So et al. (2026) Midterm drought forecasting based on dam storage prediction using deep learning algorithms

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

Identification

• Journal: Environmental Earth Sciences
• Year: 2026
• Date: 2026-09-21
• Authors: Hyeong-Yun So, Hyeon-Cheol Yoon, Tae-Gyun Kim, Se-Jeong Lee
• DOI: 10.1007/s12665-026-13138-2

Research Groups

• Department of Civil Engineering, Seoul National University
• Korea Institute of Construction Technology (KICT)

Short Summary

This study aims to improve midterm drought forecasting in South Korea by extending the prediction horizon to six months and modeling dam storage dynamics using deep learning algorithms.

Objective

• Investigate the feasibility of improving three-month drought forecasts to a six-month lead time for proactive drought preparedness in South Korea.

Study Configuration

• Spatial Scale: National scale, focusing on major dams in South Korea.
• Temporal Scale: Monthly predictions from January to June, with a focus on extending the prediction horizon by two months compared to existing three-month forecasts.

Methodology and Data

• Models used: Long Short-Term Memory (LSTM), Bidirectional LSTM (Bi-LSTM), Temporal Convolutional Network (TCN), and Convolutional Neural Network–LSTM (CNN-LSTM) architectures.
• Data sources: Observational hydrological data from major dams in South Korea.

Main Results

• The TCN architecture combined with training dataset B exhibited the highest performance among evaluated models, achieving an average Root Mean Squared Error (RMSE) of 35.440 and an error rate of approximately 6.23% relative to total storage.
• Qualitative evaluation indicates that the model effectively reproduces temporal storage variability patterns during severe drought conditions.

Contributions

• This study contributes to improving midterm drought forecasting in South Korea by extending the prediction horizon, enhancing proactive drought preparedness.
• The use of deep learning algorithms for modeling dam storage dynamics demonstrates robust quantitative predictive performance and qualitative evaluation.

Funding

• This research was supported by the Ministry of Science and ICT (MSIT) under the National Research Foundation of Korea (NRF) grant No. NRF-2020R1A2C2009446.
• Additional funding was provided by the Korea Institute of Construction Technology (KICT) through its internal research program.

Citation

@article{So2026Midterm,
  author = {So, Hyeong-Yun and Yoon, Hyeon-Cheol and Kim, Tae-Gyun and Lee, Se-Jeong},
  title = {Midterm drought forecasting based on dam storage prediction using deep learning algorithms},
  journal = {Environmental Earth Sciences},
  year = {2026},
  doi = {10.1007/s12665-026-13138-2},
  url = {https://doi.org/10.1007/s12665-026-13138-2}
}