Machine Learning for Stochastic Parameterization: Generative Adversarial Networks in the Lorenz '96 Model

Simple item page
dc.contributor.authorGagne II, David John
dc.contributor.authorChristensen, Hannah M.
dc.contributor.authorSubramanian, Aneesh C.
dc.contributor.authorMonahan, Adam H.
dc.date.accessioned2020-06-23T00:06:04Z
dc.date.available2020-06-23T00:06:04Z
dc.date.copyright2020en_US
dc.date.issued2020
dc.description.abstractStochastic parameterizations account for uncertainty in the representation of unresolved subgrid processes by sampling from the distribution of possible subgrid forcings. Some existing stochastic parameterizations utilize data‐driven approaches to characterize uncertainty, but these approaches require significant structural assumptions that can limit their scalability. Machine learning models, including neural networks, are able to represent a wide range of distributions and build optimized mappings between a large number of inputs and subgrid forcings. Recent research on machine learning parameterizations has focused only on deterministic parameterizations. In this study, we develop a stochastic parameterization using the generative adversarial network (GAN) machine learning framework. The GAN stochastic parameterization is trained and evaluated on output from the Lorenz '96 model, which is a common baseline model for evaluating both parameterization and data assimilation techniques. We evaluate different ways of characterizing the input noise for the model and perform model runs with the GAN parameterization at weather and climate time scales. Some of the GAN configurations perform better than a baseline bespoke parameterization at both time scales, and the networks closely reproduce the spatiotemporal correlations and regimes of the Lorenz '96 system. We also find that, in general, those models which produce skillful forecasts are also associated with the best climate simulations.en_US
dc.description.reviewstatusRevieweden_US
dc.description.scholarlevelFacultyen_US
dc.description.sponsorshipThis research started in a working group supported by the Statistical and Applied Mathematical Sciences Institute (SAMSI). D. J. G. and H. M. C. were funded by National Center for Atmospheric Research Advanced Study Program Postdoctoral Fellowships and by the National Science Foundation through NCAR's Cooperative Agreement AGS‐1852977. H. M. C. was funded by Natural Environment Research Council grant number NE/P018238/1. A. H. M. acknowledges support from the Natural Sciences and Engineering Research Council of Canada (NSERC) and thanks SAMSI for hosting him in the autumn of 2017. The NCAR Cheyenne supercomputer was used to generate the model runs analyzed in this paper. Sue Ellen Haupt and Joseph Tribbia provided helpful reviews of the paper prior to submission. The source code and model configuration files necessary for replicating the results of this paper can be accessed at https://doi.org/10.5281/zenodo.3663121.en_US
dc.identifier.citationGagne, D. J., Christensen, H. M., Subramanian, A. C., & Monahan, A. H. (2020). Machine Learning for Stochastic Parameterization: Generative Adversarial Networks in the Lorenz ’96 Model. Journal of Advances in Modeling Earth Systems, 12(3), 1-20. https://doi.org/10.1029/2019MS001896.en_US
dc.identifier.urihttps://doi.org/10.1029/2019MS001896
dc.identifier.urihttp://hdl.handle.net/1828/11878
dc.language.isoenen_US
dc.publisherJournal of Advances in Modeling Earth Systemsen_US
dc.subjectmachine learning
dc.subjectstochastic parameterization
dc.subjectgenerative adversarial networks
dc.subjectlorenz
dc.subjectclimate
dc.subjectweather
dc.subject.departmentSchool of Earth and Ocean Sciences
dc.titleMachine Learning for Stochastic Parameterization: Generative Adversarial Networks in the Lorenz '96 Modelen_US
dc.typeArticleen_US

Files

Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
Gagne_David_JAdvModelEarthSyst_2020.pdf
Size:
3.84 MB
Format:
Adobe Portable Document Format
Description:
Download
License bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
license.txt
Size:
1.71 KB
Format:
Item-specific license agreed upon to submission
Description:
Download

Collections

Faculty and Staff Publications
 
University of Victoria Libraries

We acknowledge and respect the Lək̓ʷəŋən (Songhees and Esquimalt) Peoples on whose territory the university stands, and the Lək̓ʷəŋən and W̱SÁNEĆ Peoples whose historical relationships with the land continue to this day.