Identification of marine ecosystem model parameters using optimization methods

Understanding the quantitative changes in marine ecosystems caused by factors such as climate change, is essential for conserving biodiversity. Marine ecosystem models can help forecast these changes. However, the parameter identification in a highly accurate forecasting model requires collecting oceanographic data and expert tuning. In this study, we applied grid search and Bayesian optimization to hyperparameter optimization to reduce the model building costs and further confirmed that the optimal parameter values can be systematically derived using both methods. The two-step Bayesian optimization, which classifies parameters based on ecosystem characteristics, proved to be the most effective.