Estimating the noise in space-geodetic positioning: the case of DORIS: Fid-Bau Portal

Estimating the noise in space-geodetic positioning: the case of DORIS

Le Bail, K.

Abstract The noise spectrum in DORIS ground- station motion is investigated by means of the Allan variance method applied to the decomposition of the 3D signal into its principal components in the time domain. Sets of weekly position time-series from 1994 to 2005 derived by three IDS Analysis Centres (IGN-JPL, INASAN, and LEGOS-CLS) for 119 stations at 69 sites are considered. The observing satellites are SPOT-2, SPOT-3, SPOT-4, and SPOT-5, TOPEX/Poseidon, and ENVISAT. Annual and semi-annual perturbations, as well as the 117.3-day term associated with the TOPEX/Poseidon orbit, are found at most stations. Their amplitudes reach up to 19.3, 23.7, and 13.3 mm, respectively, for the three analysis centres (ACs). When corrected for these components and a linear drift, the time-series dominantly show white noise (WN) at the 10–45mm level the noise level is the highest in the East direction, probably in connection with the high orbit inclinations. The noise level is minimum for the high latitude stations, mostly and intensively observed by the SPOT satellites, and the determination of the noise type is unclear; longer observation spans would be needed to decide between interannual variations and flicker noise. The improvement in positioning due to the DORIS constellation extension from three to five satellites in 2002, and the network rejuvenation program initiated in 2000, results in a decrease of the noise level by a factor of 1.7 in a WN context, both before and after the changes. One example of the benefit of studying the signal in the time eigenspace domain is the detection of anomalously large WN in the East direction for station HBKB (Hartebeesthoek, Africa) that masks the above-mentioned improvement. Studying the projection on the local frame of the second and third time-eigenspace components, a noise excess is detected in the North direction for some of the ACs. Station stability derived from our time-series analysis confirms, in general, the expected performance based on the careful technical review of the station components (antenna, pillar, etc.). The respective merits of our noise qualification method, based on direct time-series analysis in the time-eigenspace domain without any a priori statistical model, in comparison with other methods, such as the selection of a mixed-noise model by maximum likelihood estimation, are discussed.