Aggressive space mapping for optimisation of a marine ecosystem model (Q427166)

Summary: We apply the aggressive space mapping (ASM) algorithm to parameter optimisation of a one-dimensional marine ecosystem model. Such models are important to calculate the global carbon cycle and specifically the oceanic \(CO_{2}\) uptake. The aim of the optimisation is to find model parameters that minimise the misfit between model output and observational data. ASM first solves for an optimum of a computationally cheaper low-fidelity model that in our case is based on a coarser time discretisation while still guaranteeing numerical stability. Secondly, an approximate optimum of the high-fidelity model is iteratively found by a parameter mapping applying a quasi-Newton method and \textit{C.G. Broyden} update [Math. Comput. 19, 577--593 (1965; Zbl 0131.13905)]. The applicability of the ASM technique to the problem is verified by using synthetic target data. Results are compared to those of the direct high-fidelity model optimisation. We show that a very reasonable solution can be obtained while yielding a significant reduction in the total optimisation cost.