Landslides pose significant risks to communities and infrastructure, and mitigating these risks relies on understanding landslide causes and triggering processes. It has been shown that geophysical surveys can significantly contribute to the characterization of unstable slopes. However, hydrological processes can be temporally and spatially heterogeneous, requiring their related properties to be monitored over time. Geoelectrical monitoring can provide temporal and volumetric distributions of electrical resistivity, which are directly related to moisture content. To date, studies demonstrating this capability have been restricted to 2‐D sections, which are insufficient to capture the full degree of spatial heterogeneity. This study is the first to employ 4‐D (i.e., 3‐D time lapse) resistivity imaging on an active landslide, providing long‐term data (3 years) highlighting the evolution of moisture content prior to landslide reactivation and showing its decline post reactivation. Crucially, the time‐lapse inversion methodology employed here incorporates movements of the electrodes on the unstable surface. Although seasonal characteristics dominate the shallow moisture dynamics during the first 2 years with surficial drying in summer and wetting in winter, in the months preceding reactivation, moisture content increased by more than 45% throughout the slope. This is in agreement with independent data showing a significant rise in piezometric heads and shallow soil moisture contents as a result of prolonged and intense rainfall. Based on these results, remediation measures could be designed and early‐warning systems implemented. Thus, resistivity monitoring that can allow for moving electrodes provides a new means for the effective mitigation of landslide risk.

No abstract

The internal moisture dynamics of an aged (> 100 years old) railway earthwork embankment, which is still in use, are investigated using 2D and 3D resistivity monitoring. A methodology was employed that included automated 3D ERT data capture and telemetric transfer with on-site power generation, the correction of resistivity models for seasonal temperature changes and the translation of subsurface resistivity distributions into moisture content based on petrophysical relationships developed for the embankment material. Visualization of the data as 2D sections, 3D tomograms and time series plots for different zones of the embankment enabled the development of seasonal wetting fronts within the embankment to be monitored at a high-spatial resolution and the respective distributions of moisture in the flanks, crest and toes of the embankment to be assessed. Although the embankment considered here is at no immediate risk of failure, the approach developed for this study is equally applicable to other more high-risk earthworks and natural slopes.Geophysical ground imaging techniques offer the potential to complement existing approaches by spatially characterizing and monitoring the internal conditions of earthworks to provide highresolution information of subsurface property changes and hence precursors to slope failure. Resistivity imaging, or electrical resistivity tomography (ERT), holds particular promise due to its sensitivity to both lithological variations (e.g., Shevnin et al. 2007) and changes in soil moisture, which can be imaged by applying appropriate petrophysical relationships linking resistivity and saturation (e.g., Cassiani et al. 2009;Brunet et al. 2010). Twodimensional ERT is now a well-established technique for investigating natural slopes with numerous recent examples of the use of the technique for structural characterization and hydrogeological investigations (e.g

S U M M A R YIf electrodes move during geoelectrical resistivity monitoring and their new positions are not incorporated in the inversion, then the resulting tomographic images exhibit artefacts that can obscure genuine time-lapse resistivity changes in the subsurface. The effects of electrode movements on time-lapse resistivity tomography are investigated using a simple analytical model and real data. The correspondence between the model and the data is sufficiently good to be able to predict the effects of electrode movements with reasonable accuracy. For the linear electrode arrays and 2-D inversions under consideration, the data are much more sensitive to longitudinal than transverse or vertical movements. Consequently the model can be used to invert the longitudinal offsets of the electrodes from their known baseline positions using only the time-lapse ratios of the apparent resistivity data. The example data sets are taken from a permanently installed electrode array on an active lobe of a landslide. Using two sets with different levels of noise and subsurface resistivity changes, it is found that the electrode positions can be recovered to an accuracy of 4 per cent of the baseline electrode spacing. This is sufficient to correct the artefacts in the resistivity images, and provides for the possibility of monitoring the movement of the landslide and its internal hydraulic processes simultaneously using electrical resistivity tomography only.