Sinkholes Susceptibility Assessment in Urban Environment Using Heuristic, Statistical and Artificial Neural Network (ANN) Models in Evaporite Karst System: A Case Study from Lesina Marina (Southern Italy).

Sinkhole susceptibility assessment was carried out in “Lesina Marina” evaporite karst area, located in the north-eastern part of the Apulia region (southern Italy), near the Adriatic coast. Land instability due to the widespread presence of sinkholes especially in a built-up area, constitutes a complex dynamic system, structured by a sets of interacting components, controlled by several natural and anthropogenic factors, forming an integrated whole, in which physical dynamic processes evolve. Heuristic method, multivariate statistical analysis and ANN procedure were performed in order to assess sinkhole susceptibility. In the study area, sinkhole phenomenon is strictly related to the structure and stratigraphy of the evaporite rocks, the groundwater regime conditioned by tide-induced surface water and groundwater interactions, and by the presence of the complex sea-channel-lagoon system. The analysis performed by different procedures explains the relationship between datasets and models capability to predict the behaviour of the phenomenon. The performances of prediction models have been evaluated using ROC curves. The results show that the multivariate statistical model produces a more reliable accuracy.