32 Pierre-Yves Le Traon
The tidal signal is the most important variable signal in altimetric data. This signal is only partially corrected using global ocean tide models. The residual errors
are then aliased at certain periods depending on the repeat-period of the satellite.
The satellite repeat-period should be chosen to avoid aliasing near dominant oceanic periods (e.g. annual or semi-annual periods). The M2 tide is thus aliased near
60 days for TIP. This is much less of a problem than the GEOSAT 317 -day aliasing
period which can cause major problems for the interpretation of altimetric signals.
More accurate global tide models based on hydrodynamic models and/or TIP data
are now available and should be used. StiU, aliasing may occur and this has to be
taken into account when interpreting the data.
Last, but not least, comes the orbit error. This error is caused by imperfect
knowledge ofthe spacecraft position in the radial direction. It is actually the largest
error on altimetric measurements of sea surf ace topography. It depends on the quality ofthe satellite tracking system. For TIP, precise orbit determination is achieved
via three distinct tracking systems (DORIS, Laser, GPS) providing almost global
coverage of satellite orbits. The radial orbit error obtained is thus accurate to about
2 cm, compared to 10 cm accuracy for the most recent GEOSAT and ERS-1I2
orbits available. Orbit errors are long-wavelength errors (about 40000 lan) that can
be reduced by analyzing the altimeter data. This can be done via global or regional
minimization of crossover or repeat-track differences (relative to a mean or to a
given cycle). These differences do not contain any geoid signal and are dominated
by the orbit error. Crossover or repeat-track differences also contain the oceanic
signal. The main problem is thus to separate the orbit error from this oceanic signaI. We musttirst assume an a priori analytical form or an a priori spectrum ofthe
orbit error. A more empirical approach, commonly used, is to approximate the orbit
error by fitting a tirst or second degree polynomial over a given arc length. However, this method also removes the large-scale oceanic signal and is only suitable
for mesoscale studies. To minimize ocean signal removal, more complex methods
using cross-track information are required. Global crossover minimization can thus
be used to estimate the orbit error without removing too much of the large-scale
oceanic signal. More generally using inverse techniques, the orbit error signal
could be obtained through a global adjustment taking into account not only the spatial but also the temporal characteristics ofthe orbit error and oceanic signal.
Mapping
Most applications need maps (and associated formal error) of the altimetric signal (SLA) on regular space/time grids. This can be done using optimal interpolation methods which use a priori knowledge of the space and time scales of the
ocean signal. It is preferable, particularly in low eddy energy regions and when
several altimeter data sets are merged, to take into account an along-track long
wavelength error (correlated noi se, e.g. due to orbit, tidal or inverse barometer
residual errors) in the method (Le Traon et al., 1998).
The tidal signal is the most important variable signal in altimetric data. This signal is only partially corrected using global ocean tide models. The residual errors
are then aliased at certain periods depending on the repeat-period of the satellite.
The satellite repeat-period should be chosen to avoid aliasing near dominant oceanic periods (e.g. annual or semi-annual periods). The M2 tide is thus aliased near
60 days for TIP. This is much less of a problem than the GEOSAT 317 -day aliasing
period which can cause major problems for the interpretation of altimetric signals.
More accurate global tide models based on hydrodynamic models and/or TIP data
are now available and should be used. StiU, aliasing may occur and this has to be
taken into account when interpreting the data.
Last, but not least, comes the orbit error. This error is caused by imperfect
knowledge ofthe spacecraft position in the radial direction. It is actually the largest
error on altimetric measurements of sea surf ace topography. It depends on the quality ofthe satellite tracking system. For TIP, precise orbit determination is achieved
via three distinct tracking systems (DORIS, Laser, GPS) providing almost global
coverage of satellite orbits. The radial orbit error obtained is thus accurate to about
2 cm, compared to 10 cm accuracy for the most recent GEOSAT and ERS-1I2
orbits available. Orbit errors are long-wavelength errors (about 40000 lan) that can
be reduced by analyzing the altimeter data. This can be done via global or regional
minimization of crossover or repeat-track differences (relative to a mean or to a
given cycle). These differences do not contain any geoid signal and are dominated
by the orbit error. Crossover or repeat-track differences also contain the oceanic
signal. The main problem is thus to separate the orbit error from this oceanic signaI. We musttirst assume an a priori analytical form or an a priori spectrum ofthe
orbit error. A more empirical approach, commonly used, is to approximate the orbit
error by fitting a tirst or second degree polynomial over a given arc length. However, this method also removes the large-scale oceanic signal and is only suitable
for mesoscale studies. To minimize ocean signal removal, more complex methods
using cross-track information are required. Global crossover minimization can thus
be used to estimate the orbit error without removing too much of the large-scale
oceanic signal. More generally using inverse techniques, the orbit error signal
could be obtained through a global adjustment taking into account not only the spatial but also the temporal characteristics ofthe orbit error and oceanic signal.
Mapping
Most applications need maps (and associated formal error) of the altimetric signal (SLA) on regular space/time grids. This can be done using optimal interpolation methods which use a priori knowledge of the space and time scales of the
ocean signal. It is preferable, particularly in low eddy energy regions and when
several altimeter data sets are merged, to take into account an along-track long
wavelength error (correlated noi se, e.g. due to orbit, tidal or inverse barometer
residual errors) in the method (Le Traon et al., 1998).