Preprint · 2026 · Open access

Deep Learning-Based Meteorological Data Downscaling: A Comparative Study with Physics Informed CNN and a Component-Level Ablation Analysis

Johans Utama

EarthArXiv (California Digital Library) · Published 29 August 2026 · DOI 10.31223/x50n5v

Open PDFView at the publisherCite this paper

In brief

“Deep Learning-Based Meteorological Data Downscaling: A Comparative Study with Physics Informed CNN and a Component-Level Ablation Analysis” is a 2026 preprint by Johans Utama, published in EarthArXiv (California Digital Library). 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 EarthArXiv (California Digital Library).

Citations
0
References
8
Authors
1

Abstract

Meteorological statistical downscaling is a critical technique for deriving high-resolution climate information from coarse global reanalysis products. This experiment investigates the application of five representative deep learning architectures for downscaling ERA5 reanalysis 2m temperature fields from 1° to 0.25° spatial resolution (4x upscaling factor) over the China region, spanning the period 2010 to 2020. The five baseline models evaluated are Convolutional Neural Network (CNN), Generative Adversarial Network (GAN), Long Short-Term Memory network (LSTM), Vision Transformer (ViT-style), and Denoising Diffusion Probabilistic Model (DDPM). Building upon the analysis of baseline strengths and weaknesses, a sixth model is proposed, the Physics-Informed CNN (PICNN), which integrates multi-scale feature extraction, spatial attention mechanisms, and a composite physics-informed loss function incorporating mean squared error, Laplacian spatial smoothness regularization, and spatial energy conservation constraints. All models were trained on an NVIDIA GeForce RTX 5050 Laptop GPU with 8GB VRAM under computational constraints, with training epochs reduced from default values to accommodate hardware limitations. Experimental results on the held-out 2020 test set demonstrate that the proposed PICNN achieves the best performance among all models, with RMSE of 0.665 K, MAE of 0.435 K, PSNR of 43.43 dB, and SSIM of 0.976, representing improvements of 7.7% in RMSE and 8.7% in MAE over the strongest CNN baseline. To determine whether this improvement reflects the architectural design or simply the 84% increase in parameter count relative to the CNN baseline, a controlled ablation study was conducted comprising a capacity-matched plain CNN and three component-ablated PICNN variants, evaluated across three random seeds for the ablated variants. The results show that approximately 65.5% of PICNN's improvement over the CNN baseline is explained by parameter count alone. Among PICNN's three architectural components, spatial attention accounts for the majority of the remaining gain, the physics-informed loss contributes a small but consistent improvement, and the multi-scale feature extraction head shows no measurable benefit, with its ablated variant statistically indistinguishable from the full model. These findings refine the paper's central claim from a general endorsement of physics-informed design toward a specific, evidence-backed identification of which architectural choices are responsible for the observed gains. The Transformer model, despite having 139 million parameters, performed significantly worse than the CNN, highlighting the data inefficiency of attention-based architectures on moderate-sized meteorological datasets. The LSTM performed worst overall due to its inherent inability to preserve spatial structure.

Cite this paper

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

APA 7

Utama, J. (2026). Deep Learning-Based Meteorological Data Downscaling: A Comparative Study with Physics Informed CNN and a Component-Level Ablation Analysis. EarthArXiv (California Digital Library). https://doi.org/10.31223/x50n5v

MLA 9

Utama, Johans. “Deep Learning-Based Meteorological Data Downscaling: A Comparative Study with Physics Informed CNN and a Component-Level Ablation Analysis.” EarthArXiv (California Digital Library), 2026. https://doi.org/10.31223/x50n5v.

Chicago

Utama, Johans. 2026. “Deep Learning-Based Meteorological Data Downscaling: A Comparative Study with Physics Informed CNN and a Component-Level Ablation Analysis.” EarthArXiv (California Digital Library). https://doi.org/10.31223/x50n5v.

Harvard

Utama, J. (2026) ‘Deep Learning-Based Meteorological Data Downscaling: A Comparative Study with Physics Informed CNN and a Component-Level Ablation Analysis’, EarthArXiv (California Digital Library). doi:10.31223/x50n5v.

IEEE

J. Utama, “Deep Learning-Based Meteorological Data Downscaling: A Comparative Study with Physics Informed CNN and a Component-Level Ablation Analysis,” EarthArXiv (California Digital Library), 2026, doi: 10.31223/x50n5v.

BibTeX

@misc{utama2026deep,
  title     = {{Deep Learning-Based Meteorological Data Downscaling: A Comparative Study with Physics Informed CNN and a Component-Level Ablation Analysis}},
  author    = {Utama, Johans},
  howpublished = {EarthArXiv (California Digital Library)},
  year      = {2026},
  doi       = {10.31223/x50n5v},
  url       = {https://doi.org/10.31223/x50n5v}
}

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

  • physics-informed neural networks
  • ERA5 reanalysis
  • 2 m air temperature
  • spatial attention
  • physics-informed loss function
  • energy conservation constraints
  • ablation study
  • CNN vs transformer comparison

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?

Utama, J. (2026). Deep Learning-Based Meteorological Data Downscaling: A Comparative Study with Physics Informed CNN and a Component-Level Ablation Analysis. EarthArXiv (California Digital Library). https://doi.org/10.31223/x50n5v

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 EarthArXiv (California Digital Library): sign in and open the PDF from this page.

Who wrote this paper?

Johans Utama.

What is the DOI of this paper?

10.31223/x50n5v, permanently at https://doi.org/10.31223/x50n5v. 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.