Cui et al. (2026) Unraveling the long-term persistence of streamflow in China and its controlling factors

Identification

• Journal: Journal of Hydrology Regional Studies
• Year: 2026
• Date: 2026-04-02
• Authors: Tong Cui, Jan Adamowski, Masoud Zaerpour, Feng Tian, Yuping Han, Debao Lu, Jiazhong Zheng
• DOI: 10.1016/j.ejrh.2026.103397

Research Groups

• School of Hydraulic Engineering, Zhejiang University of Water Resources and Electric Power, Hangzhou, China
• Nanxun Innovation Institute, Zhejiang University of Water Resources and Electric Power, Hangzhou, China
• Faculty of Agricultural & Environmental Sciences, McGill University, Québec, Canada
• Department of Civil Engineering, Schulich School of Engineering, University of Calgary, Calgary, Canada
• Department of Hydraulic Engineering, State Key Laboratory of Hydroscience and Engineering, Tsinghua University, Beijing, China

Short Summary

This study conducted the first nationwide, spatially and seasonally resolved assessment of long-term persistence (LTP) in Chinese river streamflow using 45 years of monthly runoff data from 60 stations. It found a national annual mean Hurst coefficient of 0.710 with significant spatial and seasonal variability, primarily controlled by land cover (forest, soil texture), catchment area, and climatic factors, with their relative importance shifting seasonally.

Objective

• To quantify basin-scale long-term persistence (LTP) and its seasonal variability using 45 years of monthly runoff data from 60 stations across China and three well-established methods (Rescaled Range Analysis, Whittle Estimator, and Least Squares Variance).
• To identify the potential drivers of LTP by relating it to climate, catchment topography, size, soil, land use, and runoff attributes, employing both correlation analysis (Spearman’s correlation) and nonlinear machine learning approaches (random forests and SHapley Additive exPlanations).

Study Configuration

• Spatial Scale: Mainland China, covering 60 hydrological stations distributed across nine major river basins. Catchment areas range from 197 square kilometers to 1,699,338 square kilometers.
• Temporal Scale: 45 years (1956-2000) of monthly streamflow data. Seasonal analysis was performed for spring (March-May), summer (June-August), autumn (September-November), and winter (December-February).

Methodology and Data

• Models used:
  • Hurst coefficient estimation: Rescaled Range Analysis (R/S), Whittle Estimator (WE), Least Squares Variance (LSV).
  • Seasonality removal: Seasonal-Trend Decomposition based on Loess (STL).
  • Driver analysis: Spearman’s correlation, Random Forests (RF), SHapley Additive exPlanations (SHAP).
• Data sources:
  • Monthly streamflow data (1956-2000): Ministry of Water Resources of the People’s Republic of China.
  • Catchment boundaries, elevation, and slope: Multi-Error-Removed Improved-Terrain Digital Elevation Model (MERIT DEM) with a 3-arcsecond (~90 meters) resolution.
  • Monthly precipitation, temperature, and potential evapotranspiration (1956-2000 subset): National Tibetan Plateau Data Center at a 1-kilometer resolution.
  • Land use data (year 2000): Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences at a 1-kilometer resolution.
  • Soil data (Available Water Capacity, soil texture): Harmonized World Soil Database developed by the Food and Agriculture Organization of the United Nations at a 1-kilometer resolution.

Main Results

• The national annual mean Hurst coefficient (H) was 0.710 (average of three methods), with all 60 catchments exhibiting H > 0.5, indicating pronounced long-term persistence (LTP) in streamflow.
• Spatially, northern catchments generally showed stronger LTP than southern catchments. The Yellow River Basin exhibited the strongest LTP (mean H = 0.820), while the Pearl River Basin (0.662) and Southeast River Basin (0.611) displayed comparatively weaker LTP.
• Seasonally, winter exhibited the strongest LTP (mean H = 0.758), whereas spring and summer showed weaker persistence (mean H ≈ 0.650).
• Spearman’s correlation analysis revealed that catchment slope, precipitation, temperature, potential evapotranspiration, forest land, fine-textured soil, and specific mean discharge were negatively correlated with H. Conversely, catchment area, grassland, water bodies, unused land, and coarse- to medium-textured soils showed positive associations.
• Random Forests (RF) and SHapley Additive exPlanations (SHAP) analyses consistently identified forest land, medium- and fine-textured soils, catchment area, and key climatic factors as dominant controls of LTP, with their relative importance varying seasonally. Forest land, elevation, and specific mean discharge dominated in summer, while soil properties, precipitation, temperature, and potential evapotranspiration were most influential in spring, autumn, and winter.

Contributions

• Presents the first nationwide, spatially and seasonally resolved assessment of long-term persistence in Chinese river runoff.
• Integrates multi-method Hurst coefficient estimation with explainable machine learning (Random Forests and SHAP) to provide novel insights into streamflow predictability and catchment behavior.
• Offers a comprehensive understanding of the controls on streamflow persistence, including the identification of seasonal shifts in the dominant influencing factors.

Funding

• Nanxun Scholars Program for Young Scholars of ZJWEU (grant numbers RC2023021224)
• National Natural Science Foundation of China (grant numbers 52409046)
• China Scholarship Council (No. 202508330294)
• Joint Fund of Zhejiang Provincial Natural Science Foundation of China (grant numbers LGEZ25E090007)
• Zhejiang Key Laboratory of River-Lake Water Network Health Restoration
• Central Guidance Funds for Science and Technology Local Development Projects (grant numbers 2025ZY01091)

Citation

@article{Cui2026Unraveling,
  author = {Cui, Tong and Adamowski, Jan and Zaerpour, Masoud and Tian, Feng and Han, Yuping and Lu, Debao and Zheng, Jiazhong},
  title = {Unraveling the long-term persistence of streamflow in China and its controlling factors},
  journal = {Journal of Hydrology Regional Studies},
  year = {2026},
  doi = {10.1016/j.ejrh.2026.103397},
  url = {https://doi.org/10.1016/j.ejrh.2026.103397}
}