Magnini et al. (2025) Informativeness of teleconnections in frequency analysis of rainfall extremes

Identification

• Journal: Hydrology and earth system sciences
• Year: 2025
• Date: 2025-10-09
• Authors: Andrea Magnini, Valentina Pavan, Attilio Castellarin
• DOI: 10.5194/hess-29-5031-2025

Research Groups

• Department of Civil, Environmental, Chemical and Materials Engineering (DICAM), University of Bologna, Bologna, Italy
• ARPAE-SIMC Emilia Romagna, Bologna, Italy

Short Summary

This study proposes a reproducible framework to assess the informative content of teleconnections for regional frequency analysis of rainfall extremes in North-Central Italy. It identifies significant spatial patterns of correlation between specific climate indices (WeMOI, EA-WR) and rainfall extreme statistics, demonstrating that climate-informed regional models can improve the goodness-of-fit compared to stationary approaches.

Objective

• To develop a framework for assessing the informative content of teleconnections in the frequency analysis of hourly and daily rainfall extremes.
• To investigate the possibility of delineating robust regional zonation of the dependence on teleconnections.
• To evaluate the effect and suitability of teleconnection-informed regional frequency analysis.

Study Configuration

• Spatial Scale: North-Central Italy, utilizing 680 gauged stations. The study area is discretized into non-overlapping tiles with resolutions of 0 km (at-site), 15 km, 30 km, and 50 km for correlation analysis. Regional L-coefficient of variation (L-CV) is computed at 30 km resolution, and regional L-skewness (L-CS) at 100 km resolution.
• Temporal Scale: Rainfall data span from 1921 to 2020. Annual maximum series (AMS) of rainfall depth are analyzed for 1 hour and 24 hours durations. Sliding time windows of 30 years are used for teleconnection indices and 10 years for AMS statistics (mean and L-CV).

Methodology and Data

• Models used:
  • L-moments (mean, L-coefficient of variation, L-skewness) for characterizing rainfall extreme distributions.
  • Spearman correlation coefficient for assessing relationships between L-moments and teleconnections, with significance testing accounting for autocorrelation.
  • Generalized Extreme Value (GEV) distribution for frequency analysis, including stationary (GEV0) and doubly-stochastic (GEV1, GEV2, GEV3) models where parameters depend on teleconnections.
  • Second-order polynomial functions to model L-moments as functions of teleconnections.
  • Hierarchical regional frequency analysis approach.
  • Monte Carlo experiments for evaluating RML significance thresholds and field significance of doubly-stochastic signals.
  • Ordinary kriging for interpolating reliability index (ri) maps.
• Data sources:
  • Rainfall data: I2-RED dataset (Mazzoglio et al., 2020a), comprising 680 stations in North-Central Italy.
  • Teleconnection indices: North Atlantic Oscillation (NAO), Pacific Decadal Oscillation (PDO), East Atlantic – West Russia pattern (EA-WR), El Niño Southern Oscillation (ENSO) from NOAA Physical Sciences Laboratory; Mediterranean Oscillation Index (MOI), Western Mediterranean Oscillation Index (WeMOI) from the University of East Anglia's Climate Research Unit (CRU).

Main Results

• The Western Mediterranean Oscillation Index (WeMOI) and East Atlantic – West Russia pattern (EA-WR) show the strongest influence on the sliding mean of annual maximum rainfall, exhibiting clear spatial patterns of consistent negative correlation in the Gulf of Genoa and the North-Eastern Alps. These patterns are more pronounced for 1-hour rainfall duration.
• El Niño Southern Oscillation (ENSO) demonstrates the weakest influence on both the mean and L-CV of rainfall extremes.
• For the L-CV, PDO and MOI show promising homogeneous patterns, but overall, consistent correlation patterns are more heterogeneous and fragmented, particularly for 1-hour duration.
• Doubly-stochastic GEV models (GEV1, GEV2, GEV3) generally achieve an increased goodness-of-fit compared to stationary models (GEV0), with a significant number of sites showing improved performance based on the Ratio of Models' Likelihood (RML).
• WeMOI and EA-WR are particularly effective for GEV1 (mean-dependent) and GEV3 (mean and L-CV dependent) distributions, with the field significance of these non-stationary signals confirmed against spatial dependence.
• The variability of predicted rainfall percentiles associated with changes in teleconnections can be substantial; for example, the 100-year 24-hour rainfall depth at a specific location can vary from 170 mm to 300 mm when both mean and L-CV are modeled as functions of WeMOI.

Contributions

• Proposes a novel and reproducible framework for assessing the informative content of teleconnections in regional frequency analysis of rainfall extremes, uniquely focusing on L-moments rather than raw annual maximum series.
• Introduces a methodology for temporal (sliding windows) and spatial (tiling) aggregation of rainfall statistics, which effectively filters inter-annual variability and enhances the recognition of geographical patterns in teleconnection influence.
• Demonstrates the existence of robust regional zonation of dependence between key teleconnections (WeMOI and EA-WR) and rainfall extreme statistics (mean and L-CV) in North-Central Italy, aligning with known regional precipitation regimes.
• Shows that climate-informed regional frequency analysis models, even when employing simple polynomial functions, can significantly improve the goodness-of-fit and provide a wider, more realistic range of expected rainfall percentiles compared to traditional stationary models.
• Presents a general methodology that is highly adaptable to different geographical and climatic contexts, as well as other environmental variables such as floods and temperature.

Funding

• European Climate, Infrastructure and Environment Executive Agency (grant no. 101069928 – LIFE21-IPC-IT-LIFE CLIMAX PO).

Citation

@article{Magnini2025Informativeness,
  author = {Magnini, Andrea and Pavan, Valentina and Castellarin, Attilio},
  title = {Informativeness of teleconnections in frequency analysis of rainfall extremes},
  journal = {Hydrology and earth system sciences},
  year = {2025},
  doi = {10.5194/hess-29-5031-2025},
  url = {https://doi.org/10.5194/hess-29-5031-2025}
}