Merging recent Mean Sea Surface into a 2023 Hybrid model (from Scripps,
DTU, CLS and CNES)
Abstract
In this paper, we compute a new hybrid mean sea surface (MSS) model by
merging three recent models, CNES_CLS22, SCRIPPS_CLS22 and DTU21, and
taking advantage of their respective features. The errors associated
with these models were assessed using sea level anomalies for
wavelengths ranging from 15 to 100km from Sentinel-3A (S3A), SWOT KaRIn
during its calibration phase and ICESat-2 in the Arctic ice-covered
regions. The absolute error associated with this new Hybrid23 MSS is
estimated at 0.15 ± 0.04 cm² with S3A. The greatest improvements
observed on S3A sea level anomalies are mainly located in coastal
regions and along geodetic structures: on average, the error is reduced
by 23% within 200km along the coast and by 35% in the Indonesian
region compared with SCRIPPS_CLS22. Despite these improvements, the MSS
error still impacts significantly sea level anomalies computed from
altimetry: it explains 15% and 18% of the S3A and SWOT KaRIn
respective global variance. It becomes predominant (> 30%)
if we consider the shorter wavelengths ([15, 30km]). CNES_CLS15,
older, explains up to 88% of the variance of SWOT KaRIn at these
wavelengths. MSS errors have become a major limiting factor to the
accuracy of sea level anomalies, and hybridization even adds
sub-mesoscale errors. SCRIPPS_CLS22 and DTU21 also remain better in
certain regions of the North Atlantic above 60°N and in Arctic coastal
areas. Finally, many efforts are still required to develop the MSS to a
new level of precision, which we could soon achieve with SWOT KaRIn
during the scientific phase.