Parameter Optimization and Validation of a Marine Biogeochemical Model using a Hybrid Algorithm

Sensitivity computations, parameter identification and optimization for an 1-D marine biogeochemical model of $NPZD$ type are presented. For the optimization a hybrid algorithm combining quantum-evolutionary and local gradient-based search methods is used. It turns out to be an efficient and flexible tool for optimization and can be easily adopted for other simulation models. For the model under investigation attainable data could be exactly identified. For realistic measure ment data we argue that a certain parameter set leading to a non-optimal fit cannot be improved. Moreover we show that data uncertainty leads to a significant parameter spread. Thus we conclude that the $NPZD$ model needs to be modified or extended, maybe including a modification of external forcings and/or initial conditions