The recent steep increase in seismicity rates in Oklahoma, southern Kansas, and other parts of the central U.S. led the U.S. Geological Survey (USGS) to develop, for the first time, a probabilistic seismic hazard forecast for one year (2016) that incorporates induced seismicity. In this study, we explore a process to ground-truth the hazard model by comparing it with two databases of observations: Modified Mercalli Intensity (MMI) data from the “Did you feel it?” (DYFI?) system, and peak ground acceleration (PGA) values from instrumental data. As the 2016 hazard model was heavily based on earthquake catalogs from 2014-2015, this initial comparison utilized observations from these years. Annualized exceedance rates were calculated with the DYFI? and instrumental data for direct comparison to the model. These comparisons required assessment of the options for converting hazard model results and instrumental data from PGA to MMI for comparison with the DYFI? data. In addition, to account for known differences that affect the comparisons, the instrumental (PGA) and DYFI? data were declustered, and the hazard model adjusted for local site conditions. With these adjustments, examples at sites with the most data show reasonable agreement in the exceedance rates. However, the comparisons were complicated by the spatial and temporal completeness of the instrumental and DYFI? observations. Furthermore, most of the DYFI? responses are in the MMI II-IV range, whereas the hazard model is oriented toward forecasts at higher ground motion intensities, usually above about MMI IV. Nevertheless, the study demonstrates some of the issues that arise in making these comparisons, thereby informing future efforts to groundtruth and improve hazard modeling for induced seismicity applications.
Congratulations to Isabel on a nice paper! Check out the details here.
Update (as of 12/7/17): this paper is now available online at the journal's website!