Professional Documents
Culture Documents
Active Deformation in Algeria From Continuous GPS Measurements
Active Deformation in Algeria From Continuous GPS Measurements
net/publication/332353387
CITATIONS READS
31 856
3 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Eric Calais on 11 April 2019.
Accepted 2019 January 15. Received 2018 December 17; in original form 2018 September 12
572
C The Author(s) 2019. Published by Oxford University Press on behalf of The Royal Astronomical Society.
Active deformation in Algeria from GPS 573
Here, we use continuous GPS data from the Algerian REGAT strike slip on EW–trending faults, such as the one along which the
(REseau Gps de l’ATlas) geodetic network to directly determine Guelma pull-apart basin is currently forming (Harbi et al. 1999;
the spatial distribution of deformation along the Nubia–Eurasia Maouche et al. 2013). Active faulting is mapped over a broad area
(NU–EU) Plate boundary along most of its trace in the Western that covers the Aurès Mountains and the southern Atlas along its
Mediterranean. We show that the data is best fit by a simple kine- boundary with the stable Sahara platform, mapped by Ben Hassen
matic model where the Nubia–Eurasia oblique convergence is ac- et al. (2014) as the Gafsa right-lateral transpressive fault zone.
commodated mostly by offshore compression in the western half A recent advance in our understanding of active deformation in
of Algeria, and, in its eastern half, by onshore strike-slip combined Algeria results from exhaustive bathymetric mapping and seismic
with shortening both offshore as in the west, and inland across the reflection acquisitions across the Algerian margin following the
Atlas domain. 2003 Boumerdes earthquake (Deverchère 2005). The data shows a
series of crustal-scale reverse faults that emerge at the toe of the
Algerian continental margin, at least from longitude ∼1◦ E eastward
(Domzig et al. 2006; Yelles et al. 2009; Medaouri et al. 2014; Hamai
2 TECTONIC SETTING
EURASIAN PLATE
GEODETIC 5mm/yr
Altamimi et al., 2017
Nocquet et al., 2006
Calais et al., 2003
This work
39˚ 40˚
Fa
ia ul
Glor t
Azores Sy
ste
34˚ m
GEOLOGIC 5mm/yr
Argus et al., 2011
DeMets et al., 2015 NUBIAN PLATE
-20˚ -10˚ 0˚ 10˚ 20˚
Figure 2. Predicted velocities along the Eurasia–Nubia Plate boundary (simplified trace from Bird 2003) according to the plate angular velocities shown on
Fig. 3.
Active deformation in Algeria from GPS 575
Table 1. Angular velocities describing the Nubia–Eurasia relative plate motion. Geological (3.16 Myr averages) and geodetic (present-day) estimates are
reported.
Data set Lat. (◦ ) Lon. (◦ ) Rate 95 per cent conf. err. ell. σ -Rate
10˚
4 R E S U LT I N G V E L O C I T Y F I E L D
Among the 54 REGAT sites processed here, 12 unfortunately show
time-series with non-linear excursions that are unlikely tectonic in
origin and velocities that are significantly different from their near- 0˚
est neighbors (sites AKET, CAEH, CCOL, CKAL, CKTS, EBGR,
EBNH, EKMS, OBBL, OJGS, OSDA and OTSS). Such time-series Nocquet et al., 2006
are correlated with specific site locations (e.g. active dams) or poor
monument construction. We will not consider them further below. Calais et al.,
The 42 remaining sites show clean, linear time-series with a mean 2003 This work
long-term weighted RMS ∼1.5 mm. Velocity uncertainties decrease -10˚
as a function of observation time span to reach 0.2 mm yr−1 (one
standard deviation) for the oldest stations, which have close to 10 yr Altamimi
of continuous data, such as ABZH for instance (Fig. 4). et al., 2017
Figs 5(a) and (b) show velocities in EU-fixed and NU-fixed
frames, respectively, together with model boundary velocities. Ve-
Figure 3. Euler pole locations for Nubia–Eurasia, with their at 95 per cent
locities at all the solution sites (REGAT plus IGS) are provided in
confidence ellipse (see also Table 1), comparing the recent geological es-
Table S1. We define the EU and NU frame by minimizing veloci- timates of Argus et al. (2011) and DeMets et al. (2015) with the geodetic
ties at 65 sites with well-determined velocities, 58 located on stable estimates of Calais et al. (2003), Nocquet et al. (2006), Altamimi et al.
Eurasia, 7 on stable Nubia (sites indicated by a star in Table S1). (2017) and this work. The corresponding predicted velocities along the
Fig. 3 and Table 1 compares our NU–EU angular velocity with re- Nubia–Eurasia Plate boundary are shown on Fig. 2.
cent geodetic (Calais et al. 2003; Nocquet et al. 2006; Altamimi
et al. 2017) and geological (Argus et al. 2011; DeMets et al. 2015) In the western part of the study area, which encompasses the west-
estimates. We find an Euler pole that is consistent with previously ern Tell Atlas and western High Plateaus, velocities are collinear
published geodetic poles, but inconsistent with the geological ones, with the NW–SE Nubia motion directed N50W towards Eurasia,
located about 2500 km to the north, as previously recognized by with magnitudes of about 3 mm yr−1 , similar to those predicted by
Calais et al. (2003). the rigid EU–NU Plate motion. Deformation is therefore concen-
Velocities in the study area shows oblique convergence between trated in a narrow coastal region, while the western High Plateaus
EU and NU, consistent with the seismological and tectonic observa- appear to belong to stable Nubia. This is also readily visible in the
tions. In the EU frame (Fig. 5a), we observe velocities at the coast NU-fixed frame (Fig. 5b), which shows small residual velocities at
that decrease progressively from west (maximum 2.5 mm yr−1 at sites located in the western High Plateaus where the weighted rms
site OLHC) to east (close to zero at site CMAR). This is a first of the residuals is 0.8 mm yr−1 .
indication that the EU–NU convergence may be accommodated by East of longitude ∼3◦ E, velocities at sites located within about
offshore structures in western Algeria, but mostly on-land in the 50 km of the coast deviate from the NU–EU Plate motion direction,
eastern half of the country. with a clockwise rotation (Fig. 5a). This pattern is indicative of
576 A. Bougrine, A. K. Yelles-Chaouche and E. Calais
strike-slip faulting, as described in detail below. To the south, ve- The western profile (Fig. 6a), shows 3.5 mm yr−1 of pure
locities remain parallel to the NU–EU Plate motion, but with lower shortening between stable Nubia and site ALME on stable Eura-
magnitudes, indicating shortening within the High Atlas and/or be- sia. The distribution of the parallel-velocities along the profile
tween this region and stable Nubia, as shown when velocities are can be explained by a locked reverse fault located offshore. We
expressed in the NU-fixed frame (Fig. 5b). find that the data are best fit with a fault dipping 40◦ S, with a
10 km locking depth, and a surface trace located 140 km off-
shore within the Alboran basin. Model slip rate on this offshore
5 2-D ELASTIC STRAIN reverse fault, which could correspond to the Alboran ridge fault,
A C C U M U L AT I O N M O D E L S is 5 mm yr−1 . Note that this corresponds, given the fault dip,
to a horizontal velocity of 3.8 mm yr−1 . Profile-perpendicular
We further quantify these observations by building 2-D models of velocities, much smaller, show at most 1 mm yr−1 of slip on
elastic strain accumulation on locked faults embedded in an elastic a vertical strike-slip fault located inland, locked to a depth of
half-space using the classic back-slip approach of Savage (1983), 15 km.
further described in Segall (2010). We seek to fit the simplest fault The central profile (Fig. 6b) shows a combination of profile-
model to the velocity observations by varying fault dip, locking parallel and profile-perpendicular velocities. Profile-parallel veloc-
depth, fault location and slip rate using the weighted root mean ities are well fit by a reverse fault with 40◦ S dip, a 10 km locking
square (WRMS) of the residuals as a measure of the goodness of depth, a surface trace located 10 km offshore, close to the mar-
fit: gin toe. Slip rate on this reverse fault is 5.5 mm yr−1 (horizontal
n 2 2 1/2 velocity of 4.2 mm yr−1 ). It could correspond to the reverse fault-
ri /σi
W R M S = i=1 n 2
, (1) ing system identified offshore by recent seismic reflection profiles
i=1 1/σi (Domzig et al. 2006; Yelles et al. 2009). The velocity gradient of
where r is the model residual, σ the standard error of the observed the profile-perpendicular velocities does not coincide with that of
velocity and n the number of observations. These models do not the profile-parallel ones, hence requiring a second fault. We find
reflect the details of active faulting in Algeria, but are meant to help that the profile-perpendicular velocities are well fit by a vertical
us determine which faults are the major contributors to present- strike-slip fault locked to a depth of 15 km with a 1.3 mm yr−1
day deformation and to provide guidelines for the block models slip rate. Its surface trace, located 50 km inland, corresponds to
described in the next section. the Ghardimaou fault in Tunisia (Bahrouni et al. 2013), which ex-
We project the east and north velocity components onto directions tends westward along-strike as the North Constantine fault in Alge-
that are parallel and perpendicular to three profiles, as shown on ria (Coiffait et al. 1992; Meghraoui & Pondrelli 2012; Rabaute &
Figs 5(a) and 6. The western and central profiles WW’ and CC’ Chamot-Rooke 2014). In the following, we will call this right-lateral
have an azimuth of N140◦ E, parallel to the NU–EU Plate motion strike-slip the ‘Ghardimaou—North Constantine’ (GNC) fault
and to the observed GPS velocities in this region. They are meant (Fig. 1).
to help us identify the fault(s) responsible for the plate motion- The eastern profile (Fig. 6c) shows profile-perpendicular veloc-
parallel shortening readily visible in the velocity field. The eastern ities that are well fit by slip on the eastward continuation of the
profile EE’, perpendicular to the Algerian coast, is meant to capture GNC strike-slip fault mentioned above, with a 15 km locking depth
the strike-slip component of motion suspected from the clockwise and a 2.6 mm yr−1 slip rate. Profile-parallel velocities can be fit
rotation of velocities near the coast. by a reverse fault dipping 40◦ S that corresponds to the eastward
Active deformation in Algeria from GPS 577
Figure 5. GPS velocities shown with respect to Eurasia (top) and Nubia (bottom). Bathymetry and topography are from ETOPO1 1-min Global relief
(www.ngdc.noaa.gov). Major tectonic features are from Rabaute & Chamot-Rooke (2014). Error ellipses are 95 per cent confidence. Dashed lines are the three
profiles onto which the velocity profiles are projected, see Fig. 6. Colour-coding of the velocities indicates the profile to which they belong (blue = west, green
= centre, red = east). Top/bottom arrows show the predicted NU–EU velocities from this work.
continuation of the margin toe reverse fault mentioned for the cen- accumulation on a small number of major faults: an offshore fault
tral profile, with a 2.6 mm yr−1 slip rate (horizontal velocity of along the entire continental margin and an on-land strike-slip fault
2.0 mm yr−1 ). in the eastern part of the study area. They both correspond to active
This simple, 2-D, modelling exercise shows that GPS velocities tectonic structures identified in the geology (Bounif et al. 1987;
in northern Algeria are well explained, to first order, by elastic strain Harbi et al. 1999; Yelles et al. 2004).
578 A. Bougrine, A. K. Yelles-Chaouche and E. Calais
Figure 7. Block geometries tested in models 1–6 with the fault and block names used in the text.
6 BLOCK MODELLING which includes information from regional seismotectonic and geo-
logical studies to date, and the results from the 2-D models described
A proper kinematic interpretation of the GPS velocities requires
above. Offshore faults in the model are locked down to a depth of
accounting for both strain accumulation on locked faults and rigid
10 km and dip 40◦ S, on-land faults are locked to 15 km and are
block rotations. To do so, we use the approach of Meade & Loveless
vertical for strike-slip faults or dip 45◦ for reverse faults.
(2009) and their code ‘Blocks’, which models GPS velocities as a
We run a series of models starting from the simplest possible,
sum of rigid rotations of tectonic blocks and strain accumulation
with a two-plate geometry that we progressively complexify by
on the block-bounding faults. We define block boundaries using
adding new plates, as required by the data. The models are run—
earthquake locations from the CRAAG seismicity catalogue, the
and their statistics computed—using velocities at all sites on the EU
seismotectonic compilation of Rabaute & Chamot-Rooke (2014),
and NU plates, not only those located in Algeria. The geometry of
Active deformation in Algeria from GPS 579
38.0˚ 38.0˚
0.0˚ 2.0˚ 4.0˚ 6.0˚ 8.0˚ 0.0˚ 2.0˚ 4.0˚ 6.0˚ 8.0˚
36.0˚ 36.0˚
34.0˚ 34.0˚
5 mm/yr 5 mm/yr
mm/yr mm/yr
0 1 2 3 4 5 6 0 1 2 3 4 5 6
38.0˚
0.0˚ 2.0˚ 4.0˚ 6.0˚ 8.0˚
R = 73.5%
mu = −0.0, std = 0.7 mm/yr mu = 0.2, std = 1.2 mm/yr
2.3 30 30
2.2
2.2 2.3
2.2
36.0˚
2.2 2.2
2.2 2.1 20 20
percentage
10 10
34.0˚
5 mm/yr
0 0
−2 −1 0 1 2 −2 −1 0 1 2
EW residual (mm/yr) NS residual (mm/yr)
Figure 8. Block model 0. Coloured lines represent the fault geometry tested. (a) Model dip-slip rates resolved along fault dip. (b) Model strike-slip rates. (c)
Residual (observed minus modelled) velocities shown with black arrows; red arrows show model horizontal displacements along block-bounding faults w.r.t.
the block located north of the fault. (d) Histograms of the east and nord velocity residuals, x-axis in mm yr−1 . Solid black lines shows the corresponding normal
distribution with mean (mu) and standard deviation (std), both in mm yr−1 , indicated above each panel.
the models tested, from number 0 to 6, is shown on Fig. 7, with the where χ p21 and χ p22 are the chi-square statistics of two models with
fault and block names used in the text. Each model fit to the data is p1 and p2 degrees of freedom, respectively. We compare this ex-
quantified by its χ 2 : perimental F-ratio to the expected value of a F(p1 − p2 , p1 ) dis-
tribution for a given risk level α per cent (or a 100 − α per cent
n obs 2
confidence level) that the null hypothesis (the decrease in χ 2 is
vi − vimod
χ2 = , (2) not significant) can be rejected. We set the acceptable signifi-
σi2
i=1 cance level to 99 per cent, that is a probability of rejection less than
1 per cent.
where n is the number of observations, vobs and vmod the observed Model 0 (Fig. 8) is meant to test the plate boundary geometry
and modeled velocities and σ the standard error of the observed proposed by Meghraoui & Pondrelli (2012), where the EU–NU
velocities. We also use the variance reduction of the model defined Plate boundary in Algeria consists of a restraining bend, with most
as: of the plate motion accommodated on-land. Its variance reduction
is 74 per cent, with velocity residuals that reach up to 2 mm yr−1 ,
n 2
viobs − vimod whose distribution does not follow a normal distribution, in par-
R =1− i=1
n obs 2 . (3) ticular in the north component. The geometry tested here results
i=1 vi
in 2.2–2.3 mm yr−1 of right-lateral strike-slip on east–west trend-
ing faults, which would result in significant extension in the transfer
The R values quoted below use the GPS velocities in a EU-fixed
zone between the central and eastern segments (around 5◦ W), which
frame. Table S2 shows variance reductions in a NU-fixed frame and
is not observed.
in ITRF, for comparison.
Model 1 (Fig. 9) is an alternate end-member case where the NU–
Since model χ 2 ’s always improve when adding plates, that is
EU relative plate motion is entirely accommodated by an offshore
free parameters, we test the significance of the decrease in χ 2 from
reverse fault along the Algerian margin. Its variance reduction is
a model with fewer plates to a model with more plates using the
79 per cent, with velocity residuals that are less than 1 mm yr−1
F-ratio statistics (e.g. Stein & Gordon 1984) given by:
over most of the study area and follow a normal distribution. Model
2 fit to the data is slightly better than in model 0 in the western part
χ p1 − χ p22 / ( p1 − p2 ) of the domain, but worse in the eastern part, where a systematic
F= , (4)
χ p22 / p2
580 A. Bougrine, A. K. Yelles-Chaouche and E. Calais
38.0˚ 38.0˚
0.0˚ 2.0˚ 4.0˚ 6.0˚ 8.0˚ 0.0˚ 2.0˚ 4.0˚ 6.0˚ 8.0˚
36.0˚ 36.0˚
34.0˚ 34.0˚
5 mm/yr 5 mm/yr
mm/yr mm/yr
0 1 2 3 4 5 6 0 1 2 3 4 5 6
dip−slip strike−slip
percentage
10 10
34.0˚
5 mm/yr
0 0
−2 −1 0 1 2 −2 −1 0 1 2
EW residual (mm/yr) NS residual (mm/yr)
38.0˚ 38.0˚
0.0˚ 2.0˚ 4.0˚ 6.0˚ 8.0˚ 0.0˚ 2.0˚ 4.0˚ 6.0˚ 8.0˚
36.0˚ 36.0˚
34.0˚ 34.0˚
5 mm/yr 5 mm/yr
mm/yr mm/yr
0 1 2 3 4 5 6 0 1 2 3 4 5 6
dip−slip strike−slip
38.0˚
0.0˚ 2.0˚ 4.0˚ 6.0˚ 8.0˚
R = 91.11%
3.1
2.7 mu = −0.0, std = 0.5 mm/yr mu = 0.0, std = 0.6 mm/yr
2.3 30 30
4.0
3.9
2.7 2.7 2.7
3.9 2.7
36.0˚
3.8
20 20
percentage
10 10
34.0˚
5 mm/yr
0 0
−2 −1 0 1 2 −2 −1 0 1 2
EW residual (mm/yr) NS residual (mm/yr)
clockwise rotation is apparent. This residual rotation indicates that margin, as shown in the seismic surveys mentioned above, and on-
at least one additional block is needed to fit the data. land in the eastern part of the study area, as proposed by Meghraoui
Model 2 (Fig. 10) is a hybrid of the previous ones, where the EU– & Pondrelli (2012). This model individualizes an East Tell Atlas
NU Plate motion is accommodated both offshore along the Algerian block separated from the Nubian Plate by the GNC right-lateral
Active deformation in Algeria from GPS 581
38.0˚ 38.0˚
0.0˚ 2.0˚ 4.0˚ 6.0˚ 8.0˚ 0.0˚ 2.0˚ 4.0˚ 6.0˚ 8.0˚
36.0˚ 36.0˚
34.0˚ 34.0˚
5 mm/yr 5 mm/yr
mm/yr mm/yr
0 1 2 3 4 5 6 0 1 2 3 4 5 6
dip−slip strike−slip
percentage
1.4
1.6
1.7 10 10
1.9 34.0˚
5 mm/yr
0 0
−2 −1 0 1 2 −2 −1 0 1 2
EW residual (mm/yr) NS residual (mm/yr)
38.0˚ 38.0˚
0.0˚ 2.0˚ 4.0˚ 6.0˚ 8.0˚ 0.0˚ 2.0˚ 4.0˚ 6.0˚ 8.0˚
36.0˚ 36.0˚
34.0˚ 34.0˚
5 mm/yr 5 mm/yr
mm/yr mm/yr
0 1 2 3 4 5 6 0 1 2 3 4 5 6
dip−slip strike−slip
38.0˚
0.0˚ 2.0˚ 4.0˚ 6.0˚ 8.0˚
R = 92.63%
1.3
1.3 mu = −0.0, std = 0.5 mm/yr mu = 0.0, std = 0.6 mm/yr
1.2 30 30
3.6
3.6
2.4 2.4 2.4
3.6 2.4
36.0˚
3.5
20 20
percentage
1.1
1.2
1.3 10 10
1.1 34.0˚
1.2
1.2 5 mm/yr
0 0
−2 −1 0 1 2 −2 −1 0 1 2
1.2 EW residual (mm/yr) NS residual (mm/yr)
1.3 1.2
strike-slip fault identified from regional seismotectonic studies and model variance reduction is 91 per cent and the F-test shows that the
in the 2-D models described above. Velocity residuals decrease χ 2 decrease from model 1 is significant well above the 99 per cent
significantly in eastern Algeria compared to model 1, where they confidence level (Table 2). In this model, the EU–NU oblique
are now <1.5 mm yr−1 at all sites, except at CNGR and CNAJ. The convergence is accommodated by 3.9 mm yr−1 of oblique reverse
582 A. Bougrine, A. K. Yelles-Chaouche and E. Calais
38.0˚ 38.0˚
0.0˚ 2.0˚ 4.0˚ 6.0˚ 8.0˚ 0.0˚ 2.0˚ 4.0˚ 6.0˚ 8.0˚
36.0˚ 36.0˚
34.0˚ 34.0˚
5 mm/yr 5 mm/yr
mm/yr mm/yr
0 1 2 3 4 5 6 0 1 2 3 4 5 6
dip−slip strike−slip
percentage
2.0
2.0 1.6
2.0
1.8
1.9 10 10
2.0 34.0˚
5 mm/yr
0 0
−2 −1 0 1 2 −2 −1 0 1 2
EW residual (mm/yr) NS residual (mm/yr)
38.0˚ 38.0˚
0.0˚ 2.0˚ 4.0˚ 6.0˚ 8.0˚ 0.0˚ 2.0˚ 4.0˚ 6.0˚ 8.0˚
36.0˚ 36.0˚
34.0˚ 34.0˚
5 mm/yr 5 mm/yr
mm/yr mm/yr
0 1 2 3 4 5 6 0 1 2 3 4 5 6
dip−slip strike−slip
38.0˚
0.0˚ 2.0˚ 4.0˚ 6.0˚ 8.0˚
R = 95.28%
1.5
1.5 mu = 0.0, std = 0.4 mm/yr mu = −0.0, std = 0.4 mm/yr
1.5 30 30
2.8
2.8
2.5 2.5 2.5
2.8 2.0 2.5
36.0˚
2.7 1.9
20 20
percentage
1.9
1.8 1.7
1.8
1.9
2.0 10 10
2.1 34.0˚
0.3
0.3
0.3 5 mm/yr
0 0
−2 −1 0 1 2 −2 −1 0 1 2
0.4 EW residual (mm/yr) NS residual (mm/yr)
0.4 0.4
faulting on the continental margin fault offshore western Algeria, We note in model 2 that some of the sites south of the GNC
and partitioned between (1) ∼2.7 mm yr−1 of convergence across fault (CSVB, CABS, CBBR, CBCK, CNGR, CNAJ and CFKZ)
its continuation offshore eastern Algeria and (2) ∼2.7 mm yr−1 of and north of it (CMAR, CJIJ) still show systematic south-directed
right-lateral strike-slip motion on the GNC fault. residuals of 0.5–1.5 mm yr−1 . We therefore build model 3 (Fig. 11)
Active deformation in Algeria from GPS 583
with an additional ‘Gafsa’ block bounded to the north by the GNC which has been the locus of some historical earthquakes (Harbi et al.
fault and to the south and west by a north-dipping fault, which 2010). We find a smaller variance reduction (93 per cent) compared
we call the Gafsa fault, meant to represent the active folding and to model 3 with an increase in χ 2 , even though a block was added to
faulting structures identified along the Aures Mountains and the the model. The existence of a West Atlas block in the configuration
Saharan Atlas (Meghraoui & Pondrelli 2012; Rabaute & Chamot- of model 4 is therefore not warranted by the data.
Rooke 2014). The model variance reduction is 94 per cent and the Starting from model 3 again, we test whether adding a West Tell
F-test shows that the χ 2 decrease from model 2 is significant well Atlas block would improve the model fit (Fig. 13). This block en-
above the 99 per cent confidence level. We observe that this model compasses the restraining bend described in Meghraoui & Pondrelli
results in smaller residual velocities both within the Gafsa and (2012), a region of NE–SW–trending faults and folds described for
East Tell Atlas blocks. Block motions across the Gafsa fault zone instance in Meghraoui et al. (1986); Bouhadad (2001). It is bounded
shows 1–2 mm yr−1 of shortening, increasing southward. This is to the south by a fault that may represent the deep structure above
consistent with geological observations of active folding and reverse which active faulting and folding localized in the western Tell Atlas.
faulting along the southern edge of the Aures mountains (Rabaute Model variance reduction is 95 per cent, close to that of model 3.
& Chamot-Rooke 2014). Block motions across the GNC fault and The significance of the χ 2 decrease from model 3 is slightly larger
across the offshore reverse fault in the western part of the study area than 99 per cent. The eastern part of the study area is unchanged
are similar to model 2. Compared to that model, the convergence compared to model 3. In the west, NU–EU convergence is now par-
rate across the offshore reverse fault in the eastern part of the study titioned between ∼3 mm yr−1 of pure shortening across the offshore
area decrease slightly to ∼1.5 mm yr−1 . fault system and ∼2 mm yr−1 of oblique shortening across the Tell
Although model 3 already provides an excellent fit to the ob- fault system. This latter number is consistent with the ∼2 mm yr−1
servations, we now test in model 4 whether adding a West Atlas shortening proposed by Meghraoui et al. (1986) across the Cheliff
block would improve the fit to observations (Fig. 12). This block is basin.
bounded to the south by the Southern Atlas fault system (Fig. 1),
Table 2. Statistics of model comparisons showing χ 2 values for each model and the F-ratio tests for selected model pairs A–B. DOF: degrees of freedom,
P-value: probability that the null hypothesis—the additional blocks in model B are not warranted by the data can be rejected. 1 − P: probability that additional
blocks are warranted by the data.
A B N-plates χ 2 -A χ 2 -B DOF-A DOF-B F-ratio P-value 1−P
0 2 9390.780 196
1 2 5175.504 196
1 2 3 5175.504 2824.824 196 187 51.87 1.5648 × 10−24 >> 99.99 per cent
2 3 4 2824.824 1564.360 187 184 49.42 1.4726 × 10−23 >> 99.99 per cent
3 4 5 1564.360 2401.457 184 181
3 5 5 1564.360 1286.515 184 181 13.03 1.1562 × 10−07 >> 99.99 per cent
5 6 6 1286.515 1268.319 181 178 00.85 5.2914 × 10−01 47.09 per cent
584 A. Bougrine, A. K. Yelles-Chaouche and E. Calais
Finally, we test a model that includes the four plate boundary strike-slip fault and the Gafsa fault system at the boundary between
blocks described above (West Coast, West Atlas, East Tell Atlas, the Aures Mountains and the Saharan Atlas (Fig. 11). In terms of
Gafsa; Fig. 14). As expected, this model has the smallest χ 2 and blocks, this adds to the kinematic description of eastern Algeria
highest variance reduction (95 per cent) of all models. The signif- (1) a coastal sliver between the GNC and offshore faults and (2)
icance of the χ 2 decrease from the previous model is 47 per cent, a very low strain rate region extending from the GNC fault to the
indicating that the existence of a West Atlas block distinct from the Gafsa fault to the south. Such a model requires convergence and
Nubian Plate is not required by the current data at the 99 per cent right-lateral slip across the Gafsa fault, consistent with the active
confidence level. oblique ramp structure described in the southern Tunisian Atlas by
Saı̈d et al. (2011). Although the kinematic models described here
use a single, distinct, fault, the relative motion required by the GPS
7 DISCUSSION data between the Gafsa and the West Atlas blocks are likely more
diffuse in the actual geology, as shown by Saı̈d et al. (2011) and
7.1 NU-EU kinematics Rabaute & Chamot-Rooke (2014).
1.5 mm yr−1 in eastern Algeria where the remainder of the NU–EU latter event is also the oldest reported historical earthquake in Alge-
convergence is taken up further south and inland, between the Aures ria (Yelles-Chaouche et al. 2006; Harbi et al. 2010). Block models
Mountains and the Saharan platform. predict a combination of thrust and strike slip movement on a north-
This result is consistent with a series of observations from bathy- dipping oblique fault, consistent with focal mechanisms available in
metric and seismic reflection data that mapped a succession of this area (Fig. 1). Why the NU–EU oblique convergence is currently
reverse fault segments oriented NE–SW to E–W associated with re- partitioned in eastern Algeria may have to do with the higher obliq-
cent compressive deformation along the toe of the Algerian margin uity of the plate margin direction to the NU–EU relative motion
over a width of ∼20 km (Deverchère 2005; Domzig et al. 2006; in the east than in the west (Fig. 15)—this question is a topic for
Kherroubi et al. 2009; Yelles et al. 2009; Aidi et al. 2018). These further work.
faults have been the locus of the recent Chenoua (1989), Ain Benian
(1996), Boumerdes (2003) and Algiers (2014) earthquakes (Yelles-
Chaouche et al. 2017). They may also have been responsible for 7.5 Implications for seismic hazard
the tsunamigenic historical earthquakes of 1716 (Algiers) and 1773 Several seismic hazard maps include the territory of Algeria from
with block models, GPS velocities confirm elastic strain accumula- Bounif, A., Haessier, H. & Meghraoui, M., 1987. The Constantine (northeast
tion on an offshore reverse fault system, consistent with geological Algeria) earthquake of October 27, 1985: surface ruptures and aftershock
data showing underthrusting of the oceanic lithosphere of the Alge- study, Earth planet. Sci. Lett., 85(4), 451–460.
rian basin under the North African margin. The data do not require Buforn, E., Bezzeghoud, M., Ud as, A. & Pro, C., 2004. Seismic Sources on
the Iberia-African Plate boundary and their tectonic implications, Pure
the presence of a discrete fault zone throughout the western Tell At-
appl. Geophys., 161(3), 623–646.
las, where the 1980 El Asnam earthquake occurred, but allow for it.
Calais, E., DeMets, C. & Nocquet, J.M., 2003. Evidence for a post–3.16 Ma
In eastern Algeria, deformation is more broadly distributed. There, change in Nubia–Eurasia–North America plate motions?, Earth planet.
the data require slip on the GNC strike-slip fault and on the Gafsa Sci. Lett., 216, 81–92.
fault system, with shortening between the Aures and the Saharan Calais, E., Fleitout, L. & Lambeck, K., 2018. Eurasia-North America plate
platform. motion and Glacial isostatic adjustment, Geophys. Res. Abstracts, EGU
Although the block models and the fault slip rates derived here Gen. Assembly, 20, 10999.
from GPS velocities are still to be refined, they provide new guide- Coiffait, B., Guellal, S. & Vila, J., 1992. Carte géologique au 1/50.000 d El
lines for seismic hazard assessment in Algeria. Denser GPS mea- Aria, Sonatrach. Division Hydrocarbures.
DeMets, C., Iaffaldano, G. & Merkouriev, S., 2015. High–resolution Neo-
evolution of the Central Algerian margin from combined wide-angle and Savage, J.C., 1983. A dislocation model of strain accumulation and re-
reflection seismic data off Tipaza, Algeria, J. geophys. Res.: Solid Earth, lease at a subduction zone, J. geophys. Res.: Solid Earth, 88(B6),
118(8), 3899–3916. 4984–4996.
Maouche, S., Morhange, C. & Meghraoui, M., 2009. Large boulder accu- Schmid, R., Steigenberger, P., Gendt, G., Ge, M. & Rothacher, M., 2007.
mulation on the Algerian coast evidence tsunami events in the western Generation of a consistent absolute phase-center correction model for
Mediterranean, Mar. Geol., 262(1-4), 96–104. GPS receiver and satellite antennas, J. Geod., 81(12), 781–798.
Maouche, S., Abtout, A., Merabet, N.-E., Aı̈fa, T., Lamali, A., Bouyahiaoui, Segall, P., 2010. Earthquake and Volcano Deformation, Princeton University
B., Bougchiche, S. & Ayache, M., 2013. Tectonic and hydrothermal ac- Press.
tivities in Debagh, Guelma Basin (Algeria), J. Geol. Res., 2013(3-4), Serpelloni, E., Vannucci, G., Pondrelli, S., Argnani, A., Casula, G., Anzidei,
1–13. M., Baldi, P. & Gasperini, P., 2007. Kinematics of the Western Africa-
Mauffret, A., 2007. The Northwestern (Maghreb) boundary of the Nubia Eurasia plate boundary from focal mechanisms and GPS data, Geophys.
(Africa) Plate, Tectonophysics, 429(1-2), 21–44. J. Int., 169(3), 1180–1200.
Meade, B.J. & Loveless, J.P., 2009. Block modeling with connected fault- Stein, S. & Gordon, R.G., 1984. Statistical tests of additional plate bound-
network geometries and a linear elastic coupling estimator in spherical aries from plate motion inversions, Earth planet. Sci. Lett., 69(2), 401–
coordinates, Bull. seism. Soc. Am., 99(6), 3124–3139. 412.
are standard deviation of horizontal velocity in mm yr−1 . σ Ven are Table S2. χ 2 and variance reduction, R, computed for each model
correlation between Ve and Vn. Sites with a star are those used to in the three frames: Nubia, Eurasia and ITRF.
define the Eurasia and Nubia reference frames. Sites in bold font
Please note: Oxford University Press is not responsible for the con-
are those whose position time-series show (1) non-linear behaviour
tent or functionality of any supporting materials supplied by the
unlikely tectonic in origin and (2) velocities that are different from
authors. Any queries (other than missing material) should be di-
their nearest neighbors. These sites were excluded from the inter-
rected to the corresponding author for the paper.
pretation.