CAIG: Understanding complex marine ecosystems through artificial intelligence-enabled modeling of predator-prey interactions

This project will develop new artificial intelligence (AI) methods to understand marine food webs. Specifically, the project will study the North Atlantic right whale and the zooplankton they prey on. Ocean conditions like currents, weather, and nutrients impact their distribution. The AI approach will unite oceanographic and species data from different sources that are challenging to combine without deep learning. The framework will apply to other predator-prey interactions. The United States depends on healthy oceans for food production, energy, transportation, and commerce. The project will lay the foundation for AI-enabled approaches to ocean management that balance conservation with economic development. The research team will collaborate with federal partners to improve decision-support tools for marine ecosystem management. The research team will partner with K-12 educators for data science workshops with pre-college students. Research opportunities for undergraduate, graduate, and postdoctoral researchers will train the future STEM workforce. Marine ecosystems are shaped by complex spatiotemporal interactions among ocean dynamics, prey distributions, and predator behavior, making them difficult to predict. Meanwhile, modern ocean observing systems are generating rapidly growing volumes of multi-modal, multi-source, and high-dimensional data. Existing ecological models are not well suited to learning from these rich data streams because they typically rely on low-dimensional inputs, aggregate or low-information representations, static or weakly dynamic covariates, and restrictive parametric assumptions. As such, the overarching goal of this project is to develop a unified AI-enabled framework that learns coupled ocean-prey-predator interactions from heterogeneous ocean observing data. The project will advance interpretable, scientifically grounded AI methods that explicitly account for the physical and ecological processes underlying marine ecosystems. The resulting methods are expected to provide substantially more accurate predictions and richer scientific insights than traditional modeling frameworks, whose rigid parametric assumptions limit their ability to capture the complexity and dynamics of marine ecosystems. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.