Journal cover Journal topic
Atmospheric Chemistry and Physics An interactive open-access journal of the European Geosciences Union
Journal topic

Journal metrics

Journal metrics

  • IF value: 5.509 IF 5.509
  • IF 5-year value: 5.689 IF 5-year
    5.689
  • CiteScore value: 5.44 CiteScore
    5.44
  • SNIP value: 1.519 SNIP 1.519
  • SJR value: 3.032 SJR 3.032
  • IPP value: 5.37 IPP 5.37
  • h5-index value: 86 h5-index 86
  • Scimago H <br class='hide-on-tablet hide-on-mobile'>index value: 161 Scimago H
    index 161
Discussion papers
https://doi.org/10.5194/acp-2019-85
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.
https://doi.org/10.5194/acp-2019-85
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

Technical note 11 Apr 2019

Technical note | 11 Apr 2019

Review status
This discussion paper is a preprint. It is a manuscript under review for the journal Atmospheric Chemistry and Physics (ACP).

Technical Note: Deep Learning for Creating Surrogate Models of Precipitation in Earth System Models

Theodore Weber1, Austin Corotan1, Brian Hutchinson1,2, Ben Kravitz3,4, and Robert Link5 Theodore Weber et al.
  • 1Computer Science Department, Western Washington University, Bellingham, WA
  • 2Computing and Analytics Division, Pacific Northwest National Laboratory, Seattle, WA
  • 3Department of Earth and Atmospheric Sciences, Indiana University, Bloomington, IN
  • 4Atmospheric Sciences and Global Change Division, Pacific Northwest National Laboratory, Richland, WA
  • 5Joint Global Change Research Institute, Pacific Northwest National Laboratory, College Park, MD

Abstract. We investigate techniques for using deep neural networks to produce surrogate models for short term climate forecasts. A convolutional neural network is trained on 97 years of monthly precipitation output from the 1pctCO2 run (the CO2 concentration increases by 1 % per year) simulated by the CanESM2 Earth System Model. The neural network clearly outperforms a persistence forecast and does not show substantially degraded performance even when the forecast length is extended to 120 months. The model is prone to underpredicting precipitation in areas characterized by intense precipitation events. Scheduled sampling (forcing the model to gradually use its own past predictions rather than ground truth) is essential for avoiding amplification of early forecasting errors. However, the use of scheduled sampling also necessitates preforecasting (generating forecasts prior to the first forecast date) to obtain adequate performance for the first few prediction time steps. We document the training procedures and hyperparameter optimization process for researchers who wish to extend the use of neural networks in developing surrogate models.

Theodore Weber et al.
Interactive discussion
Status: open (until 06 Jun 2019)
Status: open (until 06 Jun 2019)
AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment
[Subscribe to comment alert] Printer-friendly Version - Printer-friendly version Supplement - Supplement
Theodore Weber et al.
Theodore Weber et al.
Viewed  
Total article views: 105 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
75 29 1 105 1 2
  • HTML: 75
  • PDF: 29
  • XML: 1
  • Total: 105
  • BibTeX: 1
  • EndNote: 2
Views and downloads (calculated since 11 Apr 2019)
Cumulative views and downloads (calculated since 11 Apr 2019)
Viewed (geographical distribution)  
Total article views: 75 (including HTML, PDF, and XML) Thereof 72 with geography defined and 3 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 
Cited  
Saved  
No saved metrics found.
Discussed  
No discussed metrics found.
Latest update: 22 Apr 2019
Publications Copernicus
Download
Short summary
Climate model emulators can save computer time but are less accurate than full climate models. We use neural networks to build emulators of precipitation, trained on existing climate model runs. By doing so, we can capture nonlinearities and how the past state of a model (to some degree) shapes the future state. Our emulator outperforms a persistence forecast of precipitation.
Climate model emulators can save computer time but are less accurate than full climate models....
Citation