The calculation of Honkura and Tanaka [1996> and that of Aceves et al. [1996>, provide important tools for the clarification of the main question of this issue, i.e., whether or not the VAN predictions can be ascribed to chance. Honkura and Tanaka's [1996> calculation showed that, in a circular area with a radius of 120 km and for the time window of 22 days, the probability P of occurrence of an earthquake (EQ) with Ms≥5.0 in Greece is less than 0.25, and even smaller for a time window of 11 days. For larger magnitude thresholds, i.e., Ms≥5.5 or Ms≥5.8 (and in view of Gutenberg-Richter relation), their P-value has to be drastically smaller. A simple comparison of these P-values with the Tables of Mulargia and Gasperini [1992, 1996a> immediately reveals that VAN-predictions cannot be ascribed to chance. Note that an inspection of the latter Tables leads to (i) the VAN success rate is 40~45% (when considering correlations with earthquakes having MEQ≥5.0 only), and (ii) the VAN alarm rates increase with the (earthquake) magnitude threshold, reaching to values of 50% and 60%, for MEQ≥5.5 and MEQ≥5.8 respectively. Another important point, which emerges from the calculation of Honkura and Tanaka [1996>, is that ''aftershocks must be treated carefully.'' This strengthens our remarks in Principles 4 and 5 of Varotsos et al. [1996a> that the inappropriate treatment of aftershocks in Mulargia and Gasperini's [1992> calculation (which was based on Poisson distribution): (i) changed drastically the values of the significance level and (ii) turned a true ''forward time correlation,'' between predictions and earthquakes, to a ''backward time association.'' The latter point is also separately checked by Honkura and Tanaka [1996> who conclude that: ''...with the backwards time correlation in mind... we could not find cases in which a high probability arises for the occurrence of an EQ of MS≥5.0 in the target area.'' In this Reply we also proceed to some necessary clarifications, concerning the calculation of the ''success rate'' and ''alarm rate'' when a prediction method has, as expected, an experimental error in the magnitude determination. ¿ American Geophysical Union 1996 |