- 129 -

ANANLYSIS OF WATER QUALITY VARIATION IN KELANI RIVER,

SRI LANKA USING PRINCIPAL COMPONET ANALYSIS

.S.W. Athukorala1, L. S. Weerasinghe1, M. Jayasooria1, D. Rajapakshe1, L. Fernando1, M. Raffeeze1

N.P. Miguntanna2
1
Department of Civil and Infrastructure, Faculty of Engineering, South Asian Institute of Technology and Medicine
(SAITM), Sri Lanka. Email: [email protected]
2
Senior Lecturer, Department of Civil & Infrastructure Engineering, Faculty of Engineering, South Asian Institute
of Technology and Medicine (SAITM), Sri Lanka.

ABSTRACT

This paper discusses the outcomes of a research project conducted to chracterise the water quality
variation in Kelani River and the influential parameters such as land use types. The water quality data
for nine parameters namely pH, Turbidity, Electrical Conductivity (EC), Chloride (Cl-), Chemical
Oxygen Demand (COD), Dissolved Oxygen (DO), Bio Chemical Oxygen Demand (BOD), Total
Coliforms, Escherichia Coli (E.Coli) collected from January 2010 to October 2012 from Kelani River
was used for the analysis. In addition to the univariate data analysis, most importantly multivariate data
analytical technique namely Principal Component Analysis (PCA) was used to understand the corelation
between parameters. Data analysis revealed that a considerable variation of the water quality due to the
different land use characteristics. However, the study identified importance of targetting the non point
sources such as stormwater runoff when design the pollution mitigation strategies and BMPs
irrespective of the type of land use. Furthermore, EC was identified as a surrogate indicator to represent
the Cl- concentration in water. In addition the use of PCA is recommended as a viable tool in water
quality studies undertake in Sri Lanka.

Key words: Kelani River, Different Land Uses, Pollutant Sources

1. INTRODUCTION
Water quality of receiving water bodies such as With the increasing attention on water quality of
rivers and lakes is critically important because it receiving water bodies, regulatory authorities
is one of the most essential resources for human strive to mitigate the adverse impacts of
existence and settlement. However, rapid growth urbanization and population growth on water
of population and increase of urban activities environments by implementing water quality
significantly influences the water quality of mitigation strategies. This mainly includes
receiving water bodies [5,12]. This is mainly due implementation of water quality legislations and
to the deterioration of water quality due to the regulations on waste disposals and the Best
higher pollutant loads resulting from various Management Practices (BMPs) such as storm
point and non-point sources of pollution. sewers wetlands and swales in the catchment
areas. In this context, accurate knowledge on
As noted by several researchers, point source of water quality variation and influential parameters
pollution primarily includes direct and is significant to develop appropriate mitigation
uncontrolled discharges from different land use strategies and BMPs.
types such as residential, industrial and
commercial land uses [5]. On the other hand On the other hand due to the increase demand on
urban storm water runoff has been recognized as potable water the need of time and cost effective
the most important non-point source of pollution treatment facilities has become to a crucial
to receiving water bodies [3,5]. importance. In this context, the identification of

SAITM Research Symposium on Engineering Advancements 2013

(SAITM – RSEA 2013)
 - 130 -

surrogate water quality indicators which can be and efficient water treatment processes.
used to represent the raw water quality is
significant in order to reduce the cost and time 2. METHODOLOGY
associate with the water quality monitoring
programs and hence the treatment processes. 2.1. Study Area

The water quality data for this study was

However, in Sri Lanka even though the point
collected on a stretch of 7 km along the Kelani
source pollution is mostly controlled by
River. Three sites namely, Intake well, Welivita
legislation the attention on implementation of the
Bridge and Kaduwela Bridge were selected to
BMPs to catchment areas is still limited.
collect water samples (Figure 1). These locations
Furthermore, the treatment processes are highly
have been selected in order to represent the
cost and time consuming. These constraints are
different land use characteristics in to understand
mainly due to the lack of understanding on the
the variability of water quality with the inherent
variability water quality of receiving water
characteristics of different land uses.
bodies with various influential parameters such
as land use, pollutant sources and their pathways
and the interrelationship between physical,
chemical and biological water quality
parameters.

This research project is conducted to understand

the water quality variation of Kelani River in the
Western Province of Sri Lanka with the type of
land uses and other influential parameters and to
identify the interrelationship between key
indicators of the water quality. Kelani River,
having the total length of 145 km and 2292 km2
of river basin is the second largest catchment
Figure 1: Location of study sites
area in the country and it is the most polluted
river in the Western Province due to rapid growth
Intake well is the place where the water is taken
of industries located in the close vicinity of the
from the river to the water treatment plant. It is
river and passes through the Colombo City
situated at a mix of industrial and residential land
which is the most populated, industrialized and
uses and it is near the main road of Mulleriyawa.
urbanized city in the country. At present around
470,000 m3 water is produced from the Kelani
Kaduwela Bridge connects two busy town areas
river for the use of the people. People in the
namely Kaduwela and Biyagama and located at
western Province depend on Kelani river for
the heart of the Kaduwela City. The traffic on the
drinking water, industrial purpose and other
bridge is considerable higher compared to the
water requirements [6].
other two locations. Welivita Bridge is located in
The outcomes generated from this study are
a residential land use, well surrounded with trees
aimed to contribute to design and conduct the
and both bank of the river are heavily covered
time and cost effective water quality monitoring
with the vegetation. During the site investigation
programs, design of the BMPs and the effective

SAITM Research Symposium on Engineering Advancements 2013

(SAITM – RSEA 2013)
 - 131 -

researchers observed a quarry near the river bank This analysis was conducted primarily to
understand the variability of the water quality
The water quality data for the selected sites were data with type of land use. Secondly, yearly
collected for the period of January 2010 to averages were obtained for each parameter
October 2012 from National Water Supply and measured at each location and analyzed for the
Drainage Board, Sri Lanka. All the samples variability of each parameter with time at each
collected were then tested for range of physical, location.
chemical and biological water quality parameters
according to the specified by Standard Methods 2.2.2. Multivariate Data Analysis
for Examination of Water and Wastewater
Due to the constraints arise when handling the
(APHA, 2001).. This includes nine water quality
number of samples and number of variables
parameters namely Turbidity (NTU), pH,
simultaneously multivariate data analysis
Electrical Conductivity (EC) Chloride (Cl-)
techniques were employed mainly to identify the
Chemical Oxygen Demand (COD) Dissolved
correlation between water quality parameters.
Oxygen (DO) Biochemical Oxygen Demand
Multivariate data analysis in this study was
(BOD) Total Coliforms E. Coli at 440C/100ml
performed using Principal Component Analysis
(PCA) which is one of the extensively used
The data obtained from the laboratory testing
pattern recognition techniques was employed
was then analyzed using two sophisticated
[5,8].
software packages available for the statistical
data analysis namely Microsoft Excel 2010 and
PCA transforms multivariate data into orthogonal
StatistiXL Version 1.9.
components called principal components (PCs)
which are uncorrelated to each other and are
2.2. Data Analysis Techniques
linear combinations of the original variables. PCs
retain the most variance within the original data
Both univariant data analysis techniques and so that transformation is achieved without loss of
multivariate data analysis techniques were information in the data set. The first principal
employed in this study to identify the variability component describes most of the data variance
of water quality with influential parameters and while the second PC, the next largest amount and
the correlation between water quality parameters. so on until as many PCs are generated depending
on the number of variables.
2.2.1. Univariate Data Analysis
In this study the PCA biplot was used to present
The univariate data analysis techniques
the outcomes of the analysis. The bi-plot shows
considered in this study are mainly included
both the loadings and the score for two selected
analysis of mean, standard deviation and analysis
principal components in parallel. The
of variability of each water quality parameter
relationships between the scores and the loadings
with land use and time. This analysis was
are often best displayed on the bi-plot. Therefore
conducted mainly in two stages.
patterns and relationships of data can be
Firstly, the average concentration of the each
identified by using objects as scores and water
water quality parameter measured for all the
quality parameters as the loadings [7,8,9]. The
three years at each location was analyzed using
angle between the lines (loadings), or, to be more
the methods of mean and standard deviation.

SAITM Research Symposium on Engineering Advancements 2013

(SAITM – RSEA 2013)
 - 132 -

precise, the cosine of the angle between the lines, 3. RESULTS AND DISCUSSION
approximates the correlation between the The results obtained from the univariate data
variables they represent. The closer the angle is analysis is shown Table 1 for all the locations in
to 90, or 270 degrees, the smaller the correlation terms of mean and standard deviation.
and the closer the angle to 0 degrees the stronger
the correlation [7].

Table 1: The average concentration of each parameter at each location

Parameter Kelani River Welivita Bridge Intake well
Average Standard Average Standard Average Standard
deviation deviation deviation
Turbidity (NTU) 14.56 21.19 11.03 14.46 11.16 14.18

pH 6.7 0.24 6.82 0.19 6.72 0.26

Electrical
Conductivity at 58.63 24.11 91.55 70.69 60.92 25.40
250C - (µS/cm)
Chloride (as Cl-)
13.40 7.43 17.18 13.72 10.88 4.33
(mg/l)
Chemical
Oxygen
19.39 21.75 17.78 27.93 11.79 10.18
Demand (COD)
– (mg/l)
Dissolved
Oxygen (DO)- 7.41 0.38 7.36 0.43 6.90 1.05
(mg/l)
Biochemical
Oxygen
1.68 1.22 2.44 2.89 2.21 2.64
Demand (BOD)
( mg/l)
Total Coliforms
30,606 49928.08 31481.25 51047.37 28546.09 46468.32
at 350C/100ml
E. Coli at
16166.47 35818.41 16575.63 39289.64 12914.22 29727.66
440C/100ml

As shown in Table 1 there is no considerable mining no attention has yet been paid on the
variation of pH between locations. However, storm water runoff as a major source of pollutant
Kaduwela bridge site show comparatively higher sources to Kelani River. Therefore, with the
values for turbidity, COD compared to other increase of roads and vehicular traffic it is of
locations. The higher Turbidity is mainly significant to generate in-depth understanding on
attributed to sand mining and the higher solids urban storm water runoff as a major pollutant
loads generated from nearby roads in the form of source to rivers and hence to design and
dust and wash-off with the runoff. Even though implement BMPs to minimize the impacts.
regulatory authorities in Sri Lanka have already Furthermore, in this context, regular street
implemented legislations to limit the sand sweeping practices could also be recommended

SAITM Research Symposium on Engineering Advancements 2013

(SAITM – RSEA 2013)
 - 133 -

to minimize the solids build-up on roads and Notably, the water quality data obtained for pH
hence to minimize the solids wash-off with the and Turbidity at all the sites whereas E-Coli for
storm water runoff to nearby water bodies such the Kaduawela and Welivita Bridge sites exceeds
as rivers and lakes. the permissible values for river water quality
compared to the guidelines given by [4].
The relatively higher COD levels at Kaduwela However, for other parameters are still within the
Bridge site could be mainly attributed to the permissible limits given by [ 4].
discharge of sewage as noted during the site
visits conducted prior to the data collection. This Figure 2 shows the PCA biplot obtained from the
is further confirmed by the findings of Mohan et Principle Component Analysis (PCA) conducted.
al. [10] who noted that high values of COD due In PCA all the water quality data at all the
to accumulation of domestic sewage at Naya locations were considered. This is due to the
Talab lake, Jodhpur, India. capability of this technique to analysis number of
samples and number of variables simultaneously
Moreover, as can be seen in Table 1 Cl-, EC, [8].
BOD, Total Coliforms and E.coli at Welivita
Bridge is higher than that compared to the other As can be seen in Figure 2 Turbidity shows a
locations. The higher value of EC can be mainly strong co-relation to E.coli and Total Coli forms.
-
attributed to the comparatively higher Cl and the This suggests that, increase of turbidity directly
quarry situates near the river bank. In this affects to the increase of micro biological
regards also, storm water runoff could play a pollution. This could be attributed to the
significant role in transporting pollutants such as presence of solids which lead to increase of
rock minerals and explosives used in the quarry turbidity provides a sustainable surface for total
to the river. On the other hand the increment of coliforms to grow up. This highlights the
BOD and E.Coli and Total Coli form level can be significance of implementation on best
mainly attributed to the organic and micro management practices such as wetlands,
biological pollution resulting from land use filtration systems and pollutant traps in the
characteristics such as domestic waste water catchment area to reduce the solids wash-off
discharges, plant debris and animal waste [13]. from the catchment generate from various
anthropogenic activities. This further suggests
-
Furthermore, higher Cl value at Welivita Bridge the importance of identification of turbidity as a
indicates the threat of eutrophication at this surrogate water quality parameter for Total
location compared to other locations studied in Coliforms in water quality monitoring and
this research project. treatments facilities associate with the Keleni
Rriver.

SAITM Research Symposium on Engineering Advancements 2013

(SAITM – RSEA 2013)
 - 134 -

Figure 2: Principal Component biplot obtained for all the locations.

Furthermore, as can be seen in Figure 2 Cl- Abyaneh et al. 2005testing procedure).

shows a strong correlation to EC. This Therefore, it is important to obtain in-depth
relationship is further confirmed by analysis of understanding on the inter dependence between
variation between EC and Chloride EC and Cl- where EC can be then used as a
concentrations as shown in Figure 3. surrogate to Cl- . This will eventually reduce the
time and cost associated with the water quality
monitoring programs and testing procedures.

Furthermore, another strong co-relation of this

plot can be seen between the parameters of BOD
and Cl-..BOD is caused by aerobic biological
organisms in water to break down organic
material present in a given water sample at
certain temperature. Increase in Cl-concentration
lead to increase the growth of in biological
-
Figure 3: Variation of Cl concentration with EC organisms. The strong correlation between Cl-
and BOD could be attributed to these reasons.
This suggests that the EC can be used as a good
indicator of the Cl- in the Kelani River. On the Additionally, the turbidity shows no correlation
other hand the measurement of EC is to EC. This confirms the importance of dissolved
-
considerably easier than the measurement of Cl fraction of solids in the water which leads to EC.

SAITM Research Symposium on Engineering Advancements 2013

(SAITM – RSEA 2013)
 - 135 -

This further suggests that the water treatment [3]. Bannerman, R.T., Owens, D.W., Dodds, R.B.
process and BMPs should also be targeted on and Hornewer, N.J. (1993). Sources of
minimizing the dissolved fraction of solids when pollutants in Wisconsin stormwater. Water
designing water treatment facilities which mostly Science and technology, 28(3-5), 241-259.
target the removal of only the suspended solids. [4]. Drinking water inspectorate ,Ergon house,
Horseferry London (2007 – 3rd edition).
4. CONCLUSION [5]. Goonetilleke, A., Thomas, E., Ginn, S. and
The water quality of Kelani River significantly Gilbert, D. (2005). Understanding the role of
depends on the different land use characteristics. land use in urban stormwater quality
However, study suggests that the importance of management. Journal of Environmental
considering the storm water runoff as a major Management, Vol. 74 (No. 1), pp. 31-42.
source of pollution to receiving water bodies [6]. IWRM, Sri Lanka, Innovative Practices for
when design the pollution mitigation strategies water and sanitation, Technical Report,
and BMPs irrespective of the type of land use. It February 2013.
is also recommended to adopt regular street [7]. Kokot,S. and Phuong, T.D. (1999). Elemental
sweeping practices to minimize the solids build- content of Vietnamese rice, Part 2.
up on roads which minimizes the solids wash-off Multivariate data analysis. Analyst, Vol. 124,
with the storm water runoff and hence to pp. 561-569.
safeguard the receiving water quality. [8]. Miguntanna N.P, Liu, A., Egodawatta P., and
Furthermore, the study revealed that the use of Goonetilleke A.,(2013), Characterising
-
EC as a surrogate indicator to Cl in water quality nutrients wash-off for effective urban
monitoring programs undertake based on the stormwater treatmentDesign, Journal of
Kelani River. In addition the use of Principal Environmental Management Vol. 120, pp.61-
Component analysis can also be recommended 67.
as a viable tool in water quality studies undertake [9]. Miguntanna N.P., (2009), Nutrients Build-Up
in Sri Lanka. And Wash-Off Processes In Urban Land
Uses, PhD Thesis, Queensland University of
5. ACKNOWLEDGMENT Technology, Brisbane, Australia.
[10]. Mohan, D., Gaur, A. and Choudhary, D.,
We convey our gratitude to the National Water 2007. Study of limnology and microbiology
Supply and Drainage Board (NSW&DB) for of Naya talab,Jaodhpur (Raj.). Proc.DAE-
BRNS Nat. Symp. Limnol., 07:64-68.
providing us the relevant data to conduct this
[11]. Ren,Y., Xiaoke, W., Zhiyun, O., Hua, Z.,
research project.
Xiaonan, D. and Hong, M. (2008). Storm
water runoff quality from different surfaces in
6. REFERENCES an urban catchment in Beijing, China. Water
[1]. American Public Health Association, Environment Research, Vol. 80 (No. 8), pp.
American Water works Association. Standard 719-724.
Methods for the Examination of water and [12]. Sharma, M.R. and Gupta, A.B., 2004.
wastewater. Washington, D.C, USA: Water Prevention and control of pollution in streams
Environment Federation , 2001, 20th Edition. of outer Himalayas
[2]. [Abyaneh et al. 2005 testing procedures.

SAITM Research Symposium on Engineering Advancements 2013

(SAITM – RSEA 2013)