Catalog errors and the M8 earthquake prediction algorithm

Abstract

The goal of earthquake prediction algorithm research efforts is to identify patterns in data that characteristically occur prior to large earthquakes. Most of these algorithms implicitly assume that variations in the data they examine are related to actual changes in physical processes occurring in the Earth. In the case of seismicity data, most of the observed variations are, in fact, related to changes in the system for detecting and reporting earthquakes and not to actual changes in the Earth. We have tested the earthquake prediction algorithm M8 using five different seismicity data sets for the southwest Pacific. We demonstrate that the alarms identified by M8 are strongly dependent on the seismicity catalog used. The alarms occur preferentially during times when magnitudes are relatively high because of systematic errors. These alarms cannot be reasonably interpreted as precursors to earthquakes or any other real change in the Earth without first understanding and correcting the errors in the seismicity catalogs. This study highlights the need for improving the quality of geophysical data, which is presently unsuitable for many studies related to earthquake prediction.