A hybrid method for winter road surface temperature prediction using improved LSTMs and stacking-based ensemble learning

Li, Wanting; Zhou, Linyi; Wu, Xianghua; Guan, Yuanhong; Guo, Yuanhao; Chen, Kun; Huang, Weiqi; Zhao, Wenqian

doi:10.5194/gmd-19-9035-2026

Articles | Volume 19, issue 18

https://doi.org/10.5194/gmd-19-9035-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/gmd-19-9035-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 19, issue 18

Development and technical paper

|

24 Sep 2026

Development and technical paper |

| 24 Sep 2026

A hybrid method for winter road surface temperature prediction using improved LSTMs and stacking-based ensemble learning

Wanting Li, Linyi Zhou, Xianghua Wu, Yuanhong Guan, Yuanhao Guo, Kun Chen, Weiqi Huang, and Wenqian Zhao

Supplement

https://doi.org/10.5194/gmd-19-9035-2026-supplement

Data sets

A hybrid method for winter road surface temperature prediction using improved LSTMs and stacking-based ensemble learning W. Li https://doi.org/10.5281/zenodo.22020890

ERA5-Land hourly data from 1950 to present J. Muñoz Sabater https://doi.org/10.24381/cds.e2161bac

Model code and software

A hybrid method for winter road surface temperature prediction using improved LSTMs and stacking-based ensemble learning W. Li https://doi.org/10.5281/zenodo.22020890

Short summary

Accurate winter road surface temperature prediction prevents icy‑road accidents. Existing methods either need rare pavement parameters or miss local weather patterns and long‑term trends. This study combines two deep‑learning models to analyse four winters of Jiangsu data. It achieves lower prediction errors in general and outperforms common models at 1, 3 and 6 h horizons. Station‑based features beat external reanalysis data.