3D structure and deformation evolution of a large deep-seated toppling revealed by GMM-based multi-source geophysical integration

Exploring the internal structure of large landslides is crucial for understanding their deformation mechanisms and conducting stability assessments. However, traditional exploration methods, such as drilling, provide only localized information and fail to reflect the spatial continuity of subsurface structures. Single geophysical methods also face challenges in accurately characterizing deep-seated structures due to inversion non-uniqueness and interpretative ambiguity. Multi-source geophysical data fusion is considered an important approach to reduce ambiguity and improve modeling reliability, but existing research largely focuses on shallow landslides, lacking effective methods for the three-dimensional reconstruction of large deep-seated rock landslides. Taking the Tizicao deep-seated toppling on the eastern edge of the Tibetan Plateau as an example, this study proposes a multi-source geophysical data fusion modeling method based on the Gaussian mixture model (GMM). This method comprehensively utilizes electrical resistivity tomography (ERT), multi-channel surface wave exploration (MASW), the horizontal and vertical spectral ratio method (HVSR) for ambient noise, and UAV photogrammetry to achieve the fusion and classification of multiple parameters such as resistivity, shear wave velocity, and structural depth. By automatically partitioning the geophysical feature space using GMM, a three-dimensional model of the Tizicao toppling is constructed. The three-dimensional model is highly consistent with the borehole results, verifying the reliability of the fusion modeling method. In addition, the deep-seated structure revealed by the three-dimensional model plays a key controlling role in the initiation of slope instability. Overall, the proposed GMM-based multi-source geophysical fusion method not only enables accurate reconstruction of the internal structure of large deep-seated rock landslides but also provides a new technical pathway for mechanism analysis and hazard prediction of large deep-seated landslides.

Keywords: Deep-seated toppling; Multi-source geophysical integration; Gaussian Mixture Model (GMM); 3D structural modeling; Deformation evolution.