Accurately simulating near-surface wind speeds is indispensable for wind energy development, particularly under extreme weather conditions. This study utilizes the Weather Research and Forecasting (WRF) model with a 6 km resolution to evaluate 80 m wind speed simulations over Europe, using the ECMWF (European Centre for Medium-Range Weather Forecasts) reanalysis version 5 (ERA5) as initial and lateral boundary conditions. Two cases were analyzed: a normal case with relatively weak winds, and an extreme case with intense cyclonic activity over 7 days, focusing on offshore wind farm regions and validated against Forschungsplattformen in Nord- und Ostsee (FINO) observational data. Sensitivity experiments were conducted by modifying key physical parameterizations associated with wind simulation to assess their impact on accuracy. Results reveal that while the model realistically captured temporal wind speed variations, errors were significantly amplified in extreme cases, with overestimation in weak wind regimes and underestimation in strong winds (approximately 1–3 m/s). The Asymmetrical Convective Model 2 (ACM2) planetary boundary layer (PBL) scheme demonstrated superior performance in extreme cases, while there were no significant differences among experiments under normal cases. These findings emphasize the critical role of physical parameterizations and the need for improved modeling approaches under extreme wind conditions. This research contributes to developing reliable wind speed simulations, supporting the operational stability of wind energy systems.

Accurate day-ahead wind speed forecasting is essential for wind power forecasting and electricity market participation under increasing renewable energy penetration. This study proposes a hybrid forecasting framework that combines raw global forecasts from GFS and IFS, the KMA KIM-RDAPS regional forecast, and dynamically downscaled GFS/IFS forecasts generated with alternative boundary-layer physics. Seven forecast members were synthesized using arithmetic averaging, performance-weighted averaging, and LightGBM-based machine learning (ML) regression. The framework was evaluated over Jeju Island, Republic of Korea, using 10 m Automatic Weather Station observations from 2023 to 2024 and 80 m meteorological mast observations from 2023. For the AWS evaluation, 2023 was used for training and validation, and 2024 was reserved for independent testing. The site-specific LightGBM synthesis achieved the most consistent improvement, reducing the median site-wise MAE across 31 AWS sites to 0.90 m s⁻¹, corresponding to a 39.2% improvement relative to the best non-downscaled member and 47.2% relative to the unweighted multi-model mean. In the 80 m mast-based diagnostic assessment, the same approach reduced derived normalized power MAE to 11.4%. These results indicate that ML synthesis of multi-source NWP forecasts can improve day-ahead wind speed and power-oriented forecast information over complex island terrain.

The roles of the drop size distribution (DSD) and turbulence in the autoconversion rate A are investigated by analyzing Lagrangian cloud model (LCM) data for shallow cumulus clouds. The correlations of DSD and turbulence with other cloud parameters are estimated, and they are applied to parameterize their effects on A. A new parameterization of A is proposed based on it, as A = αqc7/3Nc-1/3H(Rc-Rc0) with α = aNc-X(Rc-Rc0)(1+bε), where qc, Nc, and Rc are the mixing ratio, the number concentration, and the volume mean radius of cloud droplets, respectively. ε is the dissipation rate, Rc0 is the threshold value of Rc, H is the Heaviside step function, and X, a, and b are constants. Here, Nc-X(Rc-Rc0) represents the effect of DSD, via its correlation with Nc and qc, while A ∝ qc7/3Nc-1/3 represents the effect of the gravitational collisional growth for given DSD and turbulence. The correlation between turbulence and DSD makes b larger than expected based on turbulence-induced collision enhancement. The effects of DSD and turbulence and their correlations with qc and Nc explain a wide range of exponent values of qc and Nc in many existing parameterizations of A. The new parameterization is compared with the LCM data and applied to a bulk cloud model (BCM) while clarifying the difference between the cloud droplet mixing processes of the LCM and BCM. The importance of DSD and turbulence in the raindrop formation in shallow cumulus clouds are shown by comparing the results from A with and without these effects.

The activation of aerosol particles and its contribution to cloud droplet size distributions in shallow cumulus clouds are investigated using a Lagrangian cloud model (LCM), allowing explicit simulations of the activation and tracking of each simulated hydrometeor. This analysis is focused on the differences in activations for aerosol particles following central updraft from a cloud base (CB) and particles laterally entrained (LE) aloft. Strong supersaturation fluctuations induced by entrainment and turbulent mixing at cloud edges cause most LE particles to be deactivated soon after activation, substantially limiting their lifetime and growth relative to CB particles. Most net activation—the difference between activation and deactivation—occurs at the cloud base and vanishes aloft, thus making the cloud droplet concentration vertically uniform above the cloud base. Time series of various variables following individual particles confirm the distinctive nature of CB and LE activation. Activation spectra follow the theoretical prediction at high supersaturation conditions, but their magnitudes are smaller because of particle deactivation. At low supersaturation conditions, the spectra deviate from the theoretical prediction since they are affected by the transport of the CB particles activated elsewhere.