Bianco et al. (2026) A framework for generating catalogues of high-impact UNSEEN flood events

Identification

• Journal: Climate Services
• Year: 2026
• Date: 2026-01-23
• Authors: Elena Bianco, Paolo Davini, Giuseppe Zappa, Agostino Manzato, Antonio Giordani, Paolo Ruggieri
• DOI: 10.1016/j.cliser.2026.100636

Research Groups

• Dipartimento di Fisica e Astronomia ‘‘Augusto Righi’’, University of Bologna, Bologna, Italy
• Istituto di Scienze dell’Atmosfera e del Clima, Consiglio Nazionale delle Ricerche (CNR-ISAC), Torino, Italy
• Istituto di Scienze dell’Atmosfera e del Clima, Consiglio Nazionale delle Ricerche (CNR-ISAC), Bologna, Italy
• ItaliaMeteo - National Agency for Meteorology and Climate, Bologna, Italy
• School of Geography, Earth and Environmental Sciences, University of Birmingham, Birmingham, United Kingdom (current affiliation for E. Bianco)
• ARPA FVG - OSMER, Palmanova (UD), Italy (current affiliation for A. Manzato)

Short Summary

This paper introduces a modular framework that combines ensemble reforecast pooling (UNSEEN approach) with probabilistic impact modeling (CLIMADA) to generate catalogues of physically plausible, high-impact, yet unobserved flood events for European river catchments. The framework's utility is demonstrated in the Panaro watershed, Italy, showing its capacity to anticipate record-breaking historical floods and support disaster management.

Objective

• To develop and demonstrate a modular framework for generating catalogues of physically plausible, high-impact, yet unobserved (UNSEEN) flood events using ensemble prediction systems, thereby improving flood risk preparedness and emergency management in regions with limited historical data.

Study Configuration

• Spatial Scale: European river catchments (general applicability); case study focused on the Panaro watershed in Emilia-Romagna, Italy (drainage basin approximately 1775 km²). EFAS/LISFLOOD data at 1 arcminute resolution, JRC flood hazard maps at 100 meter resolution, LitPop exposure data at 30 arcseconds resolution. Atmospheric fields from SEAS5 cover 30°N–70°N and 40°W–40°E with approximately 36 km horizontal resolution and 91 vertical levels.
• Temporal Scale: Study period from 1999 to 2024 (26 years). EFAS seasonal reforecasts are initialized monthly with a lead time of 215 days (approximately 7 months). Ensemble independence is achieved after discarding the first 90 lead days. Extreme value analysis uses annual maxima of daily mean discharge.

Methodology and Data

• Models used:
  • LISFLOOD (hydrological model, used in EFAS).
  • CLIMADA (CLImate ADAptation, probabilistic impact model).
  • ECMWF Integrated Forecasting System (IFS) 34r1 (atmospheric component of SEAS5).
  • LISFLOOD-FP (flood inundation model, used for JRC hazard maps).
  • k-means clustering (for identifying recurrent synoptic patterns).
  • Generalized Extreme Value (GEV) distribution for extreme value analysis.
• Data sources:
  • European Flood Awareness System (EFAS) v5.0 historical river discharge data.
  • EFAS seasonal reforecasts (EFAS5).
  • ECMWF SEASonal forecasts version 5 (SEAS5) reforecasts (Mean Sea Level Pressure (MSLP), Geopotential Height at 500 hPa (Z500)).
  • ECMWF ERA5 reanalysis (historical MSLP and Z500).
  • European Meteorological Observations version 1 (EMO1) (gridded precipitation and temperature for LISFLOOD forcing).
  • HydroBASINS dataset of HydroSHEDS (river basin boundaries).
  • FLOod PROtection Standards database (FLOPROS) (local flood protection measures).
  • European Commission Joint Research Center (JRC) flood hazard maps (100 m resolution, nine reference return periods).
  • LitPop dataset (economic asset exposure data, combining NASA SEDAC’s Gridded Population of the World v4.10 and NASA’s Black Marble nightlight intensity).
  • World Bank (Produced capital stock for national economic indicators).
  • Zenodo (published catalogue of high-impact UNSEEN events).

Main Results

• A modular framework was successfully developed and demonstrated for generating catalogues of high-impact UNSEEN flood events.
• For the Panaro watershed, the EFAS 25-member ensemble reforecasts achieved sufficient independence after discarding 90 lead days, allowing the construction of 100 independent surrogate time series (1999-2024), yielding 2600 synthetic annual maxima.
• Bias correction using linear multiplicative scaling on annual maxima effectively aligned the statistical moments of surrogate data with historical observations.
• The pooled distribution of synthetic annual maxima was well-represented by a GEV fit, with a shape parameter (𝜉) generally between -0.2 and 0.2, centered around -0.026.
• The historical record-breaking event of November 2019 (488.3 m³/s at Bomporto) was exceeded by 40 UNSEEN events in the bias-corrected distribution.
• The highest Estimated Economic Damage (EED) identified was approximately 591.0 million USD for an UNSEEN event in October 2007, with a simulated discharge of 676.8 m³/s at Bomporto.
• Two dominant synoptic patterns conducive to high-impact floods in Emilia-Romagna were identified: a deep cyclone over central-western Italy or a deep trough over western Europe with an eastern flank over Italy.
• Analogue analysis demonstrated the framework's ability to anticipate historical floods; for example, the December 2020 historical flood's atmospheric circulation closely resembled the highest-impact UNSEEN event (October 2007), which had a 38.7% higher simulated river discharge at Bomporto and an EED of 591.0 million USD.

Contributions

• Introduces a novel, customizable, end-to-end framework that integrates ensemble reforecast pooling (UNSEEN approach) with probabilistic impact modeling (CLIMADA) to create catalogues of physically plausible, high-impact, yet unrealized flood events.
• Addresses the critical challenge of limited historical records for assessing rare, high-impact flood events by significantly expanding the sample size of extreme events through ensemble simulations.
• Demonstrates the framework's successful application to medium-sized river basins, an area often less explored by UNSEEN methodologies compared to larger national-scale systems.
• Provides a robust, physically-based foundation for exploring realistic worst-case flood scenarios and supports proactive preparedness and adaptation decision-making at the regional level.
• Highlights the added value of combining ensemble pooling with impact modeling to not only identify potential high-impact extreme events but also to understand their preconditioning synoptic drivers.
• The framework is fully enabled by open-source data from Copernicus Services, ensuring its adaptability and potential for application across various European regions with sufficient data coverage.

Funding

• NextGenerationEU as part of the PRIN PNRR 2022 project TRANSLATE - climaTe Risk informAtion from eN- SembLe weAther and climaTe prEdictions.

Citation

@article{Bianco2026framework,
  author = {Bianco, Elena and Davini, Paolo and Zappa, Giuseppe and Manzato, Agostino and Giordani, Antonio and Ruggieri, Paolo},
  title = {A framework for generating catalogues of high-impact UNSEEN flood events},
  journal = {Climate Services},
  year = {2026},
  doi = {10.1016/j.cliser.2026.100636},
  url = {https://doi.org/10.1016/j.cliser.2026.100636}
}