Zhang et al. (2025) Applicability of a Sine–Random Forest Hybrid Method for meteorological and energy variables

Identification

• Journal: Journal of Hydrology
• Year: 2025
• Date: 2025-09-12
• Authors: Siyao Zhang, Jianzhu Li, Ting Zhang, Jiyang Tian, Ping Feng
• DOI: 10.1016/j.jhydrol.2025.134238

Research Groups

• State Key Laboratory of Hydraulic Engineering Intelligent Construction and Operation, Tianjin University, Tianjin 300350, China
• China Institute of Water Resources and Hydropower Research, Beijing 100038, China
• Research Center on Flood & Drought Disaster Reduction, The Ministry of Water Resources of China, Beijing 100038, China

Short Summary

This study proposes a Sine-Random Forest Hybrid Method to reduce bias and enhance the accuracy of meteorological and energy variables in reanalysis datasets, demonstrating its effectiveness in improving agreement with measured data and capturing diurnal patterns.

Objective

• To propose and evaluate the applicability of a novel Sine-Random Forest Hybrid Method for reducing bias and enhancing the accuracy of meteorological and energy variables in reanalysis datasets (MERRA2, GLDAS 2.1, ERA5) compared to measured data.

Study Configuration

• Spatial Scale: Regional to Global (evaluation across various regions where measured data are available, applied to globally covered reanalysis datasets).
• Temporal Scale: Diurnal (analysis of diurnal patterns and maximum value occurrence times).

Methodology and Data

• Models used: Sine-Random Forest Hybrid Method, Random Forest (RF), Sine function.
• Data sources:
  • Reanalysis datasets: Modern-ERA Retrospective analysis for Research and Applications (MERRA2), Global Land Data Assimilation System (GLDAS) 2.1, European Centre for Medium-range Weather Forecasts atmospheric reanalysis (ERA5).
  • Measured data (in situ observations).

Main Results

• The Sine-RF Hybrid Method significantly improved the agreement between corrected reanalysis datasets and measured data, as indicated by maximal information coefficients (MICs) and Kling-Gupta Efficiency (KGEs).
• The method reduced MIC and KGE differences between MERRA2 and ERA5 by 10.3% and 84.7%, respectively, and between ERA5 and GLDAS by 45.6% and 70.2%, respectively.
• Amplitude differences for energy variables and phase differences for all variables were reduced after correction, with the relative mean absolute error (RMAE) decreasing by 29.1%.
• Corrected datasets effectively captured the maximum value occurrence times, particularly for energy variables.
• The Sine-RF Hybrid Method is effective for net shortwave radiation flux (SWNET), net radiation flux (NetRAD), latent heat flux (LE), sensible heat flux (H), and ground heat flux (G).

Contributions

• Introduction of a novel Sine-Random Forest Hybrid Method for bias correction of meteorological and energy variables in reanalysis datasets.
• Demonstrated significant improvement in the accuracy of reanalysis data, particularly in reducing amplitude and phase differences and better capturing diurnal patterns.
• Provided a practical and effective approach to enhance the reliability of widely used reanalysis datasets for ecohydrological and land-atmosphere studies.

Funding

• The provided paper text does not contain explicit funding information.

Citation

@article{Zhang2025Applicability,
  author = {Zhang, Siyao and Li, Jianzhu and Zhang, Ting and Tian, Jiyang and Feng, Ping},
  title = {Applicability of a Sine–Random Forest Hybrid Method for meteorological and energy variables},
  journal = {Journal of Hydrology},
  year = {2025},
  doi = {10.1016/j.jhydrol.2025.134238},
  url = {https://doi.org/10.1016/j.jhydrol.2025.134238}
}