Intermediate-term earthquake prediction using precursory events in the New Madrid Seismic Zone

Abstract Earthquake prediction may be possible for some mainshock events. The time-to-failure method described by Varnes (1989) and Bufe and Varnes (1990) uses precursory events (foreshocks) to define an accelerated energy release curve. By fitting an equation to the data, a predicted time of failure and magnitude can be calculated. Until recently, this method has been used in only a few studies in tectonically active areas, and for moderate- to large-magnitude mainshock events. Using the microearthquake network data set from the New Madrid Seismic Zone (NMSZ), which is reasonably complete for earthquakes of magnitude ≧1.5 in the area of interest, the method has yielded predicted values of past events as small as m b = 3.5. The network data set used in this evaluation covers the time interval from 29 June 1974 to 20 July 1995 for the NMSZ. There have been 36 earthquakes of magnitude ≧3.5 over the 21-yr period in which the network has been operating. Because precursory events are required for the application of the time-to-failure method, mainshocks that occurred before 1980 do not have enough data to adequately define the accelerated energy release curve. Therefore, we utilized the 26 earthquakes that occurred after 1980 and that had a magnitude ≧3.5. Sixteen of the 26 mainshock events were modeled. In most cases, the precursory sequences yielded predicted times of failure and magnitudes that were reasonably close to the actual mainshock values. The remaining mainshocks, including those occurring before 1980, could not be modeled due to either (1) not enough events to adequately define the precursory sequence or (2) interfering events that disrupt the accelerated energy release curve. In addition, two events were modeled from the Nuttli catalog (Nuttli, 1979) along with one that used a combination of both catalogs. Nineteen earthquakes with magnitudes ≧3.5 were evaluated using the time-to-failure method. The first calculation using the time-to-failure method gave predicted results with large error bounds, essentially no upper bound on the predicted magnitude. An empirical relationship between parameters has helped to constrain the range of the predicted magnitude and, to a lesser extent, the estimated time of failure. This relationship modifies the time-to-failure equation and yields predicted values for magnitudes that have an upper limit. Another empirical relationship suggests that the logarithm of the moment of the mainshock increases linearly with the logarithm of the size of the precursory event search diameter. The relative seismicity of the region also influences the optimum search diameter used to find precursory events. In addition to the evaluation of the acceleration sequences associated with the mainshocks, an analysis of the occurrence of false-positive acceleration sequences (acceleration sequences that do not end in a mainshock) was conducted. The preliminary false-positive analysis was conducted by randomly selecting potential mainshock locations. The results yielded a false-positive acceleration sequence occurrence rate of 2%. With the incorporation of the empirical relationships into the time-to-failure method, the potential for future intermediate-term earthquake predictions for the NMSZ is encouraging.