Dual-satellite crossover latitude-lumped coefficients, their use in geodesy and oceanography

Latitude-lumped coefficients (LLC) are defined, representing geopotential-orbit variations for dual-satellite crossovers (DSC). Formulae are derived for their standard errors from the covariances of geopotential field models. Numerical examples are presented for pairs of the altimeter-bearing satellites TOPEX/Poseidon, ERS 1, and Geosat, using the error matrices of recent gravity models. The DSC, connecting separate missions, will play an increasingly important role in oceanography spanning decades only when its nonoceanographic signals are thoroughly understood. In general, the content of even the long-term averaged DSC is more complex then their single satellite crossover (SSC) counterpart. The LLC, as the spatial spectra for the geopotential-caused crossover effects, discriminate these source-differences sharply. Thus, the zero-order LLC in DSC data contains zonal gravity information not present in SSC data. In addition, zero- and first-order LLC of DSC data can reveal a geocenter discrepancy between the orbit tracking of the separate satellite missions. For example, DSC analysis from orbits computed with JGM 2 show that the y-axis of the geocenter for Geosat in 1986–1988 is shifted with respect to T/P by 6–9 cm towards the eastern Pacific. Also, where the time-gap is necessarily large (as between, say, Geosat and T/P missions) oceanographic (sea-level) differences in DSC may corrupt the geopotential interpretation of the data. Most importantly, as we illustrate, media delays for the altimeter (from the ionosphere, wet troposphere and sea-state bias) are more likely sources of contamination across two missions than in SSC analyses. Again, the LLC of zero order best shows this contrast. Using the higher-order LLC of DSC for both Geosat-T/P and ERS 1-T/P as likely representation of geopotential-only error, we show by comparison with the predicted standard errors of JGM 2 that the latter's previously calibrated covariance matrix is generally valid. Received: 14 February 1996 / Accepted: 27 March 1997