Evidence of Strain Localization Preceding Rock Failure: Insights From Laboratory and Physics-Based Poroelastic Models

At present, a reliable method for forecasting earthquakes has not been developed yet, as the physical mechanisms that generate them are very complex and still not completely understood. To overcome the difficulties of retrieving direct observations and measurements in the field, here we employ laboratory and numerical models to investigate and better understand strain localization preceding mainshocks.

We perform a failure test on an intact and dry sample of Berea sandstone confined at 20 MPa with a triaxial machine (LabQuake). Employing in-house developed, conical-type and fully calibrated piezo-electric transducers (PZT), we are able to investigate the acoustic emission (AE) clouds by relocating the single events and by computing their focal mechanisms and scalar seismic moments. The PZT sensors are also used actively to allow for the construction of inhomogeneous and anisotropic velocity models. We further employ distributed strain sensing (DSS) with optical fibers to capture the heterogeneous spatial distribution of the surface strain by gluing the fibers on the sample surface. We observe AE clustering in two regions located at the top and bottom of the rock specimen throughout the majority of the experiment. As the test approaches the main failure, AE localize at one side of the sample in the lower half before obliquely propagating upwards by forming a macro-fracture. Surface strain heterogeneities are detected during the experiment, and regions of higher extensional strain correlate in time and space with rock volumes experiencing high AE activity. Numerical simulations, which are conducted using a two-dimensional continuum-based and fully coupled seismo-hydro-mechanical poro-visco-elasto-plastic modelling tool (H-MEC), are validated with both AE and DSS data. The combination of laboratory and numerical investigations allows us to individuate and study physical mechanisms (e.g., visco-plastic compaction of pores and shear banding) that explain the processes responsible for both surface strain concentration and the generated AE clouds. These findings suggest that the deformation in the interior of the sample is mainly occurring inelastically and is localized along an obliquely forming shear band. We estimate the partitioning between seismic and total deformation to be ~0.09 % and this effectively confirms the previous evidences related to the irreversible localization of strain within the rock specimen.