|
Activity Number:
|
157
- Compressing Climate Model Data: Lowering Storage Burden While Preserving Information
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, July 31, 2017 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics and the Environment
|
|
Abstract #323037
|
|
|
Title:
|
Statistical Compression and Conditional Emulation
|
|
Author(s):
|
Joseph Guinness* and Dorit Hammerling
|
|
Companies:
|
NC State University and National Center for Atmospheric Research
|
|
Keywords:
|
climate ;
massive data ;
computational algorithms ;
Gaussian process
|
|
Abstract:
|
This talk outlines recently proposed methodology for compressing massive space-time datasets resulting from the output of high resolution numerical climate simulations. The algorithms rely on the general idea of storing summary statistics--a lower-dimensional representation of the data--and estimating a statistical model for the conditional distribution of the full dataset given its lower-dimensional representation. The decompression algorithm uses the conditional distribution to generate realizations representing the original data. The choice of model and summary statistics is governed both by the quality of the conditional model and our ability to generate realizations quickly.
|
Authors who are presenting talks have a * after their name.
