Project Grant 2616259

The National Science Foundation Office of Integrative Activities awarded Louisiana State University $249,862 on October 1, 2026, under the Geosciences program (CFDA 47.050) to develop physics-informed deep learning models for understanding and predicting coastal hypoxia formation mechanisms on the Louisiana-Texas shelf. The project runs through September 30, 2029, with place of performance in Baton Rouge, Louisiana.

The work applies machine learning to coastal oxygen-loss prediction by integrating three components: U-Net and DeepLabV3+ semantic-segmentation networks to map spatial hypoxia patterns from environmental data; dynamic graph neural networks to represent oxygen-transport pathways derived from high-resolution circulation models and neural operator-based ocean digital twins; and physics-informed constraints to ensure mechanistic understanding. The project incorporates novel seafloor oxygen-flux data into the modeling framework to improve predictions of dead-zone spread and persistence—critical for forecasting the Gulf's largest dead zone, which spans an area larger than Connecticut and disrupts marine food webs.

The award funds graduate and undergraduate student training and development of educational materials for K–12 and graduate students delivered through an artificial intelligence-powered learning platform. All data, software, and trained models will be released as free, open resources to support reproducibility and enable resource managers to make faster, more reliable forecasts for fisheries management and nutrient-reduction strategy development.