In-situ observations of cloud microphysics over Arabian Sea during dust transport events

The unique in situ measurements of clouds and precipitation within the shallow and deep cumulus over the north-eastern Arabian Sea region during the Indian monsoon are illustrated in this study with a focus on droplet spectral parameters. The observational period showed a significant incursion of Arabian dust and the presence of giant cloud condensation nuclei (GCCN), modifying the cloud and precipitation spectral properties. Warm rain microphysics supported the mixed-phase development in these clouds and exhibited hydrometeors of snow, graupel and large aggregates as part of ice processes. Cloud base droplet number concentration is about 142 ± 79 cm−3 which is one third of the cloud condensation nuclei (CCN) number concentration at 0.2% supersaturation. A rapid broadening of droplet size distribution (DSD) near to the cloud base was noted in contrast to polluted continental clouds. Relationship between the relative dispersion ( ε; the ratio of DSD spectral width ( σ ) to mean radius ( rm )) and liquid water adiabatic fraction (AF) indicates that the entrainment effect has increased relative dispersion significantly (2–3 times larger) in these clouds. Effective radius ( reff ) is found to be proportional to mean volume radius ( rv ) with a proportionality constant ( β ) that varies between 1.0–1.6, depending on the spectral dispersion parameter. Drop size distributions for the small cloud droplets with size range 2–50 μ m and the large drizzle drops (or ice hydrometeors) with size range 100–6400 μ m are parameterized using the gamma function distributions useful for large-scale cloud models.


Introduction
Clouds are an important component of the global climate due to their strong influence on radiative forcing and contribution to the hydrological cycle (Ramanathan et al 1989, Browning and Gurney 1999, Chen et al 2000).Observations of aerosol and cloud microphysics using in situ measurements are necessary for studying aerosolcloud-radiation and aerosol-cloud-precipitation interactions and cloud microphysical processes which are generally parameterized in numerical weather prediction (NWP) and climate models (Golaz et al 2011, Formenti et al 2019, Gordon et al 2020, Haywood et al 2021).Starting from the aerosols activation as tiny droplets to the final formation of large raindrops, it passes through various complex mechanisms inside the clouds (Grabowski and Wang 2013, Khain et al 2013, Oh et al 2023).
Aerosol particles that are activated to droplets in a supersaturated water vapour condition are the cloud condensation nuclei (CCN), which have high affinity to water vapour (Hudson 1993, Patel and Jiang 2021, Jayachandran et al 2022).Aerosol number size distribution also plays an important role in the activation mechanism (Romakkaniemi et al 2012, Shika et al 2020).Post-activation droplet growth above the cloud base height is achieved by the condensation process of water vapour inside the clouds (Mason and Chien 1962, Paluch 1971, Vaillancourt et al 2001).The growth rate of droplets can be largely different under marine and continental conditions due to changes in droplet number concentration and moisture availability (Braham 1968, Martucci andO'Dowd 2011).Marine environments generally produce higher growth rate (i.e.larger droplet size) due to low droplet number concentration which favours higher moisture availability per droplet (Yum and Hudson 2002).However, precipitation formation in clouds cannot be achieved by the condensation growth process alone, due to its slower growth rate at larger droplet size (Devenish et al 2012).For this reason, collision-coalescence processes need to be active, which are strongly determined by the droplet size distribution (DSD) properties of the clouds (Pinsky et al 2001, Wang andGrabowski 2009).
A wide DSD is conducive to the early onset of precipitation by the collision-coalescence processes.Various mechanisms control the DSD characteristics of clouds.For example, cloud turbulence by inducing supersaturation fluctuation plays a vital role in the DSD spectral broadening effect for a fixed CCN environment (Shaw et al 1998, Chandrakar et al 2016).Entrainment-mixing of cloud-free air also influences the DSD spectral characteristics and rain initiation process (Lasher-Trapp et al 2005, Cooper et al 2013, Bera et al 2016, Mellado 2017).There can be a significant difference in DSD characteristics between the continental and marine clouds due to changes in aerosol properties (Braham 1968, Hudson and Yum 1997, 2001).Giant cloud condensation nuclei (GCCN) also favour early onset of precipitation in marine clouds by producing wider DSD leading to early collision-coalescence (Blyth et al 2003, Jensen andNugent 2017).Instead of such significance of the DSD characteristics, most of the large-scale models (where bulk microphysics schemes are used) do not specify DSD accurately and some functional forms (e.g.gamma function distribution or exponential distribution) are assumed to represent the DSD (Morrison and Grabowski 2007, Heymsfield et al 2013, Zhang et al 2023).Therefore, a detailed understanding of these properties is helpful to formulate them precisely in numerical models and to decrease the associated model uncertainty.
The primarily source of aerosols over the ocean are mainly sea-salt (e.g.NaCl) by bubble-bursting mechanism and some biogenic emissions (O'Dowd et al 2004).Long-range transport of aerosols by wind flow from the continental regions (i.e.dust particles, black carbon and organics) can also contribute to aerosol accumulation over the oceanic region (Wang et al 2020).Aerosol impact on marine clouds strongly depends on the physico-chemical properties of aerosol and regional meteorological conditions (Chen et al 2014, Feingold et al 2016).Aerosols over the oceanic region are important for the Earth's radiation balance through scattering and absorption of solar radiation (Ramanathan et al 1989, Nazarenko et al 2017, Jose et al 2020).They also impact by the indirect effects, acting as CCN or IN which modifies cloud microstructure and precipitation efficiency (Fan et al 2016).Thus, the interactions between aerosol-cloud-precipitation over the Arabian Sea (AS) regions are crucial to the global climate and regional water cycle.
The Arabian Sea has a unique geographical location and can control the climate over this region as well as the global climate.Indian summer monsoon (ISM) rainfall, which is the life-line for millions of people, is largely controlled by the moisture outflow and aerosol transport from the Arabian Sea (Nandini et al 2022).Large-scale wind circulation over the Arabian Sea changes from strong south-westerly during the summer monsoon (June-September) to moderate north-easterly in winter season (December-January), which is a consequence of the migration of Inter Tropical Convergence Zone (ITCZ).A major amount of moisture is transported over the ISM region through the south-westerly part of circulation over the AS.This wind flow pattern during the monsoon season enhances the dust aerosol concentration and impacts the cloud microphysics and precipitation over the ISM region.The Arabian Sea also serves as a genesis bed for several low-pressure systems (or tropical cyclones) which move to the surrounding continents and contribute significantly to total precipitation.This is also important for the rainfall activity over the Arabian Peninsula and the adjacent countries of north-western India (Horan et al 2023).Thus, AS has a great region of interest for studying the aerosol-climate interactions, aerosolcloud-precipitation interactions, and rainfall redistribution by low pressure systems (McFarquhar and Heymsfield 2001, Ramanathan et al 2002, Gogoi et al 2019, Wahiduzzaman et al 2022).However, in situ observation of aerosol and clouds over the AS region is largely lacking and very sparsely available that limits model accuracy due to incorrect representation.This study presents one of such rarely available dataset that contains in situ measurements of microphysics of aerosol, clouds, and precipitation inside clouds over the AS region.
A few field experiments have been conducted over the AS region such as the Indian Ocean Experiment: INDOEX during January-March, 1999 (Heymsfield andMcFarquhar 2001, Ramanathan et al 2002), Arabian Sea Monsoon Experiment: ARMEX during June-July, 2002 (Rao 2005), Integrated campaign for Aerosols, gases and Radiation Budget (ICARB) campaign during March-May, 2006(Kumar et al 2008) in order to characterize the aerosol and cloud properties over the oceanic region.Studies from these campaign-based observations concluded that dust is a dominant aerosol type, particularly in summer monsoon season, over the AS which is mostly coming from the Arabian Peninsula, Horn of Africa and Thar desert (Tiwari et al 2023).Measurements of aerosol size distribution over south-east AS showed a bimodal size distribution with nucleation and coarse modes during the south-west monsoon season (Kesti et al 2020).Using ARMEX observation over AS, Murugavel et al (2005) found that the aerosol size distribution at the top of the mixing layer is bimodal or trimodal (Aitken, accumulation and coarse mode).Rao et al (2005) using ARMEX data found bimodal aerosol size distribution with a dominant coarse mode during the monsoon season.
The most recent observational campaign (Cloud-Aerosol Interaction and Precipitation Enhancement Experiment: CAIPEEX) conducted over the Indian land and surrounding oceanic regions, Arabian Sea and Bayof-Bengal (BoB) during 2009-2015 (Kulkarni et al 2012, Konwar et al 2012a, 2012b, Deshpande et al 2014).Aircraft based measurements of growing convective clouds over the AS showed that cloud DSDs were much wider than that observed over the polluted Indo-Gangetic plain region (Konwar et al 2012b).Dust particles during transportation across the moisture rich marine environment absorb the water vapour onto the surface and act as GCCN which increases cloud droplet number concentration (CDNC) at the cloud base and enhance the possibilities of warm rain initiation despite the moderate presence of smaller CCN (Konwar et al 2012a, Deshpande et al 2014).Sathiyamoorthy et al (2013) studied cloud properties over AS using satellite based observations and found that shallow clouds are predominant during the summer monsoon season (June-September) and these clouds are unable to grow vertically due to the presence of lower tropospheric thermal inversion.Harikishan et al (2015) studies aerosol and cloud properties using satellite and reanalysis data over the AS and reported that an elevated aerosol layer exists in 2-5 km height during the monsoon break time and they also noted that dust aerosol has profound impact on cold clouds (ice-phase).Konwar et al (2012b) compared the DSDs between the continental India, AS and BoB regions and found that BoB clouds are more pristine than AS.They showed that cloud base maximum droplet number concentration in BoB is around 100 cm −3 where it is around 400-500 cm −3 in AS.McFarquhar and Heymsfield (2001) attempted to parameterize cloud microphysics using INDOEX data over the AS region for the purpose of utilization in climate models.They found that effective radius (r eff ) is proportional to the cube root of liquid water content (LWC) divided by total droplet number concentration (N d ) and the proportionality constant depends on the skewness and dispersion of DSDs.This was the only available study over the AS regions that focused on cloud microphysical parameterization using in situ measurements within clouds.
Observations from high spatial and temporal resolution airborne measurements over AS are very sparse, limiting their use in assessing the performance of cloud modelling in NWP and climate models.Most of the research studies conducted over the AS region were focusing mainly on aerosol-radiative impacts (Rajeev and Ramanathan 2002, Vinoj and Satheesh 2004, Chylek et al 2006, Jose et al 2020, Tiwari et al 2023) with satellite data, ship-based observations and numerical modelling.Studies related to cloud microphysics, especially ice and mixed-phase properties of clouds over this region, are largely lacking.The properties of mixed-phase clouds are not well understood in general and remain as an active research area.The physics of mixed-phase clouds is very important for climate (Fowler and Randall 1996, Choi et al 2010, Morrison et al 2020).However, their representation in numerical models remained poor (Klein et al 2009, Barret 2012).Therefore, studying mixedphase cloud properties over the AS region is very important.The objective of this work is to present in situ observations of aerosol and cloud microphysics (both warm and mixed-phase) over AS and parameterize some microphysical variables to be useful for large-scale models.
The subsequent sections are organized as follows.Section 2.1 describes the experiment, observation dataset and various instruments used in data acquisition.Data methods section (2.2) gives detail about the analysis methodologies used in this study.Section 3 contains the main results of the various data analyses and related discussions.Section 4 presents the summary and conclusions of the study.

Experiment and instrumentation
In situ observations presented here were collected during the CAIPEEX field experiment (phase-III) in the year 2015.More details about the CAIPEEX experiment and dataset can be found in many previous papers (Kulkarni et al 2012;Konwar et al 2012a;Bera et al 2016;Prabhakaran and Coauthors 2023).A research aircraft (Beechcraft B200) was operated from Kolhapur airport (16.66 °N, 74.28 °E) which is close to the west-coast of peninsular India (See the geographical map in supplement figure A1).The aircraft was equipped with several on board instruments for measuring aerosol, cloud, precipitation, and meteorological parameters.The observations mainly conducted during late morning and afternoon times when local convections peak their development.The observation period was 04-27th July (with a total of 25 flights observation) which is a rainy-monsoon (or south-west monsoon) season over India.Three dedicated flights (on 11, 23, 26 July) were conducted over the Arabian Sea region during this field experiment phase with a major objective to collect observations of marine aerosol and clouds.CAIPEEX research aircraft mostly targeted growing cumulus convection or embeddedconvection.Over the AS region, research aircraft probed mainly cumulus clouds which are embedded in stratus decks.These marine clouds generally propagate towards Western Ghats (WG) mountain ranges and precipitate out over the wind-ward side or introduce stratiform clouds over the leeward side of the WG.Therefore, these clouds are important for rainfall activity over the WG which receives one of the highest rainfalls in the Indian peninsula during the monsoon season.
As mentioned above, several instruments were used on board a research aircraft.A passive cavity aerosol spectrometer probe (PCASP) was used for measuring aerosol particle's number size distribution with size range 0.1 to 3.0 mm (Kleinman et al 2012).Cloud droplet probe (CDP) was used for measuring cloud droplet size spectra, number concentration and liquid water content for droplets with size range 2 to 50 mm (Lance et al 2010).Precipitation imaging probe (PIP) measured larger rain drops and ice particles with size range 100-6200 mm (Baumgardner et al 2001).Aircraft integrated meteorological measurement system (AIMMS-20) probe used for measuring air temperature and humidity, height, GPS coordinate, and 3D wind components (Beswick et al 2008).A continuous flow thermal gradient CCN counter was used for measuring the number of activated aerosol particles at varying supersaturation between 0.2%-1.2%(Roberts and Nenes 2005).All observations were conducted at a 1 Hz frequency that reveals a spatial resolution of about 100 m.More details of these instruments can be found in several previous studies from CAIPEEX (Konwar et al 2012a, Varghese et al 2016, Maheskumar et al 2018, Bera et al 2019, 2022) and we have also listed them here in table 1.All instruments were factory calibrated and validated before starting the CAIPEEX experiment campaign and also there was onsite periodic calibration to make data quality at high level.The data presented here had gone through data quality control process which includes examination of data after each flight and removal of error data or artifacts and correcting data for known instrument biases.All dataset are pre-processed before using in research study.
The Hybrid Single-Particle Lagrangian Integrated (HYSPLIT) trajectory model is used to identify the air-mass origin (backward trajectory up to 48 h) over the observation location.This helps to understand the source of aerosols which are transported to the observation location.Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) data over the AS region is used to understand some of aerosol and cloud properties.Aerosol sub-types (e.g.marine or continental type) information can be used to validate aerosol source over the location.Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2) model hourly-averaged data is used for dust column mass density and aerosol optical depth (total aerosol extinction AOD at 550 nm) spatial distribution over the AS region.Vertical profile of environmental RH was drawn using ECMWF Reanalysis v5 (ERA5) data which is due to the fact that the aircraft humidity sensor in AIIMS probe was not functioning properly at higher altitudes during these flights.Wind vector spatial pattern is drawn using ERA5 wind data.Data source locations of HYSPLIT, CALIPSO, MERRA-2, and ERA5 are mentioned at the end of the paper.

Data methods
For cloud pass observations, in-cloud samples are selected by considering a measurement threshold of CDP droplet number concentration N d > 10 cm −3 and liquid water content LWC > 0.001 g m −3 which are similar to many previous studies (Bera et al 2016, 2019, Konwar et al 2012a, Varghese et al 2023).Aerosol and CCN observations are presented for cloud-free conditions (N d < 10 cm −3 and LWC < 0.001 g m −3 ) only.
One of the main purposes of this study is to understand the clouds DSD spectra over the AS and provide a parameterization for that which can be useful for large-scale numerical models.Most of the bulk microphysics schemes available for numerical models use a parameterized DSD (such as gamma function distribution) for droplet growth, ice microphysics and rain formation related processes (Morrison and Grabowski 2007, Heymsfield et al 2013, Patade et al 2015, Bera 2021).Therefore, it is important to provide such DSD parameterization using this sparsely available dataset over a remote location which can be utilized in numerical models for improving cloud representation.For this purpose, we have attempted to parameterize DSD using gamma function distribution as given below.
Where D is the droplet diameter, N 0 is the intercept parameter, λ is the slope parameter and m is the shape parameter which is related to the dispersion of the DSD spectra.The functional parameters N 0 , m and λ can be provided for both cloud droplets (2-50 mm) using CDP data and for drizzle or rain drops (100-6200 mm) using PIP data.
We have used two different methods to estimate the gamma function parameters using CDP and PIP measurements.Method_1 assumes a large range of each parameter and then iterates computationally using a FORTRAN program to calculate the root-mean-square (RMS) error with respect to the observed DSD.The best fitting parameters of gamma distribution were obtained by minimizing the RMS error in modelled and observed size distribution.The least error gamma function is assumed as the best fitted DSD.
The second method assumes, m = e - 2 where e is the relative dispersion of the observed DSD (following Morrison and Grabowski 2007).The other two parameters (N 0 and l) are estimated using moments method derived from the observed DSD (see equation ( 5) and (6) in Patade et al 2015).The first method applies to both CDP and PIP derived DSD but the second method applies to CDP derived DSD only as its assumption m = e - 1 1 2 ( ) may not be valid for precipitation size drops.
In cloud microphysics parameterization, relative dispersion (e) and effective radius (r eff ) are two important parameters for radiative transfer models (Brenguier et al 2000, Liu et al 2008).r eff is generally calculated using the ratio between the third and second moment of the cloud droplet size distribution measured by CDP.Whereas, where AF is adiabatic fraction, the ratio of liquid water content (LWC) to adiabatic liquid water content (LWC ad ).These relationships will be useful for cloud microphysics parameterization.Figure 2 shows the NSD with mode diameters around 0.14 mm (accumulation mode) and 0.8 mm (coarse mode).It is also notable that D p > 1 mm particle (supermicron particle) number concentration is higher for the 11 July case.Accumulation mode particle number concentration is also slightly higher for the 11 July case which may lead to higher CCN concentration.Activation size spectra of CCN show a bi-modal type distribution for the 11 July case with a higher number of larger activated droplets.This can be inferred that high dust loading on 11 July causes a distinct feature on aerosol NSD and CCN activation spectrum.Supermicron dust particles produce larger size droplets in the activation spectrum of CCN.This distinct feature of CCN activation spectra diminishes at higher supersaturation (0.4% and 0.6%, results not shown) as most of the larger dust particles get activated at 0.2% SS.

Results and discussions
The presence of giant CCN in these measurements can further be illustrated by investigating CCN-N d and CCN  -r eff scatter plots which are shown in supplement figure A3.With increasing cloud base CCN number concentration (for different flights) does not lead much increase in above cloud base (300 m) N d between the cases.Most of CCN at 0.2% SS got activated to droplets at this lowest SS, indicates that large CCN were indeed present at the cloud base.This indeed facilitated the broadening of DSDs through the tail effect.Additionally, droplet effective radius at 300 m above cloud base is quite high compared to continental clouds reported in previous studies (Prabha et al 2011, Konwar et al 2012a).It is further found that r eff is higher for 11 July case (RF08) which is largely loaded by dust aerosols.This may infer that larger dust particles may act as GCCN under marine environment and produce larger droplets and wider DSDs.The impact of Arabian dust over the AS may be important for the regional climate also which is not explored in this study.The entire north Arabian Sea is significantly loaded with dust aerosols during the south-west monsoon season.These dust aerosols can act as GCCN over the marine environment due to higher moisture availability and potentially impact the cloud microphysical structure by widening the DSDs and increasing effective radius near to the cloud base.These microphysical changes directly impact the radiation properties of the clouds and thereby impact the regional climate.The dust aerosols over AS can act as GCCN and support the warm rain process at lower altitude.However, in deeper clouds, these dust aerosols can also act as ice-nucleating (IN) particles and help to form larger ice hydrometeors which may lead to heavy precipitation.This particular mechanism can be important for the mixed-phase rainfall activity over the Western Ghats regions where these transported dust aerosols can act as IN during the monsoon.
Detailed observational features of clouds on these three flights are presented in figure 3. Panels (a)-(c) show CDP measured cloud LWC vertical variation.T and RH variations with altitude are also complemented for understanding the thermodynamic constraints on cloud development over the AS.A clear temperature inversion at an altitude of 2.5 km is seen for 11 July (RF08) flight.Environmental RH decreases rapidly near to this inversion layer which poses restriction for cloud growth above this layer.So, it produced shallow clouds (depth @ 1 km) with maximum observed LWC @ 1.25 g m −3 .Flight observations on 23 and 26 July show deeper clouds with cloud depth » 3.5 and 7.5 km respectively.RH was higher over the entire altitude for 26 July flight which encountered deepest cloud with maximum observed LWC @ 2.1 g m −3 .
Altitudinal variation of aerosol and CCN number concentration (at 0.6% SS) is shown in figure 3(d).Supermicron aerosols (D p > 1 mm) number concentration variation is shown in figure 3(e).Variation of aerosol effective radius with height is shown in figure 3(f).Aerosol (total) and CCN number concentration's altitudinal variations show high values up to 4 km height for all three flights.Similar features observed for supermicron aerosol (coarse size aerosol) concentration and aerosol effective radius.It can be distinctly noted that 11 July flight (RF08) observation shows higher values of aerosol, CCN number concentrations, supermicron aerosol concentration, and aerosol effective radius up to this 4 km height level.This infers that dust aerosols dominate up to this level (in concurrence with the CALIPSO observation).CCN-supersaturation (SS) spectrum is shown in figure 3 ) parameters k and C 0 are also provided (Twomey 1959).The lowest value of k on 11 July indicates more like continental dust aerosols with larger particle size.Cloud droplet number concentration (N d ) and effective radius (r eff ) variations with height are shown in figures 3(h) and (i), respectively.Highest droplet number concentration is around 220 cm −3 which is observed slightly above (300-400 m) the cloud base height.This type of feature is generally found in continental monsoon clouds where above cloud base activation (or secondary activation) takes place (Prabha et al 2011, Bera et al 2016).Effective radius increases above cloud base and reaches a maximum (@ 16 mm) near 4 km altitude which is also the level where aerosol effective radius is maximum for 23 and 26 July flights.The height profile of droplet effective radius also indicates that drizzle formation in these clouds happened at lower altitude (@ 3 km) where it crosses 12 mm threshold (Konwar et al 2012a, Khain et al 2013).

Droplet size distribution (DSD) properties
A detailed understanding of DSD properties of these marine clouds are provided in figures 4(a)-(c).This gives contour plots of DSDs varying with cloud sample numbers at 1 Hz rate.A concurrent variation of altitude (red line) and effective droplet diameter (D eff , black dots) are also provided in the figure.The size of black dots is scaled (in a relative manner) with the DSD spectral width (s).The horizontal black line (dashed) on the figure indicates 25 mm diameter which can be considered as threshold effective diameter for rain initiation.It can be seen that DSDs are wide with multiple modes and width of DSD increases with increasing height.D eff never crossed 25 mm threshold that means no rain formation happened for 11 July (RF08) clouds.For other two flights on 23 and 26 July, D eff crossed the 25 mm threshold and rain drops formed in those clouds.The rain drop formation level is associated with bi-modal type and wider DSDs.Above the rain formation level, droplet number concentration decreases rapidly as most of the larger drops fall down from the clouds.
The lower panels (d-f) of figure 4 show height variation of (mean and standard deviation) cloud water content, DSD spectral width (s) and relative dispersion (e) parameters.Here we have shown LWC from CDP and mass concentration from PIP which accounts for both rain drops and ice hydrometeors.As inferred from DSD analysis that rain initiation occurred around 3 km height level (D eff > 25 mm) where PIP mass concentration crosses a threshold of 0.01 g m −3 .Higher values of PIP mass are observed at higher altitudes where LWC decreases as a consequence of auto-conversion of cloud droplets to rain drops or formation of ice hydrometeors.DSD spectral width increases with height for all the flight observations.Highest spectral width is found slightly above the rain formation level.It is to be mentioned that the spectral width values as found in these marine clouds are much higher than values in continental monsoon clouds reported in past studies (Kulkarni et al 2012, Prabha et al 2011, Bera et al 2019).Spectral relative dispersion also increases with height and mean values vary in between 0.3-0.65.The spectral broadening of DSDs is relatively more compared to the increase in mean radius which leads to increasing relative dispersion with height.

Correlation between vertical velocity (w) and LWC
Two-dimensional histogram of w and LWC is presented in figure 5. Color code indicates the frequency of occurrence (%).Linear fit (correlation) between these two quantities is indicated by a red line.This analysis will help to understand the cloud dynamics (represented by w) control on microphysics (LWC).For the shallow clouds on RF08, we can see that w is relatively weak and the probability distribution function (PDF) is much wider.This indicates clouds are more like mixed and diluted by entrainment of the environmental air.The correlation coefficient (Pearson correlation) and slope of the linear fit are having lower values.For the other two flights on RF19 and RF22, clouds are dynamically active with stronger updrafts and downdrafts.w-PDF is relatively narrow with most occurred frequencies near to zero value.The positive correlation indicates that stronger updrafts help to gain higher LWC at different cloud layers.The correlation coefficient and slope of the linear fit are also higher in these clouds.The deepest cloud has tighter correlation that indicates environmental impact (entrainment-mixing) in this cloud is lowest.There are also notable changes in the environmental RH during these observations (as shown in figure 3) where higher RH is seen for the deepest cloud case.The entrainment-mixing impact on clouds is also dependent on the moisture condition of the atmosphere (Derbyshire et al 2004, Stirling and Stratton 2012, Lu et al 2018).Zhao et al (2018) found that the entrainment rate and RH are negatively correlated in deep convection.A similar consequence may be expected in these observations which resulted in stronger correlation between w and LWC for the deep cloud.One may note in figure 3 that the middle troposphere became moister as the dry layer (inversion layer) noted on 11th July indeed disappeared.The large-scale moisture transport over AS by the monsoon south-westerly wind would be  responsible here.As the middle troposphere become moister, the propensity for deeper clouds also has enhanced as found earlier by Malap and Prabha (2023).There can be other environmental effects such as convective available potential energy (CAPE) which may play a role behind this stronger w -LWC correlation in deep cloud case.CAPE is expected to be higher for the deep cloud, and may lead higher buoyancy and stronger updraft velocity (w) and more LWC leading to a tighter correlation.

Rain and ice hydrometeors properties
As mentioned earlier that rain formation occurred in these clouds at lower altitudes (@ 3 km).Here we show some images from PIP in precipitating cloud layers (3-8 km) for 26 July flight observation.It is to be informed that the freezing level over the AS region is around 5 km altitude during summer-monsoon season.So, 3-5 km cloud layers contain liquid (drizzle and rain drops) hydrometeors and above layers (>5 km) contain both liquid and ice hydrometeors.Figure 6 shows images of various hydrometeors from PIP at different levels.The coexistence of ice particles and cloud/drizzle drops indicates that mixed-phase cloud processes were highly active at 7 km.
Overall, pristine ice particles are rarely seen and most of the ice particles are highly rimed.It indicates that riming is a major growth process of ice particles in the observed clouds.A few pristine needles/columns were observed around 6 km and may have been associated with riming splintering.During the descent, aggregates with diameters of about 2-3 mm were found to be present at 6.2 km.As ice particles approaches relatively warmer temperature, conditions are favourable for aggregation due to increase in sticking efficiency.At 3.7 km, numerous raindrops with diameters of 0.5-1 mm were observed.The evolution of CDP DSD and PIP particle size distribution (PSD) from warm to mixed phase regions of the cloud is also shown in the lower panel.PSD exhibits significant broadening in the mixed phase portion of the cloud when compared to warm regions.This can be attributed to efficient growth of ice particles through riming and aggregation.Arabian Sea being significantly loaded with dust particles which may act as ice nuclei (IN) and can enhance the formation of snow and graupel hydrometeors (Fan et al 2014).

Relation between microphysical parameters
Figure 7 depicts microphysical relations between e-AF and b-e.AF is a measure of dilution by dry air entrainment in clouds.Therefore, e-AF relation will be useful for microphysical parameterization of the entrainment effect.For this particular analysis, we have considered larger cloud lengths with cloud pass time >5 s (about 500 m wide clouds), so that undiluted cloud core exists in the samples.The convective core in cumulus clouds remain adiabatic (AF @ 1) which indicates no influence of dry air entrainment.However, samples with AF <1 are indicative of entrainment-mixing effect (Bera et al 2016).We have found negative relation between e and AF for all the flights over AS.This means that lateral entrainment at cumulus cloud edges makes the DSDs spectra wider (higher dispersion).e varies between 0.2 to 0.7 with AF variation between 1.0 to 0.0.A similar relationship between e-AF was found in many previous observational studies using measurements from different geographical locations (Burnet and Brenguier 2007, Lu et al 2013, Bera 2021).The correlation is stronger for shallow clouds and relatively weaker for deep clouds.This is a consequence of stronger entrainment in shallow clouds and less entrainment in deep clouds (Zhao et al 2018, Bera and Prabha 2019, Dogra et al 2023).Vertical velocity is indicated by color code that indicates that more adiabatic samples are associated with updrafts and diluted samples are associated with downdrafts.A positive non-linear relationship is found between b and e.The positive relation between b and e is tighter as both are DSDs spectral dependent parameters.Both b and e are smaller for adiabatic samples as indicated by color code.The functional forms of these relationships are provided in table 2 which can be used in model parameterization.

Parameterization of DSDs by gamma function distribution
The gamma function distribution fitting of observed DSD is achieved here using two different analysis methods as described in the data methods section.Here we attempted to fit the gamma function distribution for both small cloud droplets measured by CDP and precipitating drops and ice crystals measured by PIP.A detailed investigation of ice particle size distribution (PSD) in monsoon clouds using cloud imaging probe was provided previously by Patade et al (2015).They attempted to parameterize gamma function distribution of observed ice PSD for continental monsoon clouds over India.In this paper, we provide gamma function distribution of DSDs for small cloud droplets (2-50 mm) and also for large drops and ice crystals (0.1-6.2 mm). Figure 8 shows relationships between the three parameters of the gamma function distribution.Panels (a-c) display gamma function parameters for cloud droplets measured by CDP using method-1.The shape parameter m is dimensionless and the slope parameter l is in m −1 and the intercept parameter N 0 has unit of m -m . 4 Colors of the markers indicate temperature of the cloud samples.l and m have a linear relationship and the linear fit is shown by the line (black) plot.Fitting parameters of the linear fit are also provided in each plot (where 'a' is intercept and 'b' is slope parameter of linear fit).However, the relationships between N 0 -l and N 0 -m are not linear, and a log-scale for N 0 is required.Here we presented linear relationships between Log (N 0 )-l and Log (N 0 )-m.The intercept parameter N 0 of gamma function fitting varies over a wide range (10 15 -10 40 ) and makes it difficult to find best gamma fit using method 1 as this method uses iteration of loops for varying values of gamma function parameters.The second method (method 2) seems to be easy for getting gamma function fit of observed DSDs.One can compare method 1 and method 2 as presented in upper and middle panels, respectively.Gamma function parameters have a relatively larger range in method 2 compared to method 1. Temperature dependency of these parameters is also visible.Warmer temperature (near to cloud base) associated with higher values of these parameters.
Lower panel (g)-(i) shows relationships between gamma function parameters for large size drops and ice hydrometeors measured by PIP.Here we have used only method 1 (as method 2 leads to significant error) for the gamma function fit.The range of these parameters (N 0 , l, m) has decreased significantly which makes it more Figure 8. Scatter plots between gamma function parameters (λ, m, N 0 ) which are derived using two separate methods as described in data methods section.Upper (method-1) and middle (method-2) panels correspond to gamma function fitting using CDP droplet size distribution data and the lower (using method-1) panels correspond to gamma function fitting using PIP particle size distribution data.Linear relation established between slope (λ) and shape (m) parameters.Relations between N 0 -m, and N 0 -λ are non-linear, and we have established linear relations using Log scale for N 0 which varies widely.Coefficients of linear fit are indicated on each figure panel where 'a' is slope and 'b' is intercept of the linear fitting.Temperature variations of these relations are also indicated by the color code.appropriate for applying method 1.The linear relationships between the gamma function parameters are also shown by line plots.The values of 'a' and 'b' of the linear fits are also smaller for PIP PSDs compared to CDP DSDs.The parameters values can be directly used in large-scale models where bulk microphysics is used.A comparison of the two methods for deriving the gamma function parameters is provided in the supplement figure A4.The correlation coefficient between the fitted and observed DSDs (or PSDs) is presented with respect to temperature variation.For CDP, we can see a large scattering (0.5-1.0) in correlation coefficient values.Nevertheless, method 2 shows better correlation compared to method 1.For PIP, method 1 shows a very high correlation coefficient (> 0.9) and indicates the suitability of this method for large raindrops and ice particles.

Summary and conclusions
Microphysical characteristics of aerosol and clouds under contrasting dust loading conditions are studied over the north-eastern Arabian Sea region using in situ flight observations and a synergy of other data sets such as CALIPSO satellite observation, MERRA-2 and ECMWF reanalysis (ERA5) products.We observed that when the mid troposphere became moist, deeper convection developed in these environments which are mainly controlled by the large-scale atmospheric dynamics.Aerosol over this region is largely coming from the Arabian Desert by the westerly wind of monsoonal flow, forming a dominant aerosol type over the northern Arabian Sea (AS) region during the summer-monsoon season.The dominance of Arabian dust aerosols over the AS has been reported in many previous studies (e.g.Nandini et al 2022).Aerosol concentration was high over an altitude of 4 km where supermicron size aerosols are present.Cloud base height over the sea region is around 600-800 m above mean sea level.Warm rain process is found to be rapid in these clouds at a lower height ~3 km and under moist conditions (up to mid troposphere) these clouds can grow much deeper (top about 8 km) and activate the mixed-phase cloud processes.The unique aspect of the present study is investigating the mixed phase cloud processes over AS, which is not reported earlier from the in situ measurements.The region is typically dominated by warm rain process (Konwar et al 2012b).We report three case studies of varying environmental conditions with dust and contrasting cloud and rain microphysics.The first observation case (11 July) showed suppression of warm rain by high droplet number concentration, it is too dry and dominated by dust over the clouds along with a strong inversion layer that restricts cloud growth.While for other two cases, warm rain formed at lower altitude.PIP images have revealed ice hydrometeors types such as graupel, snow, needles, aggregates in between 6-8 km height.Droplet size distributions over the Arabian Sea are multi-modal and wider in nature and leading to a higher spectral dispersion.Relative dispersion increases with height and varies between 0.3-0.7.Entrainment effect is found to increase relative dispersion (spectral broadening) in these clouds as revealed by a negative correlation between relative dispersion and adiabatic fraction.
The observational findings reported here are comparable to the results reported in previous studies from other locations.However, the uniqueness of this study is that these in situ observations provide more detailed information about the aerosol and cloud microphysics properties, especially the mixed-phase properties of convective clouds over the AS, which was not found in the past literature.Drop size distributions measured by CDP and PIP are parameterized by the gamma function distributions, which is the most important outcome of the study as it was lacking in this region and can be applied in several large-scale models for better representation of Arabian Sea clouds.

3
and b is the effective radius ratio, a dimensionless parameter that depends on the spectral shape of the clouds DSD(Liu and  Daum 2000, Pandithurai et al 2012).We propose the functional relationships between e -AF and b e -,

3. 1 .
Wind flow and aerosol spatial features over the arabian sea Flight observations over the Arabian Sea were conducted on three different days(11, 23, 26 July 2015)  during the south-west monsoon season.Large-scale dynamical features and aerosol conditions are depicted in figure 1. Wind vectors (arrows) at 950 hPa pressure level and dust column mass density contour plots are shown in panels (a)-(c) and aerosol optical depth ( AOD at 550 nm) contour plots are shown in panels (d)-(f).Here we have used ECMWF daily mean wind data and MERRA-2 hourly averaged model data for dust column mass density and AOD which are shown for 3 h mean during afternoon hours (12:00-03:00 PM).Flight observation location was very close to the west-coast of India (about 100 km deep in Sea) as indicated by a circle on figure 1(a).Wind flow pattern shows that winds over the observation location are coming mostly from the south-westerly direction.Spatial distribution of dust column mass density indicates that northern AS is significantly loaded with dust aerosol ( ~1 g m −2 ) mainly coming from the Arabian peninsula.These dust aerosols are transported to the westcoast and even inland of India.A gradual reduction in dust column mass density over the observation location is seen during the consecutive observation days associated with strengthening of the southerly winds and rainfall activity possibly washed out some dust particles.AOD spatial pattern is well connected with the dust spatial pattern and indicates the dominance of dust aerosol in the northern AS region during the monsoon season.HYSPLIT air-mass back trajectories (48 h) at three different height levels (0.6 km, 2.5 km, and 4.1 km) for the observation location are presented in supplement figure A1.The air-mass trajectory is south-westerly at 0.6 km, westerly at 2.5 km and north-westerly at 4.1 km (except for 26 July which has westerly air-mass even at 4.1 km).These air-mass trajectories indicate the transport of dust aerosols over the observation location.Further understanding on aerosol sub-types over the AS is gained by analysing the CALIPSO satellite observation which is presented in supplementary figure A2.This indicates major aerosol sub-types are marine (sea salt), dust, and dusty-marine.It may be noted that CALIPSO satellite pass (day time) is quite far (around 600-800 km) from the observation location and mainly through the middle of AS on 23rd and 26th July.Information of cloud types is also provided in CALIPSO observations which are presented in lower panels of supplement figure A2.One can notice that mainly shallow clouds (cloud top < 2 km) persisted over the lower troposphere on 11 July.Cumuluscongestus clouds (cloud top < 6 km) observed on 23 July and deep cumulus (cloud top < 9 km) observed on 26 July as revealed from the CALIPSO.

3. 2 .
In situ observations of aerosol, CCN and clouds Aerosol particles number size distribution (NSD) and droplet size spectrum of CCN activation at 0.2% supersaturation (SS) is shown in figure 2. Aerosol NSD is measured by a PCASP probe on board the research aircraft.This probe measures the accumulation and coarse mode particles.

Figure 1 .
Figure 1.Spatial distribution of dust column mass density (a)-(c) and aerosol optical depth: AOD (e)-(f) over the Arabian Sea and adjacent continental regions using MERRA-2 dataset.Coast line is indicated by black line.Observation location is indicated by a black circle in panel (a), which is close to west-coast of Indian.Vectors in upper panel indicate wind speed (relative scale) and direction at 950 hPa pressure level.

Figure 2 .
Figure 2. (a) Aerosol number size distribution (NSD) using PCASP.(b) CCN activation droplet size spectra using CCN counter at 0.2% supersaturation.Three flight observations are indicated in panel (b) by legend.These observations correspond to sub-cloud layers (within 200 m below cloud base).Error bars indicate standard deviation of spatial mean.

Figure 3 .
Figure 3. (a)-(c) Vertical profiles of ambient temperature (T), relative humidity (RH) and cloud liquid water content (LWC) for the three flight observations RF08, RF19, and RF22 respectively.(d) Vertical profiles of aerosol (N aerosol ) and CCN (N CCN ) number concentrations for the three research flights.(e) Vertical profile of supermicron size (D p > 1 mm) aerosol number concentration. (f) Vertical profile of aerosol effective radius (r eff aerosol ( ) ). (g) CCN-supersaturation (SS) spectra with Twomey's empirical fit parameters (k and C 0 ) for the three flight observations.(h)-(i) Vertical profiles of cloud droplet number concentration (N d ) and droplet effective radius (r eff cloud ( ) ) respectively.Symbol colors in panels (d)-(i) correspond to flight observations RF08 (black), RF19 (red) and RF22 (blue), respectively.Error bars indicate standard deviation of spatial mean.

Figure 4 .
Figure 4. (a)-(c) Cloud droplet size distribution (DSD) as presented by color contours.Color label on top of the figure indicates droplet number concentration per droplet size ( - cm 3 m - m 1 ).X-axis presents sample number of observation and y-axis presents droplet size (diameter).Height of observation is presented by a red line plot and droplet effective diameter (D eff ) is presented by black dots whose size is scaled by spectral width of DSD.(d) Vertical variation of cloud water content using CDP and PIP instruments.CDP gives LWC as it measures small droplets only but PIP gives total water mass (liquid and ice) as it measures both large drops and ice crystals.(e)-(f) Vertical variation of DSD spectral width (s) and relative dispersion e = s rm ( ) respectively.In panels (d)-(f) error bars indicate standard deviation of spatial mean and colors black (RF08), red (RF19), and blue (RF22) indicate research flight numbers.

Figure 5 .
Figure 5. 2D histogram of cloud liquid water content (LWC) and vertical velocity (w) for RF08 (a), RF19 (b), and RF22 (c).Color code indicates frequency of occurrence (%).Linear fitting (red line) between these two variables is also shown on each panel.Slope of the linear fit and Pearson correlation coefficient (corr) are provided on each panel.

Figure 6 .
Figure 6.Images of various hydrometeors from PIP during ascent and descent of the aircraft inside clouds are presented in upper panels.Drop size (from CDP) and particle size (from PIP) distributions for the observed cases are shown for the warm and mixed phase region in lower panels.Height (h) of observation and air temperature (T) are indicated over each figure panels.

Figure 7 .
Figure 7. (a)-(c) Microphysical scatter plots between relative dispersion (e) and adiabatic fraction (AF = LWC/LWC a ) with vertical velocity (w) information in the color code.Linear relation between e and AF is presented by the line fit (black) and Pearson correlation coefficient (r) is indicated on each panel.(d)-(f) Microphysical scatter plots between effective radius ratio (b ) and relative dispersion (e) with AF presented by the color code.Non-linear relations between b and e are shown by the line fit (black).Flight observations RF08, RF19 and RF22 are presented in left, middle and right panels, respectively.Fitting equations for each panel are provided in table 2.

Table 1 .
Details of the instruments used in measuring aerosol, cloud and precipitation microphysics and meteorological parameters.

Table 2 .
Linear fit equations, correlation coefficients and R2 errors of different plots presented in figure 7.