2012 SEG. In the shallow water environment typical of Persian Gulf, the strong heterogeneities of subsurface and very shallow water depth (~10m) give rise to occurrence of surface waves. The presence and dispersion of surface waves depreciate the quality of geophone data to a greater extent than pressure data. Well-known conventional filtering techniques fail to sufficiently attenuate unwanted energy because these waves can resemble body-wave reflections, especially in horizontal components. Therefore, the low quality geophone data further impacts the applications of processing methods such as wavefield decomposition and subsurface imaging. Considering that seismic signals are non-stationary and surface waves have various spectra over time, we in this paper explore the differences in time-frequency variations of different seismic events to analyze surface waves. We propose a non-stationary FK filter to eliminate the effect of the surface wave from OBC record using variant-factor S transform. Analysis on field data shows that the proposed filtering could separate reflected waves from surface waves successfully on OBC geophone recordings, which could provide better data for more accurate seismic reservoir characterization.