Abstract: Seasonal signals in GNSS station coordinate time series are commonly modeled using harmonic functions with constant amplitude and phase. However, the amplitude and phase of seasonal signals are time-varying and show modulation characteristics. We propose a method to accurately estimate modulated seasonal signals in GNSS reference stations using SRIF algorithm. We here study the vertical coordinate time series of GNSS reference stations. By adjusting the standard deviation of process noise, we successfully estimate modulated seasonal signals. Through analyzing and comparing the PSDs of residuals obtained from SRIF and Hector software using different seasonal signal modeling methods, we confirm that the optimal value of the standard deviation of process noise is 0.01 mm for the annual signals. We also investigate the effects of two seasonal signal modeling methods on estimating the linear velocities, and it reveals that the maximum difference in vertical velocity estimation can reach 0.34 mm/a. This indicates that estimating modulated seasonal signals is of great significance for accurately extracting linear velocities and establishing a more precise International Terrestrial Reference Frame.