Degradation of ocean signals in satellite altimetry due to orbit error removal processes

Abstract

The accuracy of recovering sea level changes with a satellite altimeter depends critically on the error in the satellite's ephemeris and how it is removed from the height data. We examine two global ocean data sets of sea level changes:(1) climatologic monthly hydrographic variations and (2) a 1‐year wind‐driven general circulation model (GCM). We then assess their degradation after orbit error removal is applied to the two data sets. The simulations use along track data sampling for about a year of the Geosat exact repeat mission (ERM). The simulated orbit error removals are made from differences of overlapping (or collinear) passes of the simulated sea heights. Radial orbit correction algorithms include along‐track polynomials of up to one half revolution in length and global sinusoids of from one half revolution to about 4 days long with a 1 cycle/revolution (cpr) fundamental period. Orbit error removal first degrades the ocean signal along track, usually in a mild way depending on the pass length and the seasonal differences between the collinear data sets. However, the resulting distortion of yearly time series at fixed locations is generally more severe, especially when the actual signal is weak and at higher latitudes. Average degradation of both climatologic and GCM time series at fixed locations is significant in all cases and more severe with local pass polynomials than with global sinusoids. Thus we find that in even the least destructive along‐track error removal process, the global error of the derived climatologic time series (defined as rms simulated‐orbit‐error/rms signal) is greater than 0.30. However, the correlation of time series, derived versus original, and their power ratios (rms derived/rms signal) is generally high (greater than 0.78 globally) for all methods, though certain locations result in negative correlation (reversal of phase). Degradation of the more variable GCM changes (with mesoscale eddies) is generally not as severe as in climatologic series. Thus the GCM time series error distortion with long arc 1‐cpr sinusoid orbit error removal was found to be only 0.20, while with full pass bias and tilts the error was 0.53. The corresponding global errors found for climate time series were 0.31 and 0.54. Similar results should be expected for any satellite mission using collinear or crossover altimetry spanning appreciable oceanographic change.