AI for Seamless Numerical Weather-Climate Prediction
Artificial Intelligence (Al) offers an innovative approach to numerical weather climate prediction (NWCP) by training models on historical data, as opposed to solving physical equations, resulting in rapid computations compared to traditional numerical models. Advancing Al NWCP tools may present challenges and opportunities for National Meteorological and Hydrological Services (NMHSs), including impacts on the workforce and influencing future investments. However, training the Al models requires significant data and computation, frequent updates to account for climate variability and improved physical understanding to design and produce the new training datasets. It is still unknown how effectively Al NWCP models forecast extreme events and questions remain on how to best integrate Al NWCP and traditional physics based NWCP models into operational production systems that harness the best of both methods. Nevertheless, Al NWCP could revolutionize operational systems resulting in democratized access to forecast information and insights, benefitting NMHSs that previously lacked resources for in-house NWCP models. Additionally, this may result in new players entering the market, notably the private sector, including big-tech and impact-based weather startups, which may impact the role of NMHSs and the quality and reliability of weather and climate services in an expanding market. Other Al opportunities across the value cycle, such as natural language generation, computer vision or machine learning, may enhance the transformation of NWCP outputs into useful and actionable information for various users and sectors.
Text from Annex to Decision 10 (EC-78)