Related Experiment Video
Updated: Aug 13, 2026

Measuring the Structure, Composition, and Change of Underwater Environments with Large-area Imaging
Published on: April 18, 2025
LangYa: a large AI model for global ocean forecasting
Nan Yang1, Chong Wang1, Zimeng Zhao1
1Key Laboratory of Ocean Observation and Forecasting, Laboratory of Ocean Circulation and Waves, Institute of Oceanology, Chinese Academy of Sciences, Qingdao 266071, China; Qingdao Key Laboratory of AI Oceanography, Qingdao 266000, China.
LangYa, a novel AI ocean forecasting system, integrates atmospheric forcing and cross-spatiotemporal data for improved accuracy. This system enhances ocean state variable (OSV) predictions with longer lead times and greater stability.
Area of Science:
- Oceanography
- Artificial Intelligence
- Climate Science
Background:
- Accurate ocean forecasting is vital for scientific research and societal applications.
- Existing AI models improve forecasting but struggle with integrating cross-spatiotemporal and atmospheric forcing data.
- Developing a comprehensive AI system for complex ocean dynamics remains a challenge.
Purpose of the Study:
- Introduce LangYa, an AI-driven ocean forecasting system designed to integrate cross-spatiotemporal and atmospheric forcing.
- Enhance the accuracy, stability, and lead-time robustness of ocean state variable (OSV) forecasts.
- Demonstrate LangYa's potential for real-time operational deployment.
Main Methods:
- Utilized a large-language-model-based (LLM-based) time embedding for forecast lead times.
- Implemented an asynchronous cross-iterative random sampling strategy for atmospheric forcing impacts.
- Incorporated an ocean self-attention module for network stability and an adaptive loss function for thermocline dynamics.
Main Results:
- LangYa achieved 7-day Root Mean Square Errors (RMSEs) below 0.0736 m/s for currents, 0.0701 m/s for temperature, and 0.1302 psu for salinity.
- The system demonstrated superior forecast accuracy, lead-time robustness, and stability compared to existing AI and numerical models.
- Successfully produced global OSV forecasts with lead times of 1 to 7 days.
Conclusions:
- LangYa offers significant advantages for global ocean state variable forecasting.
- The system's innovative architecture addresses limitations in integrating diverse oceanographic data.
- LangYa shows strong potential for operational real-time ocean forecasting applications.
Related Concept Videos
Boundary Layer Characteristics
Application of Linearization and Approximation
Types of Global Positioning System Surveys
Buoyancy and Stability for Submerged and Floating Bodies
Eulerian and Lagrangian Flow Descriptions
The Eulerian method focuses on fixed points in space where fluid properties, such as velocity, pressure, and temperature, are observed as the fluid moves between these...
