DOE PAGES title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Accelerating Radiation Computations for Dynamical Models With Targeted Machine Learning and Code Optimization

Abstract

Atmospheric radiation is the main driver of weather and climate, yet due to a complicated absorption spectrum, the precise treatment of radiative transfer in numerical weather and climate models is computationally unfeasible. Radiation parameterizations need to maximize computational efficiency as well as accuracy, and for predicting the future climate many greenhouse gases need to be included. In this work, neural networks (NNs) were developed to replace the gas optics computations in a modern radiation scheme (RTE+RRTMGP) by using carefully constructed models and training data. The NNs, implemented in Fortran and utilizing BLAS for batched inference, are faster by a factor of 1–6, depending on the software and hardware platforms. We combined the accelerated gas optics with a refactored radiative transfer solver, resulting in clear-sky longwave (shortwave) fluxes being 3.5 (1.8) faster to compute on an Intel platform. The accuracy, evaluated with benchmark line-by-line computations across a large range of atmospheric conditions, is very similar to the original scheme with errors in heating rates and top-of-atmosphere radiative forcings typically below 0.1 K day-1 and 0.5 W m-2, respectively. These results show that targeted machine learning, code restructuring techniques, and the use of numerical libraries can yield material gains in efficiency whilemore » retaining accuracy.« less

Authors:
ORCiD logo [1]; ORCiD logo [2]; ORCiD logo [3];  [4]; ORCiD logo [5]
  1. Danish Meteorological Inst., Copenhagen (Denmark); Univ. of Copenhagen (Denmark)
  2. Univ. of Colorado, Boulder, CO (United States); National Oceanic and Atmospheric Administration (NOAA), Boulder, CO (United States)
  3. European Centre for Medium‐Range Weather Forecasts, Reading (United Kingdom)
  4. Danish Meteorological Inst., Copenhagen (Denmark); European Centre for Medium‐Range Weather Forecasts, Reading (United Kingdom)
  5. Univ. of Copenhagen (Denmark)
Publication Date:
Research Org.:
Univ. of Colorado, Boulder, CO (United States)
Sponsoring Org.:
USDOE Office of Science (SC), Biological and Environmental Research (BER); European Commission Horizon 2020
OSTI Identifier:
1734807
Alternate Identifier(s):
OSTI ID: 1786719; OSTI ID: 1852336
Grant/Contract Number:  
SC0016593; 800897; DE‐SC0016593
Resource Type:
Published Article
Journal Name:
Journal of Advances in Modeling Earth Systems
Additional Journal Information:
Journal Volume: 12; Journal Issue: 12; Journal ID: ISSN 1942-2466
Publisher:
American Geophysical Union (AGU)
Country of Publication:
United States
Language:
English
Subject:
37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CHEMISTRY

Citation Formats

Ukkonen, Peter, Pincus, Robert, Hogan, Robin J., Pagh Nielsen, Kristian, and Kaas, Eigil. Accelerating Radiation Computations for Dynamical Models With Targeted Machine Learning and Code Optimization. United States: N. p., 2020. Web. doi:10.1029/2020ms002226.
Ukkonen, Peter, Pincus, Robert, Hogan, Robin J., Pagh Nielsen, Kristian, & Kaas, Eigil. Accelerating Radiation Computations for Dynamical Models With Targeted Machine Learning and Code Optimization. United States. https://doi.org/10.1029/2020ms002226
Ukkonen, Peter, Pincus, Robert, Hogan, Robin J., Pagh Nielsen, Kristian, and Kaas, Eigil. Sun . "Accelerating Radiation Computations for Dynamical Models With Targeted Machine Learning and Code Optimization". United States. https://doi.org/10.1029/2020ms002226.
@article{osti_1734807,
title = {Accelerating Radiation Computations for Dynamical Models With Targeted Machine Learning and Code Optimization},
author = {Ukkonen, Peter and Pincus, Robert and Hogan, Robin J. and Pagh Nielsen, Kristian and Kaas, Eigil},
abstractNote = {Atmospheric radiation is the main driver of weather and climate, yet due to a complicated absorption spectrum, the precise treatment of radiative transfer in numerical weather and climate models is computationally unfeasible. Radiation parameterizations need to maximize computational efficiency as well as accuracy, and for predicting the future climate many greenhouse gases need to be included. In this work, neural networks (NNs) were developed to replace the gas optics computations in a modern radiation scheme (RTE+RRTMGP) by using carefully constructed models and training data. The NNs, implemented in Fortran and utilizing BLAS for batched inference, are faster by a factor of 1–6, depending on the software and hardware platforms. We combined the accelerated gas optics with a refactored radiative transfer solver, resulting in clear-sky longwave (shortwave) fluxes being 3.5 (1.8) faster to compute on an Intel platform. The accuracy, evaluated with benchmark line-by-line computations across a large range of atmospheric conditions, is very similar to the original scheme with errors in heating rates and top-of-atmosphere radiative forcings typically below 0.1 K day-1 and 0.5 W m-2, respectively. These results show that targeted machine learning, code restructuring techniques, and the use of numerical libraries can yield material gains in efficiency while retaining accuracy.},
doi = {10.1029/2020ms002226},
journal = {Journal of Advances in Modeling Earth Systems},
number = 12,
volume = 12,
place = {United States},
year = {2020},
month = {11}
}

Journal Article:
Free Publicly Available Full Text
Publisher's Version of Record
https://doi.org/10.1029/2020ms002226

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

Computational investigations of low-discrepancy sequences
journal, June 1997

  • Kocis, Ladislav; Whiten, William J.
  • ACM Transactions on Mathematical Software, Vol. 23, Issue 2
  • DOI: 10.1145/264029.264064

Atmospheric radiative transfer modeling: a summary of the AER codes
journal, March 2005

  • Clough, S. A.; Shephard, M. W.; Mlawer, E. J.
  • Journal of Quantitative Spectroscopy and Radiative Transfer, Vol. 91, Issue 2
  • DOI: 10.1016/j.jqsrt.2004.05.058

Accurate and Fast Neural Network Emulations of Model Radiation for the NCEP Coupled Climate Forecast System: Climate Simulations and Seasonal Predictions*
journal, May 2010

  • Krasnopolsky, V. M.; Fox-Rabinovitz, M. S.; Hou, Y. T.
  • Monthly Weather Review, Vol. 138, Issue 5
  • DOI: 10.1175/2009MWR3149.1

The correlated-k method for radiation calculations in nonhomogeneous atmospheres
journal, December 1989

  • Goody, Richard; West, Robert; Chen, Luke
  • Journal of Quantitative Spectroscopy and Radiative Transfer, Vol. 42, Issue 6
  • DOI: 10.1016/0022-4073(89)90044-7

Radiative forcing by long-lived greenhouse gases: Calculations with the AER radiative transfer models
journal, January 2008

  • Iacono, Michael J.; Delamere, Jennifer S.; Mlawer, Eli J.
  • Journal of Geophysical Research, Vol. 113, Issue D13
  • DOI: 10.1029/2008JD009944

Single-precision arithmetic in ECHAM radiation reduces runtime and energy consumption
journal, January 2020


Replicated Computational Results Certified BLIS: A Framework for Rapidly Instantiating BLAS Functionality
journal, June 2015

  • Zee, Field G. VAN; Geijn, Robert A. VAN DE
  • ACM Transactions on Mathematical Software, Vol. 41, Issue 3
  • DOI: 10.1145/2764454

Benchmark Calculations of Radiative Forcing by Greenhouse Gases
journal, December 2020

  • Pincus, Robert; Buehler, Stefan A.; Brath, Manfred
  • Journal of Geophysical Research: Atmospheres, Vol. 125, Issue 23
  • DOI: 10.1029/2020JD033483

The CAMS reanalysis of atmospheric composition
journal, January 2019

  • Inness, Antje; Ades, Melanie; Agustí-Panareda, Anna
  • Atmospheric Chemistry and Physics, Vol. 19, Issue 6
  • DOI: 10.5194/acp-19-3515-2019

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

Smartphone Pressure Collection and Bias Correction Using Machine Learning
journal, March 2018

  • McNicholas, Conor; Mass, Clifford F.
  • Journal of Atmospheric and Oceanic Technology, Vol. 35, Issue 3
  • DOI: 10.1175/JTECH-D-17-0096.1

Using Deep Neural Networks as Cost‐Effective Surrogate Models for Super‐Parameterized E3SM Radiative Transfer
journal, June 2019

  • Pal, Anikesh; Mahajan, Salil; Norman, Matthew R.
  • Geophysical Research Letters, Vol. 46, Issue 11
  • DOI: 10.1029/2018GL081646

The Radiative Forcing Model Intercomparison Project (RFMIP): experimental protocol for CMIP6
journal, January 2016

  • Pincus, Robert; Forster, Piers M.; Stevens, Bjorn
  • Geoscientific Model Development, Vol. 9, Issue 9
  • DOI: 10.5194/gmd-9-3447-2016

Optimizing Weather Model Radiative Transfer Physics for Intel’s Many Integrated Core (MIC) Architecture
journal, December 2016

  • Michalakes, John; Iacono, Michael J.; Jessup, Elizabeth R.
  • Parallel Processing Letters, Vol. 26, Issue 04
  • DOI: 10.1142/S0129626416500195

Deep learning to represent subgrid processes in climate models
journal, September 2018

  • Rasp, Stephan; Pritchard, Michael S.; Gentine, Pierre
  • Proceedings of the National Academy of Sciences, Vol. 115, Issue 39
  • DOI: 10.1073/pnas.1810286115

A Simulation of a Supercell Thunderstorm with Emulated Radiative Cooling beneath the Anvil
journal, July 2005

  • Markowski, Paul M.; Harrington, Jerry Y.
  • Journal of the Atmospheric Sciences, Vol. 62, Issue 7
  • DOI: 10.1175/JAS3497.1

Evaluation of Machine Learning Classifiers for Predicting Deep Convection
journal, June 2019

  • Ukkonen, Peter; Mäkelä, Antti
  • Journal of Advances in Modeling Earth Systems, Vol. 11, Issue 6
  • DOI: 10.1029/2018MS001561

Hyperopt: a Python library for model selection and hyperparameter optimization
journal, January 2015


Leveraging Modern Artificial Intelligence for Remote Sensing and NWP: Benefits and Challenges
journal, December 2019

  • Boukabara, Sid-Ahmed; Krasnopolsky, Vladimir; Stewart, Jebb Q.
  • Bulletin of the American Meteorological Society, Vol. 100, Issue 12
  • DOI: 10.1175/BAMS-D-18-0324.1

Initial Results From the Super‐Parameterized E3SM
journal, January 2020

  • Hannah, W. M.; Jones, C. R.; Hillman, B. R.
  • Journal of Advances in Modeling Earth Systems, Vol. 12, Issue 1
  • DOI: 10.1029/2019MS001863

Balancing Accuracy, Efficiency, and Flexibility in Radiation Calculations for Dynamical Models
journal, October 2019

  • Pincus, Robert; Mlawer, Eli J.; Delamere, Jennifer S.
  • Journal of Advances in Modeling Earth Systems, Vol. 11, Issue 10
  • DOI: 10.1029/2019MS001621

Radiance and Jacobian intercomparison of radiative transfer models applied to HIRS and AMSU channels
journal, October 2001

  • Garand, L.; Turner, D. S.; Larocque, M.
  • Journal of Geophysical Research: Atmospheres, Vol. 106, Issue D20
  • DOI: 10.1029/2000JD000184

The Full-Spectrum Correlated- k Method for Longwave Atmospheric Radiative Transfer Using an Effective Planck Function
journal, June 2010


On the Correlated k -Distribution Method for Radiative Transfer in Nonhomogeneous Atmospheres
journal, November 1992


Developments in the MPI‐M Earth System Model version 1.2 (MPI‐ESM1.2) and Its Response to Increasing CO 2
journal, April 2019

  • Mauritsen, Thorsten; Bader, Jürgen; Becker, Tobias
  • Journal of Advances in Modeling Earth Systems, Vol. 11, Issue 4
  • DOI: 10.1029/2018MS001400

A parallel Fortran framework for neural networks and deep learning
journal, March 2019


A Flexible and Efficient Radiation Scheme for the ECMWF Model
journal, August 2018

  • Hogan, Robin J.; Bozzo, Alessio
  • Journal of Advances in Modeling Earth Systems, Vol. 10, Issue 8
  • DOI: 10.1029/2018MS001364

Works referencing / citing this record:

Robustness of neural network emulations of radiative transfer parameterizations in a state-of-the-art general circulation model
journal, December 2021

  • Belochitski, Alexei; Krasnopolsky, Vladimir
  • Geoscientific Model Development, Vol. 14, Issue 12
  • DOI: 10.5194/gmd-14-7425-2021