The background anelastic model properties are tabulated in Table 1 and plotted in Figure . The model consists of 11 flat layers, each specified by a P-wave and S-wave velocity (Vp, Vs), density () and P-wave Q value. The Vp values were obtained from a combination of the averaged CMP stacking velocities (after Dix inversion) and the sonic log at the well. The Vs values were calculated above the reservoir using an average Vp/Vs ratio of 2:1, in the absence of hard shear velocity data. The Vs values in the reservoir were extrapolated from values given in an internal ARCO report, based on the same prospect area and similar reservoir conditions (Foster et al., 1990). The density values were calculated using averages from the density well log, and values predicted as a function of the Vp estimates (Gardner et al., 1974). The Q values were predicted as a function of the Vp values (Waters, 1981), except for the reservoir gas sand which was assigned a value of 25 (Batzle, 1991, pers. comm. ). The Q values were qualitatively calibrated by estimating two key normal incidence reflection coefficients from the well logs, (the reservoir top at 2960 m depth and a large shallower reflection at 1960 m depth), forward modeling their amplitude response, and comparing to observed amplitudes in the raw shot records. This calibration qualitatively justified the Q values predicted by the Waters relation.