作者: J. Douglas Zechar , Jiancang Zhuang
DOI: 10.1111/J.1365-246X.2010.04666.X
关键词: Geology 、 Data mining 、 Tracing 、 Earthquake prediction 、 Meteorology 、 Variable (computer science) 、 Pattern recognition (psychology) 、 Induced seismicity 、 Probabilistic forecasting 、 ALARM 、 Reference model
摘要: SUMMARY In 2003, the reverse tracing of precursors (RTP) algorithm attracted attention seismologists and international news agencies when researchers claimed two successful predictions large earthquakes. These had begun applying RTP to seismicity in Japan, California, eastern Mediterranean Italy; they have since applied it northern Pacific, Oregon Nevada. is a pattern recognition that uses earthquake catalogue data declare alarms, these alarms indicate expects moderate following months. The spatial extent highly variable each alarm typically lasts 9 months, although may extend time space. We examined record outcomes prospective application began, this paper we report on performance date. To analyse predictions, used recently developed approach based gambling score, simple reference model estimate prior probability target earthquakes for alarm. Formally, believe investigators did not rigorously specify first ‘successful’ advance relevant earthquakes; because issue contentious, consider analyses with without alarms. When included contentious demonstrate statistically significant skill. Under stricter interpretation, are marginally unsuccessful.