous mixing. Droplet partial evaporation is also possible in downdrafts, by which a decrease in inline image is possible. Thus small inline image is due to higher NCCN and then due to entrainment mixing and droplet evaporation. Entrainment and evaporation both contribute to the increase in ε. Freud et al. [2008] discussed robustness of effective radius at any specific level and also presented similar range of variations in the inline image as in the present study, however these changes in inline image can have variations in the relative dispersion in the clouds with constant σ. In monsoon clouds, typical values of σ vary from 1.5 μm at the cloud base to 5 μm at the cloud top (7 km above the cloud base). However, for polluted and dry premonsoon clouds σ appears to be constant with height. Large horizontal variability of σ in monsoon clouds compared to that of premonsoon cases is especially notable. Values of ε are about 0.2–0.45 (Figure 5c) similar to the continental stratus cloud observations reported in Miles et al. [2000], which are less than the range of values of ε observed during RICO (0.1–0.8) [Arabas et al., 2009] and higher than the ones reported for ACE-2 (0.2–0.4) [Pawlowska et al., 2006]. Politovich [1993] reports that for observations inside cumulus clouds, the dispersion was nearly constant with height, and was reported to be between 0.15 and 0.3. In the present study, ε increased with height in the polluted monsoon clouds, where σ also increases with height.

[22] Much higher CDNC are observed in premonsoon convective clouds (Figure 5d) compared to several of the continental and maritime stratus cases presented in Miles et al. [2000]and are comparable to deep convective clouds profiled during LBA-SMOCC (Large-Scale Biosphere-Atmosphere Experiment in Amazonia – Smoke, Aerosol, Clouds, Rainfall, and Climate) [Andreae et al., 2004; Freud et al., 2008]. It is also noted that CDNC increased with height especially from 2 to 4 km above cloud base. The question arises why there is increase in CDNC with height. Prabha et al. [2011]attributed this increase to in-cloud nucleation.Martins and Silva Dias [2009] indicated that relative dispersion in the LBA clouds varied between 0.4 and 0.6 as cloud water content increased. LWC in the premonsoon cloud is <1 g−3 while in the monsoon clouds a maximum of 3 g−3 was noted. The LWC varied significantly at any vertical level (Figure 5e), suggesting entrained or mixed cloud parcels.

4.3. Mixing and Entrainment

[23] Vertical profiles of inline image (Figure 6, top) and σ (Figure 6, bottom) are presented along with the adiabatic fraction (ADF) for all five cases. Important information from this representation is that droplets with largest inline image and σ have highest ADF, indicating less mixed samples. Cloud volumes with small inline image (<5 μm) show narrower spectra, compared to that of inline image > 5 μm. In the premonsoon conditions, ADF corresponding to drop inline image > 5 μm has σ between 1.2 and 1.8 μm. This indicates that in the premonsoon clouds, σ and inline imageare well correlated (correlation coefficient of 0.6) and both are maximum in the less diluted regions of the cloud. The correlation between inline imageand σ may also be due to dilution, where inline image and σ decreases.

Figure 6.

Vertical variations (height above mean sea level) of (top) droplet inline image and (bottom) σ of five cloud samples. The symbols are shaded based on the adiabatic fraction (%).

[24] For the monsoon cases (Figure 6), the above picture is somewhat different. inline image does not change significantly at any specific flight level in polluted monsoon cases, whereas in the clean monsoon (MON3) and in PRE cases, there is significant variation in inline image at any specific level (see Figure 6 (top) for all levels). inline image and σ are not always maximum when ADF is maximum. The correlation between inline image and σ is above 0.8 in these monsoon cases (larger that in PRE cases).

4.3.1. Relationship With CDNC in Diluted and Less Diluted Samples

[25] Data is classified by ADF greater than 0.5 and ADF between 0.1 and 0.5. ADF is the degree of cloud dilution due to the entrainment of free atmospheric air from the surroundings of the cloud. Cloud base and cloud top data are not used in this analysis.

[26] Figure 7 shows inline image (Figures 7a and 7b), σ of cloud droplet spectra (Figures 7c and 7d), LWC (Figures 7e and 7f) and relative dispersion (Figures 7g and 7h) as a function of droplet concentration CDNC. Averaged data for each 100 cm−3 CDNC bin are shown with error bars indicating the standard deviation within each bin. The right panels show less adiabatic (0.1 < ADF < 0.5) and left panels show more adiabatic (ADF > 0.5) cloud samples. This figure combines observations from 1000 m above cloud base to the altitudes below 6.0 km (to exclude data from the cloud base and the cloud top). For monsoon cases it is observed that inline image (Figure 7a) decreases as CDNC increases in both cases of dilution. It appears that inline image is much less dependent on CDNC in the premonsoon cases (PRE1 and PRE2) and it never exceeds 6 μm. For the less diluted monsoon cloud samples (Figure 7a), a strong dependence of inline image on CDNC is observed. Such negative correlation between CDNC and inline image was also noted in several earlier studies [Hudson and Svensson, 1995; Hudson and Li, 1995; Brenguier et al., 2000; Pawlowska and Brenguier, 2000].

Figure 7.

Results from five flights as a function of CDNC. First, second, third, and fourth rows show inline image, σ, and LWC, and relative dispersion (ε). Left and right panels are for less diluted (ADF > 0.5) and strongly diluted (0.1 < ADF < 0.5) cloud samples. The cloud base data and the lower 500 m of cloud data, cloud top data are screened out. Error bars indicate the variation within each concentration bin.

[27] Higher values of σ (Figures 7c and 7d) are observed in monsoon clouds (3–5 μm) as compared to premonsoon clouds (1–2 μm), irrespective of dilution. Monsoon conditions show gradual decrease in σ with increase in the CDNC for the less diluted parcels (Figure 7c). σis independent of CDNC for both diluted and less diluted premonsoon cloud samples. The near absence of collision-coalescence in premonsoon clouds keeps a constantσ. LWC increases with CDNC in the more diluted monsoon cloud parcels (Figure 7f). For the premonsoon clouds, LWC in the less diluted parcels are nearly independent of CDNC. However, in the less diluted monsoon cloud parcels, LWC increases with CDNC.

[28] ε shown in Figures 7g and 7h ranges from about 0.2 to 0.45. The overall pattern for ε seems to result from strong dependence of inline image on CDNC combined with weak dependence of σ on CDNC. There is a large spread in the relationship between ε and CDNC especially for the monsoon cases, depending on whether it is polluted or clean, i.e., ε is rather consistent with CDNC. It may be noted that for the premonsoon case, σ and inline image are nearly independent of CDNC. This leads to a nearly constant ε of 0.28 for premonsoon clouds, which is independent of CDNC. It may be noted that premonsoon cases are characterized by drier ambient conditions compared to that of monsoon cases. The drier air entrainment leads to a reduction in the CDNC and droplet evaporation at diluted regions of the cloud and a reduction in the LWC in all cases.

4.4. Spectral Broadening and Dispersion

[29] Some emphasis should be given to the identification of conditions under which σ increases mainly because cloud droplet spectral broadening is critical to precipitation formation. In Figure 8 a comparison of DSD (drop spectrum as a function of time inside the cloud) of the premonsoon cloud on 24 May (PRE2) and the monsoon cloud on 23 August (MON1) are shown. Height of observation and effective radius calculated from these measurements is presented in respective top panels. It may be noted that droplet spectra never show any sign of collision and coalescence in the premonsoon cloud (Figure 8a). This indeed is due to the second indirect effect [Albrecht, 1989; Hudson, 1993] where high NCCN leads to formation of small droplets with low collision efficiency. Also, numerous droplets in these clouds cause first indirect effect by increasing albedo. In the monsoon cloud (Figure 8b), we may note several instances with effective radius exceeding 10–12 μm. Some of these instances are also associated with multiple modes (shown with arrows in Figure 8b) due to the appearance of small (diameter < 20 μm) and large droplets in the cloud. A detailed bimodal DSD and associated cloud microphysical parameters for such a cloud pass 500–600 m away from the cloud edge is given in Figure 8c. It may be noted that strong updrafts are found to be associated with high CDNC (>400 cm−3) and high cloud LWC in these clouds [Prabha et al., 2011]. In small cumulus such as the ones observed in RICO [Gerber et al., 2008], the entrainment is observed to be taking place at much smaller spatial scales. The sampling frequency used in CAIPEEX was 1 Hz, which corresponds to a 100 m spatial resolution. Each cloud pass is over distances of 1.5–2.5 km across deep convective clouds.

Figure 8.

Drop spectrum distribution (DSD) in (a) a premonsoon cloud on 24 May and (b) monsoon cloud 23 August. Corresponding observations of height and effective radius are shown in the top panel. Color map indicates the number concentration per micrometer interval per cm3. Horizontal dashed line shows 24 μm limit. (c) Bimodal DSD from the highlighted region in Figure 5b. Legend show time of observation (UTC), height (m), CDNC (cm−3), LWC(gm−3), effective radius (μm), vertical velocity (ms−1), and adiabatic fraction corresponding to a region approximately 500–600 m away from the cloud edge.

[30] Drop spectral broadening is related to the activation/formation of droplets at the lower and higher diameter ranges of the DSD. Several theories have been discussed to explain the broadening of the DSD [Beard and Ochs, 1993; Martin et al., 1994; Hudson and Yum, 1997; Pinsky and Khain, 1997; Khain and Pinsky, 1997; Chaumat and Brenguier, 2001; Feingold and Chuang, 2002; Yum and Hudson 2005]. Spectral broadening can occur either by formation of small droplets (drop diameter < 20 μm) at the left part of DSD, or/and by generation of large droplets by efficient collisions at the right side of DSDs. Entrainment and mixing of relatively dry environmental air with the cloud [Paluch and Knight, 1984; Brenguier and Grabowski, 1993; Lasher-Trapp et al., 2005; Gerber et al., 2008] can also cause spectral broadening.

[31] Another possible mechanism of DSD broadening and formation of bi-modal and multimodal DSD is in-cloud nucleation [Prabha et al., 2011].

4.4.1. Droplets at the Tails of DSD

[32] To investigate the tail of droplet spectra, we first counted the 10 largest droplets at the tail of droplet spectra from 1 Hz data and determined the minimum diameter of the largest 10 droplets (D10). This diameter is used as a parameter characterizing the DSD tail at the higher end of the spectrum. Data considered for this analysis excludes cloud base and cloud top samples.

[33] Relationship between D10 and mean droplet diameter is presented in Figure 9 (left) for all the five cases as indicated. Adiabatic fraction is used to shade the symbols in Figure 9. It may be noted that high ADF parcels lie on the top of the diagonals and low ADF parcels on the lower part of the diagonal (Figure 9). Most important result from this analysis is that D10 is highest in the adiabatic cloud parcels. This means that the largest droplets form in adiabatic regions. This observation may support the entity type mixing proposed by Telford et al. [1984] but detailed investigation in this regard will be presented elsewhere [Prabha et al., 2012c]. In PRE cases, inline image is maximum when D10 is maximum. The decrease in inline imageand D10 in the less adiabatic parcels of PRE cases indicates the possibility of droplet partial evaporation or mixing with non-cloudy air parcels. To investigate this aspect, we have used vertical velocity observed along the flight track from the AIMMS instrument. D10, inline image, σ, ε are replotted by changing the color map to vertical velocity (Figure 10).

Figure 9.

Relationship between the (left) inline image, (middle) spectral width, (right) relative dispersion and minimum diameter of ten largest droplets at the tail of DSD. Cloud base and cloud top data was removed from this analysis. Color map is adiabatic fraction.

Figure 10.

Same as Figure 9 but for different vertical velocities in the color map.

[34] The decrease in inline image and D10 in the less adiabatic parcels of PRE cases occur concurrently with low ADF and somewhat stronger updrafts. Largest droplets (highest D10), high inline image and σ are noted for adiabatic regions with strong downdrafts (Figures 9 and 10). This may indicate that dry air is entrained into the cloud and CDNC decreases. However, in MON cases, adiabatic parcels are observed to have different ranges of D10 (Figure 9). For a fixed value of D10, inline image and σ are lower for adiabatic parcels as compared to less adiabatic parcels. Sometimes high D10 in adiabatic parcels are also observed with strong updrafts (Figure 10), corresponding to higher values of σ (indicated with ellipse in Figure 10 for MON1) and ε. Highly diluted parcels were not observed for D10 > 20 μm in MON1. It may be noted that this behavior is also associated with strong updrafts ≈10 ms−1 (Figure 10).

[35] Small slope between D10 and mean radius (<0.2 for PRE cases and >0.2 for MON cases) shows that at a given inline image, droplets in the tail are small, i.e., DSD is narrower in PRE. Figure 9 shows that in parcels with high ADF, the DSD contain a large tail. Among monsoon cases, MON3 is a clean and moist case compared to all other cases. MON3 is more ‘marine like’ (with comparatively higher droplet concentration than a typical maritime cloud). DSDs in parcels with high ADF in MON3 contain droplets as large as ≈20 μm in diameter. This supports the formation of raindrops at a lower height in the MON3 case. Another observation is that MON1 and MON2 cases have highly polluted upper layers compared to MON3. The entrained air into the cloud is polluted and some of the aerosol could act as CCNs at high supersaturations (above the cloud base leading to a decrease in inline image and increase in droplet number concentration in the updrafts).

[36] Another important result is noted in Figure 9 (middle), where D10 as a function of σis presented. The main result is that the relationship D10-σis not linear as that of D10- inline image. A relationship between D10 and ε is presented in Figure 9 (right). We see again that at a given relative dispersion, parcels with large ADF contain larger droplets while the ε is more variable in the diluted parcels.

4.4.2. Effect of Small Droplets on σ and inline image

[37] The number concentration of smallest droplets (<20 μm in diameter) and the inline image are related in Figure 11 (left). The adiabatic fraction is used as an indicator for mixing, which is shown with color scale. A similar representation with σ is shown in Figure 11 (middle) and with relative dispersion in Figure 11 (right) for all five cases. Figure 11 (left) show that inline image is maximum in adiabatic core regions. This inline image maximum is reached at high droplet concentrations with ADF maximum, corresponding to adiabatic core regions, except for a few data points in the polluted monsoon cases (MON1 and MON2). As dilution increases, the small droplet number concentration decreases. Dependencies in Figure 10 are the result of two effects: a) dilution, i.e., mechanical mixing of cloudy and droplet free volumes, leading to a decrease in the concentration of all droplets in the same proportion and b) evaporation of droplets during mixing. Evaporation can both decrease (total evaporation) and increase (partial evaporation of larger droplets) the concentration of small droplets. Since DSDs in PRE cases are narrow and contain only small droplets, dilution (likely in the updrafts) and evaporation (in the subsaturated downdrafts) lead to a marked decrease in inline image with decreasing ADF. In MON 1 and MON2 DSDs are wide and contain larger drops.

Figure 11.

Relation between the (left) inline image, (middle) spectral width, (right) relative dispersion with the smallest droplet (<20 μm) number concentration. The symbols are colored with adiabatic fraction.

[38] For a given ADF inline image in the adiabatic parcels decreases, as concentration of small droplets increases. This feature is especially pronounced in monsoon clouds. Such dependence of inline image on concentration of small droplets can be attributed to diffusional growth in ascending parcels. For instance, in adiabatic parcels increase in droplet concentration corresponds to decrease in drop size and in mean drop radius.

[39] The premonsoon cases show maximum radius corresponding to highly adiabatic parcels that have maximum small droplet number concentration. Numerous small droplet concentrations indicate that collision efficiencies are reduced and thus raindrop formation is delayed. A similar effect is also illustrated for σ with maximum σ in the more adiabatic parcels and small values of σ in less adiabatic parcels corresponding to a decrease in small droplet concentration (Figure 11, middle), which is less evident in MON2 and MON3. Relative dispersion in the PRE1, PRE2, and MON3 showed a slight increase with decreasing concentrations of small droplets (Figure 11, right). In the case of MON1, relative dispersion is independent of small droplet concentration. In addition relative dispersion is highly variable in premonsoon and monsoon clouds.

5. Discussion

[40] It appears that moisture and aerosol particle concentration for monsoon and premonsoon cloud samples have great impact on σ and inline image. General characteristics of our analysis of the entrainment effects on droplet growth is consistent with numerical studies such as Hill and Choularton [1986]. According to that study, three factors were important, humidity of the environmental air, NCCN and rate of entrainment. As a parcel ascends, SS initially increases. The droplets inside the parcel experience growth. CCN get activated and grow more rapidly, which then decreases supersaturation. This sequence of events may seem realistic in the PRE cloud observed, where strong vertical velocity is present in the diluted parcels (low ADF). Concentration of small droplets in these highly diluted parcels decreases. The decrease in mean radius and concentration may lead to further increase in supersaturation. Thus the sequence of events in these cloudy parcels may experience nucleation/evaporation in the regions of supersaturation/subsaturation, leading to a near constant σ and slight variations in inline image.

[41] It is observed that in polluted premonsoon cases σ does not change with height, however mean radii show a slight increase with altitude. The detailed analysis based on adiabatic fraction has shown a reduction in σ and inline image corresponding to a decrease in small droplet concentrations. For polluted, moist monsoon cloud samples inline image is less than that of clean monsoon cloud samples. In order to parameterize the width of the droplet spectra, it seems that a reasonable approach is to assume a constant σ in the case of premonsoon cloud samples. For monsoon cloud samples it is observed that there is an increase in σ with height. σ was nearly the same in the clean or polluted monsoon cloud samples. In monsoon clouds typical values of σ vary from 1 μm at the cloud base to 5 μm at the cloud top.

[42] The theoretical predictions of DSD evolution in an ascending adiabatic cloud parcel show that ε in clouds with higher concentration of droplets decreases with altitude faster than ε for lower droplet concentration [Pinsky et al., 2012; A. P. Khain and M. B. Pinsky, On the theory of cloud droplet diffusion growth. Part 1: Monodisperse spectra, submitted to Journal of the Atmospheric Sciences, 2012]. Thus observations of constant dispersion or increase of the dispersion with height is consistent with Khain and Pinsky [2012]. However, a continuous increase with height of σ in monsoon cloud is associated with the presence small droplets in less diluted parts of clouds (in all three cloud samples presented) needs further investigation.

[43] In a review article on the droplet growth in warm clouds, Devenish et al. [2012] argues that entrainment and mixing has a very important role in the evolution of droplet spectra. In a recent study, Bewley and Lasher-Trapp [2011] illustrate that variability in the droplet growth histories could result primarily from entrainment and could explain observed droplet width in small cumuli of RICO. They considered the inhomogeneous droplet evaporation in a numerical model. Relevance of such aspects in monsoon clouds are yet to be investigated. Our observations of premonsoon clouds as illustrated above indicate that entrainment and mixing are important in reducing the small droplet number concentration and thus changing both inline image and σ. inline image changes through two processes, by total or partial evaporation of droplets and by mixing of cloud free air with cloudy air. The diluted parcels with low ADF contain a reduced number concentration of small droplets, small inline image and σ. In adiabatic parcels higher small droplet number concentration was found. In monsoon clouds this picture is more complex as the inline image or σ may not always change by reduction of small droplet concentration. This is possibly due to stronger mixing in monsoon clouds than typically noted in small cumulus, where the cloud scale motions may also be important.

[44] The mixing in premonsoon clouds is similar to the homogeneous type of mixing described by Paluch and Knight [1984], where evaporation of droplets take place as dry and polluted outside air mixes with saturated cloudy air parcels. The vertical variation of inline image and σ did not show any drastic increase in CDNC like in the monsoon case. The monsoon cases are characterized by the continuous presence of small droplets in the mixed regions. Partial or total evaporation of droplets might occur, resulting in the reduction in size and number concentration of droplets.

6. Main Conclusions

[45] CAIPEEX data provide unique observations of deep convective clouds up to 6 km above cloud base. Such observations are rare over the Indian subcontinent, especially in highly polluted and hydrologically vulnerable locations such as the Indo Gangetic Valley. Cloud droplet spectral dispersion in the Gangetic valley region for highly polluted premonsoon and polluted and clean monsoon clouds were investigated. The premonsoon clouds developed over a dry boundary layer with cloud base height above 3 km, compared to the moist and less polluted monsoon boundary layer where cloud base height was 700–2500 m. Characteristics of the mean cloud microphysical parameters revealed the following.

[46] σ in the premonsoon clouds is independent of the total droplet number concentration and inline image increases only slightly with height. This is attributed to the aerosol loading; which has a profound influence on the decrease in drop spectrum dispersion with height. The decrease in inline image at specific levels in the diluted parcels of PRE cases leads to changes in relative dispersion. This relationship is rather different in monsoon clouds, where σ is nearly the same for polluted and clean cloud samples and is two times that of the PRE case clouds. inline image in the clean monsoon cloud is two times that of a polluted monsoon cloud, and three times that of the polluted PRE clouds. This leads to a higher relative dispersion (0.4–0.45) for polluted monsoon cloud, while making the relative dispersion (ε) in the clean monsoon cloud comparable to that of PRE cloud sample (≈0.3). The inline image and σ decreases, concurrently with increase in LWC and droplet number concentration (CDNC) in polluted monsoon cloud. Thus it can be concluded that the overall pattern of ε results from a strong dependence of inline image on CDNC and a weak dependence of σ on CDNC.

[47] During less diluted conditions, ε in a monsoon cloud depends mainly on σ and less on inline image. Polluted monsoon clouds may be considered marine like with high water vapor, while high Na or NCCN are characteristic of continental conditions. The dispersion in these clouds thus have intermediate characteristics compared to that of marine and continental cloud parcels.

[48] Adiabatic fraction is used as an indicator of mixing and to investigate droplet spectrum broadening. The number concentration of small droplets (with diameter < 20 μm) is considered to investigate the sensitivity due to droplet evaporation verses fresh nucleation or partial evaporation of large drops. Large droplet regions of the spectrum were investigated by finding the maximum diameter corresponding to the ten largest droplets (D10) at the tail of the droplet spectrum.

[49] In majority of the cases studied (except for a few data points in the clean monsoon case), inline image and σ are maximum in the less diluted parts of the cloud where number concentration of small droplets (with diameter < 20 μm) was maximum.

[50] A decrease in inline image corresponding to a decrease in the number concentration of small droplets suggests that droplets may be subjected to either partial or total evaporation as illustrated. In the diluted part of the cloud sample, either partial/total evaporation or mixing of cloud free air may be important to reduce the number concentration of small droplets.

[51] The adiabatic parcels are characterized by a stronger relationship between the mean diameter and droplet growth (D10) at the tail of DSD than that of less adiabatic parcels. In a monsoon cloud, less adiabatic parcels tend to have a higher inline image and σ, for constant D10. Observational evidence is provided for the formation of large droplets in the adiabatic parcels, that are characterized by maximum σ and inline image, and high concentration of small droplets of diameter <20 μm.


[52] The CAIPEEX project and IITM are fully funded by Ministry of Earth Sciences, Government of India, and New Delhi. A. Khain is supported by the Office of Science (BER), grant DE-SC0006788, U.S. Department of Energy. Authors acknowledge with gratitude that several scientists at IITM; their team effort and dedication made CAIPEEX a grand success. Three anonymous reviewers are thanked for their dedicated help with the reviews of this manuscript. We are grateful to Editor Steven Ghan for detailed synthesis of reviews and suggestions.