Intelligent optimizing WRF model parameters during typhoon progress via genetic intelligent algorithm

IntroductionCoastal waters of China are frequently affected by typhoons, which pose serious threats to offshore wind power, marine engineering, and coastal disaster prevention and mitigation. Therefore, accurate simulations of 10 m wind speed and sea-level pressure are of great significance for typhoon risk assessment and marine engineering safety.MethodsIn this study, a genetic algorithm was coupled with the WRF model to optimize physical parameterization schemes for typhoon hindcasting over the Taiwan Strait and adjacent waters. The root mean square error of buoy-observed 10 m wind speed was used as the fitness function, and three representative typhoons, BAILU, Hagupit, and HAIKUI, were optimized separately.ResultsThe results show that the optimal scheme combinations for the three typhoons were WSM6-Tiedtke-YSU, Lin-KF-MYJ, and WSM5-KF-MYJ, respectively. Compared with commonly used schemes, the GA_OPT scheme reduced the wind speed RMSEs of the three typhoons to 3.01, 1.93, and 2.49 m/s, respectively, and generally improved the simulation of sea-level pressure. Track-error analysis indicates that GA_OPT did not achieve the smallest track RMSE, suggesting that its improved wind speed simulation mainly resulted from the enhanced representation of wind-field structures by the optimized physical schemes, rather than from reduced track bias.DiscussionThis study demonstrates that the GA-WRF framework can efficiently optimize WRF physical parameterization schemes and provide more reliable typhoon wind-pressure hindcasting support for coastal disaster prevention and marine engineering applications.