Lin et al. (2026) Enhancing long-term reservoir inflow forecasting: an integrated approach combining switch prediction method, ensemble rainfall forecasts, and machine learning techniques

Identification

• Journal: Natural hazards and earth system sciences
• Year: 2026
• Date: 2026-07-03
• Authors: Hsuan-Yu Lin, Jhih-Huang Wang, Ming-Jui Chang
• DOI: 10.5194/nhess-26-3085-2026

Research Groups

• Department of Civil and Water Resources Engineering, National Chiayi University
• Department of Civil Engineering, National Ilan University
• Hydrotech Research Institute, National Taiwan University

Short Summary

This study proposes a novel framework that integrates the Switch Prediction Method (SPM) with machine learning techniques to forecast reservoir inflows up to 72 h in advance during typhoon events. The SPM dynamically selects and combines rainfall forecasts based on real-time performance metrics, improving predictive reliability.

Objective

• Evaluate the effectiveness of integrating ensemble rainfall forecasts using the Switch Prediction Method (SPM)
• Develop a robust forecasting framework for long-term reservoir inflow prediction under highly variable typhoon conditions

Study Configuration

• Spatial Scale: ShihMen Reservoir catchment area, approximately 763 km2
• Temporal Scale: Up to 72 h lead time

Methodology and Data

• Models used:
  • Switch Prediction Method (SPM)
  • Support Vector Machine (SVM)
  • Long Short-Term Memory (LSTM) networks
• Data sources:
  • Historical typhoon events data from 2004 to 2019
  • Meteorological forecast products, including QPF, ETQPF, CWB-WRF, WEPS-PM, and STMAS-WRF

Main Results

• The proposed SPM-LSTM-MSF framework achieved the best performance in predicting long-term reservoir inflows under typhoon conditions.
• The model demonstrated high accuracy and temporal stability, with a Mean Absolute Error (MAE) of 178.8 m3 s−1 and a Coefficient of Efficiency (CE) of 0.87 for Typhoon Soudelor.

Contributions

• This study makes a unique contribution by evaluating the effectiveness of integrating ensemble rainfall forecasts using the Switch Prediction Method (SPM).
• The proposed framework improves the robustness and reliability of long-lead inflow forecasting, providing practical value for reservoir operation under typhoon conditions.

Funding

• National Science Foundation (NSF) grant number: [insert NSF grant number]
• Ministry of Science and Technology (MOST) grant number: [insert MOST grant number]

Citation

@article{Lin2026Enhancing,
  author = {Lin, Hsuan-Yu and Wang, Jhih-Huang and Chang, Ming-Jui},
  title = {Enhancing long-term reservoir inflow forecasting: an integrated approach combining switch prediction method, ensemble rainfall forecasts, and machine learning techniques},
  journal = {Natural hazards and earth system sciences},
  year = {2026},
  doi = {10.5194/nhess-26-3085-2026},
  url = {https://doi.org/10.5194/nhess-26-3085-2026}
}