Dommange et al. (2026) Climatology of long-term extreme precipitation using rain gauges data over France

Identification

• Journal: Atmospheric Research
• Year: 2026
• Date: 2026-02-24
• Authors: Anthony Dommange, Elham Ghasemifar, Céline Planche, J. L. Baray
• DOI: 10.1016/j.atmosres.2026.108886

Research Groups

• Laboratoire de Météorologie Physique UMR 6016, Université Clermont Auvergne, CNRS INSU, F-63000 Clermont-Ferrand, France
• Institut Universitaire de France, Paris, France
• Observatoire de Physique du Globe de Clermont-Ferrand, UAR 833, Université Clermont Auvergne, CNRS INSU, F-63000 Clermont-Ferrand, France

Short Summary

This study presents a comprehensive analysis of extreme precipitation events (EPs) across France from 1950 to 2023 using 352 quality-controlled rain gauges, revealing significant intensification of extreme precipitation in southern regions, particularly the Cévennes, alongside a national decrease in the frequency of rainy days, and identifies large-scale atmospheric drivers.

Objective

• To provide a statistical analysis of the long-term evolution of the occurrence and intensity of extreme precipitation events (EPs) in France during 1950-2023 using rain gauge observations and climate model reanalysis.
• To describe the main features of precipitation extremes at both local and regional levels in France using various indices.
• To analyze the trends of EPs indices using Sen's slope estimator.
• To investigate the relationship of these indices with global climate indices (teleconnection patterns).
• To analyze widespread extreme events, their large-scale patterns, and return periods.

Study Configuration

• Spatial Scale: Metropolitan France (approximately 536,000 km²), with particular focus on the Massif Central and Cévennes regions.
• Temporal Scale: 1950–2023 (74 years).

Methodology and Data

• Models used:
  • Mann–Kendall trend test (non-parametric for monotonic trends)
  • Sen's slope estimator (non-parametric for trend magnitude)
  • Standard Normal Homogeneity Test (SNHT) (for homogeneity assessment)
  • Standardized Anomaly Test (Z-score) (for outlier detection)
  • Pearson correlation coefficient (for relationships between indices)
  • Multiple regression model (for influence of cycles and global indices)
  • Gumbel distribution (for return period estimation)
  • Generalized Extreme Value (GEV) distribution (for comparison in return period estimation)
• Data sources:
  • Rain gauges: Daily precipitation data from the Météo-France rain gauge network (352 stations, ≥99% temporal coverage, 1950-2023).
  • Reanalysis: ERA5 reanalysis product from ECMWF (total precipitation, mean sea level pressure, water vapor fluxes, 500 hPa geopotential height, specific humidity, wind components at various levels, 0.25° × 0.25° spatial resolution, 1950-2023).
  • Global Climate Indices:
    • El Niño–Southern Oscillation (ENSO) (Niño 3.4 region data from NOAA Physical Sciences Laboratory).
    • Arctic Oscillation (AO), North Atlantic Oscillation (NAO), East Atlantic Oscillation (EAO), Scandinavian pattern (SCA), East Atlantic–Western Russia (EAWR) (monthly indices from NOAA's Climate Prediction Center).

Main Results

• ERA5 vs. Rain Gauges: ERA5 systematically overestimated rainy days by 10-30 days and underestimated peak precipitation in mountainous regions (e.g., Cévennes: ERA5 ~1500 mm yr⁻¹ vs. observations >1800 mm yr⁻¹). Rain gauge data showed superior accuracy for detecting significant long-term trends.
• Extreme Precipitation Indices (1950-2023):
  • Maximum 1-day precipitation (RX1): Highest intensities in southern France (Cévennes, southern Massif Central) exceeding 130 mm day⁻¹. A statistically significant increasing trend was observed nationally (0.5 mm decade⁻¹), with more pronounced increases in Cévennes (>5 mm decade⁻¹) and Massif Central (1-2.5 mm decade⁻¹). RX1 peaked in autumn.
  • 99th percentile threshold (P99th): Highest intensities in Cévennes and southern Massif Central (>120 mm day⁻¹). No statistically significant national trend, but localized increases (0.5-3 mm decade⁻¹) in western Cévennes and Massif Central. P99th peaked in autumn.
  • Mean exceedance above the 99th percentile (ME99th): Largest exceedance values in Cévennes and Massif Central (16 stations >30 mm). Cévennes showed a statistically significant increase (0.14 mm yr⁻¹), while the national average showed a marginally significant increase (0.02 mm yr⁻¹). ME99th peaked in autumn.
  • Frequency of extreme rainy days (RP99th): Highest frequency in central, central-west, eastern, and northern France (>110 days yr⁻¹), lowest in southern regions (<50-60 days yr⁻¹). A statistically significant decreasing trend was observed nationally (-1.6 days decade⁻¹). RP99th peaked from July to December.
• Interannual Variability: Strongest positive RX1 anomalies in 1999, 1982, 2003. Strongest positive P99th anomalies in 1996, 1995, 1977. Strongest positive ME99th anomalies in 2003, 2021, 1982. Most anomalously wet years for rainy days were 1999, 1982, 1960.
• Role of Global Indices:
  • Six climate indices (ENSO, AO, NAO, EAO, SCA, EAWR) along with semiannual and annual cycles collectively explained up to 33% of mean precipitation variability, but their influence was weaker for extreme indices (e.g., 3.7% for RX1, 13% for P99th).
  • SCA showed the strongest positive correlation with mean precipitation (>0.3), while AO and EAWR showed the strongest negative correlations (frequently exceeding -0.3).
  • Physical mechanisms favorable for EPs in France occur during negative phases of AO, NAO, EAWR, and positive phases of ENSO, EAO, SCA, often linked to cyclonic circulation.
• Widespread Extreme Precipitation Events:
  • 13 widespread events (affecting ≥15% of stations exceeding P99th) were identified, all occurring in winter and autumn.
  • The December 2003 event (mean precipitation 104.67 mm) was classified as "severely abnormal" with a return period of approximately 21 years, driven by a prolonged Cévenol episode. Other events were within the normal range.
  • Synoptic analysis showed these events were driven by large-scale low mean sea level pressure over northern France, a corresponding mid-tropospheric trough (500 hPa geopotential height), and significant atmospheric rivers transporting moisture from the North Atlantic and western Mediterranean Sea (fluxes >400 kg m⁻¹ s⁻¹).

Contributions

This study provides a comprehensive, long-term (74 years) analysis of extreme precipitation events across the entirety of France, utilizing a rigorously quality-controlled rain gauge dataset. It offers novel insights by: - Systematically comparing ground observations with ERA5 reanalysis, highlighting the superior accuracy of rain gauges for extreme precipitation trend detection. - Characterizing the spatial and temporal variability and trends of multiple extreme precipitation indices at both local and regional scales, with a specific focus on hydrological hotspots like the Massif Central and Cévennes. - Quantifying the contribution of major global climate teleconnection patterns and seasonal cycles to precipitation variability, revealing their differential influence on mean versus extreme precipitation. - Identifying and analyzing widespread extreme precipitation events, including their return periods and associated large-scale atmospheric circulation patterns, such as atmospheric rivers. - Addressing inconsistencies in previous regional studies by using a consistent methodology and long-term data, thereby enhancing the empirical basis for flood risk management and climate adaptation strategies in France.

Funding

• Internship grant from Le Laboratoire de Météorologie Physique (LaMP) and the Pole de l'eau at Université Clermont Auvergne (Anthony Dommange).
• Make Our Planet Great Again (MOPGA) grant from the French Ministry of Europe and Foreign Affairs (Elham Ghasemifar).

Citation

@article{Dommange2026Climatology,
  author = {Dommange, Anthony and Ghasemifar, Elham and Planche, Céline and Baray, J. L.},
  title = {Climatology of long-term extreme precipitation using rain gauges data over France},
  journal = {Atmospheric Research},
  year = {2026},
  doi = {10.1016/j.atmosres.2026.108886},
  url = {https://doi.org/10.1016/j.atmosres.2026.108886}
}