Digital Event Horizon
HiRO-ACEv2, a new climate modeling system, offers a flexible and inexpensive way to generate high-resolution rain, wind, and pressure data for extreme events. By reducing costs and expanding capabilities, this breakthrough has the potential to significantly impact the field of climate modeling and simulation.
HiRO-ACEv2 is a new update to the High Resolution Output Ai2 Climate Emulator (HiRO-ACE) system, allowing for the simulation of extreme weather events at a fraction of the cost. The system consists of two main parts: ACE (coarse 100 km resolution) and HiRO (fine 3 km detail), with a new downscaling model (HiROv2) that predicts multiple high-resolution fields simultaneously. HiROv2 has improved skill in predicting extreme events, matching X-SHiELD's high-resolution statistics within 3% for rare events. The system accurately simulates tropical cyclones, hurricanes, and strong winds, with realistic eyes and rain bands. HiRO-ACEv2 generates global 3 km data, with statistics closely matching X-SHiELD's, at a cost of 300 times less than X-SHiELD. Future work includes training on new X-SHiELD datasets and developing a more efficient version of HiROv2, which could make the system more than ten times faster.
HiRO-ACEv2, the latest update to the High Resolution Output Ai2 Climate Emulator (HiRO-ACE) system, has been introduced to simulate extreme weather events at a fraction of the cost. This innovation, developed by Hugging Face, offers researchers a flexible and inexpensive way to generate high-resolution rain, wind, and pressure data for extreme events over extended periods.
The HiRO-ACEv2 system consists of two main parts: ACE, an AI model that operates at a coarser 100 km resolution, and HiRO, a model that sharpens those coarse outputs into fine 3 km detail. The biggest change in HiRO-ACEv2 is the new downscaling model, HiROv2, which can now predict several high-resolution fields at once, capturing a much wider range of extreme weather events. HiROv2 has improved skill in predicting extreme events, matching the target simulation's high-resolution statistics to within 3% for events that occur only once in a million data points.
One of the primary goals of HiRO-ACEv2 was to simulate extreme events well. Changes to the training process of HiROv2 improved its ability to reproduce the most intense rain, wind, and pressure values. The system accurately estimates the total energy of tropical cyclones, including storms it never saw during training, and generates hurricanes with realistic eyes and rain bands. Although winds and pressure around the eye remain somewhat more lopsided than in X-SHiELD, the system realistically generates strong winds flowing down mountain slopes.
The updated version of ACE has also been trained for this system, which shows that its coarse simulations are a solid foundation for producing long-term high-resolution climate data. When put together, HiROv2 and ACE generate a full year of global 3 km data, with statistics closely matching X-SHiELD's, even though HiRO was working from a simulation ACE ran on its own rather than from real X-SHiELD data.
Producing global high-resolution data with HiRO-ACEv2 costs roughly 300 times less than running X-SHiELD, a high-resolution global atmospheric model that explicitly simulates storms and other small-scale weather. If only one region is needed, the savings can be more than a hundred times greater, depending on the size of the area. A streamlined version of HiROv2 is also being worked on, which could make it more than ten times faster still.
Despite this significant breakthrough, there are still future work ahead. HiRO-ACEv2 only works with the version of ACE tuned to match X-ShiELD, meaning it cannot yet produce high-resolution data for other climates, such as a warmer future or the historical record in datasets like ERA5. Plans are underway to train on new X-SHiELD datasets, including one simulating a world 4°C warmer, so the system can handle different climates. Additionally, outputs like near-surface temperature and snow and ice are being planned, which would allow for even more uses, such as snowpack prediction.
To facilitate the exploration of HiRO outputs, a web demo has been created that shows pre-generated precipitation, wind, and sea-level pressure output from HiROv1 and HiROv2 side by side with X-SHiELD. This demo complements the quantitative comparisons in the tech report, offering a quick visual check of how the outputs look and allowing users to zoom in on their own region of interest to judge performance there.
The models mentioned in this article are available to download, and tutorial notebooks are available to walk through how to run them. This innovation has the potential to significantly impact the field of climate modeling and simulation, and further research and development will be crucial to unlocking its full potential.
Related Information:
https://www.digitaleventhorizon.com/articles/Revolutionizing-Climate-Modeling-with-HiRO-ACEv2-A-Breakthrough-in-Simulating-Extreme-Weather-Events-deh.shtml
https://huggingface.co/blog/allenai/hiro-acev2
https://allenai.org/blog/hiro-acev2
Published: Wed Oct 14 12:35:11 2026 by llama3.2 3B Q4_K_M
