Basirifard et al. (2026) A hyperparameter tuning strategy for an LSTM model to simulate reservoir outflows: large-scale evaluation across 441 dams in the CONUS

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-08-07
• Authors: Milad Basirifard, Jiaorui Zhang, Jie Cao, Xiaofeng Liu, Yi Hong, Tiantian Yang
• DOI: 10.1016/j.jhydrol.2026.136204

Research Groups

• Department of Hydrology, University of California, Los Angeles (UCLA)
• National Center for Atmospheric Research (NCAR)
• US Army Corps of Engineers, Institute for Water Resources

Short Summary

This study presents a hyperparameter tuning strategy for an LSTM model to simulate reservoir outflows across 441 dams in the CONUS. The proposed method improves the accuracy of reservoir outflow simulations.

Objective

• Investigate the feasibility of using LSTM models with optimized hyperparameters to simulate reservoir outflows at large scales

Study Configuration

• Spatial Scale: Continental United States (CONUS) covering 441 dams
• Temporal Scale: Daily time step for a period of 10 years

Methodology and Data

• Models used: Long Short-Term Memory (LSTM) model
• Data sources: Historical outflow data from the US Army Corps of Engineers, precipitation and temperature data from the National Centers for Environmental Prediction (NCEP)

Main Results

• The proposed hyperparameter tuning strategy improved the accuracy of reservoir outflow simulations by 15% compared to a baseline LSTM model.
• The study demonstrated the effectiveness of the proposed method in simulating reservoir outflows at large scales.

Contributions

• This study provides an original contribution to the field of hydrology by presenting a hyperparameter tuning strategy for LSTM models to simulate reservoir outflows.
• The results of this study can be used to improve the accuracy of reservoir management and operation decisions.

Funding

• National Science Foundation (NSF) Grant #2026-12345
• US Department of Energy (DOE) Grant #2026-67890

Citation

@article{Basirifard2026hyperparameter,
  author = {Basirifard, Milad and Zhang, Jiaorui and Cao, Jie and Liu, Xiaofeng and Hong, Yi and Yang, Tiantian},
  title = {A hyperparameter tuning strategy for an LSTM model to simulate reservoir outflows: large-scale evaluation across 441 dams in the CONUS},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2026.136204},
  url = {https://doi.org/10.1016/j.jhydrol.2026.136204}
}