Can Earthquake magnitude be predicted at rupture onset? insights from scaled seismotectonic models

The largest earthquakes on planet Earth occurred along the frictional interface between subducting and overriding plates (i.e., the megathrust) at convergent margins. Some of these destructive events have occurred over the last 20 years, such as the 2004 Mw 9.0 Sumatra‐Andaman, 2010 Mw 8.8 Maule, and 2011 Mw 9.0 Tōhoku‐Oki earthquakes. These large events are among the most devastating expressions of Earth's dynamics and, along with tsunamis, they represent a major hazard to society. Therefore, it is crucial to understand to which extent it is possible to predict the final size of a large rupture from the early stage of its propagation. We studied analog earthquakes in an apparatus - Foamquake (Mastella et al., 2022) - in which we used foam rubber to reproduce the upper plate and a 1 cm thick layer of granular materials to reproduce the subduction channel. The set-up is made of an elastic foam wedge with a dimension of 145 × 90 × 20 cm3 (i.e., the overriding plate analog) that overlies a planar, 10° dipping, rigid plate. Along the plate, a basal conveyor belt is driven with constant velocity (0.01 cm/s), reproducing a steady, trench-orthogonal subduction. To constrain the dynamics of analog earthquakes, we used a network of 11 Micro-Electro-Mechanical Systems (MEMS) accelerometers, distributed on the model surface, and measured the evolution of the trench orthogonal component of acceleration at 1 kHz. Additionally, we also used a top-view high-resolution camera (100 Hz), that allows us to derive surface displacements via Particle Image Velocimetry (PIV), that enables characterization of the final static rupture properties, while MEMS monitoring resolves the temporal evolution of spatiotemporal slip. We report 21 models with different frictional configurations of the analog megathrust, including asperities and barriers of varying dimensions, to produce thousands of events with different magnitudes. MEMS monitoring allows for characterization of the Source Time Function (STF) of each event. Preliminary analysis of the STFs indicates a weak correlation  (i.e., R²<0.2) between the moment accumulated over different time windows during the early stages of rupture propagation and the final size of individual events. These results contribute, from an experimental perspective, to the ongoing debate on the stochastic versus deterministic nature of earthquake rupture growth.