The new model, dubbed 'NeuralGCM,' has proven to be more accurate than other purely machine learning-based models for one to 10-day weather forecasts, outperforming top extended-range models. It also excels in projecting climate conditions over decades.
NeuralGCM, developed by Google Research, Google DeepMind, MIT, Harvard University, and the European Center for Medium-Range Weather Forecasts, uses machine learning and a neural network to train on past weather data and incorporate physics equations.
AI models like NeuralGCM offer computational power and timeliness advantages over traditional models, running relatively quickly on a laptop compared to traditional models that require powerful supercomputers and hours to process thousands of lines of code.
The integration of large-scale physics with AI components in the NeuralGCM model is highlighted as a significant novelty. Unlike other AI forecasting models, NeuralGCM combines a global circulation model with AI-driven tasks, providing more accurate forecasts.
[
Collection
]
[
|
...
]