A comprehensive examination of the current AI landscape in ocean sciences: applications, challenges, and emerging trends

Abstract

Artificial intelligence (AI) and machine learning (ML) are increasingly used in ocean sciences to analyze large, complex, and unevenly sampled datasets from satellites, autonomous platforms, sensors, vessels, and genomic tools. This review examines how AI and ML are being applied across major areas of ocean research, including physical oceanography, coastal hazards, marine biogeochemistry, ecology, remote sensing, acoustics, marine geology, and applied ocean engineering. Using a literature-synthesis workflow supported by Schlitsummarizer and full-text extraction methods, this paper identifies key applications, emerging trends, research gaps, and limitations in ocean-based ML. The review shows that ML is particularly useful for predicting surface currents, reconstructing subsurface ocean conditions, forecasting waves, estimating biogeochemical variables, classifying marine organisms and habitats, improving remote-sensing products, and supporting autonomous observing systems. Emerging trends include the growth of deep learning, physics-informed machine learning, explainable AI, data integration across observing platforms, and management-focused applications related to climate adaptation, coastal planning, fisheries, and environmental monitoring. However, major challenges remain, including sparse and biased ocean observations, limited data from deep-ocean and polar regions, validation bias, poor model transferability, class imbalance in ecological datasets, black-box model behavior, and uncertainty under changing climate conditions. Overall, this review argues that AI and ML are valuable tools for expanding ocean observation and prediction, but they must be used alongside oceanographic knowledge, careful validation, and transparent interpretation to ensure reliable scientific and management outcomes.