JOURNAL GEOLOGICAL SOCIETY OF INDIA

Vo1.74, November 2009, pp.573-578

Multivariate Statistical Analysis of Geochemical Data ofGroundwater in Veeranam Catchment Area, Tamil Nadu

M. SUVEDHAI, B. GURUGNANAM2,M. SUGANYA3and S. V ASUDEVAN4IDepartment of Geology, Alagappa Govt. Arts College, Karaikudi - 630 003

2Department of Earth Sciences, Annamalai University, Annamalai Nagar - 608 002

3No.572, P.S.P.Street, Soodamani Nagar, Karaikudi - 630 0034 Department of Geology, Bharathidasan University, Trichy

Email: swetha_earth@yahoo.co.in

Abstract: The study of hydrogeochemistry of the Mia-Pliocene sedimentary rock aquifer system in Veeranam catchment

area produced a large geochemical dataset. Groundwater samples were collected at 52 sites over 963.86 km2 area

and analyzed for major ions. The large number of data can lead to difficulties in the integration, interpretation and

representation of the results. Two multivariate statistical methods, Hierarchical cluster analysis (HCA) and Factor analysis

(FA), were applied to a subgroup of the dataset to evaluate their usefulness to classify the groundwater samples, and to

identify geochemical processes controlling groundwater geochemistry. Hydrochemical data for 52 groundwater samples

were subjected to Q- and R- mode factor and cluster analysis. R-mode analysis reveals the inter-relations among the

variables studied and the Q-mode analysis reveals the inter-relations among the samples studied. The R-mode factor

analysis shows that Ca, Mg and Cl with HC03 account for most of the electrical conductivity. total dissolved solids andtotal hardness of groundwater. The 'single dominance' nature of the majority of the factors in the R-mode analysis

indicates non-mixing or partial mixing of different types of groundwater. Both Q-mode factor and Q-mode clusteranalyses indicate an exchange between the river water and the groundwater in the vicinity. The rock water interaction

like tlood basin back swamp deposits of silty clayey formation is the major cause for the cluster \I classi fication. Cluster

classification map reveals that 58% of the study area comes under cluster \I classification.

Keywords: Groundwater, Multivariate statistical analysis. Geochemical data, Tamil Nadu.

INTRODUCTION

The objective of the study is to identify the processes

controlling the geochemical evolution of groundwater by

using two proven methods of multivariate analysis of the

geochemical data sets, namely Hierarchical cluster analysis

(HCA) and Factor analysis (FA). The relatively complex

setting and geological history of the study area, use ofHCA and FA aims at distinguishing respective roles of

geological and hydrogeological factors in this hydro-chemical evolution. We also assessed the relative

applicability and complementarities of HCA and FA

methods compared to conventional geochemical grouping

in achieving the scientific evaluations.

Multivariate statistical analysis has been successfullyapplied in a number of hydrogeochemical studies. Steinhorstand Williams (1985) used multivariate statistical analysis

of water chemistry data in two field studies to identify

groundwater sources. Usunoff and Guzma'n-Guzma'n(1989) demonstrated the usefulness of the approach in

hydrogeochemical investigations for understanding the

geological and hydrogeological state of the aquifer.

Multivariate treatment of environmental data is also widely

used to characterize and evaluate groundwater quality

(Vengosh and Keren, 1996; Suk and Lee, 1999; Helena et

al. 2000; Reghunath, 2002; Lambrakis et al. 2004;

Panagopoulos et al. 2004, Vincent Cloutier et al. 2008). It

is also useful for identifying temporal and spatial variations

caused by natural and human factors linked to seasonality.

STUDY AREA

The study area, the Veeranam catchment, occupies an

area of 963.86 km2, falling in parts of Cuddalore andPerambalur districts, Tamil Nadu. It lies between the North

latitudes 11°05'56" - 11°26' and East longitudes 79° 15'30"-

79°32' 10" (Fig. 1). Physiographically, the area is flat with

gentle slope, experiences high rainfall from the north-

east monsoon. Geologically, the area is underlain by

alluvial deposits of Early to Middle Pleistocene. Thenature and character of the alluvium have been studied,

0016-7622/2009-74-5-573/$ 1.00 <9 GEOL. SOC. INDIA

574 M. SUVEDHA AND OTHERS

79"20'O'E 79'30'O'E

A

Legend

. SamplingLocation

0 2' 8r , , , r , , , r

Kilometers

- Road

79'20'O'E

~ Water body

~River79'30'O'E

Fig,I. Map showing water sample locations.

. based on the geological sections prepared from the well logs

of the tube wells in the area. From these logs, it is evident

that alluvial deposits which form the potential aquifers

primarily consist of thick deposits of mottled sandstone,

clay and lignite deposits of Mio-Pliocene age, The

Quaternary formations are restricted to the alluvium of

Cauvery, Kollidam and their distributaries which occurs as

a isolated remnant patches over Cuddalore Formation. The

area is bounded by the river Vellar in the north and Kollidam

in south, running along the eastern to northwestern andnorthern to northeastern boundaries.

MATERIALS AND METHODS

GEOCHEMISTRY

Fifty two water samples were collected in May 2006fromdifferentshallowdug wells and deep bore wells, whichare almost uniformly distributed over the study area. Onlythose wells were selected for sampling purpose which arein constant use and approachable. After half an hourdischarge from the tube wells, the samples were collectedin air tight bottles with stoppers and subjected to chemicalanalysis to see the variations in quality parameters. Major

Unit: Concentration in ppm except pH. EC (liS Cm"), RSC and SAR

(meq I").

cations and anions were estimated by titration method.

Residual sodium carbonate (RSC) was calculated bysubtracting (Ca+Mg) from the values of carbonates and

bicarbonates expressed as epm (Eaton, 1950). Sodiumabsorption ratio (SAR), was calculated by dividing sodium

with root of half (Ca+Mg) expressed as epm (Richard, 1954).

MULTIVARIATE STATISTICAL ANALYSIS

Factor Analysis (FA)

Multivariate techniques can help to simplify and organize

large data sets and to make useful generalizations, that can

lead to meaningful insight (Laaksohmju et al. 1999). Cluster

and factor analyses are efficient ways of displaying complex

relationships among many objects (Davis, 1986). The two

methods in cluster and factor analyses, i.e. Q- and R- mode

analyses have been done for the data generated. R-mode

analysis reveals the interaction among the variables studied

and the Q-mode analysis reveals the interrelation among

the samples studied. The software packages like Statistical

Package for Social Sciences (SPSS) and STATISTICA 6

have been used to carry out the analysis. The data have been

standardized by using standard statistical procedures.

Hierarchical Cluster Analysis (HCA)

Cluster analysis comprises a series of multivariate

methods which are used to find true groups of data. In

clustering, the objects are grouped such that similar objects

fall into the same class (Danielsson et al. 1999), Hierarchical

clustering joins the most similar observations, and then

successively the next most similar observations. The levels

of similarity at which observations are merged are used to

JOUR.GEOL.SOC.INDIA, VOL.74. NOY 2009

Table 1. Mean and standard deviation of the chemical parameters ofgroundwater

Parameter Valid N Mean Minimum Maximum Stc!.Dev

Na 52 51.98 5.52 358.8 56.81K 52 11.92 0.39 117.3 24.39

Ca 52 64,17 601 170.34 42.67Mo 52 17.02 1.22 46.21 1197"CI 52 131.41 17.73 425.52 102.85

HCO, 52 205.12 42.71 842.08 14906

SO. 52 5.27 0 96.06 14.34

pH 52 6.89 6 7.7 0,43

EC 52 713.65 140 2480 482.90

TDS 52 382.28 72.96 1375.27 264.68

TH 52 230.28 20.02 565.45 146.5.6RSC 52 0.29 0 6.8 104

SAR 52 1.52 0 8.34 1.29

STATISTICAL ANALYSIS OF GEOCHEMICAL DATA OF GROUNDWATER, VEERANAM CATCHMENT, TAMIL NADU

construct a dendrogram. In this study, a standardized spaceEuclidian distance (Davis, 1986) is used. A low distance

shows the two objects are similar or "close together",

whereas a large distance indicates dissimilarity.

RESULTS AND DISCUSSION

The chloride. calcium. sodium and bicarbonate content

shows a significant difference between the medium and

maximum values, the mean values being near the quarter

values of the maximum values. It suggests that local

contamination to the groundwater system. The wide range

of bicarbonate contents, from 42.71 to 842.1 ppm is the

result of the lateral geological variations of the layers.

Box plots of the chemical concentration show that

bicarbonate, calcium, chloride and TDS have the largest

dispersions (Fig.2). The enrichment of chloride and TDSfrom values of 17.73 to 425.52 and 72.95 to 1375.27

respectively. is observed in the groundwater on the eastern

side of the study area. It is due to the high enrichment of the

flood basin back swamp deposit of silty clayey formation.The increase in the salt concentration could be associated

with different mechanisms like water rock interaction

processes.

R - Mode Factor Analysis

R-mode factor analysis of different chemical constituents

of the groundwater of Veeranam catchment area has been

carried out. All cations and anions, TDS, EC, pH andhardness have been considered for the present analysis. The

1600

. Median1400 ~- n~ 25%-75%' _h _h hh - h h _hhh- - h- - _nn_-

:::r::: Mn-Max

E 1200- -- _nn nn _nn -nn _n- - --- -- - - - .-- _h_- h__-a.a.c::= 1000 --- h-- --- nn_nnn _n _nnn_nn__h'__h-_nn--c:0..,

~c:<I>uc:0u

600 -n - .n '" -_.--. --_h - h-

600 ._nn_nnh h h nnnnnnn_nn_n

1-"-.-------

i?! . _n__-jl

L

200

Na CI HCO, so, IDS 1HK C. Mg

PARAMETffiS

Fig.2. Box and Whisker plot for chemical parameters of ground-water samples.

JOUR.GEOL.SOC.INDIA, VOL.74. NOV. 2009

575

analysis generated five factors which together account for

95.08% of variance. The rotated loadings. eigen values,

percentage of variance and cumulative percentage ofvariance of all the five factors are given in Table 2. The first

eigen value is 7.69 which accounts for 59.2o/r of the totalvariance and this constitutes the first and main factor. The

second and third eigen values are 2.33 and 1.0 I and these

account for 18% and 7.84% respectively, of the total

variance. Each of the remaining eigen values constitutes lessthan 10% of the total variance. The first factor (which

accounts for 59.2% of the total variance) is characterised

by very high loadings of Ca, Mg, CI and EC, and moderate

to high loadings of bicarbonate and pH. This factor reveals

that the EC and TDS in the study area are mainly due to Ca

and Mg and CI, though bicarbonate also plays a substantial

role in determining EC and TDS. This factor accounts for

the temporary hardness of the water. The second factor

(which accounts for 18% of the total variance) is mainly

associated with very high loading ofNa, Cl and bicarbonate,

and also with moderate loading ofTDS. This factor accounts

for the temporary salinity of the water. The loading ofbicarbonate is same as the first factor. Factors 3-5 are

characterized by the dominance of only one variable each,

such as S04 (factor 3), HCOJ (factor 4) K (factor 5), andtogether these six factors account for 17% of the totalvariance.

Q - Mode Factor Analysis

The rotated loadings, eigen values, percentage of

variance and cumulative percentage of variance of the first

three factors are given in Table 4. Q-mode factor analysis

of the 52 groundwater samples generated three factors which

Table 2. R - Mode factoranalysiswith Varimax normalized rotation

Parameter Factor I Factor 2 Factor 3 Factor 4 Factor 5

Na 0.444 0.868 0090 0.018 0 150

K 0.140 0.412 0.025 0.071 0.886

Ca 0.972 -0.035 0.118 0.089 -0.052

Mg 0.751 0.093 ..0007 0.576 0 132

CI 0.875 0.354 0.084 -0.209 0229

HC03 0566 0538 0.073 0.601 0.086

SO, 0.119 0 125 0949 0.000 0.001

pH 0.541 0.120 0.489 0327 0278

EC 0.820 0.499 0.099 0.17') 0 18(,

TDS 0.801 0.516 0.145 0.136 0.224

TH 0.959 0.006 0.084 0.258 0.007

RSC -0.128 0.879 0126 0283 0.206

SAR 0172 0935 0.046 ..0.038 0.218

Eigenvalue 7.699 2.385 1.019 (J.(,M 0592

% Total Variance 59223 1834 7.84 5.107 4.560

Cumulative % 59.22 77.57 85.41 90.524 95.08

576 M. SUVEDHA AND OTHERS

together accounted for 99.87% of the total variance

(Table 3). The three factors obtained in this way were rotated

using the Varimax procedure (Knudson et a!. 1977), which

could be more easily interpreted. The first factor (which

explains 37.96% of the total variance) was considered as

major factor controlling the relative proportions of major

element existing in the groundwater samples and had the

high loadings of almost all the samples except those fromlocation nos. 28 and 46.

On the other hand, groundwater samples from the three

locations 28, 46 and 51 had high loadings in the second

factor. As mentioned earlier, the Q-mode factor analysis

described the relative proportions of these major elementsin groundwater samples. Therefore, the relative proportions

of major elements in these groundwater samples were

controlled completely by the three factors which together

explain 99.87% of the total variance. The distribution of

wells explained by factors 2 and 3 do not conform to any

kind of spatial pattern. However, the majority of the sampleswithin factor 1 fall on either side of the main course of

the river system. This strongly suggests that there is an

exchange between the river water and groundwater in the

vicinity. This has also been discussed by Reghunath et a!.(2002).

HIERARCHICAL CLUSTER ANALYSIS

The HCA is a data classification technique. There are

different clustering techniques, but the hierarchical clustering

is the one most widely applied in Earth sciences (Davis,1986), and often used in the classification of

hydrogeochemical data (Steinhorst and Williams, 1985;Schot and van der Wal, 1992; Ribeiro and Macedo, 1995;

Gu"ler et aI., 2002). The result of the hierarchical cluster

analysis was given as a dendrogram (Fig.3). For this project,the Euclidean distance was chosen as the distance measure,

or similarity measurement, between sampling sites. The

sampling sites with the larger similarity are first grouped.

Next, group of samples are joined with a linkage rule, and

the steps are repeated until all observations have been

classified. With this geochemical dataset, Ward's methodwas more successful to form clusters that are more or less

homogenous and geochemically distinct from other clusters,

compared to other methods such as the weighted pair-group

average. Ward's method is distinct from other linkage rules,

because it uses an analysis of variance approach to evaluatethe distances between clusters (StatSoft Inc., 2004). Other

studies in their cluster analysis (Adar et a!. 1992; Schot andvan der Wal, 1992). Gu"ler et a\. (2002) also found that

using the Euclidean distance as a distance measure and

S.No

Table 3. Q - Mode factor analysis with Varimax normalized rotation

Factor 3

]

2

3

4

5

6

7

8

9

10

II

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

272829

3031

323334

353637

3839

4041

42434445

4647

48

4950

5152

Eigenvalue .,% Total Variance

Cumulative %

Factor I

0.637

0.6340.7110.5510.539

0.5010.63

0.563

0.5160.685

0.7050.5510.5450.765

0.5470.5330.635

0.6690.5780.5280.563

0.5560.502

0.5460.5830.473

0.5660.4470.7110.547

0.6520.521

0.6720.5890.679

0.6650.6480.665

0.7620.7190.708

0.7030.55

0.6710.5480.4470.685

0.7110.765

0550.524

0.7719.74

37.96237.962

Factor 2

0.614

0.5460.5210.4890.618

0.52506270589

0.6870.585

0.5320.6890.7050.522

0.5370.6960.6490.5720.665

0.6510.589

0.5890.5240.538

0.5780.619

0.5420.7540.5210.537

0.6430.708

0.5640.6380.524

0.5420.5880.533

0.3830.4550.513

0.5910.71

0.587071

07540584

0.50.522

0.710.727

0.41118.50 I

35.5873.541

0.466

0.5480.4720.6760571

0.6850.4580.5780.511

0.435

0.4690.47

0.4530375

06420.4810.418

0.4750.472

0.5450.578

0.5860.687

0.642057

0.626

0.621

0.481

0.472

0.642

0.401

0.477

0.479

0.496

0.514

0.51

0.484

0.522

052

0.525

0.484

0.394

0.44

0.418

0.441

0.481

0.436

0.494

0.375

0.44

0.444

0.486

13.694

26.335

99.876

JOUR.GEOL.SOC.INDIA. VOL.74. NOY.2009

STATISTICAL ANALYSIS OF GEOCHEMICAL DATA OF GROUNDWATER, VEERANAM CATCHMENT, TAMIL NADU

110"2J635,329"""""7173437

c: 120 '+:: "ro "0 500 .....J 4451C),c: 21= 25Co 19E 31ro 22

en4118"20"15302427"263229..614.."".

0 20 40

577

60 80 100 120

Linkage Distance

Fig.3.Dendorgram of the hierarchical cluster analysis using the Ward method.

Ward's method as a linkage rule produced the most

distinctive group.

There are three major clusters as shown in Fig. 2. Clusters

I, 2 and 3 correspond to the factors I, 2 and 3 respecti vely.

The similarity of the Q-mode cluster analysis to the Q-mode

factor analysis confirms the interpretations made using the

Q-mode factor analysis. To understand the spatialdistribution of various cluster classes, the results were taken

into GIS platform wherein spatial distribution map is

prepared (Fig.4). The salient findings of spatial distribution

map are given in the Table 4.

Table 4. Results of Cluster Classification spatial distribution map

CONCLUSION

The scientific evaluation ofthe raw data by FA and HCAleads to the conclusion that the water-rock interaction

process is the major mechanism responsible for the

groundwater salinity in the study area. The water samplesare mainly of calcium-bicarbonate type, pointing to the

JOUR.GEOL.SOC.INDIA, VOL.74, NOY. 2009

aquifer lithology dominated by calcareous sandstone and

clayey formation. The factor analysis reveals that the calcium

and magnesium concentrations are the major sources for

"'20.0'N

N

t

""'WE 79"300'E

/"-;;'-::~»

!<~ia(if@/$

P.,.,"",

"'10"'N

0 '4 8,,, ,," I

Knoon,'",

Legend

~CI'sI'"

[::::::::1 CI,sI., 2

GIJ Clu".' 379"2O"<rE 79"3O"O'E

Fig.4. Spatial distribution of map of cluster classes.

Grid code Cluster classification Area in km'

I Cluster I 127.27

2 Cluster II 533.8

3 Cluster III 302.79

578 M. SUVEDHA AND OTHERS

the hardness of groundwater. An anthropogeniccontamination was identified in both the aquifers, due tolocal pollution inputs. The Q-mode factor and cluster

analyses indicate that exchange between the river water and

ADAR. EM., ROSENTHAL,E, ISSAR, A.S. and BATELAAN,O. (1992)

Quantitative assessment of the flow pattern in the southern

Arava Valley (Israel) by environmental tracers and a mixing

cell mode!. Jour. Hydrology, v.136, pp.333-352.

DANIELSSON,A., CATO, I., CARMAN, R. and RAHM, L. (1999) Spatial

clustering of metals in the sediments of the Skagerrak/Kattegat.

Applied Geochemistry, v.14, pp.689-706.

EATON,EM. (1950) Significance of carbonate in irrigation water.

Soil Sci., v.69, pp.123-133.

GOLER, c., THYNE, G.D., MCCRAY, J.E. and TURNER, A.K. (2002)

Evaluation of graph ical and multivariate statistical methods

for classification of water chemistry data. Hydrogeology Jour.,v.lO, pp.455-474.

HELENA,B., PARDO,B., VEGA,M., BARRADO,E., FERNANDEZ,J.M.

and FERNANDEZ,L. (2000) Temporal evolution of groundwater

composition in an alluvial aquifer (Pisuerga River, Spain) by

rincipal component analysis. Water Res., v.34(3), pp.807-816.KNUDSON,E.J., DUEWER,D.L., CHRISTIAN,G.D. and LARSON,T.v.

(1977) Application of factor analysis to the study of rain

chemistry in the Puget Sound region. In: B.R. Kowalski (Ed.),

Chemometric: Theory and Application. ACS SymposiumSeries, Washington, DC, pp.80-116.

LAMBRAKIS,N., ANTONAKOS,A. and PANAGOPOULOS,G., (2004) The

use of multi component statistical analysis in hydrogeological

environmental research. Water Res., v.38, pp.1862-1872.PANAGOPOULOS, G., LAMPRAKIS, N., TSOLlS-KATAGAS, P. andPAPouLls,

D. (2004) Cation exchange processes and human activities in

inconfined aquifers. Environ. Geo!., v.46, pp.542-552.REGHUNATH,R., SREEDHARA,M.T.R. and RAGHAVAN,B.R. (2002)

the groundwater plays a dominant role in the hydrochemical

evolution of groundwater. Cluster classification map revealsthat 58°,{i of the study area comes under cluster IIclassification.

References

The utility of multivariate statistical techniques in

hydrogeochemical studies: an example from Karnataka, India.Water Res., v.36(10), pp.2437-2442.

RIBEIRO,L. and MACEDO,M.E. (1995) Application of multivariate

statistics, trend and cluster analysis to groundwater quality

in the Tejo and Sado aquifer. In: Groundwater Quality:

Remediation and Protection. Proceedings of the PragueConference, May 1995. IAHS Pub!. No.225, pp.39-47.

SCHOT,PP. and VANDERWAL,J. (1992) Human impact on regional

groundwater composition through intervention in natural flow

patterns and changes in land use. Jour. Hydrology, v.134,pp.297-313.

STATSOFTINC. (2004) STATISTICA (Data Analysis SoftwareSystem), Version 6.

STEINHORST,R.K. and WILLIAMS.R.E. (1985) Discrimination of

groundwater sources using cluster analysis, MANOVA,

canonical analysis and discriminant analysis. Water ResourcesRes., v.21, pp.1149-1156.

SUK, H. and LEE, K. (1999) Characterization of a ground water

hydrochemical system through multivariate analysis: clustering

into ground water zones. Ground Water, v.37(3), pp.358-366.VENGOSH,A. and KEREN,R. (1996) Chemical modifications of

groundwater contaminated by recharge of treated sewageeftluent. Contam. Hydro!., v.23, pp.347-360.

VINCENTCLOUTIER.,RENE LEFEBVRE.,RENETHERRIEN.,MARTINE,

M. and SAVARD.(2008) Multivariate statistical analysis of

geochemical data as indicative of the hydrogeochemical

evolution of groundwater in a sedimentary rock aqui fer systemJour. Hydrology, v.353, pp.294-313.

(Received: 28 April 2008; Revisedform accepted: 20 June 2009)

JOUR.GEOL.SOCINDIA, VOL.74, NOY.2009