Active fault reactivation poses significant hazards, and understanding their near-surface structure is crucial for mitigating seismic risk. Along the Balakot-Bagh fault (BBF), the source of the 2005 Mw 7.6 Kashmir earthquake, geomorphic evidence is gradually eroded and sedimented. Traditional Electrical Resistivity Tomography (ERT) often produces smoothly varying tomograms that obscure sharp structural boundaries. This study introduces an integrated interpretation framework that combines geological mapping, high-resolution ERT imaging, and machine learning (ML) k-means clustering to improve characterization of the BBF’s shallow deformation zone at Sar Pain (S1) and Naushahra (S2), Pakistan. Geological surveys at S1 document numerous NW–SE-trending coseismic rupture strands with vertical displacements of 0.1–3 m, defining an actively deforming damage zone, while at S2, no surface rupture is preserved due to thick alluvial cover. Inverted ERT models at both sites reveal low-resistivity anomalies associated with fractured, water-saturated materials and fault gouge; however, conventional inversions smooth sharp resistivity gradients, limiting structural interpretation. Applying k-means clustering as a post-inversion segmentation tool transforms continuous resistivity fields into discrete lithological and structural domains. The Elbow method is used to determine the optimal number of clusters to improve interpretability. The clustered models sharpen structural discontinuities, delineate fault cores and subsidiary strands at S1, and reveal concealed deformation at S2 where surface evidence is absent. This integrated interpretation framework significantly enhances the resolution and interpretability of near-surface fault architecture within a crustal-scale thrust system. The approach is particularly effective for imaging buried fault segments and has important implications for seismic hazard assessment and land-use planning.
Noor et al. (Thu,) studied this question.