OceanAI leverages open-source large language models (LLMs) to process vast oceanographic reports and data, enabling efficient analysis, summarization, and insight extraction. By automating report processing and enhancing data interpretation, OceanAI helps researchers uncover patterns, accelerate discoveries, and support informed decision-making in marine science.
OceanNet is a neural operator-based digital twin for regional sea surface height (SSH) emulation. Using a Fourier neural operator and a predictor-evaluate-corrector integration scheme, it enhances forecast stability and mitigates autoregressive error growth. Trained on historical SSH data, OceanNet provides seasonal predictions for Loop Current eddies and Gulf Stream meanders, achieving comparable accuracy to state-of-the-art models while reducing computational costs by 500,000 times.
This ensemble learning model predicts significant wave height and average wave period along the U.S. Atlantic coast. Using a stacking method, it integrates LASSO regression, support vector machine, and multilayer perceptron for improved accuracy. Trained on 20 years of NOAA buoy data, it provides forecasts at 1-, 3-, 6-, and 12-hour intervals. The inclusion of swell waves enhances long-term predictions, making it a robust alternative to traditional coastal models.
MATLAB software to identify silver perch calls in passive acoustic data using deep learning. The system employs a detection stage using signal kurtosis and signal-to-noise ratio, a feature extraction stage utilizing ResNet-50 on wavelet scalograms, and a one-vs-all SVM classifier. Trained on 6,000 perch calls and 6,000 other signals from eight sites in Pamlico Sound, NC, the classifier achieves 98.9% accuracy.
MATLAB software for identifying oyster toadfish (Opsanus tau) calls in passive acoustic data using deep learning. The classifier, trained on over 20,000 labeled signals from western Pamlico Sound, NC, automates the detection of distinctive boatwhistle calls using a combination of template matching, ResNet-50 feature extraction, and a one-vs-all SVM classifier. This tool facilitates the monitoring of toadfish vocalizations, aiding in ecological research on their distribution and behavior. Instructions and a tutorial are available in the ToadFishFinder Wiki.
Goodnight Innovation Distinguished Professor
Director, Ocean Observing and Modeling Group
Research Focus: Coastal ocean circulation dynamics; Numerical modeling and data assimilation
Director of International Affairs
Department of Marine, Earth and Atmospheric Sciences
Research Focus: AI and LLM Applications in the Geoscience;Large dataset processing and modeling
Director, Generative Intelligent Computing Lab
Department of Computer Science
Research Focus: Large Language Models, Foundation Models, Diffusion Models, AI Agents
Full Professor
Department of Marine, Earth and Atmospheric Sciences
Research Focus: Dynamic Spatio-Temporal Modeling Geovisualization Remote Sensing & Sensor Networks
Ph.D. Student
Research focuses on LLMs and diffusion models, with applications in scientific computing, multi-modal AI, and environmental modeling.
Postdoctoral Research Scholar
Observing and predicting sub-mesoscale eddies in the Gulf of Mexico using advanced machine learning techniques and remote sensing data analysis
Postdoctoral Research Scholar
Studying multi-scale ocean dynamics and applying AI/ML approaches for physical oceanography of the Gulf of Mexico, with focus on numerical modeling, data assimilation, and air-sea-ice interactions
Ph.D. Student
Leveraging Reliable Foundation Models for Scientific Discovery
Ph.D. Student
Utilizing AI and Machine Learning Models for Geoscience Applications
Developing specialized Large Language Models for marine science applications, including automated research analysis, marine species identification, and ocean parameter interpretation. Our LLMs are trained on vast marine databases to provide expert-level insights for oceanographic research.
Utilizing advanced AI models to predict marine climate patterns, ocean temperature changes, and their impact on marine ecosystems. Our research combines satellite data, ocean sensors, and machine learning to forecast climate-related ocean phenomena with unprecedented accuracy.
Applying sophisticated data mining techniques to extract meaningful patterns from vast ocean datasets. Our research focuses on discovering hidden correlations in oceanographic data, analyzing marine biodiversity patterns, and understanding complex ocean-atmosphere interactions.
Tang, S., Wang, Y., Ding, C., Liang, Y., Li, Y., & Xu, D.
European Conference on Computer Vision : 73-90 (2025)
Liu, X., Lei, B., Zhang, R., & Xu, D.
AAAI Conference on Artificial Intelligence (2025)
Chattopadhyay, A., M. Gray, T. Wu, A. B. Lowe, R. He
Scientific Reports, 14 : 21181 (2024)
DOI: 10.1038/s41598-024-72145-0
Chaichitehrani, N., R. He, and M. N. Allahdadi
Artificial Intelligence for the Earth Systems (2024)
DOI: 10.1175/AIES-D-23-0061.1
Instructor: Dr. Paul Liu - MEAS
Instructor: Dr. Del Bohnenstiehl - MEAS
Instructor: Dr. DK - CS
Instructor: Dr. DK - CS
