OBPS Workshop VIII, 14-18 Oct 2024
AI Methods for Ocean related Tasks
AI methods have not yet found their way into the field of oceans to the extent that one would expect. There may be several reasons for this, which could be related to the scarcity and sometimes noisy or unknown quality of observational data. However, the situation is different for image analysis in ocean applications, as data is sufficiently available and useful algorithms have been developed in other areas. The use of AI methods, especially machine learning, leaves room for discussion about their verification and validation. This gap can only be closed if sufficient confidence has been gained from the user side in ensuring a consistent, structured approach. In that context, considering uncertainties in the training and test data are fundamental to make AI methods robust and trustworthy, with implications for the description of EOVs.
This session aims to further deepen the discussion on AI methods, which is a very much necessary step to achieve the anticipated benefit. Embedding this discussion in the Ocean Best Practice System will lead to a wide dissemination of the outcomes of the session.