Adjoint Parameter Identification of Ecomodels

The parameter identification of ecosystem is required both in phenomenological data analysis and in modeling the spatial resolution and the temporal variability of the biotic and abiotic state variables. Moreover the structural model is generally not known because of the non-linearity of the system. An approach based only on the knowledge of the dynamical evolution of the state variables and of the associated adjoint ones is considered and a conjugate gradient procedure taking account of the data errors is proposed. Applications are shown on pelagic and neritic data. Extension to evaluation of stochastic equation errors, required by ensemble averaged conceptual models, is also discussed.