Reply to “Statistical evaluation of the VAN Method using the historic earthquake catalog in Greece,” by Richard L. Aceves, Stephen K. Park and David J. Strauss

The pioneering calculation by Aceves et al. [1996] shed light on the main question of this debate, i.e., on whether “VAN predictions can be ascribed to chance.” Aceves et al. [1996] conclude that “the VAN method has resulted in a significantly higher prediction rate than randomly sampling a PDF (probability density function) map generated from a 25 year history of earthquakes.” After investigating the totality of VAN predictions issued during the period 1987–1989, Aceves et al. [1996] found: “The prediction rate for the VAN method clearly exceeds that from the random model at all time lags between 5–22 days. At a 5 day time lag, the VAN prediction rate of 35.7% has a P‐value of less than 0.06%. This means that a random model does as well as does the VAN method less than 0.06% of the time. At 22 days, the prediction rate of 67.9% has a P‐value of less than 0.07%.” These conclusions basically coincide with those of Hamada [1993] although Aceves et al. [1996] followed different procedures. They are also in fundamental agreement with the results of Honkura and Tanaka [1996]. Another important conclusion of Aceves et al. [1996] is that, after declustering the earthquake catalog and prediction list from aftershocks, “;VAN method is still formally significant.”