Subsurface imaging of the regolith layer is an important tool for weathering zone characterization. For example, the extent of bedrock modification by weathering processes can be modelled by means of differing seismic velocities. We acquired a 360 m-long seismic profile in central Chile to characterise weathering at a semi-arid site. We used 87 3-component geophones, which continuously recorded ambient seismic noise for three days. The seismic line was centered at an 88 m deep borehole, providing core and downhole logging data for calibration. We extract Horizontal-to-Vertical Spectral Ratio (HVSR) curves along the seismic line to image the subsurface. Temporal analysis of the HVSR curves shows that the ambient noise vibrations recorded during the nighttime provide more stable HVSR curves. The trans-dimensional Bayesian Markov chain Monte Carlo (McMC) approach was used to invert the micro-tremor HVSR curves at each station to reconstruct 1D shear-wave velocity (Vs) models. The resulting individual 1D Vs models were merged to create a 2D Vs model along the linear seismic array in La Campana. The resulting Vs model shows an increase from 0.85 km/s at the surface to ca. 2.5 km/s at 100 m depth. We use the interface probability as a by-product of the Bayesian inversion to apply a more data-driven approach in identifying the different weathering layers. This method identified the boundary between saprolite and fractured bedrock at 42 m depth at the borehole, as evidenced by the interpretation of downhole logging data such as magnetic susceptibility. The resulting 2D Vs model of this site in Mediterranean climate shows a strong correlation between the interpreted weathering front at around 90-m depth and a higher precipitation rate in the study site compared to arid sites. The horizontal alignment of the weathering front indicates a correlation between the weathering front depth with topography and fractures in the bedrock.
