China Launches AI Ocean Forecasting Model Evaluation Standards and Toolkit

China News Service reported that the National Marine Environmental Forecasting Center released multidimensional evaluation standards and a supporting toolkit for artificial intelligence ocean forecasting models in Xiamen. The release marks a milestone in establishing unified testing and evaluation norms for the country's intelligent ocean forecasting models. The newly launched standard framework establishes a multidimensional evaluation structure balancing traditional forecast precision, physical consistency, and distribution consistency. It resolves past evaluation weaknesses that focused solely on forecast errors while ignoring ocean dynamic constraints, extreme events, and fine-scale features, offering an authoritative metric for development, verification, and operational access. According to the report, the standards and accompanying toolkit establish a standardized inspection and evaluation system for intelligent ocean forecasts. The framework supports standardized comparisons of intelligent temperature, salinity, and current ocean forecast models, guiding development teams to balance forecasting efficacy with ocean dynamic mechanisms. The forecasting center stated that its next step involves collaborating with relevant organizations to iteratively improve the intelligent forecast evaluation system. Plans include enriching testing datasets and extreme scenario cases to continually enhance independent innovation capacity and operational service levels in ocean forecasting.

Experts indicated that introducing unified standards and a toolkit provides industry developers with consistent verification methods. This initiative reduces repetitive testing costs, advances intelligent ocean forecasting transitions, improves the technology innovation ecosystem, and accelerates the cultivation of trustworthy artificial intelligence forecasting.