Application of apparent strong earthquake activity pattern in operational earthquake forecasting

Abstract Consulting the Catalogue of the International Seismological Centre (ISC), for the period 1904–2016 to detect the occurrence of potentially damaging earthquakes we observed that in most cases, when a high magnitude earthquake occurs (magnitude of at least 6.5), there is an increased probability that a similar high magnitude earthquake will occur within a relatively short period of weeks (less than a year). This occurs when the two events are located on latitudes of practically the same absolute value. This Apparent Strong Earthquake Pattern (ASEP) is observed in about 90% of all earthquakes of magnitudes greater than, or equal to 6.5. ASEP is evident in a high percentage of events and remains practically unchanged even after shuffling the dates of occurrence of earthquakes several times. This statistical observation is not surprising when considering the geographic distribution of earthquake epicenters and activity rates in regions of relatively frequent high magnitude earthquakes. Nevertheless, it leads us to consider the possibility of introducing ASEP in earthquake risk management and Operational Earthquake Forecasting (OEF). The main objective of this study is to assess the potential of estimating the location, magnitude, and time of occurrence of a strong earthquake after the occurrence of a similar strong earthquake in a distant area, when subsequent events conform to the apparent event pattern of strong earthquakes. The effectiveness of ASEP is quantified in terms of the ratio between the success rates obtained by applying ASEP and the probability of a randomly occurring earthquake of magnitude M in a given seismic zone within dT weeks. The effectiveness of ASEP was initially evaluated through simulations from the data from earthquake catalogues. The simulations reveal, as expected, that when the earthquakes in catalogue A are independent of the earthquakes in catalogue B, and when the occurrence of earthquakes in each catalogue are random and obey laws of Poisson distribution, the effectiveness is always lower than 1, i.e., the chance of a successful random guess is always higher than the probability of successful forecasting that follows ASEP. The opposite is observed when applying the ASEP schema to real cases. After arbitrarily choosing ten pairs of seismogenic zones, the success rates of the apparent earthquake pattern forecasts always yield a higher forecast probability, sometimes considerably higher than would a random guess. It is also observed that when selecting pairs of active zones which fail to obey ASEP requirement about the locations of events, the success rate of forecasting is, in most tested cases, zero. i.e., similar to what is observed in the simulations. Subsequently, the proposed forecasting scheme based on ASEP may be useful for OEF applications which in turn could be considered in earthquake risk management programs. These observations also suggest that the temporal behavior of strong earthquakes is not purely random.