DisastersEvaluating California earthquake forecasts
The Southern California Earthquake Center held a competition among seven different earthquake forecasts; an analysis of the competition shows that earthquake prediction remains an imperfect science, but the best forecasts are about ten times more accurate than a random prediction; the findings should help researchers both develop better earthquake forecasts and improve their tools for assessing those forecasts
Earthquake prediction remains an imperfect science, but the best forecasts are about ten times more accurate than a random prediction, according to a study published 26 September in the journal Proceedings of the National Academy of Sciences (PNAS).
In the study, UC Davis researchers compare seven different earthquake forecasts (including their own) that were submitted to a competition organized by the Southern California Earthquake Center. The findings should help researchers both develop better earthquake forecasts and improve their tools for assessing those forecasts, said Donald Turcotte, a distinguished professor of geology at UC Davis and co-author of the paper.
A University of California-Davis release reports that the center launched the competition in 2005 based on a previous forecast published by the UC Davis group in 2001. Teams were invited to forecast the probability of an earthquake of magnitude 4.95 or greater, from 1 January 2006 to 31 December 2010, in almost 8,000 grid squares covering California and bordering areas.
During this time, thirty-one earthquakes struck in twenty-two grid squares, with the largest being the magnitude 7.2 earthquake just south of the U.S.-Mexican border in April 2010. All seven forecasts showed some utility in forecasting the locations of likely earthquakes: The best forecasts were about 10 times better than a random forecast.
The forecast submitted by the UC Davis group was the most accurate in picking the locations of the earthquakes, correctly labeling seventeen of twenty-two grids and giving the highest probability of an earthquake in eight of these 17. Using a different forecasting method, Agnes Helmstetter of UCLA and colleagues gave the highest average probability of an earthquake for all twenty-two affected grids, although it did less well at assigning a higher probability to grid squares where an earthquake occurred.
“Just as there are alternative ways to forecast earthquakes, there are also alternative ways to evaluate the success of the forecasts,” Turcotte said, noting that other publications evaluating the forecasts are expected.
The UC Davis group includes professors John Rundle and Turcotte, postdoctoral researcher James Holliday, and graduate students Ya Ting Lee and Michael Sachs. Also contributing were Chien-Chih Chen, National Central University, Taiwan; Kristy Tiampo, University of Western Ontario; and Andrea Donnellan of the Jet Propulsion Laboratory, Pasadena.
— Read more in Ya-Ting Lee et al., “Results of the Regional Earthquake Likelihood Models (RELM) test of earthquake forecasts in California,” Proceedings of the National Academy of Sciences (26 September 2011) (doi: 10.1073/pnas.1113481108)
