A regional wave forecasting system in East Asia, including the Korean Peninsula, was built based on WAVEWATCH III using offshore wind forecast data from the Global Data Assimilation Prediction System. The numerical simulations were performed on the sensitivity of the interaction between input wind and wave development. The forecasts for each condition were compared and verified with the observational data of marine meteorological buoys from 1 August to 30 September 2020. The sensitivity conditions were configured to have a specific range of variables related to the directional distribution of input winds (SINA0) and variables indicating the development of input wind–wave (CDFAC) in the ST6. The results were presented by calculating the mean error and root mean square error for all observation points. Overall, as the CDFAC increased, the mean error tended to decrease according to the forecast time and the root mean square error increased. Although the effect of SINA0 at the same CDFAC was insignificant, when SINA0 increased in sections where the significant wave height decreased rapidly, the significant wave height tended to decrease. In addition, the main variables that affect the physical process of wind–wave interaction should be considered to improve wave model forecasting performance and accuracy.

A wave forecast numerical simulation was performed for Typhoon Lingling around the Korean Peninsula and in the East Asia region using sea winds from 24 members produced by the Ensemble Prediction System for Global (EPSG) of Korea Meteorological Administration (KMA). Significant wave height was observed by the ocean data buoys used to verify data of the ensemble wave model, and the results of the ensemble members were analyzed through probability verification. The forecast performance for the significant wave height improved by approximately 18% in the root mean square error in the three-day lead time compared to that of the deterministic model, and the difference in performance was particularly distinct towards mid-to-late lead times. The ensemble spread was relatively appropriate, even in the longer lead time, and each ensemble model runs were all stable. As a result of the probability verification, information on the uncertainty that could not be provided in the deterministic model could be obtained. It was found that all the Relative Operating Characteristic (ROC) curves were 0.9 or above, demonstrating good predictive performance, and the ensemble wave model is expected to be useful in identifying and determining hazardous weather conditions.

Non-linear behavioral links with atmospheric teleconnections were identified between the Indian Ocean Dipole (IOD) mode and seasonal precipitation over East Asia (EA) using statistical models. The analysis showed that the lower the lag time, the higher the correlation; more than a two-fold correlation for non-linear regression with a kernel density estimator than for the linear regression method. When the IOD peaked, a pattern of significant reductions in seasonal precipitation during the negative IOD period occurred throughout the Korean Peninsula (KP). The occurrence of the positive IOD was in line with the El Niño phenomenon and generated greater seasonal precipitation than only the positive IOD, which takes place from March to May. This change occurred more in the cold tongue El Niño than the warm pool El Niño, inducing much higher spring precipitation throughout the KP. When negative IODs and La Niña coincided, there was slightly greater precipitation from March to May compared to the sole occurrence of negative IODs. In positive (negative) IOD years, there was anti-cyclonic (cyclonic) circulation in the South China Sea (SCS), helping to transport moisture to EA. The composite precipitation anomalies in the positive (negative) IOD years show above (below) normal precipitation in southern China. In contrast, other parts of the EA experienced drier (humid) signals than normal years. In positive IOD years, the anti-cyclonic circulation strength of the Bay of Bengal and the SCS continued until autumn and spring of the following year. This shows possible remote connections between climate events related to the tropical Indian Ocean and variations in precipitation over EA.

In this study, we constructed a rapid refresh wave forecast model using sea winds from the Korea Local Analysis and Prediction System as input forcing data. The model evaluated the changes in forecast performance considering the influence of input wind–wave interaction, which is an important factor that determines forecast performance. The forecast performance was evaluated by comparing the forecast results of the wave model with the significant wave height, wave period, and wave direction provided by moored buoy observations. During the typhoon season, the model tended to underestimate the conditions, and the root mean square error (RMSE) was reduced by increasing the wind and wave interaction parameter. The best value of the interaction parameter that minimizes the RMSE was determined based on the results of the numerical experiments performed during the typhoon season. The forecast error in the typhoon season was higher than that observed in the analysis results of the non-typhoon season. This can be attributed to the variations of the wave energy caused by the relatively strong typhoon wind field considered in the wave model.

To downscale artificial intelligence (AI) medium-range forecasts over the Korean Peninsula, we developed a 1.5 km CorrDiff system trained with ERA5 inputs and UM-LDAPS analyses as targets. For 1.5 m temperature and 10 m wind, the regression stage reduced root-mean-square error (RMSE) and root-mean-square vector error (RMSVE) relative to ERA5 and increased spatial correlation. CorrDiff had smaller absolute biases and greater spatial variability than Regression but higher RMSE and RMSVE and lower spatial correlation. The most consistent improvement relative to Regression occurred for high-temperature events at or above the target 99th percentile, with higher probability of detection (POD) and critical success index (CSI) and lower conditional temperature error. Without additional training or correction, AI Forecasting System (AIFS) forecasts were downscaled through +288 h. In a July 2026 heatwave case at 47 stations, CorrDiff reduced temperature errors and alleviated the underprediction of high-temperature occurrence relative to Regression. CorrDiff had higher POD and CSI than the AIFS for temperatures ≥33 °C but higher RMSE and lower spatial correlation. Generative downscaling should therefore be evaluated for spatial variability, spatial agreement, and event-specific performance beyond deterministic errors; producing high-resolution data does not necessarily imply improved forecast performance against observations.

The Korea Meteorological Administration (KMA) has developed and is currently operating a global wave-data assimilation system that utilizes real-time observational data to improve wave forecast accuracy. In recent years, the launch of five new polar-orbiting satellites for ocean observation has doubled the number of satellites available for global wave-data assimilation, from five to ten. This study integrated these five additional satellites into the existing global wave-data assimilation system and conducted a comparative analysis of the resulting initial wave conditions and significant wave height forecasts and those form the current operational wave model. The results indicate that the root mean square error of significant wave height improved by up to 7.2% for the initial conditions and by up to 3.4% for the 24-h forecasts. Notably, forecast accuracy in semi-enclosed sea areas, such as bays, showed a significant improvement compared to that in previous methods. Based on these findings, the continued expansion of satellite observational data utilization is expected to further enhance the forecast performance of KMA’s global wave model.