Wang et al. (2025) Season-Aware Ensemble Forecasting with Improved Arctic Puffin Optimization for Robust Daily Runoff Prediction Across Multiple Climate Zones

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

Identification

• Journal: Water
• Year: 2025
• Date: 2025-12-11
• Authors: Wenchuan Wang, Xutong Zhang, Qiqi Zeng, Dongmei Xu
• DOI: 10.3390/w17243504

Research Groups

Not explicitly stated in the provided text.

Short Summary

This paper proposes a Season-Aware Ensemble Forecasting (SAEF) method that integrates multiple machine learning models with seasonal segmentation and an optimized weighting algorithm to improve daily runoff forecasting accuracy and provide physically interpretable insights into seasonal hydrological dynamics across diverse climates.

Objective

• To develop a Season-Aware Ensemble Forecasting (SAEF) method that addresses seasonal non-stationarity and inter-basin variability in runoff sequences to enhance daily runoff prediction accuracy and provide physically interpretable insights into seasonal runoff generation processes.

Study Configuration

• Spatial Scale: Dongjiang Hydrological Station (China), Elbe River (Germany), Quinebaug River basin (USA).
• Temporal Scale: Daily runoff forecasting, with annual runoff data segmented into four seasons (spring, summer, autumn, winter).

Methodology and Data

• Models used: Season-Aware Ensemble Forecasting (SAEF) method, which integrates Support Vector Machine (SVM), Least Squares Support Vector Machine (LSSVM), Long Short-Term Memory (LSTM), and Bidirectional Long Short-Term Memory (BiLSTM) models. An Improved Arctic Puffin Optimization (IAPO) algorithm is used to optimize model weights.
• Data sources: Observed runoff data from hydrological stations in the Dongjiang, Elbe, and Quinebaug river basins.

Main Results

• The SAEF method achieved average Nash–Sutcliffe Efficiency Coefficient (NSEC) values greater than 0.92 and Kling–Gupta efficiency (KGE) coefficients greater than 0.90 across the case studies.
• SAEF significantly outperformed individual models (SVM, LSSVM, LSTM, BiLSTM) with Root Mean Square Error (RMSE) reductions of up to 58.54% (vs SVM), 55.62% (vs LSSVM), 51.99% (vs LSTM), and 48.14% (vs BiLSTM).
• The framework effectively reflects seasonal runoff generation processes, such as rapid rainfall–runoff in wet seasons and baseflow contributions in dry periods, providing a physically interpretable perspective.

Contributions

• Introduces a novel Season-Aware Ensemble Forecasting (SAEF) method that integrates multiple machine learning models with a seasonal segmentation mechanism and an optimized weighting algorithm (IAPO).
• Significantly improves daily runoff forecasting accuracy across diverse climatic regions by effectively addressing seasonal non-stationarity and inter-basin variability.
• Provides a physically interpretable framework by linking data-driven ensembles with seasonal hydrological process mechanisms, enhancing understanding of runoff dynamics.
• Offers a robust and interpretable tool for operational flood control and water resource management.

Funding

Not explicitly stated in the provided text.

Citation

@article{Wang2025SeasonAware,
  author = {Wang, Wenchuan and Zhang, Xutong and Zeng, Qiqi and Xu, Dongmei},
  title = {Season-Aware Ensemble Forecasting with Improved Arctic Puffin Optimization for Robust Daily Runoff Prediction Across Multiple Climate Zones},
  journal = {Water},
  year = {2025},
  doi = {10.3390/w17243504},
  url = {https://doi.org/10.3390/w17243504}
}