Coastal modelling incorporating artificial neural networks for improved velocity prediction

ABSTRACT

Owing to the complexity of flow patterns, modelling of flow in coastal areas remains a challenge, even with the numerical techniques of today. In this paper, we have tried to address the problem of modelling the currents off the coast of Bay of Bengal. It is found that ocean modelling using coastal community models like Delft3D and MIKE21 could produce inherent errors if used to study depths of continental shelf greater than the 50 m isobaths (called Far-Shelf). This error can be partly overcome by adding a data processing tool like Artificial Neural Network (ANN) to improve velocity predictions. Predictions from such a proposed ANN model are seen to improve the correlation from as low as 0.28 for stand-alone model to more than 0.9 for ANN augmented results for Far-Shelf. This is because the ocean parameters in the Far-shelf are dependent not only on coastal processes but also on the interactions with the deep ocean. The advantage of better numerical accuracy strongly offsets the disadvantages of using such data-driven methods and will aid in getting a more accurate flow field.

KEYWORDS-:

• Artificial Neural Network
• Bay of Bengal
• correlation coefficient
• Delft3d
• ocean modelling

Disclosure statement

No potential conflict of interest was reported by the authors.