Analisi della suscettibilità ai sinkholes antropogenici nel centro urbano di Roma

The anthropogenic sinkholes in the urban area of Rome are closely linked to the network of underground cavities produced by human activities (water networks, caves, catacombs, etc.) in more than two thousand years of history. The collapse of such underground cavities and the subsequent fall down of the shallower layers of the soil may originate at surface pits up to metric dimensions. Furthermore, the formation of these pits can cause underground water leaks and failure of the hydraulic network producing the washing away of loose soil cover. The combination of both factors (e.g., cavity losses and the hydraulic network) increases this process. Over the past fifteen years the increased frequency of intense rainfall events, favors sinkhole formation; this leads to a greater risk to the population and infrastructure with damage to roads and underground utilities. Furthermore, the anthropogenic sinkholes may constitute a serious problem for the city of Rome, especially for the damages to the archaeological heritage and their preservation. In this scenario a risk assessment induced by anthropogenic sinkhole is really difficult. However, a susceptibility of the territory to sinkholes can be more easily determined as the probability that an event may occur in a given space, with unique geological-morphological characteristics, and in an infinite time. A map of sinkhole susceptibility of the Rome territory up to the ring road has been constructed by using Geographically Weighted Regression technique and geostatistics. The spatial regression model includes the analysis of more than 2600 anthropogenic sinkholes (recorded from 1875 to 2014), as well as geological, morphological, hydrological and predisposing anthropogenic characteristics (e.g., presence of cavities in the ground, the distribution network of underground etc.) of the study area. The susceptibility map was then compared with the data of ground subsidence (InSAR) to obtain a predictive model.