Kongbuchakiat et al. (2026) Lead-time-dependent effects of spatial rainfall representations on deep learning discharge forecasting in the lower Chao Phraya River

Identification

• Journal: Journal of Hydrology Regional Studies
• Year: 2026
• Date: 2026-09-15
• Authors: Pornnapus Kongbuchakiat, Natthachet Tangdamrongsub, Tanuspong Pokavanich, Thanawin Rakthanmanon, Adichai Pornprommin
• DOI: 10.1016/j.ejrh.2026.103986

Research Groups

• Department of Water Resources Engineering, Faculty of Engineering, Kasetsart University, Thailand
• Water Engineering and Management, Faculty of Civil and Environmental Engineering, Asian Institute of Technology, Thailand
• Department of Marine Science, Faculty of Fishery, Kasetsart University, Thailand
• Department of Computer Engineering, Faculty of Engineering, Kasetsart University, Thailand
• Hydro-informatics Institute, Bangkok, Thailand

Short Summary

This study evaluates the lead-time dependent benefits (1–14 days) of integrating spatial rainfall information from the Thai Meteorological Department (TMD) and Climate Prediction Center (CPC) into RNN, LSTM, and GRU architectures for downstream discharge forecasting in the lower Chao Phraya River basin. The results show that adding rainfall information depends strongly on the forecast lead time.

Objective

• Evaluate the impact of spatial rainfall representation on downstream discharge forecasting across different forecast horizons.
• Investigate how different spatial rainfall representations affect predictive performance.
• Assess whether rainfall integration can reduce temporal phase lag in discharge prediction.

Study Configuration

• Spatial Scale: The study focuses on the lower Chao Phraya River basin, with a total drainage area of approximately 158,586 km².
• Temporal Scale: The forecast lead times were categorized into short-range (1–7 days) and medium-range (8–14 days) horizons.

Methodology and Data

• Models used: RNN, LSTM, GRU architectures for downstream discharge forecasting.
• Data sources:
  • Hydrological data from the Royal Irrigation Department (RID) of Thailand.
  • Rainfall information from the Climate Prediction Center (CPC) and ground-based rain gauge observations from the Thai Meteorological Department (TMD).

Main Results

• The benefit of adding rainfall information depends strongly on the forecast lead time.
• For short-range forecasts (1–7 days), downstream discharge is controlled by upstream flow, making added rainfall information redundant.
• For medium-range forecasts (8–14 days), rainfall information is essential to capture local runoff.
• The sub-basin area average approach tended to perform better for short-range forecasts, whereas CNN-based feature extraction was more effective for the medium range.

Contributions

• This study provides new insights into the lead-time dependent benefits of integrating spatial rainfall information into deep learning architectures for downstream discharge forecasting.
• The results highlight the importance of considering temporal phase lag in discharge prediction and its impact on flood preparedness and operational decision-making.

Funding

• This research was funded by [insert funding sources and project codes].

Citation

@article{Kongbuchakiat2026Leadtimedependent,
  author = {Kongbuchakiat, Pornnapus and Tangdamrongsub, Natthachet and Pokavanich, Tanuspong and Rakthanmanon, Thanawin and Pornprommin, Adichai},
  title = {Lead-time-dependent effects of spatial rainfall representations on deep learning discharge forecasting in the lower Chao Phraya River},
  journal = {Journal of Hydrology Regional Studies},
  year = {2026},
  doi = {10.1016/j.ejrh.2026.103986},
  url = {https://doi.org/10.1016/j.ejrh.2026.103986}
}