Improving the validity and robustness of a Harmful Algal Bloom model through genetic algorithm-based optimization
To study the complex patterns governing Harmful Algal Blooms (HAB), an individual- or agent-based modeling approach was employed. The constructed model was able to mechanistically represent both the biological and the physico-chemical factors involved during events of bloom and decline of the toxic dinoflagellate, Pyrodinium bahamense var. compressum (Pbc), within Sorsogon Bay. An inherent problem with this approach, however, is that as the model complexity increases, it becomes harder to fit the model results to the in situ data during model calibration. The traditional approach to deal with this is to manually tweak the values of model parameters until the model output stabilizes or attains good agreement with field observations. However, this method is time-consuming and does not guarantee accurate results. Hence, automated calibration of the Sorsogon Bay HAB model was designed, which makes use of an optimizer in the form of a genetic algorithm (GA). In this paper, we investigate the effect of variations of the initial vertical distribution of Pbc (distribution of cells among different depths) on the model outcomes of cell increase and transport. The GA approach was used to determine the fit of the model output to field observations expressed as an optimization problem. The genetic algorithm provides values for model parameters during each simulation run and takes the error between the model output and the expected result as the objective for optimization. The GA-provided values for model parameters will then be incorporated into the model resulting to a more robust and valid representation. The calibrated model provides insights into the role of advection and spatial distribution on Pbc dynamics, and can then be more confidently used to explore various scenarios that could provide clues regarding the patterns of HAB and decline within Sorsogon Bay.
Key words: harmful algal blooms, model, Pyrodinium, Sorsogon Bay, genetic algorithm
If you wish to have the full texts of the articles of or submit articles to the Journal of Environmental Science and Management, please contact the Managing Editor at this email address: email@example.com.