A model was developed for simulating the effects of airborne silver iodide (AgI) glaciogenic cloud seeding using the weather research and forecasting (WRF) model with a modified Morrison cloud microphysics scheme. This model was used to hindcast the weather conditions and effects of seeding for three airborne seeding experiments conducted in 2016. The spatial patterns of the simulated precipitation and liquid water path (LWP) qualitatively agreed with the observations. Considering the observed wind fields during the seeding, the simulated spatiotemporal distributions of the seeding materials, AgI, and snowfall enhancements were found to be reasonable. In the enhanced snowfall cases, the process by which cloud water and vapor were converted into ice particles after seeding was also reasonable. It was also noted that the AgI residence time (>1 hr) above the optimum AgI concentration (10⁵ m⁻³) and high LWP (>100 g m⁻²) were important factors for snowfall enhancements. In the first experiment, timing of the simulated snowfall enhancement agreed with the observations, which supports the notion that the seeding of AgI resulted in enhanced snowfall in the experiment. The model developed in this study will be useful for verifying the effects of cloud seeding on precipitation.

In this study, a method for verifying the effect of cloud seeding in the case of a mixture of natural and artificial rainfall bands was proposed, and its applicability to each experimental case was evaluated. Water resources that could be secured through cloud seeding were also quantified for the experiments on forest fire prevention, drought mitigation, and dust reduction in 2020. Data on numerical simulations, radar-derived rainfall, rain gauge-derived rainfall, and weather conditions were applied. Areas with seeding and nonseeding effects were classified according to the numerical simulation results and wind system, and enhanced rainfall was determined by comparing the changes in rainfall between the two areas. The amount of water resources was determined by considering the area of the seeding effect and rainfall density. As a result, 1.74 mm (4.75 million tons) of rainfall increased from the experiment on forest fire prevention, 0.84 mm (1.30 million tons) on drought mitigation, and 2.78 mm (24.44 million tons) on dust reduction. Thus, an average rainfall of 1.0 mm could be achieved through the experiment. These results helped verify the pure seeding effect and achieve the experimental purpose.

Ground-based glaciogenic seeding experiments were conducted at the Daegwallyeong Cloud Physics Observation Site (CPOS) from 2012 to 2015 for the target area Yongpyeong, which lies 9 km away. The preseeding (NOSEED) and seeding (SEED) periods were defined based on the simulation results of AgI concentration (>10 L⁻¹) in the Weather Research and Forecast (WRF) model with the modified Morrison scheme in microphysics. It was difficult to determine whether snow enhancement via seeding occurred over the entire target area due to uncertainties associated with limitations such as observations and numerical model based on only two points (seeding and target sites). However, in three of four cases, the vertical reflectivity from micro rain radar, total concentration, and average size of snow particles observed at PARSIVEL and precipitation increased in the seeding effect time. In two of four cases, the simulated increased precipitation during the seeding effect time was also observed. In one case that did not show changes after seeding, it is analyzed that a sufficient cloud depth was not supplied to the seeding region due to the blocking effect of the Taebaek Mountains.

This study analyzed 10-h dead fuel moisture content (DFMC) data collected throughout the year 2024 at hourly intervals from 11 observation stations in Gangwon Province, South Korea, and developed a random forest–based prediction model. This model was chosen for its robust ability to capture the nonlinear relationships between diverse environmental factors and fuel moisture content, offering high predictive accuracy and interpretability. The 10-h DFMC exhibited distinct seasonal patterns, with notably low values in spring (March–May) and December, corresponding to periods of high forest fire occurrence on the Korean Peninsula. Topographical influences were evident; the highest-elevation station (DG) maintained high moisture content in February–March due to snow effects, while coastal stations (BG, GN, YG) showed greater variability in February compared to inland stations (BC, CS), likely due to frequent humidity fluctuations from sea fog and sea/land breezes, as well as higher precipitation frequency, highlighting the importance of maritime influences. The random forest model demonstrated good performance with R ² = 0.80, root-mean-square error (RMSE) = 2.73%, and mean absolute error (MAE) = 1.93% in the test dataset. Variable importance analysis revealed that relative humidity was the most influential factor (0.7562), followed by hour (0.0948), temperature (0.0506), and Julian day (0.0499), while precipitation showed unexpectedly low importance (0.0042). Model performance varied by location, with the best results at BP ( R ² = 0.86, RMSE = 2.10) and MO ( R ² = 0.85, RMSE = 2.17) stations, and the poorest at the high-elevation DG station ( R ² = 0.76, RMSE = 3.18). The developed model provides a foundation for effectively predicting fuel moisture content in areas without real-time 10-h DFMC monitoring by using only meteorological data, though future improvements should incorporate additional topographical and environmental variables to enhance prediction accuracy.