Su et al. (2025) Space-time deep hybrid boosting learning for investigating day-night hourly seamless air temperature distribution from FY-4A over China

Identification

• Journal: Journal of Hydrology
• Year: 2025
• Date: 2025-09-11
• Authors: Qin Su, Yuan Wang, Yuan‐Han Yang, Yuyu Zhou, Bingcheng Wan, Tongwen Li, Tingting Zhong, Ziyan Lu, Zunyi Xie, Hung Chak Ho, Qiangqiang Yuan
• DOI: 10.1016/j.jhydrol.2025.134245

Research Groups

• School of Atmospheric Physics, Nanjing University of Information Science and Technology, Nanjing, China
• Department of Geography, The University of Hong Kong, Hong Kong, China
• Institute for Climate and Carbon Neutrality, The University of Hong Kong, Hong Kong, China
• School of Artificial Intelligence, Nanjing University of Information Science and Technology, Nanjing, China
• Department of Civil and Environmental Engineering, Hong Kong University of Science and Technology, Hong Kong, China
• School of Geospatial Engineering and Science, Sun Yat-sen University, Zhuhai, China
• College of Long Shan, Nanjing University of Information Science and Technology, Nanjing, China
• College of Geography and Environmental Science, Henan University, Kaifeng, China
• School of Earth and Environmental Sciences, The University of Queensland, Brisbane, Australia
• Department of Public and International Affairs, The City University of Hong Kong, Hong Kong, China
• Institute of Global Governance and Innovation for a Shared Future, The City University of Hong Kong, Hong Kong, China
• School of Geodesy and Geomatics, Wuhan University, Wuhan, China

Short Summary

This study developed a Space-Time Deep Hybrid Boosting (ST-DHB) model to generate day-night hourly seamless 0.04-degree air temperature (Ta) distributions across China from Fengyun-4A data, achieving high accuracy (R² > 0.94, RMSE < 2.6 °C) and outperforming existing methods. The resulting Ta data reveals significant geographical, seasonal, and diurnal disparities in heatwave exposure, particularly in urban and farmland areas.

Objective

• To develop a Space-Time Deep Hybrid Boosting (ST-DHB) model to overcome limitations of existing air temperature (Ta) estimation methods (e.g., focus on daytime, coarse temporal resolution, insufficient spatiotemporal characteristics) and generate day-night hourly seamless 0.04-degree Ta distributions across China using Fengyun-4A data.

Study Configuration

• Spatial Scale: China, with a spatial resolution of 0.04 degrees.
• Temporal Scale: Hourly, covering both day and night seamlessly.

Methodology and Data

• Models used: Space-Time Deep Hybrid Boosting (ST-DHB) model. Comparisons were made against several widely used machine learning methods.
• Data sources: Fengyun-4A (FY-4A) satellite data, used for air temperature (Ta) estimation.

Main Results

• The ST-DHB model demonstrated high performance in estimating air temperature, with R² values of 0.946 and RMSE of 2.593 °C during daytime, and R² values of 0.958 and RMSE of 2.218 °C during nighttime.
• The developed model achieved superior metrics compared to several widely used machine learning methods and recent studies.
• Estimated Ta distributions displayed continuous spatial details and accurately captured hourly and seasonal Ta variations.
• Urban areas and farmland with large populations were identified as experiencing more severe high-temperature exposure at both spatial and temporal scales, indicating increased heatwave threats to human health.
• The estimated Ta effectively supports daytime and nighttime heatwave exposure assessment, revealing significant geographical, seasonal, and diurnal disparities of heatwaves across China.

Contributions

• Development of a novel Space-Time Deep Hybrid Boosting (ST-DHB) model for seamless day-night hourly air temperature estimation.
• Generation of a high-resolution (0.04-degree, hourly) and continuous day-night air temperature dataset for China, addressing limitations of existing methods regarding temporal resolution and day-night patterns.
• Provides a reliable estimation model and Ta dataset crucial for assessing health risks associated with day-night composite heatwave exposure.
• Identifies specific areas (urban and farmland) with higher heatwave exposure, contributing to targeted public health interventions.

Funding

[No specific funding information was provided in the abstract or affiliations section of the paper.]

Citation

@article{Su2025Spacetime,
  author = {Su, Qin and Wang, Yuan and Yang, Yuan‐Han and Zhou, Yuyu and Wan, Bingcheng and Zong, Lian and Li, Tongwen and Zhong, Tingting and Lu, Ziyan and Xie, Zunyi and Ho, Hung Chak and Yuan, Qiangqiang},
  title = {Space-time deep hybrid boosting learning for investigating day-night hourly seamless air temperature distribution from FY-4A over China},
  journal = {Journal of Hydrology},
  year = {2025},
  doi = {10.1016/j.jhydrol.2025.134245},
  url = {https://doi.org/10.1016/j.jhydrol.2025.134245}
}