C HARACTERISTICS OF SELF ‐ SIMILARITY OF SEISMICITY AND THE FAULT NETWORK OF THE S IKHOTE A LIN OROGENIC BELT AND THE ADJACENT AREAS

: We performed a comprehensive analysis of the characteristics of self‐similarity of seismicity and the fault network within the Sikhote Alin orogenic belt and the adjacent areas. It has been established that the main features of seismicity are controlled by the crustal earthquakes. Differentiation of the study area according to the density of earthquake epicenters and the fractal dimension of the epicentral field of earthquakes ( D e ) shows that the most active crustal areas are linked to the Kharpi‐Kur‐Priamurye zone, the northern Bureya massif and the Mongol‐Okhotsk folded system. The analysis of the earthquake recurrence plot slope values reveals that the highest b ‐ values correlate with the areas of the highest seismic activity of the northern part of the Bureya massif and, to a less extent, of the Mongol‐Okhotsk folded system. The increased fractal dimension values for the fault network ( D f ) correlate with the folded systems (Sikhote Alin and Mongol‐Okhotsk), while the decreased values conform to the depressions and troughs (Middle Amur, Uda and Torom). A comparison of the fractal analysis results for the fault network with the recent stress‐strain data gives evidence of their general confineness to the contemporary areas of intense compres‐ sion. The correspondence between the field of the parameter b ‐ value for the upper crustal earthquakes and the fractal dimension value for the fault network ( D f ) suggests a general consistency between the self‐similar earthquake magni‐ tude (energy) distribution and the fractal distribution of the fault sizes. The analysis results demonstrate that the self‐ similarity parameters provide an important quantitative characteristic in seismotectonics and can be used for the neotectonic and geodynamic analyses.


INTRODUCTION
The Sikhote Alin orogenic belt is located in the eastern part of the Amurian plate, its structures are stretching from the Sea of Okhotsk coast in the north to the Sea of Japan coast in the south (Fig. 1, a). From the east, the belt is separated from Sakhalin Island by the rift structure of the Tatar Strait. In the west, it is flanked by the Bureya and Khanka ancient massifs covered with Early Paleozoic continental crust, and bordered by the Jurassic Mongol-Okhotsk fold-nappe belt in the northwest [Didenko et al., 2010]. The Mesozoic-Cenozoic geodynamics of the region, its tectonic structure and recent movements were and are currently controlled by the interaction between the continental Eurasian and Amurian, the subcontinental Okhotsk and oceanic Pacific tectonic plates [Khanchuk, 2006]. In this region, seismicity is caused by the two main processes. In the eastern part, it is the Pacific Plate subduction beneath the Okhotsk and Amurian plates. In the western part, seismicity is due to the interaction between different blocks of the Amurian plate along NNE-striking Kharpi-Kur-Priamurye fault system (the northern segment of the Tan-Lu Fault System) and the Central Sikhote Alin Fault [Khanchuk, 2006;Didenko et al., 2017;Levin et al., 2008;Stepashko et al., 2018].
in the southern segment of the Tan-Lu Fault zone in the territory of China in the 1960s and 1970s. The hierarchical properties of seismicity and variousscale faults in different regions were investigated by a great number of the Russian and foreign researchers [e.g., Sadovsky, Pisarenko, 1991;Robertson et al., 1995;Turcotte, 1997;Goryainov, Ivanyuk, 2001;Sherman et al., 2001;Sadovsky, 2004;Kossobokov, Nekrasova, 2004;Nekrasova, Kossobokov, 2005;Stakhovsky, 2004Stakhovsky, , 2017Sherman, 2005Sherman, , 20122014;Ben-Zion, 2008;Torabi, Berg, 2011;Nekrasova et al., 2015;and many others]. These properties are expressed by the power laws that relate different characteristics of the fault structures and the associated seismicity. Applying fractal geometry approaches to fault tectonics considerably increases the capabilities of applied numerical methods. For example, the possibilities of such an approach were demonstrated by performing the analysis of the self-similarity characteristics for the active fault network of Eurasia in their close relationship with the seismicity characteristics .
The goal of our study is to perform a comprehensive analysis of the characteristics of self-similarity of seismicity and the fault network within the Sikhote Alin orogenic belt and the adjacent areas, and compare these characteristics between each other and with tectonic and geodynamic features of the region. In this paper, we proceed to present the results of our studies focused on seismotectonics of the region and based on approaches of the theory of dynamic systems and fractals that were initiated by Didenko et al. [2017].

INPUT DATA
The study area is bounded by 43-55 N and 129-141 E (Fig. 1). The earthquake catalog created by the Laboratory of Seismology and Seismotectonics at the Institute of Tectonics and Geophysics, Far East Branch, Russian Academy of Sciences (ITiG FEB RAS) was used as the main source of data on the seismicity of the region. It is based on the data on the Primorye and Priamurye earthquakes derived from [Kondorskaya, Shebalin, 1977] and collections [Earthquakes in Russia, 2006Earthquakes of North Eurasia, 1992Earthquakes in the USSR, 1962-1991. A total of 5177 earthquakes occurred from 1500 to 2013 have been analyzed (Fig. 1, b). The catalog compiled by ITiG FEB RAS provides the information on the earthquake origin time, hypocenter coordinates and M LH magnitude (magnitude on LH waves). The magnitudes of part of earthquakes were determined by recalculating their energy class into M LH [Rautian et al., 2007]. This catalog lists individual rare seismic events occurred before 1960, which main parameters were determined from macroseismic data. It should be noted that the number of recorded events has increased with the development of the regional network of seismic stations. The most representative datasets were obtained in 1975-1992, when the network of regional seismic stations allowed for reliable recording of up to 7-8 energy class earthquakes (M LH =2.4-2.8). It is most likely that just because of this reason the events with M≥2.4 were listed in the initial catalogs in the following years. We also used the seismic events with M LH ≥2.4 in our calculations.
Based on the analysis of earthquake hypocenter depths in the region, performed by Levin et al. [2008], Khanchuk [2006] and others, the earthquakes can be subdivided at least into two groups -crustal and mantle. The latter are grouped in the southern and southeastern parts of the region and are related to the oceanic Pacific Plate subduction beneath the Eurasian eastern margin. In this paper, we do not consider them as a subject of the study. In the region, the seismicity is mainly controlled by the crustal earthquakes caused by the recent crustal fault-block tectonic activity. Figure  1c shows the depth distribution of earthquakes based on the earthquake catalog data. It is seen that the absolute majority of the earthquake foci (a total of 4937, i. e., 95 %) are of no greater than 36 km depth. Note that in the catalog, about half of the events (a total of 2890) show the (-1) depth and thus refer to the surface earthquakes, for which it does not appear possible to determine a precise occurrence depth. Besides, the narrow maxima are confined to 5, 10, and 15 km focal depth range, which are also due to the lack of possibility to precisely determine the depths for part of the earthquakes.
In the study area, the crustal thickness varies within 14-38 km according to the model CRUST 2.0 [CRUST 2.0]. Therefore, when further analyzing seismicity, the earthquakes with the focal depths not exceeding 36 km were referred to the crustal earthquakes. Note that the upper crustal earthquakes with the focal depths of no more than 12 km are prevalent among the crustal earthquakes (see Fig. 1, c). The number of such events is as large as 4433, which amounts to 85 % of the total number of all the analyzed earthquakes, and to 90 % of the crustal ones. Such earthquake distribution is well consistent with the upper crustal thickness reaching 10-12 km in this area according to the model CRUST 2.0. This feature is important for a comparison of the characteristics of selfsimilarity of seismicity with similar characteristics for the fault network. We will mainly compare the characteristics of fault tectonics displayed at the surface (fault network) with the characteristics of the crustal and upper crustal seismicity.
To perform the analysis and to compare the characteristics of seismicity, we used the digital map of faults (vector form, see Fig. 1, b) and their descriptions pre-pared by Zabrodin et al. [2015] as the initial fault data for the region including the data on active faults, which are available at (http://itig.as.khb.ru/ppl/gis/2015mono-Fault_Tect_FE-Zabr_Ry_Gil.pdf).

ANALYSIS TECHNIQUE
The spatial structure of the epicentral field of earthquakes is rather complicated and nonuniform, and these properties are displayed in a broad scale range, that is, self-similarity, or fractality, takes place [Mandelbrot, 1983]. Without consideration to the size of an earthquake focus, a set of the earthquake foci has the character of the Cantor (point) sets. Fractal dimension D is a quantitative measure of self-similarity and a degree of complexity of a set of objects [Sadovsky, Pisarenko, 1991;Turcotte, 1997], which shows how densely and uniformly the space is filled with the elements of a set given, and is calculated from the relation: where δ is the scale of consideration, N is the number of elements, c 1 is the constant.
To analyze the fractal dimension of the epicentral field of earthquakes D e , we used the box counting (BC) method. Following that method, the analyzed set of points is covered with boxes of size , and the dependence of a kind (1) is constructed that relates the number of boxes N, where even one point of a set falls in, with a scale . A similar method is described in the patent [Klyuchevskiy et al., 2017]. The square sites were used which minimum size was determined based on the data detail and attained 0.125° by latitude, whereas the maximum size stemmed from the need to stack a number of sites within the overall domain size and to provide no less than one order of scale variation to perform the analysis. In our calculations, we took into account the boxes within which no less than one earthquake fell. The regression equation was solved using the least squares method (LSM). Figure 2, a, illustrates an example showing the calculation of the fractal dimension of the epicentral field of earthquakes D e in the region.
In terms of energy characteristics, self-similarity of the seismic regime is attested by the Gutenberg-Richter (GR) law for the magnitude distribution of a number of earthquakes which is of fundamental importance in seismology [Kasahara, 1981]: where a and b are empirical constants, N is the number of earthquakes with the magnitude exceeding M for a certain time period in a given region. This relation holds for the decay area of the distribution plot lgN(M), the earthquake recurrence plot, showing the relation between the number of weak and strong seismic events occurred in the region. The parameter b-value can be calculated using different methods, such as the maximum likelihood estimation (MLE) (e. g. [Nava et al., 2017]) or the least squares method (LSM). As mentioned above, we used the LSM in our study, a discrete step size for the earthquake magnitude attained 0.2 unit. Figure 2, a, illustrates an example of the noncumulative recurrence plot construction and the parameter b-value estimation for the region. The investigations of the distribution laws for a number of faults N (and other fault structures) according to their length L in different regions and various geodynamic environments by different authors [e.g. Sadovsky, Pisarenko, 1991;Sherman, 2005Sherman, , 2012Sherman, , 2014 and others] have clarified that these laws are described by a power-law relation of the following kind: where m is the power exponent, i. e. they are of the fractal character. To calculate the fractal dimension of the fault network D f , we used the above described box counting method and the following relation was applied: where δ is the scale of consideration, N is the box number, c 2 is the constant. During the analysis, the square sites were used whose sizes and the variation range were taken the same as for the calculation of the fractal dimension of the field of earthquake epicenters D e . in order to make the comparison of the results correct. The regression equation was solved using the least squares method. Figure 2, a, depicts an example of the fractal dimension of the fault network D f in the region.

ANALYSIS OF EPICENTER DENSITY
Before we start analyzing the characteristics of selfsimilarity of seismicity, the conventionally assumed indicator of seismic activity such as the surface density of earthquake epicenters has been analyzed. This value attains 0.33 10 -2 km -2 , on average, but it is however extremely nonuniformly distributed in the study area. To reveal the spatial pattern of seismicity, we calculated the surface density over a moving window of 2° size by latitude (approximately 160160 km at a given latitude) with a step of 0.5° (40 km) using the author's FrAnGeo program [Zakharov, , 2012. The results are presented in Fig. 3. There are practically no differences in the densities of both crustal and upper crustal earthquake epicenters, which results from complete prevalence of the number of the upper crustal and crustal seismic events over the mantle ones.
The statistics of the surface density distribution of earthquake epicenters is shown in Fig. 3, b. The distribution pattern is considerably different from the 'normal' one: the minimum value is 0, the maximum value is 2.81•10 -2 km -2 , and the median is 0.36•10 -2 km -2 . Furthermore, the distribution is bimodal. In addition to the main maximum observed at low density values consistent with the seismically inactive or weakly active zones, the explicit maximum is displayed in a range of 1.7-1.9•10 -2 km -2 , although with a lower amplitude that correlates with the areas of increased seismic activity.
The obtained results are well consistent with the earlier results presented in Didenko et al. [2017], and Stepashko et al. [2018]. They provide evidence that in terms of seismicity manifestation, the study area is rather nonuniform. Variations in the surface density of the earthquake foci (see Fig. 3, a) indicate that the most seismically active crustal areas showing the highest density values in the earthquake epicenter distribution are linked to NE-trending Kharpi-Kur-Priamurye zone, that is the northern segment of the transregional Tan-Lu Fault System [Nikolaev, 1992;Khanchuk, 2006]. Based on the highest occurrence frequency of the seismic events, four areas have been distinguished here (from south to north): (1) between the Kukan Mountain Range and the southwestern branch of the Bureya Mountain Range; (2) at the southern foothills of the Turan Range; (3) at the northeastern branch of the Bureya Range; and (4) the area between the Ezop, Yam Alin and Selemdzha Ranges. The latter area exhibits the highest density value in the earthquake epicenter distribution in the region.
In contrast to the Kharpi-Kur-Priamurye zone, we do not observe crustal zones of similar seismicity within the Sikhote Alin orogen. Here, the exceeding background seismicity can be observed in the following areas (from north to south, see Fig. 3): (1) the lower reaches of the Amur River near Nikolaevsk-on-Amur; (2) the water area of the Sea of Japan, specifically Svetlaya and Maksimovka Bays; and (3) southern Primorye. The most seismically active areas near Vladivostok and Olga Bay are most likely related to the subduction zones.

EPICENTERS
Using the FrAnGeo program [Zakharov, , 2012 and the box counting method, the fractal dimension of the distribution of earthquake epicenters D e was calculated from relation (1) for all the seismic events (а) -spatial variations in the density of the crustal earthquake epicenters  (per 100 km 2 area) compared to the structural scheme of the region (after [Zabrodin, 2017]): 1 -Precambrian platforms; 2 -Phanerozoic folded systems; 3 -Cenozoic depressions; 4 -volcano-plutonic systems; 5 -faults (figures in the circles: 1 -South Tukuringra, 2 -Paukan, 3 -Limurchan, 4 -Yitong-Yilan, 5 -Central Sikhote Alin); (b) -distribution of the surface density of earthquake epicenters, the red line shows the normal distribution with the same average value and dispersion.
Summing up, the above mentioned estimates coincide within the calculation error, and the general structure of the geometric self-similarity of the field of earthquake epicenters is mainly controlled by the crustal earthquakes. In the study area, high values of D e indicate that the structure of the epicentral field of earthquakes has a rather complicated distributed pattern in the range of two orders of spatial scale.
To reveal the spatial features of variations in the fractal dimension of the epicentral field of earthquakes, we performed more detailed moving window calculations using the author's FrAnGeo program as compared to the earlier calculations presented in Didenko et al. [2017]. The window size was 2° by latitude (approximately 160160 km), a step was 0.5° (40 km), and the range of box sizes was 2-0.125° by latitude (160-10 km). The calculated field of D e for the upper crustal earthquakes is shown in Fig. 4. Differentiation of the study area based on the fractal dimension value shows that zones of the highest values of D e , generally correlate with seismically active crustal areas determined from the surface density of the earthquake foci.

PARAMETER B-VALUE OF THE RECURRENCE PLOT
The recurrence plot was constructed for all the earthquakes recorded in the region (see Fig. 2, b). The parameter b-value was calculated from relation (2) in a magnitude range of 2.2-7.6, and b=0.44±0.03 (r=0.95). The magnitude distribution of the crustal earthquakes is appreciably different from the overall distribution: b=0.60±0.03 (r=0.97), whereas for the upper crustal earthquakes b=0.69±0.04 (r=0.97), respectively. This is because in the region, the maximum magnitudes of all the crustal earthquakes do not exceed 7, while for the upper crustal earthquakes this value is 6.4. The lack of stronger earthquakes is shown by a greater slope of the recurrence plot.
In addition to the general b-value estimation for the region, more detailed moving window calculations of the b-value field were performed as compared to the calculations reported in our previous work [Didenko et al., 2017]. The moving window size was 2° by latitude (160160 km), a step was 0.5° (40 km) and the range of box sizes used for the calculations varied from 2 to 0.125° (160-10 km).
The field of the recurrence plot slope values for the crustal earthquakes (Fig. 5, a) demonstrates that the highest absolute value (≥0.7) correlates with the areas of disjunctive faults developed in the northern Bureya massif and NE-trending Kharpi-Kur-Priamurye zone, that is the northern segment of the transregional Tan-Lu Fault System [Nikolaev, 1992;Khanchuk, 2006]. Another maximum of the parameter b-value is exhibited for the central Sikhote Alin zone. The minimum recurrence plot slope values fall in the southern and northern Sikhote Alin zones. The b-value field for the crustal earthquakes mainly differs from that for all the studied earthquakes by the presence of the b-value maximum in the Sea of Japan (at the shelf boundary). Apparently, this is because relatively strong earthquakes are rather of mantle than crustal origin in this area, which is displayed in higher b-values estimated for the crustal earthquakes.
The b-value distribution for the upper crustal earthquakes (Fig. 5, b) is quite close to the 'normal' one (except for the lowest b-value area, the minimum value is 0.20, the maximum value is 1, the average value amounts to 0.49±0.02, the standard deviation is 0.15, and the median of the distribution is 0.48). Let us point out that at b=0. 65, a step appears on the recurrence plot, which value may be used as a threshold value to distinguish the highest b-value area.
The results of comparison of the parameter b-value field and the fractal dimension of the earthquake epicenters D e are shown in Fig. 5, c. Generally, we may conclude that the increased values of both parameters are to a considerable extent spatially overlapped which is especially explicitly displayed for the Mongol-Okhotsk folded system and the Bureya massif, where the highest seismicity is observed This specific feature is most likely associated with the occurrence of a relatively large number of weak earthquakes occurred in this region,, which causes both complication of the structure of the epicentral field (frequently displayed in increasing D e ), and an increase in the recurrence plot slope b. However, such a correlation is not absolute.

NETWORK
The analysis of the fractal dimension. To perform the analysis, we have calculated the fractal dimension of all the faults derived from the database from relation (3) using the FrAnGeo program [Zakharov, , 2012 and the box-counting method. In the analysis, each fault was considered as a linear object not having its own structure. The range of box sizes used to calculate D f varied from 4 to 0.0625° by latitude (approximately 315-5 km at a given latitude). According to our calculations, the fractal dimension of the fault network is D f =1.6±0.03 (see Fig. 2c; correlation coefficient r=0.99).
In order to reveal the spatial features of variations in the fractal dimension (Fig. 6), we performed more detailed moving window calculations using the author's FrAnGeo program as compared to our earlier calculations [Didenko et al., 2017]). The window size was 2° by latitude (160160 km), a step was 0.5° (40 km) and the box sizes varied from 2 to 0.0625° by latitude (160-5 km).
We note that the fault data are irregular and are mainly available for the continental areas, the decreased values of the fractal dimension are observed in proximity to the continent-ocean boundary i.e., in the eastern part of the study area, as well as in its western part. This is, to a significant extent, due to the lack of data and a specific feature of the fractal analysis performed over a moving window and cannot be considered as evidence of a change of the fractal structure of faults in these zones.   The distribution of D f by value is complicated and differs strongly from the 'normal' one: the minimum value is 1.0, the maximum value is 1.69, the average value amounts to 1.41±0.01, the standard deviation is 0.15, the median of the distribution is 1.45, and the mode is 1.57. Simplification and generalization allowed for distinguishing three main ranges in the distribution of the fractal dimension values D f , which are separated by considerable jumps in the frequency values and form the following steps in the distribution pattern: 1-1.3, 1.3-1.5 and 1.5-1.69. The first range (D f ≤1.3) is mainly linked to the above mentioned bands of low D f values observed at the edges of the study area and cannot be comparable with the features of its structure based on these data. Therefore, we distinguish two main ranges of D f values, considering a threshold value of 1.5. Note that this value differs considerably from the average value and is close to the median of the distribution, which is due to the observed asymmetry and causes the difference from the normal distribution of D f values In our opinion, the two distinguished ranges of D f values correspond to different elements of the regional tectonic structure. Most probably, a discrepancy between the fractal dimension values for the faults of the Priamurye area (in our study, the maximum value is 1.65) and the previous results reported by Sherman et al. [2001] (1.55) is due to the difference in the input data, and, also, can be explained by somewhat different analysis techniques.

NETWORK WITH THE RECURRENCE PLOT SLOPE
The self-similar (fractal) properties of the seismic process and the medium, in which this process occurs, are expressed in parameters of the power laws that describe these properties: the fractal dimension values of the fault networks D f , the epicentral (generally speaking, and the hypocentral) fields and parameter b in the Gutenberg-Richter law. For this reason, there are strong grounds to suggest a certain relationship between them. The supposed theoretical relationship between these values was described by Kasahara [1981] and Turcotte [1997]: where c is the parameter that relates the seismic moment and the magnitude (moment magnitude), b is the slope of the earthquake recurrence plot based on the moment magnitude M w distribution. Assuming an average world value c = 1.5 [Kasahara, 1981], relation (4) is as follows: 2 .
In the previous work by Zakharov [2011], the relations between b-and D f were obtained from the seismotectonic analysis of Eurasia, which show that relation (5) approximately holds, but rather significant deviations are observed: for most regions and, on average, the coefficient relating D f and b is somewhat higher than 2 and varies within the range of 1.72.4. Such a consistency between the fractal distribution of the earthquake magnitude (and, consequently, energy) and the fractal distribution of the fault sizes quantitatively confirms the hierarchical self-similar properties of the seismotectonic process.
Quite a great number of works have been published [e.g., Oncel et al., 2001;Caneva, Smirnov, 2004;Chen et al., 2006;Stakhovsky, 2004Stakhovsky, , 2017and others], which describe practical studies and show a comparison between the fractal characteristics of the fault systems and seismicity in different regions of the world. These works demonstrate that relation (5) holds for the regions for quite a long time span, on average. At the same time, rather significant space-time variations, as well as variations due to features of the seismic regime, are possible in the relations between D f and b, The performed experiments on the acoustic emission, which serves as a model of the seismic process, have yielded the relations between D f and b close to 2 [Goebel et al., 2017].
During the analysis, we compared the fractal dimension of the faults and the parameter b-values. Since we analyzed the surface fault system, it was reasonable to examine only the upper crustal seismicity.
The superposition of the above mentioned b-value field on the D f -value field is illustrated in Fig. 6, b. To make comparison more convenient, the b-value field is shown in sparce isolines. The comparison of the fields of these parameters allows for revealing a general correlation between increased b-values (higher than an average value of 0.49) and increased D f values. Undoubtedly, this relation does not hold strictly, and the inconsistency between the D f-minimum and b maximum values observed in the southern part of the Middle Amur depression is of special significance. We suggest that such inconsistency is caused by a relatively increased seismic activity associated with the Kharpi-Kur-Priamirye zone, but the faults are weakly displayed at the surface in this zone (as earlier discussed).
In our study, we investigated the correspondence of the calculated D f and b-values to relation (5). It is necessary to take into consideration that this relation should (theoretically) hold for the recurrence plot slope based on the moment magnitude distribution of earthquakes (M w ), whereas the analyzed catalog lists the magnitudes obtained from surface waves (M LH ). Obviously, this distinction should influence the result: the coefficient of relation between D f and b will be different from 2. To more adequately estimate relation (5), we need to recalculate the magnitudes according to the scale and to calculate parameter b. To do this, we use the empirical relation between M L and M w , obtained by Munafo et al. [2016] based on the statistical analysis: where С ≈ 1.15. Since according to [Konovalov, Sychev, 2014]: 0.97 0.04 • 0.04 0.16 , i.e., M LH ≈ M L , we can use (6) to estimate M w based on the M LH data. Given the magnitude recalculation according to (6), the statistical relation D f /b tends to a 'theoretical' value of about 2, which is consistent with the previous results . This relation is also close to 2, on average, but is somewhat higher.
Thus, we may conclude that the self-similar magnitude (and, as a consequence, energy) distribution of earthquakes is generally consistent with the fractal distribution of the fault sizes. This quantitatively confirms the hierarchical self-similar properties of the seismotectonic process.

DIMENSION WITH THE MAIN TECTONIC STRUCTURES
The comparison of spatial distribution of the fractal dimension values for the fault network with the main tectonic structures of the region indicates a rather clear zonality and the confineness of certain ranges of D f values to different structures. To simplify the scheme reading (Fig. 7), we distinguished only the main threshold isoline values.
The comparison with the structural scheme of the region [Zabrodin et al., 2015;Zabrodin, 2017] (see Fig.  7, a) and the neotectonic map [Grachev, 1997] shows that the increased D f values are confined to the Phanerozoic folded systems: the Sikhote Alin (especially its central part) in the Central Sikhote Alin Fault zone, and the Mongol-Okhotsk in the South Tukuringra Fault zone and the Yitong-Yilan-Paukan fault convergence zone. The D f values observed within these fold systems exceed the average value of D f =1.42, and the value higher than 1.5 is typical of the greater part of these areas. The increased D f values are also typical of the northern Bureya massif, and, to some extent, for the adjacent areas of the Jiamusi massif. The area exhibiting the highest values (D f =1.67-1.69) is located in the northwest of the region, in areas of the Limurchan Fault, the Chayatyn and Chertov Ranges, and, also, in the northwestern Dzhagdy Range The decreased (lower than average) D f values correlate quite well with the depression and trough areas. In the first place, it is the Middle Amur depression showing the lowest values (D f =1.2 A relative local minimum is identified in areas of the Uda and Torom marginal troughs, although shifted with respect to the troughs which is probably related to their relatively small (on a scale of our analysis) sizes. In the eastern Sikhote Alin, the volcano-plutonic system also exhibits low values, but it is difficult to differentiate whether these values are due to the structural features of the fault system or, as mentioned above, can be explained by the lack of the input data.
The performed comparison helps us conclude that the increased values of the fractal dimension of the fault network are confined to the folded systems, while the decreased values correlate with the depressions and troughs. This is explained by the more intense fault formation in the folded systems due to active orogenic processes. In addition, the depressions have a thicker sedimentary cover, which prevents from distinguishing the fault structure of the basement. A certain shift of the D f -field maxima and minima with respect to the structures, which are, in particular, of a small size, can probably be explained by specific features of the analysis technique -"smearing" of the values during the averaging over a moving window.

STATE OF THE CRUST
In this study, we also compared the results of the fractal analysis of the fault network with the data on the recent stress-strain state of the crust derived from different methods, mainly from the analysis and interpretation of the data obtained by different types of remote sensing of the Earth reported by other researchers. The relations between the stress fields and deformations restored using seismic and satellite data were considered in different publications [e.g., Lukhnev et al., 2010;Petrov et al., 2008].
Let us make comparison with the deformations revealed from the GPS data interpretation and analysis described in Ashurkov et al. [2016], where the technique used is based on spline-interpolation with 3030 km box, which corresponds to the scale and degree of detail of our analysis and makes it possible to compare the results. The comparison illustrates that in general, the areas exhibiting increased values of the fractal dimension of the fault network (D f ≥1.5) correlate with the areas of relatively increased strain rates and of the second invariant (intensity) of the strain rate tensor. This results from active different-scale faulting occurred directly in the areas of intense deformations, which is displayed in the number of faults and the increased values of D f . (b) -comparison with the geodynamic and structural features of the buffer zone at the eastern front of the Amurian plate (modified after [Stepashko et al., 2018]). 1 -boundary of the Siberian platform; 2 -Amurian plate, the arrow shows its trajectory; 3 -directions of compression in the earthquake foci; 4 -boundaries of the Lower Amur crustal plate; 5 -main areas of compression (an abnormal seismicity zone at a depth of 500 km). The directions of the recent horizontal displacements according to GPS data: 6 -with respect to the Blagoveshchensk site, 7 -residual values of the displacement vectors; 8 -trajectories of earthquake migration along the boundary of the lithospheric block; 9 -clusters of weak earthquakes in Lower Priamurye; 10 -depth isolines of the subduction zone, km [Zhao, Tian, 2013]; 11 -boundaries of the lithospheric block. Рис. 7. Сопоставление расчетного поля фрактальной размерности разломной сети (D f ) с геолого-структурными особенностями региона.
The comparison with the geodynamic analysis results obtained for the eastern edge of the Amurian plate by Stepashko et al. [2018], based on absolutely different methods -the analysis of seismotectonics and the recent crustal movements in the region, also indicates good consistency between the areas of increased fractal dimension values for the fault network (D f ≥1.5) and the main zones of compression (Fig. 7, b). This means that the fractal dimension of the fault network is one more important quantitative characteristic of fault intensity and the recent stress-strain state of the crust and can be used for the geodynamic analysis.

RESULTS AND CONCLUSIONS
We performed a comprehensive analysis of the characteristics of self-similarity of seismicity and the fault network within the Sikhote Alin orogenic belt and the adjacent areas. From the analysis results, and the comparison of the fields of the obtained characteristics of self-similarity between each other and with the structural, tectonic and geodynamic features of the region, the following conclusions can be made: 1) The depth distribution of earthquake foci and the geodynamic features of the region give grounds to claim that the main features of seismicity are controlled by the crustal and upper crustal earthquakes except for the area of deep-focal earthquakes related to the subduction zone in the Sea of Japan; 2) The fractal dimension of the field of earthquake epicenters of the region was calculated. Differentiation of the study area by the density of earthquake epicenters and the fractal dimension value De provides evidence that the most active crustal areas are linked to NE-trending Kharpi-Kur-Priamurye zone, that is the northern segment of the transregional Tan-Lu Fault System, the northern part of the Bureya massif, and the Mongol-Okhotsk folded system, which agrees with our previous results [Didenko et al., 2017]; 3) The earthquake recurrence plot slope b was estimated for the region. In general, its highest value correlates with the areas of the highest seismicity in the northern area of the Bureya massif and, to a lower extent, of the Mongol-Okhotsk folded system; 4) The field of the fractal dimension of the fault network D f was calculated for the region. It has been ascertained that the increased values of the fractal dimension of the fault network D f are confined to the folded systems (Sikhote Alin and Mongol-Okhotsk), while the decreased D f values correlate with the depressions and troughs (the Middle Amur depression, and, to a lower extent, the Uda and Torom marginal troughs). This is explained by a more intense faulting in the folded systems due to active orogenic processes; 5) The comparison of the fractal analysis results for the fault network with the data on the recent stressstrain state of the crust derived from different methods [Rasskazov et al., 2014;Ashurkov et al., 2016;Stepashko et al., 2018] shows that the zones exhibiting the increased values of the fractal dimension of the fault network D f are generally confined to the areas of contemporary compression. This makes the fractal analysis of the faults an important characteristic of the stress-strain state; 6) Good agreement has been revealed between the parameter b-value field for the upper crustal earthquakes and the field of the fractal dimension of the fault network D f . We conclude that the self-similar distribution of earthquake magnitude (and, consequently, energy) is generally consistent with the fractal distribution of the fault sizes.
Our results show that the self-similarity parameters provide an important quantitative characteristic of fault intensity and the recent stress-strain state of the crust and can be used for the neotectonic and geodynamic analyses.

ACKNOWLEDGMENTS
The authors are thankful to reviewers Dr. A.V. Klyuchevskiy and Dr. V.B. Smirnov, whose constructive remarks and suggestions contributed to improving the quality of the final version of the manuscript, and E.Yu. Didenko and N.N. Kovriga, the staff members of the Institute of Tectonics and Geophysics, Far East Branch, Russian Academy of Sciences (ITiG FEB RAS), for their assistance in the preparation of the manuscript. The study was supported by the grant of the Russian Science Foundation (project No. 16-17-00015). The analysis was performed using high-performance computing systems of the Moscow State University Supercomputing Center [Sadovnichy et al., 2013].