Hani et al. (2025) Compound hydrological and thermal extremes: A nonstationary risk modeling approach for riverine ecosystems

Identification

• Journal: Journal of Environmental Management
• Year: 2025
• Date: 2025-10-11
• Authors: Ilias Hani, Taha B. M. J. Ouarda, André St‐Hilaire
• DOI: 10.1016/j.jenvman.2025.127613

Research Groups

• Canada Research Chair in Statistical Hydro-Climatology, Institut National de La Recherche Scientifique, Centre Eau Terre Environnement (INRS-ETE), Qu´ebec City, QC, Canada
• Canadian Rivers Institute, University of New Brunswick (UNB), Fredericton, NB, Canada

Short Summary

This study developed a nonstationary multivariate risk modeling framework using dynamic additive copulas to assess the joint behavior of extreme summer river water temperature (Tw) and concurrent low flow (Q) in six Atlantic salmon rivers in eastern Canada. The proposed joint nonstationary model (JNS) systematically outperformed alternative models, revealing that temporal trends significantly increased Tw extremes and teleconnections (SOI, NAO) were dominant drivers of variability, leading to elevated compound risks.

Objective

• Does accounting simultaneously for nonstationarity and interdependence improve model performance and risk estimation versus common alternatives?
• How do the temporal trend and teleconnections shape the magnitude and variability of compound Tw–Q extremes?
• Which covariates most strongly control site-specific risk structures?

Study Configuration

• Spatial Scale: Six unregulated Atlantic salmon rivers in eastern Canada, with catchment sizes ranging from 48 km2 to 18,963 km2.
• Temporal Scale: Simulated daily data covering the period 1979–2020.

Methodology and Data

• Models used:
  • Dynamic additive copula approach (rotated Archimedean copulas: Gumbel, Joe; Frank; Elliptical Gaussian Copula)
  • Generalized Additive Models for Location, Scale, and Shape (GAMLSS) framework
  • CEQUEAU (deterministic semi-distributed hydro-thermal model)
  • Univariate Mann-Kendall test (MK)
  • Multivariate trend tests: Covariance Inversion Test (CIT), Covariance Eigenvalue Test (CET), Multivariate Dependence Test (MDT)
  • L-moment ratio diagrams (L-MRD)
  • Parametric marginal distributions: Gumbel (EV), Weibull (W), Gamma (G), Lognormal (LN), Normal (N), Logistic (LO)
• Data sources:
  • Simulated daily river water temperature (Tw) and streamflow (Q) series from the CEQUEAU model (1979–2020).
  • Annual mean values of large-scale climate oscillation indices (teleconnections): North Atlantic Oscillation (NAO), Arctic Oscillation (AO), Southern Oscillation Index (SOI), Pacific Decadal Oscillation (PDO), and Pacific North American pattern (PNA) from the NOAA Physical Sciences Laboratory database.

Main Results

• A significant negative dependence was found between extreme summer river water temperature (Tw) and concurrent low flow (Q) across all rivers, with Pearson's ρp, Spearman's ρs, and Kendall's τK ranging from -0.63 to -0.4, -0.62 to -0.37, and -0.48 to -0.27, respectively. Rotated Archimedean copulas (Gumbel/Joe, 90°/270°) best captured this relationship.
• The proposed Joint Nonstationary (JNS) model systematically outperformed the univariate nonstationary (UNS) and joint stationary (JS) benchmark models across all study sites, indicated by lower AIC and BIC values. UNS systematically underestimated AND-case joint probabilities and conditional probabilities by up to 25% compared to JNS.
• Temporal trends significantly increased Tw extremes in 4 out of 6 rivers.
• Teleconnections were identified as dominant drivers of variability: negative phases of the Southern Oscillation Index (SOI, El Niño conditions) increased the variability of Tw extremes, while negative phases of the North Atlantic Oscillation Index (NAO) increased the variability of low-flow events.
• Positive phases of SOI (La Niña conditions) and NAO were associated with elevated joint and conditional exceedance probabilities, with increases of up to 66% in the Restigouche River and 45% in the Highland River.
• L-moment ratio analysis revealed shifts in the best-fit distributions for both Tw and Q extremes between early (1980–2000) and later (2001–2020) periods, indicating evolving event characteristics under climate change and variability.
• Probabilities of compound events (Tw exceeding 23 °C under low flow conditions) now exceed 70% in some rivers (e.g., Restigouche), highlighting escalating ecological risks for cold-water species.

Contributions

• Developed a novel nonstationary multivariate risk modeling framework using dynamic additive copulas to characterize the joint behavior of extreme summer river water temperature and concurrent low flow.
• Explicitly incorporated both climate change (temporal trends) and climate variability (teleconnection indices like NAO and SOI) as time-dependent covariates to model evolving marginal distributions and their interdependencies.
• Demonstrated the superior performance of the joint nonstationary model (JNS) over univariate nonstationary (UNS) and joint stationary (JS) alternatives for compound risk estimation in riverine ecosystems.
• Quantified the distinct impacts of long-term warming trends and oscillatory climate patterns on the magnitude and variability of compound hydro-thermal extremes in Atlantic salmon rivers.
• Provided a predictive framework for anticipating compound risks and informing adaptive management strategies for thermally sensitive aquatic habitats under ongoing climate change and variability.

Funding

• Mathematics of Information Technology and Complex Systems (MITACS) [grant number FR77680]
• Atlantic Salmon Research Joint Venture (ASRJV)
• Foundation for the Conservation of Atlantic Salmon (FCAS) [grant number 151560]

Citation

@article{Hani2025Compound,
  author = {Hani, Ilias and Ouarda, Taha B. M. J. and St‐Hilaire, André},
  title = {Compound hydrological and thermal extremes: A nonstationary risk modeling approach for riverine ecosystems},
  journal = {Journal of Environmental Management},
  year = {2025},
  doi = {10.1016/j.jenvman.2025.127613},
  url = {https://doi.org/10.1016/j.jenvman.2025.127613}
}