Journal article · 2026 · Open access
How far can we downscale? Resolution limits and physical interpretability of diffusion models for African precipitation
Machine Learning Earth, 2(1), 015018 · Published 1 June 2026 · DOI 10.1088/3049-4753/ae7114
In brief
“How far can we downscale? Resolution limits and physical interpretability of diffusion models for African precipitation” is a 2026 journal article by Jangho Lee and Sara Shamekh, published in Machine Learning Earth. OpenAlex counts 3 citations as of 21 November 2026, which puts it in the top 4% of works published in 2026 by citations. It is open access under a CC BY licence: the full text is free to read at Machine Learning Earth.
- Citations
- 3
- Most cited in
- 2026 (3)
- Field-weighted impact
- 4.24
- References
- 51
- Authors
- 2
Abstract
Accurate high-resolution precipitation data is critical for climate risk assessment in Sub-Saharan Africa, yet global climate models operate at coarse resolutions, failing to capture localized extremes. While deterministic deep learning models offer efficient downscaling, our evaluation shows that minimizing pointwise error inherently penalizes high-frequency features, artificially smoothing precipitation fields and suppressing extremes regardless of architectural complexity. Here, we propose a conditional denoising diffusion model for precipitation downscaling, evaluating its structural realism, physical interpretability, and resolution limits against deterministic baselines. The diffusion model successfully recovers sub-grid variability, preserving near-perfect spatial variance and heavy-tailed extreme event distributions where deterministic approaches fail. Through perturbation-based feature importance analysis, we demonstrate the generative framework acts as a context-aware emulator of atmospheric physics. It dynamically shifts predictor reliance, prioritizing low-level moisture convergence in rugged terrain and dry seasons, while favoring mid-tropospheric dynamic forcing during monsoons. Furthermore, by progressively coarsening inputs from 0.25° to 1.0°, we interrogate generative downscaling limits. As spatial constraints weaken at 1.0°, the model generates unconstrained stochastic false alarms, and the resulting ensemble smoothing directly quantifies predictive spatial uncertainty under severe information loss. A moderate 0.5° input emerges as an optimal balance, providing sufficient structural anchors to maintain both deterministic accuracy and extreme event preservation. Ultimately, this work provides a physically interpretable, resolution-aware framework for generative downscaling, offering a robust strategy for translating coarse climate projections into localized risk assessments.
Cite this paper
Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.
APA 7
Lee, J., & Shamekh, S. (2026). How far can we downscale? Resolution limits and physical interpretability of diffusion models for African precipitation. Machine Learning Earth, 2(1), 015018. https://doi.org/10.1088/3049-4753/ae7114
MLA 9
Lee, Jangho, and Sara Shamekh. “How far can we downscale? Resolution limits and physical interpretability of diffusion models for African precipitation.” Machine Learning Earth, vol. 2, no. 1, 2026, p. 015018. https://doi.org/10.1088/3049-4753/ae7114.
Chicago
Lee, Jangho and Sara Shamekh. 2026. “How far can we downscale? Resolution limits and physical interpretability of diffusion models for African precipitation.” Machine Learning Earth 2 (1): 015018. https://doi.org/10.1088/3049-4753/ae7114.
Harvard
Lee, J. and Shamekh, S. (2026) ‘How far can we downscale? Resolution limits and physical interpretability of diffusion models for African precipitation’, Machine Learning Earth, 2(1), p. 015018. doi:10.1088/3049-4753/ae7114.
IEEE
J. Lee and S. Shamekh, “How far can we downscale? Resolution limits and physical interpretability of diffusion models for African precipitation,” Machine Learning Earth, vol. 2, no. 1, p. 015018, 2026, doi: 10.1088/3049-4753/ae7114.
BibTeX
@article{lee2026downscale,
title = {{How far can we downscale? Resolution limits and physical interpretability of diffusion models for African precipitation}},
author = {Lee, Jangho and Shamekh, Sara},
journal = {Machine Learning Earth},
year = {2026},
volume = {2},
number = {1},
pages = {015018},
doi = {10.1088/3049-4753/ae7114},
url = {https://doi.org/10.1088/3049-4753/ae7114}
}
Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)
Topic and keywords
Physical Sciences › Environmental Science › Global and Planetary Change › Climate variability and models
- precipitation downscaling
- denoising diffusion probabilistic models
- resolution limit
- heavy rainfall
- sub-Saharan Africa
- physical interpretability
- feature importance analysis
- spatial uncertainty
Work with this paper
Most-cited papers in Environmental Science
- Biodiversity hotspots for conservation prioritiesNorman Myers, Russell A. Mittermeier, Cristina Goettsch Mittermeier and 2 more · Nature · 2000 · 32,720 citations
- The Tragedy of the CommonsGarrett Hardin · Science · 1968 · 26,194 citations
- River flow forecasting through conceptual models part I — A discussion of principlesJ.E. Nash, John V. Sutcliffe · Journal of Hydrology · 1970 · 24,625 citations
- The value of the world's ecosystem services and natural capitalRobert Costanza, Ralph C. d’Arge, R.S. de Groot and 10 more · Nature · 1997 · 23,508 citations
- Climate change 2007: the physical science basisSusan Solomon, Dahe Qin, Martin Manning and 5 more · 2007 · 23,383 citations
Questions about this paper
How do I cite this paper in APA style?
Lee, J., & Shamekh, S. (2026). How far can we downscale? Resolution limits and physical interpretability of diffusion models for African precipitation. Machine Learning Earth, 2(1), 015018. https://doi.org/10.1088/3049-4753/ae7114
How many times has this paper been cited?
3 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 Machine Learning Earth: sign in and open the PDF from this page.
Who wrote this paper?
Jangho Lee, Sara Shamekh.
What is the DOI of this paper?
10.1088/3049-4753/ae7114, permanently at https://doi.org/10.1088/3049-4753/ae7114. 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.