The bvalues calculated for each region.
1. Introduction
The recurrence, based in the recurrence Poincare theorem, is a fundamental property of dynamical systems that has been exploited to characterize the behavior of dynamical systems in the phase space. Recurrence is defined when an orbit visits a region of phase space that was previously visited [1]. In this context, the socalled recurrence plot (RP), introduced by Eckmann et al in [2], is a powerful tool for the visualization and analysis of the underlying dynamics of the systems in the phase space like determinism, divergence, periodicity, stationarity among others, for instance the lengths of the diagonal line structures in the RP are related to the positive Lyapunov exponent [3]. Methods based in RP have been successfully applied to wide class of data obtained in physiology, geology, physics, finances and others. RP are especially suitable for the investigation of rather short and nonstationary data [4], and complex systems [5]. For deterministic systems the analysis in the phase space is relatively direct because their solutions are represented as time series, nevertheless, for real natural systems like clime, atmosphere, dimetilsulphure production [6], some of their dynamical variables are gathered as punctual processes. Particularly, the representation of a seismic sequence as a time series is one of the most debated questions in Geophysics, nevertheless natural time analysis, introduced by in [7] by Varotsos et al considers sequences of events and, with this methodology has been possible to identify signals with noises [8] and characterize seismic processes from an order parameter properly defined [9]. Taking in consideration that faults and tectonic plates can be considered as dynamical systems which permit to apply techniques derived from the dynamical systems theory and nonlinear analysis like recurrence plots, clustering, entropy, fractality, correlation, memory among other. In this work we studied properties of the seismicity occurred in Mexico in the period 20062014, reported by Servicio Sismológico Nacional, (www.ssn.unam.mx), by considering the occurrence of events as sequences of magnitudes (time series) and determining the recurrence plots. Based on the Visual Recurrence Analysis (VRA), it is possible to get dynamical features of the seismicity. The most important seismic region in Mexico is located along the Mexican Pacific coast. In [10] was proposed a division of 19 regions of the seismo tectonic zone. They took into account seismic characteristics of the existing catalogs for the seismicity occurring in Mexico from 1899 to 2007, for details see [9] and references therein. In order to characterize the Mexican seismicity as a dynamical system driven by the tectonic movements in the Pacific edge, we proposed to investigate the qualitative dynamical features for the dispersion zone, Baja California, and for the subduction zone formed by La Ribera and Cocos plates subducting beneath the Northamerica plate. Taking into account the Geophysical structure, the seismic activity and the bvalues in the GutembergRichter law, in this work, six selected regions have been considered. These six regions are named: Baja California (Z1), JaliscoColima (Z2), Michoacán (Z3), Guerrero (Z4) and Oaxaca (Z5) and Chiapas (Z6), the first one corresponds with a dispersion zone and the other ones are subduction regions, [11]. The analyzed data set in this work correspond with the Mexican catalogue which is complete for magnitudes greater than 3, within the mentioned period. Since a geophysical point of view, the seismic activity should be individually characterized in each region because the underlying dynamics shows particular features, as is described in the next section. First, seismic data are represented as a temporal sequence of magnitudes, then, the phase space is reconstructed, by estimating the time delay and the embedding dimension. In order to distinguish some features of the underlying dynamics of each Mexican region seismic, the aim of this work is to study the recurrence plot behavior based on the visual recurrence analysis, taking into account the sequence of events (magnitudes) in time and, on the other hand, analyzing the interevents time series. Our analysis shows important differences in the recurrence maps of each region. Our finding suggest that the patterns obtained could be associated with the local geophysical structures of each subduction and dispersion zones driven by their characteristic nonlinear dynamical features of each region.
2. Methodology
2.1. Phase space reconstruction
To reconstruct the dynamics of the system is necessary to convert the time series in state vectors, being the principal step the phase space reconstruction. Takens [12] showed that it is possible to recreate a topologically equivalent picture of the original multidimensional system behavior by using the time series of a single observable variable. The idea is to estimate a time delay
the next vector is construct in phase space
where i = 1 to N – (
2.2. Mutual information function
Formally, the Mutual Information is defined, for two stochastic variables X and Y, as I(X;Y) = H(Y,X) – H(XY) – H(YX) where H(Y,X) is the joint entropy, H(XY) and H(YX) are the conditional entropies. If X represents the sequence x(t_{i}) and Y the respective sequence x(t_{i} + τ) the Mutual Information Function (MIF) is able to calculate the global correlation in a time series taking into account the linear and non linear contributions, being MIF the most frequently used to calculate the time delay and described as follows:
The mutual information I(τ) is a measure of the relative entropy between the joint distribution and the product of distributions P(x(t_{i})) and P(x(t_{i} + τ)), where P(x(t_{i}), x(t_{i} + τ)) is the joint probability of the signal measured between the times t_{i} and t_{i} + τ. The expressions P(x(t_{i})) and P(x(t_{i} + τ)) indicates the and marginal probabilities. MIF is a nonlinear generalization of the linear autocorrelation function. According to [13] the suitable value of τ is attained with the first local minimum of I(τ). When the time series is uncorrelated or random, like white noise, the next equality holds
and I(τ) = 0 [14]. Typical cases are white noise, periodic and periodic + noise and Rossler time series. In Figure 1 the MIF behavior for this typical cases are depicted.
2.3. False nearest neighbors
False Nearest Neighbors method (FNN) is an iterative algorithm which estimates the minimal embedding dimension of the system proposed by [15]. The idea is based in the uniqueness theorem of a trajectory in the phase space. A nearest neighbor P of a point Q in a phase space of ddimensional embedding is labeled false if these points are close only due to the projection from a higherdimensional (d+1)dimensional phase space. Thus, the FNNs will separate if the data is embedded in (d+1)dimensional space, while the true neighbors will remain close. When all the FNNs have been detected, then the minimal sufficient embedding dimension can be identified as the minima dimension needed to achieve zero fractions of the FNNs being the embedding dimension required. For each vector Y(i), its nearest neighbor Y(j) is looked in a mdimensional space. In order to do a comparison, the distance d((Yi).Y(j)) is calculated. By iteration of points, the condition:
If
For the mentioned examples, the false nearest neighbor fraction is calculated as an embedding function and is depicted in the Figure 3. To get the embedding dimension, first was estimated the time delay for each time series by evaluating the respective MIF. For white noise and periodic with noise time series, the embedding dimension is high and for periodic and Rossler signals, the dimension is short.
2.4. Recurrence Plot (RP)
In [2] Eckmann et al introduced the socalled Visual Recurrence Analysis (VRA) based on a graphical method designed to locate hidden recurrent patterns, nonstationarity and structural changes observed in the phase space of a dynamical system. The aim of the Recurrence Plot (RP) method is to visualize the recurrences of dynamical systems as a function of time. A brief description is as follows: assuming an orbit of the system in the phase space
were
Figure 4 shows examples of RP of the four cases described above: white noise, periodic, periodic with noise and chaotic systems.
3. Seismic regions and data set
We analyzed the whole seismic catalog of the Mexican SSN, (www.ssn.unam.mx) from 2006 to 2014 Due to geophysics features of the Mexican subduction zone, it has been described in [9 and references [20,21,24] therein], sugesting that it can be studied in segments where the six regions are: Baja California (Z1), Jalisco–Colima (Z2), Michoacán (Z3), Guerrero (Z4), Oaxaca (Z5) and Chiapas (Z6), in Figure 5 the six regions are showed.
The following panel of Figure 6 displays the seismicity within 20062014 periods for each region. It can be observed the seismicity monitored in Region Z1, Peninsula of Baja California where the Pacific–North America plate boundary in southern California and the north side of Baja California peninsula. The seismicity in Region Z2, JaliscoColima is a subduction zone located to the west, where the Rivera plate subducts at a steep angle. In the Region Z3 Michoacán, the dip angle of the Cocos plate decreases gradually toward the southeast. For the region Z4, Guerrero is bounded approximately by the onshore projection of the Orozco and O’Gorman fracture zones, the subducted slab is almost horizontal and moves under upper continental plate. The regions Z5, Oaxaca and Z6, Chiapas are located in the southeastern of Mexico, the dip of the subduction zone gradually increases to a steeper subduction in Central America.
As is showed in Figure 6, the number of earthquakes reported by the SSN in each region is different. Firstly, the seismicity in the region Z1, peninsula of Baja California, shows two periods of time, the fist one from 2006 to 2010 and the second one from 2011 to 2014, in the first period little seismic activity compared to the second is observed, however, a sudden change in the seismic activity is observed, this behavior in the seismic activity before and after 2010 can be observed in the regions Z2 and Z3 corresponding with JaliscoColima and Michoacán respectively This situation may be due to system upgrades monitoring stations. As mentioned above, the Z1 region evolve with a process in which the peninsula of Baja California is separated from the continental plate, while in the other regions, the dynamics is driven by subduction between continental plate and plates Rivera and Cocos. Specifically, JaliscoColima region (Z2) the subduction is given between La Rivera and The northAmerica plates where the stress and strain fields determine the direction of movement in which La Rivera subducts being different from the case of the Cocos plate. According with the catalogue, Guerrero (Z4), Oaxaca (Z5) and Chiapas (Z6), are the regions with a high seismic activity. It can be observed that in all regions have occurred earthquakes with magnitudes M ≈ 7. By considering this division of six regions and the seismic activity showing seismic clustering, the bvalues in the GR law were recalculated in each case by using the Aki model [17]:
where



Baja California (Z1)  3.93  1.32 
Jalisco Colima (Z2)  4.20  1.45 
Michoacán (Z3)  3.98  2.41 
Guerrero (Z4)  3.91  2.07 
Oaxaca (Z5)  3.92  1.97 
Chiapas (Z6)  4.07  1.61 
Because geophysics shaping the seismic zone as well as the various mechanisms and processes that take place in each of the regions, the statistical properties associated with the seismicity of each of these are reflected as different values to the parameters of the laws scaling as in the case of the GutenbergRichter law. This situation indicates that the local stress fields and stress must drive the interactions between different parts of a complex system, such as the Earth’s crust. So, through a study in the context of dynamic systems, where the seismic activity is considered as a response to the underlying dynamics is possible to observe different characteristics of the system and not observed directly from a statistical point of view. Although seismicity is considered, as a sequence of events whose measurable variable is the magnitude, cannot be put aside the temporal component, that is, the distribution of interevent defined as the time between two earthquakes within a region while. It is noteworthy that there are time series studies interevent by multifractal analysis for seismicity in Italy [19].
The Figure 8 shows the interevent time series of the seismicity studied. For this time series the behavior of the MIF and FNN fraction is calculated.
4. Results
In this work six seismic sequences and their respective interevent time series were analyzed by means of the RP method. Firstly the phase space is reconstructed for each sequence. The MIF and FNN fraction for the magnitudes sequences are depicted in Figures 9 and 10. In order to reconstruct the phase space, the time delay
REGION 


Baja California (Z1)  5  13 
Jalisco Colima (Z2)  2  12 
Michoacán (Z3)  3  7 
Guerrero (Z4)  3  9 
Oaxaca (Z5)  9  12 
Chiapas (Z6)  4  11 
From the behavior of MIF is possible to identify the correlation in seismic sequences. According to this behavior, the region with the lowest correlation is Z2 (JaliscoColima) and the highest correlation is Z5 (Oaxaca), as is showed in Figure 9. However, the behavior of the FNN fraction as function of the embedding dimension of the seismic sequences indicates that all of them are in phase spaces of high dimension being the shorter Z3 with
To calculate the recurrence matrix, the distances between pairs of vectors were computed in phase space. By definition this matrix is symmetric because
The distribution function associated with the distances between states in phase space allows us to observe the most likely value, and suggest a other possible criterion for determining characteristics distances among all states in the phase space and to determine the εvalue in the definition of the matrix recurrence (Ec. 7). The Figure 12 shows the pdf of the distances for the six sequences of seismic magnitudes. It is observed that the maximum values of the pdf are located in different possitions and possibly this behavior could be associated with the geophysics features of the regions: Z2, Z4, Z5 and Z6 are subduction zones where Z3 is determined by the relative motion between the Rivera plate and the Continental plate, while the other three are determined by the interaction of the Cocos plate subducting under the continental plate. For Z1 the seismic activity is produced by the movement of separation between the peninsula and the continental shelf of North America and the Z3, the seismic activity is determined by the interaction of the plates Rivera in contact with the plate subducting Cocos and both the continental plate.
On the other hand, the results of the interevent time series are showed in Figure 13 for MIF behavior and Figure 14 for FNN fraction. In Table 3 the






Baja California (Z1)  4  20 
Jalisco Colima (Z2)  5  20 
Michoacán (Z3)  4  13 
Guerrero (Z4)  5  20 
Oaxaca (Z5)  4  20 
Chiapas (Z6)  4  20 
5. Concluding remarks
It is well known that the Mexican Pacific is an important seismic region where large earthquakes have occurred with devastating consequences producing significant economic downturns and especially many human losses. Due to interactions between subduction zones and the slow separation of the Baja California Peninsula, this region is considered a complex system that evolves as consequence of many processes that occur in the interior of the Earth as well as in the areas of contact between the surfaces involved. In this context, [10] proposed a division of 19 regions of the seismo tectonic zone taking into account seismic characteristics of the existing catalogs for the seismicity occurring in Mexico from 1899 to 2007 and a seismic historic compilation from previous publications and of some catalogs. In order to distinguish some features of the underlying dynamics of each Mexican region seismic, the aim of this work is to study the recurrence plot behavior based on the visual recurrence analysis, taking into account the sequence of events (magnitudes) in time and, on the other hand, analyzing the interevents time series. Our analysis shows important differences in the recurrence maps of each region. In a similar way by considering the seismicity monitored by SSN within the period 20062014, and identifying clusters of earthquakes that can be associated with the geophysical features of the Mexican Pacific, six regions were considered to study (Figure 5). In this paper we studied dynamical features of the six seismic regions located along Mexican Pacific coast. The analyzed data set corresponds with the Mexican seismic catalogue reported by the SSN. First, sequences of magnitude of earthquakes and the interevent time series were studied. Their analysis was performed by means of the phase state reconstruction and the RP of each region. Our findings indicate a possible correlation between the RP calculated and the geophysical features characteristics in each zone (panel of Figure 11). RP displayed of BC, JC and Ch show non periodicities, correlation (not white noise structure), non stationariety. For M, G and O, RP are more similar and stability is observed. The results for the interevents time series, short correlation and large embedding dimension, suggest the possibility to establish a combination between a detremiistic model plus a stochastic noise. Our finding suggest that the patterns obtained could be associated with the local geophysical structures of each subduction and dispersion zones driven by their characteristic nonlinear dynamical features of each region.
Acknowledgments
This work was supported by the Irreversible Processes Physics Research Area of Universidad Autónoma Metropolitana, México. ARR, LRMT and RTPH thanks to Basic Sciences Department of UAM. IRR thanks UPIITAIPN México. ARR thanks the bilateral Project CNR (Italy) and CONACyT (México).
References
 1.
N. Marwan, A historical review of recurrence plots Eur. Phys. J. Special Topics 164, 3–12 (2008)  2.
Eckman, J.P., Oliffson Kamphorst, N.S. and D. Ruelle, D., Recurrence Plots of Dynamical Systems, Europhys. Lett., 4 (91, pp. 973977 (1987)  3.
Gary McGuire, G., Azar, N.B., Shelhamer, M., Recurrence matrices and the preservation of dynamical properties, Physics Letters A 237 (1997) 4347]  4.
Marwan,N., Kurths, J., Saparin, P., Generalized recurrence plot analysis for spatial data, Physics Letters A 360 (2007) 545–551]  5.
(Marwan,N., Carmen Romano, M., Thiel, M., Kurths, J., Recurrence plots for the analysis of complex systems, Physics Reports 438 (2007) 237 – 329  6.
Mendoza, B., FloresMárquez, E. L., RamírezRojas, A. and MartínezArroyo, A., Possible dependence between the total solar irradiance and dimethylsulphide, Geofísica Internacional 48 (4), 351360 (2009)  7.
Varotsos, P.A., Sarlis, N.V.,, E.S. Skordas, E.S., Spatiotemporal complexity aspects on the interrelation between Seismic Electric Signals and seismicity Pract. Athens Acad. 76 (2001) 294–321.  8.
Varotsos, P.A., Sarlis, N.V., E.S. Skordas, E.S., Attempt to distinguish electric signals of a dichotomous nature, Phys. Rev. E 68 (2003) 031106.  9.
RamírezRojas, A. and FloresMárquez, E.L., Order parameter analysis of seismicity of the Mexican Pacific coast, Physica A 392 (2013) 2507–2512  10.
Zuñiga, F.R. and Wyss, M., Inadvertent changes in magnitude reported in earthquake catalogs: Their evaluation through bvalue estimates Bull. Seismol. Soc. Am. 91 (6) (2001) 1717–1728.  11.
ValverdeEsparza, S.M.,, RamírezRojas, A., FloresMárquez, E.L. and Telesca, L., NonExtensivity Analysis of Seismicity within Four Subduction Regions in Mexico, Acta Geophysica vol. 60, no. 3, Jun. 2012, pp. 833845, DOI: 10.2478/s1160001200121  12.
Takens, F.. Detecting strange attractors in fluid turbulence. In D. Rand and L S. Young, editors, Dynamical Systems and Turbulence, pages 366{381. Springer, Berlin, 1981.  13.
Fraser A. M., Swinney H. L.: Independent coordinates for strange attractors from mutual information. Phys. Rev. A, 33 (1986), p. 1134.  14.
Abarbanel, H.D. I., Brown, R., Sidorowich, J.J. and Tsimring, L.S., The analysis of observed chaotic data in physical systems. Reviews of Modern Physics, 65 (4), (1993)13311392.]  15.
Kennel, M.B., Brown, R., and Abarbanel, H.D.I., Determining minimum embedding dimension using a geometrical construction. Physical Review A, 45(1992) 34033411.  16.
Schinkel, S., Marwan, N., Dimigen, O., Kurths, J., Confidence bounds of recurrencebased complexity measures, Physics Letters A 373 (2009) 2245–2250  17.
Aki K., Maximum likelihood estimate of b in the formula log N = abM and its confidence limits, Bull. Earthquake Res. Inst. Tokyo Univ., 43, (1965) 237–239  18.
Telesca L., Lovallo M., RamirezRojas A., FloresMarquez L., Investigating the time dynamics of seismicity by using the visibility graph approach: Application to seismicity of Mexican subduction zone, Physica A 392, (2013) 6571–6577.  19.
Telesca, L., Lapenna, V., Macchiato, M., Multifractal fluctuations in seismic interspike series, Physica A 354 (2005) 629–640.  20.
Thiel, M., Carmen Romano, M., Kurths, J., How much information is contained in a recurrence plot? Physics Letters A 330 (2004) 343–349  21.
Marwan, N. and Kurths, J., Line structures in recurrence plots, Physics Letters A 336 (2005) 349–357