Enter Note Done
Go to previous page in this tab
Session
  • Presentation | GC52B: Downscaling and Postprocessing at Weather and Climate Scales: Development and Evaluation of Methods, Products, and Applications II Oral
  • Oral
  • Share Icon
    Map Icon
    Bookmark Icon
  • GC52B-03: HiRO: A Diffusion-Based Downscaling Tool for the Ai2 Climate Emulator (ACE)
    • Schedule
    Notes
  • Location Icon267-268
    NOLA CC
    Set Timezone
    •  
    View Map

Generic 'disconnected' Message
Author(s):
W. Perkins, Allen Institute for AI (First Author)
Anna Kwa, Allen Institute for AI
Spencer Clark, Allen Institute for AI
Gideon Dresdner, Brightband
Jeremy McGibbon, Allen Institute for AI
Oliver Watt-Meyer, Allen Institute for AI
Brian Henn, Allen Institute for AI
Elynn Wu, Allen Institute for AI
James Duncan, Allen Institute for AI
Troy Arcomano, Allen Institute for AI
Christopher Bretherton, Allen Institute for AI (Presenting Author)


We introduce HiRO (High-Resolution Output), a new machine learning (ML) model that turns low-resolution atmospheric data into detailed fine-scale maps, making it easier to study local impacts of climate change. HiRO is easy to use, runs quickly and works almost anywhere on the globe. By learning from two years of fine-scale atmospheric simulation, HiRO can accurately recreate a wide range of climate features for wind and rainfall, including rare and extreme events. We also explore how HiRO can work together with another open-source ML climate prediction tool, ACE (the Ai2 Climate Emulator), to provide a combined package for climate prediction and impact assessment.



Scientific Discipline
Suggested Itineraries
Neighborhood
Type
Main Session
Discussion