An integrated method consisting of a proper orthogonal decomposition (POD)-based reduced-order model (ROM) and a particle filter (PF) is proposed for real-time prediction of low-level turbulence. The proposed method is validated using identical twin experiments of the low-level turbulence case at Shonai airport in Japan. The results of the experiment show ROM with the estimated temporal coefficients by PF reproduces the wind velocity fluctuations similar to those produced by large eddy simulation with much lower computational cost. In addition, the time evolution of the wind field obtained by ROM-PF show that turbulent spatial structure is expressed accurately in comparison with the reference flow field. It is confirmed that the root mean square error of head and crosswind velocities decreased over time by repeating data assimilation.