On November 6, the National Satellite Ocean Application Center officially launched "Kanhai," the world's first end-to-end artificial intelligence ocean model independently developed in China. This model achieves a full-loop closure from satellite remote sensing observations to three-dimensional ocean environmental predictions for the first time. It not only can reconstruct current sea conditions in real time but also accurately simulate changes in sea surface temperature, salinity, density, and flow field up to 600 meters in depth for the next 10 days, providing an unprecedented "digital ocean mirror" for marine research, disaster prevention, and ecological governance.
The core breakthrough of "Kanhai" lies in its end-to-end intelligent inference architecture. Traditional ocean prediction relies on physical models and human experience, which is time-consuming and limited in accuracy; however, "Kanhai" directly uses massive satellite remote sensing data as input, automatically learning the spatiotemporal evolution rules of the ocean system through deep learning, skipping the complicated modeling steps, and achieving "observation equals prediction." This means that when an ocean satellite flies over the South China Sea, AI can output a three-dimensional dynamic change map of the area for the next 10 days within a few hours - with accuracy far exceeding existing methods and response speed increased several times.
Image source note: The image is generated by AI, and the image licensing service provider is Midjourney.
This achievement is the result of China's long-term accumulation in the ocean observation system. Led by Academician Jiang Xingwei of the Chinese Academy of Engineering, the project was launched in 2022, integrating China's self-developed ocean satellites (such as the Haiyang No. 1 and Haiyang No. 2 series) and international satellite data, building a global, high-frequency, multi-dimensional training dataset. After more than two decades of development, China has established a world-leading ocean satellite observation network, with remote sensing data accounting for more than 70% of the global ocean observation volume, providing a solid data foundation for "Kanhai."
"Kanhai" is not only a scientific breakthrough but also a key infrastructure for addressing climate change and marine disasters. In the future, it can provide real-time warnings for risks such as red tides, abnormal sea temperatures, and strong current changes, helping fisheries avoid risks, maintain offshore wind power, plan long-distance navigation, and even provide high-precision boundary conditions for global carbon cycle research.
AIbase believes that the release of "Kanhai" marks the deep penetration of AI from consumer and manufacturing fields into Earth system science. When large models can not only understand human language but also read the "pulse" of the ocean, our understanding of this blue planet will truly enter the era of intelligent prediction. And this time, China is leading the world.
