The overall goal of this work is to investigate the performance of ecosystem models and to relate their results to existing observations in the North Atlantic. Therefore different data assimilation methods are applied. A variational adjoint technique and a micro-generic algorithm (mGA) are utilized to estimate model parameters, such that the misfit between model results and observations is minimised. Data assimilation experiments are performed with nitrogen based ecosystem models, comprising three and four state variables (NPZ- and NPZD models): dissolved inorganic nitrogen (N), phytoplankton (P), herbivorous zooplankton (Z) and detritus (D). The NPZ-model simulates mean concentrations of the different variables within the upper mixed layer, while the NPZD-model has a vertically resolved grid. Physical boundary conditions are obtained from three-dimensional simulations of the ocean's circulation in the North Atlantic, with daily mean atmospheric forcing from ECMWF-reanalysis data.

First, data assimilation experiments are conducted with observations from the Bermuda Atlantic Time-series Study (BATS) in order to optimise the NPZ-model. While applying the adjoint method different optimal parameter sets are obtained when starting from different initial parameter sets. It is shown that for parameter optimisation of an ecosystem model, the application of the mGA is superior to the performance of the adjoint method.

Second, simultaneous assimilation experiment are performed with the NPZD-model using observational data from three locations in the North Atlantic: BATS, the site of the North Atlantic Bloom Experiment (NABE) and the Ocean Weather Ship-India (OWS-INDIA). The simultaneous optimisation yields a best parameter set, which can be utilized for basin wide simulations in coupled physical-biological (general circulation) models of the North Atlantic.

The parameter set retrieved from the simultaneous optimisations produces substantial differences in the biogeochemical fluxes when compared with model results using previously published parameters. In contrast to earlier models the rapid cycling of organic matter for sustaining primary production is emphasized. Furthermore, systematic discrepancies between 14C-fixation rate and modelled primary production are identified. It is suggested that carbon based primary productivity may not be adequately represented by ecosystem models when a constant nitrogen to carbon conversion factor is assumed.

Information: http://www.awi-bremerhaven.de/People/show?mschartau