What is Langya
Langya is version 1.0 of a new-generation ocean artificial intelligence large model independently developed by the Institute of Oceanology, Chinese Academy of Sciences, focusing on ocean state variable forecasting. The model combines advanced artificial intelligence algorithms with specialized ocean science knowledge, enabling medium- and short-term high-precision forecasting of global ocean state variables. Version 1.0 of "Langya" can forecast key indicators such as temperature, salinity, and ocean currents for the next 1 to 7 days in a single run, with a spatial resolution of 1/12° and a temporal resolution of 24 hours, significantly improving the accuracy and reliability of global ocean forecasting.
Main Functions of Langya
- Medium- and short-term high-precision forecasting: The model can achieve medium- and short-term high-precision forecasting of global ocean state variables, including key indicators such as temperature, salinity, and ocean currents.
- Global coverage and high resolution: The model has global coverage capability, with a spatial resolution of 1/12° and a temporal resolution of 24 hours, improving the accuracy and reliability of global ocean forecasting.
- Multi-day continuous forecasting: It can forecast ocean state variables for the next 1 to 7 days in a single run, providing continuous data support for ocean scientific research and practical applications.
- Forecasting of complex ocean phenomena: The research and development team plans to introduce forecasting of ocean phenomena such as typhoons, precipitation, ocean waves, and sea ice in the future 2.0 version, further enhancing the ability to forecast complex ocean phenomena such as marine disasters.
Technical Principles of Langya
- Integration of artificial intelligence and ocean science: The "Langya" model deeply integrates specialized knowledge of ocean science with cutting-edge artificial intelligence algorithms. This enables the model to achieve precise forecasting of ocean state variables based on advanced artificial intelligence technology.
- High-precision forecasting capability: Version 1.0 of "Langya" has achieved medium- and short-term high-precision forecasting of global ocean state variables, and can predict key indicators such as temperature, salinity, and ocean currents for the next 1 to 7 days in a single run.
- High resolution: The model's spatial resolution reaches 1/12°, and its temporal resolution is 24 hours. This is a major technological breakthrough in the field of global ocean forecasting, significantly improving the accuracy and reliability of forecasting.
- Data-driven model architecture: The model is based on the Ocean-Specific Transformer model architecture, designs an ocean-land mask mechanism to exclude interference from land areas, and designs an ocean-specific block to capture the inherent characteristics of the ocean environment.
- Multi-source data fusion technology: The large model is trained using GLORYS12 reanalysis data, ERA5 reanalysis data, and SST satellite observation data, tested on buoy observation data and satellite observation data, and comprehensively evaluated in aspects such as ocean currents, temperature, salinity, and temperature/salinity profiles.
Application Scenarios of Langya
- Marine environmental safety assurance: By accurately forecasting ocean state variables such as temperature, salinity, and ocean currents, it provides important data support for maritime navigation safety and reduces the risk of maritime accidents.
- Addressing global climate change: The model can provide continuous data on ocean state variables, helping scientists better understand and predict global climate change trends.
- Development of marine resources: Through in-depth understanding and forecasting of the marine environment, marine resources such as fishery resources and mineral resources can be developed and utilized more effectively.
- Marine disaster prevention and mitigation: The model can predict ocean phenomena such as typhoons, precipitation, ocean waves, and sea ice, providing early warning information for coastal areas to respond to extreme weather events and reducing losses caused by natural disasters.
- Scientific Research: Provide more refined data support for marine scientific research and advance research progress in the field of marine science.