Enter Note Done
Go to previous page in this tab
Session
  • Presentation | PP31C: Applying Machine Learning to Better Reconstruct and Understand Paleoclimates and Paleoecology Poster
  • Poster
  • Share Icon
    Bookmark Icon
  • PP31C-1050: High-Resolution Paleoclimate Fields via Diffusion-Based Downscaling
    • Schedule
    Notes
  • Board 1050‚ Hall EFG (Poster Hall)
    NOLA CC
    Set Timezone

Generic 'disconnected' Message
Author(s):
Deborah Khider, University of Southern California (First Author, Presenting Author)
Armen Kemanian, Pennsylvania State University Main Campus


Understanding how climate has changed over the past 1,200 years can help scientists learn how regional environments responded to droughts, warming, and other natural variations. However, most reconstructions of past climate are only available at coarse spatial resolution, making it difficult to study impacts at local scales—such as in agriculture or water resources. In this project, we use a modern artificial intelligence technique called a diffusion model to increase the spatial detail of these past climate reconstructions. Diffusion models are commonly used to generate high-resolution images, and we adapt this approach to enhance low-resolution climate maps.


These enhanced datasets can be used to study past climate variability and its effects on societies, including questions about drought and agriculture during periods like the Medieval Megadroughts and the Mayan Collapse.




Scientific Discipline
Suggested Itineraries
Neighborhood
Type
Main Session
Discussion