A sequence of major seismic events reaching Ms 7.3, with thrust faulting mechanisms, occurred from 1891 to 2003 in the central Tell Atlas of Algeria located along the Africa–Eurasia plate boundary. Previous neotectonic investigations show that earthquake faults of the central Tell Atlas have the potential to generate large magnitude earthquakes. We calculate the level of stress change that promotes the occurrence of a seismic sequence, taking into account the earthquake fault parameters, their uncertainties, the eastward earthquake migration, the seismicity rate change, and the interseismic strain accumulation. The computed coulomb failure function (ΔCFF) includes the seismicity rate and the stress transfer with fault interaction. The ΔCFF modeling shows 0.1–0.8 bar increase on fault planes at 7 km depth with a friction coefficient μ=0.4 with stress loading lobes on targeted coseismic fault zone and location of stress shadow across other thrust‐and‐fold structures of the central Tell Atlas. The coulomb modeling suggests a distinction in earthquake triggering between moderate‐sized zones and large earthquake rupture zones. Following the 2003 earthquake, Global Positioning System, Interferometric Synthetic Aperture Radar, leveling studies, and aftershocks show that postseismic cumulative moment release amounts to 17.08%, which suggests an additional static stress change. In addition, the presence of fluid and related poroelastic deformation is considered as another parameter that favors stress increase and fault interaction. Modeling the stress change and fault interaction near major cities may contribute to a better constraint of the seismic‐hazard assessment and risk mitigation in northern Algeria.

The active tectonics along the Africa–Eurasia plate boundary in the western Mediterranean is associated with shallow large earthquakes (Mw>6.5) and stress distribution on fault ruptures. The north‐northwest–south‐southeast to northwest–southeast‐trending convergence between the tectonic plates reaches 45  mm/yr and at 23  mm/yr are accommodated across the Tell Atlas of Algeria (Serpelloni et al., 2007; Meghraoui and Pondrelli, 2012). Although the east–west‐trending seismicity may seem diffuse, large and moderate earthquakes (with Mw>5.5) in the Tell Atlas can be localized on fold‐related faults visible at the surface (Fig. 1; Meghraoui, 1988; Benouar, 1994; Ayadi and Bezzeghoud, 2015). These earthquake ruptures can be driven by the stress transfer and interaction between faults that typically result in the occurrence of earthquake sequences. The Zemmouri earthquake (21 May 2003, Mw 6.8) correlated with an offshore reverse fault revealed an average 0.5 m coastal uplift (Meghraoui et al., 2004) and afterslip that brought coulomb failure to the nearby active faults (Lin et al., 2011; Cetin et al., 2012). Previously, it was calculated that the stress on the east section of the Sahel fold increased by 0.4 bar, which may be considered as a sufficient accumulated loading for promoting an Mw>6.8 earthquake (Lin et al., 2011). A major issue for the seismic‐hazard assessment is whether the Sahel‐fault‐related fold is an analog to the El Asnam structure and may hence generate a large earthquake following the 2003 Zemmouri event.

The coulomb modeling may help in understanding the stress distribution and explain the fault interaction and implications for the seismic hazard of the central Tell Atlas. The fault interaction also depends on the inherited regional stress built up from the tectonic background and faulting system in a deformation zone. Early studies proposed the theory of static stress transfer and earthquake interactions as a triggering mechanism of earthquakes (e.g., Das and Scholz, 1981; Reasenberg and Simpson, 1992; Stein, 1999). The application of coulomb static stress modeling benefited from significant results on strike‐slip faulting (Stein et al., 1997), but studies on stress transfer with results on reverse faults remain limited. In their stress transfer study of south California, Bowman and King (2001) fix 0.4 bar as a minimum static stress‐change of sufficient value able to trigger the 1994 Northridge earthquake (Mw 6.7). Among the sequence of large earthquakes of the Tell Atlas, the 10 October 1980 El Asnam event (Mw 7.1) revealed 36  km long thrust faulting and coseismic folding (Fig. 1; Ouyed et al., 1981; Philip and Meghraoui, 1983). The sequence ends with the 2003 Zemmouri earthquake and related seismicity east of the Mitidja basin, but left the Sahel thrust fault segment as a seismic gap (Ayadi et al., 2008). Maouche et al. (2011) describe the extent of active deformation of the 60‐km‐long Sahel fold and estimate 12  mm/yr rate of vertical movements during the Late Pleistocene. Other active faults with seismogenic potential have been identified in the Tell Atlas in relation with historical earthquakes (Meghraoui, 1988).

In this work, the sequence of main historical earthquakes with Mw>5.5 is presented along with thrust fault parameters of the central Tell Atlas of Algeria. The stress change is based on the earthquake locations and their uncertainties and calculated with regard to parallel and en echelon known thrust faults. The coulomb modeling is performed for each seismic event, taking into account the coseismic and significant postseismic deformation along rupture planes of large earthquakes. The results of stress transfer, fault interaction, and related uncertainties are discussed, taking into account the rupture parameters and the elastic characteristics of coulomb modeling. The stress transfer results are presented to be incorporated in a realistic assessment of the seismic hazard and risk in northern Algeria.

The seismicity catalog of northern Algeria includes numerous earthquakes of magnitude Mw5.5 over the last decades (Benouar, 1994; Ayadi and Bezzeghoud, 2015; Fig. 1). This activity mostly shows reverse rupture mechanisms located in the central Tell Atlas specifically associated with faults in Quaternary basins. In this context of convergence along a plate boundary, en echelon fault‐related folds related to shallow seismicity can be interpreted as a transpression deformation with clockwise block rotation (Fig. 2; Jackson and Molnar, 1990; Morel and Meghraoui, 1996). Kinematic modeling (bookshelf) with clockwise block rotation upon a vertical axis and paleomagnetic results illustrate crustal deformation with simple pure shear tectonic pattern that applies to oblique plate convergence in western Mediterranean (Derder et al., 2011; Meghraoui and Pondrelli, 2012).

The seismic sequence investigated in this study includes the following earthquakes (Fig. 1): 1891 (Mw 6.5), 1910 (Mw 7.0), 1922 (Mw 6.1), 1934 (Mw 5.5), 1954 (Mw 6.7), 1959 (Mw 5.9), 1980 (Mw 7.1), 1988 (Ms 5.6), 1989 (Mw 6.0), and 2003 (Mw 6.8). Most of these earthquakes have been the site of tectonic investigations and seismic data analysis that allow for a better accuracy of the epicentral locations and related fault‐rupture characteristics (Fig. 3 and Table 1).

The 15 January 1891 Montebourg (now Gouraya) earthquake was first studied by Rothé (1950) and more recently by Maouche et al. (2008) from contemporaneous reports that indicate the zone of maximum damage (IX European Macroseismic Scale). Rothé (1950) reports coseismic surface breaks that consist of 40‐cm‐wide rupture striking northeast–southwest crossing the Montebourg coastal village showing more than 30±10  cm coastal uplift, measured from a white strip of algae (similar to those observed after the 2003 Zemmouri earthquake, Meghraoui et al., 2004). In their tectonic and geomorphologic investigations, Maouche et al. (2008) observe a 20‐km‐long and northeast‐southwest‐trending thrust fault zone oblique to the coastline and dipping southeast in agreement with the coastal uplift. Although with uncertainties, the 20±2  km fault length estimated from the size of the Quaternary basin, shallow hypocenter (10  km) from the severe damage distribution, and inferred 1.2  m slip at depth imply an estimated Mw 6.5.

The 24 June 1910 earthquake that severely damaged the city of Aumale (now Sour El Ghozlane; see Fig. 3) was affected by an Ms 6.6 event. The magnitude determination was obtained from teleseismic amplitudes and period readings, a teleseismic epicenter relocation obtained from contemporary bulletins of 22 Euro‐Mediterranean seismic stations and damage distribution from press reports to define a VIII Medvedev–Sponheuer–Karnik (MSK) zone of maximum damage (Ambraseys et al., 1991; Benouar, 1994; Table 1). The 1910 earthquake occurred southeast of the Blida thrust‐and‐fold system and can be considered as an off‐sequence seismic event. Our field investigations in tectonic geomorphology show a minimum 38‐km‐long and east‐northeast–west‐southwest‐trending, north‐dipping fault plane with reverse mechanism (Figs. 1 and 3). Using the Wells and Coppersmith (1994) relations, we obtain Mw 7.0 and infer an average 2.2 m coseismic slip at depth, and 12‐km‐wide rupture comparable to other large earthquakes of the Tell Atlas.

The 25 August 1922 Tenes‐Abou El Hassan earthquake was studied by Rothé (1950) from contemporaneous scientific and press reports that describe the maximum damage area with isoseismals reaching VIII MSK at the Cavaignac village (now Abou El Hassan located 14  km southwest of Tenes city). The earthquake epicenter relocated by Benouar (1994) a few kilometers east of Cavaignac is in agreement with the location of the northeast–southwest‐striking, 10‐km‐long reverse fault scarp (Aoudia and Meghraoui, 1995). The fault zone is related to an active fold and displays a maximum 1 m vertical offset of alluvial terraces and 0.6  m average vertical slip along strike that correspond to Mw 6.1 (Table 1).

Using seismic records of nine stations and the standard Prague formula, Benouar (1994) calculates the Ms 5.1 event and relocates the epicenter of the 7 September 1934 earthquake (Table 1). The earthquake epicenter is located at 3 km northeast of the Carnot village (now El Abadia), which is located at the northeastern end of the 1980 El Asnam earthquake fault (Fig. 3). Although fault parameters are difficult to determine for moderate earthquakes (Mw<6), we estimate 0.25±0.1  m coseismic slip on an 5‐km‐long, north‐dipping fault plane from field observations of the flexural slip faulting (Meghraoui et al., 1986). The damage distribution testifies for the shallow hypocentral depth (5  km) from which we estimate Mw 5.5 (Table 1).

The 9 September 1954 Orléansville (named El Asnam after July 1962 and now Ech Chelif) earthquake occurred on a blind fault and caused extensive damage (I0 IX MSK; Rothé, 1955) in the region. Benouar (1994) calculates the magnitude Ms 6.7 from surface waves of 17 stations and fixes the epicentral location (given by the Bureau Central de Sismologie de Strasbourg, Table 1) based on the secondary surface breaks and area of maximum damage mapped by Rothé (1955). The focal mechanism with east‐northeast–west‐southwest‐trending and north‐dipping reverse fault given by McKenzie (1972) is in agreement with the relocated 72 earthquake aftershocks (with M>4.5 from 1954 to 1987) combined with mainshock focal mechanism and Mw 6.4 proposed by Dewey (1991). In addition, the coseismic geodetic (leveling) investigations and related dislocation model suggest Mw 6.7 (M01.1×1019  N·m, Table 1) on a 21±2‐km‐long fault, striking 217°, dipping 67°, and with 3±0.2  m coseismic slip (Bezzeghoud et al., 1995).

The 7 November 1959 Bou Medfaa earthquake is moderate with Mb 4.9 (from body waves) but reaches the intensity VIII MSK (Benouar, 1994). However, along with the strike‐slip focal mechanism determination, Girardin et al. (1977) estimated a larger magnitude Ms 5.9 in accordance with the level of damage distribution. Inferred from the focal mechanism solution and the extent of the damage zone, the fault parameters show 9‐km‐long east‐southeast–west‐northwest‐trending reverse fault and 7‐km‐long fault width associated with 0.5±0.1  m left‐lateral slip, which result in Mw 5.9 (Table 1).

The 10 October 1980 El Asnam earthquake has been the site of numerous studies with extensive documentation on the magnitude determination, epicentral location, and surface faulting (e.g., Ouyed et al., 1981; Philip and Meghraoui, 1983; Bezzeghoud et al., 1995; Table 1). Because of the large earthquake magnitude (Mw 7.1), aftershocks distribution, and geodetic analysis (Bezzeghoud et al., 1995), fault parameters are well constrained (Table 1) and the mainshock epicentral location is with limited error range (less than 1 km; Ouyed et al., 1981; Yielding et al., 1989). On the basis of earthquake‐faulting mechanisms and relocated 72 earthquakes (with M>4.5 from 1954 to 1987) made by Dewey (1991), and geodetic data with dislocation modeling provided by Bezzeghoud et al. (1995), it is concluded that the 1954 and 1980 earthquake ruptures may have occurred on two different and nearly parallel fault planes with 6  km offset that also corresponds to the distance between the two earthquake epicenters (Fig. 4a).

The 30 October 1988 Oued Djer earthquake is in the same region and comparable to the 1959 earthquake. Fault parameters are obtained from the focal mechanism solution (Global Centroid Moment Tensor [CMT]) with northeast–southwest‐striking rupture and Mw 5.6, which results in 5‐km‐long and wide fault dimension, and 0.12±0.05  m slip (Table 1).

The 29 October 1989 Tipaza earthquake reached intensity VIII (MSK) with minor surface breaks, magnitude Mw 6.0, and reverse fault mechanism from the reading of P‐wave arrivals (Meghraoui, 1991) similar to the Global CMT solution. The relocated epicenter is from near‐field stations, and fault parameters with 0.4±0.1  m average slip distribution is obtained from coseismic surface ruptures and seismic source analysis (Table 1; Meghraoui, 1991; Bounif et al., 2003).

The 21 May 2003 Zemmouri earthquake has been the site of numerous studies that document the mainshock and aftershock epicenter locations (using near‐field seismic stations) and the thrust fault mechanism. The inferred 50‐km‐long fault and 1.4±0.1  m average slip were obtained from modeling of coseismic and postseismic deformation using Interferometric Synthetic Aperture Radar (InSAR; Table 1; Delouis et al., 2004; Meghraoui et al., 2004; Ayadi et al., 2008; Mahsas et al., 2008; Belabbes et al., 2009; Cetin et al., 2012).

The seismic sequence described above shows an earthquake migration probably governed by fault interaction and stress transfer comparable to other seismogenic fault zones (Stein et al., 1997; Fereidoni and Atkinson, 2015). Previous coulomb modeling in northern Algeria shows the increase of stress perturbation and takes an average coefficient of friction μ=0.4 for thrust faults that promotes earthquake triggering (Lin et al., 2011). Indeed, the static stress change is computed on fixed fault planes due to the well‐constrained earthquake fault geometries of the thrust fault system of north‐central Algeria (McKenzie, 1972; Ouyed et al., 1981; Meghraoui, 1988; Bezzeghoud et al., 1995; Meghraoui et al., 2004; Ayadi et al., 2008; Mahsas et al., 2008; Belabbes et al., 2009; Cetin et al., 2012).

Static Stress Modeling and Fault Interaction

The static stress change is computed on fixed receiver faults. Tables 1 and 2 summarize the fault source parameters (i.e., fault geometry and dimensions, average slip, and mechanism) used as an input to calculate the stress change. The computed coulomb failure function (ΔCFF; Reasenberg and Simpson, 1992) is expressed by
(1)ΔCFF=Δτμ(ΔσnΔP)
(2)ΔCFF=ΔτμΔσn,
in which τ is the shear stress, σn is the normal stress (compression positive), ΔP is the pore pressure change, and μ and μ are the coefficient of friction and the effective component, respectively. Previous studies suggest that μ is 0.6–0.9 for most geological materials (Byerlee, 1978), and the apparent friction coefficient used in triggered seismicity is defined by the combination of pore pressure and friction coefficient
(3)μ=μ(1ΔPΔσn)
(Beeler et al., 2000). The static stress‐change modeling for earthquake triggering was studied by Reasenberg and Simpson (1992) and Stein et al. (1994) where 0.1–0.8 bar were obtained from their elastic dislocation models and considered as sufficient values promoting failure in a seismogenic crust. Furthermore, Stein et al. (1994) point out that the 1‐bar stress change corresponds to more than a decade of secular stress build up in the greater Los Angeles area; they calculate that the 1933 M 6.4 Long Beach and 1952 M 7.3 Kern County shocks raised the coulomb stress at the site of the 1971 San Fernando (Mw 6.5) and 1994 Northridge (Mw 6.7) earthquakes by at least 0.1 bar.

Rupture characteristics are incorporated in our ΔCFF modeling to test different values of the effective friction coefficient μ. The ΔCFF modeling with μ=0.4 applied to all fault ruptures provides the optimum stress loading associated with the earthquake sequence. Jaumé and Sykes (1996) and Cocco and Rice (2002) explain the low value of effective friction coefficient in the presence of high fluid pressure P. As shown by Lin et al. (2011) in their modeling following the 2003 Zemmouri earthquake, the use of a relatively low effective friction coefficient μ=0.4 implies a high Skempton coefficient and high pore pressure change.

The earthquake migration and possible related fault interaction in the central Tell Atlas can be modeled to compute the stress transfer. We use coulomb 3.4 software (Toda et al., 2011) based on the conversion of dc3d code (Okada, 1992) in MATLAB functions to calculate the ΔCFF. For the modeling procedure, we assume a 0.25 typical value of Poisson ratio with 8×105  bar for the Young’s modulus and 3.3×105  bar shear modulus in the seismogenic layer (10–15 km thickness). The ΔCFF is computed on rectangular dislocations with uniform slip in homogeneous and isotropic media. The computation also includes a nonuniform slip model for the 2003 Zemmouri fault rupture (see below). A fault database with specific format was prepared to test the modeling with fixed oriented planes (see Tables 1 and 2) and related slip per event inferred from the seismic moment for each earthquake. The slip is an average estimate that matches the fault length according to Wells and Coppersmith (1994). For each earthquake fault and computed stress change, a mid‐distance location in the seismogenic zone (7 km depth) is determined as appropriate for an optimum ΔCFF calculation (Table 3). A selected area of 400  km×200  km dimension is gridded every 1 km and employed for the best‐fit interpolation of the ΔCFF modeling. The dislocation model (Okada, 1992) using a selection of source ruptures and receiver en echelon faults allows us to identify regions of stress loading and of stress shadowing (Reasenberg and Simpson, 1992; Figs. 2 and 3).

Earthquake Sequence and Static Stress Change in the Central Tell Atlas

Among the largest earthquakes (Mw6) that affected in sequence the central Tell Atlas, the 1954 Orléansville (Mw 6.7), 1980 El Asnam (Mw 7.1), 1989 Tipaza (Mw 6.0), and 2003 Zemmouri (Mw 6.8) seismic sequence suggest the northeastward migration (Fig. 1 and Table 1). In comparison with the stress transfer between juxtaposed strike‐slip faults (Stein et al., 1997), the earthquake activity in the central Tell Atlas also appears to be controlled by the en echelon pattern of fault‐related folds (Fig. 2). Here, we assume that the northeastward earthquake migration in the central Tell Atlas is due to the fault rupture interaction and related stress transfer.

In our study region, the onset of major earthquakes in the sequence are the 1954 Mw 6.7 and the 1980 Mw 7.1 earthquakes that occurred in the Chelif basin. Previously, two significant earthquakes, the 1891 earthquake Mw 6.5 and the 1922 Mw 6.1 earthquake that occurred along the coastline, can be responsible for the increased stress loading (>0.1  bar) further south in the Chelif basin (Figs. 1 and 3). The stress‐change modeling shows that the 1891 and 1922 earthquakes occurred on fault sources that favor the stress transfer on the 1954 and 1980 fault‐plane receivers. Indeed, the ΔCFF modeling of the 1891 and 1922 earthquake ruptures at 7 km depth indicates a cumulative stress loading on the 1954 fault plane with values >0.1  bar (see Table 3 and Ⓔ Figs. S1 and S2, available in the electronic supplement to this article).

Considering the 1954 stress loading only, the ΔCFF modeling at 7 km depth suggests a static stress load higher than 2 bar as sufficient to trigger the subsequent Mw 7.1 El Asnam 1980 earthquake rupture (Fig. 4a,b and Ⓔ Figs. S3 and S4). As put forward by Dewey (1991) and Bezzeghoud et al. (1995), the occurrence of the large 1954 earthquake and the existence of two different seismogenic fault segments implies a significant stress loading on the 1980 fault zone (7.9 bar in Table 3; see also the Seismotectonics and Fault Parameters section).

The cumulative stress loading caused by the 1954 Mw 6.7 and 1980 Mw 7.1 earthquakes may be at the origin of the northeastward earthquake sequence migration and the occurrence of the 1959 Mw 5.9, 1988 Mw 5.5, the 1989 Mw 6.0, and the 2003 Mw 6.8 earthquakes (Ⓔ Fig. S5). Considering the trending earthquake sequence, the 1954, 1959, 1980, and 1988 earthquakes are taken as source faults and 1989 and 2003 as the receiver faults. The ΔCFF computation denotes a positive loading value reaching 0.3 bar in the epicentral area of the 1989 earthquake fault (Ⓔ Fig. S4). The stress loading reaches 0.2 bar for the modeled receiver plane when considering the 1959 and 1988 earthquake ruptures, whereas the limit value to rupture for the 1989 earthquake average 0.6–0.8 bar. Considering the cumulative stress and the coulomb stress drop acting before and after an earthquake comparable to the 1989 seismic event, these stress loading values are in good agreement with those obtained for moderate earthquakes (M5.5) in the southern California region (Jaumé and Sykes, 1996; Bowman and King, 2001).

The large 2003 Zemmouri earthquake is the most recent seismic event of the sequence in the central Tell Atlas (Fig. 1). Our ΔCFF calculation includes all historical seismic events (see Table 1 and Ⓔ Fig. S6) in addition to the large earthquake of 24 June 1910, Mw 7.0 (Table 1). Here, the stress‐change computation also takes into account the nonuniform coseismic slip obtained from the inversion of coastal uplift, Global Positioning System (GPS), and strong‐motion inversion data (Ⓔ Fig. S7). The slip distribution is based on 128 rectangular patch dislocations, 3 m maximum slip, and 5.9×1019  N·m (Mw 7.1) seismic moment (Semmane et al., 2005). The ΔCFF calculations for uniform and nonuniform slip (see Fig. 5) indicate comparable stress loading reaching 1  bar covering the 2003 epicenter area (Fig. 6). Although the epicentral region of the 1910 seismic event occurred at 40  km south of the Zemmouri region on a comparable east‐northeast–west‐southwest‐trending fault zone (Benouar, 1994; see also the Seismotectonics and Fault Parameters section for earthquake location), the ΔCFF modeling shows that it contributes to the loading of the Mitidja region. Our predictive model for stress loading takes into account the tensorial approach applied in southern California (Jaumé and Sykes, 1996; Bowman and King, 2001) and the clock time advance based on the ratio of the maximum static stress change computed on a receiver fault and the stressing rate (Stein et al., 1997). In our case, the stress loading value necessary for triggering the 2003 earthquake reaches 1.4±0.2  bar (Fig. 5 and Ⓔ Fig. S7) taking into account the 375  yrs recurrence time necessary to nucleate an event of magnitude Mw7 in the central Tell Atlas (Beghoul et al., 2010).

From the tectonic point of view, the 2003 epicentral region limits to the east the Kabylia block, to the west the Mitidja basin and Sahel anticline (Ayadi et al., 2008), and to the southwest the Blida fold‐and‐thrust system (Figs. 3 and 5). The fault interaction and stress change indicate a clear influence of the western section of the 2003 aftershock sequence (Fig. 3 and Ⓔ Fig. S7) on the Mitidja region and specifically on the Blida fold‐and‐thrust structure which generated large (Mw 6.5 in 2 March 1825 in Ayadi and Bezzeghoud, 2015) and moderate earthquakes (Mw>4.5). For comparison, moderate‐magnitude earthquakes in the central Apennines (Italy) significantly increase (by 40%–110%) the probability of occurrence of a large earthquake on nearby seismogenic fault zones (Pace et al., 2014). In fact, the long‐term background seismicity of the Mitidja region may explain the seismic gap between the 1989 and 2003 earthquakes (Meghraoui, 1988; Maouche et al., 2011; Ayadi and Bezzeghoud, 2015; Fig. 5 and Ⓔ Fig. S8a). The 60‐km‐long Sahel‐faulted fold being a major potential source for a large magnitude event (Mw>7; Maouche et al., 2011), our ΔCFF modeling shows a positive loading that requires a higher coefficient of friction (μ>0.4) for reactivation.

We also explore the impact of the postseismic stress change of the most recent large earthquakes that occurred in Tell Atlas (1980 and 2003; see Table 1, Fig. 5, and Ⓔ Fig. S4). More specifically, we estimate the 2003 cumulative aftershock seismicity (Ayadi et al., 2008), where the aftershocks data are converted into cumulative seismic moment to obtain the postseismic strain release; in addition, we use the results of permanent scatterer InSAR (PS‐InSAR)–GPS analysis and slip model of Cetin et al. (2012) as applied to the 2003 Zemmouri earthquake (M02.4×1019  N·m) to model the postseismic stress field. The InSAR time series (Cetin et al., 2012) and maximum line of sight displacements are calculated from 2004 to 2010, and compared with the coseismic displacements (Belabbes et al., 2009). We observe that the postseismic cumulative moment release amounts to 9.16% when considering the aftershock events but reaches 17.08% when taking into account the postseismic geodetic results which suggests an additional static stress transfer. A similar comparison of seismic activity between mainshock and aftershocks of the 1980 El Asnam earthquake (M0  5.94×1019  N·m) indicates 9.59% for the postseismic moment release. A calculation of the background seismicity rate change for the area of the 2003 Zemmouri earthquake (from 2003 to 2015) provides a maximum of 2  earthquakes/yr/100  km2 (Ⓔ Fig. S8a), comparable to the results obtained by Lin et al. (2011). We compute a seismicity rate change by combining Ayadi et al. (2008) and Berkeley catalog from the period 1964–2015 (Ⓔ Fig. S8b). In addition to the earthquake catalogs of Benouar (1994) and Ayadi and Bezzeghoud (2015), we took into account the Berkeley catalog data (1964–2015) in readable format to calculate the cumulative seismic moment and the background seismicity rate. The magnitude of completeness is 3.6 (maximum‐likelihood estimation) with 0.19 uncertainty obtained by bootstrapping (Fig. 7) using Zmap software (Wiemer, 2001). The seismicity rate change represents the ratio calculated before and after the 2003 mainshocks, where the red zone in Ⓔ Figure S8b indicates 3× increase in seismicity after the 2003 earthquake, in good agreement with the postseismic stress change.

The seismotectonics and stress transfer study in north‐central Algeria shows a predominantly northeastward migrating sequence from the 1891 to 2003 earthquakes. Our ΔCFF modeling shows that the seismic sequence is apparently controlled by thrust fault interaction in the Tell Atlas. The seismotectonic analysis indicates that the uncertainty in the earthquake sequence locations is relatively low depth error range (ERH<1  km) for the 1980, 1989, and 2003 earthquakes (Ouyed et al., 1981; Yielding et al., 1989; Bounif et al., 2003; Ayadi et al., 2008). As for the 1954, 1959, and 1988 earthquakes, the Algerian seismological network admits a relatively low error range in the epicentral location (ERH<2  km; Ayadi and Bezzeghoud, 2015). Using P‐wave arrivals reported in the International Seismological Bulletin and zones of maximum damage distribution, the uncertainty of the 1910, 1922, and 1934 earthquake locations cannot exceed 5 km (Benouar, 1994). The macroseismic intensity distribution and related surface faulting and uplifted coastal region suggest a 5–10 km error range for the 1891 earthquake epicenter (Rothé, 1950). Although all seismic events are predominantly shallow, a remaining significant uncertainty that may reach ±5  km is linked with the hypocentral depth, mainly for the pre‐1988 earthquakes. Uncertainties in earthquake location and fault strike, dip, and dimensions are reduced, taking into account previous results on surface faulting for the 1891 earthquake (Maouche et al., 2008), the 1922 earthquake (Aoudia and Meghraoui, 1995), the 1954 earthquake (Rothé, 1950; Benouar, 1994, Bezzeghoud et al., 1995), the 1934 and 1980 earthquakes (Philip and Meghraoui, 1983; Meghraoui et al., 1986), and the 1989 and 2003 earthquakes (Meghraoui, 1988; Belabbes et al., 2009).

In our study, we assume a simple rectangular dislocation with uniform slip in a homogeneous medium (with typical shear‐modulus and Poisson‐ratio values) for all earthquake ruptures that subsequently reduces uncertainties in the coulomb modeling. The comparable results of computation of static stress change for the 2003 earthquake rupture reaching 1 bar for nonuniform and uniform slip may well be valid for the other earthquake ruptures of the seismic sequence. As shown in Figures 3 and 5 and Ⓔ Figure S7, the influence of the nonuniform slip model can be considered as minor with regard to the uniform slip solution. Although the 15±2  km thickness of seismogenic layer is relatively well constrained, thanks to the seismicity distribution for major earthquakes (i.e., 1980, 1989, and 2003 earthquakes), we assume that the coulomb modeling performed at 7 km depth corresponds to a realistic solution for the mechanical problem of the seismogenic crust. As thrust faults are the predominant earthquake mechanism, we use published fault parameters to calculate the coulomb stress imparted in the central Tell Atlas region. Fault geometries being well mapped and rupture parameters well constrained from field investigations in epicenter areas, we observe that uncertainties of fault parameters do not change much of the coulomb modeling results. We also note that the change in fault dip (from 30° to 70°) does not affect the amount of stress loading (ΔCFF=0.20.3  bar) for any receiver fault (Fig. 6a,b).

Among the main historical and recent earthquakes, the coulomb modeling at 7 km depth indicates that the 1980 major seismic event (Mw 7.1) significantly changes the stress field in the central Tell Atlas. The structural characteristics of tectonic blocks limited by active and seismogenic faults with en echelon geometry provide the pattern for the stress transfer modeling. A remaining question, however, is the role of fluid migration following the occurrence of large thrust earthquakes, and the study of the poroelastic deformation using constitutive equations (such as Biot’s law) that may constrain the effective stress near the fault zones. Jaumé and Sykes (1992) suggest that one explanation for the apparently low value of μ would be the presence of high fluid pressure. In our study, we show that the coulomb modeling requires an effective friction coefficient of 0.1<μ<0.4, which implies a ΔCFF=0.20.3  bar for most receiver faults (Fig. 6a and Ⓔ Fig. S6) that may explain the earthquake triggering in the sequence.

The earthquake sequence does not move further east or west in the central Tell Atlas and the coulomb modeling indicates an increased static stress in the Mitidja region and in particular on the 60‐km‐long Sahel fold‐related fault. Although the seismic sequence starts with the initial westward migration of 1891 Mw 6.5 and 1922 Mw 6.1 coastal earthquakes that loaded further south of the El Asnam (Orléansville) faulting area, we consider that the impact of the 1954 Mw 6.7 and 1980 Mw 7.1 major earthquakes with northeast migration that includes 1959, 1988, 1989, and 2003 is predominant in the static stress change (see also Table 3 and Ⓔ Fig. S5) and northeastward earthquake migration on the earthquake fault system of north‐central Algeria. The occurrence of the off‐sequence 1910 Mw 7.0 earthquake, which is apparently not consistent with the seismic sequence and related northeastward seismic migration could not be neglected; its ΔCFF helps us understand the state of stress in the Mitidja region, Blida thrust system, and Sahel‐faulted fold structure (see Fig. 3). Indeed, the occurrence of the 1910 earthquake adds to the cumulative ΔCFF and its impact on the seismic hazard of the north‐central Algeria region. Stein (1999) estimated that during an earthquake, 80% of the energy is released as seismic waves, whereas the remaining 20% is merely stored and transferred to different locations along the fault leading to specific regions becoming more susceptible to future earthquakes. Considering the 2003 postseismic strain release where the geodetic moment (that may include aseismic poroelastic and viscoelastic deformation) is twice the aftershocks moment release, the static stress increases from 0.2 to 0.4 bar on the Sahel and Blida thrust faults located in between the 1910 Mw 6.5, 1989 Mw 6.0, and 2003 Mw 6.8 earthquakes (Fig. 5). Our ΔCFF modeling indicates a significant seismic hazard and risk as the Sahel and Blida faults have the potential to generate a large magnitude earthquake near the capital city Algiers (Maouche et al., 2011).

The seismic migration that occurs along the en echelon thrust fault geometry of the Algerian section of the Eurasia–Africa plate boundary is not surprising. Our static stress‐change modeling shows a causative relation between the observed migration of earthquakes, the triggered seismic events, and the positive stress‐change computation (see Ⓔ Figs. S1–S8 for more details). Although the en echelon tectonic structures are complex, King (2007) suggests that the stress change calculated on well‐identified fixed fault planes provides better modeling than when using optimal rupture planes. In our study, the stress transfer analysis using the earthquake sequence and related fault parameters constrained from field investigations provides a better constraint on fault interaction and related failure threshold. This approach based primarily on combined geologic and seismological data is critical for the seismic‐hazard evaluation (Meghraoui and Atakan, 2014).

Because the El Asnam and Sahel thrust‐related fold present comparable geomorphological and tectonic features (thrust system with flexural slip folding and bending‐moment normal faulting), one may also consider a comparable long‐term seismogenic behavior associated with 1–2 mm/yr contraction rate across the Blida and Sahel fault systems (Meghraoui and Pondrelli, 2012). The poroelastic rebound for thrust fault mechanisms in far‐field domains is generally associated with fluid migration (i.e., pore pressure change; Whitcomb et al., 1973) and implies the ΔCFF computation as a time‐dependent phenomenon. The coulomb modeling shows that the Blida and Sahel fault regions are trapped by the positive stress loading of the earthquake sequence (Figs. 3 and 5). The static stress change and increase of promoting failure along the Sahel thrust fault is a major outcome for the earthquake‐hazard evaluation in the Mitidja region.

Seismicity data used in this article came from published sources listed in the references. MATLAB functions are from www.mathworks.com/products/matlab (last accessed January 2016). The Global Centroid Moment Tensor (CMT) Project database was searched using www.globalcmt.org/CMTsearch.html (last accessed October 2015). Some plots were made using the Generic Mapping Tools (GMT) v.4.2.1 (www.soest.hawaii.edu/gmt, last accessed January 2016; Wessel and Smith, 1998).

The authors wish to thank F. Ousadou (Centre de Recherche en Astronomy Astrophysique et Géophysique [CRAAG], Algiers) for help in preparing the seismic database. We are grateful to Robert Simpson (U.S. Geological Survey [USGS]) for reading an early version of the article and Luke Griffiths (Institut de Physique du Globe [IPG] Strasbourg) for checking the new version. We thank Ross Stein, Jian Lin, and Shinji Toda for sharing the Coulomb 3.4 software and for discussion during a previous work. We also thank the three anonymous reviewers for their comments and suggestions on the article. K. J. benefits from a scholarship from the Algerian Ministry of Higher Education and Scientific Research (MESRS). This research program was funded by the Direction of Research at MESRS and the Direction Europe de la Recherche et Coopération Internationale (DERCI‐CNRS) with Institut National des Sciences de l'Univers‐Unité Mixte de Recherche (INSU‐UMR) 7516 IPG Strasbourg. Some figures were prepared using the public domain Generic Mapping Tools (GMT) software (Wessel and Smith, 1998).