Shahi (2026) Atmospheric drivers and modeling challenges of Mediterranean flood-inducing precipitation events: Insights from high-resolution ICON-NWP simulations

Identification

• Journal: Weather and Climate Extremes
• Year: 2026
• Date: 2026-09-22
• Authors: Namendra Kumar Shahi
• DOI: 10.1016/j.wace.2026.100957

Research Groups

• Institute of Meteorology and Climate Research – Troposphere Research (IMKTRO), Karlsruhe Institute of Technology (KIT), Eggenstein-Leopoldshafen, Germany

Short Summary

This study uses high-resolution ICON-NWP simulations to analyze the atmospheric drivers of two catastrophic Mediterranean heavy precipitation events (HPEs) in southern France, demonstrating the added value of convection-permitting models in representing localized precipitation while highlighting persistent challenges in accurately reproducing precipitation extremes.

Objective

• To provide a detailed, process-based analysis of the physical mechanisms governing two catastrophic flood-inducing heavy precipitation events (HPEs) in the northwestern Mediterranean region of southern France (12–13 November 1999 and 14–15 October 2018).
• To assess the modeling fidelity of these HPEs using a high-resolution, convection-permitting ICON modeling framework.
• To evaluate the performance of ERA5-driven high-resolution simulations in reproducing the intensity, spatial distribution, and temporal evolution of precipitation.
• To investigate the underlying physical mechanisms driving these events, with particular emphasis on the interactions between mesoscale convective systems and the large-scale synoptic circulation.

Study Configuration

• Spatial Scale:
  • Outer model domain (ICON-LAMR2B8): Approximately 10 km horizontal resolution, covering a large portion of western Europe and the adjacent Atlantic region.
  • Inner model domain (ICON-LAMR2B10): Approximately 2.5 km horizontal resolution, centered over Western Europe and the Mediterranean (southern France).
  • Vertical resolution: 65 levels for the 10 km domain (model top at 23.5 km), 60 levels for the 2.5 km domain (model top at 22 km).
• Temporal Scale:
  • Simulation periods: 12–13 November 1999 and 14–15 October 2018.
  • Spin-up time: 7 days for 10 km simulations, 3 days for 2.5 km nested simulations.
  • Integration time steps: 30 seconds for the 10 km domain, 10 seconds for the 2.5 km domain.
  • Data output: 3-hourly for driving data, hourly for ICON-LAM_R2B10 for temporal evolution analysis.

Methodology and Data

• Models used:
  • ICON (Icosahedral Nonhydrostatic) Numerical Weather Prediction (ICON-NWP) model (release 2024.07) in limited area mode (ICON-LAM), utilizing a two-step nesting approach.
  • Physics packages from the operational regional model COSMO and the ECMWF Integrated Forecast System (IFS).
  • Specific parameterizations: ecRad scheme for radiation, Probability Density Function (PDF) scheme for cloud cover, Single-Moment Cloud Microphysics Scheme, Lott and Miller scheme for sub-grid-scale orographic drag, wave dissipation at critical levels for non-orographic gravity-wave drag, prognostic turbulent kinetic energy (TKE) scheme for vertical diffusion and turbulent transport, tiled TERRA scheme for surface processes.
  • Convection parameterization: Mass-flux scheme (shallow and deep) for the 10 km domain; deep convection deactivated, shallow convection parameterized for the 2.5 km domain.
• Data sources:
  • Driving Data: ERA5 reanalysis (3-hourly initial and lateral boundary conditions).
  • Observational Data for Evaluation:
    • Météo-France station-based rainfall observations (Réseau d’Acquisition de Données et d’Observation Météorologiques Étendues - RADOME network).
    • E-OBS daily gridded land-only precipitation dataset (version 33.0e, 0.1° x 0.1° spatial resolution, approximately 11 km).
    • EMO-1 (European Meteorological Observations) gridded precipitation dataset (1 arc-minute spatial resolution, approximately 1.5 km).
  • Methodology: Process-based diagnostics (e.g., potential vorticity (PV, 1 PVU = 10⁻⁶ K m² kg⁻¹ s⁻¹), geopotential height, winds, vertical velocity, mean sea level pressure (hPa), vertically integrated moisture transport (VIMT) and its divergence (VIMFD)), Fractions Skill Score (FSS), Contiguous Rain Area (CRA) method, Quantile-Quantile (Q-Q) plots with upper-tail relative-bias heatmaps, point-to-point comparison.

Main Results

• Process-based understanding of atmospheric drivers:
  • 1999 Event (12–13 November): Characterized by cyclonic Rossby wave breaking (CWB) transforming a positively tilted PV streamer into a negatively tilted, "hook-shaped" configuration, resulting in a quasi-stationary upper-level cut-off low system. This sustained downstream moisture transport and deep-tropospheric ascent.
  • 2018 Event (14–15 October): Driven by the remnants of Hurricane Leslie phasing with the downstream flank of a high-amplitude Rossby wave breaking (RWB) streamer, establishing a quasi-stationary frontal system. This large-scale interaction dynamically altered the tropospheric structure and enriched the low-level inflow.
  • Common mechanisms for both events: Both events featured atmospheric river-like moisture plumes fueled by persistent southeasterly Mediterranean flow, with enhanced mid-tropospheric upward motion (negative vertical velocity in Pa s⁻¹) and low-level moisture convergence establishing favorable conditions for sustaining extreme precipitation.
• Model Performance (ICON-LAMR2B10 vs. coarser resolutions and observations):
  • Convection-permitting simulations (2.5 km resolution) demonstrate clear added value in representing localized precipitation intensity, spatial structure, and statistical distribution compared to coarser-resolution simulations (10 km and ERA5).
  • ICON-LAMR2B10 substantially reduced upper-tail precipitation biases relative to Météo-France station observations: approximately 25% to 35% for the 1999 event, and below 10% (with slight overestimation at extreme percentiles) for the 2018 event, outperforming E-OBS and EMO-1 for 2018. The coarser ICON-LAMR2B8 showed 45% to 55% bias.
  • Spatial verification (FSS): ICON-LAMR2B10 showed greater spatial agreement with EMO-1, with FSS values exceeding 0.65 for 1999 and 0.8 for 2018 at a 10 km neighborhood scale, approaching unity at approximately 60 km and 30 km, respectively.
  • Object-based verification (CRA): ICON-LAMR2B10 exhibited smaller displacement errors (e.g., approximately 33 km versus 43 km for 1999), reduced precipitation-volume underestimation (e.g., approximately -42% versus -55% for 1999), and lower root-mean-square (RMS) pattern errors compared to ICON-LAMR2B8.
• Persistent Challenges: Despite improvements, accurately reproducing the exact intensity and precise spatial location of observed precipitation extremes remains challenging, with displacement errors of approximately 27.1 km (1999) and 16.4 km (2018) for ICON-LAM_R2B10 compared to Météo-France stations. Gridded observational products (E-OBS, EMO-1) also show uncertainties and tend to underestimate local maxima due to spatial smoothing.

Contributions

• First application of a single, standardized, high-resolution (approximately 2.5 km) convection-permitting ICON modeling framework to analyze and compare two distinct catastrophic Mediterranean heavy precipitation events (1999 and 2018), providing a consistent basis for evaluation.
• Extends the understanding of these benchmark events by providing a broader synoptic-scale perspective on their atmospheric drivers, complementing previous studies focused on mesoscale and boundary-layer processes.
• Quantitatively demonstrates the clear added value of convection-permitting models (ICON-LAMR2B10) over coarser resolutions (ICON-LAMR2B8, ERA5) in representing the intensity, spatial distribution, and statistical properties of extreme precipitation in a complex topographical region.
• Highlights the persistent challenges in accurately reproducing the precise location and magnitude of precipitation extremes, even with convection-permitting resolutions, and underscores inherent uncertainties in gridded observational products.
• Advances process-based understanding by identifying common dynamical mechanisms (atmospheric river-like moisture plumes, deep-layer ascent, low-level moisture convergence) crucial for sustaining extreme precipitation in the Mediterranean, despite differing synoptic origins.

Funding

• Open access funding enabled and organized by Projekt DEAL.

Citation

@article{Shahi2026Atmospheric,
  author = {Shahi, Namendra Kumar},
  title = {Atmospheric drivers and modeling challenges of Mediterranean flood-inducing precipitation events: Insights from high-resolution ICON-NWP simulations},
  journal = {Weather and Climate Extremes},
  year = {2026},
  doi = {10.1016/j.wace.2026.100957},
  url = {https://doi.org/10.1016/j.wace.2026.100957}
}