“Cellular Seismology”: Does Past Seismicity Delineate Zones Where Future Large Earthquakes are Likely to Occur?

Alan Kafka
Weston Observatory
Department of Earth and Environmental Sciences
Boston College

Although earthquake prediction is not (currently?) possible, people of course still want to plan as well as they can for the impact of future large earthquakes. An obvious component of such planning, if it were possible, would be knowing where future large earthquakes are likely to occur.

Plate tectonics provides a very successful model for describing where the vast majority of earthquakes occur, i.e., at or near plate boundaries. The development of the theory of plate tectonics in the 1950s and 60s depended to a large extent on the fact that most earthquakes repeatedly occur within the same zones, zones that we now know are plate boundaries. But one of the fundamental questions that still remains to be answered is: What causes some (occasionally quite large) earthquakes to occur outside of those zones, in locations that are not near plate boundaries?

Figure 1: Global seismicity for two 19-year time periods, 1973-1991 and 1992-2010 (NEIC, magnitude 5 and greater). For any time period long enough to represent a good statistical sample, the vast majority of earthquakes map out plate boundaries, demonstrating that on a global scale most earthquakes keep occurring in the same places.

These non-plate boundary (NPB) earthquakes are either strictly “intraplate” earthquakes (i.e., those that occur deep in the interiors of plates) or earthquakes that occur in diffuse zones near, but not on, plate boundaries. Although we know that, on a global scale, most earthquakes repeatedly occur in well-defined zones, we don’t know if that is the case for NPB earthquakes. Do NPB earthquakes repeatedly occur in specific zones, or do the zones they occur in “migrate” over time from one place to another such that they might eventually occur essentially anywhere in plate interiors? If NPB earthquakes occur in specific zones that remain stationary over time, then given a long enough record of past seismicity, we should be able to discern where large NPB earthquakes will occur in the future. But if NPB seismicity migrates over time, then past seismicity will not be a useful indicator of where future large NPB earthquakes will occur.

If we are to ever understand the cause of NPB earthquakes, we will have to resolve this question of whether or not the locations where these NPB earthquakes are occurring are persistent. And we will have to determine the characteristics of Earth processes in those zones that are seismically active, as compared to zones that aren’t. My attempt to answer these types of questions has led me to invent the method known as “Cellular Seismology” (CS). Details of my CS research have been published in scientific/technical papers (e.g., Kafka, 2002; Kafka, 2007; and Kafka and Ebel, 2011). What follows here is my explanation of the essence of CS in (what I hope to be) simple, non-technical language.

CS is an intentionally simple method of systematically investigating the relationship between locations of past and future earthquakes in a given region. The name “Celluar Seismology” was chosen because it is analogous to a cellular phone system, with past earthquakes acting analogously to a cell phone tower. The cell tower is associated with a circular zone, extending some radius away from the tower, within which cell phones can receive a signal from the tower. Analogously, we envision that some circular zone surrounding the epicenter of a past earthquake is a zone that presumably has the necessary geophysical characteristics to generate future earthquakes.

CS involves analyzing what seismologists refer to as “earthquake catalogs”, i.e., databases of times, locations and magnitudes of earthquakes in a given region. To implement CS we construct circles of a given radius around each epicenter in an earthquake catalog (which we call the “Pre-CAT”), and investigate the percentage of later-occurring earthquakes (in what we call the “Post-CAT) that were located within that radius of at least one previous earthquake. These Post-CAT earthquakes that occurred near a Pre-CAT earthquake are referred to as “hits.”

We then systematically analyze the observed percentages of hits in an attempt to discern the extent to which patterns emerge in the relationship between locations of past and future earthquakes. We have found what seems to be a stable pattern of at least 2/3 to 3/4 of future earthquakes occurring near past earthquakes in most regions, and we are now investigating how the patterns compare and contrast for different regions.

Figure 2: Hypothetical region, showing how Cellular Seismology works. When a red (Post-CAT) earthquake occurs within a green zone, i.e., a region surrounding a Pre-CAT earthquake, that Post-CAT earthquake is referred to as a “hit.”

Figure 3: Cellular Seismology results for Northeastern United States. This analysis is shown for the case of earthquakes with magnitude of 3 and greater. The circle radius around the Pre-CAT earthquakes is chosen so as to fill 33% of the map area within the blue polygon. For 33% map area, there are 84% hits for this case.

Figure 4: Cellular Seismology results for California. This analysis is shown for the case of earthquakes with magnitude of 4 and greater. The circle radius around the Pre-CAT earthquakes is chosen so as to fill 33% of the map area within the blue polygon. For 33% map area, there are 93% hits for this case.

This CS research has been occurring within the context of a resurgence of interest among some seismologists in earthquake forecasting. It is well accepted among most seismologists that “earthquake prediction” (which refers to predicting the specific time, place and magnitude of an earthquake) is not currently possible. But some level of “earthquake forecasting” (which refers to more long-term estimates of the probability of earthquakes occurring within some region) is considered to be a reasonably attainable goal.

We applied CS to earthquakes in California where there have been recently published earthquake forecasts based on past seismicity. The results are somewhat counter-intuitive. We are finding (so far?) that there does not appear to be anything in the record of past seismicity that is any more predictive of where future earthquakes are likely to occur other than the simple notion that future earthquakes tend to occur near past earthquakes.

The underlying philosophy behind CS is to prefer the simplest (most parsimonious) approach to analyzing any phenomenon of interest. We tried using more complicated approaches to analyzing the relationship between past seismicity and later-occurring earthquakes, and found that more complicated methods showed insufficient gain in predictability to warrant any more complicated approach to this problem than the simple CS method. Thus, we argue that before invoking a complicated solution to predicting locations of future earthquakes, that complicated approach should be checked to see if it performs any better than CS, which we think is a reasonable, least astonishing, hypothesis.

References:

• Kafka, A.L (2002). Statistical analysis of the hypothesis that seismicity delineates areas where future large earthquakes are likely to occur in the central and eastern United States, Seismological Research Letters, 73(6), 990-1001.
• Kafka, A.L. (2007). Does seismicity delineate zones where future large earthquakes will occur in intraplate environments?, In Continental Intraplate Earthquakes: Science, Hazard, and Policy Issues, Geological Society of America Volume, Geological Society of America Special Paper 425, Edited by S. Stein and S. Mazzotti, 35–48, doi: 10.1130/2007.2425(03).
• Kafka, A.L. and J.E. Ebel (2011). Proximity to past earthquakes as a least astonishing hypothesis for forecasting locations of future earthquakes, Bulletin of the Seismological Society of America, 101(4), August 2011, doi: 10.1007/s00024-011-0315-1.