African Journal of Environmental Science and Technology Vol. 4(8), pp.

526-534, August 2010

Available online at http://www.academicjournals.org/AJEST
ISSN 1991-637X © 2010 Academic Journals

Full Length Research Paper

Multivariate statistical characterization of groundwater

quality in Ain Azel plain, Algeria
Lazhar Belkhiri1*, Abdurrahman Boudoukha2, Lotfi Mouni3 and Toufik Baouz4
1
Department of hydraulics, University of Hadj Lakhdar Batna, Batna 05000, Algeria.
2
Laboratoire de recherche en hydraulique appliquée Université de Hadj Lakhdar Batna, Batna 05000, Algeria.
3
Laboratoire de technologie des matériaux et de génie des procédés de l'université de Bejaia,
Targa-Ouzemour 06000, Alegria.
4
Laboratory of Organic Materials, University of Bejaia, Targa- Ouzemour 06000, Algeria.
Accepted 25 May, 2010

Multivariate statistical techniques, cluster and principal component analysis were applied to the data on
groundwater quality of Ain Azel plain (Algeria), to extract principal factors corresponding to the
different sources of variation in the hydrochemistry, with the objective of defining the main controls on
the hydrochemistry at the plain scale. Q-mode hierarchical cluster analysis grouped 54 groundwater
samples into three clusters, that is, relatively less saline water (group 1), mixed water (group 2) and
blended water (group 3), based on the similarity of groundwater quality characteristics. Principal
component analysis, applied to the data sets of the three different groups obtained from hierarchical
cluster analysis, resulted in four, five and three latent factors explaining 83.21, 83.36 and 87.30% of the
total variance in groundwater quality data sets of group 1, group 2 and group 3, respectively. The
varifactors obtained from PCA indicate that the parameters responsible for groundwater quality
variations are mainly related to presence and dissolution of some carbonate, dolomitic and evaporite
minerals; natural processes and water-rock interaction in the three water types. The results of this
study clearly demonstrate the usefulness of multivariate statistical analysis in hydro chemical.

Key words: Cluster analysis, principal component analysis, hydrochemistry, Ain Azel plain, Algeria.

INTRODUCTION

In the current world economic paradigms, sustainable natural waters. Some minerals like carbonates and
socioeconomic development of every community evaporites dissolve quickly and change the composition
depends much on the sustainability of the available water of water faster, while other minerals like silicates dissolve
resources. Water of adequate quantity and quality is more slowly and have less conspicuous effects on the
required to meet growing household, industrial and composition of water. Temperature also plays a vital role
agricultural needs. Groundwater quality is a very in controlling the chemical and biological composition of a
sensitive issue, which transcends national boundaries. It freshwater body. Previous studies (Frape et al., 1984;
is influenced by many factors, including atmospheric Garrels and McKenzie, 1967; Hem, 1989; Hartman et al.,
chemistry, the underlying geology, the vegetation (or 2005) have revealed that the chemistry of natural waters
organic matter decay) and anthropogenic agents. The can often be traced to the reaction of these waters with
solubility of minerals in water places an upper limit on the sediments or rocks through which they flow. Based on
maximum amounts of certain species of chemicals in catchments studies in the USA, Walling (1980) observed
differences in the weathering mechanisms of different
rocks. Walling (1980) concluded that total dissolved
solids in the water from limestones, volcanics and sand
*Corresponding author. E-mail: [email protected]. and gravel is almost independent of the amount of runoff.
These differences in behaviour accrue from the
Abbreviations: CA, Cluster analysis; FA, factor analysis; differences in the solubilities of the minerals present in
EC, electrical conductivity; PCA, principal component these rocks (Yidana et al., 2008).
analysis. Multivariate statistical techniques, cluster analysis (CA)
 Belkhiri et al. 527

and factor analysis (FA), are effective means of fed by stream water coming from different relief’s
manipulating, interpreting and representing data surrounding the depression inter-mountainous of Ain
concerning groundwater pollutants and geochemistry. Azel. The plain hosts a large number of water-wells with
They are frequently employed to characterize the quality depths varying from 8 - 38 m. Most of these wells supply
of groundwater. Liu et al. (2003) adopted FA to evaluate water for drinking and irrigation. The direction of
groundwater quality in a black foot disease area, Taiwan. groundwater flow around Ain Azel plain is from south to
Two factors, seawater salinization and arsenic pollutant, north and from west to east. In general, the groundwater
were considered and their factor scores were mapped. flows toward center of the plain (Figure 2) and development
Love et al. (2004) applied FA to distinguish several this map shows a piezometric depression, which
groundwater signatures, including uncontaminated coincides with high density of the well in the center of the
groundwater, agricultural activities, mining activities and plain. The limits of south and west are of imposed entering
sewage pollution. Olmez et al. (1994); Mahknecht et al. flux; however the limit of east is of null flux. The pumping
(2003); Farnham et al. (2003) used FA to discuss tests on different wells showed high transmissivity (30 - 36
geochemical evolution and mineralization and m2/day) indicating high yields (Belkhiri, 2005).
groundwater contamination. Additionally, the CA was also
used to interpret the hydrochemical data based on factor
scores, for example, Suk and Lee (1999); Reghunath et MATERIALS AND METHODS
al. (2002); Kim et al. (2005). This study comprises
Sample collection and analysis
application of multivariate statistical techniques to
groundwater quality data set obtained from groundwater 18 wells currently in use were selected based on the preliminary
of Ain Azel plain in Algeria. field survey carried out to understand the overall distribution of the
various types of wells in the studied area (Figure 1). The selected
wells are used for domestic, agricultural, and domestic/agricultural
DESCRIPTION OF THE STUDIED AREA purposes and are uniformly distributed over the area of concern.
Groundwater samplings were performed three times in 2004: June,
September and December and a total of 54 groundwater samples
The studied area is located in the east of Algeria. Most of were collected during this period. The samples were collected after
inhabitants (more than 30000 inhabitants) are centered pumping for 10 min. This was done to remove groundwater stored
on the town of Ain Azel. The principal activity in this area in the well. These samples were collected using 4 - 1 acid-washed
is the production of cereals (barley and corn) and its polypropylene containers. Each sample was immediately filtered on
climate is semi-arid. The amount of rainfall varies from site through 0.45 µm filters on acetate cellulose. Filtrate for metals
analyses were transferred into 100-cm3 polyethylene bottles and
place to place and from season to season, ranging from immediately acidified to pH < 2 by the addition of MerckTM ultrapure
346 mm/yr in the plains (altitude < 1000 m) to 534 mm/yr nitric acid (5 ml 6 N HNO3). Samples for anions analyses were
in the mountains (altitude > 1000 m). The annual maximum collected into 250-cm3 polyethylene bottles without preservation. All
˚
and minimum temperatures are 21.8 and 8.5 C, the samples were stored in an ice chest at a temperature of < 4˚C
respectively (Belkhiri, 2005). and later, transferred to the laboratory and stored in a refrigerator at
a temperature of < 4˚C until analyzed (within 1 week). Immediately
According to many authors (Savornin, 1920; Galcon,
after sampling, pH and electrical conductivity (EC) were measured
1967; Guiraud, 1973; Vila, 1980) the area of concern is in the field using a multi-parameter WTW (P3 MultiLine pH/LF-
distinguished by two geological sets. In the South, SET).
autochthonous Jurassic and Cretaceous carbonate Subsequently, the samples were analyzed in the laboratory for
lithologies are mainly observed in Djebels Boutaleb, their chemical constituents such as calcium, magnesium, sodium,
Djebel Hadjar Labiod and Fourhal (Figure 1). In the potassium, chloride, bicarbonate, sulfate, lead and iron. This was
achieved using standard methods as suggested by the American
North, allochthonous Jurassic and Cretaceous lithologies Public Health Association (APHA, 1989; 1995). Ca, Mg, HCO3 and
are observed in Djebels Kalaoun and Sekrine. A Cl were analyzed by volumetric titrations. Concentrations of Ca and
lithostratigraphic study of Ain Azel area identifies the Mg were estimated titrimetrically using 0.05 N EDTA and 0.01 N
following two aquifer formations. The first is a carbonate and those of HCO3 and Cl by H2SO4 and AgNO3 titration,
and sandy formation of about 700 m thick presenting respectively. Concentrations of Na and K were measured using a
fracture porosity and constitutes the Barremian formation. flame photometer (Model: Systronics Flame Photometer 128) and
that of sulfate by turbidimetric method (Clesceri et al., 1998).
The second is an alluvial formation of about 250 m thick Standard solutions for the above analysis were prepared from the
corresponding to a Mio-Plio-Quaternary formation with respective salts of analytical reagent grades. Trace metals were
interstitial porosity (Boutaleb, 2001; Belkhiri, 2005). A determined by Graphite Furnace Atomic Absorption
mineralogical study shows that the metalliferous minerals, Spectrophotometer (Perkin-Elmer AAnalyst 700) using multi
particularly the sphalerite, the gangue minerals and the element Perkin-Elmer standard solutions. The accuracy of the
chemical analysis was verified by calculating ion-balance errors
dolomite contain variable quantities of trace metals such
where the errors were generally around 10%.
as Zn, Pb, Cd, Fe and Cu (Boutaleb, 2001).
The studied area is situated in the alluvial plain of the
Mio-Plio-Quaternary (Figure 1) showing a heterogeneous Multivariate statistical analysis
continental detrital sedimentation (Boutaleb, 2001;
Belkhiri, 2005; Attouche, 2006). This aquifer is directly In recent times, multivariate statistical methods have been applied
528 Afr. J. Environ. Sci. Technol.

o o o
20 0 20

MEDITERRAN
MEDITERRANEAN SEA NORTHERN EAN SEA
Spain Morocco Tunisia
Alger Annaba 30
o ATLANTIC
ATLANTIC OCEAN
Setif
OCEAN
Algeria Libya Egypt

RE
D
Mauritania

SE
Morocco Tunisia Mali

A
Niger
Chad
Sudan

ALGERIA
Libya
o
0
Study area

INDIAN

Mauritania OCEAN

SOUTHERN
ATLANTIC
OCEAN

Mali
Niger
o
30

o
5 30 '
292
DJ. Sekrine
DJ. Kalaoun
Mio-Plio-Quaternary (studied aquifer)
290
Miocene (congloerate, clay and marl)
1 18
6 Cenomanian and Turonian (limestone)

288 Kef el Djebas 2 17

Aptian (carbonate)
7 5
Aptian (marl)
14
Barremian (carbonate and sandstone)
286 3 16
Mine of Kherzet 4 Neocomian (sandstone and marl)
Youssef 15 12
13
Jurassic (carbonate)
284 11
iod

Ain Azel 10 Lias (marl)

8
ab

9
rL

Triassic (evaporite)
Annfif
ja

Oued
ad

282 Wells
.H

Mine of Chaabet
DJ

el Hamra
DJ. FOURHAL
o
280 DJ. BOUTALEB o
47 ' 35 47 '
744 746 748 750 752 754 756 758 760 762 764
o
5 30 '

Figure 1. Map showing water sampling locations and geology of the studied area.

widely to investigate environmental phenomena e.g., (Anazawa et and hydrogeochemical processes (Cameron, 1996; Duffy and
al., 2003; Anazawa and Ohmori, 2005; Güler and Thyne, 2004; Brandes, 2001; Gupta and Subramanian, 1998). Momen et al.
Laaksoharju et al., 1999). The combined use of principal (1996) used cluster analysis and PCA to identify the temporal and
component analysis (PCA) and cluster analysis enabled the spatial variation of water chemistry in Lake George in New York.
classification of water samples into distinct groups on the basis of Tariq et al. (2005) similarly applied multivariate techniques to trace
their hydrochemical characteristics. Multivariate statistical tools metal levels in tannery effluents in Pershawar in Pakistan. Their
have been successfully used to study and classify different study involved samples from tannery effluents, groundwater and
sediment types (Huisman and Kiden, 1998; Tebens et al., 2001), soils and with the aid of multivariate tools they were able to
 Belkhiri et al. 529

o
5 30 '
292
DJ. Sekrine
DJ. Kalaoun
290

288 Kef el Djebas

Flowpath

286
Mine of Kherzet

ath
Youssef

owp
284
d

l
Ain Azel

F
io
ab
rL

Annfif
ja

Oued
ad

282
.H

Mine of Chaabet
DJ

el Hamra
DJ. FOURHAL
o
280 DJ. BOUTALEB o
35 47 ' 35 47 '
744 746 748 750 752 754 756 758 760 762 764
o
5 30 '

Wells Flow direction Water table contour (m)

0
90
Figure 2. Potentiometer surface map in December, 2004 (unit: m).

correlate important chemical species of the three media and by using the software STATISTICA®.
established significant relationships. Using factor analysis, Zeng
and Rasmussen (2005) attributed the variations in quality of water
from Lake Lanier in Georgia to anoxia associated with lake
Factor analysis
stratification. This paper dwells on the strength of multivariate
techniques to characterize the hydrochemical variations along the
Ain Azel plain. It employs the combined use of cluster and factor Factor analysis is a multivariate analytical technique, which derives
analysis to assess the spatial and temporal variations of a subset of uncorrelated variables called factors that explain the
groundwater chemistry. variance observed in the original dataset (Anazawa and Ohmori,
2005; Brown, 1998). Factor analysis is used to uncover the latent
structure of a set of variables. In technical terms, common factor
Cluster analysis analysis represents the common variance of variables, excluding
unique variance and is thus a correlation-focused approach seeking
Cluster analysis was used to determine if the samples can be to reproduce the intercorrelation among the variables. On the other
grouped into statistically distinct hydrochemical groups that may be hand, components (from PCA) reflect both common and unique
significant in the geologic context. A number of studies used this variance of the variables and may be seen as a variance-focused
technique to successfully classify water samples (Alther, 1979; approach that reproduces both the total variable variance with all
Williams, 1982; Farnham et al., 2000; Alberto et al., 2001; Meng components as well as the correlations. PCA is far more commonly
and Maynard, 2001). Comparisons based on multiple parameters used than principal factor analysis (PFA). However, it is common to
from different samples were made and the samples were grouped use ‘‘factors’’ interchangeably with ‘‘components’’ in multivariate
according to their ‘similarity’ to each other. Classifications of analysis. Factor analysis can be performed on any kind of scientific
samples according to their parameters are known as Q-mode data to establish a pattern of variation among variables or reduce
classifications. In the present study Q-mode HCA was used to large data sets into factors for easy handling and interpretation.
classify the samples into distinct hydrochemical groups. The Ward’s The total number of factors generated from a typical factor
linkage method (Ward, 1963) was used in this analysis. A classification analysis indicates the total number of possible sources of variation
scheme using Euclidean distance (straight line distance between in the data. Factors are ranked in order of merit. The first factor or
two points in c-dimensional space defined by c variables) for component has the highest eigenvector sum and represents the
similarity measurement, together with Ward’s method for linkage, most important source of variation in the data. The last factor is the
produces the most distinctive groups where each member within least important process contributing to the chemical variation.
the group is more similar to its fellow members than to any member Factor loadings on the factor loadings tables are interpreted as
outside the group (Güler et al., 2002). All 11 hydrochemical variables correlation coefficients between the variables and the factors. In
measured (consisting of EC, pH, Ca, Mg, Na, K, Cl, SO4, HCO3, Pb this research, PCA was applied to chemical data from the Ain Azel
and Fe) were utilized in this analysis. For statistical analysis, all the plain to extract the principal factors corresponding to the different
variables were log-transformed and more closely correspond to sources of variation in the data. Here, PCA was selected for the
normally distributed data. Subsequently, they were standardized to reasons stated above. In order to maximize the variation among the
their standard scores (z-scores) as described by Güler et al. (2002). variables under each factor, the factor axes were subsequently
Hydrochemical results of all samples were statistically analyzed varimax rotated.
 530 Afr. J. Environ. Sci. Technol.

Table 1. Statistical summary of hydrochemical parameters of groundwater.

EC pH Ca Mg Na K Cl SO4 HCO3 Pb Fe
Min 545 6.7 64.13 35.88 18.5 6.14 56.80 14 131.76 0.03 0.023
June (n = 18) Max 1783 7.9 232.46 181.20 143.7 18.02 337.25 309 1348.10 1.83 0.338
Mean 1012 7.1 134.13 84.20 61.7 12.92 166.66 114 392.70 0.61 0.149
Std.Dev. 374 0.3 45.76 41.54 37.9 3.20 94.82 82 256.52 0.53 0.082

Min 470 6.8 64.12 35.88 16.7 5.14 56.80 18 176.90 0.29 0.067
Sept (n = 18) Max 2203 7.6 256.51 184.20 145.3 19.01 330.15 312 1342.00 1.79 0.789
Mean 1092 7.2 132.84 87.42 61.3 12.11 160.18 118 421.58 0.91 0.410
Std.Dev. 533 0.2 46.67 46.42 38.2 3.49 87.58 82 240.37 0.45 0.211

Min 629 6.8 62.52 35.45 16.3 4.20 63.90 37 183.00 0.02 0.089
Dce (n = 18) Max 2578 7.9 250.09 192.21 145.4 20.20 337.25 368 1335.90 0.29 0.645
Mean 1304 7.1 128.49 83.96 60.3 12.04 170.99 134 433.32 0.09 0.325
Std.Dev. 604 0.3 45.53 39.96 38.3 3.47 94.91 90 242.17 0.07 0.133

Min 470 6.7 62.52 35.45 16.3 4.20 56.80 14 131.76 0.02 0.023
Total (n = 54) Max 2578 7.9 256.51 192.21 145.4 20.20 337.25 368 1348.10 1.83 0.789
Mean 1136 7.1 131.82 85.19 61.1 12.36 165.94 122 415.87 0.54 0.295
Std.Dev. 518 0.3 45.18 41.94 37.4 3.35 90.85 84 242.38 0.53 0.185
All values are in mg/l except pH and EC (µ.Siemens/cm).

RESULTS AND DISCUSSION during the three campaigns. For this reason the three
campaigns were grouped as one.
Hydrochemical characteristics

Statistics of the chemical compositions of the groundwater Cluster analysis

samples are shown in Table 1. The groundwater samples
of the study area have pH values ranging from 6.7 - 7.9, Cluster analysis suggests three groups of groundwater’s
which indicate that the groundwater is slightly alkaline. The (Figure 3 and Table 2). EC seems to be a major
electrical conductivity (EC) values ranged 470 - 2,578 distinguishing factor, which increases with concentrations
µS/cm. The order of abundance of the major cations is increasing in all major-ions following the order: Group 1,
Ca Mg > Na > K and all samples exceeded the desirable 2 and 3 (Table 2). The Group 1 is composed of the wells
limit of Ca for drinking water (75 mg/l) except sample 10, 3, 4, 5, 6, 7, 10, 12 and 13, and concerns 44% of the
but only 78% of them exceed that of Mg (50 mg/l). The water samples. This type of water is relatively fresh with a
abundance of the major anions is HCO3 Cl > SO4 and mean EC of 782 µS/cm, which is the characteristic of less
almost 28% of the samples exceeded the desirable limit of saline water (Ca-Mg-HCO3). This group is basically
Cl (200 mg/l), but the sulfate concentrations are all below bicarbonate dominated, however; calcium and magnesium
health guidelines (WHO, 1993). The concentration of lead are also present.
for the three campaigns ranges from 0.02 - 1.83 mg/l and Group 2 is represented by the wells 1, 8, 11, 15, 16 and
all samples exceeded the WHO guideline limit of 0.01 17, and it occupies 24% of the water samples. The
mg/l. In the case of iron, the concentration in many of the electrical conductivity for this group is 1.062 µS/cm,
samples was higher than the WHO permitted limit of 0.3 which is the characteristic of mixed water (Mg-Ca-HCO3-
mg/l and the percent samples above the WHO limit Cl). This water type 2 is bicarbonate and chloride dominated
ranges from 60 - 83% for the three campaigns. Since in and it also has low concentrations of sulfate. This water
this region, except for agricultural activity no other type can be interpreted as the first step of water-rock
anthropogenic activities were being carried out in this interactions occurring in dilute solutions, explained by the
region and hence the high Fe concentration in these abundance of carbonate and evaporite formations in the
waters could not be assigned to the anthropogenic aquifer.
activities but rather the water-rock interaction should be Group 3 includes samples: 2, 9, 14 and 18, where the
the reason for the higher value of Fe observed in the EC is 1.956 µS/cm which is the characteristic of blended
plain. It is clearly observed that the concentrations of the water (Mg-Ca-Cl-HCO3). Chloride content is also high with
major elements do not undergo an appreciable change respect to bicarbonate concentration. Based ongeological
 Belkhiri et al. 531

Figure 3. Dendrogram of the hydrochemical samples.

Table 2. Mean parameter values of the three principal water groups (determined from HCA).

Group EC pH Ca Mg Na K Cl SO4 HCO3 Pb Fe

G1(n = 24) 782 7.0 108.93 54.98 26.3 10.81 82.83 98 352.03 0.45 0.319
G2 (n = 18) 1062 7.1 120.08 89.44 78.0 12.37 198.05 127 342.62 0.54 0.336
G3(n = 12) 1956 7.3 195.22 139.25 105.4 15.43 284.01 164 653.41 0.70 0.185
pH (standard units), EC (µ.Siemens/cm) and mean (mg/l).

and geographical position, discharges of industrial waste- elevation.

water and contamination of the groundwater by seawater
were excluded as the source of high concentrations. A
local contamination by septic tanks, sewage systems, Factor analysis
agricultural fertilizers and dissolution of the formations
clay, marl and salt and could be responsible for this The Kaiser criterion (Kaiser, 1960) was applied to
532 Afr. J. Environ. Sci. Technol.

Table 3. Rotation PCA loading matrix

Group 1 2 3
Component 1 2 3 4 1 2 3 4 5 1 2 3
EC 0.92 0.22 -0.18 0.06 0.20 -0.09 -0.02 0.92 -0.14 -0.13 0.69 0.62
pH -0.16 -0.28 0.37 0.78 0.80 0.21 -0.22 0.14 0.07 -0.96 -0.07 -0.07
Ca 0.81 0.05 0.02 -0.27 -0.86 0.25 0.01 0.11 0.21 0.44 0.84 -0.15
Mg 0.81 0.26 0.12 -0.28 0.13 0.87 0.20 0.05 0.03 0.68 0.62 0.10
Na 0.34 -0.82 0.06 0.26 -0.28 0.37 -0.59 0.37 0.43 0.93 0.29 -0.07
K 0.23 -0.26 -0.46 0.63 0.05 -0.14 0.07 -0.14 0.93 0.24 -0.93 -0.04
Cl 0.88 -0.38 -0.07 0.06 0.78 0.27 0.07 0.36 0.25 0.93 0.18 -0.20
SO4 -0.47 0.06 -0.15 0.77 0.05 0.14 0.94 -0.04 0.07 0.91 -0.35 -0.06
HCO3 0.40 0.78 0.00 -0.36 -0.40 0.07 0.84 0.09 0.06 -0.91 0.05 0.32
Pb -0.01 0.09 0.89 0.00 0.35 0.74 -0.02 -0.35 0.05 -0.66 0.05 -0.23
Fe 0.23 0.78 0.36 0.13 -0.16 0.72 -0.04 0.05 -0.16 0.00 -0.02 0.91
% of variance 32.32 20.86 12.24 17.80 22.13 19.93 18.47 11.80 11.02 49.72 24.45 13.14
Cumulative % 32.32 53.17 65.41 83.21 22.13 42.06 60.54 72.34 83.36 49.72 74.17 87.30
The bold values indicate absolute component loadings higher than 0.5, which are considered significant contributors to the variance in
the hydrochemistry.

determine the total number of factors for each data set in in the hydrochemistry.
this analysis. Under this criterion, only factors with
eigenvalues greater than or equal to 1 will be accepted as
possible sources of variance in the data, with the highest Group 2
priority ascribed to the factor that has the highest
eigenvector sum. The rationale for choosing 1 is that a
Five significant PCs explain 83.36% of the total variation
factor must have a variance at least as large as that of a
in the hydrochemistry. Most of the variance is contained
single standardized original variable to be acceptable.
in the PC1 (22.13%), which is associated with the
variables pH, Ca and Cl (Table 3). PC 2 represents 19.93%
of the total variation in the hydrochemistry and has high
Group 1
loadings for Mg, Pb and Fe. The variables Na, HCO3 and
SO4 contribute most strongly to the third component
Four principal components (PC) were extracted and
(PC3) that explains 18.47% of the total variance. PC1,
rotated using the varimax normalization (Kaiser, 1960).
PC2 and PC3 are assumed to be indicative of the natural
An initial run using the Kaiser criterion (Kaiser, 1960)
processes and water-rock interaction. PC4 and PC5,
resulted in five principal components. However, it was
respectively, show high loadings for EC and K.
observed that the fifth factor would not constitute a
unique source of variance in the hydrochemistry since it
had only one loading greater than 0.50. It was therefore
dropped and four factors were chosen for varimax Group 3
rotation. The results of (Table 3) show that the four PC
account for more than 83.21% of the total variance, which Three PCs explain 87.30% of the total variance (Table 3).
is quite good and can be relied upon to identify the main Most of the variance is contained in the PC1 (49.72%),
sources of variation in the hydrochemistry. PC1 which is associated with the variables pH, Mg, Na, Cl,
represents about 32.32% of the variance and has high SO4, HCO3 and Pb and probably shows the result of
absolute loadings for EC, Ca, Mg and Cl and probably mineral water reactions in the plain. The aquifer is
shows the result of mineral water reactions in the area. sedimentary rocks, which range from sandstones to
PC 2, which accounts for 20.86% of the total variance, conglomerates to clays with varying mineralogical
contains high loadings for Na, HCO3 and Fe. PC1and compositions. The variables EC, Ca and K contribute
PC2 represents the major geochemical processes taking most strongly to the third component (PC2), which
place, which is most probably the presence and explains 24.45% of the total variance and probably
dissolution of some carbonate, dolomitic and evaporite represents the presence of carbonate minerals in the
minerals in the aquifer. PC3 and PC4, respectively, show aquifer. PC3 explains 13.14% of the variance and is
high loadings for Pb and pH, K and SO4. PC3 and PC4, mainly related to Fe and assumed to be indicative of the
respectively, account for 12.24 and 17.80% of the variation water-rock interaction.
 Belkhiri et al. 533

CONCLUSION examination of water and wastewater, 20th edn. Am. Public Health
Assoc., Am. Water Works Assoc., Water Environ. Fed., Wash.
Duffy CJ, Brandes D (2001). Dimension reduction and source
Multivariate statistical techniques including cluster and identification for multispecies groundwater contamination. J.
principal component analysis can successfully be used to Contaminant Hydrol., 48: 151-165.
derive information from the data set about the possible Farnham IM, Johannesson KH, Singh AK, Hodge VF, Stetzenbach KJ
(2003). Factor analytical approaches for evaluating groundwater
influences of the environment on groundwater quality and
trace element chemistry data. Analytica Chimica Acta., 490(1-2): 123-
also identify natural groupings in the set of data. These 138.
methods are important to avoid misinterpretation of Farnham IM, Stetzenbach KJ, Singh AK, Johannesson KH (2000).
environmental monitoring data due to uncertainties. In Deciphering groundwater flow systems in Oasis Valley, Nevada,
Using trace element chemistry, multivariate statistics, and
this study, multivariate statistical methods were applied to Geographical Information System. Math. Geol., 32: 943-968.
data set obtained from Ain Azel plain, Algeria. Frape SK, Fritz P, McNutt RH (1984). Water-rock interaction and
Interpretation of analytical data showed that the chemistry of groundwaters from the Canadian Shield. Geochim.
abundance of the major ions is as follows: Ca Mg > Na Cosmochim. Acta., 48: 1617-1627.
Galcon J (1967). Recherches sur la géologie et les gîtes métallifères du
> K and HCO3 Cl > SO4. Three major water types are
Tell Sétifien, Doct Thesis Sc. Nat. Publ. Serv. Geol., from Algeria.
suggested by the Q-mode HCA analysis. The samples Garrels RM, MacKenzie FT (1967). Origin of the chemical compositions
from the area were classified as less saline water (Ca- of some springs and lakes, In: Equilibrium Concepts in Natural
Mg-HCO3), mixed water (Mg-Ca-HCO3-Cl) and blended Waters. Am. Cancer Soc., Wash., DC.
Guiraud R (1973). Evolution post-triasique de l’avant pays de la chaîne
water (Mg-Ca-Cl-HCO3). The varifactors obtained from
Alpine de l’Algérie, d’après l’étude du bassin d’El Eulma et les
PCA indicate that the parameters responsible for régions voisines, Thesis Sc. Nat. Nice, France.
groundwater quality variations are mainly related to Güler C, Thyne GD (2004). Hydrologic and geologic factors controlling
presence and dissolution of some carbonate, dolomitic surface and Groundwater chemistry in Indian wells-Owens Valley
area, southeastern California, USA. J. Hydrol., 285: 177-198.
and evaporite minerals in relatively less saline water type; Güler C, Thyne GD, McCray JE, Turner AK (2002). Evaluation of
natural processes and water-rock interaction in mixed graphical and multivariate statistical methods for classification of
water type; water-rock interaction in blended water type. water chemistry data. Hydrogeol. J., 10: 455-474.
The results of this study clearly demonstrate the Gupta LP, Subramanian V (1998). Geochemical factors controlling the
chemical nature of water and sediments in the Gomte River, India.
usefulness of multivariate statistical analysis in hydro Envron. Geol., 36: 102-108.
chemical. Hartman J, Berna Z, Stuben D, Henze N (2005). A statistical procedure
for the analysis of seismotechtonically induced hydrochemical
signals: A case study from the Eastern Carpathians, Romania.
REFERENCES Tectonophys., 405: 77-98.
Hem JD (1989). Study and interpretation of the chemical characteristics
Alberto WD, Del PDM, Valeria AM, Fabiana PS, Cecilia HA, De Los of natural water. US Geol. Surv. Water-Supply Paper, p. 2254.
Angeles BM (2001). Pattern recognition techniques for the evaluation Huisman DJ, Kiden P (1998). A geochemical record of late Cenozoic
of spatial and temporal variations in water quality. A case study: sedimentation history in Southern Netherlands. Geologie en
Suquıa River Basin (Cordoba-Argentina). Water Res., 35: 2881-2894. Mijnbouw, 76: 277-292.
Alther GA (1979). A simplified statistical sequence applied to routine Kaiser HF (1960). The application of electronic computers to factor
water quality analysis: a case history. Ground Water, 17: 556-561. analysis. Educ. Psychol. Measurement, 20: 141-151.
Anazawa K, Ohmori H (2005). The hydrochemistry of surface waters in Kim JH, Kim RH, Lee JH, Cheong TJ, Yum BW, Chang HW (2005).
Andesitic Volcanic area, Norikura volcano, central Japan. Multivariate statistical analysis to identify the major factors governing
Chemosphere, 59: 605-615. groundwater quality in the coastal area of Kimje, South Korea.
Anazawa K, Ohmori H, Tomiyasu T, Sakamoto H (2003). Hydrol. Processes, 19(6): 1261-1276.
Hydrochemistry at a volcanic summit area, Norikura, central Japan. Laaksoharju M, Gurban I, Skarman C, Skarman E (1999). Multivariate
Geochimica et Cosmochimica Acta., 67 (18S): 17. mixing and mass balance (M3) calculations, a new tool for decoding
APHA-AWWA-WPCF (1995). Standard methods for the examination of hydrogeochemical information. Appl. Geochem., 14: 861-871.
water and waste water (19th ed.). New York, USA. Liu CW, Lin KH, Kuo YM (2003). Application of factor analysis in the
APHA (1995). Standard methods for the examination of water and assessment of groundwater quality in a blackfoot disease area in
wastewater, 19th edn. Am. public Health Assoc. Wash., DC. Taiwan. Sci. Total Environ., 313: 77-89.
APHA (1989). Standard methods for examination of water and Love D, Hallbauer D, Amos A, Hranova R (2004). Factor analysis as a
wastewater, 17th edn. Am. Public Health Assoc., Wash. DC. tool in groundwater quality management: two Southern African case
Attoucheik L (2006). Etude géochimique des rejets du complexe minier studies. Phys. Chem. Earth., 29(15-18): 1135-1143.
de Kherzet Youssef (SETIF) et son impact sur l’environnement, Mahlknecht J, Steinich B, Navarro de Leon I (2003). Groundwater
Thesis, IST. USTHB, Algeria. chemistry and mass transfers in the Independence aquifer, central
Belkhiri L (2005). Etude hydrogéologique et problème de la qualité des Mexico, by using multivariate statistics and mass-balance models.
eaux souterraines de la plaine de Ain Azel, Wilaya de Sétif Est Meng SX, Maynard JB (2001). Use of statistical analysis to formulate
Algérien. Thesis, University of Batna, Algeria. conceptual models of geochemical behavior: water chemical data
Boutaleb A (2001). Les minéralisations à Pb-Zn du domaine Sétifien- from Butucatu aquifer in Sao Paulo State, Brazil. J. Hydrol., 250: 78-
Hodna : Gitologie, pétrographie des dolomies, microthermométrie et 97.
implications métallogéniques. Ph.D. thesis statement in mining Momen B, Eichler LW, Boylen CW, Zehr JP (1996). Application of
geology, IST. USTHB, Algeria. multivariate statistics in detecting temporal and spatial patterns of
Brown CE (1998). Applied Multivariate Statistics in Geohydrology and water chemistry in Lake George, New York. Ecol. Modelling, 91: 183-
Related Sciences. Springer, New York. 192.
Cameron EM (1996). The hydrochemistry of the Fraser River, British Olmez I, Beal JW, Villaume JF (1994). New approach to understanding
Columbia: seasonal variation in major and minor components. J. multiple-source groundwater contamination: Factor analysis and
Hydrol., 182: 209-215. chemical mass balances. Water Res., 28(5): 1095-1101.
Clesceri LS, Greenberg AE, Eaton AD (1998). Standard methods for the Reghunath R, Murthy TRS, Raghavan BR (2002). The utility of
534 Afr. J. Environ. Sci. Technol.

multivariate statistical techniques in hydrogeochemical studies: an Vila JM (1980). La chaîne Alpine d’Algérie orientale et des confins
example from Karnataka, India. Water Res., 36(10): 2437-2442. Algéro-Tunisiens, Thesis of doctorate be-Sc. Nat. Paris VI, France.
Savornin J (1920). Etude géologique du Hodna et du plateau Sétifien, Walling DE (1980). Water in catchment ecosystems. In: Gower, A.M.
Thèse Sc. Nat. Lyon. France. (Ed.), Geochemistry, Groundwater and Pollution. A. A. Balkema,
STATISTICA® 5.0 for Windows (1998). StatSoft, Inc.,Tulsa OK. USDA, Rotterdam.
Natural Resources Conservation Services, 1999. Soil taxonomy: A Ward JH (1963). Hierarchical grouping to optimize an objective function.
basic system of soil classification for making and interpreting soil J. Am. Stat. Assoc., 69: 236-244.
surveys. Agric. Handbook No. 436, p. 871. WHO (World Health Organization) (1993). Guidelines for drinking water
Suk HJ, Lee KK (1999). Characterization of a ground water quality, vol.1, Recommendations (2nd edn.), Geneva, WHO, pp 130.
hydrochemical system through multivariate analysis: clustering into Williams RE (1982). Statistical identification of hydraulic connections
ground water zones. GroundWater. 37(3): 358-366. between the surface of a mountain and internal mineralized sources.
Tariq RS, Shah MH, Shaheen N, Khalique K, Manzoor S, Jaffar M Groundwater., 20: 466-478.
(2005). Multivariate analysis of trace metal levels in tannery effluents Yidanaa SM, Ophoria D, Banoeng-Yakubob B (2008). A multivariate
in relation to soil and water: a case study from Peshawar, Pakistan. J. statistical analysis of surface water chemistry data-The Ankobra
Environ. Manage., 79: 20-29. Basin, Ghana. J. Environ. Manage., 88: 697-707.
Tebens L, Veldkamp A, Kroonenberg SB (2001). The impact of climate Zeng X, Rasmussen CD (2005). Multivariate statistical characterization
change on the bulk and clay geochemistry of fluvial residual channel of water quality in Lake Lanier, Georgia, USA. J. Environ. Qual. 34:
infillings: the Late Weichselian and Early Holocene River Meuse 1980-1991.
sediments (The Netherlands). J. Quaternary Sci., 13 : 345-356.