Natural Time Investigation of the Effect of Significant Data Loss on Identifying Seismic Electric Signals

Abstract

In many cases of geophysical and/or geological interest, like the case of SES observations, it happens that for substantial parts of the time of data collection high noise prevents any attempt to extract a useful signal so that such data are removed from further analysis. The appearance of such a noise may be either random or periodic. It is the objective of this chapter to examine whether the remaining data allow the identification of the SES activities (critical dynamics) characterized by infinitely ranged temporal correlations. The following two cases are treated here. First, the effect of the random removal of data segments of fixed length on the scaling properties of SES activities. Second, the appearance of a periodic noise like in Japan, where the electric field measurements at some sites are seriously contaminated by high noise – due mainly to leakage currents from DC driven trains –during the period 06:00 to 22:00 LT every day, i.e., around 70–80% data loss. Here, we show that, in both cases, the identification of a SES activity becomes possible with probability around 70% even after severe data loss (e.g., 70-80 %). This is achieved by combining natural time analysis and DFA as follows: the remaining data is first represented in natural time and then analyzed in order to deduce the quantities k1, S and S- as well as the exponent a from the slope of the log-log plot of the DFA analysis in natural time. We then examine whether the latter slope has a value close to unity and/or the conditions k1 ≈0.070 and S,S− <Su are obeyed.