EPSCoR Research Fellows: NSF: Advancing Adaptive Ocean Modeling Using Machine Learning Methods
Ente: EPSCoR RII: EPSCoR Research Fe
Scadenza: 2028-04-30
Importo max: 295.993 EUR
Paese: US
Descrizione
This Research Infrastructure Improvement (RII) EPSCoR Research Fellows project provides a fellowship to an Associate Professor, Dr. Michal Kopera, and training for a graduate student at Boise State University. This work is conducted in collaboration with Dr. Pierre Lermusiaux at Massachusetts Institute of Technology. Through the fellowship, the PI will develop new artificial intelligence tools that help ocean models automatically decide where higher resolution is needed, improving both accuracy and efficiency. The project will address how to allocate limited computing power in large-scale ocean simulations so that key features, such as currents, are captured without excessive cost. The project combines mathematics, ocean science, and machine learning to replace traditional resolution adjustment methods with an adaptive, machine-learning-based approach. More efficient ocean models can improve weather forecasting, coastal resilience planning, and natural disaster preparedness. The fellowship will also strengthen research capacity in Idaho by training students in scientific machine learning and strengthening collaboration between Boise State and Massachusetts Institute of Technology.
This project will integrate machine learning with adaptive-grid ocean modeling to improve the efficiency and predictive capability of large-scale simulations. It will develop a deep reinforcement learning framework that replaces heuristic mesh refinement criteria for systems of partial differential equations governing ocean circulation. The intellectual contribution lies in formulating a refinement agent that dynamically adjusts the mesh based on learned policies rather than user-defined thresholds, while operating within prescribed computational resource constraints. The methodology will include designing reward functions for nonlinear shallow water systems discretized using high-order discontinuous Galerkin methods, training and validating refinement agents on benchmark ocean test cases, and deploying this approach within the multi-layer h-NUMO model, a hydrostatic version of the Non-hydrostatic Unified Model of the Ocean. The fellowship will advance research infrastructure by expanding faculty expertise in scientific machine learning, developing collaboration with a national research institution, and providing interdisciplinary training for a doctoral student. Project activities will integrate research, curriculum development, and workforce preparation in computational mathematics and Earth system modeling. This project is supported by the EPSCoR Research Infrastructure Improvement Program: EPSCoR Research Fellows, which supports early- and mid-career investigators in eligible jurisdictions to develop collaborations at the nation’s private, government or academic research institutions.
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
Istituzione: Boise State University
Sede: BOISE, ID
PI: Michal Kopera
Settori: EPSCoR RII: EPSCoR Research Fe
Vai al bando originale
Registrati gratis su Bandolo per trovare bandi compatibili con la tua azienda.