Much of this mismodeling occurs at the once-per-orbital-revolution frequency. Accordingly, constant and one-cycle-per-revolution empirical accelerations are used to accommodate the effect [ Colombo, 1989]. This technique has been used successfully for T/P in both the SLR/DORIS [ Tapley et al., 1994a] and the GPS [ Bertiger et al., 1994] dynamic orbit solutions. Furthermore, in the GPS ``reduced dynamic'' approach, the converged dynamic orbit is used as an a priori ephemeris and constant, three-dimensional stochastic accelerations spaced at short intervals are estimated, virtually eliminating any residual dynamic force modeling error in the resulting orbits [ Wu et al., 1991; Yunck et al., 1990, 1994]. The extensive temporal and spatial tracking coverage provided by DORIS and GPS permit this routine recovery of extensive empirical/stochastic force model parameters. The use of these empirically determined effects, although effective, should not be considered as a substitute for correct modeling and can misposition the satellite during times of data anomalies. These approaches have allowed sub-4 cm radial accuracy to be achieved on T/P. Empirical parameterization in the dynamic approach is not a panacea for dynamic model deficiencies, however, since these parameters can absorb the signals arising from the geophysical processes one is trying to observe.