High-resolution local earthquake tomography of seismogenic structures along the Rhone-Simplon fault zone (Swiss Alps)

In this study, we explore the potential to image the seismic velocity structure of moderate-sized, upper-crustal seismogenic fault zones by means of local earthquake tomography (LET) methods. The study region, located in the transition zone between the Central and Western Alps, represents one of the most seismically active and hazardous areas within the Alpine Arc. Over the past 500 years, several damaging earthquakes with Mw up to 6.2 are documented in historical earthquake catalogs in the vicinity of the Rhone-Simplon Fault (RSF), the dominant tectonic feature of this region. In particular, two major seismogenic structures are imaged by instrumental seismicity on either side of the RSF. To the north, seismicity occurs in the approximately NE-SW striking, 30–40 km long Rawil Fault Zone (RFZ). To the south, diffuse seismicity occurs within the hanging wall of the Pennine Basal Thrust, forming the Penninic Fault Zone (PFZ).

Owing to the dense instrumentation and above-average seismic activity in the study region, the Pg and Sg travel-time data recorded since the year 2000 by the Swiss Seismological Service is of exceptionally high quality and allows for high-quality 3D LET images of the uppermost crust, potentially imaging the damage zones of seismogenic faults. Relative double-difference locations of a recent earthquake sequence within the RFZ, on the other hand, indicate that the width of this fault zone is only on the order of 1 km. Imaging such narrow fault zones with standard LET methods therefore requires model parametrizations with grid spacing of few kilometers and less. Such dense grid spacing, however, poses several challenges in terms of model resolution and the reliability of LET inversion results, especially in less well constrained parts of the model.

To minimize such effects, we therefore tested and applied two different LET inversion strategies to derive 3D Vp and Vs models. The first strategy follows the common approach to compute a minimum 1D model as initial model for the 3D LET. The second strategy uses a coarser 3D regional LET model as initial model for the high-resolution 3D inversion. Synthetic tests suggest that a minimum image resolution of 5x5x3 km can be achieved with the current data, covering a region of about 125x125 km between 0 and 10-15 km depth. The 3D Vp and Vs models derived with the two initial-model strategies are remarkably similar within well resolved parts of the model. This similarity indicates that the anomalies in these parts are well constrained by the data and the solution is stable with respect to differences in the two initial models. On the other hand, the results suggest that results derived with the 3D initial model are more reliable in regions of reduced resolution. Additional synthetic tests were performed to document the potential resolution for hypothetical damage-zone scenarios and to support the interpretation of the derived models presented in this study.