The dynamic response analysis of structures subjected to a stochastic wind field is carried out in the time domain by a step-by-step integration approach. The loading is represented by simulated time histories of the aerodynamic force. The auto-regressive and moving average (ARMA) recursive models are utilized to simulate time series of wind loads. Depending on the system dynamic characteristics, the time-integration schemes require that the time increment should not exceed a prescribed value. This study focuses on the development of ARMA representation based procedures to simulate realizations of wind loading with small time increments required by the time-integration schemes. A three-stage-matching method and a scheme which combines ARMA and digital interpolation filters are presented to efficiently generate correlated time histories of wind loads at the prescribed time increments.