The variations of elastic parameters and attenuation of rock under deformation are of great importance for a number of geophysical problems. Here, we estimate elastic parameters and their evolutions during laboratory rock deformation experiments, while developing a Monte-Carlo full-waveform fitting method. The transducer-transducer one-source one-station active seismic data of dry and water-saturated samples are obtained from Zaima and Katayama (2018, https://doi.org/10.1029/2018JB016377). The synthetic seismic data are modeled using the spectral element method. We first performed a trial-and-error estimate of the boundary conditions in order to suppress its influence on waveform matching. The synthetic seismic data are generated using equivalent homogeneous models with different combinations of elastic and anelastic parameters. We then compared the synthetics with the laboratory experimental data. Based on these comparisons, we obtain the time-lapse variations of seismic velocities and attenuation of rock samples during deformation, which are then interpreted as crack development. Our simultaneous estimation of elastic and anelastic parameters allowed us to detail the dynamics prior to the rock failure.

We quantitatively evaluate transducer-transducer one-source one-station active seismic waveform data, in order to monitor time-lapse changes of elastic and anelastic structure during deformation experiments in laboratory. The experiment data of dry and water-saturated sample are provided by Zaima and Katayama (2018, https://doi.org/10.1029/2018JB016377). A transducer receiver, at the mid-point of cylindrical rock sample, is located on the antipodal position of the transducer source, emitting compressional and shear waves. Due to the extremely underdetermined nature of inverse problem, we limit the number of unknowns to be four: global P- and S- wave velocities and their corresponding anelastic attenuation factors, which can represent the micro-cracks nucleation during the loading and before the appearance of the largest crack that causes the fracture. We first performed a trial-and-error search for a realistic boundary condition in three-dimensional seismic waveform modeling using spectral-element method, in order to fit the synthetic data with the observed waveforms. We then generated synthetic data for 6000 combinations of elastic and anelastic parameters, in order to conduct Monte-Carlo waveform inversion based on the cost functions using waveform misfit and zero-lag cross-correlation. We obtained the time-lapse changes in velocity and attenuation during the deformation, which are then linked to crack development. Compared with the wet experiment, the dry experiment has a larger change in both the velocity and attenuation. However, regardless of the configuration, global seismic wave speeds rise first and then decrease during the experiments. The quality factor shows roughly the same trend.