Allen Institute for AI introduced HiRO-ACEv2. It simulates extreme rain, wind and pressure events at 3 km resolution. The system costs about 1/300th of X-SHiELD. HiROv2 now predicts multiple high-resolution fields at once. It improved accuracy for rare extreme events. A full year of global 3 km data was generated. Hurricane energy and structure match X-SHiELD closely. Regional runs can cut costs over a hundredfold. A faster HiROv2 version is in development. Future work includes support for warmer climates and new outputs like temperature. Models and notebooks are available for download. A web demo compares HiROv2 outputs with X-SHiELD.
What changed
HiRO-ACE used a downscaling model that only predicted rainfall at 3 km resolution.
What it unlocks
Generating global 3 km climate data for extreme weather events at far lower cost.
- 3 km resolution
- 1/300th the cost of X-SHiELD
- 3% accuracy for rare events
- 10-year simulation
What you need to act on it
- HiROv2 requires ACE tuned to match X-SHiELD
- Current version limited to X-SHiELD-based climates
Sources