China's new marine AI models make headway

JINAN — Ten meters below the ocean surface off Wuzhizhou Island in South China's Hainan province, a small robotic inspector glides silently among branching corals. Through its cameras, hawksbill turtles and parrotfish appear against a reef that, a decade ago, was nearly lifeless.

The island's marine ranch once faced coral collapse. Since 2010, artificial reefs have helped restore the habitat, but monitoring remained difficult.

"We used to rely on manual diving. It's inefficient and risky," said Wang Aimin, a professor at Hainan University and chief scientist at the Hainan Provincial Engineering Research Center for Modern Marine Ranching.

The shift came through collaboration with Robotfish, a marine technology firm based in Qingdao in Shandong province. Its underwater robot can operate for hours with minimal disturbance, capturing close-up, multi-angle images of coral health. Today, seven information monitoring stations form a round-the-clock network tracking water temperature, salinity and reef activity.

"If something goes wrong, we can act immediately," Wang said.

China's homegrown marine "black tech" — from intelligent robots to ocean AI models — is increasingly being deployed in ecological protection, disaster forecasting and shipping services, as the country pushes digital and intelligent transformation in its marine sector. Monitoring coral reefs is relatively benign compared with harsher ocean conditions. GPS signals do not penetrate seawater. Pressure increases by one atmosphere every 10 meters. Biofouling, corrosion and strong currents can degrade equipment within weeks.

"One cannot simply adapt land-based tech, but must rebuild it from first principles," said Wang Fan, director of the Chinese Academy of Sciences' Institute of Oceanology.

He said the ocean remains one of the least instrumented places on Earth, with marine data still sparse and costly to collect.

This challenge is driving the development of marine AI models in hubs such as Qingdao, known as China's "capital of marine science".

At a recent Digital Earth conference, the Institute of Oceanology of the Chinese Academy of Sciences unveiled LangYa 2.0, an ocean AI model that forecasts typhoons, precipitation, sea ice, storm surges, internal solitary waves and mesoscale eddies.

The model's six vertical systems translate forecasts into formats usable by policymakers and coastal communities. Wang Fan said the system correctly predicted several sudden-turning typhoons last year, improving 24-hour forecast accuracy by more than 10 percent.

"With lightweight versions of large models, we can deploy them locally," he said, adding that this could offer lower-cost forecasting tools for developing regions.

In the Strait of Malacca, one of the world's busiest shipping lanes, fourth-generation underwater cleaning robots from ZhiZhen Marine Science and Technology (Weihai) Co scrape barnacles and algae from vessel hulls.

"Biofouling increases drag and fuel consumption, and can shorten a ship's lifespan. Traditional manual cleaning is inefficient, risky and no longer fits modern shipping," said Qin Mingda, head of the company's cleaning robot project.

The robots operate at depths of up to 150 meters and can clean 2,000 square meters per hour, work that would require several skilled divers. In 2025, they cleaned over 1,000 ships in Malacca waters, according to the company.

Officials have highlighted marine technology as a strategic priority in the 15th Five-Year Plan (2026-30), with emphasis on innovation and digital transformation.