EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Evaluating the Generalization of an Ensemble Learning Model for Global Horizontal Irradiance Estimation in Guangxi Province Using FY-4A Satellite Data

Jiaqiu Hu, Yiming Qin, Qian Ye, Kui Huang, Houjian Zhan, Jian Tang, Jie Lin, Yixin Zhuo, Huanxing Qi

Journal of Environmental & Earth Sciences · 15.04.2026 · 143-156

DOI: 10.30564/jees.v8i4.13005 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Environmental & Earth Sciences, online erschienen am 15.04.2026. Verfasst von Jiaqiu Hu und 8 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This study investigates the application of the Extreme Gradient Boosting (XGBoost) ensemble learning algorithm for estimating global horizontal irradiance (GHI) based on satellite data in Guangxi Province, China. By synergistically integrating top-of-atmosphere (TOA) reflectance and brightness temperature data from 14 spectral bands of the Fengyun-4A (FY-4A) Advanced Geosynchronous Radiation Imager (AGRI) and European Centre for Medium-Range Weather Forecasts Reanalysis v5 (ERA5) reanalysis meteorological variables—including relative humidity, planetary boundary layer height, and surface pressure—we developed an all-sky model to predict hourly surface solar radiation. The model was trained on data from 159 ground stations across China in 2018 and incorporates 31 features covering satellite observations, geographical parameters, and meteorological variables. Validation was conducted using independent observational data from three additional ground stations in Guangxi (Guilin, Nanning, and Beihai) that were withheld from training, yielding Root Mean Square Error (RMSE) of approximately 126–150 W/m2 and Correlation Coefficient (CC) of 0.80–0.84, confirming strong spatial generalization. Seasonal analysis revealed that the model performed best in winter and least accurately in summer, attributable to the complexity of convective cloud dynamics in the subtropical monsoon climate of Guangxi. Feature importance analysis identified brightness temperature at 7.42 μm, solar zenith angle, relative humidity, and TOA reflectance at 0.47 μm as the most influential predictors, consistent with the physical mechanisms governing atmospheric transmissivity. These findings demonstrate that the direct data-driven satellite–machine learning framework offers a computationally efficient and scalable alternative to semi-empirical approaches for regional solar resource assessment.

Zitieren

APA

Hu, J., Qin, Y., Ye, Q., Huang, K., Zhan, H., Tang, J., Lin, J., Zhuo, Y., & Qi, H. (2026). Evaluating the Generalization of an Ensemble Learning Model for Global Horizontal Irradiance Estimation in Guangxi Province Using FY-4A Satellite Data. Journal of Environmental & Earth Sciences, 143-156. https://doi.org/10.30564/jees.v8i4.13005

Vancouver

Hu J, Qin Y, Ye Q, Huang K, Zhan H, Tang J, et al. Evaluating the Generalization of an Ensemble Learning Model for Global Horizontal Irradiance Estimation in Guangxi Province Using FY-4A Satellite Data. Journal of Environmental & Earth Sciences. 2026:143-156. doi:10.30564/jees.v8i4.13005

BibTeX
@article{Hu2026,
  author = {Hu, Jiaqiu and Qin, Yiming and Ye, Qian and Huang, Kui and Zhan, Houjian and Tang, Jian and Lin, Jie and Zhuo, Yixin and Qi, Huanxing},
  title = {Evaluating the Generalization of an Ensemble Learning Model for Global Horizontal Irradiance Estimation in Guangxi Province Using FY-4A Satellite Data},
  journal = {Journal of Environmental & Earth Sciences},
  year = {2026},
  pages = {143--156},
  doi = {10.30564/jees.v8i4.13005}
}
Kostenlos starten →