Preprint · 2025 · Open access

Generating ensembles for convective-scale precipitation downscaling using generative machine learning

P. Jyoteeshkumar Reddy, Richard J. Matear

Published 23 May 2025 · DOI 10.22541/essoar.174802957.72376154/v1

Open PDFView at the publisherCite this paper

In brief

“Generating ensembles for convective-scale precipitation downscaling using generative machine learning” is a 2025 preprint by P. Jyoteeshkumar Reddy and Richard J. Matear. OpenAlex counts 2 citations as of 21 November 2026. It is open access: the full text is free to read.

Citations
2
Most cited in
2026 (2)
References
32
Authors
2

Abstract

High-resolution precipitation information is essential for understanding localised extremes. However, global reanalysis products such as ERA5 cannot resolve fine-scale features due to their coarse (0.25°) spatial resolution. While convection-scale regional reanalyses, such as BARRA-C2 (4.4 km resolution), offer improved realism, they are computationally expensive and impractical for generating large ensembles, which are essential for quantifying uncertainty. To address this limitation, we use a generative machine learning approach based on the Relativistic R1-R2 Regularized GAN (R3GAN) to use ERA5 atmospheric variables to predict convective-scale precipitation fields, using BARRA-C2 as the training dataset. Using Tasmania as a test region, we evaluate the performance of the R3GAN model against ERA5, a baseline UNET model trained with mean squared error loss, and the BARRA-C2 target. The R3GAN model successfully captures the distribution of precipitation and spatial power spectral density, effectively reproducing extreme rainfall events missed by ERA5 and UNET. With R3GAN model we explore ensemble predictions by utilising a novel framework that systematically varies three sources of generative variability: noise seed, input noise variance, and model epoch. We show that an ensemble considering the three terms produces the best ensemble prediction. Our results highlight the potential of R3GAN model for ensemble downscaled prediction of high-resolution precipitation.

Cite this paper

Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.

APA 7

Reddy, P. J., & Matear, R. J. (2025). Generating ensembles for convective-scale precipitation downscaling using generative machine learning. https://doi.org/10.22541/essoar.174802957.72376154/v1

MLA 9

Reddy, P. Jyoteeshkumar, and Richard J. Matear. “Generating ensembles for convective-scale precipitation downscaling using generative machine learning.” 2025. https://doi.org/10.22541/essoar.174802957.72376154/v1.

Chicago

Reddy, P. Jyoteeshkumar and Richard J. Matear. 2025. “Generating ensembles for convective-scale precipitation downscaling using generative machine learning.” https://doi.org/10.22541/essoar.174802957.72376154/v1.

Harvard

Reddy, P. J. and Matear, R. J. (2025) ‘Generating ensembles for convective-scale precipitation downscaling using generative machine learning’. doi:10.22541/essoar.174802957.72376154/v1.

IEEE

P. J. Reddy and R. J. Matear, “Generating ensembles for convective-scale precipitation downscaling using generative machine learning,” 2025, doi: 10.22541/essoar.174802957.72376154/v1.

BibTeX

@misc{reddy2025generating,
  title     = {{Generating ensembles for convective-scale precipitation downscaling using generative machine learning}},
  author    = {Reddy, P. Jyoteeshkumar and Matear, Richard J.},
  year      = {2025},
  doi       = {10.22541/essoar.174802957.72376154/v1},
  url       = {https://doi.org/10.22541/essoar.174802957.72376154/v1}
}

Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)

Topic and keywords

Physical Sciences › Earth and Planetary Sciences › Atmospheric Science › Meteorological Phenomena and Simulations

  • precipitation downscaling
  • generative adversarial networks
  • ensemble forecasting
  • uncertainty quantification
  • ERA5 reanalysis
  • heavy rainfall
  • Tasmania

Work with this paper

Most-cited papers in Earth and Planetary Sciences

All highly cited Earth and Planetary Sciences papers

Questions about this paper

How do I cite this paper in APA style?

Reddy, P. J., & Matear, R. J. (2025). Generating ensembles for convective-scale precipitation downscaling using generative machine learning. https://doi.org/10.22541/essoar.174802957.72376154/v1

How many times has this paper been cited?

2 times, according to OpenAlex on 21 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.

Is this paper free to read?

Yes. It is open access: sign in and open the PDF from this page.

Who wrote this paper?

P. Jyoteeshkumar Reddy, Richard J. Matear.

What is the DOI of this paper?

10.22541/essoar.174802957.72376154/v1, permanently at https://doi.org/10.22541/essoar.174802957.72376154/v1. Sign in and open it from this page.

Bibliographic data from OpenAlex (CC0), last checked 21 November 2026. ResearchGuru does not host this paper; once you are signed in, the links open it at the publisher or the repository that does.