Global couplled ocean-atmosphere reanalysis and forecast with pelagic ecosystem

The system used to produce the data set is the 4-dimensional variational (4D-VAR) coupled data assimilation system based on a coupled ocean-atmosphere global model, which was developed by Sugiura et al. (2008) and the JAMSTEC-K7 consortium. It also incorporates a biogeochemical model (NPZDC model), which is the pelagic ecosystem model based on nitrogen cycle (Masuda et al. 2013). (See the references for details of the system and models.) In the 4D-VAR approach, optimized 4-dimensional analysis fields are sought by minimizing a cost function on the basis of adjoint method for physical parameters and Green's function approach (Menemenlis et al. 2005) for biogeochemical ones. The assimilated observations for the atmospheric component are the PREPBUFR data set (air temperature, specific humidity, and wind vectors) and 10 m scalar wind stress from SSMIS satellite observations. The assimilated observations for the oceanic physical component are OISST, and temperature and salinity profile data from the Argo and GTSPP. The assimilated data for the biogeochemical component are climatological mean monthly nitrate from WOA05, chlorophyll-a from SeaWiFS, and chlorophyll-a from WOA98 which is converted to detritus. The procedure of the production of the data set is as follows: using the assimilation window of 3 months, the assimilation experiment was conducted for every 3-months period from January-March 2010. Using the optimized control variables and parameters within the assimilation windows, the forecast experiments were performed for 3 years from the beginning of the assimilation windows (thus, the first 3 months correspond to the state estimation and the remaining 2 years and 9 months correspond to the pure forecast). In the forecast experiments, 10 ensemble forecast runs were also conducted using the atmospheric initial conditions shifted to 1-5 days before/after.