Gonzalez-Mora et al. (2025) A climate-informed statistical framework to indirectly estimate trends in future seasonal high flows in snow-dominated watersheds using short-term climate variability indices

Identification

• Journal: Journal of Hydrology
• Year: 2025
• Date: 2025-10-21
• Authors: Andrés F. Gonzalez-Mora, Étienne Foulon, Alain N. Rousseau
• DOI: 10.1016/j.jhydrol.2025.134441

Research Groups

• Institut national de la recherche scientifique (INRS). Centre Eau Terre Environnement (ETE), Québec, Canada

Short Summary

This study developed a climate-informed statistical framework to indirectly estimate future seasonal high flow trends in snow-dominated watersheds using short-term climate variability indices (SCIs). It found that future high flow variability can be anticipated using highly correlated SCIs, with a single SCI explaining at least 50% of the variability.

Objective

• To use a climate-informed statistical framework to indirectly estimate the temporal variability of seasonal high flow indices (HFI) using a set of short-term climate variability indices (SCI) characterizing likely causative mechanisms over different aggregated look-back periods.
• To prove that the temporal variability of projected seasonal high flows (maximum stream flows) can be indirectly estimated through the temporal analysis of relevant short-term climate variability indices.

Study Configuration

• Spatial Scale: Two snow-dominated watersheds in Southern Quebec, Canada: Bécancour River watershed (2597 km²) and Yamaska River watershed (4843 km²).
• Temporal Scale:
  • Climate simulations: 1955 to 2100.
  • Reference periods for bias correction: 1971–1994 and 1997–2002.
  • Streamflow simulations: January 1st, 1957 to December 30th, 2099.
  • Analysis periods: 1997–2022 (baseline), 2023–2048 (near-term), 2049–2074 (mid-term), and 2075–2099 (long-term).
  • SCI look-back periods: 1 to 5 days, 1 to 3 weeks, 1 to 6 months, and 8, 10, 12, and 24 months.

Methodology and Data

• Models used:
  • Hydrological model: HYDROTEL (semi-distributed, physically-based).
  • GIS software for watershed delineation: PHYSITEL.
  • Climate models: Ensemble of 138 climate simulations from CMIP5 (Coupled Model Intercomparison Project Phase 5), including Climex CRCM5-LE, CMIP5 GCMs, CORDEX-NA RCMs, and Ouranos CRCM5, under RCP 4.5 and 8.5 scenarios.
  • Optimization algorithm for calibration: Asynchronous Parallel Pareto Archived Dynamically Dimensioned Search (ParaPADDS) multi-objective search algorithm (OSTRICH v22.01.04).
  • Statistical tests: Pearson and Spearman correlation coefficients, partial correlations, Mann–Kendall (MK) trend test, non-parametric bootstrap resampling (1000 repetitions).
  • Bias correction methods: Quantile mapping for climate data, Daily Bias Correction (DBC) for streamflow.
• Data sources:
  • Climate data: Ensemble of 138 climate simulations (GCMs/RCMs) under RCP 4.5 and 8.5 scenarios, bias-corrected using a 10-km reference grid.
  • Hydrological data: Daily streamflow records from the Quebec Hydrological Expertise Centre (CEHQ).
  • Calibration meteorological data: Daily Surface Weather and Climatological Summaries Data (Daymet V4 R1) at 1-km grid, upscaled to 5-km.
  • Physiographic data, soil types, river network features (for PHYSITEL).
  • Short-term climate variability indices (SCIs): 31 indices categorized into precipitation-based (SCIPR), temperature-based (SCIT), combined precipitation/temperature (SCIPR+T), and drought/flood (SCIDF).
  • High flow indices (HFIs): Snow-free maximum stream flow (Qmaxsf) and snow-cover maximum stream flow (Qmaxsc).

Main Results

• The HYDROTEL hydrological model showed satisfactory calibration performance (KGE and R² > 0.77, NSE > 0.77, PBIAS < 10%), accurately representing observed streamflow regimes and high flows. Bias correction successfully matched observed streamflow distributions.
• For the snow-free season, a single SCI could explain at least 50% of the maximum stream flow (Qmaxsf) variability. Key SCIs included cumulative total precipitation (cPRCP) over 3-4 days, climatic demands (CD) over 5 days or 2 weeks, and the Effective Drought Index (EDICD) over 150-180 days.
• For the snow-cover season, a single SCI explained approximately 40% of the maximum stream flow (Qmax_sc) variability. Important SCIs were cumulative snowfall (cSnow) over 10 months and snowmelt and rainfall amounts (CSR) over 5 days.
• The Effective Drought Index (EDI) demonstrated strong correlations with summer/fall peak flows, suggesting its utility beyond traditional low flow monitoring.
• Partial correlation analysis indicated that time did not significantly influence the relationships between SCIs and HFIs, with changes in explained variability being less than 1% compared to simple correlations.
• Strong SCI-HFI correlations observed during the baseline period largely persisted across near-, mid-, and long-term future horizons under both RCP 4.5 and 8.5, demonstrating transferability.
• Under RCP 8.5, the correlation between Qmaxsc and cumulative snow decreased over time, while correlations with total precipitation or EDIPRCP strengthened, reflecting a climate-driven shift from snowfall to rainfall.
• Significant positive trends in SCIs (e.g., cPRCP, EDICD) and HFIs (Qmaxsf) were predominantly observed for snow-free conditions, particularly under RCP 8.5.
• Significant negative trends in SCIs (e.g., cSnow, TBZ, CSR) and HFIs (Qmax_sc) were primarily found for snow-cover conditions, especially under RCP 8.5, aligning with projected temperature increases and reduced snow accumulation.
• The degree of agreement between trends in highly correlated SCIs and HFIs was higher under the more extreme RCP 8.5 scenario, indicating an intensification of the streamflow regime due to greater climate disturbances.

Contributions

• Development and validation of a climate-informed statistical framework for indirectly estimating future seasonal high flow trends, offering an alternative to complex hydrological modeling, especially in data-limited regions.
• Identification of specific short-term climate variability indices (SCIs) and their optimal aggregated look-back periods that are strongly correlated with seasonal high flows in snow-dominated watersheds.
• Demonstration of the Effective Drought Index (EDI)'s effectiveness as a proxy for assessing high flow variability, expanding its application beyond low flow monitoring.
• Confirmation of the transferability and robustness of SCI-HFI relationships across different future climate horizons and scenarios.
• Provision of insights into the evolving drivers of high flows under climate change, highlighting shifts in the relative importance of snow-related versus rainfall-related mechanisms.

Funding

• Natural Sciences and Engineering Research Council of Canada Discovery Grant Program (RGPIN/06757-2019, A.N. Rousseau, principal investigator).
• cQ2 and INFO-Crue projects by Ouranos.

Citation

@article{GonzalezMora2025climateinformed,
  author = {Gonzalez-Mora, Andrés F. and Foulon, Étienne and Rousseau, Alain N.},
  title = {A climate-informed statistical framework to indirectly estimate trends in future seasonal high flows in snow-dominated watersheds using short-term climate variability indices},
  journal = {Journal of Hydrology},
  year = {2025},
  doi = {10.1016/j.jhydrol.2025.134441},
  url = {https://doi.org/10.1016/j.jhydrol.2025.134441}
}