Spatial and Temporal Variation of Seismic Attenuation During Hydraulic Fracturing of a Sandstone Block Subjected to Triaxial Stress

Abstract

Q factors estimated with spectral ratios describe the attenuation experienced by seismic arrivals in their travel from the source(s) to the receiver(s). When Q is not the same along the travel path of each arrival, the formulation must incorporate two Q factors making the problem indeterminate. Here we show that by assuming that Q is a random variable, the parameters of its probability distribution can be recovered through a mapping workflow using the spectral ratios. An advantage of such a workflow is that uncertainties in source location and velocity model can also be incorporated by representing these elements as random variables. We apply such a methodology to an acoustic emission (AE) data set from a laboratory experiment of hydraulic fracturing in a Colton sandstone block. We assume that the probability distribution of Q is normal. Anisotropic mean (μQ) values correlate with the triaxial stress conditions imposed at the boundaries of the block, with higher stress corresponding with less attenuation. The μQ values estimated within time intervals in the direction perpendicular to the hydraulic fracture plane correlate with stress buildup at the borehole during fracture propagation and borehole pressure relaxation or drop during fracture closure. The standard deviations (σQ) reflect the attenuation heterogeneity within the sampled regions. An increase in heterogeneity is correlated with pressure buildup at the borehole, which we interpret to reflect a dynamic stress gradient formed from the hydraulic fracture toward the surroundings.

Deep Compressed Seismic Learning for fast location and moment tensor inferences with natural and induced seismicity

“…In both cases, the standard deviation was 4% of the mean. Sampling velocities in this way is meant to capture uncertainties in their variation that results from heterogeneities and stress-induced anisotropy in the rock [84]. While drawing propagation velocities from the probability distributions, it was ensured that the v P /v S ratio ranged within the interval (1.45, 2.0), which was empirically selected as reasonable.…”

Deep compressed seismic learning for fast location and moment tensor inferences with natural and induced seismicity

“…The method also takes into account the varying nature of the block velocities in response to the imposed stress variations 80 . This is a state-of-the-art method used in multiple previous projects [81][82][83][84] ; therefore, we use it as a benchmark to compare the location part of the solution from deepCoSeL.…”

Seismic and Microseismic Signatures of Fluids in Rocks: Bridging the Scale Gap

“…Assuming that wave attenuation (Q S or Q P ) in a heterogeneous rock formation is a random variable, Vera Rodriguez and Stanchits () show that the parameters of its probability distribution function can be retrieved through a mapping workflow using the Q S /Q P ratio. This allows for the estimation of Q from microseismic events induced by the hydraulic fracturing of a decimetric block of porous Colton sandstone in a true triaxial stress vessel in the laboratory, while accounting for uncertainties in source location and velocity model.…”