NeuralGCM revolutionises climate modeling with machine learning

NeuralGCM revolutionises climate modeling with machine learning is profiled by BTW Media because published evidence links it to internet infrastructure, governance, operational dependencies, or market visibility.

• A team of researchers at Google and the European Centre for Medium-Range Weather Forecasts (ECMWF) have developed a new climate model called NeuralGCM that uses AI to improve the speed and accuracy of weather forecasts.
• By using neural networks to augment traditional physical simulations, NeuralGCM has made significant advances in modelling and predicting climate processes.

OUR TAKE
Researchers from Google and ECMWF introduce an innovative climate model called NeuralGCM. NeuralGCM uses neural networks to augment traditional High Performance Computing (HPC) focusing on small-scale climate processes such as clouds and precision variations that are difficult for traditional models to accurately simulate. By using historical weather data collected by ECMWF for training, NeuralGCM has achieved significant results in improving forecast speed and accuracy, especially in high-resolution simulations, outperforming existing climate models and bringing new hope and possibilities to the field of climate science and weather forecasting.
-Rae Li, BTW reporter

What happened

NeuralGCM focuses on small-scale climate processes that are difficult for traditional models to capture accurately, such as clouds and accuracy changes, and demonstrates the potential to improve forecast speed and accuracy by being trained using historical weather data collected by the ECMWF. It was developed using Google’s JAX machine learning framework, which enables the model to run natively on accelerators such as TPUs or GPUs, resulting in significant improvements in speed and efficiency.

Google claims that the 1.4-degree resolution version of the NeuralGCM model is more than 3,500 times faster than the X-SHiELD model. The source code and weights for the NeuralGCM model have been released under a non-commercial licence on GitHub for public use. The researchers hope to eventually incorporate other aspects of the Earth’s climate system, such as the oceans and the carbon cycle, into the model, allowing NeuralGCM to make predictions on longer time scales, going beyond weather forecasting to the level of climate prediction.

Also read: What does DataRobot do? Automating AI and machine learning

Also read: What is DataRobot: Revolutionising machine learning and AI

Why it’s important

The NeuralGCM model can be indicated the great potential of AI in improving the predictive capabilities of climate models. NeuralGCM models are able to simulate and predict climate phenomena more accurately, especially those small-scale processes that are difficult to capture by traditional models. They can help to improve the accuracy and speed of weather forecasting and have important implications for understanding the complex climate system, responding to climate change and formulating related policies.

In addition, the development and open-source release of the NeuralGCM model has provided the climate science community with a new tool that facilitates interdisciplinary collaboration and knowledge sharing. By using high-performance computing resources, such as TPUs or GPUs, NeuralGCM models are able to run at faster speeds, which helps researchers to perform climate simulations and analyses more efficiently.