crossover catalogue is generated in two steps. First, approximated crossovers are derived
by crossing master arcs for ascending and descending subsatellite tracks. Second, the
precise location is determined by least squares solving simultaneously for latitude,
longitude and precise crossing times. The significant wave height and wind speed
crossover differences are estimated as well.
The crossover catalogue is then used to estimate the radial orbit error that is modeled by
piecewise cubic polynomials. These polynomials are set up for every ascending and
descending track with constraints ensuring continuity up to the first derivative at track
transitions. The sequence of cubic polynomials is analyzed globally. The polynomial
coefficients are estimated by a rigorous least squares adjustment with an inversely
proportional weighting to the crossover time differences.
The radial orbit error estimation was done separately for ERS-1.ALT.ORP2 and ERSl.ALT.QLOPR. For ERS-l the analysis was performed for batches of 35 days. For
TOPEX data no orbit error adjustment was applied due to the high orbit quality.
Merging of altimeter data for different satellites causes systematic errors that must be
removed before the gridding procedure. There are several reasons for the inconsistencies
between different satellite sources:
a global absolute vertical datum is not available, which means that global biases
exist between different satellite altimeter data
the time periods for ERS-1.ALT.OPR2, ERS-1.ALT.QLOPR and TOPEX GDR
are not the same, resulting in different atmospheric conditions and thus various sea
states and ocean circulation
the satellite heights above a reference ellipsoid are based on different gravity
models and station coordinates causing long wavelength systematic errors
the external range corrections, for example ocean tides (see table 1), differ for
each data type, resulting in short and long wavelength systematic errors
In order to remove most of the systematics between different satellite data a polynomial
approach was chosen which removes trackwise long wavelength differences to a reference
sea surface. For the reference sea surface, an ERS-l two year mean sea surface
(SSH_L94101) was taken (due to the above mentioned systematic errors when merging
different satellite sources a single mission sea surface was chosen). This model, which is
an ESA standard product, is based on ERS-1.ALT.OPR2 from the multidisciplinary phase
(35 day repeat) and the second ice phase (3 day repeat) and has a spatial resolution of 6' .
Each half revolution of the altimeter data is compared to this reference sea surface. The
differences are represented by third degree polynomials. The polynomials are adjusted to
the altimeter data, which results in a removal of long wavelength systematic errors. Short
wavelength differences (up to few hundred km) remain in the data and must be removed
by the gridding procedure.
A sea surface height model is obtained by interpolation of along-track sea surface heights
corrected for the orbit error and the systematic differences to the reference sea surface to
an equiangular earth [lXed grid. For each grid node a local plane is estimated by a least
211
by crossing master arcs for ascending and descending subsatellite tracks. Second, the
precise location is determined by least squares solving simultaneously for latitude,
longitude and precise crossing times. The significant wave height and wind speed
crossover differences are estimated as well.
The crossover catalogue is then used to estimate the radial orbit error that is modeled by
piecewise cubic polynomials. These polynomials are set up for every ascending and
descending track with constraints ensuring continuity up to the first derivative at track
transitions. The sequence of cubic polynomials is analyzed globally. The polynomial
coefficients are estimated by a rigorous least squares adjustment with an inversely
proportional weighting to the crossover time differences.
The radial orbit error estimation was done separately for ERS-1.ALT.ORP2 and ERSl.ALT.QLOPR. For ERS-l the analysis was performed for batches of 35 days. For
TOPEX data no orbit error adjustment was applied due to the high orbit quality.
Merging of altimeter data for different satellites causes systematic errors that must be
removed before the gridding procedure. There are several reasons for the inconsistencies
between different satellite sources:
a global absolute vertical datum is not available, which means that global biases
exist between different satellite altimeter data
the time periods for ERS-1.ALT.OPR2, ERS-1.ALT.QLOPR and TOPEX GDR
are not the same, resulting in different atmospheric conditions and thus various sea
states and ocean circulation
the satellite heights above a reference ellipsoid are based on different gravity
models and station coordinates causing long wavelength systematic errors
the external range corrections, for example ocean tides (see table 1), differ for
each data type, resulting in short and long wavelength systematic errors
In order to remove most of the systematics between different satellite data a polynomial
approach was chosen which removes trackwise long wavelength differences to a reference
sea surface. For the reference sea surface, an ERS-l two year mean sea surface
(SSH_L94101) was taken (due to the above mentioned systematic errors when merging
different satellite sources a single mission sea surface was chosen). This model, which is
an ESA standard product, is based on ERS-1.ALT.OPR2 from the multidisciplinary phase
(35 day repeat) and the second ice phase (3 day repeat) and has a spatial resolution of 6' .
Each half revolution of the altimeter data is compared to this reference sea surface. The
differences are represented by third degree polynomials. The polynomials are adjusted to
the altimeter data, which results in a removal of long wavelength systematic errors. Short
wavelength differences (up to few hundred km) remain in the data and must be removed
by the gridding procedure.
A sea surface height model is obtained by interpolation of along-track sea surface heights
corrected for the orbit error and the systematic differences to the reference sea surface to
an equiangular earth [lXed grid. For each grid node a local plane is estimated by a least
211