Improving Absolute Hypocenter Accuracy With 3D Pg and Sg Body‐Wave Inversion Procedures and Application to Earthquakes in the Central Alps Region

Abstract

Accuracy of hypocenter location, in particular focal depth, is a precondition for high‐resolution seismotectonic analysis of natural and induced seismicity. For instance, linking seismicity with mapped fault segments requires hypocenter accuracy at the sub‐kilometer scale. In this study, we demonstrate that inaccurate velocity models and improper phase selection can bias absolute hypocenter locations and location uncertainties, resulting in errors larger than the targeted accuracy. To avoid such bias in densely instrumented seismic networks, we propose a coupled hypocenter‐velocity inversion procedure restricted to direct, first‐arriving, mainly upper‐crustal Pg and Sg phases. On the basis of synthetic tests and selected ground‐truth events we demonstrate that a sub‐kilometer hypocenter accuracy can be achieved by regional‐scale, three‐dimensional Pg and Sg velocity models combined with dynamic phase selection and a non‐linear location algorithm. The tomographic inversion uses about 60,000 Pg and 30,000 Sg quality‐checked phases of local earthquakes in the Central Alps region. The derived models image the VP and VS structure of the Central Alps upper crust at unprecedented resolution, including small‐scale anomalies such as those caused by Subalpine Molasse units below the Alpine front. The relocation procedure is applied to more than 18,000 earthquakes and the relocated hypocenters reveal previously unrecognized seismogenic structures, for instance in the Swiss Molasse basin south of Bern. The ML 4.6 Urnerboden earthquake of 2017 is used as an example to demonstrate how the derived 3D velocity structure and relocated hypocenters can be jointly interpreted to constrain the lithology hosting upper‐crustal seismicity in the Central Alps.

New insights into the Rhône–Simplon fault system (Swiss Alps) from a consistent earthquake catalogue covering 35 yr

“…In addition, phase selection and network geometry can impact the hypocentre accuracy as well, leading to systematic errors, especially in focal depths (e.g. Diehl et al 2021b). As a result, a realistic assessment of hypocentre accuracy from the formal errors provided by single-event location algorithms is virtually impossible.…”

“…In addition, it solves for station-delay times, which compensate the effects of near-surface 3-D velocity heterogeneities. As the inversion performs a joint hypocentre determination (JHD), it is capable of detecting systematic errors and outliers in the arrival-time data (Maurer et al 2010;Kianimehr et al 2018;Rezaeifar & Kissling 2018;Diehl et al 2021b).…”

“…To compute an accurate P-wave velocity (V P ) model, P phases observed at epicentral distances >140 km were excluded, as they tend to bias the focal depth estimation as documented by Diehl et al (2021b) and are often contaminated by gross misidentifications. A series of JHD inversions were performed to identify potential errors in the arrival-time data, following the procedures of Diehl et al (2021b) (see Text S1 in the supplementary data for details). A total of 29 531 P phases passed this quality filter and were used in the minimum 1-D model calculation for the modern era.…”

“…Hutton et al 2010;Nanjo et al 2010) as well as location quality (e.g. Bondár et al 2004;Diehl et al 2021b). As a result, hypocentre parameters and associated location uncertainties reported in catalogues covering periods of several decades are highly inconsistent and therefore hamper high-resolution seismotectonic interpretation at scales of kilometres and below.…”

“…Diehl et al 2017;Küperkoch et al 2018). However, the improvement of absolute hypocentre location accuracy as well as its uncertainty assessment remain challenging, even with modern state-of-the-art location algorithms (Husen & Hardebeck 2010;Diehl et al 2021b). As documented by Diehl et al (2021b), hypocentre accuracy depends mainly on the velocity model used, and the targeted subkilometre scale requires highly accurate velocity models in combination with dense seismic networks.…”