Assessments of the performance of the 2017 one-year seismic-hazard forecast for the central and eastern United States via simulated earthquake shaking data

As a result of wastewater injection from nonconventional oil and gas production, the central and eastern United States experienced a dramatic increase in seismicity. To better characterize the resulting hazard, the U.S. Geological Survey (USGS) began producing one-year seismic-hazard models intended to capture both natural and induced seismicity as of 2016. In its first year, we found that the map performed very well, demonstrating both a good match between the observed and expected number of exceedances, and between observed and predicted shaking. We repeat this analysis for the 2017 map, using “Did You Feel It?” data to explore the map's performance in different regions of the country. We find that the 2017 model performed well, but not as well as the previous year's model. We explore the likelihood of observing the performance seen in 2017, by simulating earthquake shaking realizations using the assumptions of the 2017 hazard model, including a- and b-values, locations of induced earthquakes, and ground-motion models (GMMs). These simulations indicate a low likelihood of this decrease in performance happening by chance if the assumptions in the hazard model were appropriate. Hence, it is likely that the USGS one-year seismic model's performance reflects a reduction in wastewater injection rates, possibly due to regulatory and economic pressures. Future models could benefit from better modeling how seismic rates may change year-to-year with variations in wastewater injection rates and locations, and improved GMMs.