It is argued that, to remove the satellite orbital errors, the crossover adjustment method that represents orbital errors in terms of Fourier series is preferable to the conventional bias-and-tilt crossover adjustment method. However, no matter what technique is used in the crossover adjustment, orbital errors will introduce a geographically dependent error (mostly a large-scale slope which, in terms of spherical harmonics, is represented by the degree 1 terms) into the derived sea surface topography. This is because the responsible orbital errors have left no traces in crossover differences and therefore can never be removed. However, it is possible to reduce this fundamental uncertainty substantially if the record length is long enough. Its implications on observing the basin-scale ocean circulation by satellite altimetry are discussed. We also propose new methods for observing the large-scale oceanic variability (the conventional approach limited us to observing the mesoscale variability).