Comparison of the pyGIMLi and BERT packagesfor landslide and cavity detection: A case studyfrom Tghat-Oued Fez, Morocco

Abstract

In the present study, two-dimensional (2D) electrical resistivity tomography (ERT) and vertical electricalsounding (VES) techniques were adopted to predict and detect landslides and subsurface voids occurrencesin the Tghat-Oued Fez district of Fez city, Morocco. The purpose of this research is to examine theeAects of electrical-based applied geophysics methods for data inversion and modelling using open-sourcealgorithms for spatial distribution and shape of sliding and underground cavities as well as fractured zonesin the study area. The data acquisition was based on nine ERT proBles using a dipole–dipole electrodearray conBguration and 72 VES soundings laid out on an area in which underground cavities wereexpected. 2D electrical sections, derived from electrical resistivity tomography measurements, in terms ofelectrical resistivity contrast, correctly highlighted the site’s well-deBned lithology, identiBed the presenceof cavities with variable geometry, and were capable of inferring the sliding surface. VES-based measurements,in conjunction with the ERT technique, mapped the potential presence of near-surface voidsinside a conductive marly formation caused by high electrical resistivity anomalies and delimited theconglomerate roof layer. The joint combination of these two electrical-based geophysical methods alongwith the numerical modelling and inversion of the data structure and parameters using the useful featuresof the open-source python-based environment pyGIMLi and BERT, have demonstrated their capabilityto infer near-surface voids in the study area. The geophysical survey results suggest that the integratedgeophysical approach is a reliable and capable geophysical tool and could be eAectively used for landslide,and cavity detection and to assess the risk of those subsidence-prone areas. Thus, this approach could beeasily applied and reproduced in those areas with similar characteristics.