World
AI ocean forecast model evaluation standards and toolkit released in Xiamen
The National Marine Environmental Forecasting Center released multidimensional evaluation standards and a supporting toolkit for artificial-intelligence ocean forecasting models in Xiamen on November 5. The system balances forecast accuracy, physical consistency and distribution consistency, aiming to unify standards for research and development verification, performance validation and operational admission for intelligent ocean models.
The National Marine Environmental Forecasting Center released multidimensional evaluation standards and a supporting toolkit for artificial-intelligence ocean forecasting models in Xiamen on November 5. The issuing organization said this marks a landmark new achievement in China’s unified evaluation standards for intelligent ocean forecasting models.
The field of intelligent ocean forecasting has long faced problems such as inconsistent evaluation criteria and insufficient testing of physical plausibility. The new standards build a multidimensional evaluation framework that balances traditional forecast accuracy, physical consistency and distribution consistency, addressing past shortcomings of looking only at forecast errors while ignoring ocean dynamic constraints, as well as the difficulty of measuring errors in extreme events and fine-scale features.
According to the introduction, the standards and toolkit establish for the first time a standardized inspection and assessment system for intelligent ocean forecasting models. They can support standardized comparison of intelligent models for ocean forecasts such as temperature, salinity and currents, guiding research and development teams to pay attention to both forecast performance and ocean dynamic mechanisms, and improving the reliability of AI forecasting results.
Relevant experts believe that unified evaluation standards and verification tools can reduce redundant industry assessment costs, promote the intelligent transformation of ocean forecasting, and accelerate the development of trustworthy AI ocean forecasting technology. The National Marine Environmental Forecasting Center said its next step will be to work with relevant institutions to continuously iterate the evaluation system and enrich assessment datasets and extreme-scenario cases.