Abstract: Standalone mesoscale numerical weather prediction models often neglect two key aspects when simulating wind speed: first, the influence of sea surface changes due to wave propagation and current evolution on wind turbulence; second, the ability to provide second-scale fluctuating wind speeds. To address these limitations, this study develops a multi-scale nested wind-wave-current numerical simulation framework based on WRF-SWAN-ROMS-LES. Using this approach, we compared and analyzed the differences in wind speed simulation between a coupled air-sea mesoscale atmospheric model and a standalone atmospheric model. The effectiveness of microscale large-eddy simulation (LES) in reproducing turbulent fluctuations is also verified. Results show that the wind-wave-current coupled model achieves higher accuracy in characterizing offshore wind speed evolution. Specifically, at a height of 38 m, the standard deviation between the coupled model and radar observations is 8.8% lower compared to the standalone atmospheric model. In the microscale simulation, the introduction of three-dimensional spatiotemporal information such as mesoscale atmospheric velocity, potential temperature, and pressure successfully drives the small-scale large-eddy simulation. The output consistently lies within the central region of the LES fluctuating wind speed curve while maintaining second-scale high-frequency fluctuation characteristics.