Godet et al. (2026) Quantifying the added value of impact-based warnings for flash flood monitoring using innovative multi-source impact data

Identification

• Journal: International Journal of Disaster Risk Reduction
• Year: 2026
• Date: 2026-02-16
• Authors: Juliette Godet, Eric Gaume, Pierre Javelle, Thomas Dias, P Nicolle, Olivier Payrastre
• DOI: 10.1016/j.ijdrr.2026.106058

Research Groups

• GERS-LEE, Univ. Gustave Eiffel, IFSTTAR, Bouguenais, France
• RECOVER, INRAE, Université d’Aix-Marseille, Aix-en-Provence, France

Short Summary

This study quantifies the added value of impact-based warnings (IBW) over traditional hazard-based warnings (HBW) for flash flood monitoring in the French Mediterranean region. Using multi-source impact data over a 13-year period, it demonstrates that IBW significantly reduce false alarms and improve the prioritization of affected areas, especially at finer spatial scales.

Objective

• To quantify the added value of impact-based warnings (IBW) compared to hazard-based warnings (HBW) for flash floods, specifically in terms of identifying and prioritizing affected areas and reducing false alarms.

Study Configuration

• Spatial Scale: French Mediterranean region (61 main river basins, approximately 42,000 square kilometers drained area, 20,000 kilometers of river network). Analysis conducted at two resolutions: river reach scale (mean sub-catchment area of 4 square kilometers) and municipality scale (mean municipality area of 20 square kilometers).
• Temporal Scale: Continuous 13-year period (2010–2022) for validation, with nowcasts simulated from September 2009 to September 2023.

Methodology and Data

• Models used:
  • Hazard-Based Warning (HBW) model: Vigicrues Flash monitoring system (reservoir-based rainfall-runoff model, 1 square kilometer resolution, Gumbel fitting for discharge return periods, SHYREG method for larger quantiles).
  • Impact-Based Warning (IBW) model: HBW coupled with a flood library (eight flood maps for return periods from 2 to 1000 years, computed using a 2D Shallow Water Equations model at 5 meter resolution) and a building exposure layer (BD-TOPO 2020, building considered flooded if water level exceeds 15 centimeters).
  • Flood-Based (FB) Model: Outputs the real-time percentage of the forecasted flooded area within the 1000-year flood envelope.
  • Hybrid Models: Combine both HB and IB criteria (e.g., simultaneous exceedance of discharge return period and number of affected buildings thresholds).
• Data sources:
  • Rainfall input: ANTILOPE reanalysis rainfall product.
  • Impact validation data (multi-source):
    • Legislative decrees attesting to natural disasters (CatNat decrees from GASPAR database, municipality scale).
    • Insurance claim records (CCR database, river reach scale).
    • Fire and rescue service operation logs (SDIS30 and SDIS06 databases, point data aggregated to river reach and municipality scales).
  • Ancillary data: BD-TOPO building shapefile (2020), 1000-year flood extent map.

Main Results

• Impact-based warnings (IBW) demonstrate a clear added value over hazard-based warnings (HBW), particularly at finer spatial scales.
• At the river reach scale, IBW reduced the number of false alarms by a factor of 2 to 3 compared to HBW, for a similar hit rate (Probability of Detection of 60% to 70%) when validated with insurance and fire and rescue service data.
• The reduction in false alarms by IBW is more pronounced at the river reach scale (ratio of 2 to 2.7) than at the municipal scale (ratio of 1.3 to 1.7).
• Validation using CatNat decrees did not show the same reduction in false alarms for IBW, as the IB model failed to identify flooded buildings for nearly 40% of CatNat decrees, likely due to CatNat data including surface flooding impacts not accounted for by the river-focused IBW.
• Fire and rescue service operations data yielded low hit rates (below 0.30) for both HBW and IBW, attributed to the data capturing a broader range of impacts, many of which are outside the scope of river flash flood warnings.
• The Flood-Based (FB) model, which relies solely on flood extent, performed significantly worse than HBW and IBW, confirming the importance of incorporating exposure information.
• For insurance data, hybrid HB-IB models achieved the best Critical Success Index (CSI) scores, effectively reducing false alarms without severely impacting the number of hits.
• Event-scale analysis confirmed that the IB model consistently outperforms the HB model in both detecting and ranking impacted areas, with the majority of events showing higher CSI scores for IBW.

Contributions

• Provides new quantitative evidence of the operational potential and added value of impact-based warnings (IBW) for flash flood monitoring.
• Offers the first systematic, long-term (13 years), and large-scale (French Mediterranean region) comparison of hazard-based and impact-based warning models using diverse impact data.
• Demonstrates that IBW significantly reduce false alarms and improve the spatial prioritization of affected areas, particularly at high spatial resolutions.
• Highlights critical limitations and inconsistencies in existing impact data sources, underscoring the need for systematic and harmonized post-event impact data collection for robust IBW evaluation.

Funding

The authors acknowledge support and valuable datasets provided by the following organizations: * Université de Montpellier * French Ministry of the Interior and Overseas, General Directorate for Civil Security and Crisis Management (DGSCGC) * Local fire and rescue services (SDIS), particularly SDIS30 (Gard) and SDIS06 (Alpes-Maritimes) * Road authorities within the Departmental Councils, especially the Departmental Councils of Gard and Var (CD30, CD83) * Orange telecommunications team (participated in the ANR DISCRET project) * Visov organization * Roya Reconstruction Mission (MIRV) * Departmental Directorate for Territories and the Sea of Alpes-Maritimes (DDTM06) * French Center for Studies and Expertise on Risks, Environment, Mobility and Urban Planning (CEREMA) for Murex database access * CCR (Caisse Centrale de Réassurance)

Citation

@article{Godet2026Quantifying,
  author = {Godet, Juliette and Gaume, Eric and Javelle, Pierre and Dias, Thomas and Nicolle, P and Payrastre, Olivier},
  title = {Quantifying the added value of impact-based warnings for flash flood monitoring using innovative multi-source impact data},
  journal = {International Journal of Disaster Risk Reduction},
  year = {2026},
  doi = {10.1016/j.ijdrr.2026.106058},
  url = {https://doi.org/10.1016/j.ijdrr.2026.106058}
}