Reducing Orbit Error for a Better Estimate of oceanic Variability from Satellite Altimetry

Abstract

The variable ocean dynamic topography is generally estimated from the satellite altimeter signal once the orbit error has been removed. To compute the orbit error, the most conventional technique is to fit a polynomial function (zeroth, first, or second degree) over lengths of several thousand kilometers to each altimetric profile. However, the method induces significant errors. To reduce them, one needs a more detailed representation of the orbit error spectrum and to take account of the spatial and temporal characteristics of the signal and noise. This can be achieved by the form of optimal analysis known as ?inverse theory.? If a realistic statistical description of the altimeter signal components (i.e., oceanic variability and orbit error) is provided, the inverse formalism optimally separates the components. Although the whole set of altimeter data is reduced to the data at the intersections of ascending and descending ground tracks (crossover points), the method remains quasi-optimal. The authors highlight the effectiveness of the method by applying it to the altimeter data for the Brazil-Malvinas confluence area, a few thousand kilometers wide. The authors compare the orbit error estimates to those of the most conventional method that is a method set to a similar environment (short-are analyses). With a homogeneous oceanic variability of 15 cm rms and a nominal orbit error of 30 cm rms, the error on the estimation is reduced to 2 cm all along the altimetric profiles. Taking into account the nonhomogeneous characteristics of the variability signal improves the estimation. It can he further improved simply by adding to the selected altimeter dataset the crossover points one orbital revolution away. For the Geosat satellite, they are at the same latitude but 25°25;prime; farther west or cast. The results encourage the use of the inverse method for orbit error reduction. The method is good at separating signals once the a priori parameters are well defined. Unlike polynomial fits, it does not remove other residual environmental terms.