Generative and deterministic deep learning models comparison for fine-scale precipitation retrievals from infrared brightness temperature
Abstract
Accurate precipitation estimation at fine spatial scales is critical for hydrology, agriculture, and climate studies. Infrared brightness temperatures from geostationary satellites offer excellent temporal coverage over continental-scale domains. However, because these measurements primarily characterize cloud-top properties rather than precipitation processes near the surface, their correlation with rainfall intensity remains limited, making quantitative precipitation estimation challenging. In this study, we conduct a systematic inter-comparison of state-of-the-art deep learning models for high-resolution precipitation retrieval from Meteosat Second Generation infrared brightness temperatures over metropolitan France. These models include deterministic U-Nets, transformer-based architectures, conditional GANs, and diffusion models. We construct a curated dataset spanning 2008--2023, combining M{é}t{é}o-France radar mosaics as reference with multi-channel infrared observations, and design preprocessing and sampling strategies to address the heavy-tailed, intermittent nature of rainfall. Our results show that deterministic models provide robust mean estimates and excel in pixel-wise accuracy, but systematically underestimate extreme precipitation. In contrast, generative models better capture the full precipitation distribution, including rare and heavy rainfall events, producing more realistic spatial structures at the cost of reduced pixel-wise fidelity. These results highlight a trade-off between pixel-wise accuracy and precipitation variability, showing that generative approaches are advantageous for extreme-event detection and probabilistic applications. This work establishes a reproducible framework for evaluating infrared- based precipitation retrieval methods and provides guidance for designing models that balance precision, variability, and extreme-event representation.
Related
- Adaptive observation weighting in TCKF1D-Var for ground-based multi-sensor thermodynamic retrievals prior to nocturnal heavy precipitation over China
- High-resolution Calibrated Probabilistic Hourly Precipitation from a Deterministic Forecast
- Development of a nonparametric spatial standardized precipitation–temperature index for regional drought characterization
- Polar front jet regulates tropical sea surface temperature impacts on South China extreme precipitation
- Climate models lack decadal variability in precipitation because of a weak atmospheric response to sea surface temperature anomalies
BibTeX
@misc{meignin2026generative,
title = {{Generative and deterministic deep learning models comparison for fine-scale precipitation retrievals from infrared brightness temperature}},
author = {Matthieu Meignin and Cécile Mallet and Nicolas Viltard},
howpublished = {HAL},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.09859},
eprint = {2610.09859},
archivePrefix = {arXiv},
url = {https://hal.science/hal-05649885v2},
}
RIS
TY - UNPB
TI - Generative and deterministic deep learning models comparison for fine-scale precipitation retrievals from infrared brightness temperature
AU - Matthieu Meignin
AU - Cécile Mallet
AU - Nicolas Viltard
PY - 2026
DA - 2026/10/05
PB - HAL
DO - 10.48550/arxiv.2610.09859
UR - https://hal.science/hal-05649885v2
AB - Accurate precipitation estimation at fine spatial scales is critical for hydrology, agriculture, and climate studies. Infrared brightness temperatures from geostationary satellites offer excellent temporal coverage over continental-scale domains. However, because these measurements primarily characterize cloud-top properties rather than precipitation processes near the surface, their correlation with rainfall intensity remains limited, making quantitative precipitation estimation challenging. In this study, we conduct a systematic inter-comparison of state-of-the-art deep learning models for high-resolution precipitation retrieval from Meteosat Second Generation infrared brightness temperatures over metropolitan France. These models include deterministic U-Nets, transformer-based architectures, conditional GANs, and diffusion models. We construct a curated dataset spanning 2008--2023, combining M{é}t{é}o-France radar mosaics as reference with multi-channel infrared observations, and design preprocessing and sampling strategies to address the heavy-tailed, intermittent nature of rainfall. Our results show that deterministic models provide robust mean estimates and excel in pixel-wise accuracy, but systematically underestimate extreme precipitation. In contrast, generative models better capture the full precipitation distribution, including rare and heavy rainfall events, producing more realistic spatial structures at the cost of reduced pixel-wise fidelity. These results highlight a trade-off between pixel-wise accuracy and precipitation variability, showing that generative approaches are advantageous for extreme-event detection and probabilistic applications. This work establishes a reproducible framework for evaluating infrared- based precipitation retrieval methods and provides guidance for designing models that balance precision, variability, and extreme-event representation.
ER -