Manco et al. (2025) Identifying recurring patterns of extreme daily precipitation using K-means algorithm: Uncovering spatial shift driven by climate change over the Italian Peninsula

Identification

• Journal: Weather and Climate Extremes
• Year: 2025
• Date: 2025-12-26
• Authors: I. Manco, O.M. Feitosa, M. Raffa, P. Schiano, G. Rianna, P. Mercogliano
• DOI: 10.1016/j.wace.2025.100849

Research Groups

• CMCC Foundation - Euro-Mediterranean Center on Climate Change, Italy
• INPE - National Institute for Space Research, Brazil
• Department of Physics and Astronomy, University of Bologna, Bologna, Italy

Short Summary

This study applies k-means clustering to high-resolution climate projections (VHR-PRO_IT) to identify and characterize recurring spatial-temporal patterns of extreme daily precipitation over the Italian Peninsula, revealing significant shifts and increased variability under future climate scenarios (RCP4.5 and RCP8.5).

Objective

• To characterize recurring patterns of extreme daily precipitation events across the Italian Peninsula during the historical period (1981–2005) using k-means clustering.
• To analyze the potential changes in these precipitation patterns and their spatial distribution under future climate scenarios (RCP4.5 and RCP8.5) for the period 2035–2065.
• To associate the identified recurring precipitation clusters with the atmospheric dynamics that impact the Italian Peninsula.

Study Configuration

• Spatial Scale: Italian Peninsula, with a resolution of 2.2 km.
• Temporal Scale: Historical period (1981–2005) and future period (2035–2065) under RCP4.5 and RCP8.5 scenarios, analyzed at daily frequency.

Methodology and Data

• Models used: K-means clustering algorithm, Principal Component Analysis (PCA) for sensitivity analysis of cluster number. Climate projections from the COSMO-CLM model.
• Data sources: VHR-PRO_IT dataset, a very-high-resolution climate projection dataset for Italy. Extreme daily precipitation events were cataloged using a two-step thresholding process (98th and 95th percentiles).

Main Results

• The k-means clustering identified 13 distinct extreme daily precipitation patterns, with complex orography, geographical location, and maritime influence playing decisive roles.
• Significant seasonal variations were observed, with the highest precipitation intensities occurring in winter, particularly in mountainous regions.
• Future projections (2035–2065) indicate an overall increase in rainfall variability across the Italian Peninsula, with the standard deviation increasing by approximately 19% under RCP4.5 and 17% under RCP8.5.
• Intensification of precipitation is projected for the Eastern Alps and northern Apennines, while a decreasing trend is observed in Sicily, Sardinia, and along the Tyrrhenian coast during summer.
• Spatial reorganization of precipitation patterns is evident, including the splitting of some historical clusters and the disappearance of others, alongside a general southward displacement of cluster centroids for centro-south regions, suggesting a strengthening of maxima over southern Italy.
• Winter projections show increased intensity and variability, with the Western Alps emerging as a distinct region with higher median values (e.g., approximately 21 mm in RCP4.5 compared to 15.8 mm historically).
• Summer projections indicate a general decrease in precipitation across the peninsula, particularly along the Tyrrhenian coast and in Sardinia/Sicily (median values approaching 0 mm), with localized exceptions like Mount Etna and specific Alpine regions showing increases.
• Autumn projections show increased mean values for each cluster, more pronounced under RCP4.5, with significant increases in Piemonte (e.g., from 41.4 mm to 55.0 mm under RCP4.5) and the Trentino Alps.
• Spring projections exhibit the smallest variations, with a slight overall decrease in mean under RCP4.5 and a slight increase under RCP8.5, alongside a more homogeneous distribution in Lombardy and enhanced precipitation in Apulia, Sicily, and southern Calabria.

Contributions

• Provides the first application of convection-scale climate simulations with clustering analysis to identify mesoscale precipitation features and their evolution under future climate projections (RCP4.5 and RCP8.5) over the Italian Peninsula.
• Offers a detailed, objective, and data-driven classification of extreme daily precipitation regimes, accounting for complex orography and seasonal specificity at very high resolution (2.2 km).
• Delivers critical insights into localized changes in precipitation intensity and spatial shifts, which are essential for regional climate adaptation planning and civil protection.
• Proposes a homogeneous, data-driven framework for delineating alert zones, enhancing consistency and transparency in risk communication and emergency planning.
• Supports risk-management strategies in financial and insurance sectors by identifying climatologically homogeneous and highly susceptible areas.
• Establishes a robust foundation for detailed local studies, including the calibration of stochastic precipitation generators and the derivation or updating of intensity-duration-frequency (IDF) curves.

Funding

• 3 ICSC - Centro Nazionale di Ricerca in High-Performance-Computing, Big Data and Quantum Computing, funded by European Union -NextGenerationEU.

Citation

@article{Manco2025Identifying,
  author = {Manco, I. and Feitosa, O.M. and Raffa, M. and Schiano, P. and Rianna, G. and Mercogliano, P.},
  title = {Identifying recurring patterns of extreme daily precipitation using K-means algorithm: Uncovering spatial shift driven by climate change over the Italian Peninsula},
  journal = {Weather and Climate Extremes},
  year = {2025},
  doi = {10.1016/j.wace.2025.100849},
  url = {https://doi.org/10.1016/j.wace.2025.100849}
}