Professional Documents
Culture Documents
DOI 10.1007/s12517-011-0509-y
ORIGINAL PAPER
Introduction
The 21 May 2003 Boumerdes (NE of Algeria) earthquake
(Mw 06.8) was one of the strongest destructive event that
occurred in the last decade in the Maghreb region (Northern
Africa). This last is known as a tectonic active zone where the
Eurasian plate is subducting the African plate with a velocity
of about 6 mm per year. That tectonic activity induces a
persistent compressional state not only along the continental
Mediterranean coast but also offshore (Benouar and Laradi
1996; Bouhadad et al. 1994).
The 21 May 2003 earthquake strongly struck the quite
balneal city of Boumerdes and its surrounding places like
the capital Algiers and Zemmouri city, located respectively
at 50 km NW and 15 km NE. The main shock (Mw 06.8)
which lasted between 30 and 40 s was followed by devastating
effects in terms of victims, injured people and collapsed infrastructures (Edward Curtis 2004).
The epicentre location was calculated at (latitude, 36.90 N
and longitude 3.71 E), at10 km approximately in offshore of
Zemmouri city (Fig. 1). The focal mechanism study exhibited
a trust-faulting process (strike 54, dip 47, rake 76). It
should be specified that neither geometry of this fault nor its
precise localisation were unsuspected and therefore not studied before the 21 May 2003 earthquake.
It was known that a more or less long aftershock sequence
follows the major event of earthquakes. That is observed
particularly for the shallow earthquakes like the Boumerdes
21 May 2003 (10 km deep). Aftershock series are due to the
upheaval of the local tectonic state induced by the stress
change of the main shock. The fundamental interest that a
Arab J Geosci
Fig. 1 Overview of the May 21, 2003 earthquake: a seismological retrospective (16751999) around the current epicentral area (from EERI
modified), b main features of the Boumerdes earthquake; b main features of the May 21, 2003, earthquake
recorded aftershock sequence offers for analysis is its combined temporal and spatial distribution. Then one can use the
one-dimensional or the multivariate statistical methods to
explore and to interpret the data displayed in the recorded
aftershock sequence (Bounif et al. 2004). Because of their
privileged temporal localization, i.e., after the main shock and
before the next earthquake, the aftershocks may be considered
as foreshocks for the seismological prevision approach.
Arab J Geosci
(7 min after the major event). For many people, this first
strong aftershock was felt like a second earthquake because of
additional hazard caused to population. Indeed, severe structural damage had accompanied the occurrence of this event.
For all the studied sequence, the aftershocks are shallow
(10 km). A summary statistical review is given in Figure 2.
Outliers detection in the data set
We have mentioned above that the psychological and material impact of the first strongest aftershock (Ms 05.7) which
occurred 7 min after the main shock (Mw 06.8). That aftershock magnitude may be considered relatively close to the
main shock knowing that the region of Boumerdes belongs,
until the event of 21 May 2003, to a seismically quite region
in north of Algeria. The first idea to justify this relatively
high value of the aftershock consists in supposing another
source for this last. Indeed, jointly with the offshore fault
responsible for the main shock (the Zemmouri fault), there
exists an older and better studied fault located at the NE near
the epicentral area, the WNW-ESE Thenia fault. This fault
(Fig. 3) is considered as one of the major faults close to the
city of Boumerdes liable to generate a strong earthquake.
The seismogeological interpretations made later included
this strong aftershock to the normal seismic activity which
occurs after the main event. Analysing the values included
in the aftershock sequence, the major magnitude of the main
event (Mw 06.8) may be described as natural failed value.
The asked question consists to know if the strongest aftershock recorded magnitude (M05.7) is another failed value
which could suppose the occurrence of another earthquake.
The calculated quartiles of the aftershock sequence are as
follow: the first quartile Q1 03.6, the second quartile (median) Q2 04.2 and the third quartile Q3 04.6. The third quartile
Q3 indicates that 75% of the aftershock magnitudes are
located under the magnitude 4.6. The inter-quartile IQ is
defined as: IQ0Q3 Q1 01. Figure 4 shows the spatial distribution of the aftershock sequences after distribution of
magnitudes in subclasses.
The presumed high failed values included in the aftershock sequence can be detected by using the Tukey criterion
which stipulates that such value is identified as failed in
given data must firstly belonging to this data and secondly
be greater than the result Q3 +1.5IQ. Applying this criterion
we obtain here the imaginary magnitude M06.1 which does
not belong to the aftershock sequence. Then only the main
shock magnitude Mw 06.8 may be characterized as failed in
the whole of the recordings.
Arab J Geosci
nt t p
dn
K
lt
dt
c t p
Arab J Geosci
Arab J Geosci
b M
Here b
a and b are constants for a region and N
defines the cumulative number of events with magnitude
greater or equal to a reference magnitude M. This number
is currently replaced simply by N as the number of
earthquakes taken in a specified time interval. By putting
a log10b
a, the relationship 4 takes the most used form
log10 N m M a bm
We specified in our previous analysis the importance attached to b value of the GutenbergRichter law and to the
Omoris temporal parameter p. More recently, another descriptive factor of seismological phenomenon has been introduced: the fractal dimension D (Scholz and Mandelbrot
1989; Palmer 1988).
Formally, there is no direct relationship between the b value
and the space fractal dimension because these two parameters
are influenced by identical seismological characteristics.
However, practical studies on the spatial variations of these
two parameters, showed a narrow interrelationship between
these obvious three quantities mainly between b value and
fractal dimension D.
Moreover, the empirical GutenbergRichter needs a better justification for becoming a mirror of the seismic context
of the considered zone and the fractal dimension can be an
active parameter here.
Contribution of the fractal dimension in seismological
studies
It is now accepted that the aftershocks sequence grows in
a fractal environment with non-integer dimension D (Barton
and La Pointe 1995).
Arab J Geosci
The fractal dimension on the spatial distribution of epicentres or hypocentres may be interpreted as an indicator of
geological heterogeneities within a fracturing material. Thus,
the fractal dimension provides information on the seismic
stability of a region, and its variations are interpreted as a
dynamic evolution of tectonic states (Smalley et al. 1987). It
was also noticed that the fractal dimension is scale invariance
which allows efficient statistical analysis of seismological
parameters to quantify the dimension of the seismic distribution
like the level of clusterization. This dimension characterizes the
degree of filling up of the considered space environment by
fractal structures. Knowledge of the fractal dimension becomes
a predictor of fracture process. For example a fractal dimension
close to 2 implies the presence of a plan being filled up by
fractal structures and the events can be distributed randomly or
in an organized basis in a two-dimensional space.
As mentioned above the spatial fractal dimension D of
earthquakes is correlated with the slope b of the Gutenberg
Richter law, regardless of the size of the earthquake. The
formula most commonly used is the Akis formula which
establishes a possible relationship between b and D as follows
D 3b=c
The variogram analysis describes the continuity or the appearance of roughness of the data. It begins by calculating
an experimental variogram followed by a real model fitted
to the data.
The experimental variogram is calculated by averaging
one-half of the squared differences of the values over all pairs
of observations of a separation distance and direction data
(Houlding 2000; Webster and Oliver 2001). Graphically, the
variogram is presented in the form of a continuous curve. It
used to highlight an eventual clustering of data or a random
distribution of values of the studied variable (Guo and Ogain
1997).
Three main functions are commonly used in geostatistics to
describe the spatial correlation of observations (Fig. 14): the
correlogram, the covariance and the semi-variogram also called
Arab J Geosci
Nh
X
1
Zi Z i h2
2N u; h i1
1
2N h
N h
X
Zi Z i h2
i1
Components of a variogram
The variogram is a graph which describes the variations of
the semi-variance versus separation distance. When the autocorrelation is included in the analysis, the curve of semivariance is located immediately below for the short distances of separation therefore a higher autocorrelation. The
semi-variance curve is characterized by three typical elements: a range, a nugget variance and a sill (Cressi 1991).
The nugget variance C0 is the intercept of the variogram
corresponding to a zero separation distance. The variogram
vs. the separation distance h is an increasing curve which
reaches its maximum at a specified value C. This value is
achieved for a value of the lag h0a called range. Beyond the
value of the range, the variogram follows a horizontal
asymptotical behaviour called sill (S). The total sill of the
variogram is C+C0 (Fig. 16).
Near the origin the parameter C0 estimates the continuity
of the phenomenon while the sill (S) characterizes the randomness of data so that the range a and the parameter C
characterizes the structural aspect of spatial distribution of
data. The level of irregularity of the data is highlighted by
the slope of the tangent line to the experimental variogram
including two or three points closest to the origin. Two
extremes cases can occur: for a near-vertical tangent the
data is assumed very irregular and at contrary for nearhorizontal tangent, the data has a greater regularity.
Arab J Geosci
Sill C0 +C00.27420
Proportion C0CC 0:98
Range a00.1230
10
where
h
h 3 i
ah 0:2706 0:1845h 0:5 0:1230
0:0036
The numerical value C0 +C00.27420 is simply the sample
variance. The analytical method of calculating the same
Fig. 17 Theoretical and experimental spherical variogram of the aftershock sequences of the Boumerdes earthquake
Arab J Geosci
a
2
11
References
Barton CC, La Pointe PR (1995) Fractal in the earth sciences. Plenum
Press, New York
Benouar D, Laradi N (1996) A reappraisal of the seismicity of the
Maghreb countriesAlgeria-Morocco-Tunisia. Journal of Natural Hazard Vol. 13 No. 2
Bouhadad Y et al. (1994) The Boumerdes (Algeria) earthquake of May
21, 2003 (Mw 06.8). Ground deformation and intensity. Journal of
Seismology vol.8 No. 4
Bounif A et al. (2004) Zemmouri (Algeria) Earthquake Mw6.8
relocation and aftershock sequence analysis. Geophys. Res.
Lett. 31
Cressi N (1991) Spatial statistics. Wiley, New York
Edward Curtis L (2004) Zemmouri Algeria Mw 06.8 earthquake of
May 21, 2003. ASCE Publications
Guo Z, Ogain O (1997) Statistical relations between the parameters of
aftershock in time, space and magnitude
Gutenberg B, Richter CF (1942) Earthquake magnitude intensity,
energy and acceleration. Bull Seismol Soc Am 32
Houlding SW (2000) Practical geostatistics modelling and spatial
analysis. Springer, New York, p 160
Kagan YY (1996) Comment on the Gutenberg law or characteristic
earthquake distribution, which is it? Wesnousky SG. Bull Seismol
Soc Amer. 86
Kagan YY (1999) The universality of the frequency magnitude relationship. J Pure Appl Geophys 155
Nanjo K, Nayahama H, Satomoura M (1998) Rates of decay and the
fractal structure of active fault systems. Tectonophysics 287
Palmer MW (1988) Fractal geometry: a tool for describing patterns of
plan communities. Vegetation Revue 102
Scholz CH, Mandelbrot B (1989) Book: Fractal in geophysics. Edited
by Basel Boston Birkhauser
Smalley RF et al. (1987) A fractal approach to the clustering of
earthquake. Bull Seismol Soc Amer 77
Utsu T, Ogata Y, Matsura S (1989) The centenary of the Omori
formula for decay law of aftershock activity. Journal of
Physics Earth 43
Webster R, Oliver MA (2001) Geostatistics for environmental scientists. Wiley, Chichester, p 271