A Test of the Earthquake Gap Hypothesis in Mexico

Allen L. Husker, Maximilian J. Werner, & Miguel Santoyo

Submitted August 16, 2021, SCEC Contribution #11642, 2021 SCEC Annual Meeting Poster #267

The seismic gap hypothesis to predict earthquakes holds that the seismic cycle implies quasi-regular, characteristic earthquakes that follow the build-up and release of stress. Mexico's subduction zone produces many earthquakes, and the hypothesis has been widely cited in Mexico to predict the location of future earthquakes. The Guerrero seismic gap has received the greatest attention, like the Tokai gap in Japan. However, no analysis of the outcome of any predictions of the hypothesis in Mexico has been done to-date. This work evaluates the prediction by Nishenko and Singh (1987), who provided relatively well-defined probabilities, areas (segments) and timeframes that allow for an evaluation. The time frames were for 5 years, 10 years and 20 years after 1986, and there were a total of 11 segments evaluated. The probabilities rely on precise occurrence times of past characteristic earthquakes to define segments, but we find that their definition of characteristic earthquakes is imprecise and at times appears arbitrary. The one defining feature of the earthquakes that does not change is that they are Ms greater than or equal to 7.4. We discuss some of their other decisions building their catalog to explain how we evaluate their prediction. We create catalogs of earthquakes based on their probabilities of earthquake occurrence for each segment. We also generate null model earthquake catalogs using the average number of earthquakes that occur in the subduction zone but are randomly distributed along the segments. We find that the null model performed better than the seismic gap hypothesis prediction. The catalogs generated with the Nishenko and Singh (1987) model only correctly predicted the outcome in 48% of the segments at the longest time frame of 20 years compared to 73% of the segments for the null model. The gap hypothesis also greatly over-predicted the total number of segments to have a characteristic earthquake: Ms greater than or equal to 7.4 earthquakes were predicted to occur in 6 of the 11 segments over the 20-year timeframe, but only 1 actually had an earthquake. We conclude that the gap hypothesis performed worse than a uniformly random forecast. This suggests that earthquake activity does not follow a simple cycle of stress build-up and release in segments and that other factors must influence the so-called seismic cycle.

Husker, A. L., Werner, M. J., & Santoyo, M. (2021, 08). A Test of the Earthquake Gap Hypothesis in Mexico. Poster Presentation at 2021 SCEC Annual Meeting.

Related Projects & Working Groups
Earthquake Forecasting and Predictability (EFP)