A simultaneous optimization of sea ice model parameters by genetic algorithm

• Main Authors: Sumata, Hiroshi, Kauker, Frank, Gerdes, Rüdiger, Karcher, Michael, Köberle, Cornelia
• Format: Conference Object
• Language: unknown
• Published: 2016
• Subjects: Sea ice
• Online Access: https://epic.awi.de/id/eprint/42416/; https://hdl.handle.net/10013/epic.49098

Improvement/optimization of a sea ice model is of great significance for understanding the sea ice physics and for understanding the Arctic climate system and its linkage to the global climate. For better representation of modeled sea ice properties, we develop a parameter optimization system for a couped ocean-sea ice model. Since the sensitivities of dynamic and thermodynamic parameters of sea ice models are interrelated, the system handles both sets of parameters simultaneously. The system also handles a long assimilation window of 33 years. Such a long time window has never been tested by other algorithms (e.g., adjoint method, EnKF). Since the cost function defined by the model - data misfit may have an ill-shaped structure (multiple local minima), we apply an algorithm, which can find the global minimum of an ill-shaped function. A micro-Genetic Algorithm is one of the possible solutions to optimize sea ice model parameters.