Informativeness of teleconnections in local and regional frequency analysis of rainfall extremes

Numerous studies have established strong long-range relationships (teleconnections) between global climatic indexes and precipitation across diverse geographical regions worldwide. Typically, these investigations focus on the number of wet days or cumulative rainfall over specific seasons or the entire year, while only a few explicitly explore the informative value of teleconnections in describing the frequency regime of sub-daily rainfall annual maxima. Furthermore, most studies analyze the correlation between rainfall characteristics and teleconnection index values at individual gauge stations within the same season and without considering any time lag. 

Our study provides a comprehensive assessment of the potential and informative content of teleconnections for representing and modeling the frequency regime of rainfall extremes, addressing the limitations mentioned above. Our dataset comprises annual maximum series (AMS) of sub-daily rainfall depth recorded between 1921 and 2022 at approximately 2300 rain gauges spanning a large and climatically diverse region in Northern Italy. Based on a comprehensive literature review, we selected six global climate indexes and evaluated their correlation with time series of gridded regional L-moments, statistical measures characterizing the distribution of sub-daily rainfall extremes. In analyzing the spatial patterns of gridded L-moments, we considered time aggregation intervals (durations) ranging from 1 to 24 hours, discretization of the study region with tile sizes (resolutions) up to 100 km, and time lags in teleconnections up to 30 years. Our results reveal significant spatial patterns in the teleconnections, with the Western Mediterranean Oscillation Index exhibiting stronger relationships. The robustness of these spatial patterns is confirmed by their limited sensitivity to the chosen grid resolution and time lag, likely arising from the utilization of time series of spatially smoothed statistics of AMSs (gridded L-moments) rather than raw annual sequences of rainfall maxima. Consequently, our research suggests promising pathways for climate-informed local and regional frequency analysis of rainfall extremes.