Schartau, M. M.. Institut fuer Meereskunde Kiel, firstname.lastname@example.org
Willebrand, J. M.. Institut fuer Meereskunde Kiel, email@example.com
PARAMETER OPTIMIZATION AT BATS USING THE ADJOINT METHOD
Assimilation experiments with BATS-measurements (1989-1996) were performed to find the optimal parameter values of a zero-dimensional NPZ model. An optimization scheme based on discrete adjoint equations has been formulated which can be applied with low computational cost. Experiments with different parameterizations of the entrained nutrients and changes in mixed layer depth variations were performed to reveal linear dependencies of certain parameters as well as to give information about model sensitivities.
Therefore a Monte Carlo method with hundreds of individual optimizations at a time have been applied. The optimized parameter values that resulted from the random a priori guesses were used to determine a posteriori error-matrix
which is compared with the discrete Hessean matrix close to the best parameter solutions.
Identical twin experiments showed that different ecosystem model interpretations are obtained depending on the physical boundary conditions in the absence of zooplankton information and primary production data. A monthly mean mixed layer (ML)
depth variability may result in model parameters corresponding to a bottom up control of the ecosystem, whereas the inclusion of daily ML depth variability leads to top down control.
Additional data of primary production constrains the dynamical control of the ecosystem but does not give a definite solution of the grazing parameters.
Day: Monday, Feb. 1
Time: 04:45 - 05:00pm
Location: Eldorado Hotel