skip to main content
DOE PAGES title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Deep learning to represent subgrid processes in climate models

Abstract

The representation of nonlinear subgrid processes, especially clouds, has been a major source of uncertainty in climate models for decades. Cloud-resolving models better represent many of these processes and can now be run globally but only for short-term simulations of at most a few years because of computational limitations. Here we demonstrate that deep learning can be used to capture many advantages of cloud-resolving modeling at a fraction of the computational cost. We train a deep neural network to represent all atmospheric subgrid processes in a climate model by learning from a multiscale model in which convection is treated explicitly. The trained neural network then replaces the traditional subgrid parameterizations in a global general circulation model in which it freely interacts with the resolved dynamics and the surface-flux scheme. The prognostic multiyear simulations are stable and closely reproduce not only the mean climate of the cloud-resolving simulation but also key aspects of variability, including precipitation extremes and the equatorial wave spectrum. Furthermore, the neural network approximately conserves energy despite not being explicitly instructed to. Finally, we show that the neural network parameterization generalizes to new surface forcing patterns but struggles to cope with temperatures far outside its training manifold. Ourmore » results show the feasibility of using deep learning for climate model parameterization. In a broader context, we anticipate that data-driven Earth system model development could play a key role in reducing climate prediction uncertainty in the coming decade.« less

Authors:
ORCiD logo; ;
Publication Date:
Research Org.:
Univ. of California, Irvine, CA (United States); Columbia Univ., New York, NY (United States)
Sponsoring Org.:
USDOE Office of Science (SC), Advanced Scientific Computing Research (ASCR) (SC-21). Scientific Discovery through Advanced Computing (SciDAC); NSF Extreme Science and Engineering Discovery Environment (XSEDE)
OSTI Identifier:
1468882
Alternate Identifier(s):
OSTI ID: 1547350
Grant/Contract Number:  
SC0012152; SC00-12548; SC0014203; SC0012548
Resource Type:
Published Article
Journal Name:
Proceedings of the National Academy of Sciences of the United States of America
Additional Journal Information:
Journal Name: Proceedings of the National Academy of Sciences of the United States of America Journal Volume: 115 Journal Issue: 39; Journal ID: ISSN 0027-8424
Publisher:
Proceedings of the National Academy of Sciences
Country of Publication:
United States
Language:
English
Subject:
54 ENVIRONMENTAL SCIENCES; climate modeling; deep learning; subgrid parameterization; convection

Citation Formats

Rasp, Stephan, Pritchard, Michael S., and Gentine, Pierre. Deep learning to represent subgrid processes in climate models. United States: N. p., 2018. Web. doi:10.1073/pnas.1810286115.
Rasp, Stephan, Pritchard, Michael S., & Gentine, Pierre. Deep learning to represent subgrid processes in climate models. United States. doi:10.1073/pnas.1810286115.
Rasp, Stephan, Pritchard, Michael S., and Gentine, Pierre. Thu . "Deep learning to represent subgrid processes in climate models". United States. doi:10.1073/pnas.1810286115.
@article{osti_1468882,
title = {Deep learning to represent subgrid processes in climate models},
author = {Rasp, Stephan and Pritchard, Michael S. and Gentine, Pierre},
abstractNote = {The representation of nonlinear subgrid processes, especially clouds, has been a major source of uncertainty in climate models for decades. Cloud-resolving models better represent many of these processes and can now be run globally but only for short-term simulations of at most a few years because of computational limitations. Here we demonstrate that deep learning can be used to capture many advantages of cloud-resolving modeling at a fraction of the computational cost. We train a deep neural network to represent all atmospheric subgrid processes in a climate model by learning from a multiscale model in which convection is treated explicitly. The trained neural network then replaces the traditional subgrid parameterizations in a global general circulation model in which it freely interacts with the resolved dynamics and the surface-flux scheme. The prognostic multiyear simulations are stable and closely reproduce not only the mean climate of the cloud-resolving simulation but also key aspects of variability, including precipitation extremes and the equatorial wave spectrum. Furthermore, the neural network approximately conserves energy despite not being explicitly instructed to. Finally, we show that the neural network parameterization generalizes to new surface forcing patterns but struggles to cope with temperatures far outside its training manifold. Our results show the feasibility of using deep learning for climate model parameterization. In a broader context, we anticipate that data-driven Earth system model development could play a key role in reducing climate prediction uncertainty in the coming decade.},
doi = {10.1073/pnas.1810286115},
journal = {Proceedings of the National Academy of Sciences of the United States of America},
number = 39,
volume = 115,
place = {United States},
year = {2018},
month = {9}
}

Journal Article:
Free Publicly Available Full Text
Publisher's Version of Record
DOI: 10.1073/pnas.1810286115

Citation Metrics:
Cited by: 14 works
Citation information provided by
Web of Science

Save / Share:

Works referenced in this record:

Could Machine Learning Break the Convection Parameterization Deadlock?
journal, June 2018

  • Gentine, P.; Pritchard, M.; Rasp, S.
  • Geophysical Research Letters, Vol. 45, Issue 11
  • DOI: 10.1029/2018GL078202

Marine boundary layer clouds at the heart of tropical cloud feedback uncertainties in climate models
journal, January 2005


What favors convective aggregation and why?: WHAT FAVORS CONVECTIVE AGGREGATION
journal, July 2015

  • Muller, Caroline; Bony, Sandrine
  • Geophysical Research Letters, Vol. 42, Issue 13
  • DOI: 10.1002/2015GL064260

The vertical distribution of cloud feedback in coupled ocean-atmosphere models: VERTICAL DISTRIBUTION OF CLOUD FEEDBACK
journal, June 2011

  • Soden, Brian J.; Vecchi, Gabriel A.
  • Geophysical Research Letters, Vol. 38, Issue 12
  • DOI: 10.1029/2011GL047632

Structure of the Madden–Julian Oscillation in the Superparameterized CAM
journal, November 2009

  • Benedict, James J.; Randall, David A.
  • Journal of the Atmospheric Sciences, Vol. 66, Issue 11
  • DOI: 10.1175/2009JAS3030.1

Deep learning
journal, May 2015

  • LeCun, Yann; Bengio, Yoshua; Hinton, Geoffrey
  • Nature, Vol. 521, Issue 7553
  • DOI: 10.1038/nature14539

Rediscovery of the doldrums in storm-resolving simulations over the tropical Atlantic
journal, November 2017


Using Ensemble of Neural Networks to Learn Stochastic Convection Parameterizations for Climate and Numerical Weather Prediction Models from Data Simulated by a Cloud Resolving Model
journal, January 2013

  • Krasnopolsky, Vladimir M.; Fox-Rabinovitz, Michael S.; Belochitski, Alexei A.
  • Advances in Artificial Neural Systems, Vol. 2013
  • DOI: 10.1155/2013/485913

The Formulation and Atmospheric Simulation of the Community Atmosphere Model Version 3 (CAM3)
journal, June 2006

  • Collins, William D.; Rasch, Philip J.; Boville, Byron A.
  • Journal of Climate, Vol. 19, Issue 11
  • DOI: 10.1175/JCLI3760.1

What Are Climate Models Missing?
journal, May 2013


Using Machine Learning to Parameterize Moist Convection: Potential for Modeling of Climate, Climate Change, and Extreme Events
journal, October 2018

  • O'Gorman, Paul A.; Dwyer, John G.
  • Journal of Advances in Modeling Earth Systems, Vol. 10, Issue 10
  • DOI: 10.1029/2018MS001351

Restricting 32-128 km horizontal scales hardly affects the MJO in the Superparameterized Community Atmosphere Model v.3.0 but the number of cloud-resolving grid columns constrains vertical mixing
journal, August 2014

  • Pritchard, Michael S.; Bretherton, Christopher S.; DeMott, Charlotte A.
  • Journal of Advances in Modeling Earth Systems, Vol. 6, Issue 3
  • DOI: 10.1002/2014MS000340

Deep moist atmospheric convection in a subkilometer global simulation
journal, September 2013

  • Miyamoto, Yoshiaki; Kajikawa, Yoshiyuki; Yoshida, Ryuji
  • Geophysical Research Letters, Vol. 40, Issue 18
  • DOI: 10.1002/grl.50944

Evaluation of the convection-resolving climate modeling approach on continental scales: CONVECTION-RESOLVING CONTINENTAL-SCALE SIMULATION
journal, May 2017

  • Leutwyler, David; Lüthi, Daniel; Ban, Nikolina
  • Journal of Geophysical Research: Atmospheres, Vol. 122, Issue 10
  • DOI: 10.1002/2016JD026013

Convectively Coupled Equatorial Waves: Analysis of Clouds and Temperature in the Wavenumber–Frequency Domain
journal, February 1999


Spread in model climate sensitivity traced to atmospheric convective mixing
journal, January 2014

  • Sherwood, Steven C.; Bony, Sandrine; Dufresne, Jean-Louis
  • Nature, Vol. 505, Issue 7481
  • DOI: 10.1038/nature12829

Prognostic Validation of a Neural Network Unified Physics Parameterization
journal, June 2018

  • Brenowitz, N. D.; Bretherton, C. S.
  • Geophysical Research Letters, Vol. 45, Issue 12
  • DOI: 10.1029/2018GL078510

Rainfall From Resolved Rather Than Parameterized Processes Better Represents the Present‐Day and Climate Change Response of Moderate Rates in the Community Atmosphere Model
journal, April 2018

  • Kooperman, Gabriel J.; Pritchard, Michael S.; O'Brien, Travis A.
  • Journal of Advances in Modeling Earth Systems, Vol. 10, Issue 4
  • DOI: 10.1002/2017MS001188

The double ITCZ bias in CMIP5 models: interaction between SST, large-scale circulation and precipitation
journal, January 2015


A cloud resolving model as a cloud parameterization in the NCAR Community Climate System Model: Preliminary results
journal, September 2001

  • Khairoutdinov, Marat F.; Randall, David A.
  • Geophysical Research Letters, Vol. 28, Issue 18
  • DOI: 10.1029/2001GL013552

The Resolution Dependence of Explicitly Modeled Convective Systems
journal, April 1997


Convective self-aggregation feedbacks in near-global cloud-resolving simulations of an aquaplanet: AGGREGATION FEEDBACKS IN NEAR-GLOBAL CRM
journal, November 2015

  • Bretherton, Christopher S.; Khairoutdinov, Marat F.
  • Journal of Advances in Modeling Earth Systems, Vol. 7, Issue 4
  • DOI: 10.1002/2015MS000499

A Probabilistic Bulk Model of Coupled Mixed Layer and Convection. Part I: Clear-Sky Case
journal, June 2013

  • Gentine, Pierre; Betts, Alan K.; Lintner, Benjamin R.
  • Journal of the Atmospheric Sciences, Vol. 70, Issue 6
  • DOI: 10.1175/JAS-D-12-0145.1

Global-scale convective aggregation: Implications for the Madden-Julian Oscillation: GLOBAL-SCALE CONVECTIVE AGGREGATION
journal, October 2015

  • Arnold, Nathan P.; Randall, David A.
  • Journal of Advances in Modeling Earth Systems, Vol. 7, Issue 4
  • DOI: 10.1002/2015MS000498

Moist Static Energy Budget of MJO-like Disturbances in the Atmosphere of a Zonally Symmetric Aquaplanet
journal, April 2012


Earth System Modeling 2.0: A Blueprint for Models That Learn From Observations and Targeted High-Resolution Simulations: EARTH SYSTEM MODELING 2.0
journal, December 2017

  • Schneider, Tapio; Lan, Shiwei; Stuart, Andrew
  • Geophysical Research Letters, Vol. 44, Issue 24
  • DOI: 10.1002/2017GL076101

Climate goals and computing the future of clouds
journal, January 2017

  • Schneider, Tapio; Teixeira, João; Bretherton, Christopher S.
  • Nature Climate Change, Vol. 7, Issue 1
  • DOI: 10.1038/nclimate3190

Effects of explicit convection on global land-atmosphere coupling in the superparameterized CAM: EXPLICIT CONVECTION ON GLOBAL LAND-ATMOSPHERE COUPLING
journal, August 2016

  • Sun, Jian; Pritchard, Michael S.
  • Journal of Advances in Modeling Earth Systems, Vol. 8, Issue 3
  • DOI: 10.1002/2016MS000689

The Art and Science of Climate Model Tuning
journal, March 2017

  • Hourdin, Frédéric; Mauritsen, Thorsten; Gettelman, Andrew
  • Bulletin of the American Meteorological Society, Vol. 98, Issue 3
  • DOI: 10.1175/BAMS-D-15-00135.1

Clouds, circulation and climate sensitivity
journal, March 2015

  • Bony, Sandrine; Stevens, Bjorn; Frierson, Dargan M. W.
  • Nature Geoscience, Vol. 8, Issue 4
  • DOI: 10.1038/ngeo2398

Resolution Dependence of the Diurnal Cycle of Precipitation Simulated by a Global Cloud-System Resolving Model
journal, January 2016

  • Yashiro, Hisashi; Kajikawa, Yoshiyuki; Miyamoto, Yoshiaki
  • SOLA, Vol. 12, Issue 0
  • DOI: 10.2151/sola.2016-053

Random Forests
journal, January 2001