Yıldız et al. (2025) Evaluating global precipitation datasets over Sicily: From daily estimates to extreme events

Identification

• Journal: Journal of Hydrology Regional Studies
• Year: 2025
• Date: 2025-12-19
• Authors: Mehmet Berkant Yıldız, Fabio Di Nunno, Giovanni de Marinis, Francesco Granata
• DOI: 10.1016/j.ejrh.2025.103062

Research Groups

University of Cassino and Southern Lazio, Department of Civil and Mechanical Engineering (DICEM), Frosinone, Cassino, Italy.

Short Summary

This study evaluates 11 global daily precipitation datasets against ground observations across Sicily (2003–2023) to assess their accuracy and ability to represent extreme events. It found that blended products (MSWEP, ERA5-Land, HydroGFD) performed best for daily estimates, but all datasets consistently underestimated the magnitude and frequency of severe extreme precipitation, particularly in mountainous regions.

Objective

• To identify which global precipitation datasets most accurately reproduce daily rainfall over Sicily compared to ground-based observations.
• To assess how well these datasets represent extreme precipitation, including events associated with major floods.

Study Configuration

• Spatial Scale: Sicily, the largest island in the Mediterranean, spanning approximately 25,700 square kilometers.
• Temporal Scale: 2003–2023 (20 years of daily observations).

Methodology and Data

• Models used:
  • Statistical metrics: Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Percent Bias (PBIAS), Pearson correlation coefficient (r), Nash–Sutcliffe Efficiency (NSE), Modified NSE (mNSE), Index of Agreement (d), Modified Index of Agreement (md), and Kling–Gupta Efficiency (KGE).
  • Extreme value analysis: Generalized Extreme Value (GEV) distribution, Peaks Over Threshold (POT) method (using Generalized Pareto Distribution, GPD).
• Data sources:
  • 11 global daily precipitation datasets:
    • Satellite-based: GPM IMERG, TRMM, PERSIANN (PERSIANN/PERS, PERSIANN/CCS, PERSIANN/CDR, PERSIANN/PDIR-Now).
    • Reanalysis: ERA5-Land, GLDAS-2.
    • Blended: MSWEP V2, HydroGFD3.0.
    • Gauge-based gridded: E-OBS.
  • Ground-based observations: Servizio Informativo Agrometeorologico Siciliano (SIAS) meteorological stations (point-to-pixel approach).

Main Results

• Overall Performance: MSWEP V2 consistently delivered the highest overall accuracy for daily precipitation estimates, followed by ERA5-Land and HydroGFD3.0. Satellite-based datasets (PERSIANN family, TRMM) showed consistently poor performance.
• Spatial Variability: Dataset accuracy varied spatially, with higher performance in southern and central-southern Sicily (e.g., Delia station, r = 0.79 for MSWEP) and lower performance in northern and northeastern mountainous areas (e.g., Linguaglossa station near Mount Etna, r = 0.70 for MSWEP).
• Extreme Events: While MSWEP accurately represented moderate extreme events, it systematically underestimated the magnitude and frequency of severe precipitation events.
  • GEV analysis showed that MSWEP often simplified bimodal empirical distributions into single peaks and significantly underestimated return levels for rare events (e.g., 100-year return level underestimated by approximately 50 mm at Delia, and by approximately 900 mm at Linguaglossa).
  • POT analysis revealed a systematic underrepresentation of extreme overflow events in MSWEP, with observed 99th percentile thresholds extending to 200–300 mm in mountainous areas, while MSWEP remained confined to 80–100 mm.
• Error Sources: The underestimation of intense rainfall in mountainous and convective regions is attributed to limitations in satellite retrieval algorithms, simplified representations of convection and terrain interactions in reanalysis models, and residual errors from data blending.

Contributions

• The first island-wide intercomparison of 11 global daily precipitation datasets over Sicily, including satellite, reanalysis, blended, and observational products.
• A novel linkage between general performance metrics and formal extreme-value analysis, providing insights into how datasets capture extreme precipitation events.
• Spatially explicit assessments of dataset performance across Sicily’s diverse microclimatic conditions.

Funding

Not explicitly mentioned in the provided text.

Citation

@article{Yıldız2025Evaluating,
  author = {Yıldız, Mehmet Berkant and Nunno, Fabio Di and Marinis, Giovanni de and Granata, Francesco},
  title = {Evaluating global precipitation datasets over Sicily: From daily estimates to extreme events},
  journal = {Journal of Hydrology Regional Studies},
  year = {2025},
  doi = {10.1016/j.ejrh.2025.103062},
  url = {https://doi.org/10.1016/j.ejrh.2025.103062}
}