Precipitation enhancement in stratocumulus clouds through airbourne seeding: sensitivity analysis by UCLALES–SALSA

Abstract. Artificial enhancement of precipitation via hygroscopic cloud seeding is investigated with a numerical large-eddy simulation model coupled with a spectral aerosol-cloud microphysics module. We focus our investigation on marine stratocumulus clouds and evaluate our model results by comparing them with recently published results from field observations. Creating multiple realizations of a single cloud event with the model provides a robust method to detect and attribute the seeding effects, which reinforces the analysis based on experimental data. Owing to the detailed representation of aerosol-cloud interactions, our model successfully reproduces the microphysical signatures attributed to the seeding, that were also seen in the observations. Moreover, the model simulations show up to a 2–3 fold increase in the precipitation flux due to the seeding, depending on the seeding rate and injection strategy. However, our simulations suggest that a relatively high seeding particle emission rate is needed for a substantial increase in the precipitation yield, as compared with the estimated seeding concentrations from the field campaign. In practical applications, the seeding aerosol is often produced by flare burning. It is speculated, that the required amount of large seeding particles suggested by our results could pose a technical challenge to the flare-based approach.



Introduction
Water scarcity is a cause for increasing concern in arid as well as in semi-arid regions (WWAP, 2019). This has revived the interest in research and investments in weather modification efforts, particularly those related to rain enhancement (Flossmann computational mesh with cyclic lateral boundary conditions, including a prognostic description of the three Cartesian wind components. The advection of momentum is based on fourth-order difference equations with time stepping scheme based on the leapfrog method. The prognostic scalar variables include the liquid water potential temperature and variables describing water vapor and condensate amounts, depending on the model configuration. The scalar advection uses a second order fluxlimited scheme and the time integration is performed with a simple Eulerian forward time stepping method. 70 The aerosol and cloud microphysics are represented by the SALSA module. The design of SALSA is described in detail in Kokkola et al. (2008) and Tonttila et al. (2017), but key features are also summarized below, including a description of some important updates. The general layout of the current SALSA module is illustrated in Figure 1. SALSA is a bin microphysics model, describing the size distribution and composition of particles in up to four different categories: aerosol particles, two sets of liquid hydrometeors (cloud droplets and a separate regime dedicated for droplets and precipitation dominated by collision-75 coalescence growth) and ice (not used in the current work and thus not shown in Figure 1). For aerosol particles and cloud droplets it is possible to use two parallel sets of bins in order to describe externally mixed aerosol populations with separate size distributions.
For particle composition (in all categories), the model uses up to five different aerosol constituents (sulfate, sea salt, organic carbon, black carbon, dust) plus water. Each parameter (particle number and mass of each constituent) in each bin comprises 80 a prognostic scalar variable in UCLALES-SALSA. The model solves all the key microphysical processes for all particle categories, including cloud activation, coagulation and collection processes as well as the condensation of water vapor on both aerosol particles and hydrometeors (see Tonttila et al. (2017) for further details).
The design of SALSA is an attempt 1) to find a compromise between computational cost and model accuracy and, in particular, 2) to be able to track both non-activated and activated particle populations. The latter in particular is fulfilled by 85 setting both the aerosol and cloud droplet bin limits according to the dry aerosol or CCN particle diameter, as illustrated in Figure 1. This allows to preserve the characteristics of the aerosol size distribution at the aerosol -cloud droplet interface. As seen in past model experiments (Tonttila et al., 2017;Boutle et al., 2018), it also provides an adequate description of the droplet size in order to solve the in-cloud microphysical processes. The ability to track the aerosol size distribution in and outside of clouds brings the model a step further from more straight-forward 1-d bin model designs, where information about activated 90 aerosol is not explicitly tracked. This is also a very important feature for this study, since tracking the evolution of the seeding particles within the cloud is key in order to capture their influence on cloud microphysics and rain formation as well as to analyze and understand the underlying processes.
An important upgrade not described in previous papers concerns the description of precipitation. In the current version, the size distribution for drizzle and rain is described by 20 mass doubling bins (in the wet diameter space) starting from 20 µm.

95
Moreover, instead of a parameterized autoconversion process, we now determine the transition from cloud droplet to drizzle regime directly from the coagulation code. If the droplet diameter resulting from the collision-coalescence between two cloud droplets exceeds the 20 µm limit, the outcome is moved to a drizzle bin with appropriate diameter range. Obviously, the 20 µm limit separating cloud droplets and drizzle is quite small. The decision for this limit was made in favor of more accurate resolution of the droplet size in wet diameter space, where the droplet growth begins to be dominated by collision-coalescence.

100
This helps to ensure a smooth transition from cloud droplets through intermediate sized drizzle (Glienke et al., 2017) to full size drizzle drops and eventually rain. It also provides a more accurate method to describe the growth and transport of GCCN particles within the diameter space, than what was possible with the previously used parameterized autoconversion method (Tonttila et al., 2017).

Model setup 105
Marine stratocumulus clouds have been used as the setting for numerous field experiments looking into aerosol-cloud interactions. In a recent study (Jung et al., 2015, henceforward J15), rain enhancement by hygroscopic cloud seeding was investigated from airbourne observations, that took place southwest of Monterey, California, on 3rd August, 2011. We will use this case as the basis for the model experiments in this paper. The meteorological data and aerosol characteristics presented in J15 are used to provide the initial conditions for UCLALES-SALSA. We exploit this setup to investigate the ability of UCLALES-SALSA 110 to simulate the cloud microphysical response to hygroscopic seeding in a series of experiments, consisting of a control run and three cloud seeding runs with different seeding strategies, whose details are outlined below.

Initial conditions
The measurements reported in J15 took place starting from 16 UTC. We match the model time with this period, which mainly affects the radiative budget at cloud top via the solar zenith angle. The initial profiles of potential temperature and the total 115 water mixing ratio are shown in Figure 2, where the moisture profile is saturated between the levels of approximately 300 m and 650 m. The strength of the potential temperature inversion at the top of the cloud (after condensation of the excess moisture from Figure 2) is approximately 8.5 K. For simplicity, the horizontal wind is initialized to u = −12 m s −1 , v = 0 m s −1 , following the standard meteorological notation, which corresponds to the observed wind speed in J15. Surface energy fluxes and the large-scale subsidence are prescribed according to Ackerman et al. (2009) who focus on a similar stratocumulus case.

120
The model grid resolution is set to 50 m (10 m in the vertical) and the domain size is 5 km across laterally, extending up to The initial aerosol size distribution is described in the model as a sum of lognormal modes. J15 reported the accumulation mode aerosol concentration in the range of 200 to 800 cm −3 , with CDNC approximately 150 − 200 cm −3 . Therefore, we set the initial accumulation mode concentration to 200 cm −3 . Additional 400 cm −3 particles were allocated in the Aitken mode, 125 which can be viewed as a typical feature in marine boundary layer (e.g. Zheng et al., 2018). However, for the current setup, the Aitken mode particles reside mostly below the critical size for droplet activation -pilot test runs showed little sensitivity in the results for the Aitken mode number concentration. In addition, natural GCCN particles consisting of sea salt, which are virtually omnipresent in marine boundary layers (Jensen and Nugent, 2017), were assigned to the background aerosol, with mode diameter at 1 µm and concentration at 1 cm −3 . The lognormal size distribution parameters for the initial model aerosol 130 population are summarized in Table 1.

Cloud seeding
For the seeding aerosol, we use sea salt as a proxy for the particle composition. We use SALSA's parallel bin regimes to describe the seeding aerosol initially as externally mixed from the background. The seeding aerosol is then allowed to interact with the background particles through coagulation and cloud collection. We will assume two modes for the seeding particles, 135 with mean diameters at 1.5 µm and 8 µm, which roughly correspond to the mode mean diameters shown in J15, but are also more generally representative of the powdered salt size ranges presumed effective in terms of precipitation enhancement (Rosenfeld et al., 2010).
In the model experiments, the seeding emissions are performed in-cloud, near the cloud top altitude, at approximately 580 m, as in J15. Two methods for the release of the seeding particles are considered: i) a domain-wide instantaneous injection of 140 the aerosol at the specified layer and ii) an explicit Lagrangian point source emission. In the former, the particles are assumed to occupy a layer of 50 m in depth. For the latter, the trajectory and speed of the emission source are prescribed. We assume the seeding aerosol plume to immediately occupy a cross-sectional area of 50 × 50 m 2 . Further dilution of the plume is controlled dominantly by grid-scale mixing. To mimic the airborne emissions, we assume a speed of 60 m s −1 for the source, whose trajectory is illustrated in Figure 3. The seeding source propagates towards the positive x-direction along a trajectory that 145 covers the entire extent of the domain in y-direction. Since the seeding proceeds against the horizontal wind (in x-direction), the seeding is stopped at the middle of the domain. This is to avoid seeding the same plume twice, because the rear of the plume is advected back into the domain from the right due to the cyclic boundary conditions. Our experimental strategy is to create two parallel sets of realizations: a control simulation without seeding and a set of experiments with seeding particle emissions. This allows to account for both the direct microphysical effects, as well as the 150 dynamical feedbacks potentially caused by the aerosol perturbation. We must take care, that the initial states for the control simulations and sensitivity experiments are identical. Therefore, a "master" model run is performed, from which the model state is saved to a "restart" file prior to the time of the seeding procedure. The length of the "master" run is 4 hours, to allow the boundary-layer mixing and precipitation process to settle in a quasi-steady-state (i.e. spinup). All our model experiments, including both the control and seeding experiments, are initialized from this restart file, which ensures that any differences in 155 the experiments can be attributed directly to the seeding.
The model experiments performed in this study are summarized in Table 2. The seeding experiments Seed1 and Seed2 use the moving point emission source along the trajectory shown in Figure 4. The total seeding rate in the former experiment is set as 1.5 × 10 11 s −1 and in the latter as 1.5 × 10 12 s −1 . With the assumed initial plume cross-section and the aircraft speed, this yields an initial plume concentration of approximately 1 cm −3 and 10 cm −3 , respectively, which will be rather quickly 160 diluted by in-cloud mixing and drizzle formation. In the experiment Seed3 the seeding particles are injected instantaneously in a domain-wide slab at the same altitude as the point emissions in Seed1 and Seed2. The initial concentration in Seed3 is set as 1 cm −3 . Comparing the model results between the point source and domain slab emissions and different emission rates helps to understand the impact of the total emitted mass to the efficacy of the seeding, since it critically affects the concentration in the diluted population. Note, that we do not consider the possible aerodynamic effects caused by the aircraft.

Sensitivity of precipitation
The overall precipitation rate in the simulated case is rather low. The observed estimate reported in J15 is about 0.05 mm h −1 on average, sampled around the cloud base height. In order to evaluate the seeding efficacy in UCLALES-SALSA, we want to match the simulated precipitation rate as close as possible to the measurements. We find that in the current case, the precipitation rate is strongly controlled by two key aspects in the model setup: the background aerosol and the vertical resolution. In case 170 of the former, we focus here on the role of natural GCCN, due to their overall small concentration, but potentially strong impact on drizzle formation. For the latter, vertical resolution of the model is known to have a substantial influence on the representation of entrainment mixing at cloud top (Stevens et al., 2005). This is expected to modulate the rate of evaporation of cloud droplets due to the entrained dry air from above the cloud, which affects the mean droplet size and the size distribution width and, therefore, drizzle formation via the collision-coalescence process. Pilot model simulations are performed to test the 175 choices made in our simulation setup.
First, Figure 4 illustrates the effect of GCCN concentration on the simulated precipitation in pilot experiments. With 1 cm −3 (used in the experiments), we obtain a precipitation rate much closer to the observed estimate, than with a lower concentration (0.1 cm −3 ). Similar concentrations of GCCN were also measured in the field campaign according to J15 and the good agreement with the observed precipitation rate (0.05 mm h −1 ) warrants the use of the background aerosol size distribution detailed 180 in Table 1.
Second, Figure 5 shows the effect of vertical resolution on the simulated precipitation rate. It is evident, that 20 m spacing in the model vertical grid yields lower precipitation rate, than the simulations with 5 m or 10 m grid spacings. That said, the difference seen between 5 m and 10 m grid spacings is minor. Even though 10 m resolution has been shown to be inadequate to fully represent the effects of e.g. entrainment mixing, considering this result and the fact that the UCLALES-SALSA is 185 computationally a rather expensive model, we find that the 10 m vertical grid spacing is justified. Figure 6 shows the flux of precipitating water sampled at an altitude of 300 m, which is close to the cloud base height. In the control case, the precipitation rate is approximately 0.03 − 0.05 mm h −1 , which agrees well with the corresponding measured 190 precipitation rate in J15. It is evident, that the seeding emission rate and seeding strategy have a strong influence on the seeding efficiency. The experiment Seed1 (with a low emission rate) yields a positive, but a rather weak signal in terms of the precipitation flux. Instead, Seed2 (with high emission rate) approximately doubles the precipitation flux within an hour after the start of the seeding, as compared to the control run. Likewise, even stronger (by a factor of 3) increase in precipitation is shown by Seed3, where the seeding rate is the same as in Seed1, but the total injected mass is much larger because it is applied 195 to all grid points across the target layer. Therefore, it is important to consider the impact of the seeding strategy on the diluted plume concentration. Figure 7 shows the estimated domain mean profiles of the seeding particle concentration. Since the seeding takes place in-cloud, practically all the particles are activated very quickly. Therefore, the concentration is analysed as the number of cloud droplets formed by the seeding particles, 15 minutes after the seeding. This gives the plumes some time to dilute; with 200 further delay, the plume concentration starts to be more and more affected by precipitation, after which it is difficult to robustly distinguish the particles associated with seeding in our modelling setup. Moreover, taking the profile as domain average is justified because the trajectory of the source in Seed1 and Seed2 is set to cover the entire width of the domain in y direction, as depicted in Figure 3, and it does not take long for the turbulent diffusion to spread the seeding across the domain area. Figure 7, clearly the highest average concentration is seen for Seed3 and the lowest for Seed1, which indeed 205 emphasizes the connection between the diluted plume concentration to the seeding efficacy. Therefore, as Figure 7 suggests, in order to produce an approximately 2-fold precipitation yield, the concentration in the diluted plume should be on the order of 0.01 cm −3 .

Microphysical signatures
The main hypothesized pathways to precipitation enhancement include 1) the reduction of CDNC due to increased competition 210 for water vapor induced by the large seeding particles and 2) enhanced collision-coalescence (Segal et al., 2004;Rosenfeld et al., 2010). Thus, to better understand the reasons for the simulated enhanced precipitation shown in Figure 6, we will analyze more closely the cloud microphysical effects induced by the seeding.
To start with, our model shows very little change in in-cloud relative humidity between the experiments. This is in line with releasing the seeding aerosol close to the cloud top height, following the experimental approach in J15. Consequently, should be rather high, on the order of 0.01 cm −3 , in order to generate an approximately 2-fold precipitation enhancement. This concentration is on the high end of the diluted plume concentrations estimated by J15.
To increase the confidence to our results, the simulated seeding effects on the cloud microphysical properties were inspected.
The model simulations skillfully reproduced the seeding effects on the hydrometeor size distribution, as compared with J15.

255
The model size distributions indicated a temporal propagation of increased concentrations first in the drizzle size range and later towards the larger rain drop sizes, which was expected as the result of an enhanced collision growth of droplets. Accordingly, the model also showed a decrease in the total CDNC by up to 10 % due to the seeding emissions. However, again the notable microphysical effects only become visible with the aforementioned 0.01 cm −3 diluted seeding plume concentration.
In addition, J15 reported a somewhat larger decrease in CDNC after the seeding. Arguably, estimating the seeding efficacy 260 from the model by comparing multiple realizations provides a more robust estimate (at least in terms of the data sampling) than estimating the effects of seeding aerosol from temporally consecutive sets of observations. Given the qualitatively very similar seeding effects between the observations and our model, but a larger decrease in CDNC in the observations than in our model, we speculate that other factors than the seeding alone might be playing a role in the observed cloud layer. These include possible meteorological variations as well as aerodynamic effects from the aircraft used for seeding, which we do not account 265 for in our simulations. The latter has been studied in the context of ice and mixed-phase clouds (Kärcher, 2018;Moisseev et al., 2019), but in terms of warm cloud seeding, a dedicated study would be required.
Flare burning is a commonly used and cost effective method to produce the seeding particles, but they have the tendency to produce particles in a wide range of sizes, with significant contribution from sub-micron particles. Our results highlight the importance of the seeding plume concentrations. Even though the plume concentration can be controlled to an extent both by 270 the emission rate as well as the seeding strategy (e.g. flight plan, use of multiple aircraft), the results corroborate the view presented in recent literature (Rosenfeld et al., 2010), that obtaining a high enough concentration of large (> 1 µm) CCN from flares in terms of efficient precipitation enhancement might be challenging with currently used practical applications.     Figure 10. The contribution of hydrometeors to the total precipitation rate at cloud base as a function of size before (-300 s) and after (1800 s and 3600 s) the particle injection in the seeding experiments.