An Experiment of Ocean State Estimate by Using Argo Data and a 4D-VAR Data Assimilation System

The in-situ Argo data is assimilated into a 4-dimensional variational data assimilation system in order to investigate its effects on the estimation of the ocean state. A time-varying oceanic reanalysis dataset is obtained which is dynamically consistent with both the ocean circulation model and the field observation. The assimilation result exhibits more realistic features of the ocean circulation processes than that obtained only from the ocean circulation model, showing the effectiveness of our assimilation model and the great impact of the Argo data. The 4D-VAR data assimilation system used in this study is constituted on the basis of a strong constraint formalism by using the GFDL Modular Ocean Model (MOM3) and its adjoint, as well as the 4-dimensional variational method. An optimization problem is solved to minimize the cost of the model result and the observational data by controlling the initial condition of the model variables and the air-sea heat, fresh water and momentum fluxes. In the experiment, only the Argo data of temperature and salinity profiles from Jan. 2001 to Jun. 2004 is used. A finer global model is selected in which the horizontal resolution is 1 degree in both longitude and latitude, with 36 vertical levels spaced from 10m near the sea surface to 400m at the bottom. Using climatological monthly forcing, a stable ocean state is firstly calculated. Then, a first guessed field is generated through a 24-year integration with NCEP2's monthly forcing started from 1980. Finally, the Argo data assimilation is carried out. It is revealed that both seasonal and interannual variations in the ocean state are significantly improved through the assimilation, although the coverage of Argo float is still very sparse. Phenomenon such as the El Nino event of 2002 is well reproduced by the assimilation, which agrees with the results derived from the objective analysis with Argo float and TRITON buoy data. To get more realistic estimation of the ocean state, other ocean observational data should be included into the assimilation model.