Temporal variations of the fractal properties of seismicity in the western part of the north Anatolian fault zone: possible artifacts due to improvements in station coverage

Abstract. Seismically-active fault zones are complex natural systems exhibiting scale-invariant or fractal correlation between earthquakes in space and time, and a power-law scaling of fault length or earthquake source dimension consistent with the exponent b of the Gutenberg-Richter frequency-magnitude relation. The fractal dimension of seismicity is a measure of the degree of both the heterogeneity of the process (whether fixed or self-generated) and the clustering of seismic activity. Temporal variations of the b-value and the two-point fractal (correlation) dimension D_c have been related to the preparation process for natural earthquakes and rock fracture in the laboratory These statistical scaling properties of seismicity may therefore have the potential at least to be sensitive short- term predictors of major earthquakes. The North Anatolian Fault Zone (NAFZ) is a seismicallyactive dextral strike slip fault zone which forms the northern boundary of the westward moving Anatolian plate. It is splayed into three branches at about 31oE and continues westward toward the northern Aegean sea. In this study, we investigate the temporal variation of D_c and the Gutenberg-Richter b-value for seismicity in the western part of the NAFZ (including the northern Aegean sea) for earthquakes of M_s > 4.5 occurring in the period between 1900 and 1992. b ranges from 0.6-1.6 and Dc from 0.6 to 1.4. The b-value is found to be weakly negatively correlated with D_c (r=-0.56). However the (log of) event rate N is positively correlated with b, with a similar degree of statistical significance (r=0.42), and negatively correlated with D_c (r=-0.48). Since N increases dramatically with improved station coverage since 1970, the observed negative correlation between b and D_c is therefore more likely to be due to this effect than any underlying physical process in this case. We present this as an example of how man-made artefacts of recording can have similar statistical effects to underlying processes.