Read: 1549
Researchers at Google have developed NeuralGCM, an innovative weather prediction system that merges traditional physics-basedwith techniques. This new model combine the speed and computational efficiency of systems with the accuracy and physical grounding of conventional general circulationGCMs. It represents a departure from the ongoing debate among meteorologists about whether or more traditional methods are superior for forecasting.
NeuralGCM integrates classic GCM calculations that determine large-scale atmospheric changes essential to making predictions. Then, it employsalgorithms that excel in correcting errors on smaller scalestypically those related to cloud formations and regional microclimates like San Francisco's fog. The researchers clm that this selective integration ofenhances the model’s performance across various scales.
The outcome is a system capable of generating accurate forecasts for weather conditions spanning from one day up to two weeks ahead, matching or even surpassing the accuracy of real-time predictions by the European Centre for Medium-Range Weather Forecasts ECMW. This achievement potentially revolutionizes how we predict complex climate events and long-term atmospheric changes that are currently too computationally demanding using traditional methods.
The significance of NeuralGCM lies beyond enhancing local weather forecasting. It offers a powerful tool to address costly computational bottlenecks in large-scale climate research, such as predicting tropical cyclones more efficiently or modeling the impacts of climate change years ahead. Asbasedconsume fewer resources and can run with less than 5,500 lines of code compared to conventional systems requiring nearly 377,000 code lines for initialization alone, the potential applications ext far beyond academic research.
skepticism in weather forecasting has diminished, as advancements have demonstrated their utility. However, many experts awt a clearer direction on howcan reshape forecasting methodologies and enhance our understanding of climate dynamics further.
Acknowledgments:
was updated to clarify that Stephan Hoyer is affiliated with Google Research rather than Google DeepMind.
Citation:
O'Donnell, James 2024. Google's NeuralGCM: A Fusion of Physics-Basedandfor Enhanced Weather Prediction. MIT Technology Review. https:www.mitreview.comneuralgcm-goggle--prediction
:
MIT Technology Review reserves the right to modify articles as necessary to mntn factual accuracy, while preserving the original meaning and intent.
Illustration by Rose Wong
Founded at MIT in 1899, MIT Technology Review is an indepent media company that covers emerging technologies and their impact on society. We publish insights, analysis, interviews, reviews, live events, and more for a global audience of technology and business leaders.
Explore our history
MIT Alumni News
Advertise with us
International Editions
Custom content solutions
Privacy policy
Cookie Policy
This article is reproduced from: https://www.technologyreview.com/2024/07/22/1095184/a-new-weather-prediction-model-from-google-combines-ai-with-traditional-physics/
Please indicate when reprinting from: https://www.58es.com/Weather_forecast/NeuralGCM_Weather_Prediction_Google.html
NeuralGCM: Fusion of AI and Physics for Weather Prediction Enhanced Accuracy with Googles NeuralGCM Model Revolutionizing Climate Research with NeuralGCM NeuralGCM: Reducing Computational Demands in Forecasting AI vs Traditional Methods: A New Era in Weather Prediction NeuralGCM: Streamlining Complex Atmospheric Changes Modeling