Preliminary Results of Cloud Seeding Experiments for Air Pollution Reduction in 2020

In previous studies, cloud seeding has been primarily used to increase precipitation; however, its use for the purpose of reducing particulate matter concentration has not yet been adequately addressed. In this study, we investigated the effect of cloud seeding on the reduction of fine dust concentration by conducting two airborne experiments in west Korea along the Yellow Sea coast on November 1, 2020. Numerical simulations showed that the seeding material diffused in Seoul, Gyeonggi-do, and Chungcheongnam-do as observed from the prevailing wind direction. We also observed enhanced radar reflectivity in these areas, and analysis of aircraft observation data revealed that the average concentration of cloud, drizzle, and precipitation particles increased after seeding as compared with the observations before and during seeding. Further, the precipitation particles had large diameters after seeding; particularly, PM10 (particulate matter with size ≤ 10 µm) concentration tended to decrease owing to precipitation in the areas affected by cloud seeding; however, it remained unchanged in the unaffected areas. Although it is necessary to conduct further experiments to obtain more comprehensive results, the findings of this study highlight the possibility of reducing fine dust concentration in the atmosphere via cloud seeding.


Introduction
Studies have shown that particulate matter (PM) and yellow dust affect air quality in the Korean Peninsula (Park et al. 2012). It has also been observed that yellow dust generated in China is picked up by the westerlies and transported to the Korean Peninsula (Park et al. 2012;Kim et al. 2015). Therefore, strategies to reduce fine dust concentration in the atmosphere so as to improve air quality have been widely explored (Chun et al. 2001;Chung et al. 2001;Yi et al. 2001;Lim et al. 2004;Lee et al. 2005;Shin et al. 2005).
In Korea, the average annual PM 10 (particulate matter with size ≤ 10 µm) concentration showed a decreasing trend until 2012 owing to the implementation of an active air management policy since the 1990s to mitigate air pollution (Kim and Lee 2018;Yoo et al. 2020). However, the concentration of fine dust has been on the rise since 2013, which has become a social and environmental problem (Song 2018;Yeo and Kim 2019;Yoo et al. 2020). The increase in fine dust concentration in Korea since 2013 could be attributed to smog transported from China and an increase in the use of diesel vehicles in Korea (GRI 2013, KIPF 2017. The Ministry of Environment has proposed various policies for reducing fine dust concentration and established measures for reducing fine dust concentration with a focus on reducing NOx emissions by restricting coal power plant and vehicle operations . However, air quality in Korea has not significantly improved (Oh and Chung 2007;Woo 2009;Kim 2010;Lim et al. 2013).
In addition to these policies, it has been reported that the washout effect of precipitation, which offers the possibility to remove numerous pollutants from the atmosphere within a short period, can be employed to reduce fine dust concentration (Wesely and Hicks 2000). Several studies have been conducted to estimate the amount of fine dust that is washed out by precipitation (Park and Choi 1999;Wesely and Hicks 2000;Yang et al. 2011Yang et al. , 2012Lim et al. 2013 Chung et al. 2018;Park et al. 2020aPark et al. , 2020b. For example, Lim et al. (2013) conducted a study to quantitatively estimate the decrease in air pollutant concentration by evaluating the cleaning contribution of precipitation. In particular, it has been observed that the fine dust concentration-reducing effect of precipitation depends on its intensity and amount. Specifically, the reduction rate of fine dust concentration was found to be higher when precipitation intensity was ≥ 5.0 mm/h. However, it became insignificant when the intensity was < 5.0 mm/h (Park et al. 2020a(Park et al. , 2020b. Additionally, Ouyang et al. (2015) confirmed that fine dust concentration significantly reduces with an increase in accumulated precipitation. It has also been observed that a 1-mm increase in precipitation intensity brings about 4%-28% and 15%-35% decreases in PM 2.5 (particulate matter ≤ 2.5 µm in size) and PM 10 concentrations, respectively (Ouyang et al. 2015;Olszowski 2016).
The cloud seeding technology can be used to increase precipitation intensity given that it offers the possibility to activate cloud development and precipitation condensation based on the artificial sprinkling of condensation nuclei or ice nuclei, which act as cloud seeds to insufficiently developed clouds to produce precipitation (NIMS 2018). Silver iodide (AgI), which can act as an ice nucleus, is generally used for clouds with temperatures < 0 °C, while calcium chloride (CaCl 2 ), which is a hygroscopic material, can act as a condensation nucleus for clouds with temperatures above 0 °C. Studies on cloud seeding have been conducted worldwide, e.g., in the United States, Japan, China, and Thailand (Silverman and Sukarnjanaset 2000;Hashimoto et al. 2008;Geerts et al. 2010Geerts et al. , 2013Ohtake et al. 2014;Pokharel and Geerts;Li et al. 2017;Flossmann et al. 2018).
In these previous studies, the focus was primary on the use of cloud seeding to increase precipitation. For example, in Israel, a cloud seeding experiment primarily for increasing precipitation and not reducing PM concentrations, was conducted in the desert (Rosenfeld and Farbstein 1991, Nirel and Rosenfeld 1995, Givati and Rosenfeld 2005. Thus, reducing fine dust concentration based on the washout effect of precipitation cloud seeding is an emerging research field. In this study, we analyzed the fine dust concentration reducing effect of cloud seeding based on the washout effect of induced precipitation, i.e., the aim of this study was to investigate the effect of cloud seeding on the reduction of fine dust concentration. To this end, we conducted two airborne experiments in west Korea along the Yellow Sea coast.
The remainder of this paper is structured as follows: in Section 2, we describe the airborne cloud seeding experiments conducted in 2020. In Section 3, we provide the results of the cloud seeding experiments and also describe the effect of precipitation on find PM concentration reduction. Finally, in Section 4, the results of the cloud seeding experiment are discussed and techniques that may be applied in future to analyze the reduction of fine dust concentration resulting from cloud seeding are suggested.

Airborne Cloud Seeding Experiments
In this study, cloud seeding experiments were conducted twice over the west coast of Korea on November 1, 2020 using an aircraft to explore the possibility of reducing fine dust concentration by inducing precipitation. The airborne cloud seeding was performed in the morning (11:26-12:20 KST) and in the afternoon (15:33-16:29) in the west sea near Taean. Thereafter, the effect of the experiments was observed using an aircraft (the procedure is shown in Fig. 1). The cloud seeding experiments were conducted in the order B → S → A (before seeding, B; during seeding, S; and after seeding, A) and the seeding experiments and observation paths were marked according to the wind direction over the experimental area. Experiment 1 was conducted at 0.7 km above the ground, while experiment 2 was conducted while the aircraft was in motion according to the position of the cloud given that it was distributed at an altitude that was lower than the observation altitude before this experiment. CaCl 2 was used as the seeding material, and 18 flares of the substance (1 flare = 180 g) were seeded in each experiment. Table 1 shows the average flight altitude, speed, temperature, and total amount of CaCl 2 used for cloud seeding on the day of the experiment (November 1, 2020).

Results and Discussion
The effect of cloud seeding on fine dust concentration was analyzed via cloud seeding rainfall numerical simulation, radar reflectivity analysis, ground precipitation analysis, and cloud particle analysis (Ku et al. 2020, Jung et al. 2021. Figure 2 shows the procedure of airborne cloud seeding experiment for the reduction of the fine dust concentration.

Diffusion of the Seeding Material
Via numerical simulation, we confirmed the area over which the cloud seeding material spread and assessed the success of the cloud seeding experiment. The numerical simulations were performed using the weather research and forecasting (WRF) model (v. 3.8) based on the modified Morrison cloud microphysical method (Chae et al. 2018). Data from the unified model of the local data assimilation and prediction system analysis field of the Korea Meteorological Administration (KMA) obtained at 3-h intervals (resolution = 1.5 km) were used as the input data. Figure 3(a) and (b), corresponding to Experiments 1 and 2, respectively, shows the distribution of the vertically accumulated CaCl 2 from the surface up to an altitude of 3 km based on the numerical simulation model. The black line in this figure represents the aircraft seeding line, while blue points represent the locations of the automatic weather station (AWS) rain gauges of the KMA ( Fig. 3(a) and (b)). The simulation of the spread of the seeding material under the influence of southwest and west winds in Experiments 1 and 2 from the seeding area to Gyeonggi-do, Seoul, and Gangwon-do are shown in Fig. 3(c). From this figure, it is evident that seeding precipitation did not occur in all areas where the seeding material had spread. Therefore, the simulated diffusion ranges of the spread of the cloud seeding material were analyzed based on the difference between the SEED (seeding) and NOSEED (non-seeding) areas.
Further, Fig. 3(c) shows the 1-h cumulative precipitation difference between the SEED and NOSEED areas for Experiments 1 and 2. In this figure, SEED represents the results obtained following the diffusion of the seeding material in the experimental area, while NOSEED represents results obtained in the absence of the seeding material. Further, the difference between SEED and NOSEED (SEED − NOSEED) denoted the enhanced precipitation. Experiment 1 showed the diffusion range of the cloud seeding material in Seoul and northern Gyeonggido due to the southwesterly winds, while Experiment 2 showed the diffusion range of the cloud seeding material in Chungcheongnam-do due to the westerly winds.
The ascending air current in the experimental area and within the range of the simulated diffusion was analyzed given that it rapidly brought about an increase the concentration of cloud particles. This is an important factor in cloud seeding experiments because it promotes collision and coalescence between the seeding material and cloud particles (Silverman and Sukarnjanaset 2000). Thus, the ascending air current facilitates the rapid growth of cloud particles into precipitated particles (NIMS 2018). Figure 4 shows an ascending air current of 700 hPa between 11:00 and 19:00 KST on November 1, 2020. The air current ascended in the west sea and Seoul from 11:00 KST, and was particularly strong in northern Gyeonggido from 13:00 to 14:00 KST. Further, it affected the three areas until 16:00 KST, and gradually weakened from 17:00 KST onward. The path of this ascending air current was found to be consistent with the area where cloud seeding was conducted (Fig. 4). Thus, the rate of increase of the seeding-induced precipitation in the study area was numerically simulated.

Analysis of precipitation data
The observed AWS data were analyzed to verify the seeding effect on reducing fine dust concentration. The hourly cumulative precipitation before the start of the experiment from 13:00 to 20:00 KST is shown in Fig. 5(a). The precipitation band from the northwest affected Seoul, Gyeonggi-do, and Incheon at 13:00 KST. Further, this precipitation band strengthened from 13:00 to 15:00 KST, and reached 10 mm in Seoul. Further, the precipitation band coincided with the simulated time and space simulated for the seeding material (Fig. 3). Subsequently, it gradually weakened as it moved southeast.
The change in radar reflectivity in the seeding-affected area was also analyzed (Fig. 5(b)). Thus, we observed that the precipitation band moved eastward over time, and the radar reflectivity was strong in Seoul and Gyeonggi-do until 15:00 KST. Subsequently, it moved to southern Gyeonggido, after which also gradually moved southeast, affecting Chungcheong-do. Additionally, we observed that the radar reflectivity weakened until 18:00 KST, strengthened again at 19:00 KST, and then strengthened again in the form of a precipitation band in Chungcheong-do until 20:00 KST.

Analysis of airborne observation data
The KMA/National Institute of Meteorological Sciences (KMA/NIMS) atmospheric research aircraft (NARA) is equipped with a liquid water content (LWC)-100 sensor, cloud combination probe, and cloud condensation nuclei counter. Cloud and precipitation particles were observed in the sky above the experimental area using the abovementioned meteorological observation equipment, and cloud particle changes were analyzed by dividing them into three categories, namely, before, during, and after the seeding experiment.
The average concentrations of cloud, drizzle, and precipitation particles before, during, and after the seeding experiment were compared (Fig. 6). Data obtained based on Experiment 2 obtained during and after the experiment were analyzed making use of the same observation altitude; the observation altitude before the experiments was different. The cloud particles observed after both experiments were found to be the largest, with diameters of 10 µm to approximately 3,000 cm −3 . Additionally, the average concentration of the particles also increased after seeding. Similar observations were made for the drizzle particles. Further, the precipitation particles were observed to have larger diameters after seeding than before seeding, suggesting that the cloud and drizzle particles existed in the cloud before coming in contact with the seeding material and growing into larger precipitation particles.
The average concentrations (total number of aerosols for each particle size observed divided by the total number of observations) of cloud, drizzle, and precipitation particles were compared before, during, and after the experiment. In both experiments, the seeding level and altitude of the clouds differed depending on the location (Section 3.2). Therefore, the average concentrations of the cloud particles before and after Experiment 1 were compared since the same altitude was used for all the three instances (before, during, and after seeding). Data from Experiment 2 obtained before seeding was excluded from the analysis because the altitude for this instance differed from that employed during and after the experiment. Therefore, only average concentrations during and after the experiment (same observation altitude) were compared ( Table 2). The average concentration of fine particles was calculated using an LWC of 0.01 g·cm −3 or higher (Table 2). Thus, we observed that the overall average concentration of cloud, drizzle, and precipitation particles increased significantly after seeding compared to the observations made before and during seeding. In Experiment 1, the average concentrations of cloud particles before and after seeding were 373 and 891 cm −3 , respectively, i.e., an increase of approximately 140%. Further, the average concentrations of drizzle particles were 0.5 and 1.4 L −1 before and after seeding, respectively, showing an increase of 180%. Precipitation particles also showed an increase of 300% (from 0.1 L −1 before the experiment to 0.4 L −1 after the experiment). In Experiment 2, the average concentrations of cloud, drizzle, and precipitation particles during and after the experiments showed increases of 70%, 160%, and 200%, respectively.

Reduction of fine dust concentration
Numerical simulations of seeding material, ascending air current, radar data, and aircraft observation data were analyzed to confirm the effect of cloud seeding. The results obtained suggested that cloud seeding was useful for reducing fine dust concentration in Seoul, northern Gyeonggido, and Chungcheongnam-do. Therefore, PM 10 time-series data were analyzed for areas affected by the cloud seeding experiments.
The analysis of the distribution of PM 10 in the Korean Peninsula on November 1, 2020 (Fig. 7) showed that fine dust arrived the Korean Peninsula from the northwest. Thus, the PM 10 concentration started to increase in Seoul at 11:00 KST. The fine dust then moved southeast until 15:00 KST, affecting Chungcheong-do and Gyeongsangbuk-do. After 15:00 KST, it was also observed that fine dust further arrived from the northwest, resulting in another increase in PM 10 concentration in Seoul.
To analyze the PM 10 time-series due to cloud seeding, the influence of cloud seeding in Experiments 1 and 2 on find dust concentration was determined using the previously obtained simulation (Fig. 3) and ascending air current (Fig. 4) data. Specifically, data from PM 10 measurement stations in areas affected by the cloud seeding experiments corresponding to a period of up to 3 h, which is the reaction time of CaCl 2 , were selected. Further, PM 10 data corresponding to during Experiment 1 were selected for the time-series analysis of PM 10 concentration in Seoul, Goyang, and Gyeonggi-do, whereas PM 10 data corresponding to Experiment 2 were selected for analysis of PM 10 concentration in Asan, Hongseong, and Chungcheongnam-do. Simulation results showed that Taean in Chungcheongnamdo and Yangyang in Gangwon-do were not affected by the cloud seeding experiments. Rainfall, occurred during the experiment in Taean, but not in Yangyang. Figure 8 shows the area affected by the cloud seeding experiments as well as the locations of the PM 10 measurement stations. Figure 9 shows the PM 10 time-series data and numerically simulated precipitation for the cloud seeding-affected and unaffected areas. The time of the cloud seeding effect was determined with reference to the diffusion field of the seeding material based numerical simulation (Fig. 3), considering the time when the seeding material affected the area (Fig. 9). The reaction time of the seeding material was thus, noted. Even when the diffusion of seeding material in the area affected by cloud seeding was numerically simulated until a later time than the reaction time of the seeding material, it was uncertain whether cloud seeding affected the area at that time. Therefore, it was necessary to determine the time after which the effect of the cloud seeding could be observed by considering the maximum reaction time of the seeding material. The reaction time of CaCl 2 in this experiment was up to 3 h. Thus, PM 10 concentrations were analyzed for up to 3 h after the seeding experiment. Figure 9(a) shows the changes in PM 10 concentration in the area affected by cloud seeding in Experiment 1. Precipitation occurred at the SEED time at both the Seoul and Goyang stations. Specifically, in Seoul, after the seeding material was advected, the PM 10 concentration decreased until after 15:00 KST. This decrease continued until 16:00 KST, after which an increased was again observed. Similarly, the PM 10 concentration in Goyang decreased until 15:00 KST, after which it increased. The numerical simulation of the seeding materials was also conducted during this period.
The results obtained in Experiment 2 were similar to those obtained in Experiment 1 (Fig. 9b). Notably, in Experiment 2, precipitation occurred when cloud seeding started affecting the area. The PM 10 reduction rate observed in Experiment 2 was lower than that observed in Experiment 1. Moreover, PM 10 concentration decreased during the SEED period and increased thereafter in Hongseong and Asan. Thus, cloud seeding-induced precipitation brought about a decrease in PM 10 concentration. Figure 9c shows the changes in PM 10 concentration in the area unaffected by cloud seeding. In Taean, precipitation occurred during the SEED period ( Fig. 9(a) and (b)). However, the PM 10 concentration continued to increase, which can be attributed to thermal power plants in this area, whose emissions can affect PM 10 concentration more significantly than cloud seeding. Therefore, it was not possible to accurately determine the effect of the seeding experiment in Taean. Further, as precipitation did not occur in Yangyang, no significant differences in PM 10 concentrations during the SEED periods in Experiments 1 and 2 were observed. However, decreases in PM 10 concentrations under the influence of precipitation due to cloud seeding were observed.

Conclusions
Two cloud seeding experiments were conducted on November 1, 2020 to estimate the fine dust concentration reduction effect of induced precipitation. The results obtained can be summarized as follows: The numerical simulation of the spread of the seeding material in Experiment 1 showed that the seeding material spread throughout Gyeonggi-do, Seoul, and Gangwon-do under the influence of the southwest wind. Additionally, precipitation occurred in Seoul and northern Gyeonggi-do due to the seeding material during its estimated effect time. In Experiment 2, we observed that the seeding material spread to Chungcheong-do under the influence of the west wind, and precipitation occurred in some areas. In particular, an Radar reflectivity analysis and AWS data showed that the precipitation band strengthened as it moved toward Seoul, Gyeonggi-do, and Chungcheongnam-do due to the effect of cloud seeding. Specifically, Experiment 1 showed that the radar reflectivity strengthened in Seoul and Gyeonggi-do until 15:00 KST, and in Experiment 2, it weakened gradually Chungcheongnam-do until 18:00 KST but strengthened again at 19:00 KST, and affected this area until 20:00 KST.
Both Experiments 1 and 2 showed that the average concentration of cloud, drizzle, and precipitation particles observed using NARA increased after seeding compared with the observations made before seeding. In Experiment 1, the average concentration of cloud, drizzle, and precipitation particles increased by 140%, 180%, and 300%, respectively after seeding, and in Experiment 2, the average concentration of cloud, drizzle, and precipitation particles increased by 70%, 160%, and 100%, respectively, after seeding. In particular, precipitation particles had a larger diameter after seeding than before or during seeding.
In the areas affected by cloud seeding in the two experiments, PM 10 concentration decreased owing to the combined effect of precipitation and cloud seeding. In Experiment 2, PM 10 concentration decreased in areas with normal air quality affected by cloud seeding, but the reduction effect was not significant. Further, the cloud seeding effect was well demonstrated via the numerical simulation of the seeding material and the analysis of radar and aircraft observation data. Furthermore, the PM 10 concentration decreased during the effective time of the cloud seeding material, suggesting that fine dust can be reduced using this technique. Yangyang showed no changes in PM 10 concentration during the experiment.
Based on the analysis of numerical simulation data as well as ground and aircraft observation data obtained before, during, and after the seeding experiments, cloud seeding was found to be effective for reducing fine dust concentration. Changes in PM 10 were analyzed by selecting the areas that were affected and unaffected by cloud seeding. However, an accurate estimation of the decrease in the concentration of the fine dust particles owing to cloud seeding was not possible. This is because the experiment was conducted during a period of normal precipitation. Further, Experiment 2 could not be performed at a certain altitude because of the weather conditions in the experimental area; hence, the analysis of the aerial data before and after the experiment was limited. Additionally, in two experiments, the CCN (Cloud Condensation Nuclei) counter could not be used to observe clouds due to a temporary breakdown. Even though further experiments are required to obtain more comprehensive results, this study highlights the possibility of reducing fine dust concentration via cloud seeding.
Presently, it is not scientifically clear how to distinguish between natural rainfall and cloud seeding-induced rainfall (Bruintjes 1999). Notwithstanding, in this study, the occurrence of cloud seeding rainfall was suggested based on: (1) the growth of cloud particles after the experiment (Table 2 and Fig. 6) and (2) the spatio-temporal correspondence between natural rainfall and simulated rainfall (Figs. 3 and 5). As shown in Fig. 9, the simulated rainfall and AWSbased rainfall occurred during the same time period, and it was considered as cloud seeding rainfall though it is indirect verification.
In future studies, cloud seeding experiments at a certain altitude should be perform so as to obtain comprehensive aircraft data. Furthermore, additional experiments are required to develop strategies by which the effect of induced precipitation on reducing fine dust concentration can be enhanced. Therefore, additional research and technology are needed. For example, the establishment of a numerical model for verifying cloud seeding is required. It is also necessary to develop a strategy for the quantitative estimation of the reduction amount of fine dust resulting from precipitation and for the analysis of the precipitation components in the area affected by cloud seeding.