Abstract: Owing to the problem that the computational efficiency of traditional spectral representation (SR) method by Deodatis is restricted to the times of Cholesky decompositions, Hermite interpolation is included in the decomposition of cross-spectral density matrix. A reduced method for wind field simulation is derived accordingly. With the reduced method, the time of Cholesky decomposition in the traditional SR method can be decreased from n×N to n×2k (2k<N), hence the efficiency will be sharply increased. Taking the wind field simulation of the girder of a long-span triple-tower suspension bridge as an example, a 4096s-long fluctuating wind field is simulated by traditional SR method, third-order Lagrange interpolation method and the derived Hermite interpolation method. The comparison of the computational efficiency and its corresponding accuracy indicate that both the third-order Lagrange interpolation method and the Hermite interpolation method can prominently improve the computational efficiency of SR method. The efficiency of the Hermite interpolation method is slightly lower than that of the third-order Lagrange one, but its computing accuracy of Matrix H is on a much higher level which shows great superiority in wind field simulation. When the reduced method with Hermite interpolation is adopted, both the power spectral density and correlation functions of the simulated wind field can well conform to the targets. Based on the work above, the suggested boundaries for the interpolation intervals are given according to an optimization analysis.