Robust Estimation of Rupture Properties at Propagating Front of Laboratory Earthquakes

Earthquake rupture propagation along a fault interface depends on the governing friction law. Although a variety of friction laws have been proposed based on measurements made in the lab, uncertainty remains regarding to what extent those measurements capture the true rupture properties. By comparing laboratory earthquakes with numerically simulated ruptures, we find that certain properties inside the rupture front may not always be accessible to direct measurements in the real world, due to a lack of resolution caused by the shrinking size of rupture front zone (Lorentz contraction). Instead, we propose an indirect method to robustly estimate rupture properties of laboratory earthquakes, by simultaneously fitting both the trajectory of rupture front and the detailed strain waveform shape using numerical synthetics. Under the current data resolution and on the basis of a slip-weakening friction model, our method can provide upper bound constraints on slip-weakening distance, in addition to regular constraints on fracture energy. Application of this new method to 26 laboratory earthquakes shows that rupture properties (directivity, propagation speed, fracture energy, and frictional strength) can fluctuate over earthquake cycles. Post hoc examination further suggests that our direct measurements largely reflect the deformation features outside the rupture front. As the same issue on the feasibility of direct measurements inside the rupture front may exist in other studies, our work questions the accuracy of previously derived friction laws, addresses the need for calibrating measurement location and resolution, and provides a means for examining the self-consistency of estimated rupture properties. Plain Language Summary The physics at propagating front of earthquake ruptures is the key to understanding the mechanism of earthquakes. While it is generally assumed that rupture properties can be directly estimated by measurements made close to the fault during laboratory earthquakes, our careful analysis shows that this is not always the case, because the size of rupture front zone can dynamically evolve and decrease below the resolution level of direct measurements. Instead of pursuing direct estimation, we assume a friction model on the fault and aim to infer rupture properties in the framework of the assumed model. By fitting multiple features of data close to the fault using synthetic waveforms generated by numerical simulations, we can constrain the assumed model and improve the estimation of rupture properties. Application of this method to 26 laboratory earthquakes reveals diversity and fluctuation of rupture properties over earthquake cycles. Our work provides an alternative way to study fault friction and dynamic ruptures, which can be further used to enhance the physical understanding of rupture behaviors in the lab and in nature.