Simulating Earthquake Ground-Motion with Salvus

The simulation of earthquake ground-motion is an important task in seismic hazard assessment. In particular, simulations can help to fill the data gap that naturally exists due the rare occurence of high-magnitude earthquakes in regions of moderate or low seismicity. Large pools of simulated data can thus be used to establish so-called ground-motion prediction equations (GMPEs) that enable the estimation of statistical properties of ground-motions (e.g., their expected mean and standard deviations) for a potential site at risk. This is crucial for the estimation of seismic hazard.

The Salvus software suite provides the necessary tools to accurately simulate earthquake ground-motions:

• Arbitrary complex inputs: 3D velocity models, optionally including attenuation and anisotropy.
• Topography: Salvus can accurately simulate the effects of topography on seismic wave propagation.
• Fast computations: The massively parallel implementation of Salvus (on both CPUs and GPUs) enables the rapid simulation of ground-motions for large catalogs of potential earthquakes and input models.
• Earthquake source models: Salvus can simulate the seismic response for simple point sources as well as kinematic fault ruptures.

Intensity Measures Implemented in Salvus

Salvus enables the user to output accurate time histories (i.e., seismograms) at arbitrary locations inside the simulation domain. In addition to that, Salvus comes with convenience functions that enable the user to automatically compute common intensity measures used in seismic risk analysis, including:

• Peak Ground Velocity (PGV) and Peak Ground Acceleration (PGA).
• Arias Intensity.
• (Pseudo-)Spectral Acceleration, Velocity, and Displacment.
• Significant Duration.
• Macroseismic Intensity.

Case Study: Simulating The Historic Basel Earthquake

In 1356, a significant earthquake occured in Basel. Intense shaking destroyed all churches and castles within a 30 km radius. The earthquake's moment magnitude was estimated to lie in the range of 6.0 to 7.1. Here, we use Salvus to simulate the ground motions throughout Switzerland caused by such an earthquake and illustrate how Salvus can accurately handle increasingly complex input models. In the following we assume the earthquake had a magnitude of 6.6.

We simulate the earthquake on a multi-segment fault located in Basel that, on average, strikes NNE-SSW and dips 75° to the East. Salvus is capable of reading common earthquake fault rupture formats and converting those to the required Salvus sources.

Figure: Salvus visualization of the fault used to simulate the historic Basel earthquake.

In a first iteration, we simulate the Basel earthquake using a simple velocity model that only varies with depth and has a flat topography. Simulations are performed up to a frequency of 1 Hz. On a single GPU (NVIDIA RTX 4000 SFF ADA), the simulation for a domain spanning over all of Switzerland takes about 2 minutes to complete. We then visualize the shaking intensity in terms of macroseismic intensity, derived from the peak ground velocity according to the relations provided by Faenza & Michelini (2010). In the near-source region, the intensity reaches degree IX.

Figure: 3D seismic velocity model (left) and simulated macroseismic intensity for a Magnitude 6.6 earthquake on the Basel fault (right).

It is straightforward to include topography in the Salvus simulations through the use of a digital elevation model. Salvus can automatically generate an unstructured simulation mesh that follows the topography. Repeating the previous simulation with realistic topography leads to significantly different ground motion intensities, especially in the alpine regions. This demonstrates the importance of taking topography into account when simulating earthquake ground-motions.

Figure: 3D seismic velocity model with added topography (left) and resulting macroseismic intensity (right).

Additionally, spatial variations in the seismic velocity can lead to focusing and defocusing effects of the wavefield, which in turn alter the shaking intensity. Here, we adapt a realistic 3D velocity model of Switzerland provided by Diehl et al. (2021). to evaluate such effects. Note that the distribution of macroseismic intensity is again significantly altered. This can prove critical for the evaluation of seismic hazard.

Figure: Realistic seismic velocity model of Switzerland (left) and resulting macroseismic intensity (right).

Time histories (seismograms) can be readily extracted from the Salvus simulation output at arbitrary locations inside the simulation domain. The following time histories were computed for a seismic station located close to Basel for the three simulations described above, illustrating the differences in the local site response when considering increasingly complex input models.

Figure: Simulated time histories at a station close to Basel.

For further analysis with third-party seismic hazard analysis software, Salvus allows users to write the simulated intensity measures to a simple spreadsheet at user-specified locations inside the simulation domain. Additionally, the spreadsheets will include information on the earthquake source and the velocity model used in the simulations.

Figure: Optional spreadsheet output of shaking intensity measures.

References

• L. Faenza & A. Michelini (2010): Regression analysis of MCS intensity and ground motion parameters in Italy and its application in ShakeMap. Geophysical Journal International. 180(3), pp. 1138-1152. doi: 10.1111/j.1365-246X.2009.04467.x
• T. Diehl et al. (2021): Improving Absolute Hypocenter Accuracy With 3D Pg and Sg Body-Wave Inversion Procedures and Application to Earthquakes in the Central Alps Region. Journal of Geophysical Research: Solid Earth, 126, e2021JB022155. doi: 10.1029/2021JB022155

Funding

This work has received funding from the European Union's Horizon Europe research and innovation programme under grant agreement no. 101058129 (DT-Geo). Visit the DT Geo website for more information.