water quality assessment and apportionment of pollution sources

7/25/2019 Water Quality Assessment and Apportionment of Pollution Sources

1/20

Analytica Chimica Acta 538 (2005) 355374

Water quality assessment and apportionment of pollution sources ofGomti river (India) using multivariate statistical techniques

a case study

Kunwar P. Singh a,, Amrita Malika, Sarita Sinha b

a Environmental Chemistry Section, Industrial Toxicology Research Centre, Post Box 80, MG Marg, Lucknow 226 001, Indiab National Botanical Research Institute, Rana Pratap Marg, Lucknow 226 001, India

Received 31 October 2004; received in revised form 30 January 2005; accepted 1 February 2005

Available online 9 March 2005

Abstract

Multivariate statistical techniques, such as cluster analysis (CA), factor analysis (FA), principal component analysis (PCA) and discriminant

analysis (DA) were applied to the data set on water quality of the Gomti river (India), generated during three years (19992001) monitoring at

eight different sites for 34 parameters (9792 observations). This study presents usefulness of multivariate statistical techniques for evaluation

and interpretation of large complex water quality data sets and apportionment of pollution sources/factors with a view to get better information

about the water quality and design of monitoring network for effective management of water resources. Three significant groups, upper

catchments (UC), middle catchments (MC) and lower catchments (LC) of sampling sites were obtained through CA on the basis of similarity

between them. FA/PCA applied to the data sets pertaining to three catchments regions of the river resulted in seven, seven and six latent

factors, respectively responsible for the data structure, explaining 74.3, 73.6 and 81.4% of the total variance of the respective data sets. These

included the trace metals group (leaching from soil and industrial waste disposal sites), organic pollution group (municipal and industrial

effluents), nutrients group (agricultural runoff), alkalinity, hardness, EC and solids (soil leaching and runoff process). DA showed the best

results for data reduction and pattern recognition during both temporal and spatial analysis. It rendered five parameters (temperature, totalalkalinity, Cl, Na and K) affording more than 94% right assignations in temporal analysis, while 10 parameters (river discharge, pH, BOD,

Cl, F, PO4, NH4N, NO3N, TKN and Zn) to afford 97% right assignations in spatial analysis of three different regions in the basin. Thus,

DA allowed reduction in dimensionality of the large data set, delineating a few indicator parameters responsible for large variations in

water quality. Further, receptor modeling through multi-linear regression of the absolute principal component scores (APCS-MLR) provided

apportionment of various sources/factors in respective regions contributing to the river pollution. It revealed that soil weathering, leaching and

runoff; municipal and industrial wastewater; waste disposal sites leaching were among the major sources/factors responsible for river quality

deterioration.

2005 Elsevier B.V. All rights reserved.

Keywords: Gomti river; Water quality management; Cluster analysis; Factor analysis; Principal component analysis; Discriminant analysis; Source apportion-

ment

1. Introduction

The surface water quality is a matter of serious concern

today. Rivers due to their role in carrying off the munici-

pal and industrial wastewater and run-off from agricultural

Corresponding author. Tel.: +91 522 2508916; fax: +91 522 2628227.

E-mail address:kpsingh [email protected] (K.P. Singh).

land in their vast drainage basins are among the most vul-

nerable water bodies to pollution. The surface water quality

in a region is largely determined both by the natural pro-

cesses (precipitation rate, weathering processes, soil erosion)

and the anthropogenic influences viz. urban, industrial and

agricultural activities and increasing exploitation of water

resources [1,2]. The municipal and industrial wastewater dis-

charge constitutes the constant polluting source, whereas, the

0003-2670/$ see front matter 2005 Elsevier B.V. All rights reserved.

doi:10.1016/j.aca.2005.02.006


2/20

356 K.P. Singh et al. / Analytica Chimica Acta 538 (2005) 355374

surface run-off is a seasonal phenomenon, largely affected by

climate in the basin. Seasonal variations in precipitation, sur-

face run-off, ground water flow and water interception and

abstraction have a strong effect on river discharge and subse-

quently on the concentration of pollutants in river water [3].

Since, rivers constitute the main inland water resources for

domestic, industrial and irrigation purposes, it is imperativeto prevent and control the rivers pollution and to have reliable

information on thequality of water for effective management.

In view of the spatial and temporal variations in the hydro-

chemistry of rivers, regular monitoring programs are required

for reliable estimates of the water quality. This results in a

huge and complex data matrix comprised of a large number

of physico-chemical parameters, which are often difficult to

interpret and draw meaningful conclusions[4]. Further, for

effective pollution control and water resource management,

it is required to identify the pollution sources and their quan-

titative contributions.

The Gomti river, a major tributary of the Ganga river sys-

temin India,originates from a natural reservoir in theforested

area (elevation of about 200 m; North latitude 28

34

andEast longitude 8007) in Uttar Pradesh. The river traverses

a total distance of about 730 km before finally merging with

the Ganga river near Varanasi. It drains a catchments area

of about 25,800 km2. Kathna, Sarayan, Reth, Luni, Kalyani

and Sai rivers are the tributaries of the Gomti river. Lucknow

(population about 3.5 million), Sultanpur (population about

0.2 million) and Jaunpur (population about 0.2 million) are

Fig. 1. Map showing the water quality monitoring sites on the Gomti river.


3/20

K.P. Singh et al. / Analytica Chimica Acta 538 (2005) 355374 357

the three major urban settlements on the banks of the river

(Fig. 1).The river serves as a major source of domestic water

supply of the Lucknow city, the State capital of Uttar Pradesh.

Subsequently, the river receives back the untreated domestic

wastewater from Lucknow city (about 450 mld), Jagdishpur,

Sultanpur and Jaunpur towns and effluents from a few indus-

tries (distilleries, sugar mills, chemical and others) directlyduring its course. The river during its course of about 730 km

receives pollution load both from the point and non-point

sources. It receives agricultural run-off from its vast catch-

ments area directly or through its tributaries and wastewater

drains (45 numbers). The Gomti river has been identified as

one of the most polluted rivers in India. Asa part of the Ganga

Action Plan, a vast database on its quality has been gener-

ated through regular monitoring of the river during the last

about a decade for its regeneration and management under

the National Rivers Conservation Program (NRCP).

The application of different multivariate statistical tech-

niques viz., cluster analysis (CA), discriminant analysis

(DA), principal component analysis (PCA)/factor analysis(FA), source apportionment by multiple linear regression on

absolute principal component scores (APCS-MLR) for inter-

pretation of the complex databasesoffers a better understand-

ing of water quality in the study region. These techniques also

permit identification of the possible factors/sources that are

responsible for the variations in water quality and influence

the water system and in apportionment of the sources, which,

thus offers valuable tool for developing appropriate strategies

for effective management of the water resources[510].

In the present paper, the large data base obtained during

the three years monitoring program (9792 observations) was

subjected to different multivariate statistical techniques witha view to extract information about the similarities or dissim-

ilarities between the sampling sites, identification of water

quality variables responsible for spatial and temporal varia-

tions in river water quality, the hidden factors explaining the

structure of the database, the influence of the possible sources

(natural and anthropogenic) on the water quality parameters;

and the source apportioning for estimation of the contribution

of possible sources on the concentration of determined water

quality parameters of the Gomti river.

2. Methods

2.1. Monitoring sites

In the present study, total eight sites, namely Neemsar

(site 1), Bhatpur (site 2), Gaughat (site 3), Mid-Lucknow

(site 4), Pipraghat (site 5), Gangaganj (site 6), downstream

of Sultanpur (site 7) and Jaunpur (site 8) were selected on the

Gomti river under the river quality-monitoring network. The

sampling network was designed to cover a wide range of de-

terminants at key sites, which reasonably represent the water

quality of the river system accounting for tributary and inputs

from wastewater drains that have impact on downstream wa-

ter quality. The first three sites (13) are located in the area

of relatively low river pollution and are upstream of the Luc-

know city. Other three sites (46) are located in the region

of high river pollution as there are a number of wastewater

drains (27 numbers) and two highly polluted tributaries emp-

tying in to the river in this stretch. The last two sites (7 and

8) are in the downstream region of moderate river pollutionas the river considerably recovers in the course (Fig. 1).

2.2. Sampling and chemical analysis

Water samples were collected each month at three points

(1/4, 1/2 and 3/4) across the river width at all the eight

sites with a view to monitor changes caused by the sea-

sonal hydrological cycle during the study period (January

1999December 2001). Sampling, preservation and trans-

portation of the water samples to the Laboratory were as per

standard methods[11]. The Gomti river discharge was mea-

sured at each of the eight sites along with sampling followingthe areavelocity method using the calibrated water current

meter. Water temperature was measured on thesite using mer-

cury thermometer. All other parameters were determined in

laboratory following the standard protocols[11]. The sam-

ples were analysed for 33 parameters, namely temperature,

pH, electrical conductivity, total alkalinity, total hardness,

calcium hardness, total solids, total dissolved solids, total

suspended solids, dissolvedoxygen, 5-days biochemical oxy-

gen demand, chemical oxygen demand, ammonical nitrogen,

nitrate nitrogen, total kjeldahl nitrogen, chloride, fluoride,

sulfate, phosphate, sodium, potassium, calcium, magnesium,

total coliform, faecal coliform, cadmium, chromium, iron,

manganese, copper, lead, zinc and nickel. Different water

quality parameters, their units and methods of analysis are

summarized inTable 1.The analytical data quality was en-

sured through careful standardization, procedural blank mea-

surements, spiked and duplicate samples. The ionic charge

balance of each sample was within5%. The laboratory also

participated in regular national program on analytical quality

control (AQC). Basic statistics of the 3-years data set on river

water quality is summarized inTable 2.

2.3. Data treatment and multivariate statistical methods

Correlation structure between the variables was stud-

ied using the Spearman R coefficient as a non-parametric

measure of the correlation between the variables [12]. In

the present study, the temporal variations of the river wa-

ter quality parameters (Table 1) were evaluated through

season-parameter correlation matrix using the Spearman

non-parametric correlation coefficient (Spearman R). The

water quality parameters were grouped in three different sea-

sons (winter, summer and monsoon) and each assigned a

numerical value in the data file, which as a variable corre-

sponding to the season was correlated (pair by pair) with all

the measured parameters.


4/20


Table 1

Water quality parameters, abbreviations, units and analytical methods as measured during 19992001 for the Gomti river water

Variables Abbreviations Units Analytical methods

Discharge Dis m3 s1 Current meter

Temperature Temp C Mercury thermometer

Electrical conductivity EC S cm1 Electrometric

PH pH pH unit pH-meter

Total solids TS mg l1 GravimetricTotal dissolved solids TDS mg l1 Gravimetric

Total suspended solids TSS mg l1 Gravimetric

Total alkalinity T-Alk CaCO3mg l1 Titrimetric

Total hardness T-Hard CaCO3mg l1 Titrimetric

Calcium hardness Ca-Hard CaCO3mg l1 Titrimetric

Dissolved oxygen DO mg l1 Winkler azide method

Biochemical oxygen demand BOD mg l1 Winkler azide method

Chemical oxygen demand COD mg l1 Dichromate reflex method

Chloride Cl mg l1 Spectrophotometric

Fluoride F mg l1 Spectrophotometric

Phosphate PO4 mg l1 Spectrophotometric

Sulphate SO4 mg l1 Spectrophotometric

Potassium K mg l1 Flame photometer

Sodium Na mg l1 Flame photometer

Calcium Ca mg l1 Flame AASMagnesium Mg mg l1 Flame AAS

Ammonical nitrogen NH4N mg l1 Spectrophotometric

Nitrate nitrogen NO3N mg l1 Spectrophotometric

Total kjeldahl nitrogen TKN mg l1 Spectrophotometric

Total coliform T. Coli MPN/100 ml Multiple tube method

Feacal coliform F. Coli MPN/100 ml Multiple tube method

Cadmium Cd mg l1 ICP-OES

Chromium Cr mg l1 ICP-OES

Iron Fe mg l1 ICP-OES

Manganese Mn mg l1 ICP-OES

Lead Pb mg l1 ICP-OES

Copper Cu mg l1 ICP-OES

Zinc Zn mg l1 ICP-OES

Nickel Ni mg l1 ICP-OES

Multivariate analysis of the river water quality data setwas

performed through CA, DA, PCA, FA and APCS-MLR tech-

niques. DA was applied on raw data, whereas, CA, PCA and

FA were applied on experimental data standardized through

z-scale transformation in order to avoid misclassification due

to wide differences in data dimensionality [8,9]. Standardiza-

tion tends to minimize the influence of difference of variance

of variables and eliminates the influence of different units of

measurement and renders the data dimensionless.

2.3.1. Cluster analysis

Cluster analysis groups the objects (cases) into classes

(clusters) on the basis of similarities within a class and dis-

similarities between different classes. The results of CA help

in interpreting the data and indicate patterns [3,10].In hier-

archical clustering, clusters are formed sequentially by start-

ing with the most similar pair of objects and forming higher

clusters step by step. Hierarchical agglomerative CA was per-

formed on the normalized data set (mean of observations

over the whole period) by means of the Wards method us-

ing squared Euclidean distances as a measure of similarity

[13].Cluster significance was determined using the criterion

of 0.66Dmax[5].

Cluster analysis was applied to the river water quality data

set with a view to group the similar sampling sites (spatial

variability) spread over the river stretch and in the resulted

dendrogram, the linkage distance is reported as Dlink/Dmax,

which represent the quotient between the linkage distance

for a particular case divided by the maximal distance, mul-

tiplied by 100 as a way to standardize the linkage distance

represented ony-axis[9,10,12].

2.3.2. Discriminant analysis

Discriminant analysis determines the variables that dis-criminate between two or more naturally occurring groups.

It constructs a discriminant function (DF) for each group [14]

as in Eq.(1):

f(Gi) = ki +

n

j=1

wijpij (1)

whereiis the number of groups (G),kithe constant inherent

to each group, n the number of parameters used to classify

a set of data into a given group, wj the weight coefficient,

assigned by DA to a given selected parameter (pj).


5/20

Table 2

Range, mean and S.D. of different water-quality parameters at different locations of the Gomti river during 19992001

Parameters Site 1 Site 2 Site 3 Site 4 Site 5 Site 6

Discharge (m3 s1) Range 6.137.3 6.336.5 5.645.2 6.952.6 8.853.8 14.358.5

Mean 15.70 16.84 16.85 22.49 22.08 28.95

S.D. 8.93 8.48 11.36 14.48 13.80 14.39

Temperature (C) Range 18.035.0 15.033.0 15.033.0 16.035.0 17.034.0 16.033.0

Mean 26.71 26.21 26.44 26.71 26.44 26.44

S.D. 5.26 5.36 5.29 5.66 5.33 5.25 EC (S cm1) Range 208.3443.3 210.0462.0 150.0460.3 216.7511.7 230.0588.3 256.7606.7

Mean 356.80 379.30 369.38 411.43 452.49 458.18

S.D. 60.32 69.38 78.64 81.82 93.52 103.20

PH Range 7.98.7 7.98.7 8.08.8 6.08.7 7.48.7 7.48.7

Mean 8.34 8.33 8.36 8.05 7.92 7.98

S.D. 0.17 0.18 0.18 0.40 0.27 0.28

TS (mg l1) Range 182.5372.0 200.0359.3 226.7414.9 205.3432.9 186.0440.0 268.0470.1

Mean 282.5 285.97 287.65 311.73 323.83 330.12

S.D. 44.57 39.21 40.40 53.13 56.58 48.56

TDS (mg l1) Range 142.3300.0 173.3302.7 173.3300.0 165.9366.7 148.7386.7 185.7380.0

Mean 236.60 246.22 247.84 265.98 276.87 283.08

S.D. 35.57 37.56 27.82 44.99 48.90 45.03

TSS (mg l1) Range 5.686.0 16.086.7 16.1108.9 8.7138.0 13.9124.4 11.3167.2

Mean 44.95 39.58 38.39 45.30 47.23 48.00

S.D. 22.16 18.37 21.10 29.94 26.15 33.97

T-Alk (mg l1) Range 99.3238.0 85.3258.0 110.7261.3 112.0274.7 113.3292.0 120.0289.3

Mean 189.13 196.48 197.70 206.68 213.98 216.88

S.D. 40.12 47.21 47.69 47.58 48.83 50.30

T-Hard (mg l1) Range 52.0230.7 45.3236.0 49.3244.0 78.7254.7 60.0262.7 46.7280.0

Mean 157.67 168.58 170.56 184.04 193.92 193.28

S.D. 49.53 51.96 56.48 50.61 55.95 58.89

Ca-Hard (mg l1) Range 44.0149.3 40.0156.0 37.3156.7 12.7169.3 53.3184.0 40.0200.0

Mean 89.80 89.05 90.31 105.24 112.65 112.50

S.D. 28.18 26.63 31.3 34.00 29.42 32.09

DO (mg l1) Range 3.610.6 3.810.5 4.110.4 0.07.8 0.05.4 1.96.3

Mean 7.41 7.19 7.17 4.05 0.95 3.68

S.D. 1.77 1.78 1.81 2.02 1.52 1.00

BOD (mg l1) Range 0.88.9 1.19.5 1.110.7 5.431.5 6.335.8 4.724.7

Mean 3.35 3.56 3.55 14.09 18.93 12.95

S.D. 1.74 1.86 1.83 6.17 7.70 4.33

COD (mg l1) Range 2.619.3 6.420.0 6.221.5 11.858.3 8.176.3 8.239.4

Mean 10.76 11.89 11.88 29.33 38.68 27.52

S.D. 4.10 3.36 3.17 10.93 14.23 7.50


6/20

Table 2 (Continued)

Parameters Site 1 Site 2 Site 3 Site 4 Site 5 Site 6

Cl (mg l1) Range 0.75.8 1.06.3 0.210.3 2.017.3 3.323.7 3.321.3

Mean 2.99 3.86 4.45 7.35 11.26 11.74

S.D. 1.21 1.47 2.05 3.51 5.04 5.41

F (mg l1) Range 0.10.9 0.090.91 0.170.97 0.170.91 0.191.05 0.161.01

Mean 0.39 0.41 0.42 0.47 0.53 0.52

S.D. 0.16 0.18 0.17 0.17 0.20 0.20

PO4(mgl1) Range 0.020.15 0.020.34 0.0310.0 0.150.55 0.091.19 0.12.07

Mean 0.06 0.09 0.37 0.23 0.44 0.49

S.D. 0.04 0.06 1.65 0.09 0.30 0.43

SO4(mgl1) Range 3.511.8 2.912.5 3.813.8 4.153.6 4.218.8 6.522.5

Mean 7.65 8.53 9.10 12.94 13.48 15.13

S.D. 2.29 2.42 2.58 7.43 3.57 4.01

K (mg l1) Range 2.210.2 3.09.2 3.39.1 3.222.2 3.626.7 3.911.8

Mean 4.18 4.89 5.08 7.01 7.45 6.59

S.D. 1.42 1.62 1.45 3.89 3.88 2.05

Na (mg l1) Range 11.439.3 7.943.7 15.446.8 15.471.1 17.385.5 19.078.7

Mean 27.72 29.80 31.92 37.70 43.42 43.37

S.D. 6.56 8.01 7.55 10.39 12.61 12.90

Ca (mg l1) Range 17.659.7 16.062.4 14.962.7 5.167.7 21.373.6 16.080.0

Mean 35.92 35.62 36.12 42.10 45.06 45.00

S.D. 11.3 10.65 12.52 13.60 11.77 12.83

Mg (mg l1) Range 1.535.5 1.333.6 0.1641.6 1.441.3 1.636.2 1.640.6

Mean 16.29 19.09 19.26 18.91 19.51 19.39

S.D. 9.92 9.47 12.15 10.31 9.24 10.84

NH4N (mg l1) Range 0.00.26 0.00.42 0.040.51 0.124.57 0.091.66 0.051.37

Mean 0.05 0.14 0.19 0.62 0.51 0.34

S.D. 0.08 0.10 0.11 0.77 0.40 0.29

NO3N (mg l1) Range 0.020.72 0.030.84 0.00.85 0.041.50 0.072.23 0.112.14

Mean 0.16 0.16 0.17 0.50 0.83 0.85

S.D. 0.17 0.14 0.16 0.38 0.60 0.58

TKN (mg l1

) Range 0.044.35 1.554.25 0.163.84 2.544.85 2.947.95 2.457.25 Mean 2.84 2.86 2.59 3.72 4.15 3.95

S.D. 0.89 0.68 0.85 0.53 1.18 0.99

T. Coli (MPN/100ml) Range 201.42E06 277.0E + 04 331.7E + 05 409.7E + 09 5.4E+ 058.9E+ 10 1.6E+ 038.4E+ 0

Mean 4.7E + 04 1.1E + 04 8.2E + 03 9.7E + 08 2.8E + 09 5.1E + 05

S.D. 2.4E + 05 1.9E + 04 2.8E + 04 2.6E + 09 1.5E + 10 1.4E + 06

F. Coli (MPN/100ml) Range 201.4E + 06 277.0E + 04 331.7E + 05 409.7E + 09 5.4E+ 058.9E+ 10 1.6E+ 031.7E+ 0

Mean 4.7E + 07 1.1E + 04 8.1E + 03 9.7E + 08 2.7E + 09 9.4E + 05

S.D. 2.4E + 05 1.9E + 04 2.8E + 04 2.6E + 09 1.5E + 10 3.0E + 06


7/20

Cd (mg l1) Range 0.00.003 0.00.007 0.00.002 0.00.004 0.00.006 0.00.002

Mean 0.0003 0.001 0.0004 0.001 0.001 0.0003

S.D. 0.0006 0.001 0.0005 0.001 0.001 0.0005

Cr (mg l1) Range 0.00.048 0.00.054 0.00.019 0.00.014 0.00.048 0.00.049

Mean 0.006 0.004 0.003 0.003 0.006 0.007

S.D. 0.01 0.009 0.004 0.004 0.009 0.010

Fe (mg l1) Range 0.07517.064 0.05920.397 0.03214.32 0.079.736 0.1813.338 0.010.494

Mean 3.211 2.847 2.234 1.982 2.446 2.472

S.D. 3.346 3.754 3.120 2.299 2.714 2.917

Pb (mg l1) Range 0.00.107 0.00.114 0.00.077 0.00.097 0.00.124 0.00.142

Mean 0.024 0.021 0.017 0.019 0.021 0.022

S.D. 0.023 0.022 0.017 0.019 0.022 0.025

Cu (mg l1) Range 0.00.102 0.00.139 0.00.179 0.00.108 0.00.369 0.00.136

Mean 0.018 0.015 0.014 0.016 0.034 0.016

S.D. 0.027 0.027 0.035 0.023 0.071 0.031

Mn (mg l1) Range 0.0030.601 0.0041.50 0.0030.432 0.0140.507 0.0020.426 0.0030.318

Mean 0.120 0.136 0.075 0.082 0.097 0.111

S.D. 0.111 0.247 0.085 0.098 0.080 0.076

Zn (mg l1) Range 0.0020.325 0.0040.430 0.00.166 0.00.086 0.0180.289 0.0010.455

Mean 0.093 0.075 0.042 0.030 0.086 0.073

S.D. 0.088 0.100 0.039 0.021 0.056 0.086

Ni (mg l1) Range 0.0040.057 0.0040.082 0.0020.038 0.0040.033 0.0070.058 0.0060.035

Mean 0.016 0.016 0.014 0.014 0.016 0.016

S.D. 0.009 0.012 0.008 0.006 0.008 0.008


8/20


In this case study, three groups both for temporal (three

seasons) and spatial (three sampling regions) evaluations

have been selected. DA was applied to raw data by using the

standard, forward stepwise and backward stepwise modes

to construct DFs to evaluate both the spatial and temporal

variations in river water quality. The sites (spatial) and the

seasons (temporal) were the grouping (dependent) variables,while all the measured parameters constituted the indepen-

dent variables.

2.3.3. Principal component analysis/factor analysis

Principal component analysis provides information on the

most meaningful parameters, which describe whole data set

rendering data reduction with minimum loss of original in-

formation[3,12,15]. It is a powerful technique for pattern

recognition that attempts to explain the variance of a large set

of inter-correlated variables and transforming into a smaller

set of independent (uncorrelated) variables (principal com-

ponents). The principal component (PC) is expressed as:

zij= ai1x1j+ ai2x2j+ ai3x3j+ + aimxmj (2)

wherea is the component loading, z the component score,x

the measured value of a variable, i the component number,j

the sample number, andmthe total number of variables.

Factor analysis attempts to extract a lower dimensional

linear structure from the data set. It further reduces the con-

tribution of less significant variables obtained from PCA and

the new group of variables known as varifactors (VFs) is ex-

tracted through rotating the axis defined by PCA. In FA, the

basic concept is expressed in Eq.(3),

zji = af1f1i + af2f2i + af3f3i + + afmfmi + efi (3)

wherezis the measured value of a variable,athe factor load-

ing, f the factor score, e the residual term accounting for

errors or other sources of variation, i the sample number, j

the variable number, andmthe total number of factors.

The two methods, PCAand FA, in principle, are expressed

as similar equations, however, the difference lies in the fact

that in earlier one, the PC is expressed as a linear combination

of measured variables, while, in case of FA, measured vari-

able is expressed as a combination of factors and the equation

contains the residual term and thus, a VF can include unob-servable, hypothetical, latent variables[3,10,12,15].

Principal component analysis/factor analysis was per-

formed on correlation matrix of rearranged data (all observa-

tions for each group of sites), so that it explains the structure

of the underlying data set. The correlation coefficient matrix

measures how well the variance of each constituent can be

explained by relationship with each of the others[8].PCA

of the normalized variables (water quality data set) was per-

formed to extract significant PCs and to further reduce the

contribution of variables with minor significance; these PCs

were subjected to varimax rotation (raw) generating VFs.

2.3.4. Receptor modeling (APCS-MLR)

Receptor modeling approach based on multi-linear regres-

sion of the absolute principalcomponent score (APCS-MLR)

is a widely employed statistical technique for source appor-

tionment of environmental contaminants in air pollution stud-

ies [5,6,16,17]. It has recently been applied to water pollution

source apportionment also[9].It is based on the assumptionthat the total concentration of each contaminant is made up

of the linear sum of elemental contributions from each of the

jpollution source components collected at the receptor site,

Zjk =

p

j=1

wijpjk (4)

where zjk is the normalized concentration of contaminant

(variable),j the number of pollution sources, wij the factor

loadings, the coefficient matrix of the components relating

the pollution sources to their elemental concentrations; and

pjkthe factor scores, the value of thejth sources components

on observationkin Eq.(4).Bothwijandpjkare dimension-less.

Since,zjkin Eq. (4) is normalized valueof variables, it can-

not be used directly for computation of quantitative source

contributions, the normalized factor scores determined in

Eq. (4) were converted to un-normalized APCS following

the method reported elsewhere[18].The contribution from

each factor was then estimated by multiple linear regression

(MLR), using the APCS values as the independent variables

and the measured concentration of the particular contaminant

as the dependent variable, as

Mjk = ai0 +

p

j=1

Aij(APCS)jk (5)

whereMjkis the contaminants concentration; ai0the average

contribution of the jth contaminant from sources not deter-

mined by PCA/FA,Aijthe linear regression coefficient for the

ith contaminant and thejth factor, and (APCS)jkthe absolute

factor score for thejth factor with thekth measurement. The

values forMjk, ai0andAijhave the dimensions of the original

concentration measurements.

After determination of the number and identity of pos-

sible sources influencing the river water quality in three

different catchments regions (UC, MC and LC) by using

PCA/FA (Table 7), source contributions were computed

through APCS-MLR technique. Quantitative contributions

from each source for individual parameter/contaminant were

compared with their measured values.

3. Results and discussion

3.1. Spatial similarity and site grouping

Cluster analysis was applied to detect spatial similarity

for grouping of sites under the monitoring network. It ren-


9/20


Fig. 2. Dendogram showing sampling site clusters on the Gomti river.

dered a dendrogram (Fig. 2),grouping all the eight sampling

sites on the river into three statistically significant clusters

at (Dlink/Dmax) 100 < 70. The clustering procedure gener-ated three groups of sites in a very convincing way, as the

sites in these groups have similar characteristic features and

natural background source types. Cluster 1 (sites 13), clus-

ter 2 (sites 46) and cluster 3 (sites 7 and 8) correspond to

a relatively low pollution, very high pollution and moderate

pollution regions, respectively. It implies that for rapid as-

sessment of water quality, only one site in each cluster may

serve as good in spatial assessment of the water quality as

the whole network. It is evident that the CA technique is

useful in offering reliable classification of surface waters in

the whole region and will make possible to design a future

spatial sampling strategy in an optimal manner. Thus, thenumber of sampling sites in the monitoring network will be

reduced, hence cost without loosing any significance of the

outcome. There are other reports[5,6,9,10,12], where simi-

lar approach has successfully been applied in water quality

programs.

3.2. Spatial and temporal variations in river water

quality

The temporal variations of the river water quality parame-

ters (Table 2) were evaluated through season-parameter cor-

relation matrix, which showed that all the measured param-

eters were found significantly (p < 0.05) correlated with the

season, except TSS, BOD, COD, F, SO4,KandTKN.Among

these, temperature exhibited highest correlation (Spearman)

coefficient (R = 0.69). Other parameters exhibiting high cor-

relation with season were total alkalinity (R =0.55), total

hardness (R =0.52), Na (R =0.52) and EC (R =0.51).

The season-correlated parameters can be taken as represent-

ing the major source of temporal variations in water qual-

ity. In view of the source types in the river catchments,

these correlations can be explained on the basis of seasonal

features in the monitoring region. Wide seasonal variations

in temperature and river discharge round the year can be

attributed to the high seasonality in various water quality

parameters.

Temporal variations in water quality were further eval-

uated through DA. Temporal DA was performed on raw

data after dividing the whole data set into three seasonal

groups (winter, summer and monsoon). Discriminant func-

tions (DFs) and classification matrices (CMs) obtained fromthe standard, forward stepwise and backward stepwise modes

of DA are shown in Tables 3 and 4. In forward stepwise

mode, variables are included step-by-step beginning with the

more significant until no significant changes are obtained,

whereas, in backward stepwise mode, variables are removed

step-by-step beginning with the less significant until no sig-

nificant changes are obtained. The standard DA mode, con-

structed DFs including 31 parameters are shown in Table 3.

The coefficients for the total coliform bacteria group were

zero. Both the standard and forward stepwise mode DFs

using 31 and 21 discriminant variables, respectively, ren-

dered the corresponding CMs assigning 97% cases correctly

(Tables 3 and 4). However, in backward stepwise mode DAgave CMs with 94% correct assignations using only five dis-

criminant parameters (Tables 3 and 4) with a little different

match for each season compared with the forward stepwise

mode. Forward stepwise DA showed that temperature, alka-

linity, Cl, Na and K are followed by a second group of param-

eter formed by discharge, pH, TDS, T-Hard, Ca-Hard, DO,

BOD, F, SO4, PO4, NO3N, TKN, Cr, Pb, Cu and Ni but less

significant as could be seen from the difference in percentage

of correct assignations between the backward and forward

DA modes (Tables 3 and 4).Further, a much less significant

third group of remaining 10 parameters is evident from the

standard mode DA assignations. Thus, the temporal DA re-sults suggest that temperature, alkalinity, Cl, Na and K are

the most significant parameters to discriminate between the

three different seasons, which means that these five param-

eters account for most of the expected temporal variations

in the river water quality (Table 3). This also suggests that

the anthropogenic pollution, which is the major river pol-

lution problem, mainly due to discharge of wastewater into

the river does not discriminate between the seasons and is a

regular source throughout the year. The trend obtained was

also supported by the analysis of the results on the raw data

set.

As identified by DA, box and whisker plots of the selected

parameters showing seasonal trends are given in Fig. 3. The

variation of temperatureshows a clear-cutseasonal effect. To-

tal alkalinity during three seasons showed increasing trend

in summers over winters and decline during monsoon and

this may be attributed to enhanced weathering process dur-

ing the summers in the catchments, which during monsoon

months declines due to excessive dilution. Chloride, sodium

andpotassium ions in the river water showedsimilar trends of

increase during summers over winters and declining during

the monsoon.

Spatial DA was performed with the same raw data set

comprised of 31 parameters after grouping into three major


10/20


Table 3

Classification functions (Eq.(1)) for discriminant analysis of temporal variation in Gomti river water

Parameters Standard mode Forward stepwise mode Backward stepwise mode

Winter

coefficientaSummer

coefficientaMonsoon

coefficientaWinter

coefficientaSummer

coefficientaMonsoon

coefficientaWinter

coefficientaSummer

coefficientaMonsoon

coefficienta

Discharge 0.236 0.194 0.347 0.243 0.196 0.352

Temperature 0.018 1.510 1.319 0.351 1.141 0.980 3.349 4.645 4.444PH 174.33 172.77 168.36 166.47 164.70 160.69

EC 0.140 0.159 0.150

TS 0.268 0.272 0.231

TDS 0.532 0.502 0.457 0.276 0.247 0.239

TSS 0.616 0.636 0.583

T-Alk 0.947 1.091 0.883 0.958 1.113 0.901 0.549 0.664 0.480

T-Hard 0.175 0.195 0.188 0.122 0.137 0.131

Ca-Hard 0.262 0.468 0.139 0.394 0.253 0.541

DO 0.880 0.782 0.796 0.620 0.495 0.525

BOD 3.539 3.718 3.446 3.857 3.889 3.684

COD 0.054 0.175 0.103

Cl 4.312 3.964 3.953 3.823 3.443 3.463 0.721 0.392 0.336

F 12.748 17.463 14.319 7.539 11.758 9.208

SO4 0.580 0.329 0.476 0.611 0.308 0.450

PO4 1.560 1.621 1.341 1.438 1.520 1.230K 1.515 0.849 0.918 1.841 1.273 1.283 0.045 0.751 0.446

Na 2.009 2.541 2.275 1.408 1.898 1.664 0.536 1.002 0.808

NH4N 14.100 13.524 12.996

NO3N 0.744 0.424 1.282 2.838 3.944 4.958

TKN 10.395 9.409 10.695 10.047 8.970 10.301

T. Coli 0.000 0.000 0.000

Cd 2724.57 3147.42 3049.37

Cr 251.54 260.71 127.294 408.08 422.14 269.94

Fe 2.293 2.372 2.063

Pb 225.93 217.66 169.348 266.83 259.30 220.02

Cu 142.91 118.11 118.847 230.23 210.45 204.71

Mn 3.657 1.062 5.869

Zn 26.81 30.240 26.118

Ni 1207.24 1210.55 1003.92 1288.99 1313.35 1167.95

Constant 877.165 914.368 838.95 825.301 858.663 788.357 82.364 129.214 92.562a Discriminant function coefficient for winter, summer and monsoon seasons correspond towijas defined in Eq.(1).

Table 4

Classificationmatrixfor discriminantanalysisof temporalvariationin Gomti

river water

Monitoring seasons % Correct Season assigned by DA

Winter Summer Monsoon

Standard DA mode

Winter 96.9 93 2 1

Summer 96.9 3 93 0

Monsoon 96.9 0 3 93

Total 96.9 96 98 94

Forward stepwise DA mode

Winter 95.8 92 3 1

Summer 96.9 3 93 0

Monsoon 96.9 0 3 93

Total 96.5 95 99 94

Backward stepwise DA mode

Winter 93.8 90 5 1

Summer 94.8 5 91 0

Monsoon 93.8 0 6 90

Total 94.1 95 102 91

classes of UC, MC and LC as obtained through CA. The site

(clustered) was the grouping (dependent) variable, while all

the measured parameters constituted the independent vari-

ables. Discriminant functions and classification matrices ob-

tained from the standard, forward stepwise and backward

stepwise modes of DA are shown in Tables 5 and 6. Sim-

ilar to the temporal DA, the standard DA mode constructs

DFs including 31 parameters (Table 5),the coliform bacte-

ria group coefficients are zero again. Both the standard and

forward stepwise mode DFs using 31 and 22 discriminant

parameters, respectively, rendered the corresponding CMs

assigning more than 97% cases correctly (Tables 5 and 6).

The backward stepwise mode DA gave CMs with slightly

less than 97% correct assignations using only 10 discriminant

parameters (Tables 5 and 6).Backward stepwise DA shows

that discharge, pH, BOD, Cl, F, SO4, NH4N, NO3N, TKN

and Zn are the discriminating parameters in space. The cor-

rect assignations (97%) by DA for three different site clusters

(UC, MC and LC) further confirmed the adequacy of DA and

the grouping pattern coincides with our previous spatial CA.

Both CA and DA predict important differences in water qual-


11/20


ity due to impact from Lucknow city. DA shows that there

are significant differences between these three regions (UC,

MC and LC), which are expressed in terms of 10 discrimi-

nating parameters. Hence, DA rendered a considerable data

reduction.

Box and whisker plots of some selected discriminating pa-

rameters identified by spatial DA (backward step mode) were

constructed to evaluate different patterns associated with spa-

tial variations in river water quality (Fig. 4).Mean discharge

of the river shows a steady increase with the river course. It

Fig. 3. Temporal variations: (a) temperature; (b) total alkalinity; (c) chloride; (d) sodium; (e) potassium in Gomti river water.


12/20


Table 5

Classification functions (Eq.(1)) for discriminant analysis of spatial variations in Gomti river water

Parameters Standard DA mode Forward stepwise DA mode Backward stepwise DA mode

UCa

coefficientbMCc

coefficient

LCd

coefficient

UC

coefficient

MC

coefficient

LC

coefficient

UC

coefficient

MC

coefficient

LC

coefficient

Discharge 0.680 0.971 1.245 0.630 0.911 1.188 0.294 0.519 0.764

Temperature 0.232 0.032 0.116 0.296 0.101 0.198EC 0.107 0.112 0.102

PH 173.59 167.799 173.171 169.112 163.578 168.991 158.038 152.04 158.002

TS 0.124 0.146 0.117 0.243 0.224 0.237

TDS 0.403 0.407 0.392

TSS 0.438 0.444 0.393 0.045 0.045 0.012

T-Alk 0.453 0.438 0.488 0.409 0.402 0.440

T-Hard 0.152 0.150 0.151

Ca-Hard 0.112 0.152 0.112 0.080 0.120 0.079

DO 0.430 0.639 0.453 0.246 0.452 0.277

BOD 1.237 1.988 1.414 1.703 2.413 2.011 1.781 2.367 2.022

COD 0.314 0.293 0.395

Cl 3.540 3.125 2.458 2.847 2.413 1.805 1.408 0.912 0.278

F 5.880 4.311 12.407 19.104 17.433 24.854 15.044 11.116 19.842

SO4 1.733 1.278 0.964

PO4 1.353 1.500 1.988 1.484 1.633 2.103 1.202 1.338 1.792K 1.463 1.355 01.297

Na 1.565 1.458 1.505 1.325 1.209 1.244

NH4N 5.033 8.741 6.304 5.488 9.058 7.027 2.951 7.299 4.725

NO3N 8.698 5.159 9.651 8.847 4.935 9.753 14.644 12.619 17.054

TKN 8.032 9.097 6.321 8.040 9.066 6.288 6.091 6.530 4.120

T. Coli 0.000 0.000 0.000

Cd 2261.17 2564.128 3045.26 953.68 1279.52 1751.65

Cr 97.198 26.087 108.666

Fe 0.698 0.857 1.730 0.297 0.067 0.770

Pb 167.217 154.37 156.351

Cu 119.202 155.499 167.689 82.956 122.64 132.41

Mn 7.490 10.66 21.85 16.704 13.678 1.372

Zn 41.558 33.678 46.851 42.200 34.681 47.987 43.615 39.167 54.37

Ni 260.491 316.299 321.337 458.318 525.86 495.55

Constant 819.742 792.008 849.818 789.624 763.29 821.70 674.153 645.499 699.582a Upper catchments includes sites 13.b Coefficients for different catchments correspond to wijas defined in Eq.(1).c Middle catchments includes sites 46.d Lower catchments includes sites 7 and 8.

is due to the inputs from its tributaries and 45 wastewater

drains emptying into the river directly. Trends for pH, BOD,

NH4N, NO3N and TKN suggest for high load of dissolved

organic matter in the MC region added by 28 wastewater

drains from Lucknow town pouring about 450 mld of raw

wastewater in to the river directly leading to anaerobic con-

ditions which results in formation of ammonia and organic

acids. Hydrolysis of these acidic materials causes a decrease

of water pH in this region. Similar trends of spatial variations

observed for BOD, NH4N and TKN suggest vast difference

in pollution load and sources in three catchments regions of

the river. Chloride and zinc show increasing trends from UC

to LC and this reflects additive nature of the two relatively

conservative constituents. Similar trends of spatial variations

observed for BOD, NH4N and TKN suggest vast difference

in pollution load and sources in three catchments regions of

the river. An increase in F level in river water at MC sug-

gests its origin in municipal and industrial wastewater in this

zone.

3.3. Data structure determination and source

identification

Principal component analysis/factor analysis was applied

to the normalized data sets (26 variables) separately for

the three different spatial regions viz. UC, MC and LC,

as delineated by CA technique, to compare the composi-

tional patterns between the analyzed water samples and to

identify the factors that influence each one. The input data

matrices (variables cases) for PCA/FA were [26 108]

for UC and MC, and [26 72] for LC regions. PCA of

the three data sets evolved seven PCs each for first two

regions (UC and MC) and six PCs for last region (LC)

with eigenvalue >1, explaining 74.4, 73.6 and 81.4% of

the total variance in respective water quality data sets.

Equal numbers of VFs were obtained for three regions

through FA performed on the PCs. Corresponding VFs,

variables loadings and variance explained are presented in

Table 7.


13/20


For the dataset pertaining to UC, among the seven VFs,

the VF1 explaining 31.6% of total variance has strong neg-

ative loadings (>0.70) on Cd, Cr, Fe, Mn and Ni and mod-

erate negative loading on Pb. Thus, it represents the metal

group. VF2 explaining 11.4% of the total variance has strong

positive loadings on organic pollution parameters, BOD and

COD, and moderate loading on TKN, thus, basically repre-

sents the organic pollution group. VF3 has strong positive

loadings on TS and TDS and moderate positive loadings on

Na and SO4, which can be interpreted as a mineral compo-

nent of the river water. This clustering of variables points to

a common origin for these minerals, likely from dissolution

of limestone and gypsum soils in the river catchments [3].

It may be noted that gypsum is widely used as soil modi-

Fig. 4. Spatial variations: (a) discharge; (b) pH; (c) BOD; (d) chloride; (e) fluoride; (f) phosphate; (g) ammonical nitrogen; (h) total kjeldahl nitrogen; (i) zinc;

(j) nitrate-nitrogen in Gomti river water.


14/20


Fig. 4. (Continued).

Table 6

Classification matrix for discriminant analysis of spatial variations in Gomti

river water

Monitoring regions % Correct Regions assigned by DA

UCa MCb LCc

Standard DA mode

UCa 99.1 107 1 0

MCb 96.3 2 104 2

LCc 95.8 2 1 69

Total 97.2 111 106 71

Forward stepwise DA mode

UCa 99.1 107 1 0

MCb 97.2 1 105 2

LCc 95.8 2 1 69

Total 97.6 110 107 71

Backward stepwise DA mode

UCa 99.1 107 1 0

MCb 95.4 3 103 2

LCc 95.8 2 1 69

Total 96.9 112 105 71

a Upper catchments includes sites (13).b Middle catchments includes sites (36).c Lower catchments includes sites (7 and 8).

fier in the river catchments. VF4 explaining relatively lower

variance (6.45%) has strong negative loadings on alkalin-

ity, hardness and DO and moderate negative loadings on Na

and Cl ions. This VF represents natural sources of these pa-

rameters in catchments from soil weathering and subsequent

run-off. VF5 has strong loading on Cu and moderate loading

on NO3N. This indicates influence of industrial activities

VF6 has strong negative loading on K and moderate nega-

tive loading on Cl. This VF also represents natural source

as these ions have origin in the catchments soils. VF7 has

strong positive loading on NH4N. This represents influenceof agricultural runoff from the soil as nitrogenous fertilisers

are extensively used in this region.

For the data set pertaining to water quality in MC region,

among the seven VFs, VF1 explaining 27% of total variance

has strong positive loadings on TS and TDS and moderate

loadings on Na and K. This basically represents the solids

group. This clustering point to common sources of natural

processes of dissolution of soil constituents mainly carbon-

ates. VF2, explaining about 18% of the variance, has strong

positive loadings on metals (Cr, Fe, Mn and Ni), whereas, a

moderate loading on Pb, Cu and Zn. Thus, it basically repre-


15/20


Table 7

Loadings of experimental variables (26) on significant principal components for (a) UC dataset, (b) MC dataset, (c) LC dataset

Variables VF1 VF2 VF3 VF4 VF5 VF6 VF7

UC dataset (seven significant principal components)

EC 0.123 0.026 0.257 0.834 0.098 0.073 0.047

TS 0.028 0.017 0.876 0.184 0.089 0.004 0.055

TDS 0.157 0.008 0.870 0.223 0.044 0.128 0.073

T-Alk 0.374 0.053 0.257 0.777 0.160 0.244 0.013

T-Hard 0.182 0.093 0.092 0.774 0.163 0.010 0.047

DO 0.246 0.344 0.004 0.812 0.048 0.120 0.032

BOD 0.028 0.891 0.097 0.103 0.058 0.001 0.091

COD 0.076 0.868 0.010 0.106 0.097 0.120 0.182

Cl 0.235 0.033 0.140 0.599 0.010 0.593 0.092

F 0.079 0.314 0.154 0.017 0.492 0.398 0.282

PO4 0.031 0.037 0.263 0.222 0.180 0.349 0.018

SO4 0.096 0.177 0.617 0.091 0.399 0.143 0.037

K 0.204 0.004 0.148 0.022 0.127 0.844 0.028

Na 0.336 0.008 0.515 0.593 0.075 0.297 0.060

NH4N 0.039 0.103 0.031 0.045 0.022 0.041 0.868

NO3N 0.267 0.189 0.051 0.420 0.657 0.152 0.061

TKN 0.144 0.544 0.120 0.006 0.286 0.326 0.198

T. Coli 0.009

0.177

0.316 0.211 0.134 0.155

0.481Cd 0.721 0.160 0.024 0.084 0.064 0.010 0.181

Cr 0.832 0.028 0.020 0.198 0.128 0.029 0.139

Fe 0.825 0.028 0.119 0.378 0.248 0.149 0.001

Pb 0.674 0.193 0.152 0.010 0.494 0.212 0.124

Cu 0.305 0.072 0.060 0.135 0.824 0.042 0.016

Mn 0.895 0.016 0.181 0.181 0.145 0.032 0.056

Zn 0.414 0.419 0.280 0.218 0.320 0.020 0.301

Ni 0.908 0.025 0.002 0.247 0.002 0.105 0.014

Eigenvalue 8.21 2.97 2.40 1.68 1.62 1.31 1.15

% Total variance 31.58 11.40 9.23 6.45 6.25 5.04 4.42

Cumulative % variance 31.58 42.98 52.21 58.66 64.91 69.95 74.37

MC dataset (seven significant principal components)

EC 0.357 0.187 0.080 0.329 0.041 0.802 0.081

TS 0.876 0.084 0.084 0.028 0.024 0.191 0.050

TDS 0.787 0.092 0.150 0.211 0.002 0.409 0.067T-Alk 0.425 0.209 0.034 0.200 0.033 0.806 0.079

T-Hard 0.070 0.058 0.015 0.027 0.108 0.900 0.076

DO 0.100 0.031 0.725 0.257 0.176 0.032 0.236

BOD 0.265 0.068 0.893 0.005 0.095 0.027 0.113

COD 0.146 0.108 0.892 0.158 0.145 0.082 0.023

Cl 0.409 0.046 0.091 0.609 0.163 0.523 0.108

F 0.229 0.072 0.123 0.265 0.689 0.145 0.129

PO4 0.268 0.050 0.162 0.607 0.090 0.186 0.084

SO4 0.095 0.137 0.003 0.470 0.142 0.272 0.271

K 0.512 0.195 0.407 0.329 0.306 0.008 0.164

Na 0.630 0.216 0.099 0.347 0.0001 0.478 0.125

NH4N 0.020 0.160 0.011 0.012 0.066 0.003 0.859

NO3N 0.063 0.092 0.159 0.821 0.101 0.041 0.040

TKN 0.036 0.093 0.168 0.395 0.620 0.037 0.109

T. Coli 0.090 0.016 0.297 0.241 0.154 0.145 0.358

Cd 0.123 0.055 0.324 0.291 0.258 0.236 0.234

Cr 0.161 0.754 0.043 0.179 0.010 0.120 0.021

Fe 0.173 0.820 0.065 0.125 0.002 0.425 0.004

Pb 0.142 0.614 0.004 0.022 0.546 0.051 0.072

Cu 0.036 0.685 0.174 0.117 0.441 0.039 0.008

Mn 0.011 0.849 0.186 0.021 0.021 0.249 0.045

Zn 0.153 0.624 0.323 0.229 0.007 0.062 0.135

Ni 0.091 0.850 0.055 0.295 0.015 0.146 0.020

Eigenvalue 6.99 4.72 2.19 1.68 1.40 1.08 1.06

% Total variance 26.89 18.17 8.41 6.47 5.40 4.14 4.08

Cumulative % variance 26.89 45.06 53.47 59.94 65.34 69.48 73.56


16/20


Table 7 (Continued)

Variables VF1 VF2 VF3 VF4 VF5 VF6 VF7

LC dataset (six significant principal components)

EC 0.636 0.424 0.019 0.376 0.446 0.101

TS 0.161 0.184 0.010 0.894 0.043 0.012

TDS 0.408 0.236 0.023 0.823 0.048 0.023

T-Alk 0.600 0.458 0.050 0.434 0.435 0.075

T-Hard 0.324 0.366 0.064 0.364 0.571 0.035DO 0.120 0.481 0.314 0.194 0.631 0.006

BOD 0.076 0.040 0.885 0.074 0.039 0.140

COD 0.122 0.071 0.863 0.177 0.175 0.007

Cl 0.821 0.217 0.037 0.276 0.283 0.110

F 0.646 0.183 0.427 0.131 0.042 0.235

PO4 0.857 0.038 0.014 0.195 0.194 0.108

SO4 0.452 0.490 0.262 0.074 0.327 0.262

K 0.833 .098 0.175 0.045 0.055 0.178

Na 0.649 0.359 0.019 0.417 0.310 0.080

NH4N 0.090 0.054 0.148 0.217 0.743 0.244

NO3N 0.160 0.182 0.091 0.020 0.200 0.724

TKN 0.182 0.300 0.758 0.168 0.137 0.175

T. Coli 0.111 0.432 0.473 0.329 0.075 0.312

Cd 0.008 0.710 0.105 0.209 0.027 0.457

Cr 0.127 0.887 0.152 0.193 0.057 0.274Fe 0.309 0.821 0.108 0.169 0.189 0.330

Pb 0.155 0.630 0.398 0.003 0.026 0.558

Cu 0.053 0.414 0.247 0.048 0.003 0.784

Mn 0.249 0.844 0.098 0.142 0.152 0.261

Zn 0.124 0.802 0.042 0.132 0.143 0.122

Ni 0.242 0.898 0.058 0.086 0.055 0.203

Eigenvalue 11.03 3.97 2.01 1.71 1.43 1.01

% Total variance 42.44 15.28 7.74 6.57 5.50 3.87

Cumulative % variance 42.44 57.73 65.47 72.04 77.54 81.41

Bold-faced values represent strong loadings.

sents a toxic metals group. VF3 has strong positive loadingboth on BOD and COD, whereas, a negative strong loading

on DO. It is, thus, a group of purely organic pollution indica-

tor parameters. This VF represents anthropogenic pollution

sources and can be explained that high levels of dissolved or-

ganic matterconsume large amounts of oxygen, which under-

goes anaerobic fermentation processes leading to formation

of ammonia and organic acids. Hydrolysis of these acidic

materials causes a decrease of water pH values [3]. VF4

has strong positive loading on NO3N and moderate positive

loading on Cl and PO4. Thus, it represents the nutrients group

of pollutants which points to some source of wastewater and

runoff. VF5 has moderate positive loading on F, while mod-

erate negative loadings on TKN and Pb. This region (MC) isknown for high fluoride in soils and groundwater. VF6 has

strong positive loadings on EC, alkalinity and hardness and

a moderate positive loading on Cl. This basically represents

the salts group. VF7 has strong negative loading on NH4N

alone, and thus indicating the influence of domestic waste

and agricultural runoff.

Lastly, for the data set representing the LC region, among

total six significant VFs, the first one (VF1) explaining about

42.5% of the total variance has strong positive loadings on

K, Cl, and PO4 and moderate positive loadings on EC, al-

kalinity, F and Na. It basically represents the ionic group

of salts. The phosphate has its origin in soils due to use ofphosphatic fertilisers in this region. VF2 explaining more

than 15% of total variance has strong negative loadings on

Cd, Cr, Fe, Mn, Zn and Ni. It also has moderate negative

loading on Pb. Thus, it basically represents the toxic met-

als group. VF3, a group of organic pollution indicators, has

strong positive loadings on BOD, COD and TKN. VF4 has

strong positive loadings on TS and TDS. This factor loaded

with solids indicates towards their origin in run-off from the

fields with high load of solids and waste disposal activi-

ties. VF5 has strong negative loading on NH4N, whereas,

a moderate negative loadings on hardness and DO. VF6

has strong positive loadings on NO3N and Cu. It also has

moderate positive loading on Pb and represents industrialwaste.

From the PCA/FA loadings, it is evident that for all the

three regions, major groups of parameters (strong loadings)

emerged are the trace metals group (leaching from soil, and

industrial waste disposal sites), organic pollution group (rep-

resenting influencesfrom pointsources suchas municipaland

industrial effluents), nutrient parameters group (represent-

ing influences from non-point sources such as agricultural

runoff and atmospheric deposition), alkalinity, hardness, EC

and solids group (soil leaching/erosion followed by runoff

process).


17/20


Table 8

Source contribution to Gomti river water in the (a) UC region, (b) MC region, (c) LC region

Parameters Source type Observed

mean (O)

Measured

mean (M)

Ratio

(O/M)

R2

SM S1 S2 S3 S4 S5 S6 S7

UC regiona

EC (S cm1) 30.26 5.88 105.37 251.23 392.74 368.49 1.07 0.80

TS (mg l1

) 44.22 2.32 211.35 32.64 290.54 285.4 1.02 0.81TDS (mg l1) 52.22 173.42 32.67 258.31 243.6 1.06 0.86

T-Alk (mg l1) 23.68 67.62 150.47 241.77 194.44 1.24 0.90

T-Hard (mg l1) 5.37 28.37 175.54 209.24 165.6 1.26 0.68

DO (mg l1) 4.28 6.22 10.50 7.26 1.45 0.86

BOD (mg l1) 4.23 0.09 4.32 3.49 1.24 0.83

COD (mg l1) 3.08 0.74 10.08 13.90 11.51 1.21 0.83

Cl (mg l1) 0.82 3.46 4.28 3.77 1.14 0.80

F (mg l1) 0.38 0.02 0.06 0.46 0.41 1.14 0.61

PO4(mg l1) 0.08 0.07 0.03 0.17 0.17 1.00 0.28

SO4(mg l1) 0.78 8.42 0.54 0.09 9.80 8.43 1.17 0.61

K (mg l1) 1.90 1.33 0.48 3.71 3.50 1.06 0.79

Na (mg l1) 18.64 15.90 0.82 35.36 29.81 1.19 0.83

NH4N (mg l1) 0.15 0.15 0.13 1.15 0.77

NO3N (mg l1) 0.12 0.05 0.01 0.18 0.17 1.08 0.75

TKN (mg l1

) 1.00 1.43 0.57 0.02 0.19 3.21 2.76 1.16 0.56T. Coli (MPN/100 ml) 3129 78206 81335 22365 3.64 0.45

Fe (mg l1) 1.07 2.21 3.27 2.76 1.18 0.92

Pb (mg l1) 0.033 0.013 0.046 0.02 2.21 0.82

Cu (mg l1) 0.016 0.024 0.041 0.016 2.54 0.80

Mn (mg l1) 0.19 0.191 0.11 1.73 0.89

Zn (mg l1) 0.011 0.038 0.047 0.017 0.114 0.07 1.63 0.67

Ni (mg l1) 0.025 0.025 0.015 1.62 0.90

Cr (mg l1) 0.006 0.001 0.007 0.005 1.60 0.77

MC regionb

EC (S cm1) 80.25 165.97 5.83 18.34 208.33 478.71 440.70 1.09 0.93

TS (mg l1) 55.25 227.45 7.00 3.43 0.88 27.71 1.76 323.48 321.89 1.00 0.82

TDS (mg l1) 37.36 179.52 5.34 5.78 52.07 280.06 275.31 1.02 0.87

T-Alk (mg l1) 17.12 101.57 5.74 107.45 231.88 190.41 1.22 0.92

T-Hard (mg l1) 47.43 18.76 135.40 2.81 204.4 190.41 1.07 0.84

DO (mg l1) 2.19 0.53 0.26 0.19 0.26 3.42 2.90 1.18 0.69

BOD (mg l1) 8.71 0.72 4.60 0.02 0.79 0.49 0.50 15.82 15.3 1.03 0.90

COD (mg l1) 6.47 8.71 2.08 8.33 1.13 2.20 2.73 0.19 31.84 31.84 1.00 0.88

Cl (mg l1) 9.42 2.91 12.32 10.12 1.22 0.86

F (mg l1) 0.3 0.22 0.08 0.59 0.51 1.16 0.66

PO4(mg l1) 0.33 0.07 0.11 0.51 0.39 1.32 0.52

SO4(mg l1) 5.17 2.49 1.47 0.94 3.97 0.97 15.01 13.85 1.08 0.42

K (mg l1) 6.32 1.06 0.65 8.03 7.02 1.14 0.69

Na (mg l1) 34.80 2.49 13.28 50.57 41.50 1.22 0.82

NH4N (mg l1) 0.48 0.04 0.52 0.49 1.07 0.77

NO3N (mg l1) 0.16 0.17 0.08 0.07 0.27 0.06 0.81 0.73 1.12 0.73

TKN (mg l1) 2.59 0.17 0.14 0.12 0.22 0.73 3.97 3.94 1.01 0.59

T. Coli (MPN/100 ml) 3.8E+09 2.3E+08 2.0E+09 3.5E+09 9.5E+09 1.5E+10 0.63 0.33

Cd (mg l1) 0.0005 0.0001 0.0001 0.0001 0.0007 0.0003 2.49 0.69

Fe (mg l1) 2.40 1.71 4.12 2.27 1.81 0.90

Pb (mg l1) 0.005 0.02 0.014 0.001 0.039 0.021 1.87 0.71

Cu (mg l1) 0.035 0.002 0.001 0.005 0.043 0.022 1.96 0.71

Mn (mg l1) 0.052 0.114 0.166 0.097 1.71 0.82

Zn (mg l1) 0.029 0.064 0.016 0.009 0.118 0.063 1.85 0.59

Ni (mg l1) 0.006 0.004 0.011 0.02 0.016 1.29 0.84

Cr (mg l1) 0.008 0.001 0.001 0.009 0.005 1.72 0.64

LC regionc

EC (S cm1) 36.18 89.58 234.04 152.86 512.65 425.83 1.20 0.94

TS (mg l1) 57.56 11.01 1.85 271.28 341.69 318.44 1.07 0.86

TDS (mg l1) 9.05 27.11 4.12 241.61 7.80 0.54 290.23 270.68 1.07 0.90

T-Alk (mg l1) 46.15 147.68 81.55 275.38 206.01 1.34 0.96

T-Hard (mg l1) 27.24 135.64 117.12 280.00 183.76 1.52 0.70

DO (mg l1) 2.97 1.89 3.77 8.62 7.02 1.23 0.78


18/20


Table 8 (Continued)

Parameters Source type Observed

mean (O)

Measured

mean (M)

Ratio

(O/M)

R2

SM S1 S2 S3 S4 S5 S6 S7

BOD (mg l1) 5.31 1.02 6.33 5.31 1.19 0.82

COD (mg l1) 1.52 0.82 0.59 13.92 2.24 19.09 16.18 1.18 0.83

Cl (mg l1) 5.21 5.73 3.76 14.70 11.63 1.26 0.89

F (mgl1

) 0.39 0.15 0.54 0.43 1.25 0.71PO4 (mg l1) 0.01 0.13 0.10 0.25 0.23 1.06 0.82

SO4 (mg l1) 6.79 2.90 4.53 2.11 5.11 21.44 18.11 1.18 0.69

K (mgl1) 2.72 1.92 1.09 0.46 6.19 5.45 1.18 0.69

Na (mg l1) 11.58 24.92 13.45 49.94 40.74 1.23 0.83

NH4N (mg l1) 0.01 0.07 0.15 0.23 0.18 1.26 0.69

NO3N (mg l1) 0.18 0.07 0.08 0.09 0.41 0.38 1.10 0.63

TKN (mg l1) 0.62 0.31 1.70 0.62 0.05 3.30 2.87 1.15 0.78

T. Coli. (MPN/100 ml) 6937 18722 28576 22647 86880 72160 1.20 0.63

Cd (mg l1) 0.002 0.0003 0.0018 0.0008 2.38 0.77

Fe (mg l1) 2.18 5.17 0.39 7.73 4.62 1.67 0.96

Pb (mg l1) 0.027 0.01 0.006 0.043 0.027 1.57 0.89

Cu (mg l1) 0.011 0.011 0.027 0.049 0.033 1.49 0.85

Mn (mg l1) 0.051 0.19 0.014 0.005 0.261 0.156 1.67 0.90

Zn (mg l1) 0.041 0.194 0.008 0.244 0.138 1.76 0.71

Ni (mg l1

) 0.006 0.019 0.003 0.001 0.03 0.022 1.33 0.92Cr (mg l1) 0.001 0.012 0.003 0.001 0.017 0.008 2.2 0.94

a SM: miscellaneous sources; S1: industrial waste; S2: municipal and industrial waste; S3: soil runoff 1; S4: weathering and runoff; S5: agricultural runoff;

S6: soil runoff 2; S7: waste site runoff.b SM: miscellaneous sources; S1: municipal and industrial waste 1; S2: industrial waste; S3: municipal and industrial waste 2; S4: agricultural runoff; S5:

soil runoff 1; S6: soil runoff 2; S7: waste site runoff.c SM: miscellaneous sources; S1: agricultural runoff; S2: industrial waste 1; S3: municipal and industrial waste; S4: soil runoff; S5: waste site runoff; S6:

industrial waste 2.

The results from temporal PCA/FA suggested that most

of the variations in water quality are explained by the

soluble salts (natural), toxic metals (industrial), nutrients

(non-point) and organic pollutants (anthropogenic). Inthis study, FA did not result in much data reduction, as

we still need 1417 parameters (about 5565% of the

26 parameters) to explain 7482% of the data variance

in three catchments regions (Table 7ac). However, FA

served as a means to identify those parameters, which have

greatest contribution to temporal variation in the river water

quality and suggested possible sets of pollution sources

in each of the catchments regions of the river. Similar

approach based on PCA/FA for evaluation of temporal and

spatial variations in water quality has earlier been used

[3,9,10,12,15].

However, from the PCA/FA results, it may convincingly

be presumed that in all the three catchments regions under

study, the river pollution is mainly from soil weathering and

agricultural run-off, leaching from solid waste disposal sites

and domestic and industrial wastewater disposal. These find-

ings are also supported by the catchments source/activity in-

ventory.

3.4. Source apportionment

Results of receptor modeling through APCS-MLR

for source apportionment in three different catchments

regions of the Gomti river as the contributions of the possible

sources (identified through PCA/FA) in various water quality

parameters are presented inTable 8ac. As evident from the

correlation coefficients, the multiple regression exhibited

good adequacy between the measured and predicted values.Further, the ratio of mean observed and measured values of

almost all the water quality variables suggest goodness of

the receptor modeling approach to the source apportionment

of river water. MLR on APCS showed that except for a few,

most of the parameters are influenced by sources (PCs) iden-

tified as industrial waste disposal site leaching, domestic and

industrial waste disposal, agricultural runoff, weathering and

runoff from catchments. However, the miscellaneous uniden-

tified sources in all the three regions also contribute to the

river water pollution for most of the water quality variables.

In all the three regions (UC, MC and LC), major sources

influencing the river water quality are industrial waste

disposal sites leaching, domestic and industrial waste, soil

weathering and agricultural land runoff. Further, the uniden-

tified miscellaneous sources also contribute significantly

to the river water quality variations (Table 8ac). Relative

contributions of different sources for major water quality

variables identified for the region through DA (Fig. 5ac)

suggest that the miscellaneous unidentified sources along

with the natural weathering, industrial waste sites leaching,

domestic and industrial wastewater, agricultural soil leaching

and runoff from catchments are the major sources/factors

contributing most to the river water quality in all the three

regions.


19/20


Fig. 5. (a) Source contributions for selected water quality parameters of Gomti river in the UC region: SM = miscellaneous sources; S1 = industrial waste;

S2 = municipal and industrial waste; S3 = soil runoff 1; S4 = weathering and runoff; S6 = soil runoff 2; S7 = waste site runoff. (b) Source contributions for

selected water quality parameters of Gomti river in the MC region: SM = miscellaneous sources; S1 = municipal and industrial waste 1; S2 = industrial waste;

S3 = municipal and industrial waste 2; S4 = agricultural runoff; S5 = soil runoff 1; S6 = soil runoff 2; S7 = waste site runoff. (c) Source contributions for selected

water quality parameters of Gomti river in the LC region: SM = miscellaneous sources; S1 = agricultural runoff; S2 = industrial waste 1; S3 = municipal and

industrial waste; S4 = soil runoff; S5 = waste site runoff; S6 = industrial waste 2.

4. Conclusions

In this study, hierarchical CA grouped the sampling sites

into three clusters of similar characteristics reflecting the wa-

ter quality characteristics. The extracted grouping informa-

tion can be used in reducing the number of sampling sites

without missing much information. FA/PCA helped in iden-

tifying the factors/sources responsible for river water quality

variations in three differentregions.VFs obtained from FA in-

dicate that the parameters responsible for water quality varia-

tions aremainly related to trace metals(leaching from soil and

industrial waste sites), soluble salts (natural), organic pollu-

tion and nutrients (anthropogenic). DArendered an important

data reduction as it uses only five parameters (temperature,

total alkalinity, Cl, Na and K) affording more than 94% right

assignations in temporal analysis, while 10 parameters (river

discharge, pH, BOD, Cl, F, PO4, NH4N, NO3N, TKN and

Zn) to afford 97% right assignations in spatial analysis of

three different regions in the basin. Thus, DA allowed reduc-

tion in dimensionality of the large data set, delineating a few

indicator parameters responsible for large variations in water

quality. Further, receptor modeling through multi-linear re-

gression of the absolute principal component scores (APCS-

MLR) provided apportionment of various sources/factors in

respective regions contributing to the river pollution. It re-

vealed that agricultural soil weathering, leaching and runoff;

municipal and industrial wastewater; waste disposal sites

leaching were among the major sources/factors responsible

for river quality deterioration. Thus, this study presents use-

fulness of multivariate statistical techniques in water qual-

ity assessment, identification and apportionment of pollution

sources/factors with a view to get better information about


20/20


the water quality and design of monitoring network/strategy

for effective management of water resources.

Acknowledgements

The authors thank the National River Conservation Direc-

torate (NRCD), Ministry of Environment and Forests, Gov-

ernment of India for financial support and Director, ITRC,

Lucknow for encouragement. Suggestions and help provided

by Prof. V. Simeonov (Faculty of Chemistry, University of

Sofia, Bulgaria) and Prof. D.A. Wunderlin (Facultad de Cien-

cias Quimicas, Universidad National de Cordoba, Argentina)

in multivariate analysis of data are thankfully acknowledged.

References

[1] S.R. Carpenter, N.F. Caraco, D.L. Correll, R.W. Howarth, A.N.

Sharpley, V.H. Smith, Ecol. Appl. 8 (1998) 559.[2] H.P. Jarvie, B.A. Whitton, C. Neal, Sci. Total Environ. 210/211

(1998) 79.

[3] M. Vega, R. Pardo, E. Barrado, L. Deban, Water Res. 32 (1998)

3581.

[4] W. Dixon, B. Chiswell, Water Res. 30 (1996) 1935.

[5] P. Simeonova, V. Simeonov, G. Andreev, Cen. Eur. J. Chem. 2 (2003)

121.

[6] V. Simeonov, P. Simeonova, R. Tsitouridou, Chem. Eng. Ecol. 11

(2004) 450.

[7] K. Bengraine, T.F. Marhaba, J. Hazard. Mater. B 100 (2003) 179.

[8] C.W. Liu, K.H. Lin, Y.M. Kuo, Sci. Total Environ. 313 (2003) 77.

[9] V. Simeonov, J.A. Stratis, C. Samara, G. Zachariadis, D. Voutsa,

A. Anthemidis, M. Sofoniou, Th. Kouimtzis, Water Res. 37 (2003)

4119.

[10] K.P. Singh, A. Malik, D. Mohan, S. Sinha, Water Res. 38 (2004)

3980.

[11] APHA, Standard Methods for the Examination of Water and Wastew-

ater, 20th ed., American Public Health Association, Washington, DC,

1998.

[12] D.A. Wunderlin, D.M.D. Pilar, A.M. Valeria, P.S. Fabiana, H.A.

Cecilia, B.M. De Los Angeles, Water Res. 35 (2001) 2881.

[13] M. Otto, Multivariate methods, in: R. Kellner, J.M. Mermet, M. Otto,

H.M. Widmer (Eds.), Analytical Chemistry, Wiley/VCH, Weinheim,

Germany, 1998, p. 916.

[14] R.A. Johnson, D.W. Wichern, Applied Multivariate Statistical Anal-

ysis, 3rd ed., Prentice-Hall International, Englewood Cliffs, New Jer-

sey, USA, 1992, p. 642.

[15] B. Helena, R. Pardo, M. Vega, E. Barrado, J.M. Fernandez, L. Fer-nandez, Water Res. 34 (2000) 807.

[16] Y.S. Fung, L.W.Y. Wong, Atmos. Environ. 29 (1995) 2041.

[17] E. Swietlicki, R. Krejei, Nucl. Instr. Meth. Phys. B 109/110 (1996)

519.

[18] G.D. Thurston, J.D. Spengler, Atmos. Environ. 19 (1985) 9.

water quality assessment and apportionment of pollution sources

Documents