مصطفی علامه زاده

In this paper, the dynamics of the seismic activity and fractal structures in magnitude time series of Sarpol-e Zahab earthquakes are investigated. In this case, the dynamics of the seismic activity is analyzed through the evolution of the scaling parameter so-called Hurst exponent. By estimating the Hurst parameter, we can investigate how the consecutive earthquakes are related. It has been observed that more than one scaling exponent is needed to account for the scaling properties of earthquake time series. Therefore, the influence of different time-scales on the dynamics of earthquakes is measured by decomposing the seismic time series into simple oscillations associated with distinct time-scales. To this end, the Empirical Mode Decomposition (EMD) method was used to estimate the locally long-term persistence signature derived from the Hurst exponent. As a result, the time dependent Hurst exponent, H(t), was estimated and all values of H>0.5 was obtained, indicating that a long-term memory exists in earthquake time series. The main contribution of this paper is estimating H(t) locally for different time-scales and investigating the long-memory behavior that exists in the non-stationary multifractal time-series. The time-dependent scaling properties of earthquake time series are associated with the relative weights of the amplitudes at characteristic frequencies. The superiority of the method is the simplicity and the accuracy in estimating the Hurst exponent of earthquakes in each time, without any assumption on the probability distribution of the time series. We investigated the negative and positive autocorrelations that exist between consecutive seismic activities by estimating the time-dependent Hurst parameter, H * (t). To this end, the empirical mode decomposition and the Hilbert-Huang transform are applied. Using this method, the seismic activities are studied locally, and the autocorrelation between consecutive earthquakes are estimated in each time. We have investigated the superiority of the estimator by simulation. Furthermore, the method is applied in estimating H * (t) of earthquakes occurred in Sarpol-e Zahab during November 12, 2017 to January 20, 2018. By estimating the Hurst parameter locally, and considering the values of H * (t) that are greater than 0.5, we identify that the long memory behavior exist in consecutive earthquakes during 25/11/2017 and 13/1/2018. It is also seen that, in spite of small values of H * (t) for some times, which shows the stochastic behavior in earthquakes, the local Hurst parameters are tending to be greater than 0.5, and after this patterns, a peak in magnitude is seen. This paper shows that the combination of the EMD and its associated Hilbert spectral analysis offers a powerful tool to uncover the time-dependent scaling patterns of consecutive seismic activities data. In this paper, we investigate the long-range correlations and trends between consecutive earthquakes by means of the scaling parameter so-called locally Hurst parameter, H(t), and examine its variations in time, to find a specific pattern that exists between foreshocks, main shock and the aftershocks. The long-range correlations are usually detected by calculating a constant Hurst parameter. However, the multi-fractal structure of earthquakes caused that more than one scaling exponent is needed to account for the scaling properties of such processes. Thus, in this paper, we consider the time-dependent Hurst exponent, to realize scale variations in trend and correlations between consecutive seismic activities, for all times. We apply the Hilbert-Huang transform to estimate H(t) for the time series extracted from seismic activities occurred in Iran. The superiority of the method is discovering some specific hidden patterns that exist between consecutive earthquakes by studying the trend and variations of H(t). Estimating H(t) only as a measure of dependency may lead to misleading results, but using this method, the trend and variations of the parameter is studying to discover hidden dependencies between consecutive earthquakes. Recognizing such dependency patterns can help us in prediction of mainshocks.

Seismicity pattern studies are one of the effective tools in the interpretation of variation in seismic sequence. The study of variations of seismicity parameters as a function of time indicates that the temporal distribution of events is not uniform, and these parameters can give quantitative information about the seismic patterns of different regions. In this research, to investigate temporal variations of seismicity pattern in the Zagros fold and thrust belt, the Schreider algorithm is applied. This algorithm that introduced by Schreider (1990) to detect seismic quiescence has been used in different parts of the world. For this purpose, four earthquakes with Mw≥6 that have recently been occurred in Zagros have been studied. At first, a complete catalogue from the period of 2000 to 2017 within a circular area has been selected. Then, the catalogues are homogenized to ML and the Minimum magnitude of completeness are computed (Mc=3.4). To perform Schreider algorithm, the time between consecutive earthquakes ( ) should be calculated. The smoothness procedure is used to evaluate a convolution function of . The smoothness of this function is done by Gaussian function. In the R radius, smoothness parameter (s) has controlled the extent of surrounding earthquakes to detect smooth values. The kth seismic event is related with the temporal convolution T(k) that decrease and increase indicate seismic activity or low seismic activity, respectively. The numbers of earthquakes that are located in the nearest distance to main shock determine the l parameter. The value of l is determined when the function f(n,s) is approximately zero. Therefore, the function T(k) depends on the s and l parameters. To investigate temporal variations of seismicity, in addition to the temporal convolution (T(k)) plot the magnitude-time, number-time and space-time plots have drawn. The results show that before the 2006 Silakhor and 2014 Mormori earthquakes, both of which occurred in the north part of Zagros, the precursory doughnut pattern is seen. Several years before the 2013 Dashti and 2005 Qeshm earthquakes, both of which occurred in the south part of Zagros, the seismic quiescence is seen for which ended with the sudden occurrence of these earthquakes. The result of sensitivity analysis showed that smoothing parameters of Schreider algorithm have a significant influence on the algorithm outcomes, and these parameters should be selected with more accuracy. The results of this paper show that Schreider algorithm can demonstrate precursory seismic quiescence before the occurrence of large earthquakes due to the intelligent usage of t parameter.

Peak ground acceleration is one of the most important factors that needs to be investigated in order to predict the devastation potential resulting from earthquakes in reconstruction sites. Besides, the maximum level of shaking control is subjected criteria that can be worth considering. In this research, a training algorithm based on gradient descent and Levenberg-Marquart (Train LM) were developed and employed by using strong ground motion records. The Artificial Neural Networks (ANN) algorithm indicated that the fitting between the predicted PGA values by the networks and the observed PGA values were able to yield high correlation coefficients of 0.78 for PGA.
From a deterministic point of view, the determination of the strongest level of shaking that is expected at a site has long been a significant consideration in earthquake engineering. Besides, knowledge of the maximum physically possible ground motions allows a meaningful truncation of the distribution of ground motion residuals, and as a result, leads to falling of the values computed in probabilistic seismic hazard analysis (Strasser and Bommer, 2009).
The peak ground acceleration parameter is often estimated by the attenuation of relationships and by using regression analysis. PGA is one of the most important parameters, often analyzed in studies related to damages caused by earthquakes (Gullo and Ercelebi, 2007). It is mostly estimated by the attenuation of equations and is developed by a regression analysis of powerful motion data.
Kerh and Chaw (2002) used software calculation techniques to remove the lack of certainties in declining relations. They used the mixed gradient training algorithm of Fletcher-Reeves’ back propagation error (Fletcher and Reeves, 1964). They applied three neural network models with different inputs including epicentric distance, focal depth and magnitude of the earthquakes. These records were trained and then the output results were compared with available nonlinear regression analysis.
In this article, to estimate strong ground motion acceleration component in an area, four artificial neural networks with different algorithms were used, including General Regression Neural Network (GRNN), Nonlinear Auto Regression neural network (NARX), Feed-Forward Back-Propagation error (FFBP) and General Feed-Forward Neural Network (GFFNN). Input vectors of neural networks include four parameters, which have key effects in occurrence of an earthquake in an area. The parameters include magnitude of moment, rupture distance of earthquake center, mechanism of faults, and ranking of site. Output vector has only one component: maximum peak ground acceleration for an earthquake in an area is used as a target output.
After different tests, GRNN network has maximum output correlation coefficient (0.87) and General Feed-forward Back-Propagation error neural network (FFBP) has the least (0.41). Besides, GRNN network had the least mean square error (0/014), and Back-Propagation network had 0.125. In this research, GRNN neural network is the best neural network, which can estimate possible peak acceleration more than 1g in an area.
Artificial neural networks are a set of non-linear optimizer methods which do not need certain mathematical models in order to solve problems. In regression analysis, PGA is calculated as a function of earthquake magnitude, distance from the source of the earthquake to the site under study, local condition of the site and other characteristics that are linked to the earthquake source such as slippery length and reverse, normal or wave propagation. In non-linear regression methods, non-linear relations that exist between input and output parameters are expressed as estimations, through statistical calculations within a specified relationship (Douglas, 2003).

In this paper an approach is presented to predict the concentration and the trend of seismic pattern and clusters of earthquakes. The method is based on Copulas and artificial Neural Networks that have attractedmuch attention in spatial statistics over the past few years. They are used as a flexible alternative to traditional methods for non- Gaussian spatial modeling and interpolation. This methodology shows how it can be predicted aftershocks distribution in a Bayesian framework by assigning priors to all model parameters. The Gaussian spatial copula model is equivalent to trans-Gaussian kriging with transformation function. A restriction of the Gaussian copula is that it models not only a symmetric but even a radials symmetric dependence, where high and low quartiles have equal dependence properties. Experimental results show that the proposed models are superior to predict and identify seismic risk at high seismicity areas.

In this article, we showed the results of pattern recognition by using numerical simulations models using Copula methods for earthquake prediction for central Alborz (Tehran region) and compared them with the results of seismogenic nodes. The goal of this paper is to identify where earthquakes with M>6.0 can occur and define the assemblage of geological-geomorphological features that discriminate such sites from areas of lower seismic potential. It is demonstrated that the synthetic clustering in space and time of earthquakes are useful for seismic hazard assessment and intermediaterange earthquake in Tehran region. In these experiments, we will present a Copula methods that can finds the unknown location of the possibility major earthquakes by using foreshocks, Seismic silence, and Doughnut pattern in this region. Our model and computer simulations obtain these synchrony results across in Tehran region located in center of Alborz in Irananian plate.

In recent years, considerable research has been devoted to the study of seismic data carrying information about the complex events that lead to earthquakes. Much of the recent geophysical research associated with earthquakes has centered on investigating spatial and temporal patterns of aftershocks in local and regional seismicity data (Kanamori, 1981). Notable examples include characteristics of earthquakes and temporal clustering (Frohlich, 1987; Press and Allen, 1995; Dodge et al., 1996). The principal properties of aftershock sequences are described empirically (e.g. Utsu, 1961; Omari, 1894). Although much of this work represents important attempts to describe these characteristic patterns using empirical probability density functions, none of these observations or methodologies systematically identifies all possible seismicity patterns. The quantification of all possible space-time patterns would seem to be a necessary first step in the process of identifying which patterns are precursory to large events, leading to the possible development of new approaches in forecast methodology. It is difficult for most scientists to understand why events with large magnitude such as Bandar-Abbas earthquake are not preceded by at least some causal process. This paper presents an investigation on spatial and temporal fluctuations of earthquake form time series, using the inverse statistical analysis, a concept borrowed from turbulence. Inverse statistics propose to invert the structure function equation, and instead consider averaged moments of distance between two points, giving a difference value between those points. In the inverse statistics method is used another well-known method that called Level Crossing method (LC). In level crossing method, no scaling feature is explicitly required and this is the main advantage of this method in estimating the statistical information of the series. Level crossing is based on stochastic processes that grasp the scale dependence of the time series. The analysis is also performed for the shuffled time series of earthquakes. Since the application of shuffling data destroys the correlation and demonstrates whether there are actually correlations among time (distance) series. We report some significant differences between the results of these two sets of data.

This study continues exploring methodologies to improve earthquake-explosion discrimination using regional amplitude such as ARMA coefficients. The earliest simple source models predicted P wave amplitudes for explosions should be much larger than for earthquakes across the body wave spectrum. However، empirical observations show that the separation of explosions from earthquakes using regional P amplitudes is strongly frequency dependent، with relatively poor separation at low frequencies (~1 Hz) and relatively good separation at high frequencies (>~3 Hz). The authors of this paper demonstrate this by using closely located pairs of earthquakes and explosions recorded on common، publicly available stations at test sites around the world. It was shown that this pattern appeared to have little dependence on the point source variability revealed by P-waves and surface waves. Therefore، S-waves resulted by explosions can be predicted from the P-wave models via a frequency dependent transfer function. A recent model by Allamehzadeh (2011) shows that the explosion ARMA coefficients can be modeled using the P-wave spectra with the corner fre-quency reduced by the ratio of the wave velocities، which seems to imply a direct generation of S-waves in the source region. A number of nuclear tests were examined from around the world in the light of these models. In this paper، the problem of seismic source classification is considered at regional distances. The neural network classifier is fed by Nonlinear AutoRegressive Moving Average (NARMA) coefficients extracted from raw data as feature set. The authors have devised a supervised neural system to discriminate between earthquakes and chemical explosions applying filter coefficients obtained from the windowed P-wave phase (15 sec). First، the recorded signals are preprocessed to cancel out instrumental and attenuation site effects and obtain a compact representation of seismic records. Then، a Quadratic Neural Network (QNN) system was used to extract nonlinear ARMA coefficients for feature extraction in the discrimination problem. These coefficients were then applied to a probabilistic network for the purpose of training and classification. The results have shown that the overall combination of this feature extraction and classifier structure leads to a suitable seismic discrimination performance. The events tested include 36 chemical explosions at the Semipalatinsk test site in Kazakhstan and at test site in China as well as 61 earthquakes (mb = 5. 0-6. 5) recorded by the Iranian National Seismic Network (INSN). In this study، we implemented feature extraction-feature discrimination approach to the seismic discrimination problem by using Neural Networks. As the discrimination features، various power or ARMA characteristics of seismo-grams were used that are typically different for earthquakes and explosions. The discrimination problem was solved by application of statistical pattern recognition techniques.