An Update on Coastal Artificial Intelligence and the AI2ES NSF AI Institute

The use of artificial intelligence (AI) to model the coastal environment started as early as the nineties. The ability of AI models to describe nonlinear relationships between variables makes the approach highly valuable including to describe the sometimes complex dynamical processes of the coastal environment at the intersection of land, ocean, and atmosphere. To further harness the power of AI ethically and equitably the National Science Foundation (NSF) has been creating NSF AI institutes. One of the inaugural NSF institutes, the AI institute for Research on Trustworthy AI in Weather, Climate, and Coastal Oceanography (AI2ES) is in part focusing on coastal processes and their forcings. Past and present applications of AI to predict coastal processes will be shared including new deep learning (DL) models designed to better capture air-sea and air-land interactions to predict more accurately coastal fog, inundation, and other coastal processes. The DL models ingest several types of data, numerical weather predictions, coastal measurements, and satellite imagery. EXplainable AI, or XAI, methods are being developed along with the new DL architectures with the aim to estimate the relative importance of the models’ respective input features, to guide further model development, and, potentially, gain scientific insights into the processes. The talk will end by sharing a couple of successful implementations of coastal AI models.