AVO analysis has been used with some success in seismic exploration to directly detect the presence of hydrocarbons. AVO inversion essentially implies a least squares fitting of reflection coefficients (seismic amplitude) as functions of source-receiver offsets in the moveout corrected seismograms, assuming that the background velocity is known accurately. Unlike conventional approaches, we carry out the background velocity and AVO inversion in the plane wave (intercept time-ray parameter or T-p) domain. Normal moveout analysis in the plane wave domain results in interval velocity estimates and the T-p data are closer approximations to the plane wave reflection coefficients. Having determined background velocity and fractional changes from an AVO inversion, we carry out a full wave from inversion in which we use full elastic waveform modeling that includes all internal multiples and converted waves. We apply this multi-stage seismic waveform inversion approach to a suite ofCMP gathers from a 2D seismic line collected offshore of the east coast of the United States; a region in which all occurence of gas hydrates has been reported. Gas hydrates have the economic potential of being tapped as a fuel source and also have the potential to act as a greenhouse agent if freed into the atmosphere. In seismic sections, the base of gas hydrate zone is marked by bright, highamplitude relfections, which follow the seafloor topography and are called bottom-simulating reflectors (BSR). BSRs have negative polarity with respect to the seafloor reflection and in a common shot or a CDP gather, the amplitude increases with offset. Our analysis was aimed at deriving a high resolution seismic velocity structure for the gas hydrates and sediments below. At locations where a BSR exists, we identify a low velocity zone that coincides with the BSR. We also identify several thin low velocity zones beneath the BSR, interpreted to be due to the presence of free gas. We compare and contrast our results with the velocity function derived from zero-offset VSP data collected during the ODP drilling Leg 164 at holes 997 located NE of our seismic line. The general trend of the two independent estimates of velocity is in good agreement. The lowP wave velocity zones show no change of shear wave velocity, indicating the presence of free gas. This was confirmed by drilling in the nearby area. However, the VSP-derived velocity model was obtained by the application of smoothing in the traveltime inversion of the VSP data in which only a smooth velocity model was sought. The resulting model shows a nearly 200m thick low velocity zone (continuous free gas) which may have been caused by artifacts due to smoothing. Unlike the VSP model, our result shows several thin low velocity layers.