Accuracy of the earth's gravity field models

Abstract

Accuracy tests on the most recent GEM (Goddard Earth Model) gravity models for the representation of the Earth's gravity field, using specially devised statistical techniques of comparative evaluation, show that there is steady improvement in these models with time. On this comparative basis, the accuracy of determination for the spherical harmonic coefficients of the Earth's gravity field is ∼ 100% for n = 2–6, 90–99% for n = 7–10, 55–80% for n = 11–14 and ⪡ 50% for n ⩾ 15, deteriorating rapidly with increasing n. The higher degree coefficients corresponding to n ≥ 15 do not seem to be determined accurately enough to be useful from a geophysical standpoint, though their cumulative contribution is undoubtedly useful for specific orbital computations. The estimated errors are 0.3 mGal for n = 2–6, 1.5 mgal for the frequency range n = 2–10, 3 mGal for n = 2–14 and 5–6 mGal for n = 2–22. These error estimates, especially the ones for the higher frequency range, may have been affected by possible errors in the comparison standards used for this evaluation. Consequently, some of the higher degree coefficients of recent GEM models may be more accurate than predicted by these tests. Due to the inherent deficiency of the comparison standards, the errors given in this paper should be treated as error estimates. The steady and progressive improvement, shown by the various GEM gravity models when tested against comparison standards 10E and WGS 72, i.e. the more recent a gravity model, the better it tests against the comparison standards in contrast to its predecessors, is remarkable, as the comparison standards themselves are several years older than the gravity models tested here. This clearly validates our choice of comparison standards, as well as the premises and predictions of our evaluation techniques. It also demonstrates the power and potential of these techniques, which only seem to be limited by the level of accuracy of the available standard of comparison.