At locations where ascending and descending tracks cross over oceans, the satellite is able to sample the sea surface on the same location, but at different epochs. These locations are generally known as crossovers. Because measurements are available at the same location, the height of the reference surface can be eliminated by computing the difference between the two fully-corrected sea surface heights. Thus, in contrast to the altimeter height residuals, these crossover height differences are free of errors in the reference surface. In addition, if the timelaps between the respective height measurements is some, correction errors will be highly correlated and partly drop out of the height difference.
The generation of crossover height differences is an iterative process, where first the location of all crossovers and the corresponding epochs are estimated analytically by crossing all ascending and descending tracks during a selected period. Subsequently, crossovers that do not contain sufficient altimeter data around the two epochs (e.g., those over land) are rejected. The positions and epochs of the remaining ones are re-estimated at the crossing of two linear fits through four measurement locations along either track around the estimated position. Cubic spline fits through the relative sea heights along the tracks yield the interpolated relative sea height at either epoch.
The difference between the respective interpolated relative sea heights (referred to as crossover height difference) is only partly due to actual sea level change between the two epochs. The remaining part is caused by errors in the applied altimeter range corrections or the computed orbit altitude, of which the latter is the largest effect for ERS-1. If these errors are different on crossing arcs, they will affect the crossover height differences. Hence, the crossover difference is primarily a measure of the non-geographically correlated radial orbit error.