Abstract: The estimation of the aeroacoustic characteristics of the high-speed train is of great importance in train design. Due to the extreme slender shape of the full marshaling of the train (generally 8 or 16 cars), the calculation of the far filed acoustics involves massive consumption of the CPU time which even makes the optimization unpractical. This paper numerically studies the influence of train length on its aeroacoustics characteristics by considering the acoustic contribution car by car. The nonlinear acoustic solver (NLAS) is used to calculate the acoustic source at the acoustic surface, and the FW-H analogy method is used to integrate on the acoustic surface to obtain the far-field results. Three different marshaling of the trains, i.e. 3 cars, 4 cars, and 6 cars, are studied and compared. The far-field acoustic level and frequency profile along the train is obtained. The results show that for different middle cars, the far-field acoustic profile along the length is very similar in both quantity and shape, except the offset in position. For the cars in the same position of different marshaling, both the sound pressure level distribution and the frequency profile are also very close to each other. Thus the key aeroacoustics characteristics of a long marshaling train could be estimated with a much smaller marshaling such as 3 cars. With the superposition of the acoustic surface data from short marshaling simulation, the aeroacoustics characteristics of long marshaling could be obtained. The comparison of the superposed results and the results calculated directly from long marshaling simulation is close enough for engineering use. This demonstrates that the proposed novel approach for estimating aeroacoustics of long marshaling could not only reduce the computational cost significantly but also be as accurate as direct simulation. This paper might provide a handy tool for engineering practice in this region.