Li et al. (2026) A coupled flood forecasting framework of HYDRUS-1D and Xinanjiang models based on the transfer of physically-based soil moisture fields

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-09-18
• Authors: Qiaoling Li, Yang Wang, Hongde Li, Xingwen Liu, Zhijia Li, Yang Xiao
• DOI: 10.1016/j.jhydrol.2026.136447

Research Groups

• College of Hydrology and Water Resources, Hohai University, Nanjing, China
• Wuling Power Corporation Ltd., Changsha, China

Short Summary

This study develops a coupled flood forecasting framework (XAJ-H1D) by integrating the physically-based HYDRUS-1D model with the conceptual Xinanjiang model to improve initial soil moisture estimation. The framework demonstrates enhanced accuracy in event-based flood forecasting by transferring physically-based soil moisture fields, significantly reducing errors in flood volume and peak discharge.

Objective

• To address the limitations of error accumulation and spatial homogenization in initial soil moisture estimation by the Xinanjiang (XAJ) model, this study aims to utilize the physically-based HYDRUS-1D (H1D) model to support a more physically consistent initialization of the XAJ model and propose a flood forecasting framework (XAJ-H1D) based on the transfer of physically-based soil moisture fields.

Study Configuration

• Spatial Scale: Near-dam area of the Wuqiangxi Reservoir, extending to basin-scale through 34 in-situ and virtual stations, with soil moisture averaged within sub-basins.
• Temporal Scale: 8 flood events during 2023 and 2024, with particular focus on the early flood season.

Methodology and Data

• Models used: HYDRUS-1D (H1D), Xinanjiang (XAJ) model.
• Data sources: Reliable soil moisture observations from a monitoring network after sensor calibration, used for H1D calibration.

Main Results

• The optimized HYDRUS-1D model accurately captured point-scale soil moisture dynamics with a mean root mean square error (RMSE) of 0.025 m³⋅m⁻³.
• The XAJ-H1D framework reduced the mean relative error of flood volume from 12.99% to 5.27% across 8 flood events.
• The mean relative error of peak discharge was reduced from 17.35% to 9.00%.
• The mean Nash-Sutcliffe efficiency (NSE) increased from 0.53 to 0.75 compared with conventional XAJ initialization.
• More significant improvements were observed during the early flood season or under relatively dry antecedent conditions.

Contributions

• Proposes a novel coupled flood forecasting framework (XAJ-H1D) that integrates a physically-based soil moisture model (HYDRUS-1D) with a conceptual hydrological model (Xinanjiang).
• Establishes a water balance-based mechanism for transferring physically-based soil moisture fields to improve the initialization of conceptual hydrological models.
• Utilizes a virtual–actual integration strategy to extend point-scale physically-based soil moisture simulations to a basin scale.
• Demonstrates significant improvements in flood forecasting accuracy (flood volume, peak discharge, and NSE) by mitigating initialization bias in conceptual models.

Funding

Not specified in the provided text.

Citation

@article{Li2026coupled,
  author = {Li, Qiaoling and Wang, Yang and Li, Hongde and Liu, Xingwen and Li, Zhijia and Xiao, Yang},
  title = {A coupled flood forecasting framework of HYDRUS-1D and Xinanjiang models based on the transfer of physically-based soil moisture fields},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2026.136447},
  url = {https://doi.org/10.1016/j.jhydrol.2026.136447}
}