• Skip to Content
  • Skip to Main Navigation
  • Skip to Search

Indiana University Bloomington Indiana University Bloomington IU Bloomington

Open Search
  • People
  • Research
  • Research Highlights
  • Contact Us

Climate Modeling at IU

  • Home
  • People
  • Research
  • Research Highlights
  • Contact Us
  • Search
  • Home
  • Research Highlights
  • Machine Learning for Climate Predictability

Machine Learning for Climate Predictability

Tuesday, September 22, 2020

By using an 18-layer residual recurrent neural network (left), we were able to obtain a surrogate of climate model precipitation that generally matches the output of the climate model:  the map shows differences between the climate model and the surrogate.

Climate models are the gold standard for conducting studies of what the climate might look like in the future.  But they are very computationally expensive – running a single simulation out to the end of the century can take weeks on a supercomputer.  Is there a faster way to get information that might not be perfect but is good enough?

Working with colleagues from Western Washington University and Pacific Northwest National Laboratory, Professor Ben Kravitz of Indiana University trained a deep learning algorithm on climate model output to create a surrogate – a model of precipitation over the 21st century that roughly matches the output of the climate model but can run in seconds.  Our model outperformed several metrics of skill for forecasts out to 120 months.  In doing this, not only did we successfully build a surrogate, but we also learned more about how well deep learning works in the geosciences.

Read the article

Climate Modeling at IU resources

  • Earth and Atmospheric Sciences

Indiana University

Accessibility | College Scorecard | Open to All | Privacy Notice | Copyright © 2025 The Trustees of Indiana University