Tramblay et al. (2026) Regional process-based analysis of trends in flood peaks and volumes using hourly data

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-06-29
• Authors: Yves Tramblay, Patrick Arnaud, Pierre Javelle
• DOI: 10.1016/j.jhydrol.2026.135949

Research Groups

• Espace-Dev, Univ. Montpellier, IRD, Montpellier, France
• RECOVER, INRAE, Université D’aix-Marseille, Aix-en-Provence, France

Short Summary

This study developed a process-based regional framework using hourly data for 829 French catchments to analyze trends in flood peaks and volumes across different flood types, revealing that flood volumes exhibit more frequent and stronger increasing trends, especially for flash floods in mountainous and Mediterranean regions, compared to flood peaks.

Objective

• To investigate contrasting trends in flood peaks and volumes across different flood types and hydro-climatic regions in France using high-resolution hourly data, addressing the limitations of daily discharge data in capturing short-duration events and disentangling trends in various flood-generation mechanisms.

Study Configuration

• Spatial Scale: 829 small to medium-sized catchments across France, each with an area less than 500 km². Catchments were grouped into four homogeneous hydro-climatic regions.
• Temporal Scale: Hourly discharge data, radar rainfall, and reanalysis-based soil moisture data covering the period between 1997 and 2024 (average of 26 years of data per basin, minimum 15 years).

Methodology and Data

• Models used:
  • Peaks-over-threshold (POT) approach for flood event extraction.
  • Lyne and Hollick recursive digital filter for baseflow separation.
  • Ascending hierarchical classification with Principal Component Analysis (PCA) for regionalization, guided by Silhouette Index and Gap Statistic.
  • DBSCAN algorithm for spatio-temporal declustering of flood events.
  • Quantile regression for trend detection in median (τ = 0.5) and extreme (τ = 0.90) flood quantiles.
  • Cluster bootstrap with Bias-Corrected and Accelerated (BCa) method for uncertainty assessment.
• Data sources:
  • Hourly instantaneous discharge data from hydro.eaufrance.fr.
  • Hourly precipitation data from COMEPHORE reanalysis (1 km spatial resolution).
  • Antecedent soil moisture (Soil Wetness Index, SWI) from SAFRAN-ISBA-MODCOU (SIM) reanalysis (8 km spatial resolution, daily time step).
  • Catchment physiographic and climatic characteristics (e.g., altitude, Flashiness Index, mean annual rainfall, mean annual temperature, response time, catchment area).

Main Results

• Significant contrasts were observed in flood trends between different regions and flood types.
• Few statistically significant trends were detected for flood peaks across the study area.
• Flood volumes exhibited more frequent and pronounced significant increasing trends compared to flood peaks.
• The strongest increases in flood volumes were found for flash floods in mountainous and Mediterranean regions, with extreme events (τ = 0.90) showing increases of nearly 50% in mountainous areas.
• For slow-onset floods, significant positive volume trends were observed in temperate regions, correlating with increases in both rainfall totals and duration.
• Flood volumes consistently demonstrated a higher sensitivity to change than peak flows across all regions and flood types.
• At the national scale, increases in flood volume were primarily driven by longer flood durations, which in turn were linked to increases in the duration of rainfall events, rather than increases in rainfall amounts or intensity.
• Four distinct hydro-climatic regions were identified: Temperate Large, Temperate Small, Mediterranean, and Mountainous, each with unique hydrological regimes.
• The average spatial extent of clustered flood events ranged from 100-150 km in Atlantic and continental regions (1 and 2) to 50-70 km in Mediterranean and mountainous regions (3 and 4).

Contributions

• This study provides one of the first large-scale assessments of flood changes across different flood-generating processes using high-resolution hourly observations, revealing heterogeneity not captured by daily or aggregated analyses.
• It demonstrates that flood trend detection is critically dependent on temporal resolution, flood type, and regional context.
• The research highlights that volume-based metrics offer complementary and often stronger signals for assessing evolving flood hazards compared to traditional peak-based indicators.
• It provides new evidence on the evolution of flash flood events, underscoring the value of high-frequency hydrological observations for understanding changing flood risks.
• The findings advocate for a shift beyond traditional daily peak analyses, emphasizing the importance of integrating high-frequency data and mechanistic understanding into future flood risk studies, infrastructure design, and water resource management.

Funding

Not explicitly mentioned in the provided text.

Citation

@article{Tramblay2026Regional,
  author = {Tramblay, Yves and Arnaud, Patrick and Javelle, Pierre},
  title = {Regional process-based analysis of trends in flood peaks and volumes using hourly data},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2026.135949},
  url = {https://doi.org/10.1016/j.jhydrol.2026.135949}
}