Wang et al. (2026) Integrating Physical-Based Xinanjiang Model and Deep Learning for Interpretable Streamflow Simulation: A Multi-Source Data Fusion Approach across Diverse Chinese Basins

Identification

• Journal: Hydrology and earth system sciences
• Year: 2026
• Date: 2026-09-01
• Authors: Zhaocai Wang, Nannan Xu, Wei Guo Song, Xingxing Zhang, Junhao Wu, Xi Chen
• DOI: 10.5194/hess-30-5521-2026

Research Groups

• College of Marine Ecology and Environment, Shanghai Ocean University
• Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences
• State Key Laboratory of Estuarine and Coastal Research, East China Normal University

Short Summary

This study presents a hybrid streamflow simulation model that combines the physical-based Xinanjiang (XAJ) model with the data-driven Temporal Convolutional Network (TCN)-Gated Recurrent Unit (GRU) model through nonlinear ensemble. The XAJ-TCN-GRU model exhibits promising results in simulating streamflow in four different basins in China, achieving high Nash-Sutcliffe Efficiency (NSE) values.

Objective

• To develop an interpretable hybrid streamflow simulation model that integrates physical mechanisms and data-driven methods for accurate and reliable streamflow prediction.

Study Configuration

• Spatial Scale: Four river basins in China with diverse hydrological characteristics.
• Temporal Scale: Daily streamflow simulation.

Methodology and Data

• Models used:
  • Xinanjiang (XAJ) model
  • Temporal Convolutional Network (TCN)
  • Gated Recurrent Unit (GRU)
  • Random Forest (RF)
• Data sources:
  • Multi-source data, including precipitation, temperature, humidity, and streamflow observations.

Main Results

• The XAJ-TCN-GRU model achieved high NSE values of 0.991, 0.971, 0.984, and 0.986 for the Wuding River, Chu River, Jianxi River, and Qingyi River, respectively.
• The model outperformed other benchmark models in terms of streamflow simulation, flood simulating, and interval simulation.

Contributions

• This study presents a robust hybrid integration framework that combines physical mechanisms with data-driven methods to enhance both accuracy and interpretability in streamflow prediction.
• The XAJ-TCN-GRU model demonstrates remarkable effectiveness and robustness in practical hydrological simulation scenarios, especially in flood prediction across diverse basins.

Funding

• This research was funded by the National Natural Science Foundation of China (Grant No. 52109215).

Citation

@article{Wang2026Integrating,
  author = {Wang, Zhaocai and Xu, Nannan and Song, Wei Guo and Zhang, Xingxing and Wu, Junhao and Chen, Xi},
  title = {Integrating Physical-Based Xinanjiang Model and Deep Learning for Interpretable Streamflow Simulation: A Multi-Source Data Fusion Approach across Diverse Chinese Basins},
  journal = {Hydrology and earth system sciences},
  year = {2026},
  doi = {10.5194/hess-30-5521-2026},
  url = {https://doi.org/10.5194/hess-30-5521-2026}
}