The impact of GPS radio occultation (RO) refractivity measurement on cyclone/ typhoon weather prediction was explored. The nonlocal operator that assimilates the integrated retrieved refractivity along an assumed straight raypath was employed in WRF 3DVAR, which provides increments at the model initial time. We have reviewed general evaluation of such impact on Asian cyclone/typhoon predictions in recent years (2007 and 2008) and do find some positive or marginal benefits when the available GPS RO data have been assimilated. The impact on track prediction of the Cyclone Gonu (2007) over Western Indian Ocean appears appealing when 56 GPS RO soundings are assimilated. A recent typhoon Fung-Wong, which made landfall at southeast Taiwan near the end of July 2008, was also simulated and the prediction of local severe rainfall over Taiwan was improved as well when available GPS data were assimilated. When extended to the other cyclone such as the disastrous Myanmar Cyclone Nargis (2008), the initial cyclone was not well recovered by assimilation of GPS RO soundings at remote distances from the cyclone, in spite of inclusion of an initial bogus vortex or a cycling assimilation. Thus, the sensitivity to different resolution and sounding locations of GPS RO data for a targeted typhoon has been investigated, based on an Observing System Simulation Experiments (OSSEs) study for Typhoon Krosa (2007) in which the RO soundings surrounding the typhoon center are generated using a 2-D raytracing model with the atmospheric state from a high-resolution MM5 model run. These soundings are then assimilated into the degraded model initial conditions to recover the “observed” typhoon structure. The OSSE results substantiate that the biased track during landfall on Taiwan can be improved after assimilation of these soundings. This study indicates an unequivocal impact of dense or diluted GPS RO soundings on typhoon prediction.