García-Gamero et al. (2026) Predicting hydrological drought at global scale: an analysis of the CEMS seasonal forecasts

Identification

• Journal: Natural Hazards
• Year: 2026
• Date: 2026-04-01
• Authors: Vanesa García-Gamero, Carmelo Cammalleri, Alessandro Ceppi, Christel Prudhomme, Arthur Ramos, Juan Camilo Acosta Navarro, Andrea Toreti
• DOI: 10.1007/s11069-025-07751-w

Research Groups

• Politecnico di Milano, Dipartimento di Ingegneria Civile e Ambientale (DICA), Milan, Italy
• European Centre for Medium-Range Weather Forecasts (ECMWF), Reading, UK
• European Commission, Joint Research Centre, Ispra, Italy

Short Summary

This study evaluates the performance of the Copernicus Emergency Management System (CEMS) seasonal forecasts in detecting global hydrological drought, demonstrating high skill for 1- to 3-month horizons and identifying key drivers of predictability and the utility of the signal-to-noise ratio (SNR) for forecast reliability.

Objective

• To exploit CEMS seasonal forecasts for hydrological drought forecasting, quantifying forecast skill at different temporal scales across the globe, and detecting the main sources of predictability.

Study Configuration

• Spatial Scale: Quasi-global coverage (90° N to 60° S), with raw data at 0.05° spatial resolution, aggregated to 0.5° resolution for analysis.
• Temporal Scale: Evaluation period from 1991 to 2022. Raw data at daily time steps, aggregated to monthly values. Seasonal forecasts with a temporal horizon up to 7 months, analyzed for Standardized Streamflow Index (SSI) accumulation periods of 1, 3, and 6 months.

Methodology and Data

• Models used: LISFLOOD hydrological model, ECMWF Seasonal Forecast System version 5 (SEAS5).
• Data sources:
  • CEMS GloFAS-ERA5 (G-E) dataset: LISFLOOD simulations forced by ECMWF ERA5 reanalysis (used as observational reference proxy).
  • CEMS GloFAS-SEAS5 (G-S) dataset: LISFLOOD simulations forced by ECMWF SEAS5 seasonal reforecasts (ensemble of 25-51 members).
  • ECMWF ERA5 reanalysis (for initial conditions and meteorological forcings variability).

Main Results

• The ensemble mean of the Standardized Streamflow Index (SSI) shows high skill at 1- and 3-month time horizons, remaining informative up to 6 months ahead (average Pearson r > 0.6 for 6-month horizon).
• Forecasts consistently outperform trivial persistence forecasts (using the latest available reanalysis state), with the highest added value observed, on average, in the Northern Hemisphere, particularly at higher latitudes (above 50° N) for 3- and 6-month lead times in spring–summer.
• Seasonality in average river discharge is a main driver of skill differences between seasons, with higher skill generally observed when average river discharge is lower.
• Inter-annual variability in initial conditions (Soil Wetness Index) plays a dominant role in forecast skill during summer (JJA) and spring–summer (MAM-JJA).
• Inter-annual variability in meteorological forcings (precipitation) primarily influences forecast skill during winter (DJF) and autumn–winter (SON-DJF).
• The signal-to-noise ratio (SNR) of the ensemble forecast is identified as a key metric to operationally assess and communicate forecast reliability, with higher average skill associated with SNR values greater than 2.

Contributions

• Provides the first comprehensive global assessment of the skill of CEMS seasonal forecasts for hydrological drought detection.
• Quantifies the forecast skill at various temporal scales and identifies the main drivers of predictability (initial conditions, meteorological forcings, seasonality).
• Demonstrates the significant added value of seasonal hydrological drought forecasts over persistence-based approaches.
• Establishes the Signal-to-Noise Ratio (SNR) as a valuable intrinsic measure for assessing the reliability of hydrological drought forecasts in near real-time, supporting operational decision-making.
• Supports the operational implementation of new seasonal hydrological forecast indicators within the CEMS portfolio.

Funding

• Horizon Europe project SEED-FD (Strengthening Extreme Events Detection for Flood and Drought) HORIZON‐CL4‐2023‐SPACE‐01‐32 (CUP: D43C23003660006-2023).

Citation

@article{GarcíaGamero2026Predicting,
  author = {García-Gamero, Vanesa and Cammalleri, Carmelo and Ceppi, Alessandro and Prudhomme, Christel and Ramos, Arthur and Navarro, Juan Camilo Acosta and Toreti, Andrea},
  title = {Predicting hydrological drought at global scale: an analysis of the CEMS seasonal forecasts},
  journal = {Natural Hazards},
  year = {2026},
  doi = {10.1007/s11069-025-07751-w},
  url = {https://doi.org/10.1007/s11069-025-07751-w}
}