The laser-based short-are technique has been developed in order to avoid lo
cal errors which affect the dynamical orbit computation, such as those due
to mismodeling in the geopotential. It is based on a geometric method and c
onsists in fitting short arcs (about 4000 km), issued from a global orbit,
with satellite laser ranging tracking measurements from a ground station ne
twork. Ninety-two TOPEX/Poseidon (T/P) cycles of laser-based short-are orbi
ts have then been compared to JGM-2 and JGM-3 T/P orbits computed by the Pr
ecise Orbit Determination (POD) teams (Service d'Orbitographie Doris/Centre
National d'Etudes Spatiales and Goddard Space Plight Center/NASA) over two
areas: (1) the Mediterranean area and (2) a part of the Pacific (including
California and Hawaii) called hereafter the U.S. area. Geographically corr
elated orbit errors in these areas are clearly evidenced: for example, -2.6
cm and +0.7 cm for the Mediterranean and U.S. areas, respectively, relativ
e to JGM3 orbits. However, geographically correlated errors (GCE) which are
commonly linked to errors in the gravity model, can also be due to systema
tic errors in the reference frame and/or to biases in the tracking measurem
ents. The short-are technique being very sensitive to such error sources, o
ur analysis however demonstrates that the induced geographical systematic e
ffects are at the level of 1-2 cm on the radial orbit component. Results ar
e also compared with those obtained with the GPS-based reduced dynamic tech
nique. The time-dependent part of GCE has also been studied. Over 6 years o
f T/P data, coherent signals in the radial component of T/P Precise Orbit E
phemeris (POE) are clearly evidenced with a time period of about 6 months.
In addition, impact of time varying-error sources coming from the reference
frame and the tracking data accuracy has been analyzed, showing a possible
linear trend of about 0.5-1 mm/yr in the radial component of T/P POE.