Assessing the variability in hydraulic fracturing-induced seismicity occurrence between North American shale plays

Abstract

We examine differences in the occurrence rates of hydraulic fracturing-induced seismicity (HF-IS) between different shale plays in North America. While previous studies have investigated variations in HF-IS prevalence within individual plays, no studies have compared HF-IS occurrence between different plays. Our study examines associations between hydraulic fracturing and induced seismicity across 12 shale plays: the Barnett, Bakken, Duvernay, Eagle Ford, Fayetteville, Haynesville, Horn River, Marcellus, Montney, Niobrara, Utica, and Woodford. We compile earthquake catalogues for each play using a combination of data from published studies and regional and national earthquake databases. For plays in the United States we compile well data from the FracFocus database, while for Canadian plays we compile well data provided by provincial regulators. We use a rate-change approach to identify potential cases of HF-IS, where increases in earthquake rates above a background level in spatial and temporal proximity to active wells is taken to indicate a likelihood that seismic events are induced by hydraulic fracturing. We find very large variations, of several orders of magnitude, in the occurrence rates of HF-IS between different plays: for some plays we find an average rate of one M ≥ 2.0 event per 3 wells, while some plays do not have any plausible cases of HF-IS despite hydraulic fracturing of many thousands of wells, and injection of several hundred million m3 of fluid. We interpret these variations with respect to the underlying geological conditions. We find that geomechanical connections into basement rocks may promote HF-IS, but there are plays without such connections that have also generated significant levels of induced seismicity. We find that HF-IS is more prevalent in plays with high pore pressure gradients and strike-slip to compressive stress regimes. A multiple linear regression finds statistically significant fit between these two parameters in tandem and the observed rates of HF-IS occurrence.