Journal of Water Resource and Protection, 2012, 4, 576-589
http://dx.doi.org/10.4236/jwarp.2012.48067 Published Online August 2012 (http://www.SciRP.org/journal/jwarp)
Impact of Land Use and Aquatic Plants on the Water
Quality of the Sub-Tropical Alpine Wetlands in India:
A Case Study Using Neuro-Genetic Models
Malabika Biswas Roy1, Pankaj K. Roy2, Asis Mazumdar2, Mrinmoy Majumder2, Nihar R. Samal3
1Gandhi Centenary B. T. College, Habra, West Bengal State University, West Bengal, India
2School of Water Resources Engineering, Jadavpur University, Kolkata, India
3Department of Civil Engineering, National Institute of Technology, Durgapur, India
Received April 30, 2012; revised June 1, 2012; accepted June 12, 2012
ABSTRACT
The suspended and dissolved waste in the incoming storm water of wetlands largely depends on the adjacent land use
which can influence the quality of the water body. The micro- and macro-floral population of a wetland can absorb,
convert, transform and release different organic or inorganic elements, which can also change or impact the overall
quality of the wetland water. The present study investigates the influence of the land use and the plant species in the
waterbed on the water quality of a high-altitude, sub-tropical wetland in India. The estimation capabilities of
neuro-genetic models were utilized to identify the inherent relationships between the Biochemical Oxygen Demand
(BOD), Dissolved Oxygen (DO), chlorine (Cl) and Chemical Oxygen Demand (COD) with the land use and wetland
zoology. A thematic map of the quality parameters was also generated based on the identified relationship to observe
the influence that the morphological and biological diversity in and around the study area has on the quality parameters
of the wetland. According to the results, the BOD, COD and Cl were found to vary with differences in land use and the
presence of different plant species, whereas the DO was found to be largely invariant with changes in these parameters.
The reasons may be contributed to the impact of uncontrolled eco-tourism activities around the wetland.
Keywords: Wetland; Neural Network; Water Quality; Land Use; Aquatic Plants
1. Introduction
Nowadays, maintaining the water quality of freshwater
wetlands has become a significant issue because, for this
kind of wetland, municipal and industrial wastewater
discharge constitutes a constant polluting source, whereas
surface run-off is a seasonal phenomenon [1]. For this
reason, the water environment quality issue is a subject
of ongoing concerned for the development of an econ-
omy in any country [2]. Naturally, a well-designed wa-
ter-quality monitoring plan should preserve scarce re-
sources by minimizing the redundancy of nearby moni-
toring stations and the plethora of possible variables
monitored, while at the same time maximizing the in-
formation content of the collected data [3]. However, it is
also true that to construct a well-designed water-quality
monitoring plan, detailed testing of the water quality is
essential for any freshwater wetland. In this paper, a
multivariate statistical characterization of water quality
of a tropical, freshwater lake in eastern India is dis-
cussed.
Some studies are available that deal with multiple pur-
poses, including water quality monitoring. The useful-
ness of multivariate statistical techniques demonstrated
[4] for the evaluation and interpretation of large complex
water-quality data sets and the apportionment of pollu-
tion sources/factors with the intention of obtaining better
information about the water quality and the design of the
monitoring network for the effective management of wa-
ter resources. A data set of 10 years analyzed [5] surface
water quality data pertaining to a polluted river using
partial least squares (PLS) regression models. In their
study, both the unfold-PLS and N-PLS (tri-PLS and
quadri-PLS) models were applied to the multivariate,
multi-way data array with the intention of assessing and
comparing their predictive capabilities for the biochemi-
cal oxygen demand (BOD) of a river water in terms of
their relative mean squared errors of cross-validation,
prediction and variance captured. However, the principal
component analysis applied [6] 16 water-quality pa-
rameters that had been collected monthly over a 6-year
period in an effort to describe the spatial dependence and
inherent variations of water quality patterns in the Flor-
C
opyright © 2012 SciRes. JWARP

M. B. ROY ET AL. 577
ida Bay-Whitewater Bay area. Moreover, Researcher [7]
tried to establish how the last years of artificial manage-
ment have affected the ecosystem of the Tablas de
Daimiel National Park, a Spanish continental wetland. To
carry out this study he analyzed the water physico-
chemical characteristics over the period 1995-1997, and
using different statistical techniques, these data were
compared with those obtained from a survey conducted
from 1974-1975, which represented the original situation.
However, geochemical characteristics and the apparent
ages of sampled groundwater were used [8] to determine
which of the two regionally extensive bedrock aquifers,
the lower bedrock aquifer or the upper bedrock aquifer,
is a more likely source of water discharging into the
springs in the Nine Springs watershed, which is located
in south-central Dane County, WI. Even, the potential of
systematic and formalized interdisciplinary research
concepts and methods for sustainable water and wetland
policy and management were reviewed and examined [9],
as advocated by the recently adopted European Water
Framework Directive. However, the multivariate data
analysis applied [10] large water quality data sets on the
Buyuk Menderes River Basin to analyze the surface wa-
ter contamination and establish correlations between wa-
ter quality parameters. Later, the water quality was ex-
amined [11] of the Tahtali River Basin in Turkey. In this
study, multivariate statistical methods, including factor,
principal component and cluster analyses, were applied
to surface water quality data sets obtained from the
Tahtali River Basin. The factor and principal components
analyses results revealed that the surface water quality
was mainly controlled by agricultural uses and domestic
discharges. The cluster analysis generated two clusters.
Moreover, a strategy was presented [3] to reduce the
measured parameters, locations, and frequency without
compromising the quality of the monitoring program.
Even so, the surface water quality data sets were ana-
lyzed [2] to obtain from the Xiangjiang watershed, which
were generated over 7 years (1994-2000) and monitored
12 parameters at 34 different profiles with the help of
multivariate statistical methods, including factor, princi-
pal component and cluster analysis. The multivariate
statistical methods, i.e., cluster analysis (CA) and dis-
criminate analysis (DA) were used [12] to assess tempo-
ral and spatial variations in the water quality of the wa-
tercourses in the Northwestern New Territories of Hong
Kong over a period of five years (2000-2004) using 23
parameters at 23 different sites. The principal component
analysis (PCA) was used [13] to reduce the data dimen-
sionality from the 18 original physico-chemical and
microbiological parameters which determined in drinking
water samples to six principal components that explained
about 83% of the data variability to analyze 126 drink-
ing water samples taken from a city water network in
North Moravia, the Czech Republic, over the course of
six months, according to a monitoring plan. In addition,
some samples were collected [14] to analyze the pa-
rameters such as Temperature, pH, DO, Conductivity,
Turbidity, Total Suspended Solids (TSS), Nitrate, Phos-
phate, COD and BOD from ten sampling locations dis-
tributed along the Juru Estuary in the Penang state of
Malaysia to analyze the water quality data with help of
CA and descriptive statistics. However, the environ-
mental economics for wetland construction, restoration
and preservation, and the net ecosystem services values
of constructed, human-interfered and natural wetlands
explored [15] as a comparative study for the case of a
typical human-interfered wetland in Wenzhou, China.
Nonetheless, Turner [16] emphasized an integrated wet-
land research framework, which suggests that a combina-
tion of economic valuation, integrated modeling, stake-
holder analysis, and multi-criteria evolution can provide
complementary insights into sustainable and welfare-
optimizing wetland management and policy. Furthermore,
Ouyan [17] applied principal component analysis (PCA)
and principal factor analysis (PFA) techniques to evalu-
ate the effectiveness of the surface water quality-moni-
toring network in a river, where the variables are evalu-
ated at the monitoring stations. The objective of his study
was to identify the monitoring stations that are important
in assessing the annual variations of river water quality.
Detenbeck [18] even developed a method that evaluated
the cumulative effect of wetland mosaics on the water
quality, which was applied to 33 lake watersheds in the
seven-county region surrounding Minneapolis-St. Paul,
Minnesota. Wayland [19] compared biogeochemical data
from three synoptic sampling events, which represents
the temporal variability of base flow chemistry and land
use, using R-mode factor analysis. At the same time, a
meta-analysis was described [20] to estimate relation-
ships between the non-use components of willingness to
pay (WTP) for surface water quality improvements and a
combination of resource, context, and study design at-
tributes, where these attributes include estimated use
values for identical improvements. The relationships
between water quality and six different land uses were
investigated [21] to offer practical guidance in the plan-
ning of future urban developments. In terms of safe-
guarding the water quality, high-density residential de-
velopment, which results in a smaller footprint than
sparse development; should be the preferred option ac-
cording to his study. The R-mode factor analysis and
Q-mode cluster analysis were applied [22] to a set of
1349 groundwater analyses to determine the factors con-
trolling the groundwater composition and the main re-
sulting water types. The PCA was used [23] to assess the
degree of contamination and spatial distribution of heavy
metals, such as Ag, As, Cd, Co, Cr, Cu, Hg, Ni, Pb, Sr,

M. B. ROY ET AL. 578
Zn, and nutrients (Org-C, Tot-N and Tot-P) in different
areas of Taihu Lake in China.
Wetlands are also believed to play a significant role in
global climate change by acting as a source of an atmos-
pheric greenhouse gases, such as methane, carbon, and
nitrogen [24]. Global biodiversity is also enhanced by
wetlands, which are vital for the survival of dispropor-
tionately large number of threatened and endangered
species [25]. However, uncontrolled domestic discharges
caused by rapid urbanization are threatening the surface
water quality [26] of wetlands. Consequently, various
physico-chemical and microbiological parameters of wa-
ter and different biogeochemical cycles of wetlands are
affected intensely. In such a situation, the present wet-
land is really important because it represents a major
group of Indian wetlands that are endangered by a lack of
appreciation of the importance of their role and as soft
target of developers.
The land use and quality of water in the wetlands are
correlated. The runoff from the industries, roads nor-
mally has higher concentration of metallic compounds
and dissolved wastes than from hills or forests whereas
organic compounds are found in higher concentration in
runoff from the latter. The type of floral species that re-
side in the wetland bottom can also influence the water
characteristics. Shrubs can uptake pollutants from the
muck layer, small algae can increase dissolved oxygen.
The microbes can increase the BOD by decomposing the
organic wastes. Table 1 depicts the impact of different
land use on the water quality of wetlands as observed by
[27,28] and many other scientists.
The floral population found in the wetlands was also
found to be influential in controlling water quality of
such water bodies. The quality of the wetlands can be
accessed by observing the presence of different kind of
plant species. Some examples of such indications are
shown in Table 2.
1.1. Impact of Major Quality Parameters on the
Wetland Quality
As per recommendation by the APHA [29] the observa-
tion of the following parameters can yield an overview of
the overall water quality of water bodies. The indicative
properties of such parameters can give a clear idea about
the overall quality of the wetlands.
1.1.1. Biochemical Oxygen Demand (BOD)
Oxygen is used for respiration in animals. Fish require
the highest concentrations of oxygen. If the dissolved
oxygen falls below 5 ppm (part per million), fish are the
first to suffer and die. Then, the population of bacteria
rises to abnormal levels. Imbalances between species are
a sign of water pollution. Substances that consume dis-
solved oxygen and add to the biochemical oxygen de-
mand are pollutants. Such substances come from human
Table 1. Relationship betw ee n land use and the water quality of w a te r bodies, as documented in diffe re nt sc ientific ar ticles.
Land Use Possibility of Water Contaminant Reason
Road (R)
High concentrations of chloride, nitrate and pesticides
can be observed in the adjacent wetlands. The amount of
concentration depends on the population and road
density of the contributing area.
Salt-treated roads, surface runoff from adjacent residential,
agricultural and industrial regions, population density and soil
porosity can vary the intensity of contamination.
Forest (F) The lowest chloride and nutrient concentrations can be
observed in wetlands adjacent to high density forests.
As mixing of residential and industrial wastes with surface as
well as groundwater increases the chloride and nutrient con-
centration, absence of the same has decreased the extent of
contamination.
Road & Agriculture Presence of pesticides and herbicides.
The fertilizers applied in the adjacent agricultural area can
contaminant surface runoff as well as seepage from aquifers,
thereby increasing the toxicity and nutrient content of the
wetland water.
Road & Industry Presence of volatile organic compounds like Trichloro-
ethane.
The effluents from petroleum and organic industries, if mixed
with ground and surface water, can severely contaminate the
wetland water.
Road & Residential
Domestic sewage can increase the concentration of
Nitrates, Chlorides and dissolved nutrients. The munici-
pal sewage water will have an elevated concentration of
organic compounds and ammonia which can deplete the
Singh et al., (2006) DO of the water body.
High population densities and human activity can contribute to
the contamination of water bodies.
Hill
The surface runoff that flushes in from a forest may filter
the dead bodies of macro-phytes. The mixing of such
runoff with the water body is quick and also increases
the diffusion of oxygen from the atmosphere due to the
aeration of surface water by high flow velocities. The
lowest Chloride and nutrient concentrations can be
observed in wetlands adjacent to high density forests in
the hills.
Because the mixing of residential and industrial wastes with
surface and ground water increases the chloride and nutrient
concentration of the wetland, the absence of the same has
decreased the amount of the contaminants.

M. B. ROY ET AL. 579
Table 2. Relationship between plant species and the water quality of wetlands as documented in different scientific and
government reports.
Types of Aquatic Species Symptoms Water Quality Reason
Rooted floating leaved
plants like the water lily
Increased
growth
Sediment contaminants may reach plant bodies.
The Chloride or Nitrate concentration will be re-
duced due to absorbance by the aquatic plants. The
DO will also be reduced under extreme conditions.
Presence of forest or agriculture fields can
increase the growth of such plants because
the surface runoff can flush in nutrient rich
water and sometimes act as a carrier agent of
the rooted plants.
Submerged plants Increased
growth
Decrease in the DO due to the respiration of the
submerged plants. The dead bodies of such plants
will attract microorganisms, which will increase
the BOD of the water body.
Such plants can trap nutrient-attached sediments
from reaching the algal population, thereby pre-
venting the occurrence of algal blooms.
Increase in the nutrient concentration due to
human activities like the deposition of
wastes, industrial effluents, etc.
The aeration of water, which will increase
the DO in the epilimnion and create an envi-
ronment conducive for the germination of
such plants.
Free Floating Plants like
water hyacinth (Eichhornia
crassipes)
Increased
growth
Increase in the DO but because the growth of free
floating plants is normally aggressive, the native
species of the water body face severe depletion.
The chloride or nitrate concentration will decrease
due to the absorbance of the metallic ions by such
plants.
Increase in the nutrient concentration due
to human activities, like the deposition of
wastes, industrial effluents etc.
Algae like Spirogyra sp.,
blue green algae, etc.
Increased
growth
Minor increase in the DO, but the lake color and
odor will change. Some types of filamentous algae
may produce scums or mats. Due to the parasitism
of microorganisms, lake water will show higher
BOD values.
Increase in the nutrient concentration due to
human activities.
waste. The amount of dissolved oxygen used up during
oxidation by bacteria of the organic matter in a sample of
water is called the biochemical oxygen demand (BOD).
Water is rated as pure if the BOD is 1 ppm or less, fairly
pure with a BOD of 3 ppm and suspect when the BOD
reaches 5 ppm.
1.1.2. Chemical Oxygen Demand (COD)
The chemical oxygen demand (COD) test is used to in-
directly measure the amount of organic compounds in
water that can be oxidized with both organic and inor-
ganic oxidizing agents. The regulated amount of COD
for surface water is generally 200 - 1000 mg/L but differs
with respect to country and state.
1.1.3. Dissolved Oxygen (DO)
Dissolved oxygen (DO) is the oxygen that is dissolved in
water by diffusion from the surrounding air or aeration of
water. Fish and aquatic animals cannot split oxygen from
water (H2O) or other oxygen-containing compounds.
Only green plants and some bacteria can do that through
photosynthesis and similar processes. Virtually all of the
oxygen we breathe is manufactured by green plants. A
total of three-fourths of the earth’s oxygen supply is
produced by phytoplankton in the oceans. If water is too
warm, there may not be enough oxygen in it. When there
are too many bacteria or aquatic animal in the area, they
may overpopulate, and consume the DO in great amounts.
Oxygen levels also can be reduced through the over-fer-
tilization of water plants by run-off from farm fields
containing phosphates and nitrates (the ingredients in
fertilizers). Under these conditions, the numbers and
sizes of water plants increase a great deal. Then, if the
weather becomes cloudy for several days, respiring
plants will use much of the available DO. When these
plants die, they become food for bacteria, which in turn
multiply and use large amounts of oxygen.
Numerous scientific studies suggest that 4 - 5 parts per
million (ppm) of DO is the minimum amount that will
support a large, diverse fish population. The DO level in
good fishing waters generally averages about 9 parts per
million (ppm). When DO levels drop below about 3 parts
per million, even hardy fish will die.
1.1.4. Chlori n e (Cl)
Chlorine is used as disinfectant due to its capacity of
oxidation. Chlorine can be found as free chlorine or as
total chlorine. Free chlorine is highly toxic for aquatic
inhabitants and microbes due to its highly oxidizing na-
ture. Aquatic animals can tolerate up to 1 mg/L of free
chlorine and fish will usually die if more than 0.36 mg/L
of chlorine is found in the water. However, low concen-
trations of chlorine, i.e. less than 0.1 mg/L, can improve
the quality of the water body. The total chlorine can rep-
resent the salinity of a water body where an excess
amount of chloride can harm the aquatic inhabitants.
The present investigation aims to identify the rela-
tionship between the land use, type of aquatic plant and
the above water quality parameter. A brief introduction
and the common methodology followed in achieving the
objectives through neural networks are discussed next.

M. B. ROY ET AL. 580
1.2. Artificial Neural Network (ANN)
An (ANN) is a flexible mathematical structure that is
capable of identifying complex nonlinear relationships
between input and output data sets. Artificial Neural
Networks (ANNs) offer a relatively quick and flexible
means of modeling, and as a result, the application of
ANN modeling has been widely reported in various hy-
drological studies [30-32]. In the context of hydrological
forecasting, recent papers have reported that ANNs may
offer a promising alternative for rainfall-runoff modeling
[33-36], stream flow prediction [37-40], reservoir inflow
forecasting [41,42] and the prediction of water quality
parameters [43]. All the papers reported a high degree of
satisfaction with the neural network estimations.
Artificial neural networks are viable computational
models for a wide variety of problems. These include
pattern classifications, speech synthesis and recognition,
adaptive interfaces between humans and complex physic-
cal systems, function approximation, image compression,
associative memory, clustering, forecasting and predict-
tion, combinatorial, combinatorial optimization, nonlin-
ear system modeling, and control. These networks are
“neural” in the sense that they may have been inspired by
neuroscience but not necessarily because they are faithful
models of neurobiological or cognitive phenomena. In
fact, the majority of the networks covered in this book
are more closely related to traditional mathematical and
statistical models, such as non-parametric pattern classi-
fiers, clustering algorithms, nonlinear filters, and statis-
ticcal regression models than they are to neurobiology-
cal models.
1.2.1. Mathematical Representation of Ar t i fi ci al
Neural Network
An (ANN) (see Figure 1) is a flexible mathematical
Figure 1. A schematic diagram of an artificial neural net-
work.
structure that is capable of identifying complex nonlinear
relationships between input and output data sets. The
ANN model of a physical system can be considered as n
input neurons


,
12 n
x
xx, h hidden neurons


,zz z

,
12 n and m output neurons 12 n
y
yy. Let tj
be the bias for neuron zj and fk for neuron yk. Let wij be
the weight of the connection from neuron xi to zj and beta
is the weight of the connection zj to yk. The function that
ANN calculates is:



1
kA jjkk
ygzb fjh


 (1)
In which,



1
jA jijj
zf xwt i h


 (2)
where gA and fA are the activation functions [44].
The development of an artificial neural network, as pre-
scribed by ASCE [45] follows the following basic rules:
1) Information must be processed at many single ele-
ments called nodes.
2) Signals are passed between nodes through connec-
tion links, and each link has an associated weight that
represents its connection strength.
3) Each of the nodes applies a non-linear transforma-
tion called an activation function to its net input to de-
termine its output signal.
The numbers of neurons contained in the input and
output layers are determined by the number of input and
output variables of a given system. The size or number of
neurons of a hidden layer is an important consideration
when solving problems using multilayer feed-forward
networks. If there are fewer neurons within a hidden
layer, there may not be enough opportunity for the neural
network to capture the intricate relationships between
indicator parameters and the computed output parameters.
A network with too many hidden layer neurons not only
requires a large computational time for accurate training
but may also result in overtraining. A neural network is
said to be “over-trained” when the network focuses on
the characteristics of individual data points rather than
just capturing the general patterns present in the entire
training set. The network building procedure is divided
into three phases, which are described next in a broad
way.
1.3. Study Area
Mirik Lake (26˚54'N to 26.9˚N—latitude and 88˚10'E to
88.17˚E—longitude) is situated in a valley encircled by
hill ridges with an extensive natural drainage network.
This lake is located at an altitude of 1767 meters above
the sea level. It is 49 km from Darjeeling and falls in the
state of West Bengal in India. On the western side close
to Mirik Lake flows the Mechi River, which demarcates
the Indo-Nepal border. The climate is pleasant all year

M. B. ROY ET AL.
JWARP
581
round with temperatures of a maximum 30˚C in the sum-
mer and a minimum of 2˚C in winter. Mirik Lake is sur-
rounded by the Mirik bazaar, Thana-line, Krishnanagar,
Pratapgaon, and Mahendragaon (wards nos. 2, 3, 5, 7 and
8 respectively). Its rich biodiversity, location on a mi-
gratory bird route, and vast areas of suitable habitats for
multiple bird species make the lake important center for
wildlife [46]. Mirik Lake as a whole contains multi-
farious features for boating, recreation, jogging, fairs,
picnics and many other activities. The total of pollution
load is drained from the surface runoff carrying the do-
mestic and municipal sewage for the entire Mirik Lake.
Some other sources of pollution, such as the outflow
from hotels carrying waste, human excreta from poor
sanitation, washing clothes, bathing, etc., are drained into
the Lake. Figures 2 and 3 show the location of sampling
points on Mirik Lake and the land use map of Mirik Lake,
respectively.
and the characteristics of the wetland bottom, on the wa-
ter quality of wetlands. A neuro-genetic model was de-
veloped to estimate the interrelation between the former
with the latter. The study can reveal the answers to ques-
tion such as:
1) How does the quality parameter vary from the pe-
ripheral region to the central region of the wetlands? The
1.4. Objective
The present study investigates the influence of land use Figure 2. Location of sampling points in mirik lake.
Figure 3. Land use map of mirik lake.

M. B. ROY ET AL. 582
quality of the water in the peripheral region will be sig-
nificantly influenced by the adjacent land use, but the
water quality in the central region will be impacted by
the floral population of the wetland.
2) What are the relationship between land use and the
quality of stored water and why do these relationships
exist?
3) What are the relationships that exists between aquatic
plants and the quality parameters of the water and why
do these relationships exist?
2. Methodology
2.1. Sample Collection and Analysis
During study period, the water-spread area of Mirik Lake
was about 162,256 m2, the width was 163 m, the depth
was 1.65 m and the total volume of the water was about
2,677,224 m3. The point sources of pollution of this lake
are the four main drains for the domestic and municipal
sewage. The non-point sources of pollutions include the
outflow of hotels, clothes washing, bathing and surface
run off from the surrounding areas.
Samples (Figure 2) were collected from all the sam-
pling points on the same day at different times. The DO,
pH, temperature and turbidity were measured on spot in
the field with Rugged Field Kit HQ Series Portable Me-
ters (HQd/IntelliCALTM-8505300-HACH). Samples were
brought to Kolkata for further physico-chemical (BOD,
COD and chloride) and bacteriological analysis at the
School of Water Resources Engineering, Jadavpur Uni-
versity as per the standard method [30].
2.2. Selection of the Land Use Class
A land use map (Figure 3) of the study area was devel-
oped with the help of satellite imagery and a ground
tooth sample survey. The major land uses within the 500
m diameter around the lake were identified as: Hill (H),
Forest (F), Pond (P) and Road (R) as can be observed
from Figure 2, which also showed the collection points
for surface water.
2.3. Identification of Major Aquatic Species of
the Lake
The aquatic species of the lake were identified from both
the ground tooth survey and microscopic tests of the col-
lected samples. The three major types of planktons iden-
tified were divided into the following classes: Rooted
Floating Leaves or Shrubs (S), Submerged Plants or
Shrubs (SS) (Table 2), and Clear Water (C), which
represents water with no noticeable floral population.
The identified land use and plant classes along with
the value of BOD, COD, DO and Cl parameters from the
collected samples, a neural network model was devel-
oped with each quality parameter as output and land use,
plant classes, distance from the wetland periphery and
other 3 quality parameter as input. The other parameters
were included as input to educate the model about the
interrelationship if any in-between the parameters which
also help to validate the model output.
2.4. Development of the Neuro-Genetic Models
2.4.1. Network Building Procedure
Selection of Network Topology
Neural networks can be of different types, like feed for-
ward, radial basis function, time lag delay etc. The type
of network is selected with respect to the knowledge of
input and output parameters and their relationships. Once
the type of network is selected, selecting the network
topology is the next concern. A trial and error method is
generally used for this purpose, but many studies now
prefer the application of a genetic algorithm [47]. Ge-
netic algorithms are search algorithms based on the me-
chanics of natural genetics and natural selection. The
basic elements of natural genetics—reproduction, cross-
over, and mutation—are used in the genetic search pro-
cedure. A GA can be considered to consist of the fol-
lowing steps [48]:
1) Select an initial population of strings.
2) Evaluate the fitness of each string.
3) Select strings from the current population to mate.
4) Perform crossover (mating) for the selected strings.
5) Perform mutation for selected string elements.
6) Repeat steps 2) - 5) for the required number of gen-
erations.
The genetic algorithm is a robust method of searching
for the optimum solution to complex problems, such as
the selection of an optimal network topology, where it is
difficult or impossible to test for optimality. The basics
of GAs have already been discussed by many authors
[47,49,50]. Hence the details of the basic procedures of
GAs are not discussed in the present literature.
2.4.2. Training Phase
To encapsulate the desired input output relationship, the
weights are adjusted and applied to the network until the
desired error is achieved. This is called as “training the
network”.
2.4.3. Testing Phase
After training is completed, some portion of the available
historical dataset is fed to the trained network and a
known output is estimated out of them. The estimated
values are compared with the target output to compute
the MSE. If the value of MSE is less than 1%, then the
network is said to be sufficiently trained and ready for
estimation (see Figure 4). The dataset is also used for

M. B. ROY ET AL. 583
Data
Preprocessing
I N P U T L A Y E R(x
i
)
H I D D E N L A Y E R (z
j
)
O U T P U T L A Y E R
w
ij
y
k
f
A
(x
i
)
b
jk
z
j
t
j
f
k
g
A
(x
i
)
If,
(y
k
-d
k
)<desired MSE
NO
NO
Training Algorithm like
QP, CGD and BBP to
update Weight
Training Algorithm like
QP, CGD and BBP to
update Weight
YES
MODEL DEVELOPMENT SUCCESS FUL
Figure 4. The basic methodology followed for the develop-
ment of a Neural Network [40].
cross-validation to prevent over-training during the train-
ing phase [44].
3. Result & Discussion
In total four neuro-genetic models were prepared with
one of four quality parameters as the output. The models
were trained with the Quick Propagation and Conjugate
Gradient Descent training algorithms, and the perform-
ance analysis of the results from the models revealed QP
trained models for BOD and DO and CGD trained mod-
els for COD and Cl as the best trained neuro-genetic
model (see Table 3). The estimation work was carried
out with the help of the model trained with the selected
algorithm.
3.1. Discussion
According to the Table s 1, 2 and 4 following observation
and discussions are made to analyze the impact of land
use and aquatic plants on the quality of wetland water.
According to the results (Figure 5), eight combina-
tions of land use and types of wetland bottoms were
identified. The results showed that the BOD is worst (i.e.,
highest concentration) for the combination of road and
submerged floral species and the combination of clear
water and forest yielded the best concentration of the
BOD (i.e., lowest concentration). The combination of S
and R showed a moderate concentration of BOD in the
lake water. BOD concentrations of a lake are known to
increase in the presence of organic content, which invites
micro-bacteria. The bacteria, with the help of dissolved
oxygen, decay the organic matter. The presence of or-
ganic content can thus reduce the DO of a lake. Lowering
the DO concentration can severely impact the fish popu-
lation and other floral and faunal colonies of the lake
because these species depend on the lake DO for food
production and respiratory activities. In the present in-
vestigation, the presence of submerged species in the
water may increase the organic content of the water. The
dead bodies of such species can increase the BOD con-
centration. The road adjacent to the lake will indicate
heavy depositions of organic wastes in the pond from the
incoming population, and because the lake is famous for
eco-tourism, the influx of temporary population is very
high. However, the absence of any biological species in
the lake water and forest in the adjacent areas has re-
duced the deposition of organic wastes in the water. The
forest had acted as a filter for removing organic matter,
such as the dead bodies of animals and large trees from
Table 3. The specification used and results achieved from the neuro-genetic models developed for the present study.
Network Input Hidden Layer Output Training Algorithm Training MSE Testing MSE MSE r SD
BOD 6.00 7.00 1.00 QP 0.05 0.08 0.06 0.870.96
BOD 6.00 3.00 1.00 CGD 1.21 1.56 1.02 0.760.87
COD 6.00 6 1.00 QP 7.87 7.68 5.67 0.760.88
COD 6.00 3.00 1.00 CGD 5.45 5.88 4.95 0.880.90
DO 6.00 3 1.00 QP 0.05 0.045 0.05 0.980.89
DO 6.00 3.00 1.00 CGD 0.07 0.08 0.08 0.761.20
Cl 6.00 4 1.00 QP 0.02 0.04 0.06 0.890.95
Cl 6.00 16.00 1.00 CGD 0.01 0.03 0.01 0.980.96

M. B. ROY ET AL. 584
Table 4. The impact of characteristics of adjacent land use
and wetland bed on the four quality parameters considered
for the present investigation.
Combination LUB LUA COD Cl BODDO
1 S R 26.27 9.55 8.685.67
1 S R 95.85 7.87 5.045.96
1 S R 37.56 5.50 9.655.54
1 S R 63.90 4.06 9.245.62
2 N H 45.72 5.23 9.535.82
3 SS H 18.83 19.53 11.055.22
3 SS H 143.91 12.74 2.524.42
3 SS H 55.11 8.51 12.024.52
3 SS H 42.82 21.62 14.805.62
2 C R 368.87 29.54 5.555.79
4 C F 30.55 20.20 0.244.97
2 C R 84.43 26.41 9.84 4.91
5 C P 16.10 28.13 23.294.63
2 C R 207.98 21.93 10.32 4.66
2 C H 32.58 15.07 8.92 4.84
6 S H 41.42 21.16 8.31 5.16
7 SS H 30.24 22.54 8.52 4.93
7 SS H 16.29 15.54 24.534.97
7 SS R 56.94 26.65 21.744.59
8 SS R 30.51 26.23 20.594.56
the surface runoff coming through the forest. Hence, the
BOD of such areas were found to be lower than those of
other areas that are widely visited by the tourist and lo-
cals who are economically dependent on the former for
their sustainability, but the locals and tourists had also
converted the lake into a basket for their waste materials.
The presence of non-biodegradable wastes can in-
crease the COD concentration of the lake. According to
the results (see Figure 6), the combination of clean water
and road yielded the higher values of COD, and a com-
bination of SS and H yielded the lower values of the pa-
rameter. The absence of shrubs or submerged species can
reduce the BOD of a water body, but the presence of
non-biodegradable wastes can increase the COD of the
same. The presence of roads has only smoothed the path
of such wastes with the surface runoff coming into the
lake unhindered. That may be the reason for high con-
centrations of COD when the road is present within the
500 m of the lake. The results from the model also
showed that the COD (mg/L) is lower (≤75 mg/L)
whenever shrubs or submerged shrubs are present in the
wetland bed, but the biochemical oxygen demands
(BODs) of such areas are found to be more than 5 mg/L.
The reason can be attributed to the presence of abiotic
bacteria present in the plankton population of the lake.
These bacteria produce their food with the help of oxy-
gen bonded to metallic ions. That is why the CODs in
such areas are lower because the chemical oxygen is not
used for decaying inorganic wastes, but BOD is more
than 5 mg/L because the decomposition of organic waste
is done by the abiotic bacterial population.
In case of the DO (see Figure 7), the identified com-
binations yielded no noticeable differences, but the DO is
found to be more for shrubs and road combination and
less for submerged shrubs and hill combinations. The
probable reason can be attributed to ecotourism events,
such as boating and fishing, which are rampant in the
lake and may aerate the lake water. Again, presence of
shrubs can also maintain the oxygen content of the lake
due to the ribosomal bacteria present in the roots, which
release oxygen during nutrient uptake. The relative in
crease of the DO in presence of shrubs may be the results
of such nutrification procedures.
From the prediction of chlorine (see Figure 8) it can
be observed that the concentration is higher for SS and R
combinations, whereas the same is lower for S and R
combination with respect to the other combinations iden-
tified. The presence of a road within 500 m of the lake
can contribute to the increase in chlorine concentration.
The lake, in case of the present investigation, is a popular
place for eco-tourism and the generated organic as well
as inorganic waste are deposited in the lake. The surface
runoff from the adjacent high altitude land also brings
dissolved chlorine due to the uncontrolled use of fertiliz-
ers in the adjacent floriculture. The unhindered surface
(due to the road area) from these areas along with depo-
sition of wastes by the tourists can cumulatively influ-
ence the increase in chlorine concentration of the lake.
However, because water plants are popular for their
chlorine uptake, the presence of shrubs has reduced the
chlorine concentration and absence of the same has al-
lowed the concentration to rise. The toxic byproducts
generated from a submerged algal population can in-
crease chlorine content, so the presence of submerged
planktons and the absence of shrubs may have allowed
the chlorine concentration to rise.
From Table 4 and the discussions above, it can be ob-
served that SS is an influential factor, which may impact
the quality of water, because most of the cases the pres-
ence of submerged shrubs had caused the differences in
the concentration of water quality parameters of the lake.
The road and hills, present in the area within 500 m of

M. B. ROY ET AL. 585
3 to
6 to
16 t
27 t
37 t
47 t
5
15
o 26
o 36
o 46
o 50
50 and ab
BOD Color
+WWC-4
WWC-3
+WWC-5
+WWC-6
+WWC-8
+WWC-9
+WWC-10
+WWC-11
+WWC-12
+WWC-13
+WWC-14
+WWC-15
+WWC-16
+WWC-17
+WWC-18
WWC-19
WWC-20
WWC-21
WWC-2
WWC-1
WWC-30
WWC-26
WWC-27
+WWC-28
Figure 5. Thematic map of the BOD concentration generated from the identified relationships between the land use, wetland
bed and the quality par a me ter.
3 to 50
51 to
151 to
201 to
301 to
351 to
401 and
150
200
300
350
400
ab
COD Color
WWC-06 WWC-08
WWC-09
WWC-10
WWC-11
WWC-12
WWC-13
WWC-14
WWC-15
WWC-16
WWC-17
WWC-18
WWC-19
WWC-20
WWC-21
WWC-04
WWC-03
WWC-02
WWC-01 WWC-36
WWC-38 WWC-34
WWC-30
WWC-26
WWC-27
WWC-28
Figure 6. Thematic map of the COD concentration generated from the identified relationships between the land use, wetland
bed and the quality par a me ter.

M. B. ROY ET AL. 586
53 to 10
3 to 4
2 to 3
1 to 2
0.5 to 1
0 to 0.5400
0
DO Color
WWC-06
WWC-08
WWC-09
WWC-10
WWC-11
WWC-12
WWC-13
WWC-14
WWC-15
WWC-16
WWC-17
WWC-18
WWC-19
WWC-20
WWC-21
WWC-04
WWC-03
WWC-02
WWC-01 WWC-36
WWC-38 WWC-34
WWC-30
WWC-26
WWC-28
WWC-27
Figure 7. Thematic map of the DO concentration generated from the identified relationships between the land use, wetland
bed and the quality par a me ter.
3 to 5
6 to 15
16 to 26
27 to 36
37 to 46
47 to 50
50 and ab
Chlorine Color
WWC-06
WWC-08
WWC-09
WWC-10
WWC-11
WWC-12
WWC-13
WWC-14
WWC-15
WWC-16
WWC-17
WWC-18
WWC-19
WWC-20
WWC-21
WWC-04
WWC-03
WWC-02
WWC-01 WWC-36
WWC-38 WWC-34
WWC-30
WWC-26
WWC-27
WWC-28
Figure 8. Thematic map of the Cl concentration generated from the identified relationships between the land use, wetland
ed and the quality par a me ter. b

M. B. ROY ET AL. 587
the lake are also identified as an influencing factor on the
quality of wetland water.
4. Conclusion
The present study investigates the relationship between
water quality parameters and adjacent land use and the
aquatic plants of the wetland with the help of neuro-ge-
netic models. The model results showed that the sub-
merged shrubs along with road and hills within the 500
m of the lake have a quantifiable relationship with the
water quality of the lake. The DO was found to be least
source of problems, and the BOD was found to be a
highly correlated parameter with the inputs among the
considered four parameters, which the study has assumed
to be representative of overall quality of the wetland. The
ecotourism activities, which are common in and around
the lake due to the geo-morphology and biodiversity of
the region was also found to be affecting the quality of
the lake water. That is why the anthropogenic impacts
coming from both tourist and local population on the
quality of water may be considered as the next objective
for the overall economic and environmental sustainabil-
ity of the wetlands, which is treated as the major source
of income of the local population. The identification of
the correlation can be performed for other lakes also to
obtain a better conclusion and generalization of the ob-
servations. The neuro-genetic models were found to be
suitable for the identification of relationships, which is
eminent from the low MSE achieved by all the models
trained with different training algorithms.
REFERENCES
[1] U. C. Panda, S. K. Sundaray, P. Rath, B. B. Nayak and D.
Bhatta, “Application of Factor and Cluster Analysis for
Characterization of River and Estuarine Water Systems—
A Case Study: Mahanadi River (India),” Journal of Hy-
drology, Vol. 331, No. 3-4, 2006, pp. 434-445.
[2] Q. Zhang, Z. Li, G. Zeng, J. Li, Y. Fang, Q. Yuan, Y.
Wang and F. Ye, “Assessment of Surface Water Quality
Using Multivariate Statistical Techniques in Red Soil
Hilly Region: A Case Study of Xiangjiang Watershed,
China,” Environmental Monitoring and Assessment, Vol.
152, No. 1-4, 2009, 2008, pp. 123-131.
[3] X. Zeng and T. C. Rasmussen, “Multivariate Statistical
Characterization of Water Quality in Lake Lanier, Geor-
gia, USA,” Journal of Environmental Quality, Vol. 34,
No. 6, 2005, pp. 1980-1991.
[4] K. P. Singh, A. Malik and S. Sinha, “Water Quality As-
sessment and Apportionment of Pollution Sources of
Gomti River (India) Using Multivariate Statistical Tech-
niques—A Cases Study,” Analytica Chimica Acta, Vol.
538, No. 1-2, 2005, pp. 355-374.
[5] K. P. Singh, A. Malik, N. Basant and P. Saxena, “Multi-
Way Partial Least Squares Modeling of Water Quality
Data,” Analytica Chimica Acta, Vol. 584, No. 2, 2007, pp.
385-396.
[6] N. J. Boyer, J. W. Fourqurean and R. D. Jones, “Spatial
Characterization of Water Quality in Florida Bay and
Whitewater Bay by Multivariate Analyses: Zones of
Similar Influence,” Estuaries and Coasts, Vol. 20, No. 4,
1997, pp. 743-758. doi:10.2307/1352248
[7] J. J. Berzas, L. F. Garciâa, R. C. Rodriâguez and P. J.
Martiân-Alvare, “Evolution of the Water Quality of a
Managed Natural Wetland: Tablas Dé Daimiel National
Park (Spain),” Pergamon, Vol. 34, No. 12, 2000, pp.
3161-3170.
[8] S. K. Swanson, J. M. Bahr, M. T. Schwar and K. W. Pot-
ter, “Two-Way Cluster Analysis of Geochemical Data to
Constrain Spring Source Waters,” Chemical Geology,
Vol. 179, No. 1-4, 2001, pp. 73-91.
doi:10.1016/S0009-2541(01)00316-3
[9] R. Brouwer, S. Georgiou and R. K. Turner, “Integrated
Assessment and Sustainable Water and Wetland Manage-
ment. A Review of Concepts and Methods,” Integrated
Assessment, Vol. 4, No. 3, 2003, pp. 172-184.
[10] H. Boyacioglu and O. Gunduz, “Application of Factor
Analysis in the Assessment of Surface Water Quality in
Buyuk Menderes River Basin,” European Water, Vol. 9-
10, 2005, pp. 43-49.
[11] H. Boyacioglu, “Water Pollution Sources Assessment by
Multivariate Statistical Methods in the Tahtali Basin,
Turkey,” Environmental Geology, Vol. 54, No. 2, 2008,
pp. 275-282.
[12] F. Zhou, Y. Liu and H. C. Guo, “Application of Multi-
variate Statistical Methods to the Water Quality Assess-
ment of the Watercourses in the Northwestern New Ter-
ritories, Hong Kong,” Environmental Monitoring and
Assessment, Vol. 132, No. 1-3, 2007, pp. 1-13.
doi:10.1007/s10661-006-9497-x
[13] P. Praus, “Urban Water Quality Evaluation Using Multi-
variate Analysis,” Acta Montanistica Slovaca, Vol. 12,
No. 2, 2007, pp. 150-158.
[14] A. F. M. Al-karkhi, M. A. Hossain and I. Norli, “Applica-
tion of Cluster Analysis for Water Quality Parameters-Juru
Estuary (Malaysia),” International Conference on Envi-
ronmental Research and Technology, 2008.
[15] Z. M. Chen, G. Q. Chen, B. Chen, J. B. Zhou, Z. F. Yang
and Y. Zhou, “Net Ecosystem Services Value of Wetland:
Environmental Economic Account,” Communications in
Nonlinear Science and Numerical Simulation, Vol. 14,
No. 6, 2009, pp. 2837-2843.
[16] R. K. Turner, “Valuation of Wetlands in a Landscape and
Institutional Perspective,” Ecological Economics, Vol. 35,
2000, pp. 1-6.
[17] Y. Ouyang, “Evaluation of River Water Quality Monitor-
ing Stations by Principal Component Analysis,” Water
Research, Vol. 39, No. 12, 2005, pp. 2621-2635.
doi:10.1016/j.watres.2005.04.024
[18] N. E. Detenbeck, C. A. Johnston and G. J. Niemi, “Wet-
land Effects on Lake Water Quality in the Minneapolis/St.
Paul Metropolitan Area, Netherlands,” Landscape Ecol-
ogy, Vol. 8, No. 1, 1993, pp. 39-61.

M. B. ROY ET AL. 588
doi:10.1007/BF00129866
[19] K. G. Wayland, D. T. Long, D. W. Hyndman, B. C. Pija-
nowski, S. M. Woodhams and S. K. Haack, “Identifying
Relationships between Basesflow Geochemistry and Land
Use with Synoptic Sampling and R-Mode Factor Analy-
sis,” Journal of Environmental Quality, Vol. 32, No. 1,
2003, pp. 180-190. doi:10.2134/jeq2003.0180
[20] R. J. Johnston, E. Y. Besedin and R. F. Wardwell, “Mod-
eling Relationships between Use and Nonuse Values for
Surface Water Quality: A Meta-Analysis,” Water Re-
sources Research, Vol. 39, No. 12, 2003, p. 1363.
doi:10.1029/2003WR002649
[21] A. Goonetillekea, E. Thomas, S. Ginn and D. Gilbert,
“Understanding the Role of Land Use in Urban Storm
Water Quality Management,” Journal of Environmental
Management, Vol. 74, No. 1, 2005, pp. 31-42.
[22] P. P. Sehot and J. V. D. Wal, “Human Impact on Re-
gional Groundwater Composition through Intervention in
Natural Flow Patterns and Changes in Land Use,” Jour-
nal of Hydrology, Vol. 134, No. 1-4, 1992, pp. 297-313.
[23] Q. Wenchuan, M. Dickman and W. Sumin, “Multivariate
Analysis of Heavy Metal and Nutrient Concentrations in
Sediments of Taihu Lake, China,” Hydrobiologia, Vol.
450, No. 1-3, 2001, pp. 83-89.
[24] M. Jansson, R. Anderson, H. Berggren and L. Leonardson,
“Wetlands and Lakes as Nitrogen Traps,” JSTOR: Ambio,
Vol. 23, No. 6, 1994, pp. 320-325.
[25] W. J. Mitsh and J. G. Gossolink, “Wetlands,” 2nd Edition,
Van Nostrand Reinhold, New York, 1993.
[26] H. Boyacioglu, “Development of Universal Water Quality
Index and Environmetric Analysis in Surface Water Qual-
ity Assessment with Field Implementation and Valida-
tion,” M.Sc. Thesis, Karlsruhe University, Karlsruhe, 2006.
[27] M. V. Croft and P. Chow-Fraser, “Use and Development
of the Wetland Macrophyte Index to Detect Water Qual-
ity Impairment in Fish Habitat of Great Lakes Coastal
Marshes,” Journal of Great Lakes Research, Vol. 33, No.
sp3, 2007, pp. 172-197.
doi:10.3394/0380-1330(2007)33[172:UADOTW]2.0.CO;
2
[28] D. Eckhardt, “Land Use and Water Quality,” 2003.
http://www.waterencyclopedia.com/La-Mi/Land-Use-and
-Water-Quality.html
[29] American Public Health Association (APHA), “Standard
Methods for the Examination of Water and Wastewater,”
19th Edition, APHA, Washington, 1995.
[30] Q. Zhang and S. J. Stanley, “Realtime Water Treatment
Process Control with Artificial Neural Networks,” Jour-
nal of Environmental Engineering, Vol. 125, No. 2, 1999,
pp. 153-160.
doi:10.1061/(ASCE)0733-9372(1999)125:2(153)
[31] T. R. Neelakantan and N. V. Pundarikanthan, “Neural
Network Based Simulation-Optimization Model for Res-
ervoir Operation,” Journal of Water Resource, Planning
and Management, Vol. 126, No. 2, 2000, pp. 57-64.
doi:10.1061/(ASCE)0733-9496(2000)126:2(57)
[32] C. Ray and K. K. Klindworth, “Neural Networks for Ag-
richemical Vulnerability Assessment of Rural Private
Wells,” Journal of Hydrologic Engineering, Vol. 5, No. 2,
2000, pp. 162-171.
doi:10.1061/(ASCE)1084-0699(2000)5:2(162)
[33] K. Hsu, H. V. Gupta and S. Sorooshian, “Artifical Neural
Network Modeling of the Rainfall-Runoff Process,” Jour-
nal of Water Resource Research, Vol. 31, No. 10, 1995,
pp. 2517-2530. doi:10.1029/95WR01955
[34] D. A. Fernando and A. W. Jayawardena, “Runoff Fore-
casting Using RBF Networks with OLS Algorithm,”
Journal of Hydrologic Engineering, Vol. 3, No. 3, 1998,
pp. 203-209.
doi:10.1061/(ASCE)1084-0699(1998)3:3(203)
[35] A. S. Tokar and P. A. Johnson, “Rainfall-Runoff Model-
ing Using Artificial Neural Networks,” Journal of Hy-
drologic Engineering, Vol. 4, No. 3, 1999, pp. 232-239.
doi:10.1061/(ASCE)1084-0699(1999)4:3(232)
[36] S. Y. Liong, S. T. Khu and W. T. Chan, “Derivation of
Pareto Front with Genetic Algorithm and Neural Net-
work,” Journal of Hydrologic Engineering, Vol. 6, No. 1,
2001, pp. 52-61.
doi:10.1061/(ASCE)1084-0699(2001)6:1(52)
[37] T. A. Clair and J. M. Ehrman, “Using Neural Networks to
Assess the Influence of Changing Seasonal Climates in
Modifying Discharge, Dissolved Organic Carbon, and
Nitrogen Export in Eastern Canadian Rivers,” Journal of
Water Resources Research, Vol. 34, No. 3, 1998, pp.
447-455. doi:10.1029/97WR03472
[38] C. E. Imrie, S. Durucan and A. Korre, “River Flow Pre-
diction Using Neural Networks: Generalization beyond
the Calibration Range,” Journal of Hydrology, Vol. 233,
No. 3-4, 2000, pp. 138-153.
doi:10.1016/S0022-1694(00)00228-6
[39] M. Majumder, P. K. Roy and A. Mazumdar, “Optimiza-
tion of the Water Use in the River Damodar in West Ben-
gal in India: An Integrated Multi-Reservoir System with
the Help of Artificial Neural Network,” Journal of Engi-
neering, Computing and Architecture, Vol. 1, No. 2, 2007,
Article ID: 1398.
[40] M. Majumder, “Watershed Modeling of River Damodar
with the Help of Neural Network and Genetic Algo-
rithm,” Ph.D. Thesis, Jadavpur University, Kolkata, 2010.
[41] S. K. Jain, A. Das and D. K. Srivastava, “Application of
ANN for Reservoir Inflow Prediction and Operation,”
Journal of Water Resource, Planning and Management,
Vol. 125, No. 5, 1999, pp. 263-271.
doi:10.1061/(ASCE)0733-9496(1999)125:5(263)
[42] P. Coulibaly, F. Anctil and B. Bobee, “Daily Reservoir
Inflow Forecasting Using Artificial Neural Networks with
Stopped Training Approach,” Journal of Hydrology, Vol.
230, No. 3-4, 2000, pp. 244-257.
doi:10.1016/S0022-1694(00)00214-6
[43] H. R. Maier and G. C. Dandy, “Empirical Comparison of
Various Methods for Training Feed-Forward Neural Net-
works for Salinity Forecasting,” Journal of Water Re-
sources Research, Vol. 35, No. 8, 1999, pp. 2591-2596.
doi:10.1029/1999WR900150
[44] K. P. Sudheer, “Knowledge Extraction from Trained Neu-
ral Network River Flow Models,” Journal of Hydrologic
Engineering, Vol. 10, No. 4, 2005, pp. 264-269.

M. B. ROY ET AL. 589
doi:10.1061/(ASCE)1084-0699(2005)10:4(264)
[45] ASCE, “Task Committee on Application of Artificial Neu-
ral Networks in Hydrology. Artificial Neural Networks in
Hydrology I: Preliminary Concepts,” Journal of Hydro-
logic Engineering, Vol. 5, No. 2, 2000, pp. 115-123.
[46] C. F. Iscen, O. Emiroglu, S. Ilhan, N. Arslan, V. Yilmaz
and S. Ahiska, “Application of Multivariate Statistical
Techniques in the Assessment of Surface Water Quality
in Uluabat Lake, Turk,” Environmental Monitoring and
Assessment, Vol. 144, No. 1-3, 2008, pp. 269-276.
doi:10.1007/s10661-007-9989-3
[47] J. A. Ahmed and A. K. Sarma, “Genetic Algorithm for
Optimal Operating Policy of a Multipurpose Reservoir,”
Journal of Water Resources Management, Vol. 19, No. 2,
2005, pp. 145-161. doi:10.1007/s11269-005-2704-7
[48] D. H. Burn and J. S. Yulianti, “Waste-Load Allocation
Using Genetic Algorithms,” Journal of Water Resource,
Planning and Management, Vol. 127, No. 2, 2001, pp.
121-129.
doi:10.1061/(ASCE)0733-9496(2001)127:2(121)
[49] Q. J. Wang, “The Genetic Algorithm and Its Application
to Calibrating Conceptual Rainfall-Runoff Models,” Jour-
nal of Water Resources Research, Vol. 27, No. 9, 1991,
pp. 2467-2471. doi:10.1029/91WR01305
[50] R. Wardlaw and M. Sharif, “Evaluation of Genetic Algo-
rithms for Optimal Reservoir System Operation,” Journal
of Water Resources Planning and Management, Vol. 125,
No. 1, 1999, pp. 25-33.
doi:10.1061/(ASCE)0733-9496(1999)125:1(25)