Journal article · 2026 · Open access

Robust deep learning-based downscaling of mean and extreme precipitation over the Indian subcontinent

Midhun Murukesh, Pankaj Kumar

Environmental Research Climate, 5(2), 025029 · Published 19 May 2026 · DOI 10.1088/2752-5295/ae6885

Open full textCite this paper

In brief

“Robust deep learning-based downscaling of mean and extreme precipitation over the Indian subcontinent” is a 2026 journal article by Midhun Murukesh and Pankaj Kumar, published in Environmental Research Climate. OpenAlex counts 1 citations as of 21 November 2026. It is open access under a CC BY licence: the full text is free to read at Environmental Research Climate.

Citations
1
Most cited in
2026 (1)
Field-weighted impact
1.41
References
57
Authors
2

Abstract

Downscaling bridges the gap between coarse-resolution global climate models and local impacts by improving the representation of regional variability and localized extremes. However, precipitation downscaling remains challenging due to its stochastic nature and highly skewed distribution. In this study, we investigate the generalization capability of a deep learning approach, the Super-Resolution Deep Residual Network (SRDRN), for gridded precipitation downscaling over the Indian subcontinent, considering four distinct regions: Central India (CI), Southern Peninsular India (SP), Northwest India (NW), and Northeast India (NE), with NE chosen as the transfer domain. The SRDRN is trained using samples from CI, SP, and NW with three loss functions: mean squared error (MSE), mean absolute error (MAE), and weighted MAE (WMAE), resulting in three model variants. All variants consistently outperform the baseline bias correction and spatial disaggregation (BCSD) method across standard pixel-wise metrics, including Kling–Gupta efficiency, root MSE (RMSE), and percent bias (PBIAS), and effectively capture localized extreme precipitation, preserving key indices such as the 99th percentile magnitude (R99), monthly maximum 1-day precipitation (Rx1Day), consecutive wet days (CWDs), consecutive dry days (CDDs), and the annual count of very heavy precipitation days (R20mm). Among the variants, SRDRN-WMAE shows the highest skill in representing extreme precipitation intensity, while SRDRN-MAE provides the most balanced performance across both intensity-based and persistence-based indices. In a zero-shot transfer setting, all SRDRN variants applied over NE without any domain-specific training consistently outperform BCSD, especially in capturing extreme precipitation, despite BCSD being explicitly calibrated for the NE region. The results indicate that SRDRN models learn transferable representations of precipitation structure that generalize effectively beyond their training domains. Overall, the results demonstrate that SRDRN is a robust, generalizable, and computationally efficient framework for high-resolution precipitation downscaling, with strong potential for applications in data-sparse and climatologically complex regions.

Cite this paper

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

APA 7

Murukesh, M., & Kumar, P. (2026). Robust deep learning-based downscaling of mean and extreme precipitation over the Indian subcontinent. Environmental Research Climate, 5(2), 025029. https://doi.org/10.1088/2752-5295/ae6885

MLA 9

Murukesh, Midhun, and Pankaj Kumar. “Robust deep learning-based downscaling of mean and extreme precipitation over the Indian subcontinent.” Environmental Research Climate, vol. 5, no. 2, 2026, p. 025029. https://doi.org/10.1088/2752-5295/ae6885.

Chicago

Murukesh, Midhun and Pankaj Kumar. 2026. “Robust deep learning-based downscaling of mean and extreme precipitation over the Indian subcontinent.” Environmental Research Climate 5 (2): 025029. https://doi.org/10.1088/2752-5295/ae6885.

Harvard

Murukesh, M. and Kumar, P. (2026) ‘Robust deep learning-based downscaling of mean and extreme precipitation over the Indian subcontinent’, Environmental Research Climate, 5(2), p. 025029. doi:10.1088/2752-5295/ae6885.

IEEE

M. Murukesh and P. Kumar, “Robust deep learning-based downscaling of mean and extreme precipitation over the Indian subcontinent,” Environmental Research Climate, vol. 5, no. 2, p. 025029, 2026, doi: 10.1088/2752-5295/ae6885.

BibTeX

@article{murukesh2026robust,
  title     = {{Robust deep learning-based downscaling of mean and extreme precipitation over the Indian subcontinent}},
  author    = {Murukesh, Midhun and Kumar, Pankaj},
  journal   = {Environmental Research Climate},
  year      = {2026},
  volume    = {5},
  number    = {2},
  pages     = {025029},
  doi       = {10.1088/2752-5295/ae6885},
  url       = {https://doi.org/10.1088/2752-5295/ae6885}
}

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
  • heavy rainfall
  • gridded precipitation
  • deep learning
  • zero-shot transfer learning
  • Indian subcontinent
  • regional climate variability
  • loss function comparison

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?

Murukesh, M., & Kumar, P. (2026). Robust deep learning-based downscaling of mean and extreme precipitation over the Indian subcontinent. Environmental Research Climate, 5(2), 025029. https://doi.org/10.1088/2752-5295/ae6885

How many times has this paper been cited?

1 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 Environmental Research Climate: sign in and open the full text from this page.

Who wrote this paper?

Midhun Murukesh, Pankaj Kumar.

What is the DOI of this paper?

10.1088/2752-5295/ae6885, permanently at https://doi.org/10.1088/2752-5295/ae6885. 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.