The application of the Boolean Stochastic Generation Method to model seepage under levees in heterogenous soils

Soil heterogeneity plays an important role in incrementing the uncertainty about the reliability of geotechnical engineering works, especially levees. The presence of thin layers of soils more permeable than the surrounding matrix significantly changes the seepage flow net below and within the dike. However, the detection of these layers is extremely difficult. Consequently, to evaluate the overall safety of dikes it may be useful to adopt stochastic computation methods. The paper presents the calibration of a seepage model of an operational river embankment subject to sand boils. The levee, located along the Adige River (South Tyrol, Italy), has been monitored since 2016. Piezometers and spot temperature sensors are installed in the lateral banks while a distributed fiber optic sensors (FOS), buried in a 350 m long trench at the toe of the levee, is used to check temperature distribution in the subground. The Boolean Stochastic Generation (BoSG) method addressess the influence of layers of material with different properties with respect of the surrounding soil. 360 soil configurations were generated for a two-dimensional groundwater flow model of the levee and confronted with the monitoring data of two piezometers. This analysis permits to identify the configuration that has effects more congruent with piezometers data, which in this case is a configuration with a major presence of lenses on the waterside respect the fieldside. This evidence could guide strategies for remedial works.