Three cloud seeding events carried out on 21, 23, and 24 August in 2019 are selected here for the evaluation of seeding signatures and plausible links to microphysical properties. Instruments for the measurement of flare particles, aerosol, and cloud properties were operated on a Beechcraft B200 aircraft. This aircraft was equipped with flare racks located under both the wings and the belly. The flare racks in the wings are used for warm cloud seeding operations (Mather et al., 1997), while the belly is utilized for cold cloud seeding operations (French et al., 2018; Friedrich et al., 2020). The temperature (T, °C), relative humidity (RH %), wind speed (m s-1), and wind directions were measured with the Airborne Integrated Meteorological Measurement System (AIMMS-20). The DSD in the size range of 2–50 µm was measured with a cloud droplet probe (CDP-2) manufactured by Droplet Measurement Technologies, LLC, USA. The bulk microphysical properties are derived from the measured DSDs, e.g. the total number concentration (Nt, cm-3) and liquid water content (LWC, g m-3). The effective radius (re, µm) was calculated from the ratio between the third and second moments of the DSDs (Martin et al., 1994). The precipitation imaging probe (PIP) was used to document drizzle drops in the clouds over the size range of 100–6200 µm. The technical specifications of these instruments are shown in Table 1. The uncertainties associated with the CDP and single-particle light-scattering instruments like the CDP have been well characterized and documented (Baumgardner, 1983; Baumgardner et al., 2001, 2016; Lance et al., 2010). In water droplets the sizing uncertainty is ±20 % and counting accuracy ±16 %, which propagates into a LWC uncertainty of ±38 %.

Cloud properties are altered by the entrainment of cloud-free air masses at the edges of the cloud; hence to minimize the influences of entrainment and mixing processes in the seeded and non-seeded clouds, only clouds with near-adiabatic or slightly diluted cloud parcels are considered to evaluate cloud microphysical properties. Only cloud passes with LWC in the range of 0.75 < LWC/LWCmax < 1 (Konwar et al., 2021) were selected for this study. Here, LWCmax represents the maximum measured value of LWC during a cloud pass. Note that this cloud regime may be considered the cloud core, typically located within the strongest updraft zone. Our main aim is to select the DSDs located within the cloud core regime. Note that in most naturally developing clouds the LWCmax values are less than the adiabatic LWC (LWCad) values because of the entrainment of drier air, mixing, precipitation fallout, and radiative heating/cooling (Korolev et al., 2007). The maximum adiabatic fraction, AFmx= LWCmax/ LWCad, indicates the extent of dilution that has occurred in the cloud core regime. During their development and dissipation stages clouds undergo significant changes; therefore, it is practically impossible to find two clouds identical in all states, let alone their lifetimes. It is to be noted that the AF values may not accurately represent the mixing state when CC is significant and drizzle particles form within the clouds. Additionally, studies of the seeding effect using parcel model simulations without the inclusion of mixing processes indicate a significant change in the LWC profile compared to the non-seeded cloud (Konwar et al., 2023). Such changes in LWC values at different vertical distances from the cloud base of the seeded clouds do not necessarily imply the true dilution rate in the observations. Since the cloud seeding flare produces high concentrations of small-sized particles, they can be activated into cloud droplets in strong updraft regimes with high supersaturation (Konwar et al., 2023; Prabhakaran et al., 2023). In a parcel model simulation, small aerosols released from flares are found to be activated due to an increase in supersaturation when the collision–coalescence process is active (Konwar et al., 2023). For details on the nucleation process within the zone of intense collision, where a rapid decrease in drop concentration leads to an increase in supersaturation, readers are referred to Pinsky and Khain (2002). At a given height, however, seeding does not change the adiabatic value, but activation of new particles at a given level due to seeding can alter the AF. Another aspect is that near the cloud base the LWCad values are quite small (e.g. < 1 g m-3); therefore any small change in the measured LWC could indicate a large change in AF. With this background information in mind, the DSDs for seed cloud (SCl) and non-seed cloud (NSCl) conditions are compared at different vertical distances above the cloud base (D*, km). The lowest unbroken visible section of a convective cloud was selected as the cloud base. The cloud top is defined as the maximum altitude attained by these clouds at any given moment during their development.

We utilized an mAMS to analyse the chemical compositions of residual particles from cloud droplets, specifically to trace flare particles within the seed clouds. The CVI is manufactured by Brechtel Manufacturing Inc. (BMI, Model 1204, https://www.brechtel.com/, last access: 20 March 2024). The cloud droplets were passed through the CVI to obtain the droplet residuals that were sampled by the mAMS. Through the use of inertial impaction, the CVI inlet allows cloud hydrometeors with aerodynamic diameters larger than a certain size to pass through, depending on the velocity of the counterflow. A warm, particle-free dry nitrogen gas is directed towards the inlet against the direction of the ambient airflow. This causes a separation in the incoming free-stream air, with particles > 7 µm in the sampled air having enough inertia to penetrate the counterflow and join the sample flow. The CVI adjusted flow rates with its internal software based on true air speed (TAS) obtained from the AIMMS-20. The cut size is a function of various factors, e.g. air pressure, air speed, and the average angle of attack, and is known to have an uncertainty of approximately ±1 µm. The heated air evaporates cloud droplets, and the remaining dried residuals enter the mAMS where their chemical compositions are classified. Details of the operational principles of the CVI can be found in Ogren et al. (1985, 1987), Noone et al. (1988), Shingler et al. (2012), Golderger et al. (2020), and references therein.

The mAMS measured the residual particles with vacuum aerodynamic diameters of less than 1 µm, sampling through an aerodynamic lens. The aerosol sample stream is intermittently blocked to measure background signals. The aerosol signal is the difference between unblocked (“open”) measurements and those obtained during the blocked (“closed”) period. The mAMS sampled 10 s of closed signal for every 110 s of open signal. The heater, operated at 600 °C, vaporized the sample, electron impacts ionized the vapours, and the resultant ions were extracted into the mass analyser for the measurement of chemical composition and mass distributions (Jayne et al., 2000; DeCarlo et al., 2006; Canagaratna et al., 2007; Drewnick et al., 2015; Giordano et al., 2018; Salcedo et al., 2006).

Ice Crystal Engineering (ICE) Inc. (USA) manufactured the hygroscopic flares used in this work. The flares were composed of an aggregated mixture of potassium perchlorate (KClO4) and calcium chloride (CaCl2) (Hindman, 1978; Bruintjes et al., 2012).

For non-refractory ambient aerosol species (i.e. NH4, NO3, SO4) aerosol concentrations are obtained from the difference between the open and closed signals. The vaporization of non-refractory aerosol species at 600 °C is typically completed on the timescale of hundreds of microseconds; however, semi-refractory species such as metals and salts may take minutes to completely vaporize (Canagaratna et al., 2007; Salcedo et al., 2006).

As discussed below, the Cl, HCl, and K from the KClO4 and CaCl2 in flares are semi-refractory species which exhibit slow vaporization. These slow-vaporizing species were analysed using only the open signals. The background signal was calculated from measurements obtained immediately before the cloud intercept of interest.

CaCl2, the seeding component in the flares, has a melting point of 774 °C. Laboratory measurements of atomized CaCl2, primarily detected as Cl and HCl ions, exhibit the same slow vaporization seen in refractory salts (Drewnick et al., 2015). Figure 1 shows a comparison of vaporization timescales of CaCl2, NH4NO3, and (NH4)2SO4 obtained with an AMS during laboratory measurements of CaCl2 in solution with H2O which had been atomized and passed through a drier before sampling. This behaviour differs from that observed from non-refractory NH4NO3 and (NH4)2SO4, which were present as tracers.

The seeded cloud pass shown in Fig. 2a illustrates a single seeded cloud pass. The K and Cl time series have a delayed decay to background compared to sulfate or nitrate. The relative intensity shown in Fig. 2b highlights the delayed response in the decay of the two flare-associated species (K, Cl).

An exponential decay was fit to each cloud intercept, from the signal peak to five e-folding times. The average decay exponential (τ) for Cl and K across all seeded cloud intercepts is shown in Table 2.

For each slowly vaporizing species, a new corrected time series was created. The start, stop, and maximum total mass times were identified for each cloud pass (Fig. 3). For each species, a background signal was determined from measurements during the non-cloud period preceding each pass. This background was subtracted from the signal observed during each cloud intercept.

The cloud intercept time series peak at the same time as the uncorrected series. However, the tails were corrected to decay within five e-folding times while preserving the total mass. The equations used in these calculations are shown below.

The measured mass from the start of the pass to the end of the slow-vaporization regime was scaled by the ratio of the total area divided by the area of fast vaporization (Eq. 1). 1Conc.Areacorrected(t)StartEnd+(5τ)=Conc.t-Conc.Background⋅AreaPeak+TailAreaPeak The decay of this normalized mass is adjusted to the exponential decay fit (Table 2) to the slow-vaporized mass (Eq. 2). This decay extends from the cloud pass peak to the end of the normal vaporization period plus five e-folding times (Giordano et al., 2018). 2Conc.TailCorrected(t)PeakEnd+(5τ)=Conc.AreaCorrectedt⋅e-1τt This decay-corrected time-shifted time series is normalized to the unmodified slow-vaporizing total mass (Eq. 3). 3Conc.CorrectedtStartEnd=Conc.TailCorrected(t)⋅AreaPeakAreaPeak+AreaPeak+Tail Finally, we applied an enhancement factor correction to the mAMS data resulting from the ambient aerosol concentration being concentrated in the CVI by following Shingler et al. (2012).

Although many aerosol species readily vaporize at 600 °C, some semi-refractory materials in nature do not. Submicron aerosol particles in the troposphere, which contain Cl, are rarely semi-refractory and vaporize quickly in the mAMS. However, Cl in seeded clouds was found to vaporize slowly. The Cl measured in clouds seeded using CaCl2 and KClO4 exhibited the same slow vaporization (Fig. 2) as atomized CaCl2 in the laboratory (Fig. 1). The majority of atmospheric Cl-containing aerosols are non-refractory. In our study the slowly vaporizing Cl was only observed in seeded clouds; thus, we assume that the source of the slow-vaporizing Cl was from the flare material. Aerosol K is uncommon except as super-micron mineral dust. As shown in Fig. 2b, slowly vaporizing signals of Cl and K were observed in the campaign during seeded cloud intercepts.

The combination of the isolation of cloud residuals by the CVI and the presence of K and semi-refractory Cl allows for the discrimination of the particles containing the flare combustion products.

The element Ca was also present in the flare. The boiling point of Ca of 1484 °C at ambient pressure means that this species was not vaporized inside the AMS and is thus considered a refractory species. Since Ca could not be observed in our study, the focus remained on the other species present.

As previously discussed, the time series of semi-refractory Cl and K signals are corrected to account for the difference in the decay response of slowly vaporizing species in the mAMS. Figure 3 depicts the corrected (K*) and uncorrected semi-refractory K signals in the mAMS measurements for a seeded cloud pass, defining the periods for the start, peak, end, and tail of the pass.

A vertical profile of cloud residual aerosols within the same cloud taken before and after seeding provides a platform for measuring and observing cloud physical and chemical changes. The resultant mAMS measurements from one such experiment, on 23 August 2019, with three cloud passes of the same cloud before and three passes after seeding are shown in Fig. 4.

In the mid-level, all chemical species were found in higher quantities in the seeded cloud than in the non-seeded cloud. Cl and K concentrations were significantly increased for all seeded cloud passes compared to non-seeded cloud passes. The measurement of the flare chemical species in the seeded cloud indicates that the mAMS could successfully identify the cloud droplets that contain seeding material.

An additional observation is the increased NO3 and SO4 concentration in the cloud drops of seeded clouds at upper heights. We hypothesized that the increased concentrations of these two chemical species could be linked with the activation of the flare particles and other organics while mixing with the naturally available NO3 and SO4 aerosols. The increased concentration of NO3 in the seeded cloud may also be due to the presence of more LWC. The additional water drives nitric acid (HNO3) from gas to liquid NO3 (Wang and Laskin, 2014).

This example highlights the ability of the mAMS to identify flare-associated species, by both increased concentration and time response, in order to confirm the presence of seeding material in cloud droplet residuals.