Li et al. (2026) Quantifying the uncertainty contribution in runoff projection and the time scale effects

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-01-31
• Authors: Zhanling Li, Yingtao Ye, Cong Xie, Xiaoyan Zhai
• DOI: 10.1016/j.jhydrol.2026.135067

Research Groups

• Key Laboratory of Groundwater Conservation of MWR, China University of Geosciences (Beijing), Beijing 100083, China
• School of Water Resources and Environment, China University of Geosciences (Beijing), Beijing 100083, China
• Ninghai County Water Conservancy Bureau, Ningbo 315600, China
• Research Center on Flood and Drought Disaster Reduction of the Ministry of Water Resources, China Institute of Water Resources and Hydropower Research, Beijing 100038, China

Short Summary

This study quantifies the uncertainty contributions of various factors (modeling chain and internal variability) to runoff projections and explores their time scale effects in the Upper Heihe River Basin (UHRB) and Upper Yalong River Basin (UYRB), China. It reveals that internal variability dominates short-term uncertainty (less than 20 years), while the modeling chain becomes the primary source for longer-term projections (over 35 years).

Objective

• To quantify the uncertainty contributions of different factors (General Circulation Models (GCMs), climate scenarios (SSPs), bias correction methods (BCs), forecast models (FMs), and internal variability of the hydrological system) to runoff projections.
• To explore how these uncertainty contributions respond to changes in hydrological forecasting periods, identifying their time scale effects.

Study Configuration

• Spatial Scale: Upper Heihe River Basin (UHRB) (10,009 km²) and Upper Yalong River Basin (UYRB) (32,925 km²) in China.
• Temporal Scale: Historical period: 1980–2014. Future projection period: 2026–2100. Forecasting periods analyzed range from 5 years to 75 years.

Methodology and Data

• Models used:
  • General Circulation Models (GCMs): 10 CMIP6 models (ACCESS-CM2, ACCESS-ESM1-5, CanESM5, CMCC-ESM2, MIROC6, MPI-ESM1-2-LR, MRI-ESM2-0, NorESM2-LM, NorESM2-MM, TaiESM).
  • Climate Scenarios: Shared Socioeconomic Pathways (SSP2-RCP4.5 and SSP5-RCP8.5).
  • Bias Correction Methods: Empirical Quantile Mapping (EQM) and Linear Scale (LS).
  • Forecast Models (FMs): Soil and Water Assessment Tool (SWAT), Water And Snow balance MODeling system-Evapotranspiration (WASMOD-E), Random Forest (RF).
  • Uncertainty Quantification: Time-series Analysis of Variance (ANOVA) using the QUALYPSO method.
• Data sources:
  • Historical Measured Data:
    • Digital Elevation Model (DEM) (90 m resolution, from gscloud.cn).
    • Land use data (1 km resolution, from resdc.cn).
    • Soil data (1 km resolution, from fao.org).
    • Daily meteorological data (precipitation, maximum/minimum temperatures, relative humidity, wind speed, sunshine hour) from gauge stations (1980–2014).
    • Daily flow data from hydrological stations (1980–2014).
  • GCM Data: Gridded daily meteorological data (precipitation, maximum/minimum temperatures, relative humidity, wind speed, short-wave radiation) for historical (1980–2014) and future (2026–2100) periods from 10 CMIP6 GCMs (Centre for Environmental Data Analysis, UK).

Main Results

• For short-term hydrological forecasting (less than 20 years), the internal variability of the system is the main source of uncertainty, explaining more than 50% of the total uncertainty in Qmean, Q10, and Q90 projections for both basins. At a 10-year scale, internal variability contributes approximately 75%.
• As the time scale increases, the contribution from internal variability gradually weakens, while that from the modeling chain strengthens.
• When the time scale reaches 35 years, the impacts of internal variability and the modeling chain on forecast results reach a relatively balanced state, and contributions tend to stabilize.
• For 75-year long-term hydrological forecasting, the modeling chain is the main source of uncertainty, explaining 80–87% of the total uncertainty in Qmean, Q10, and Q90 projections for both basins.
• Within the modeling chain for 75-year projections:
  • For the UYRB, the General Circulation Model (GCM) contributes most to the total uncertainty (37% for Qmean, 23% for Q10, 41% for Q90).
  • For the UHRB, the Forecast Model (FM) contributes most (50% for Qmean, 55% for Q10, 34% for Q90).
  • Climate scenarios (SSP) and bias correction methods (BC) contribute minorly, less than 5% and 3% respectively.
  • Interactions between factors contribute about 20% of the total uncertainty, with the interaction between GCM and FM being the largest (6–15%).
• The Random Forest (RF) model showed inherent limitations in extrapolation, leading to suppressed variance in projected runoff time series when future climate data exceeded historical training boundaries.

Contributions

• This study proposes a dynamic analysis framework for uncertainty attribution over time, moving beyond traditional static uncertainty decomposition.
• It identifies the main influencing factors of uncertainty across different forecasting periods (short-term, medium-term, long-term).
• It determines a critical time scale (approximately 35 years) at which the dominant mechanism affecting hydrological projection uncertainty shifts from internal variability to the modeling chain.
• The findings provide new insights for improving hydrological projection design and dynamically adjusting risk management strategies under climate change, emphasizing the need for flexible plans for short-term forecasts and multi-model ensemble averaging for medium- to long-term forecasts.

Funding

• National Natural Science Foundation of China (No. 42171047)
• National Natural Science Foundation of China (No. 41101038)
• Belt and Road Special Foundation of the State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering (2021nkms03)

Citation

@article{Li2026Quantifying,
  author = {Li, Zhanling and Ye, Yingtao and Xie, Cong and Zhai, Xiaoyan},
  title = {Quantifying the uncertainty contribution in runoff projection and the time scale effects},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2026.135067},
  url = {https://doi.org/10.1016/j.jhydrol.2026.135067}
}