Dual-Stream Downscaling and Flow-Constrained Correction of Meteorological Fields over Complex Terrain

Abstract High-resolution near-surface meteorological fields are essential, yet conventional numerical simulation is computationally expensive and data-driven methods often lack physical consistency. Here, a lightweight framework for meteorological-field downscaling and observation-driven correction is presented as a cost-effective alternative. In the downscaling stage, multivariate synoptic-scale information is fused with high-resolution terrain by a dual-stream generative adversarial network, and the fields are refined in a reconstruction-driven cascade. In the correction stage, a simplified UNet learns additive corrections supervised directly by station observations, and a flow-constrained regularizer enforces physical consistency. In independent tests, the downscaling stage reduces the 0.25° four-variable error by 54.0–73.5% relative to bicubic interpolation, and the full-field correction reduces the station MAE of T2m, V10, and U10 by 17.81%, 27.59%, and 21.06%, respectively. The full training-to-inference workflow can be completed within a few days on an ordinary personal computer, making it practical for operational fine-scale meteorological services.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-13
DOI
https://doi.org/10.21203/rs.3.rs-11134908/v1
Primary Topic
Meteorological Phenomena and Simulations
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Dual-Stream Downscaling and Flow-Constrained Correction of Meteorological Fields over Complex Terrain

Ao Shen, Wei Zhong, Hongrang He, Ying Yuan
5 links
Research Square
Meteorological Phenomena and Simulations
preprint

Dual-Stream Downscaling and Flow-Constrained Correction of Meteorological Fields over Complex Terrain

Ao Shen, Wei Zhong, Hongrang He, Ying Yuan
preprint en

Abstract

Abstract High-resolution near-surface meteorological fields are essential, yet conventional numerical simulation is computationally expensive and data-driven methods often lack physical consistency. Here, a lightweight framework for meteorological-field downscaling and observation-driven correction is presented as a cost-effective alternative. In the downscaling stage, multivariate synoptic-scale information is fused with high-resolution terrain by a dual-stream generative adversarial network, and the fields are refined in a reconstruction-driven cascade. In the correction stage, a simplified UNet learns additive corrections supervised directly by station observations, and a flow-constrained regularizer enforces physical consistency. In independent tests, the downscaling stage reduces the 0.25° four-variable error by 54.0–73.5% relative to bicubic interpolation, and the full-field correction reduces the station MAE of T2m, V10, and U10 by 17.81%, 27.59%, and 21.06%, respectively. The full training-to-inference workflow can be completed within a few days on an ordinary personal computer, making it practical for operational fine-scale meteorological services.

Research Square
National University of Defense Technology (CN)
Climate action
Meteorological Phenomena and Simulations
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon