Abstract: A global ocean data assimilation system named ZFL_GODAS is developed in this study. The system is a sub-system of an operational system for short-range climate forecasting, and provides the global ocean initial state field for coupled ocean-atmosphere models. It can assimilate observations such as satellite altimetry, sea surface temperature (SST), in situ temperature, and salinity from Argo, XBT, TAO, and other sources. ZFL_GODAS uses LICOM1.0 (a global OGCM developed by LASG/IAP) as its ocean model. It is an ensemble optimal interpolation system that uses an ensemble of a series of model states from a free LICOM running to estimate the background error covariances (BECs). The ensemble-based BECs are multivariate and inhomogeneous and they can reflect the length scales, anisotropy, and covariability of oceanic physical processes. To enhance the efficiency, a set of distinctive, efficient, and load-balanced parallelized Ensemble Optimal Interpolation (EnOI) programs have been developed.
The performance of ZFL_GODAS is evaluated by comparing its results with a range of satellite-derived and in situ observations. We compare the interannual variability of SST and sea surface height, the evolution of the SST anomaly at the equator, and the root-mean-square error of model results of SST, sea level anomaly, and sub-surface temperature and salinity. We also show the five-year-mean profiles for temperature bias, salinity, and zonal velocity. We find that the ocean data assimilation shows a very positive impact on the modeled fields. We can preliminarily conclude that ZFL_GODAS performs well, so it can provide a desirable global ocean initial state for the ocean model component of the climate forecasting system, and provide effective reanalysis data for improving our understanding of the oceans.