Preprint · 2026 · Open access

Diffusion and flow matching for precipitation downscaling over India: a benchmark against statistical and deterministic methods

Shailesh Kumar Jha, Priyank J. Sharma, Anurag Mishra, Saksham Joshi, Vivek Vardhan Gupta

Research Square · Published 31 August 2026 · DOI 10.21203/rs.3.rs-10772032/v1

Open PDFView at the publisherCite this paper

In brief

“Diffusion and flow matching for precipitation downscaling over India: a benchmark against statistical and deterministic methods” is a 2026 preprint by Shailesh Kumar Jha and 4 co-authors, published in Research Square. OpenAlex counts 0 citations as of 21 November 2026. It is open access under a CC BY licence: the full text is free to read at Research Square.

Citations
0
References
83
Authors
5

Abstract

High-resolution precipitation data are essential for flood-risk assessment and water-resource management, yet coarse climate models cannot resolve local rainfall variability and extremes. Conventional statistical and deep-learning downscaling methods often estimate a conditional mean, favouring central tendency and suppressing extremes in the highly skewed precipitation distribution. Unlike GANs, prone to training instability and mode collapse, we use probability-path generative models, which sample from the full conditional distribution, preserving variability and rare events. We evaluate five methods, Constructed Analogues, Weather Typing, SRCNN, denoising diffusion probabilistic modelling (DDPM), and Flow Matching, for 10× daily precipitation downscaling from 1.0° to 0.1° over India in a perfect-model framework. Only the generative models reproduce the observed distribution and upper tail, recovering the generalized extreme-value shape parameter to within 0.004 and the 10-year return level to within 1%, whereas the non-generative methods underestimate the return level by 10–67%. Flow Matching matches or exceeds diffusion in spatial and spectral fidelity while achieving similar distributional accuracy with substantially fewer model evaluations. Both generative models also provide better probabilistic skill than deterministic methods, though their ensembles remain underdispersed. These results show that representing the full conditional distribution, not its expectation, is key to recovering precipitation extremes in high-ratio downscaling.

Cite this paper

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

APA 7

Jha, S. K., Sharma, P. J., Mishra, A., Joshi, S., & Gupta, V. V. (2026). Diffusion and flow matching for precipitation downscaling over India: a benchmark against statistical and deterministic methods. Research Square. https://doi.org/10.21203/rs.3.rs-10772032/v1

MLA 9

Jha, Shailesh Kumar, et al. “Diffusion and flow matching for precipitation downscaling over India: a benchmark against statistical and deterministic methods.” Research Square, 2026. https://doi.org/10.21203/rs.3.rs-10772032/v1.

Chicago

Jha, Shailesh Kumar, Priyank J. Sharma, Anurag Mishra, Saksham Joshi, and Vivek Vardhan Gupta. 2026. “Diffusion and flow matching for precipitation downscaling over India: a benchmark against statistical and deterministic methods.” Research Square. https://doi.org/10.21203/rs.3.rs-10772032/v1.

Harvard

Jha, S. K. et al. (2026) ‘Diffusion and flow matching for precipitation downscaling over India: a benchmark against statistical and deterministic methods’, Research Square. doi:10.21203/rs.3.rs-10772032/v1.

IEEE

S. K. Jha, P. J. Sharma, A. Mishra, S. Joshi, and V. V. Gupta, “Diffusion and flow matching for precipitation downscaling over India: a benchmark against statistical and deterministic methods,” Research Square, 2026, doi: 10.21203/rs.3.rs-10772032/v1.

BibTeX

@misc{jha2026diffusion,
  title     = {{Diffusion and flow matching for precipitation downscaling over India: a benchmark against statistical and deterministic methods}},
  author    = {Jha, Shailesh Kumar and Sharma, Priyank J. and Mishra, Anurag and Joshi, Saksham and Gupta, Vivek Vardhan},
  howpublished = {Research Square},
  year      = {2026},
  doi       = {10.21203/rs.3.rs-10772032/v1},
  url       = {https://doi.org/10.21203/rs.3.rs-10772032/v1}
}

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
  • flow matching
  • denoising diffusion probabilistic models
  • heavy rainfall
  • return levels
  • generalized extreme value distribution
  • high-resolution precipitation
  • India

Work with this paper

Most-cited papers in Environmental Science

All highly cited Environmental Science papers

Questions about this paper

How do I cite this paper in APA style?

Jha, S. K., Sharma, P. J., Mishra, A., Joshi, S., & Gupta, V. V. (2026). Diffusion and flow matching for precipitation downscaling over India: a benchmark against statistical and deterministic methods. Research Square. https://doi.org/10.21203/rs.3.rs-10772032/v1

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 Research Square: sign in and open the PDF from this page.

Who wrote this paper?

Shailesh Kumar Jha, Priyank J. Sharma, Anurag Mishra, Saksham Joshi, Vivek Vardhan Gupta.

What is the DOI of this paper?

10.21203/rs.3.rs-10772032/v1, permanently at https://doi.org/10.21203/rs.3.rs-10772032/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.