Dissertation · 2026 · Open access

Deep learning for meteorological downscaling: from deterministic super-resolution to probabilistic and multi-variable models

Fabio Merizzi

AMS Dottorato Institutional Doctoral Theses Repository (University of Bologna) · Published 7 August 2026 · DOI 10.48676/unibo/amsdottorato/12638

Open full textView at the publisherCite this paper

In brief

“Deep learning for meteorological downscaling: from deterministic super-resolution to probabilistic and multi-variable models” is a 2026 dissertation by Fabio Merizzi, published in AMS Dottorato Institutional Doctoral Theses Repository (University of Bologna). OpenAlex counts 0 citations as of 21 November 2026. It is open access: the full text is free to read at AMS Dottorato Institutional Doctoral Theses Repository (University of Bologna).

Citations
0
Authors
1

Abstract

High-resolution meteorological data is essential for understanding and predicting atmospheric processes, yet their production remains limited by the high computational cost of numerical downscaling models. In recent years, deep learning has emerged as a viable and highly efficient alternative, capable of producing high-quality, high-resolution fields at a fraction of the computational expense. This thesis explores the landscape of deep learning approaches for climate downscaling, analyzing their limitations and introducing new architectural designs and methodologies that enhance accuracy, represent uncertainty, and exploit inter-variable relationships. Alongside establishing a general methodological framework that maps existing neural downscaling approaches, the thesis comprises three original studies that collectively advance the state of data-driven downscaling. The first applies diffusion models to the super-resolution of near-surface wind speed from ERA5 to CERRA, benchmarking a range of architectures and introducing an ensemble diffusion formulation that achieves state-of-the-art accuracy while reproducing fine-scale variability. The second develops a theoretical link between the number of reverse diffusion steps and model variance, enabling calibrated uncertainty control and efficient ensemble generation. This framework is validated against existing ensemble reanalyses and extended to regions lacking ensemble information, positioning diffusion models as practical tools for uncertainty-aware meteorological modeling. The third study introduces a multi-variable, multi-task Vision Transformer that jointly downscales six meteorological variables, demonstrating that shared spatial representations improve both accuracy and efficiency compared to single-variable approaches. Together, these contributions provide new methodological insights into the use of generative and transformer-based architectures for climate and weather applications. They highlight how deep learning can evolve from merely reproducing existing models to discovering new representations of atmospheric variability and coherence.

Cite this paper

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

APA 7

Merizzi, F. (2026). Deep learning for meteorological downscaling: from deterministic super-resolution to probabilistic and multi-variable models. AMS Dottorato Institutional Doctoral Theses Repository (University of Bologna). https://doi.org/10.48676/unibo/amsdottorato/12638

MLA 9

Merizzi, Fabio. “Deep learning for meteorological downscaling: from deterministic super-resolution to probabilistic and multi-variable models.” AMS Dottorato Institutional Doctoral Theses Repository (University of Bologna), 2026. https://doi.org/10.48676/unibo/amsdottorato/12638.

Chicago

Merizzi, Fabio. 2026. “Deep learning for meteorological downscaling: from deterministic super-resolution to probabilistic and multi-variable models.” AMS Dottorato Institutional Doctoral Theses Repository (University of Bologna). https://doi.org/10.48676/unibo/amsdottorato/12638.

Harvard

Merizzi, F. (2026) ‘Deep learning for meteorological downscaling: from deterministic super-resolution to probabilistic and multi-variable models’, AMS Dottorato Institutional Doctoral Theses Repository (University of Bologna). doi:10.48676/unibo/amsdottorato/12638.

IEEE

F. Merizzi, “Deep learning for meteorological downscaling: from deterministic super-resolution to probabilistic and multi-variable models,” AMS Dottorato Institutional Doctoral Theses Repository (University of Bologna), 2026, doi: 10.48676/unibo/amsdottorato/12638.

BibTeX

@phdthesis{merizzi2026deep,
  title     = {{Deep learning for meteorological downscaling: from deterministic super-resolution to probabilistic and multi-variable models}},
  author    = {Merizzi, Fabio},
  year      = {2026},
  doi       = {10.48676/unibo/amsdottorato/12638},
  url       = {https://doi.org/10.48676/unibo/amsdottorato/12638}
}

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

  • climate downscaling
  • super-resolution
  • diffusion models
  • surface wind speed
  • ERA5 reanalysis
  • CERRA
  • uncertainty calibration
  • multi-task learning

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?

Merizzi, F. (2026). Deep learning for meteorological downscaling: from deterministic super-resolution to probabilistic and multi-variable models. AMS Dottorato Institutional Doctoral Theses Repository (University of Bologna). https://doi.org/10.48676/unibo/amsdottorato/12638

How many times has this paper been cited?

0 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 at AMS Dottorato Institutional Doctoral Theses Repository (University of Bologna): sign in and open the full text from this page.

Who wrote this paper?

Fabio Merizzi.

What is the DOI of this paper?

10.48676/unibo/amsdottorato/12638, permanently at https://doi.org/10.48676/unibo/amsdottorato/12638. 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.