Efficient Analog-Initialized Latent Transport for Spatially Coherent Probabilistic Downscaling from Global to Kilometer Scales
Abstract
Probabilistic downscaling must represent kilometer-scale structure left unresolved by a deterministic regional prediction. We introduce an analog-initialized latent transport method that uses historical regional residuals as a meteorologically informed empirical source. A frozen graph neural network predicts the deterministic state from the Integrated Forecasting System (IFS). For each input, residuals from 20 training dates with similar IFS conditions initialize the ensemble. A conditional flow-matching model transports these fields in a carefully designed 65-factor, 100,000-node latent space, and a grid decoder reconstructs 81 atmospheric channels including precipitation at 2.5km spacing. On a season- and cycle-balanced 2025 test, ensemble CRPS improves over deterministic mean absolute error for all 81 variables. Matched analogs improve point and spatial errors over Gaussian initialization and reduce spectral error compared with covariance matching. Comparisons with released CorrDiff, a local CorrDiff-mini model and full grid EDM provide context on the 2023 validation sample. Total training of the downscaling model takes approximately 12h on one NVIDIA H200, and inference takes about 12s per 81-field member. The frozen networks also generate fields over France when supplied with local static data and a high-resolution IFS-derived analog bank. The method combines empirical initialization with whole-domain latent transport for computationally practical regional ensembles.
Related
- Earth observation embeddings are effective sub-grid descriptors for probabilistic weather downscaling
- Less is more: error-distance scaling relation for data-efficient kilometer-scale downscaling of extreme heat
- Using GIS, remote-sensing and linear modeling to prioritize spatially explicit action against erosion and sediment transport within Lake Victoria watershed in Kenya
- Efficient, Geometry-only Prediction of Advection-dominated Particle Transport through Saturated Soils
- Deep learning-based downscaling of PM2.5, NO2 and O3 over Europe across emission-reduction scenarios
BibTeX
@misc{miralles2026efficient,
title = {{Efficient Analog-Initialized Latent Transport for Spatially Coherent Probabilistic Downscaling from Global to Kilometer Scales}},
author = {Ophélia Miralles},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.03736},
eprint = {2610.03736},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.03736},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Efficient Analog-Initialized Latent Transport for Spatially Coherent Probabilistic Downscaling from Global to Kilometer Scales
AU - Ophélia Miralles
PY - 2026
DA - 2026/10/01
PB - arXiv
DO - 10.48550/arxiv.2610.03736
UR - https://arxiv.org/abs/2610.03736
AB - Probabilistic downscaling must represent kilometer-scale structure left unresolved by a deterministic regional prediction. We introduce an analog-initialized latent transport method that uses historical regional residuals as a meteorologically informed empirical source. A frozen graph neural network predicts the deterministic state from the Integrated Forecasting System (IFS). For each input, residuals from 20 training dates with similar IFS conditions initialize the ensemble. A conditional flow-matching model transports these fields in a carefully designed 65-factor, 100,000-node latent space, and a grid decoder reconstructs 81 atmospheric channels including precipitation at 2.5km spacing. On a season- and cycle-balanced 2025 test, ensemble CRPS improves over deterministic mean absolute error for all 81 variables. Matched analogs improve point and spatial errors over Gaussian initialization and reduce spectral error compared with covariance matching. Comparisons with released CorrDiff, a local CorrDiff-mini model and full grid EDM provide context on the 2023 validation sample. Total training of the downscaling model takes approximately 12h on one NVIDIA H200, and inference takes about 12s per 81-field member. The frozen networks also generate fields over France when supplied with local static data and a high-resolution IFS-derived analog bank. The method combines empirical initialization with whole-domain latent transport for computationally practical regional ensembles.
ER -