k-Means clustering as tool for multivariate geophysical data analysis. An application to shallow fault zone imaging

We present the results of an integrated imaging approach for two-dimensional high-resolution magnetotelluric and seismic profiles. These were carried out in the seismically active intermontane basin of Pantano di San Gregorio Magno (southern Italy), along a line across the surface rupture of the 1980, M6.9, earthquake. We focus on the application of the post-inversion k-means clustering technique to the univariate resistivity and P-wave velocity models, which were obtained previously through independent inversions. Five cluster classes are recognized, allowing a joint two-dimensional section to be imaged in terms of homogeneous zones from a geo-structural point of view. Two distinct local relationships between electrical resistivity and seismic velocities are inferred. In this way, the hanging and footwall zones have been retrieved, and are characterized according to the different fracturing degrees. The case dealt with here can be viewed as a successful example of how cluster analysis can be a promising auxiliary tool that provides bridging towards the integration of distinct geophysical methods. •A multivariate dataset is analyzed (p-wave velocities and electrical resistivities).•Univariate models have been post-processed using a clustering procedure.•A unique lithological model has been finally obtained.•The presence of a hanging and the footwall fault zone has been inferred.•A linear relationship between velocities and resistivities has been deduced.