Abstract
Artificial intelligence has improved the accuracy and efficiency of weather forecasting, surpassing traditional numerical weather prediction models. However, the coarse spatial resolution of global weather forecasting systems limits their ability to capture fine-scale surface heterogeneity and localized extremes, particularly in regions with complex terrain or urban heat island effects. Here, we introduce SR-Weather, a deep learning-based super-resolution framework that converts coarse 0.25° forecasts into 1-km surface air temperature fields using MODIS-derived temperature targets and high-resolution auxiliary inputs. SR-Weather outperforms existing super-resolution methods by explicitly incorporating spatial context, such as topography, impervious surface fraction, and seasonal climatology maps of air temperature. When SR-Weather was applied to the FuXi global weather forecast, the 7-day forecast error in South Korea decreased by more than 20%, which was comparable to the 1-day forecast error from low-resolution prediction using simple spatial interpolation. In addition, SR-Weather effectively reconstructs missing pixels in MODIS-derived air temperature maps under heavy cloud contamination by leveraging auxiliary variables and climatologically smoothed fields. Although validated over South Korea, the framework relies on globally available MODIS products and minimal auxiliary inputs, making it feasible to retrain for other regions. These results indicate that SR-Weather is a scalable and high-fidelity tool for enhancing machine learning-based weather forecasts at fine spatial scales.
Similar content being viewed by others
Data availability
The FuXi 0.25° forecast data were obtained from https://weatherbench2.readthedocs.io/. The ERA5 2 meter temperature (T2M) was from https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-pressure-levels?tab=form. The MODIS/Terra 1 km Land Surface Temperature (LST) product was downloaded from https://www.earthdata.nasa.gov/data/catalog/lpcloud-mod21a1d-061. The Shuttle Radar Topography Mission (SRTM) Void Filled Global 3 arc-second (2012 release) digital elevation model was from https://earthexplorer.usgs.gov/. MODIS/Terra+Aqua Land Cover Type was from https://www.earthdata.nasa.gov/data/catalog/lpcloud-mcd12q1-061. The Automated Synoptic Observing System (ASOS) and Automatic Weather Station (AWS) were obtained from https://data.kma.go.kr.
References
Qian, Y. et al. Urbanization impact on regional climate and extreme weather: Current understanding, uncertainties, and future research directions. Adv. Atmos. Sci. 39, 819–860 (2022).
Hsu, A., Sheriff, G., Chakraborty, T. & Manya, D. Disproportionate exposure to urban heat island intensity across major US cities. Nat. Commun. 12, 2721 (2021).
Li, Y. et al. Green spaces provide substantial but unequal urban cooling globally. Nat. Commun. 15, 7108 (2024).
Watt-Meyer, O. et al. Neural network parameterization of subgrid-scale physics from a realistic geography global storm-resolving simulation. J. Adv. Model. Earth Syst. 16, e2023MS003668 (2024).
Stevens, B. et al. DYAMOND: the DYnamics of the Atmospheric general circulation Modeled On Non-hydrostatic Domains. Prog. Earth Planet. Sci. 6, 1–17 (2019).
Leutwyler, D., Lüthi, D., Ban, N., Fuhrer, O. & Schär, C. Evaluation of the convection-resolving climate modeling approach on continental scales. J. Geophys. Res. Atmos. 122, 5237–5258 (2017).
Prein, A. F. et al. A review on regional convection-permitting climate modeling: Demonstrations, prospects, and challenges. Rev. Geophys. 53, 323–361 (2015).
Termonia, P. et al. The CORDEX.be initiative as a foundation for climate services in Belgium. Clim. Serv. 11, 49–61 (2018).
Korea Meteorological Administration. Numerical Data Application Manual (KMA, Seoul, 2011).
Dowell, D. C. et al. The High-Resolution Rapid Refresh (HRRR): An hourly updating convection-allowing forecast model. Part I: Motivation and system description. Weather Forecast 37, 1371–1395 (2022).
Baldauf, M. et al. Operational convective-scale numerical weather prediction with the COSMO model: Description and sensitivities. Mon. Weather Rev. 139, 3887–3905 (2011).
Dirmeyer, P. A. et al. Verification of land–atmosphere coupling in forecast models, reanalyses, and land surface models using flux site observations. J. Hydrometeorol. 19, 375–392 (2018).
Bi, K. et al. Accurate medium-range global weather forecasting with 3D neural networks. Nature 619, 533–538 (2023).
Lam, R. et al. Learning skillful medium-range global weather forecasting. Science 382, 1416–1421 (2023).
Chen, L. et al. FuXi: A cascade machine learning forecasting system for 15-day global weather forecast. npj Clim. Atmos. Sci. 6, 190 (2023).
Mardani, M. et al. Residual corrective diffusion modeling for km-scale atmospheric downscaling. Commun. Earth Environ. 6, 124 (2025).
Balsamo, G. et al. Satellite and in situ observations for advancing global Earth surface modelling: A review. Remote Sens 10, 2038 (2018).
Meng, Q. et al. GLOSTFM: A global spatiotemporal fusion model integrating multi-source satellite observations to enhance land surface temperature resolution. Remote Sens. Environ. 319, 114640 (2025).
Zou, R., Wei, L. & Guan, L. Super resolution of satellite-derived sea surface temperature using a transformer-based model. Remote Sens 15, 5376 (2023).
Chen, X., Wang, X., Zhou, J., Qiao, Y. & Dong, C. Activating more pixels in image super-resolution transformer. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 22367–22377 (Vancouver, BC, Canada, 2023).
Ledig, C. et al. Photo-realistic single image super-resolution using a generative adversarial network. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 4681–4690 (Honolulu, HI, USA, 2017).
Yasuda, Y., Onishi, R., Hirokawa, Y., Kolomenskiy, D. & Sugiyama, D. Super-resolution of near-surface temperature utilizing physical quantities for real-time prediction of urban micrometeorology. Build. Environ. 209, 108597 (2022).
Li, S., Wan, H., Yu, Q. & Wang, X. Downscaling of ERA5 reanalysis land surface temperature based on attention mechanism and Google Earth Engine. Sci. Rep. 15, 675 (2025).
Liang, M. et al. A high-resolution land surface temperature downscaling method based on geographically weighted neural network regression. Remote Sens 15, 1740 (2023).
Weng, Q., Fu, P. & Gao, F. Generating daily land surface temperature at Landsat resolution by fusing Landsat and MODIS data. Remote Sens. Environ. 145, 55–67 (2014).
Zhang, T., Zhou, Y., Zhu, Z., Li, X. & Asrar, G. R. A global seamless 1 km resolution daily land surface temperature dataset (2003–2020). Earth Syst. Sci. Data 14, 651–664 (2022).
Shi, W. et al. Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 1874–1883 (Las Vegas, NV, USA, 2016).
Islam, M. A., Kowal, M., Jia, S., Derpanis, K. G. & Bruce, N. D. Global pooling, more than meets the eye: Position information is encoded channel-wise in CNNs. 2021 IEEE/CVF International Conference on Computer Vision (ICCV), 793–801 (Montreal, QC, Canada, 2021).
Buster, G., Cox, J., Benton, B. N. & King, R. N. Estimating the impacts of increasing temperatures and the efficacy of climate adaptation strategies in urban microclimates with deep learning. Urban Climate 64, 102603 (2025).
Yang, Q. et al. Local off-grid weather forecasting with multi-modal Earth observation data. arXiv:2410.12938 (2024).
Byun, J. & Paik, K. The development process of the Korean coastal mountain range: Examination from spatial distribution of knickzones. Prog. Phys. Geogr. Earth Environ. 45, 541–563 (2021).
Tsai, C. L., Kim, K., Liou, Y. C., Lee, G. & Yu, C. K. Impacts of topography on airflow and precipitation in the Pyeongchang area seen from multiple-Doppler radar observations. Mon. Weather Rev. 146, 3401–3424 (2018).
Hwang, Y., Ryu, Y. & Qu, S. Expanding vegetated areas by human activities and strengthening vegetation growth concurrently explain the greening of Seoul. Landsc. Urban Plan. 227, 104518 (2022).
Bae, J. S., Joo, R. W. & Kim, Y. S. Forest transition in South Korea: reality, path and drivers. Land Use Policy 29, 198–207 (2012).
Yoo, C., Im, J., Park, S. & Quackenbush, L. J. Estimation of daily maximum and minimum air temperatures in urban landscapes using MODIS time series satellite data. ISPRS J. Photogramm. Remote Sens. 137, 149–162 (2018).
Wan, Z., Hook, S. & Hulley, G. MODIS/Terra Land Surface Temperature/Emissivity Daily L3 Global 1 km SIN Grid (Version 6.1) [MOD11A1]. NASA EOSDIS Land Processes DAAC (2021). https://doi.org/10.5067/MODIS/MOD11A1.061.
Wan, Z. New refinements and validation of the collection-6 MODIS land-surface temperature/emissivity product. Remote Sens. Environ. 140, 36–45 (2014).
Rasp, S. et al. Weatherbench 2: A benchmark for the next generation of data-driven global weather models. J. Adv. Model. Earth Syst. 16, e2023MS004019 (2024).
Reuter, H. I., Nelson, A. & Jarvis, A. An evaluation of void-filling interpolation methods for SRTM data. Int. J. Geogr. Inf. Sci. 21, 983–1008 (2007).
Friedl, M. & Sulla-Menashe, D. MODIS/Terra+Aqua Land Cover Type Yearly L3 Global 500 m SIN Grid (Version 6.1) [MCD12Q1]. NASA EOSDIS Land Processes DAAC (2022). https://doi.org/10.5067/MODIS/MCD12Q1.061.
Sulla-Menashe, D., Gray, J. M., Abercrombie, S. P. & Friedl, M. A. Hierarchical mapping of annual global land cover 2001 to present: The MODIS Collection 6 Land Cover product. Remote Sens. Environ. 222, 183–194 (2019).
Woo, S., Park, J., Lee, J. Y. & Kweon, I. S. CBAM: Convolutional block attention module. European Conference on Computer Vision (ECCV), 3–19 (Munich, Germany, 2018).
Liu, G. et al. Image inpainting for irregular holes using partial convolutions. European Conference on Computer Vision (ECCV), 85–100 (Munich, Germany, 2018).
Liang, J. et al. SwinIR: Image restoration using swin transformer. 2021 IEEE/CVF International Conference on Computer Vision (ICCV), 1833–1844 (Montreal, QC, Canada, 2021).
Acknowledgements
This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea Ministry of Science and ICT (MSIT) (NRF-2022M3K3A1094114 and RS-2025-02310080).
Author information
Authors and Affiliations
Contributions
H. Park, S. Park, D. Kang and J.-H. Kim designed the research. H. Park and D. Kang compiled the data. H. Park, S. Park, D. Kang and J.-H. Kim developed the methodology. H. Park, S. Park and D. Kang conducted analyses and prepared the figures. H. Park, S. Park and D. Kang wrote the first draft of the manuscript, and all authors contributed in the writing of the final version of the manuscript.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Park, H., Park, S., Kang, D. et al. A super-resolution framework for downscaling machine learning weather prediction toward 1-km air temperature. npj Clim Atmos Sci (2026). https://doi.org/10.1038/s41612-026-01328-5
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41612-026-01328-5
