Zhao et al. (2026) Enhancing Flood Prediction in Mountainous Watersheds across Diverse Climates Using Spatiotemporal Variable Source-Mixed Runoff Model

Identification

• Journal: Water Resources Management
• Year: 2026
• Date: 2026-01-01
• Authors: Xuantao Zhao, Aidi Huo, Changjun Liu, Xia Jia, Lei Wen, Qi Liu
• DOI: 10.1007/s11269-025-04445-7

Research Groups

• School of Water and Environment, Chang’an University, Xi’an, China
• Key Laboratory of Subsurface Hydrology and Ecological Effects in Arid Region of the Ministry of Education, Chang’an University, Xi’an, China
• Research Center on Flood and Drought Disaster Reduction, Institute of Water Resources and Hydropower Research, Beijing, China
• Guizhou Hydraulic Research Institute, Guiyang, China

Short Summary

This study introduces and evaluates a novel Spatiotemporal Variable Source-Mixed Runoff (SVSMR) model for flood prediction in mountainous watersheds across diverse climates, demonstrating its superior accuracy and cross-climatic adaptability compared to five established hydrological models by effectively representing climate-dependent runoff generation mechanisms.

Objective

• To validate the SVSMR model's applicability across diverse climatic regions and highlight its advantages in mountain flood forecasting.
• To quantify the SVSMR model's improvements in simulation accuracy and forecasting success rates through a multi-model comparison.
• To identify the dominant mechanisms driving mountain flood formation under different climatic conditions.

Study Configuration

• Spatial Scale: 15 mountainous watersheds in China, with drainage areas ranging from 21.3 square kilometers to 427.1 square kilometers (average 218.4 square kilometers). Watersheds are categorized into semi-arid/semi-humid (8) and humid (7) climatic zones. The minimum Hydrological Response Unit (HRU) area was set to 10 square kilometers, with an average HRU area of 14.8 square kilometers.
• Temporal Scale: An average of 27 years of observational records per watershed. Daily precipitation and temperature data were collected for non-storm periods, while hourly precipitation and discharge data were obtained for storm periods. Model simulations were performed with an hourly time step, with 15-day hourly timescales completing within 5 seconds.

Methodology and Data

• Models used:
  • Spatiotemporal Variable Source-Mixed Runoff (SVSMR) model (novel, semi-distributed)
  • Antecedent Precipitation Index (API) (lumped, empirical)
  • Xinanjiang (XAJ) model (lumped, conceptual)
  • Dahuofang (DHF) model (lumped, conceptual)
  • TOPMODEL (semi-distributed)
  • HEC-HMS (semi-distributed)
• Data sources:
  • Hydrometeorological data: China Meteorological Data Service Centre (56 rain gauges, 9 county-level stations), China Institute of Water Resources and Hydropower Research (storm-specific datasets).
  • Underlying surface data: Digital Elevation Model (DEM) at 30-meter spatial resolution (Chinese Academy of Sciences Geospatial Data Cloud), soil texture data from China Soil and Terrain Database (SOTER v1.0) at 1:1 million scale, and land use classifications from the 2009 Global Land Cover map at 300-meter resolution.

Main Results

• The SVSMR model consistently demonstrated superior overall flood simulation performance across all four evaluation metrics (PBIAS, RMSE, r, and NSE) compared to the five established hydrological models during both calibration and validation periods.
• SVSMR exhibited robust cross-climatic adaptability, achieving 'Good' agreement (NSE > 0.5, EQP < 20%, ETP ≤ 1 hour) between observations and simulations for 76.6% of flood events in semi-arid/semi-humid regions and 89.0% in humid regions.
• Parameter sensitivity analysis revealed distinct regional patterns: routing and surface runoff parameters were most influential in semi-arid/semi-humid regions, while interflow parameters (accounting for 83.1% of total sensitivity) dominated in humid regions.
• Runoff component analysis indicated climate-dependent mechanisms: infiltration-excess runoff prevailed in semi-arid/semi-humid watersheds, whereas saturation-excess runoff coupled with rapid interflow (contributing 40–70%) dominated in humid watersheds.
• The SVSMR model demonstrated high computational efficiency, completing 15-day hourly simulations within 5 seconds (average 2.3 seconds for 15 watersheds), making it suitable for real-time operational flood forecasting.

Contributions

• Development and comprehensive validation of a novel Spatiotemporal Variable Source-Mixed Runoff (SVSMR) model that significantly enhances flood prediction accuracy and cross-climatic adaptability in mountainous watersheds.
• Quantitative evidence of the SVSMR model's superior performance over five widely-used hydrological models across diverse climatic conditions, addressing a critical limitation of conventional models.
• Identification and quantification of climate-dependent dominant runoff generation mechanisms and corresponding parameter sensitivities, providing mechanistic insights for improved hydrological modeling.
• Advancement of operational flood forecasting capabilities through a computationally efficient model that requires minimal underlying surface datasets and incorporates expert knowledge for parameter initialization, particularly beneficial for data-scarce regions.

Funding

• National Key R&D Program of China (Grant No. 2023YFC3006705)
• National Natural Science Foundation of China (Grant No. 42261144749, 42377158)
• National Foreign Expert Individual Human Project (Category H) (H20240400)

Citation

@article{Zhao2026Enhancing,
  author = {Zhao, Xuantao and Huo, Aidi and Liu, Changjun and Jia, Xia and Wen, Lei and Liu, Qi},
  title = {Enhancing Flood Prediction in Mountainous Watersheds across Diverse Climates Using Spatiotemporal Variable Source-Mixed Runoff Model},
  journal = {Water Resources Management},
  year = {2026},
  doi = {10.1007/s11269-025-04445-7},
  url = {https://doi.org/10.1007/s11269-025-04445-7}
}