Zheng et al. (2026) Cross-scale separation of climate and human impacts on runoff using a dual-step refined time-varying attribution model

Identification

• Journal: Environmental Modelling & Software
• Year: 2026
• Date: 2026-01-02
• Authors: Ziqin Zheng, Zengchuan Dong, Wenzhuo Wang, Jinyu Meng, Hao Ke, You Zhang
• DOI: 10.1016/j.envsoft.2026.106855

Research Groups

• College of Hydrology and Water Resources, Hohai University, Nanjing, Jiangsu, China
• Jiangsu Water Conservancy Digital Center, Nanjing, Jiangsu, China
• School of Atmospheric Sciences, Nanjing University of Information Science and Technology, Nanjing, China

Short Summary

This study develops a dual-step refined time-varying attribution model based on the Budyko framework to precisely separate the dynamic impacts of climate change and human activities on runoff across multiple spatio-temporal scales. The model significantly enhances the accuracy of runoff change separation, achieving improvements of 11.42 %–33.46 % at annual scales and 5.06 %–6.84 % at multi-year average scales.

Objective

• To develop and apply a dual-step refined time-varying attribution model to precisely separate the dynamic evolution of climate change and human activity impacts on runoff across multiple spatial and temporal scales.
• To contribute a general modeling framework for attribution analysis that enhances the accuracy of dynamic attribution and is applicable across various scales.

Study Configuration

• Spatial Scale: Lixia River Basin, Jiangsu Province, China (24,050 km²), further divided into three subregions: Lixia River Plain, Doubei District, and Dounan District.
• Temporal Scale: Annual scale (1986–2017) and multi-year average scales (before and after runoff mutation points), using an 11-year sliding window for smoothing hydrometeorological sequences. Data period: 1981–2022.

Methodology and Data

• Models used:
  • Dual-step refined time-varying attribution model based on the Budyko framework.
  • Step 1: Bidirectional Decomposition Method (BDM) and Runge-Kutta-improved Elasticity Coefficient Method (IECM) to revise traditional methods.
  • Step 2: Hierarchical Progressive Time-Varying Attribution Model (HBDM and HIECM) to separate intertwined effects.
  • Parametric Budyko equations: Fu Baopu formula, Mezentsev-Choudhury-Yang formula, and Zhang Lu empirical formula.
  • Stepwise multiple regression analysis with linear, exponential, and logarithmic functional forms for dynamic simulation of the Budyko parameter (ω).
• Data sources:
  • Meteorological data: Precipitation (meteorological stations), average temperature (meteorological stations), potential evapotranspiration (National Tibetan Plateau Science Data Center).
  • Runoff data: Jiangsu Water Conservancy Digital Center.
  • Human activity data: Land use (30 m annual land cover dataset CLCD), Normalized Vegetation Index (NDVI) (AVHRR GIMMS-3G+), soil moisture content (GLEAM4.1a).

Main Results

• The dual-step refined time-varying attribution model enhanced the accuracy of runoff change separation by 11.42 %–33.46 % at annual scales and by 5.06 %–6.84 % at multi-year average scales.
• The second step of the model demonstrated particularly significant improvements in separation accuracy: 10.24 %–33.08 % at annual scales and 3.90 %–5.64 % at multi-year average scales.
• At annual scales, climate change was the main cause of runoff changes in most periods, showing significant fluctuations correlated with precipitation. Human activities exhibited continuous impacts, with a shift from negative to positive and then negative impacts on runoff, consistent with anthropogenic modification intensity.
• At multi-year average scales, climate change was the dominant factor in runoff change for the Lixia River Basin, Doubei District, and Dounan District, leading to an increase of 18.18 mm (54.27 %), an increase of 20.92 mm (58.88 %), and a decrease of 26.97 mm (53.08 %), respectively. Human activities were the dominant factor in the Lixia River Plain, causing an increase of 28.66 mm (57.30 %) in runoff.
• The dynamic simulation of the time-varying parameter ω (ωch, considering both climate and human factors) achieved an average Root Mean Square Error (RMSE) of 0.08, relative error (RE) of 2.46 %, and Nash-Sutcliffe Efficiency (NSE) of 0.93, outperforming models considering only anthropogenic factors (ωh).
• The simulation of ω showed spatial heterogeneity, with higher accuracy at smaller spatial scales.

Contributions

• Proposes a novel dual-step refined time-varying attribution model that significantly enhances the accuracy of separating climate and human impacts on runoff by mitigating inherent errors and addressing non-linearity and order dependence.
• Introduces a hierarchical progressive structure to precisely characterize and separate the dynamic interweaving effects of climatic and anthropogenic drivers on watershed characteristics (e.g., Budyko parameter ω).
• Provides a flexible and accurate modeling strategy for capturing evolving system behavior through dynamically simulated parameters, rather than static calibration constants.
• Offers a generalizable modeling framework applicable to a wide range of environmental modeling problems, including water resources assessment and climate impact analysis, beyond the specific case study.

Funding

• Jiangsu Water Science and Technology Project of Water Resources Department of Jiangsu Province, China (WRD, China) (Grant No. 2023027)

Citation

@article{Zheng2026Crossscale,
  author = {Zheng, Ziqin and Dong, Zengchuan and Wang, Wenzhuo and Meng, Jinyu and Ke, Hao and Zhang, You},
  title = {Cross-scale separation of climate and human impacts on runoff using a dual-step refined time-varying attribution model},
  journal = {Environmental Modelling & Software},
  year = {2026},
  doi = {10.1016/j.envsoft.2026.106855},
  url = {https://doi.org/10.1016/j.envsoft.2026.106855}
}