However, gravity field model errors induce regional vertical offsets of the orbit that are constant from repeat cycle to repeat cycle, and may be common to both the ascending and descending pass. Thus, a considerable part of the gravity-induced radial orbit error is purely geographically correlated and can not be observed from crossover height differences. But also other statistical analyses (orbital differences, altimeter height residuals, collinear track analyses) do not provide any information on this part of the radial orbit error. This phenomenon is known as the non-observability of the geographically-correlated radial orbit error. Estimates for this contribution thus comes from statistical analyses of the estimated gravity model error (See Section ).

Figure 6 displays the global distribution of the 50,143 crossovers that could be generated from the edited OPR02 data sets for Cycle 9. This Figure also presents the RMS crossover height difference per 20 x 20 cell, based on the operational DUT orbits. In order to eliminate spurious data, editing at a level of 3.5 times the global crossover RMS of 20.2 cm was applied.

Some higher crossover differences are influenced by sea level variability caused by strong meandering and variable currents, like the Circumpolar Current, Kuroshio, and Gulf Stream. Generally, at higher latitudes, the crossover RMS is higher than over equatorial regions (See also Figure 7).

**Figure 7:**
Crossover height difference RMS as a function of their geodetic latitude.
The crossovers are based on the D-PAF orbit (dotted line),
DUT JGM-2 operational orbit (continuous line), or the JGM-2 orbit including
SLR and ERS-1/TOPEX crossovers as tracking data (dashed line).
Dots indicate the number of crossovers per bin.

This might be due to a remaining local degradation of the radial orbit error, but meandering of currents, and the resulting sea level variability, should not be excluded as a cause. In coastal regions, imperfection in the modelling of tides contribute to a regional increase of the crossover height differences. At very high latitudes (above 60 North and below 60 South), errors in the atmospheric corrections and interference by ice sheets will strongly affect the statistics. Low values like in the Eastern Pacific Ocean, where no sizable currents are present, are most indicative of the real radial orbit errors.

**Figure 8:**
Crossover height difference RMS
as a function of the time interval
between the ascending and descending pass forming each crossover.
The crossovers are based on the D-PAF orbit (dotted line),
DUT JGM-2 operational orbit (continuous line), or the JGM-2 orbit including
SLR and ERS-1/TOPEX crossovers as tracking data (dashed line).
Dots indicate the number of crossovers per bin.

Figure 8 portrays the distribution of the crossover RMS as a function of the time laps between the ascending and descending passes forming the crossover.

The crossover RMS , however, shows a significant increase during the first ten days, and then stabilises. The RMS values start out with 17-18 cm for a time interval less than 2 days, and increase to an average of 20-23 cm for intervals longer than ten days. The discrepancy of 12-14 cm (in RSS sense) is likely to be caused by ocean variability, propagation correction errors, and tide model errors. All these have decorrelation periods of some 2 to 10 days, and thus contribute to the crossover height differences only at longer time intervals. The magnitude of the discrepancy is compatible with generally accepted values of 6, 4, and 4 cm for the ocean variability, tide model errors, and propagation correction errors, respectively.

Thu 16 Feb 1995