Neftissov et al. (2025) An Advanced Ensemble Machine Learning Framework for Estimating Long-Term Average Discharge at Hydrological Stations Using Global Metadata

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Water
• Year: 2025
• Date: 2025-07-14
• Authors: Alexandr Neftissov, Аndrii Biloshchytskyi, Ilyas Kazambayev, Serhii Dolhopolov, Tetyana Honcharenko
• DOI: 10.3390/w17142097

Research Groups

Not specified

Short Summary

This study develops a machine learning framework, utilizing a weighted Meta Ensemble model, to accurately estimate long-term average (LTA) discharge using global hydrological station metadata.

Objective

• To create a robust, accurate, and scalable data-driven solution for estimating long-term average discharge, specifically targeting data-scarce or ungauged basins.

Study Configuration

• Spatial Scale: Global (utilizing GRDC station metadata)
• Temporal Scale: Long-term average (LTA)

Methodology and Data

• Models used: Custom Deep Neural Network (DNN), XGBoost, LightGBM, CatBoost, and a weighted Meta Ensemble model; Bayesian optimization for hyperparameters and SHAP for interpretability.
• Data sources: Global Runoff Data Centre (GRDC) hydrological station metadata.

Main Results

• The weighted Meta Ensemble model outperformed individual models, achieving a Coefficient of Determination ($R^2$) of 0.954 on the independent test set.
• SHAP analysis revealed that catchment area and geographical attributes are the most significant predictors of LTA discharge.

Contributions

• Introduces a high-performance ensemble machine learning approach that significantly improves the estimation of LTA discharge compared to baseline models, providing a powerful tool for large-scale hydrological analysis in ungauged regions.

Funding

Not specified

Citation

@article{Neftissov2025Advanced,
  author = {Neftissov, Alexandr and Biloshchytskyi, Аndrii and Kazambayev, Ilyas and Dolhopolov, Serhii and Honcharenko, Tetyana},
  title = {An Advanced Ensemble Machine Learning Framework for Estimating Long-Term Average Discharge at Hydrological Stations Using Global Metadata},
  journal = {Water},
  year = {2025},
  doi = {10.3390/w17142097},
  url = {https://doi.org/10.3390/w17142097}
}