Probabilistic earthquake location in complex three‐dimensional velocity models: Application to Switzerland

Abstract

[1] We present a new approach to determine precise and reliable hypocenter locations in the tectonically complex region of Switzerland. A three-dimensional (3-D) P wave velocity model to be used for earthquake relocation is obtained by simultaneously inverting arrival times of local earthquakes for hypocenter locations and 3-D P wave velocity structure. A 3-D P wave velocity model derived from controlled source seismology (CSS) is used as an initial reference model. The final 3-D model thus combines all available information from both CSS and local earthquake data. The probabilistic, nonlinear formulation of the earthquake location problem includes a complete description of location uncertainties and can be used with any kind of velocity model. In particular, the combination of nonlinear, global search algorithms, such as the Oct-Tree Importance Sampling, with probabilistic earthquake location provides a fast and reliable tool for earthquake location. The comparison of hypocenter locations obtained routinely by the Swiss Seismological Service (SED) to those relocated in the new 3-D velocity model using a probabilistic approach reveals no systematic shifts but does exhibit large individual shifts in some epicenter locations and focal depths. We can attribute these large shifts in part to large uncertainties in the hypocenter location. Events with a low number of observations (<8) and no observation within the critical focal depth distance typically show large location uncertainties. Improved hypocentral locations, particularly for mine blasts and earthquakes whose routine hypocenter locations had been questionable, confirm that improved velocity model and probabilistic earthquake location yield more precise and reliable hypocenter locations and associated location uncertainties for Switzerland.

Real-time Earthquake Location Using Kirchhoff Reconstruction

“…Stronger peaks will occur at the appropriate depth when contributions from reflected arrivals are correlated. Husen et al (2003) provide the rule of thumb that resolution of the event depth requires at least one station within a distance equal to 1.5 times the source focal depth.…”

“…Global search algorithms have made it computationally practical to use probabilistic earthquake location and maintain a complete description of the uncertainties contributing to the solution (Lomax, 2000;Husen et al, 2003). These methods, however, are not well suited to real-time applications, because, in general, they do not deal with the association problem and hence are sensitive to interference by other earthquakes.…”

Local earthquake tomography of Scotland

“…This layer is poorly constrained, with few ray paths, and will couple strongly with the station corrections. Husen et al, (2003) state that "station delays are only meaningful if more than 10 observations per station exist", such corrections are marked in white in the figure and are not interpreted. Otherwise, the corrections seem reasonable with positive corrections in the Midland Valley and around the Solway Firth/Vale of Eden, indicating that the near surface seismic velocity close to these stations is lower than average for that part of the model.…”

The role of topography and lateral velocity heterogeneities on near-source scattering and ground-motion variability

“…We choose this particular set of models since T3 has the roughest topography and, therefore, the largest scattering potential. Following available studies for the alpine region in Switzerland (Husen et al 2003;Steimen et al 2003;Roten et al 2008), we adopt the 1D velocity model shown in Fig. 2.…”