Probability of Reactivation of a Sand Boil near a River Embankment of the Po River (Italy)

This paper presents a reliability analysis referred to backward erosion piping phenomena. The study is carried out on a cross-section of the major Italian watercourse, the Po River, where evidence of piping effects, namely a sand boil of remarkable size, has been periodically detected during past high-water events. The initiation of the process has been analysed by means of a two-dimensional (2D) finite element (FE) model of the groundwater flow beneath the river embankment, to obtain an effective description of the hydraulic gradient variation close to the sand boil due to the rise of the river water level, and thus capture the conditions triggering piping. Detailed stratigraphic soil profiling, as well as careful geotechnical characterization of the riverbank sediments and the foundation subsoil, have been carried out. In particular, representative mean values of the saturated permeability, and the relevant standard deviation, have been obtained from statistical analysis of the estimates provided by empirical correlations applied to cone penetration tests (CPT) profiles. The methodology adopted for reliability analysis relies on the use of fragility curves. The probability of backward erosion piping initiation as a function of the floodwater elevation has been therefore calculated assuming that hydraulic conductivity and a few geometrical parameters affecting the problem could be considered as random variables in the FE model. Then, the probability of reactivation has been estimated using the Taylor’s Series first-order second moment (FOSM) method. The same procedure has been developed by applying the well-known blanket theory. Both analyses result in a high probability of reactivation of the sand boil, also for moderate river levels. In addition, fragility curves have been interpreted in the light of the available field observations. The interpretation indicates that, despite the significant variability of the water levels recorded at the on set of reactivations, the blanket theory tends to overestimate the probability of sand boil initiation, while the FE analysis results are closer to the field observations collected in the last 20 years