Ai2 and Google back open-source climate model for long-term projections

From weather forecasts to climate projections

NeuralGCM+ will build on Google Research’s NeuralGCM, a hybrid system that combines physical equations for large scale dynamics with artificial intelligence to represent smaller-scale processes such as cloud formation and rainfall.

While NeuralGCM has been developed for forecasts over weeks to months, Ai2 will focus on climate trends over years to decades. Planned applications include studying how rainfall and water availability could change through mid-century in regions such as the southwestern United States.

The expanded model is intended to simulate the atmosphere and ocean together, including how oceans absorb and redistribute heat. Ai2 says this is a major influence on long-term climate.

The case for faster climate simulations

Conventional climate models have helped scientists understand global warming, but their computational demands limit the number of experiments researchers can run. Ai2 says faster and less expensive simulations could allow scientists to explore more possible futures and investigate persistent uncertainties in regional temperature and rainfall trends.

Google Research’s existing NeuralGCM demonstrates the potential for speed. In a public benchmark, an atmosphere-only version simulated a year of atmospheric conditions for the entire globe in eight minutes on a single TPU.

Ai2 also points to its Climate Emulator, or ACE, family of models. In one study, ACE2 completed experiments on how ocean-temperature patterns affect climate roughly 100 times faster than the conventional atmospheric model it was trained to emulate.

Three research priorities for NeuralGCM+

Ai2 plans to develop NeuralGCM+ across three connected areas. First, it will link NeuralGCM with an ocean and sea-ice model to produce stable simulations spanning decades. A key goal is to reduce regional temperature and rainfall errors when simulating the past 50 years, compared with conventional climate models.

Second, the team will add more established physics and test whether the model can reproduce temperature and rainfall changes in warmer climates that were not represented in its training data. Conventional climate-model simulations will serve as a reference.

Third, Ai2 plans to make the tools easier to use and extend. Building on NeuralGCM’s open-source code and available model weights, it expects to release models and modeling frameworks with documented workflows for outside researchers to train and adapt models and contribute code. The team also plans to improve performance on widely used GPUs.

An open platform for climate research

Ai2 says its long-term ambition is a shared research platform that climate modeling laboratories can use and improve together. The stated goal is to put these tools within reach of more scientists and help communities make better-informed decisions about water planning and climate adaptation.

“Much of the progress in Earth system AI modeling has focused on weather forecasting, in the range of days to months,” Oliver Watt-Meyer, Ai2’s Climate Modeling Lead, said in the announcement. “But communities also need to understand what could happen over decades as they plan for a changing climate.”

The partnership matters because it combines Google Research’s existing hybrid modeling approach with Ai2’s focus on longer-term projections and open research. However, Ai2 says demonstrating reliable future projections remains a central research task, so NeuralGCM+ is a development effort rather than a finished forecasting system.