| Abstract |
The characterization of seismicity in areas of extraction and injection is becoming an increasingly important topic in public and seismological community discussions (see Majer, 2010 for example). The injection-seismicity-related work of Ake et al. (2005), Shapiro et al., (2010) and Shapiro and Dinske (2009) has shown that it may be possible to determine injection-induced earthquake magnitude exceedance probabilities as a function of injection parameters, site characteristics, and background seismicity rates. Here we extend and further validate the injection volume and earthquake magnitude exceedance relations outlined in Shapiro et al. (2010) with a more extensive Paradox Valley (Northern Colorado) injection and seismic dataset first documented in Ake et al. (2005). The predicted Mmax (1.2 at the 95% exceedance probability) is remarkably close to the observed Mmax (0.9 at the 92% exceedance probability) for the Paradox Valley 1991 14-day 11,000 m3 initial injection sequence. The predicted Mmax (5.7 at the 95% exceedance probability) and observed Mmax (4.3 at the 55% exceedance probability) for the Paradox Valley 1996-2000 >2,000,000 m3 injection sequence is arguably a less robust result in predicting injection-related ground motion exceedance probabilities. Nonetheless these results demonstrate that the methodology is useful for estimating a range of ground motion exceedance probabilities. Sensitivity testing reveals that the key parameters in these injection-related seismicity hazard estimates are seismogenic permeability (or hydraulic diffusivity), the maximum pore pressure necessary to create displacement along randomly-oriented cracks, Cmax, poroelastic uniaxial storage coefficient, and cumulative injection volume at the time of the end of injection; predicted induced Mmax is much less sensitive to crack density, background earthquake activity rate, and earthquake recurrence b value. The seismogenic permeability of Talwani et al. (2007) is not only important to determine the potential for onset of any induced seismicity, but it determines the potential reservoir volume which has the strongest effect on estimated Mmax. Consequently, while it is feasible to estimate the most important other parameters (Cmax and poroelastic uniaxial storage coefficient) through drilling, sampling, and lab testing prior to injection, real-time seismicity monitoring during initial injection is essential to confirm or update the seismogenic permeability to obtain robust estimates of Mmax. However, it is critical that these observations be evaluated in the context of well bore conditions and local transients that may not reflect conditions in the far field where larger events are likely to occur. |