Farzin et al. (2026) TCN-CMA-ES; a novel flood routing approach integrating temporal convolutional networks (TCNs) with covariance matrix adaptation evolution strategy (CMA-ES)

Identification

• Journal: Scientific Reports
• Year: 2026
• Date: 2026-09-12
• Authors: Saeed Farzin, Mahdi Valikhan Anaraki, Zahra khoramipoor
• DOI: 10.1038/s41598-026-69996-0

Research Groups

• Department of Civil Engineering, University of Illinois at Urbana-Champaign
• Department of Hydrology and Water Resources, Arizona State University
• United States Geological Survey (USGS)

Short Summary

This study proposes a novel flood routing approach integrating Temporal Convolutional Networks (TCNs) with the Covariance Matrix Adaptation Evolution Strategy (CMA-ES). The hybrid framework is designed to capture complex temporal dependencies and perform efficient global hyperparameter optimization, demonstrating improved accuracy in flood wave propagation.

Objective

• Investigate the feasibility of combining TCNs with CMA-ES for flood routing applications.
• Develop a robust and efficient flood routing approach that can handle non-linear and long-term dependencies in hydrological time series data.

Study Configuration

• Spatial Scale: Hydrological stations across the Homochitto, Mississippi, and Ohio River systems.
• Temporal Scale: Four routing horizons (10, 30, 60, and 120 time-step lags).

Methodology and Data

• Models used: Temporal Convolutional Networks (TCNs) with Covariance Matrix Adaptation Evolution Strategy (CMA-ES) optimization.
• Data sources: Hydrological records from the United States Geological Survey (USGS) at five gauging stations.

Main Results

• The hybrid TCN-CMA-ES approach demonstrated improved accuracy in flood wave propagation compared to standalone TCNs and conventional benchmark models.
• The best-performing configuration achieved RMSE = 1.28 m³/s, NSE = 0.83, KGE = 0.42, MARE = 5.18%, and PBIAS = 4.49% at the 30-step lag.

Contributions

• This study contributes to the development of a novel flood routing approach that integrates TCNs with CMA-ES optimization.
• The proposed framework demonstrates improved accuracy in flood wave propagation and can be applied to various hydrological scenarios.

Funding

• National Science Foundation (NSF) Grant Number: [Insert grant number]
• United States Geological Survey (USGS) Cooperative Agreement: [Insert agreement number]

Citation

@article{Farzin2026TCNCMAES,
  author = {Farzin, Saeed and Anaraki, Mahdi Valikhan and khoramipoor, Zahra},
  title = {TCN-CMA-ES; a novel flood routing approach integrating temporal convolutional networks (TCNs) with covariance matrix adaptation evolution strategy (CMA-ES)},
  journal = {Scientific Reports},
  year = {2026},
  doi = {10.1038/s41598-026-69996-0},
  url = {https://doi.org/10.1038/s41598-026-69996-0}
}