How essential are Argo observations to constrain a global ocean data assimilation system?

Type Article
Date 2016
Language English
Author(s) Turpin Vincent1, 3, Remy E.1, Le Traon Pierre-YvesORCID1, 2
Affiliation(s) 1 : Mercator Ocean, Parc Technol Canal,8-10 Rue Hermes, F-31520 Ramonville St Agne, France.
2 : IFREMER, ZI Pointe Diable, Technopole Brest Iroise, F-29280 Plouzane, France.
3 : Inst Pierre Simon Laplace, LOCEAN, 4 Pl Jussieu, F-75252 Paris, France.
Source Ocean Science (1812-0784) (Copernicus Gesellschaft Mbh), 2016 , Vol. 12 , N. 1 , P. 257-274
DOI 10.5194/os-12-257-2016
WOS© Times Cited 14
Abstract Observing system experiments (OSEs) are carried out over a 1-year period to quantify the impact of Argo observations on the Mercator Ocean 0.25° global ocean analysis and forecasting system. The reference simulation assimilates sea surface temperature (SST), SSALTO/DUACS (Segment Sol multi-missions dALTimetrie, d'orbitographie et de localisation précise/Data unification and Altimeter combination system) altimeter data and Argo and other in situ observations from the Coriolis data center. Two other simulations are carried out where all Argo and half of the Argo data are withheld. Assimilating Argo observations has a significant impact on analyzed and forecast temperature and salinity fields at different depths. Without Argo data assimilation, large errors occur in analyzed fields as estimated from the differences when compared with in situ observations. For example, in the 0–300 m layer RMS (root mean square) differences between analyzed fields and observations reach 0.25 psu and 1.25 °C in the western boundary currents and 0.1 psu and 0.75 °C in the open ocean. The impact of the Argo data in reducing observation–model forecast differences is also significant from the surface down to a depth of 2000 m. Differences between in situ observations and forecast fields are thus reduced by 20 % in the upper layers and by up to 40 % at a depth of 2000 m when Argo data are assimilated. At depth, the most impacted regions in the global ocean are the Mediterranean outflow, the Gulf Stream region and the Labrador Sea. A significant degradation can be observed when only half of the data are assimilated. Therefore, Argo observations matter to constrain the model solution, even for an eddy-permitting model configuration. The impact of the Argo floats' data assimilation on other model variables is briefly assessed: the improvement of the fit to Argo profiles do not lead globally to unphysical corrections on the sea surface temperature and sea surface height. The main conclusion is that the performance of the Mercator Ocean 0.25° global data assimilation system is heavily dependent on the availability of Argo data.
Full Text
File Pages Size Access
Publisher's official version 18 5 MB Open access
Top of the page