Francisco Martínez Álvarez, co-author of the study and senior lecturer at the UPO, and Antonio Morales Esteban, co-author of the study and senior lecturer at the US, have spent time on the Iberian Peninsula studying patterns of behavior that appear before an earthquake occurs, and utilized clustering techniques to help forecast larger earthquakes more efficiently. 

"Using mathematical techniques, we have found patterns when medium-large earthquakes happen, that is, earthquakes greater than 4.4 on the Richter scale," said Álvarez. 

To do this, researchers used data on 4,017 earthquakes that were between a 3 and a 7 on the Richter scale compiled by the Instituto Geográfico. All earthquakes used in this data occurred between 1978 and 2007 on the Iberian Peninsula and in surrounding waters.

Researchers focused on two seismogenic regions that had the most data, which were the Alboran Sea and the Western Azores-Gibraltar fault region. They added clustering techniques to this data and observed the magnitude of seismic movement, the time that has passed since the last earthquake and the change in b-value from one earthquake to another. A high b-value means most of the earthquakes are small and the surrounding land has a low level of resistance, while the opposite is true for a low b-value.  

Through this method, Álvarez and Esteban found patterns that could allow for better forecasts of earthquakes. Their forecasts summarize both sensitivity, which is the probability that an earthquake will occur after the detected patterns occur, and specificity, which is the probability that an earthquake will not occur when none of the patterns have occurred.

Through this forecasting, researchers have found a sensitivity of 79.31 percent and a specificity of 90.38 percent for the Western Azores-Gibraltar fault region, and a sensitivity of 90 percent and a specificity of 82.56 percent for the Alboran Sea region. This means that the probability of an earthquake occurring in these regions immediately after the patterns is high.

"We have discovered the strong relationship between earthquakes and the parameter b-value, recording accuracy rates of more than 80 percent," said Esteban. "After the calculations had been performed, providing the circumstances and sequences we have determined to be forerunners occur, we obtain a significant success probability." 

The next step for these researchers is to apply other mathematical techniques to the same data in order to discover other common events and patterns that may help better their forecasting system. 

This study will be published in Expert Systems with Applications.