STATISTICAL EVALUATION OF THE SES PREDICTIONS ISSUED IN GREECE - ALARM AND SUCCESS RATES

Earthquake prediction, based on continuous measurements of the electric field of the Earth, has been systematically carried out by Varotsos and co-workers in Greece. Predictions were issued by telegrams to Greek authorities. Statistical evaluation of these predictions was made by comparing telegram prediction information with earthquake parameters determined by the US Geological Survey. The space-time domain selected for this evaluation was 20-24-degrees-E, 36-41-degrees-N, from January, 1987, to June, 1989. Successfully predicted earthquakes are defined here as those that occurred within 22 days of the prediction, within 100 km of the predicted epicenter and with a magnitude difference (predicted minus true) not greater than 0.7. Twelve predictions were issued for earthquakes with expected M(s) greater-than-or-equal-to 5.3. Following our criteria, ten were successful, two were not. The magnitudes of the predicted earthquakes varied from 4.1 to 5.5. For earthquakes with M(b)(USGS) greater-than-or-equal-to 5.0, the ratio of the predicted to the total number of earthquakes is 6/12 (50%) and the success rate of the prediction is a (50%), with a probability gain of a factor of 4. With a confidence level of 99.8%, the possibility of this success rate being explained by a random model of earthquake occurrence, taking into account a regional factor which includes high seismicity in the prediction area, can be rejected. For earthquakes with M(b)(USGS) < 4.5, a statistical significance could not be obtained because of the large number of small earthquakes.