Preprints
https://doi.org/10.5194/egusphere-2022-559
https://doi.org/10.5194/egusphere-2022-559
08 Jul 2022
 | 08 Jul 2022

Emulating Aerosol Optics with Randomly Generated Neural Networks

Andrew Geiss, Po-Lun Ma, Balwinder Singh, and Joseph C. Hardin

Abstract. Atmospheric aerosols have a substantial impact on climate and remain one of the largest sources of uncertainty in climate forecasts. Accurate representation of their direct radiative effects is a crucial component of modern climate models. Direct computation of the radiative properties of aerosols is far too computationally expensive to perform in a climate model however, so optical properties are typically approximated using a parameterization. This work develops artificial neural networks (ANNs) capable of replacing the current aerosol optics parameterization used in the Energy Exascale Earth System Model (E3SM). A large training dataset is generated by using Mie code to directly compute the optical properties of a range of atmospheric aerosol populations given a large variety of particle sizes, wavelengths, and refractive indices. Optimal neural architectures for shortwave and longwave bands are identified by evaluating ANNs with randomly generated wirings. Randomly generated deep ANNs are able to outperform conventional multi-layer perceptron style architectures with comparable parameter counts. Finally, the ANN-based parameterization is found to dramatically outperform the current parameterization. The success of this approach makes possible the future inclusion of much more sophisticated representations of aerosol optics in climate models that cannot be captured through simple expansion of the existing parameterization scheme.

Andrew Geiss et al.

Status: final response (author comments only)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
  • RC1: 'Comment on egusphere-2022-559', Anonymous Referee #1, 15 Jul 2022
  • RC2: 'Comment on egusphere-2022-559', Anonymous Referee #2, 10 Aug 2022
  • CEC1: 'Comment on egusphere-2022-559', Juan Antonio Añel, 23 Aug 2022
    • AC1: 'Reply on CEC1', Andrew Geiss, 06 Oct 2022

Andrew Geiss et al.

Data sets

Aerosol Optics ML Datasets Andrew Geiss https://doi.org/10.5281/zenodo.6762700

Model code and software

Aerosol Optics ML Code Andrew Geiss https://doi.org/10.5281/zenodo.6767169

Andrew Geiss et al.

Viewed

Total article views: 768 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
603 150 15 768 7 6
  • HTML: 603
  • PDF: 150
  • XML: 15
  • Total: 768
  • BibTeX: 7
  • EndNote: 6
Views and downloads (calculated since 08 Jul 2022)
Cumulative views and downloads (calculated since 08 Jul 2022)

Viewed (geographical distribution)

Total article views: 722 (including HTML, PDF, and XML) Thereof 722 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 
Latest update: 26 Mar 2023
Download
Short summary
Atmospheric aerosols play a critical role in Earth's climate, but it is too computationally expensive to directly model their interaction with radiation. This work develops a new neural network based parameterization of aerosol optical properties for use in the Energy Exascale Earth System model that is much more accurate than the current one, and introduces a unique model optimization method that involves randomly generating neural network architectures.