General circulation models (GCMs) form the backbone of weather and climate prediction, leveraging numerical solvers for large-scale dynamics and parameterizations for smaller-scale processes such as cloud formation. Despite ongoing improvements, GCMs face significant challenges, including persistent errors, biases, and uncertainties in long-term climate projections and extreme weather events. Recent machine learning (ML) models have had remarkable success in short-term weather forecasts. Still, they lack stability for long-term predictions and do not provide calibrated uncertainty estimates, limiting their utility.
GoogleAI proposes NeuralGCM to address limitations in climate and weather prediction using general circulation models (GCMs). Traditional GCMs, which rely on physics-based simulations, are computationally intensive and struggle to achieve long-term stability and accurate ensemble forecasts. These GCMs combine numerical solvers for large-scale atmospheric dynamics with empirical parameterizations for smaller-scale processes such as cloud formation. Machine learning models, trained on historical data such as ECMWF’s ERA5, have demonstrated impressive short-term weather prediction capabilities at lower computational costs, but fail in long-term prediction and ensemble accuracy.
GoogleAI's NeuralGCM is a hybrid model that combines a differentiable solver for atmospheric dynamics with machine learning components to parameterize physical processes. This model aims to leverage the advantages of traditional GCMs and machine learning approaches, delivering stable and accurate forecasts on multiple time scales with significant computational efficiency.
NeuralGCM integrates a differentiable dynamical core with a learned physics module, which uses a neural network to predict the effects of unresolved atmospheric processes. The end-to-end training approach involves backpropagation through multiple simulation steps, gradually increasing the deployment duration from 6 hours to 5 days. This method ensures that the model accounts for interactions between the learned physics and the large-scale dynamics, improving stability and accuracy.
Experiments were conducted to evaluate the performance of NeuralGCM against best-in-class models such as ECMWF-HRES and ensemble prediction systems, as well as machine learning models such as GraphCast and Pangu. For 1- to 15-day weather forecasts, NeuralGCM achieves comparable accuracy, and the stochastic version shows lower error and better ensemble mean predictions. In climate simulations, NeuralGCM accurately tracks climate metrics over several decades and simulates emergent phenomena such as tropical cyclones, with significant computational savings.
In conclusion, NeuralGCM successfully addresses the limitations of both traditional GCM models and pure machine learning models, and provides a stable and accurate hybrid approach to climate and weather prediction. By combining differentiable solvers with machine learning parameterizations, NeuralGCM improves large-scale physics simulations essential for understanding and predicting the Earth system, while offering significant computational efficiency.
Review the Paper and DetailsAll credit for this research goes to the researchers of this project. Also, don't forget to follow us on twitter.com/Marktechpost”>twitter and join our Telegram Channel and LinkedIn GrAbove!. If you like our work, you will love our Newsletter..
Don't forget to join our Over 47,000 ML subscribers on Reddit
Find upcoming ai webinars here
Pragati Jhunjhunwala is a Consulting Intern at MarktechPost. She is currently pursuing her Bachelors in technology from Indian Institute of technology (IIT) Kharagpur. She is a technology enthusiast and has a keen interest in the field of software applications and data science. She is always reading about the advancements in different fields of ai and ML.
<script async src="//platform.twitter.com/widgets.js” charset=”utf-8″>