Human-Induced Earthquakes: The Performance of Questionnaire Schemes

Abstract

ABSTRACT Questionnaire schemes are commonly used to investigate whether or not certain earthquakes were induced by industrial activity. Such schemes are subjective and cannot be assumed to give the “right” answer in a scientifically rigorous sense. They only yield current expert opinion on how strongly existing data support an induced or natural cause. Work to optimize a standard generic questionnaire is ongoing. To this end, we designed and compared three schemes that produce measures of data quality and support for human induction. One scheme is a generalization of an existing questionnaire for assessing fluid-injection-associated earthquakes. A second scheme is purely subjective, and a third scheme is purely objective. Because questionnaires are opinion-dependent, different analysts produce different results. We tested the three schemes on 55 diverse cases from the Human-Induced Earthquake Database with the maximum magnitude earthquakes M 4.1–7.9. The results of three analysts correlate with each other at the r ∼ 0.4–0.9 level. Higher correlations were found between schemes than between analysts. A simple, rapid, five-question Likert scale correlated well (r = 0.79) with results from a sophisticated, time-consuming scheme. Measures of data quality were uncorrelated with Mmax, and support for human induction correlated weakly negatively with Mmax. One scheme identified an earthquake sequence not proposed to have been human-induced but that has induced characteristics. New mechanisms of induction are still being recognized, and it is important that questionnaire schemes do not preclude new developments in future.