Seismicity-based rate forecast for induced earthquakes in the central U.S.

Abstract:

The U.S. Geological Survey is developing a one-year seismic hazard model for induced earthquakes. Zones of induced seismicity have been defined by reviewing the current scientific literature, and by identifying unusual patterns of seismicity. We focus our calculations on earthquakes that have occurred in Oklahoma, southern Kansas and northern Texas (OK-KS_S-TX_N). This region has a high activity rate, and has produced a large number of earthquakes for the likelihood tests. We apply likelihood tests to smoothed seismicity models in order to test the ability of seismicity-based rate models to predict the spatial distribution and absolute rate of induced earthquakes.

Since 2009, the OK-KS_S-TX_N region has experienced an over 200-fold increase in monthly activity rates. Regional activity rates exhibit monthly variations of up to 50 percent, as well as strong spatio-temporal variations. Models were computed from 6- and 12-month sub-catalogs, dating from 2009-2014 and are tested against earthquakes that occurred in the first six months of 2015; we also test the effect of various smoothing distances (5-50 km). Likelihood results show a strong correlation between information gain and the use of seismicity from the most recent part of the earthquake catalog. Information gains from the smoothed seismicity models developed from 2013 and 2014 are about five times better than the smoothed seismicity models developed from the earthquakes that occurred from 2009-2012. Because we use the rate-normalized formulation for the likelihood calculation, these results also imply that the spatial distribution of recent seismicity is significantly different than the seismicity occurring as recently as three years ago. Our results indicate that the one-year forecast model will have the greatest information gain when it is based on the prior one or two years of seismicity.