Sequential assimilation of a year-long time-series of SeaWiFS chlorophyll data into a 3D biogeochemical model on the French Mediterranean coast

The objective of this paper was to explore the potentialities of sequential statistical estimation methods to assimilate ocean color observations in a primary production model coupled to a 3D hydrodynamic model. The study site was the gulf of Fos-Rhone delta region on the French Mediterranean coast. The high rate of primary production generally observed in this area is mainly due to strong nutrient inputs of the Rhone River. The assimilation method is derived from the singular evolutive extended Kalman filter (SEEK), which uses an error subspace represented by multivariate empirical orthogonal functions (EOF). SeaWiFS chlorophyll data were assimilated by the ecosystem model during a simulation performed under realistic meteorological conditions for the year 2001. An 'adaptive' computing method of the EOF was applied in order to lower the instabilities of the filter. Data assimilation system permitted to reduce the mean absolute error between model and data from 1.51 to 0.77 mg m−3 thanks to the SEEK filter, showing a substantial 49% gain. Efficiency of the SEEK filter was then investigated considering several areas of interest inside the modelled domain. Finally, impact of the assimilation scheme on non-observed variables was illustrated and discussed. Throughout this experimentation the data assimilation system showed its potential regarding operational systems.