Extrapolation of populations of small earthquakes to predict consequences of low- probability high impact events: The Pohang case study revisited

Abstract

The magnitude (MW) 5.5 Pohang, Korea, earthquake on 15 November 2017, induced by the Pohang Engineered Geothermal Systems (EGS) project, caused one fatality and ∼US$300 million of economic consequences. The Commission, appointed by the Korean Government to investigate this earthquake, has made public a release of data including magnitudes of the smaller earthquakes associated with the well stimulations. On the basis of this earthquake population, it has been proposed that a significant probability of such losses was predictable beforehand, and that the project should have been suspended, implying that its developer was remiss for not doing so. This argument depends on the low b-value estimated, ∼0.61. However, three factors are shown to contribute, individually or in combination, to inaccuracy of these magnitude determinations: the low recording bandwidth of the permanent seismograph stations in the area; miscalibration of the formula for determining local magnitudes in Korea; and the relation used to estimate magnitudes of smaller events from larger events by template matching. These factors all cause underestimation of magnitudes of the smallest events documented, resulting in underestimation of b-values. The true b-values are higher, being 1.12 for the earthquakes associated with the August 2017 stimulation of well PX-1; similar values are estimated for the other well stimulations. A consequence of this analysis is that the probability of any earthquake as large as MW = 5.5, predicted ahead of its occurrence by extrapolation using b-values, was much lower than has been claimed. This analysis highlights the need for agreed workflow specifications for reporting datasets like this, where the data might influence prosecution of EGS developers, as well as agreed specifications for acceptable economic risk arising from EGS projects.