TechPrufer
Get started
Hugging Face
Hugging Face logoHugging Face · Feature trail

High-Resolution Climate Simulation

1 post · Oct 2026

What this trail covers

Hugging Face is developing capabilities for high-resolution climate simulation, enabling the generation of detailed weather data at resolutions like 3 km. This involves AI models that downscale coarser climate outputs, improving the accuracy of predicting extreme weather events such as rain, wind, and pressure. The system aims to significantly reduce the computational cost compared to traditional atmospheric models, making climate research more accessible and scalable. Future work includes expanding the system to handle different climate scenarios and adding more output fields like temperature and snow/ice.

All Hugging Face trails
  1. HiRO-ACEv2: Simulating extreme weather at 3 km resolution for a fraction of the costThis post introduces HiRO-ACEv2, an updated system for simulating extreme weather at 3 km resolution. Key changes include a new downscaling model (HiROv2) capable of predicting multiple high-resolution fields simultaneously (beyond just rainfall) and improved skill in predicting extreme events. The training methodology for HiROv2 was modified to enhance its ability to reproduce intense rain, wind, and pressure values, achieving within 3% accuracy for rare events. The system also demonstrates improved skill in estimating tropical cyclone energy and generating realistic storm structures. An updated ACE model was also trained, showing good alignment with X-SHiELD for hurricanes, average climate patterns, and distributions of weather variables over a 10-year simulation. The combined HiRO-ACEv2 system generated a full year of global 3 km data with statistics closely matching X-SHiELD, at approximately 1/300th the cost. Future work includes training on new climate datasets and adding more output fields.