Search the RAMMB website
Citation: Ebert-Uphoff, I., and K. Hilburn, 2023: The outlook for AI weather prediction. Nature, News and Views, July 5, 2023. https://doi.org/10.1038/d41586-023-02084-9. (Full article access: https://rdcu.be/df8LB)
Summary: This Nature News & Views article discusses two new Artificial Intelligence (AI) models for global weather forecasting published in this issue of Nature that seek to replace traditional physics-based numerical weather prediction models.
Detailed summary: This article places into context two new AI models for global weather forecasting. These new models demonstrate the enormous potential that AI holds for weather prediction, speeding up the process of forecasting by up to a factor of 10,000. Such increases in computational speed could yield immense benefits, such as increasing spatial resolution of models, generating large ensembles, and the potential to integrate additional physical processes (such as fire spread). However, our article also spells out several potential risks of AI-based models, including potential under-prediction of extreme events, not representing meteorological features with as much clarity as their numerical counterparts, inconsistencies between different fields, and the risk of unpredictable behavior when encountering conditions the model was not trained on. Given both the potential benefits and risks associated with AI models for weather prediction, this article issues a call to action for weather forecasters to get involved in evaluating and interpreting such systems to ensure that they meet the needs of forecasters and thus contribute to – and not endanger – public safety. (POC: Imme Ebert-Uphoff and Kyle Hilburn, CIRA, iebert@colostate.edu, kyle.hilburn@colostate.edu, Funding: NSF and CIRA ML strategic funds)