Haghizadeh et al. (2026) Integrating HEC-RAS with machine learning and deep learning models as a rapid tool for river water level prediction

Identification

• Journal: Scientific Reports
• Year: 2026
• Date: 2026-09-15
• Authors: Ali Haghizadeh, Leila Ghasemi, Sanaz Vahidimanesh
• DOI: 10.1038/s41598-026-70828-4

Research Groups

• Department of Water Engineering, University of Tehran, Iran
• Lorestan Province Water Authority, Iran
• Institute of Geophysics and Geomatics, University of Science and Technology, Iran

Short Summary

This study presents a hybrid approach combining HEC-RAS with machine learning models to predict water surface elevation in the Khorram Abad River. The Random Forest model achieved an R2 of 0.991, outperforming other machine learning algorithms.

Objective

• To develop a rapid and accurate tool for predicting water surface elevation in rivers using a hybrid approach combining HEC-RAS with machine learning models.
• To investigate the importance of channel bed elevation and discharge in controlling water surface elevation.

Study Configuration

• Spatial Scale: River reach scale, 10 km long, Khorram Abad River, Lorestan Province, Iran.
• Temporal Scale: Eight return periods (2 to 500 years) simulated using HEC-RAS.

Methodology and Data

• Models used: HEC-RAS version 6.0, Random Forest, Support Vector Machine, Deep Neural Network, Recurrent Neural Network.
• Data sources: Hydrological data from Bahramju station, geometric data from cross-sections, hydraulic data generated using HEC-RAS.

Main Results

• The Random Forest model achieved an R2 of 0.991, with accuracy comparable to the physical model but with near-instantaneous execution.
• Channel bed elevation (35%) and the logarithm of discharge (24%) were the two most important predictors.
• The model identified dominant physical controls in the training data, not spurious correlations.

Contributions

• This study presents a hybrid approach combining HEC-RAS with machine learning models to predict water surface elevation in rivers.
• The Random Forest model outperformed other machine learning algorithms and achieved high accuracy.
• The study highlights the importance of channel bed elevation and discharge in controlling water surface elevation.

Funding

• This research was funded by the Lorestan Province Water Authority, Iran (Project Code: LWAP-2020).
• Additional funding was provided by the University of Tehran, Department of Water Engineering (Grant Number: UT-WER-2022).

Citation

@article{Haghizadeh2026Integrating,
  author = {Haghizadeh, Ali and Ghasemi, Leila and Vahidimanesh, Sanaz},
  title = {Integrating HEC-RAS with machine learning and deep learning models as a rapid tool for river water level prediction},
  journal = {Scientific Reports},
  year = {2026},
  doi = {10.1038/s41598-026-70828-4},
  url = {https://doi.org/10.1038/s41598-026-70828-4}
}