Li et al. (2026) Multi-Scale and Interpretable Daily Runoff Forecasting with IEWT and ModernTCN

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

Identification

• Journal: Water
• Year: 2026
• Date: 2026-01-09
• Authors: Qing Li, Zhou Yunwei, Yongshun Zheng, Chu Zhang, Tian Peng
• DOI: 10.3390/w18020183

Research Groups

Not explicitly stated in the provided text.

Short Summary

This study proposes a novel framework, IPBT-IEWT-SHAP-ModernTCN, for daily runoff prediction by integrating multi-scale decomposition, interpretable feature selection, and advanced deep learning, incorporating upstream-downstream hydrological correlation. The framework significantly enhances prediction accuracy, stability, and generalization compared to benchmark methods, providing an efficient tool for water resource management.

Objective

• To propose a runoff prediction framework that incorporates upstream–downstream hydrological correlation information and integrates Improved Empirical Wavelet Transform (IEWT), SHAP-based interpretable feature selection, Improved Population-Based Training (IPBT), and the Modern Temporal Convolutional Network (ModernTCN) to enhance forecasting accuracy and model robustness for daily runoff.

Study Configuration

• Spatial Scale: Basin scale, focusing on upstream-downstream hydrological correlations, demonstrated with a case study at the Hankou station.
• Temporal Scale: Daily (daily runoff series, daily runoff sequence, daily runoff prediction).

Methodology and Data

• Models used: Improved Empirical Wavelet Transform (IEWT), SHAP (SHapley Additive exPlanations) method for feature selection, Improved Population-Based Training (IPBT) for hyperparameter optimization, Modern Temporal Convolutional Network (ModernTCN) for forecasting. Benchmark methods included LSTM, iTransformer, and TCN.
• Data sources: Daily runoff sequences, upstream-downstream hydrological correlation information, and multi-source basin features.

Main Results

• The proposed IPBT-IEWT-SHAP-ModernTCN model significantly outperforms benchmark methods (LSTM, iTransformer, TCN) in terms of accuracy, stability, and generalization.
• Achieved a root mean square error (RMSE) of 342.14.
• Achieved a mean absolute error (MAE) of 251.01.
• Achieved a Nash–Sutcliffe efficiency (NSE) of 0.9992.
• The method effectively captures the nonlinear correlation characteristics between upstream and downstream hydrological processes.

Contributions

• Introduction of a novel integrated framework (IPBT-IEWT-SHAP-ModernTCN) for daily runoff prediction, combining multi-scale decomposition, interpretable feature selection, and advanced deep learning.
• Incorporation of upstream-downstream hydrological correlation information to improve prediction accuracy and model robustness.
• Demonstration of superior performance in accuracy, stability, and generalization compared to existing benchmark models.
• Provides an efficient and widely adaptable framework for scientific water resources management by effectively capturing complex nonlinear hydrological correlations.

Funding

Not explicitly stated in the provided text.

Citation

@article{Li2026MultiScale,
  author = {Li, Qing and Yunwei, Zhou and Zheng, Yongshun and Zhang, Chu and Peng, Tian},
  title = {Multi-Scale and Interpretable Daily Runoff Forecasting with IEWT and ModernTCN},
  journal = {Water},
  year = {2026},
  doi = {10.3390/w18020183},
  url = {https://doi.org/10.3390/w18020183}
}