Holistic genetic optimization of a Generalized Multiple Discrete Interaction Approximation for wind waves

A key element of wind wave models is the parameterization of the resonant nonlinear interactions between spectral wave components. In a companion paper a new Generalized Multiple Discrete Interaction Approximation (GMD) has been developed. The present paper addresses the optimization of the free parameters of the GMD. A holistic optimization approach is used where full model integration results are optimized. Fifteen objective metrics are used, defined to measure the accuracy of a model using the GMD relative to a model using the full (exact) interactions. Due to the large number of free parameters to be optimized, and due to the existence of many local error minima in parameter space, traditional error mapping or steepest descent search algorithms are not suitable to optimize the GMD. The focus of the present study is on establishing genetic optimization techniques as a feasible and economical way to optimize the free parameters in the GMD. The behavior of the GMD with optimized parameters is outside the scope of this study, and is discussed in detail in the companion paper.

► Holistic full-model optimization is used for interactions in wind wave models.
► Genetic optimization allows for optimization of more than 20 free parameters.
► Genetic optimization is used together with mapping and steepest descent algorithms.
► This is the first time genetic optimization is used in wind wave models.