The performance of five global Earth gravitational models, published
after 1995, was examined through tests with data (mostly) withheld from the
development of these models. We considered the models: JGM-3 (Tapley et al.,
1996), GRIM4-C4 (Schwintzer et al., 1997), TEG-3 (Tapley et al., 1997), EGM96
(Lemoine et al., 1998) and GPM98A (Wenzel, 1998). The test data that we used for
model evaluations include satellite tracking measurements acquired over several
spacecraft at various inclinations and altitudes, geoid undulations ( or height
anomalies) determined from GPS positioning and leveling observations, Dynamic
Ocean Topography (DOT) information implied by an ocean circulation model, as
well as hydrographic estimates of (relative) DOT. Over 9307 GPS/leveling geoid
undulation values distributed over North America, Europe and Australia, EGM96 (to
degree 360) outperforms all other models tested, yielding a standard deviation of the
undulation differences of ±37.2 cm. Considering that the available GPS/leveling
data are located over some of the best surveyed areas (gravimetrically), this value is
consistent with the predicted ( commission plus omission) geoid error of EGM96,
whose global rms value equals ±45.3 cm. Over the ocean, the performance of
EGM96 is superior to that of a11 other models tested, as judged by the results of
comparisons with both the POCM_4B circulation model DOT output and with the
hydrographic DOT estimates. GPM98A was found to be inaccurate over medium
wavelengths, and is not considered suitable for orbit determination applications.