Using Deep Neural Networks as Cost-Effective Surrogate Models for Super-Parameterized E3SM Radiative Transfer

Anikesh Pal; Salil Mahajan; Matthew R. Norman

doi:10.1029/2018GL081646

Using Deep Neural Networks as Cost-Effective Surrogate Models for Super-Parameterized E3SM Radiative Transfer

Anikesh Pal, Salil Mahajan, Matthew R. Norman

Research output: Contribution to journal › Article › peer-review

51 Scopus citations

Abstract

Deep neural networks (DNNs) are implemented in Super-Parameterized Energy Exascale Earth System Model (SP-E3SM) to imitate the shortwave and longwave radiative transfer calculations. These DNNs were able to emulate the radiation parameters with an accuracy of 90–95% at a cost of 8–10 times cheaper than the original radiation parameterization. A comparison of time-averaged radiative fluxes and the prognostic variables manifested qualitative and quantitative similarity between the DNN emulation and the original parameterization. It has also been found that the differences between the DNN emulation and the original parameterization are comparable to the internal variability of the original parameterization. Although the DNNs developed in this investigation emulate the radiation parameters for a specific set of initial conditions, the results justify the need of further research to generalize the use of DNNs for the emulations of full model radiation and other parameterization for seasonal predictions and climate simulations.

Original language	English
Pages (from-to)	6069-6079
Number of pages	11
Journal	Geophysical Research Letters
Volume	46
Issue number	11
DOIs	https://doi.org/10.1029/2018GL081646
State	Published - Jun 16 2019

Funding

This research used resources of the Oak Ridge Leadership Computing Facility, which is a DOE Office of Science User Facility supported under contract DE-AC05-00OR22725. This research was partially supported by the Energy Exascale Earth System Model (E3SM) project, funded by the U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research. In the originally published version of this paper, there was an error present in the abstract; also, the following statement was omitted from the acknowledgement section: “This research was partially supported by the Energy Exascale Earth System Model (E3SM) project, funded by the U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research.” These errors have since been corrected, and the present version may be considered the authoritative version of record.

Keywords

deep neural networks
general circulation models
radiation models

Access to Document

10.1029/2018GL081646

Cite this

@article{6d86252ed8ff4d6b9c2639f51f07d560,

title = "Using Deep Neural Networks as Cost-Effective Surrogate Models for Super-Parameterized E3SM Radiative Transfer",

abstract = "Deep neural networks (DNNs) are implemented in Super-Parameterized Energy Exascale Earth System Model (SP-E3SM) to imitate the shortwave and longwave radiative transfer calculations. These DNNs were able to emulate the radiation parameters with an accuracy of 90–95% at a cost of 8–10 times cheaper than the original radiation parameterization. A comparison of time-averaged radiative fluxes and the prognostic variables manifested qualitative and quantitative similarity between the DNN emulation and the original parameterization. It has also been found that the differences between the DNN emulation and the original parameterization are comparable to the internal variability of the original parameterization. Although the DNNs developed in this investigation emulate the radiation parameters for a specific set of initial conditions, the results justify the need of further research to generalize the use of DNNs for the emulations of full model radiation and other parameterization for seasonal predictions and climate simulations.",

keywords = "deep neural networks, general circulation models, radiation models",

author = "Anikesh Pal and Salil Mahajan and Norman, {Matthew R.}",

year = "2019",

month = jun,

day = "16",

doi = "10.1029/2018GL081646",

language = "English",

volume = "46",

pages = "6069--6079",

journal = "Geophysical Research Letters",

issn = "0094-8276",

publisher = "Wiley Open Access",

number = "11",

}

TY - JOUR

T1 - Using Deep Neural Networks as Cost-Effective Surrogate Models for Super-Parameterized E3SM Radiative Transfer

AU - Pal, Anikesh

AU - Mahajan, Salil

AU - Norman, Matthew R.

PY - 2019/6/16

Y1 - 2019/6/16

N2 - Deep neural networks (DNNs) are implemented in Super-Parameterized Energy Exascale Earth System Model (SP-E3SM) to imitate the shortwave and longwave radiative transfer calculations. These DNNs were able to emulate the radiation parameters with an accuracy of 90–95% at a cost of 8–10 times cheaper than the original radiation parameterization. A comparison of time-averaged radiative fluxes and the prognostic variables manifested qualitative and quantitative similarity between the DNN emulation and the original parameterization. It has also been found that the differences between the DNN emulation and the original parameterization are comparable to the internal variability of the original parameterization. Although the DNNs developed in this investigation emulate the radiation parameters for a specific set of initial conditions, the results justify the need of further research to generalize the use of DNNs for the emulations of full model radiation and other parameterization for seasonal predictions and climate simulations.

AB - Deep neural networks (DNNs) are implemented in Super-Parameterized Energy Exascale Earth System Model (SP-E3SM) to imitate the shortwave and longwave radiative transfer calculations. These DNNs were able to emulate the radiation parameters with an accuracy of 90–95% at a cost of 8–10 times cheaper than the original radiation parameterization. A comparison of time-averaged radiative fluxes and the prognostic variables manifested qualitative and quantitative similarity between the DNN emulation and the original parameterization. It has also been found that the differences between the DNN emulation and the original parameterization are comparable to the internal variability of the original parameterization. Although the DNNs developed in this investigation emulate the radiation parameters for a specific set of initial conditions, the results justify the need of further research to generalize the use of DNNs for the emulations of full model radiation and other parameterization for seasonal predictions and climate simulations.

KW - deep neural networks

KW - general circulation models

KW - radiation models

UR - http://www.scopus.com/inward/record.url?scp=85067385764&partnerID=8YFLogxK

U2 - 10.1029/2018GL081646

DO - 10.1029/2018GL081646

M3 - Article

AN - SCOPUS:85067385764

SN - 0094-8276

VL - 46

SP - 6069

EP - 6079

JO - Geophysical Research Letters

JF - Geophysical Research Letters

IS - 11

ER -

Using Deep Neural Networks as Cost-Effective Surrogate Models for Super-Parameterized E3SM Radiative Transfer

Abstract

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this