Multiscale Granger dependencies in the precipitation network of the island of Sicily

The island of Sicily has been displaying unusual rainfall behavior and unexpected extreme precipitation events in recent decades. In this study, we investigate the Granger causal (GC) dependencies in the network of precipitation measurement sites of Sicily at different timescales (every 10 min, 1 h, 6 h, 12 h, and 24 h). We study, across seasons and years, different parameters that characterize the GC dependencies: the total in/out-degree of nodes, the total in/out strength of nodes, the total number of links in the network, the number of eastward/westward links, the strength of eastward/westward links, and the maximum strength of links. We then investigate GC statistic intensities, focusing on the temporal evolution of maximum values over multiple timescales. Our study of precipitation patterns in Sicily indicates that, since 2013, the southern regions near Mount Etna (Catania, Siracusa, and Ragusa) have been increasingly affected, while the western areas (Trapani, Palermo, and Agrigento) have been the most affected. Granger causality networks reveal scale-invariant dependencies, with stronger and sparser connections at timescales that extend beyond 6 h, with a notable westward flow of predictive information. These patterns, which are consistent across seasons, suggest localized perturbation fronts, with stronger links indicating a more significant influence on westward predictions. This study highlights shifts in Sicily's water cycle that call for adaptive management strategies in the face of the increasing frequency of extreme events.