Artifactual log‐periodicity in finite size data: Relevance for earthquake aftershocks

Abstract

The recently proposed discrete scale invariance and its associated log‐periodicity are an elaboration of the concept of scale invariance in which the system is scale invariant only under powers of specific values of the magnification factor. We report on the discovery of a novel mechanism for such log‐periodicity relying solely on the manipulation of data. This “synthetic” scenario for log‐periodicity relies on two steps: (1) the fact that approximately logarithmic sampling in time corresponds to uniform sampling in the logarithm of time; and (2) a low‐pass‐filtering step, as occurs in constructing cumulative functions, in maximum likelihood estimations, and in de‐trending, reddens the noise and, in a finite sample, creates a maximum in the spectrum leading to a most probable frequency in the logarithm of time. We explore in detail this mechanism and present extensive numerical simulations. We use this insight to analyze the 27 best aftershock sequences studied by Kisslinger and Jones [1991] to search for traces of genuine log‐periodic corrections to Omori's law, which states that the earthquake rate decays approximately as the inverse of the time since the last main shock. The observed log‐periodicity is shown to almost entirely result from the “synthetic scenario” owing to the data analysis. From a statistical point of view, resolving the issue of the possible existence of log‐periodicity in aftershocks will be very difficult as Omori's law describes a point process with a uniform sampling in the logarithm of the time. By construction, strong log‐periodic fluctuations are thus created by this logarithmic sampling.