Reconstructing subsurface fracture geometries in rock slope instabilities through ambient vibration-based numerical modelling inversion

Detailed engineering-geological models are crucial for assessing landslide hazards, yet their reliability is limited when poorly defined fracture networks control slope failure mechanisms. Traditional surveying techniques often fail to accurately constrain fracture extents, resulting in oversimplified and uncertain boundary conditions. We address these limitations by integrating array-based ambient vibration modal analysis with numerical modelling to invert for the subsurface geometry of fracture-controlled rock slope instabilities. We applied our approach at two case studies exhibiting similar toppling failure mechanisms. Linear seismic arrays were deployed to record ambient vibrations and derive resonance frequencies and 3D mode shapes using the Frequency Domain Decomposition technique. We then constructed 3D finite-element models representing the unstable rock volumes with their rear boundaries segmented into regular grids to simulate thousands of unique fracture configurations. Model results were compared with field-derived modal parameters using a multi-metric similarity ranking score evaluating resonance frequency and mode shape consistency. Results revealed ensembles of top-performing models that reproduced the observed resonance modes and converged toward fracture geometries consistent with field-estimated fracture depths. Inversion stability increased with the number of resonance modes considered, highlighting the need for multiple constraints. Our results demonstrate that integrating ambient vibration field surveys with numerical modal analysis can support quantitative description of subsurface boundary conditions in unstable rock slopes, providing a robust framework for improved landslide structural characterization and monitoring.