Skillful Kilometer-Scale Regional Weather Forecasting via Global and Regional Coupling

Weiqi Chen; Wenwei Wang; Qilong Yuan; Lefei Shen; Bingqing Peng; Jiawei Chen; Bo Wu; Liang Sun

cs.LGMar 30, 2026

Skillful Kilometer-Scale Regional Weather Forecasting via Global and Regional Coupling

Authors: Weiqi Chen, Wenwei Wang, Qilong Yuan, Lefei Shen, Bingqing Peng, Jiawei Chen, Bo Wu, Liang Sun

Organizations: DAMO Academy, Alibaba Group Hangzhou, China · Northwest Polytechnical University Xi’an, China · Zhejiang University Hangzhou, China · Institute of Atmospheric Physics, Chinese Academy of Sciences Beijing, China

Abstract

Data-driven weather models have advanced global medium-range forecasting, yet high-resolution regional prediction remains challenging due to unresolved multiscale interactions between large-scale dynamics and small-scale processes such as terrain-induced circulations and coastal effects. This paper presents a global-regional coupling framework for kilometer-scale regional weather forecasting that synergistically couples a pretrained Transformer-based global model with a high-resolution regional network via a novel bidirectional coupling module, ScaleMixer. ScaleMixer dynamically identifies meteorologically critical regions through adaptive key-position sampling and enables cross-scale feature interaction through dedicated attention mechanisms. The framework produces forecasts at $0.05^\circ$ ( $\sim 5 \mathrm{km}$ ) and 1-hour resolution over China, significantly outperforming operational NWP and AI baselines on both gridded reanalysis data and real-time weather station observations. It exhibits exceptional skill in capturing fine-grained phenomena such as orographic wind patterns and Foehn warming, demonstrating effective global-scale coherence with high-resolution fidelity. The code is available at https://anonymous.4open.science/r/ScaleMixer-6B66.

Figures & tables

Figure 1. Left: The Architecture of Global-Regional Weather Forecasting Model: Synoptic-scale context ( $\mathcal{M}_{\mathrm{global}}$ ) drives mesoscale regional refinement ( $\mathcal{M}_{\mathrm{regional}}$ ) via ScaleMixer, ensuring cross-scale coupling and consistency. Right: ScaleMixer Module: Bidirectional Cross-Scale Coupling via Key Position Identification and encoding. Key components include (1) key position identification, (2) coupling regional dynamics with global context via global-to-position and position-to-regional attention, and (3) global token adaptation incorporating regional features.

Variable	Latitude-weighted RMSE								Latitude-weighted ACC
Variable	IFS-HRES	Baguan	$\mathcal{M}_{\mathrm{global}}$	$\mathcal{M}_{\mathrm{regional}}$	OneForecast	LAM	ScaleMixer	$\Delta\mathrm{RMSE}$	IFS-HRES	$\mathcal{M}_{\mathrm{global}}$	$\mathcal{M}_{\mathrm{regional}}$	OneForecast	LAM	ScaleMixer	$\Delta\mathrm{ACC}$
T2M	1.815	1.928	1.991	2.452	1.571	1.742	1.382	$\downarrow$ 23.86%	0.862	0.881	0.845	0.897	0.901	0.921	$\uparrow$ 6.84%
U10	1.934	1.956	1.967	2.631	2.251	1.889	1.644	$\downarrow$ 14.99%	0.721	0.753	0.716	0.744	0.742	0.793	$\uparrow$ 9.99%
V10	1.928	1.937	1.970	2.886	2.316	1.905	1.617	$\downarrow$ 16.13%	0.723	0.751	0.713	0.732	0.737	0.785	$\uparrow$ 8.57%
Q	0.807	0.611	0.811	1.243	0.714	0.676	0.559	$\downarrow$ 30.73%	0.774	0.768	0.771	0.751	0.781	0.812	$\uparrow$ 4.91%
P	13.27	22.97	10.27	4.241	2.545	2.14	1.874	$\downarrow$ 85.88%	0.887	0.902	0.894	0.899	0.912	0.920	$\uparrow$ 3.72%

Table 1. Average RMSE and ACC across

\Delta t=1\sim 48

hours lead time regional weather hindcast at 0.05 ∘ resolution. The best results are bolded .

Panel of Figure 2 — Figure 2. ScaleMixer demonstrates superior deterministic forecasting skill compared to IFS-HRES at 0.05° resolution. Seven surface variables (T2M, U10, V10, Q, P, TCC, and SSRD) are evaluated using latitude-weighted RMSE (lower values indicate superior performance). (a) Hindcast results show ScaleMixer outperforms IFS-HRES across all variables during 2024/10–2024/12. (b) Operational forecasts confirm ScaleMixer maintains superiority performance (2025/01–2025/04).

Figure 3. Left: Temporal evolution of 10m wind speed predictions initialized at 2024/10/30 12 UTC over the Hengduan Mountains (25.0–35.0°N, 95.0–105.0°E), China. Black arrows represent wind flow fields. ScaleMixer resolves enhanced resolution of orographic wind heterogeneity (peaking >10 m/s at crests and <2 m/s in valleys). Right: Corresponding temperature fields. Foehn effects are illustrated in the picture, characterized by 4–8°C leeward warming relative to windward slopes through adiabatic compression processes. ScaleMixer captures fine-grained temperature gradients, contrasting with IFS-HRES exhibiting spatial smoothing forecasts.

Figure 4. Station distrbution map. The station observation dataset contains 2216 weather stations across China, which record hourly observations of meteorological variables such as temperature, air pressure, and wind speed.

Variable	IFS-HRES	Baguan	ScaleMixer	$\Delta\mathrm{RMSE}$
T2M	3.158	3.108	1.822	$\downarrow$ 41.4%
D2M	2.812	2.655	1.789	$\downarrow$ 36.4%
WS10	1.401	1.402	1.329	$\downarrow$ 5.10%

Table 2. Average RMSE across

\Delta t=1\sim 48

hours lead time regional weather operational forecasting against weather station observations.

Model Variant	Configuration Details	T2M RMSE	U10 RMSE
ScaleMixer	Adaptive Sampling + Bidirectional	1.382	1.644
Variant A	Random Sampling + Bidirectional	1.605 (+16.1%)	1.882 (+14.5%)
Variant B	Fixed Uniform Grid + Bidirectional	1.512 (+9.4%)	1.795 (+9.2%)
Variant C	Adaptive Sampling + Unidirectional	1.468 (+6.2%)	1.721 (+4.7%)
Variant D	No Interaction (Standalone)	1.991 (+44.1%)	1.967 (+19.6%)

Table 3. Ablation study of ScaleMixer components on 48-hour forecast performance.

Metric	Regional Encoder Layers ( $k$ )
Metric	$k=2$	${k=4}$	$k=8$
T2M RMSE	1.485	1.382	1.379
U10 RMSE	1.752	1.644	1.641
Inference Time	22ms	28ms	51ms

Table 4. Sensitivity analysis of Regional Encoder Layers (

k

Appendix figures & tables16 assets

Supplementary material from the paper’s appendix.

Appendix

Module	Hyperparameter	Description	Value
Global Model $\mathcal{M}_{\mathrm{global}}$	$P$	Patch size of global tokens	$6$
	$d$	hidden dimension	$1536$
	$M$	Number of Transformer encoder layers of in the global model	$24$
	Heads	Number of attention heads	$8$
	MLP ratio	Expansion factor for MLP	$4.$
	Depth of prediction head	Number of deconvolution layers of the final prediction head	$2$

Appendix

Table 5. Default hyperparameters of the framework.

Type	Variable Name	Abbrev.	Description	Pressure Levels
Static Variable	Land-sea mask	$lsm$	Binary mask distinguishing land (1) from sea (0)	N/A
	Orography	$oro$	Height of Earth’s surface	N/A
	Latitude	$lat$	Latitude of each grid point	N/A
Surface Variable	2 metre temperature	$t2m$	Temperature measured 2 meters above the surface	Single level
	10 metre U wind component	$u10$	East-west wind speed at 10 meters above the surface	Single level
	10 metre V wind component	$v10$	North-south wind speed at 10 meters above the surface	Single level

Appendix

Table 6. Summary of ECMWF variables utilized in the ERA5 and operational analysis dataset with

0.25

° resolution. The variables

lsm

and

oro

are constant and invariant with time.

Type	Variable Name	Abbrev.	Description
Surface Variable	2 metre temperature	$T$	Temperature measured 2 meters above the surface
	10 metre U wind component	$U$	East-west wind speed at 10 meters above the surface
	10 metre V wind component	$V$	North-south wind speed at 10 meters above the surface
	Surface specific humidity	$Q$	Mixing ratio of water vapor to total air mass at 2 meters above the surface
	Surface pressure	$P$	Pressure of the atmosphere on the surface of land, sea and in-land water
	Total cloud cover	$TCC$	Cloud occurring at different model levels through the atmosphere

Appendix

Table 7. Summary of variables utilized in CLDAS with

0.05

° resolution.

Figure 5. 2 metre temperature forecasts over China

Figure 6. 2 metre temperature forecasts over a subregion of latitudes in $[30,40]$ and longitudes in $[95,115]$

Figure 7. surface pressure forecasts over China

Figure 8. surface pressure forecasts over a subregion of latitudes in $[30,40]$ and longitudes in $[95,115]$

Figure 9. 10 metre Wind speed U component forecasts over China

Figure 10. 10 metre wind speed U component forecasts over a subregion of latitudes in $[16.3,26]$ and longitudes in $[115,125]$

Figure 11. 10 metre Wind speed V component forecasts over China

Figure 12. 10 metre wind speed V component forecasts over a subregion of latitudes in $[16.3,26]$ and longitudes in $[115,125]$

Figure 13. surface solar radiation downwards forecasts over China

Figure 14. surface solar radiation downwards forecasts over a subregion of latitudes in $[35,50]$ and longitudes in $[115,135]$

Figure 15. total cloud cover forecasts over China

Figure 16. total cloud cover forecasts over a subregion of latitudes in $[35,50]$ and longitudes in $[115,135]$

Figure 17. RMSE and ACC for some key variables of various AI-based weather models on Weatherbench2

Weather Forecasting

Weather forecasting predicts future atmospheric conditions, from local storms to global patterns, using physical models, observations, and machine learning.

See all 176 papers

Learning Kilometer-Scale Weather Prediction with Global-Regional Alignment
Oct 8, 2026Guowen Li, Yang Liu, Yujie Wang +5Weather Forecasting Representation Alignment
Tsinghua University · The Chinese University of Hong Kong · Sun Yat-Sen University +1
StretchCast: Global-Regional AI Weather Forecasting on Stretched Cubed-Sphere Mesh
Mar 28, 2026Jin FengWeather Forecasting Time Series Forecasting
Institute of Urban Meteorology, China Meteorological Administration (IUM, CMA), Beijing, China · LeadSee Team, Beijing Meteorological Service
From Global to Local: Efficient Regional Weather Downscaling with Global Weather Foundation Model
Jul 3, 2026Wiktor Kamzela, Jakub Kubiak, Adam Dobosz +5Spatiotemporal Forecasting Weather Forecasting
Pozna´n Supercomputing and Networking Center · Pozna´n University of Technology

Oct 8, 2026J/K move · Enter open · S save

Learning Kilometer-Scale Weather Prediction with Global-Regional Alignment

Guowen Li, Yang Liu, Yujie Wang, Qiuyan Sun+4

Tsinghua University · The Chinese University of Hong Kong · Sun Yat-Sen University+1

Weather Forecasting Representation Alignment

Open paper AskPDF arXiv

Kilometer-scale regional weather forecasting is essential for local weather warnings and weather-sensitive decisions. Existing data-driven approaches often rely on numerical forecasts for large-scale guidance or require additional training of global forecasting components. Pretrained global weather models offer an efficient source of large-scale forecasts, motivating their reuse to guide high-resolution regional prediction. However, this coupling requires aligning global and regional representations across different grids and integrating global guidance with local interactions to advance regional states. We propose ScaleCast, a regional forecasting framework that addresses these challenges through Global-Regional Alignment. Its Global-Regional Conversion module aligns joint global and regional representations with regional locations, while the Global-Regional Alignment and Dynamics block combines aligned guidance with regional neighborhood interactions. Experiments using ERA5 global analyses on a 0.25-degree grid and CERRA regional reanalysis at 5.5 km spacing demonstrate improved regional forecasts across surface and upper-air variables, with a single trained model supporting multiple global forecast drivers (i.e., Pangu-Weather, GraphCast, and HRES) without specific retraining. Fine-tuning on HRRR at 3 km spacing further demonstrates the framework's adaptability to a different regional domain and spatial resolution. Windstorm case studies show improved cyclone positioning and core-pressure estimates, while comparisons with HadISD station observations show closer agreement with local temperature and humidity changes.

Abstract

Weather Forecasting

Explore similar work

Learning Kilometer-Scale Weather Prediction with Global-Regional Alignment

StretchCast: Global-Regional AI Weather Forecasting on Stretched Cubed-Sphere Mesh

From Global to Local: Efficient Regional Weather Downscaling with Global Weather Foundation Model