Exploring earthquake recurrence and nucleation processes with Foamquake and a variety of asperity configurations

The short seismic record with respect to the return time of large subduction earthquakes and the spatial fragmentation of available geophysical data represent unfavourable conditions for robust hazard assessment. Over the last decade, data from scaled seismotectonic models have become useful in filling the observational gaps of seismic and geodetic networks. Such models allow reproducing hundreds of analogue seismic cycles in a few minutes of experimental time and with the advantage of known and controllable boundary conditions. 

Here we present experimental results from Foamquake – an established 3D seismotectonic model, which simulates megathrust subduction. Recent technical advances in experimental monitoring have allowed us to include into our research a high-frequency camera to record model surface deformation at 50 Hz and a network of 5 accelerometers (located on the model surface) that measure the three components of acceleration at 1 kHz. To analyse the camera data, we used particle image velocimetry (PIV) to derive surface displacements, such as in a dense, homogeneously distributed geodetic network spanning updip to scaled depths that are often offshore and, therefore, typically under-monitored in natural subduction zones.

We performed 33 experiments exploring 10 different geometrical configurations of asperities along the analog megathrust. In particular, we varied the number of asperities, their size, location, and extra normal load. We observed that the rupture pattern of analogue earthquakes predictably changes as the extra normal load varies and the distribution of asperity configurations becomes more complex. Depending on the number and size of the asperities and the size of the barrier between them, we noticed different ratios between full and partial ruptures with different recurrence time (Rt) intervals. In some experiments we detected cascades of ruptures. We used the coefficient of variation (CoV) of recurrence time to quantify analog earthquakes periodicity. Most of our models display quasi-periodic analog earthquakes recurrence with CoV<0.5, but multi-asperity experiments with variable-size and extra normal load lean toward random behaviour as testified by CoV~0.8.

Future investigations include the following steps – exploring this great volume of data using machine learning, looking for spatial and temporal relationships between accelerometer and PIV displacements, and tracking in detail the aseismic processes that may precede and follow earthquake rupture.