GSA Annual Meeting in Seattle, Washington, USA - 2017

Paper No. 29-14
Presentation Time: 11:45 AM

PERFORMANCE OF USGS ONE-YEAR EARTHQUAKE HAZARD MAP FOR NATURAL AND INDUCED SEISMICITY IN THE CENTRAL AND EASTERN UNITED STATES


BROOKS, Edward M1, STEIN, Seth1, SPENCER, Bruce D.2 and SALDITCH, Leah1, (1)Earth and Planetary Sciences, Northwestern University, 2145 Sheridan Road, Evanston, IL 60208, (2)Department of Statistics and Institute for Policy Research, Northwestern University, Evanston, IL 60208, eddie@earth.northwestern.edu

Seismicity in the central United States has dramatically increased since 2008 due to the injection of wastewater produced by oil and gas extraction. In response, the USGS created a one-year probabilistic hazard model and map for 2016 to describe the increased hazard posed to the central and eastern United States. Using the intensity of shaking reported to the “Did You Feel It?” system during 2016, we assess the performance of this model. Assessing the performance of earthquake hazard maps for natural and induced seismicity is conceptually similar but has practical differences. Maps that have return periods of hundreds or thousands of years— as commonly used for natural seismicity— can be assessed using historical intensity data that also span hundreds or thousands of years. Several different features stand out when assessing the USGS 2016 seismic hazard model for the central and eastern United States from induced and natural earthquakes. First, the model can be assessed as a forecast in one year, because event rates are sufficiently high to permit evaluation with one year of data. Second, because these models are projections from the previous year thus implicitly assuming that fluid injection rates remain the same, misfit may reflect changes in human activity. Our results suggest that the model was very successful by the metric implicit in probabilistic hazard seismic assessment: namely, that the fraction of sites at which the maximum shaking exceeded the mapped value is comparable to that expected. The model also did well by a misfit metric that compares the spatial patterns of predicted and maximum observed shaking. This was true for both the central and eastern United States as a whole, and for the region within it with the highest amount of seismicity, Oklahoma and its surrounding area. The model performed least well in northern Texas, over-stating hazard, presumably because lower oil and gas prices and regulatory action reduced the water injection volume relative to the previous year. These results imply that such hazard maps have the potential to be valuable tools for policy makers and regulators in managing the seismic risks associated with unconventional oil and gas production.