In order to validate and improve the MSS calculation, several tests were performed:

• Look at the differencies between the T/P, ERS-1, and Geosat mean profiles heights and corresponding interpolated values of the MSS.
• Compare the MSS gradients with the along track mean profiles to verify the restitution of the short vawelengths of the geoid.
• Compare the CLS_SHOM v. 98.2 MSS to MSS calculated with other dataset (e.g. GEOSAT geodetic dataset), in particular the OSU95 MSS (Yi, 1995) and the GRGS MSS (Cazenave et al., Mazzega et al.,1996).
• Analyse the impact of using the CLS_SHOM v. 98.2 MSS to reference satellite data (instead of using the classical method of referencing SLA to a mean profile).
• Analyse the impact of using the CLS_SHOM v. 98.2 MSS to calculate the geoid cross-track errors, instead of the OSU95 MSS.

FIRST CONCLUSIONS

• The validation tests performed are not independant:
  

  1) we have compared the MSS value to the mean profiles used to estimate the surface, it is an internal quality test ;
  

  2) we have looked at the differences with other MSS, it is an external relative test (we do not know the absolute precision of the other MSS).

• Because the T/P and ERS-1 mean profiles are accurate, and because the differences of these profiles with the CLS_SHOM v. 98.2 MSS are low (~2 cm rms), we can conclude that the accuracy of the MSS under the satellite ground tracks is good, and the short wavelength well represented (slope differences below 2mm/km). Note that in area where few T/P or ERS-1 mean profile data can constrain the MSS estimation, we assume that the accuracy treshold should be of the order of the geodetic data precision, i.e., 6.5 cm rms.

• MSS differences indicate that the OSU MSS is less accurate than the CLS_SHOM v. 98.2 MSS. Visual comparisons clearly show the trackiness contaminating the OSU MSS. While comparisons between the GRGS and CLS_SHOM v. 98.2 MSS reveal the so called "orange skin effect". The CLS_SHOM v. 98.2 MSS seems to well reproduce the geoid short wavelength along the satellite ground tracks, but might be smoother than the GRGS MSS in between (note that the CLS_SHOM v. 98.2 MSS does not contain GEOSAT geodetic data which provide higher spatial resolution).

• The CLS_SHOM v. 98.2 MSS have a really homogeneous oceanic signal content. This was one of the main objective of the work: by referencing to the T/P mean profile, and by reducing the oceanic variability of the ERS-1 data, the oceanic content of the CLS_SHOM v. 98.2 MSS is the mean height of the ocean during 1993-94-1995.

• The impact of the use of the MSS to reference SLA has been inferred. It shows that using the MSS does not degrade the SLA: 2 cm rms differences are observable in some areas with the SLA classically referenced to the mean profiles. These differences are either caused by MSS biases (like geographical correlated errors) or mean profile accuracy (near the cost and the island the mean profiles are less precise).

• The impact of the use of the CLS_SHOM v. 98.2 MSS instead of the OSU95 MSS for the geoid cross track error has been evaluated. RMS differences at crossover points of the T/P mean profile corrected using in the correction the CLS_SHOM v. 98.2 or OSU95 MSS are 1.5 and 1.8 cm rms respectively. CLS_SHOM v. 98.2 is thus more accurate at the track vicinity.