Abstract: Based on the frequency-wavenumber spectrum representation correlating with the standard orthogonal random variables, a hybrid approach of frequency-wavenumber spectrum and a random function for simulating the continuous spatio-temporal stochastic field is proposed by introducing the random function of standard orthogonal random variable sets. Meanwhile, the simulation efficiency of the proposed approach is greatly enhanced by employing Fast Fourier Transform (FFT) algorithm technique. Benefiting from the proposed approach, the probability characteristics of the spatio-temporal stochastic field can be described on the probability density level with only two elementary random variables. Therefore, the complete representative point sets with assigned probabilities of the elementary random variables can be obtained through the number-theoretical method. As a result, the dimension reduction representation of the continuous spatio-temporal stochastic field can be realized. Numerical examples indicate that when using the same number of samples and taking the efficiency and accuracy into consideration at the same time, the proposed approach have a similar simulation result to the conventional frequency-wavenumber spectrum representation. However, the smallest number of the elementary random variables is needed in the proposed approach, which leads to a smaller number of representative samples with a complete probability set. Consequently, it could naturally be combined with the probability density evolutionary method (PDEM) to carry out the accurate analysis of stochastic dynamic response and dynamic reliability assessment of engineering structures. Finally, combining the Kaimal fluctuating wind velocity spectrum with Davenport spatial coherence function, a numerical example of simulation for horizontal-fluctuating-wind velocity continuous stochastic field is presented to verify the accuracy and superiority of the proposed approach.