Hydraulic properties calibration by high resolution monitoring data and 1D to 3D groundwater flow modeling of the Carozzo Landslide

A reliable modeling of a landslide activation and reactivation requires a representative geological and engineering geological characterization of the affected materials. Beyond the material strength, landslide reactivation is sensitive to groundwater pressure distributions, that are generated by some external perturbation (recharge) and by the hydraulic properties of the materials. Drainage stabilization works generally involve drilling of a large number of drains and, therefore, minimize the total length is of primary concern to reduce the costs.

Aim of this work was the calibration of material properties for the optimization of drainage elements to be built for the slope stabilization and the construction of a shallow tunnel crossing a landslide. The case study is represented by the 4.0 · 10⁵ m³ Carozzo landslide (La Spezia, Liguria, Italy) which affects some marly and sandstone formation. During the tunnel excavation a monitoring network consisting of five DMS columns for displacements and piezometric head multilevel measurements was installed. The monitoring provided a series of piezometric head recession curves following some recharge events. The series of data generated in response of a unique perturbation (rainfall recharge event) were chosen to calibrate the material properties through a multi-step approach, starting from a 1D model and progressively approaching a complete 3D model.

The 1D simplified approach applies the solution by Troch et al. (2003) that considers a homogeneous landslide material, with constant slope and a progressive change in the slope width. In this model a storage function considers the amount of water stored in a slope section. By imposing the continuity equation and the Darcy law a second order of partial differential equation is solved by integration in space and time. By taking the initial conditions from piezometric measurements and assuming a constant rainfall recharge, the piezometric level and the outflow rate were computed and compared with the local piezometric level time history, by changing the hydraulic conductivity and the storage function value.

Successively, a groundwater flow FEM numerical model (in 2D and 3D) was developed considering the landslide geometry and internal zonation, including the presence of the excavated part of the tunnel. The model domain was divided into sub-zones according to the available geological surveys to account for internal variations of the material properties. The steady-state simulation of the water flow allowed to estimate the equivalent hydrogeological parameters of each subdomain. The hydraulic head distribution obtained under steady-state conditions was used as initial condition for the transient-state simulation. The recharge from precipitation was also included in the water balance by means of daily rainfall time-series. Finally, the model parameters were calibrated in transient state by comparing measured data and simulated results.

The minimum error between simulated and measured piezometric heads under transient conditions was obtained through the 3D configuration. Calibrated hydraulic conductivities in the 3D solution are up to an order of magnitude lower than the 1D solution because of the homogenous assumption of the model. The internal zonation of the landslide body and the modeling of a low-conductivity shear zone were essential to explain the pressure differences inside the body.