Comparison of Finite Element Methods and Satellite InSAR for Monitoring Deformations of a Large Tailings Dam

Abstract

Following the recent catastrophic failure of several mine tailings dams there has been much interest in the use of numerical modeling and remote sensing for monitoring the safety and stability of these structures. This thesis presents a case study that investigates the accuracy of InSAR measurements and the predictive capabilities of finite element models using ground truth surface and sub-surface monitoring data applied to the Zelazny Most (SW Poland) copper tailings storage facility. This site has a well-documented history of lateral deformations in a critical section (XVIE) of the East dam that have been attributed to a deep-seated translation mechanism of shearing through the underlying Pliocene, glacial clays. Since 2014, operators of the facility have constructed a series of stabilizing berms at this critical section. We investigated the accuracy of InSAR over this period, ending in 2019, by analyzing 186 ascending Sentinel-1 C-band images and 219 descending images using Persistent Scatterer Interferometry and SARProzTM software, comparing results with two surface geodetic benchmarks. Finite element analyses of the structure required a 2D model of section XVIE. We developed and integrated a stratigraphic model for the foundation soils, the complete construction history of the dam (since 1975), and selected input parameters for constitutive models to represent the soil behavior (foundation soils, tailings, dyke and berm materials) using PlaxisTM software. Our results show that InSAR achieves very consistent agreement with geodetic measurements for vertical (Up-Down) and lateral (E-W) surface deformations, over a time period where construction was limited to raising of the dyke near the crest of the dam and berm construction at the toe. The InSAR data are also insightful in showing relatively uniform lateral deformations occurring over the face of the dam, consistent with the interpreted translational failure mechanism. In contrast, it has proved much more challenging to predict subsurface deformations by FE analyses. The computed movements reflect accumulation of deformations over multiple stages of construction and involve shearing through the complex foundation stratigraphy. We were able to achieve credible estimates of lateral deformations within the range of laboratory shear strength properties published in the literature and using the Hardening Soil (HS) model for non-linear shear stress-strain properties. However, the predictions of surface settlements and lateral deformation are much less reliable and depend on undocumented properties of the tailings, phreatic conditions in the tailings and details of the construction history.