Abstract.Existence of a self-affine long range persistence in the seismic noise time series evidences that the current state of system is not in the pure diffused regime and transition from coherent to incoherent motion is still on progress. Rate of this 15 evolving transition can be indirectly linked to the degree of heterogeneity of medium, thus in this paper we tried to gain an insight into the heterogeneity of the medium by analyzing the width, extreme and asymmetrical trend of multifractal spectrum of seismic records. Nonetheless, toward high frequency ranges a seismic signal itself loses its stationarity short while after its recording onset time. Experimentally, the long-range correlation of a stationary time series (with 0 < ℎ(2) < 1) can be discerned from a non-stationary process (with ℎ(2) > 1) by examining the values of scaling exponent ℎ(2), 20 however, changing in the fractal properties in the crossover time scales in time series don't permit us to ascribe a single amount for ℎ(2) and without executing additional analysis on the stationarity length of signals, direct calculation of such long range correlation and fractal dimensions might be biased. Hence, in this paper we examined the inherent stationarity of a signal relative to the different observation scales in the stochastic contexts before feeding the signal into the cycle of DFA.This method is based on the comparison between global and local features of the original signal and its synthesized time-25 frequency surrogates; therefor it can effectively improve the accuracy of results. Our approach proves the existence of a high-velocity anomalous feature in the right flank of Sahand inactive volcano where it is surrounded by heterogeneous lowvelocity structures and extended to the shallower than ~3 km depth beneath this region at the northwestern of Iran.

Although research on seismic interferometry is now entering a phase of maturity, earthquakes are still the most troublesome issues that plague the process in real applications. To address the problems that arise from spatially scattered and temporally transient enormous earthquakes, preference is usually given to the use of time-dependent weights. However, small earthquakes can also have a disturbing effect on the accuracy of interpretations if they are persistently clustered right next to the perpendicular bisector of the line joining station pairs or in close proximity to one of the stations. With regard to the suppression of these cluster earthquakes, commonly used solutions for dealing with monochromatic microseismic cluster events (e.g., implementing a band-reject filter around a comparatively narrow frequency band or whitening the amplitude spectra before calculating the cross-spectrum between two signals) may not have the necessary efficiency since earthquake clusters are generally a collection of events with different magnitudes, each having its own frequency and energy contents. Therefore, the only solution left in such a situation is to use stronger non-linear time-dependent weights (e.g., square of the running average or one-bit normalization), which may cause Green’s function amplitude information to be lost. In this paper, by simulating the records of a benchmark earthquake MN 5.2 with the help of empirical Green’s functions (EGF) obtained after the Ahar-Varzeghan Earthquake Doublet (MN 6.4 and MN 6.3), it is shown that the amplitude-unbiased phase cross-correlation is a relatively efficient approach in the face of the issues concerning long-standing cluster events.