and Software
Vol. 4, No. 2, 1 J une 2014
International Academy of Ecology and Environmental Sciences
Computational Ecology and Software
ISSN 2220721X
Volume 4, Number 2, 1 J une 2014
EditorinChief
WenJ un Zhang
Sun Yatsen University, China
International Academy of Ecology and Environmental Sciences, Hong Kong
Email: zhwj@mail.sysu.edu.cn, wjzhang@iaees.org
Editorial Board
Ronaldo Angelini (The Federal University of Rio Grande do Norte, Brazil)
Andre Bianconi (Sao Paulo State University (Unesp), Brazil)
Bin Chen (Beijing Normal University, China)
Daniela Cianelli (University of Naples Parthenope, Italy)
Alessandro Ferrarini (University of Parma, Italy)
Yanbo Huang (USDAARS Crop Production Systems Research Unit, USA)
Istvan Karsai (East Tennessee State University, USA)
Vladimir Krivtsov (HeriotWatt University, UK)
Lev V. Nedorezov (University of Nova Gorica, Slovenia)
Fivos Papadimitriou (Environmental and Land Use Consultancies, Greece)
George P. Petropoulos (Institute of Applied and Computational Mathematics, Greece)
Vikas Rai (J azan University, Saudi Arabia)
Santanu Ray (Visva Bharati University, India)
Kalle Remm (University of Tartu, Estonia)
Rick Stafford (University of Bedfordshire, UK)
Luciano Telesca (Institute of Methodologies for Environmental Analysis, Italy)
Bulent Tutmez (Inonu University, Turkey)
Ranjit Kumar Upadhyay (Indian School of Mines, India)
Ezio Venturino (Universita di Torino, Italy)
Michael J ohn Watts (The University of Adelaide, Australia)
Peter A. Whigham (University of Otago, New Zealand)
ZhiGuo Zhang (Sun Yatsen University, China)
Editorial Office: ces@iaees.org
Publisher: International Academy of Ecology and Environmental Sciences
Address: Flat C, 23/F, Lucky Plaza, 315321 Lockhart Road, Wanchai, Hong Kong
Tel: 008526555 7188
Fax: 008523177 9906
Website: http://www.iaees.org/
Email: office@iaees.org
Computational Ecology and Software, 2014, 4(2): 8288
IAEES www.iaees.org
Article
Spatial risk assessment of alien plants in China using biodiversity
resistance theory
YouHua Chen
Department of Renewable Resources, University of Alberta, Edmonton, T6G 2H1, Canada
Email: haydi@126.com
Received 24 February 2014; Accepted 27 March 2014; Published online 1 June 2014
Abstract
In the present study, the potential occurrence risk of invasive plants across different provinces of China is
studied using disease risk mapping techniques (empirical Bayesian smoothing and PoissonGamma model).
The biodiversity resistance theory which predicts that highbiodiversity areas will have reduced risk of species
invasion serves as the base for performing spatial risk assessment of plant invasion across provinces. The
results show that, both risk mapping methods identified that northeastern part of China have the highest
relative risk of plant invasion. In contrast, southwestern and southeastern parts of China, which have high
woody plant richness, are predicted to possess low relative risks of plant invasion. Through spatial regression
analysis (simultaneous autoregression model), nine environmental variables representing energy availability,
water availability, seasonality, and habitat heterogeneity are used to explain the relative risk of plant invasion
across provinces of China. The fitting results suggest that, PRECrange and TEMrange are the most two
important covariates correlated with the occurrence risks of alien plants at provincial level in China. As
indicated by Morans I index, spatial regression analysis can effectively eliminate the potential biases caused
by spatial autocorrelation.
Keywords spatial ecology; risk assessment; invasion biology; plant distribution.
1 Introduction
Alien plants may cause severe ecological problems onto local ecosystems because they can better compete for
the resources than native species (Zhang and Chen, 2011). Biodiversity hotspots are believed to be more
resistant to species invasion: more diverse communities would have less species invasions (Law and Morton,
1996; Stachowicz et al., 2002; Zhang, 2011, 2012a). This biodiversity resistance theory can be an option for
performing risk assessment of species invasions.
Modeling plant invasion risk in China has been done in some previous studies recently (Bai et al., 2013;
Liu et al., 2005; Wu et al., 2006; Feng and Zhu, 2010). However, all these studies only rely on the richness
ComputationalEcologyandSoftware
ISSN2220721X
URL:http://www.iaees.org/publications/journals/ces/onlineversion.asp
RSS:http://www.iaees.org/publications/journals/ces/rss.xml
Email:ces@iaees.org
EditorinChief:WenJunZhang
Publisher:InternationalAcademyofEcologyandEnvironmentalSciences
Computational Ecology and Software, 2014, 4(2): 8288
IAEES www.iaees.org
information of alien plants, the dependence of alien plant diversity and native plant diversity has never been
considered yet. As such, one may utilize the biodiversity resistance theory to quantify relative risk of alien
plants by explicitly incorporating the information of native plant diversity.
In the present study, we perform the relative risk assessment of alien plants in China by using spatial
disease mapping methods (including empirical Bayesian smoothing and PoissonGamma model) so as to
incorporate richness information of native woody plant species.
2 Materials and Methods
2.1 Invasive plant diversity data
The provincelevel 127 alien plant richness data are collected from previous studies (Chen, 2013; Wang et al.,
2009; Wu et al., 2006; Yan et al., 2012). The resultant data include provincial distributional information and
physiological trait patterns of each invasive plant found in China.
2.2 Woody plant diversity data
The population size at risk when performing risk mapping is required as an entry. In the present study, we
regard the population size at risk as the number of native woody plant species. As such, similar to the disease
transmission model, areas where there are higher population sizes would have less relative risk of disease
outbreak. In our modeling of relative risk estimation of plant invasion, we hold the recognition that areas with
high biological diversity should have less plant invasion risk (Law and Morton, 1996; Stachowicz et al., 2002).
2.3 Risk mapping using empirical Bayesian smoothing and PoissonGamma model
We use both empirical Bayesian smoothing and PoissonGamma model for predicting the outbreak risk of
invasive plants over the provinces of China. For testing whether the relative risk of plant invasion is
significantly higher than 1, we utilize Poisson exact test.
2.4 Identification of important covariates associated with the occurrence risk of alien plants
We use generalized linear models to identify important covariates which are closely related to the occurrence
risk of alien plants at provincial level in China. The fitting model is given by,
log( ) X
where is the relative risk of alien plants across various provinces of China calculated as above; is the
associated coefficients for the covariates presented in terms of columns in the covariate matrix X.
We standardize all the covariates in X so that the estimated coefficients are standardized partial
coefficients. We then choose those covariates with high values of as the important ones for explaining the
occurrence risk of alien plants at provincial level in China.
However, the above model is spatially implicit, for which it can not be able to partial out the influence of
spatial autocorrelation and may inflate Type I errors. As such, we implemented the alternative model which
explicitly incorporate the spatial influence in the model as,
log( ) log( ) X W
where W is the matrix with each element
ij
w =1 if provinces i and j are adjacent (i.e., they share some
boundaries), otherwise
ij
w =0. measures the strength of spatial autocorrelation for the response variable
(logtransformed relative risk). This is the formula of spatial simultaneous autoregression lag model (SAR)
(Dormann et al., 2007; Dormann, 2007). Implication of SAR model has been done using R (R Development
Core Team, 2011) package spdep (Bivand et al., 2013).
The following environmental variables at provincial level are used for fitting the above model (Qian, 2013):
mean annual temperature (TEM, C), mean temperature of the coldest month (TEMmin, C), mean
83
Computational Ecology and Software, 2014, 4(2): 8288
IAEES www.iaees.org
temperature of the warmest month (TEMmax, C), annual precipitation (PREC, mm), annual actual
evapotranspiration (AET, mm), annual potential evapotranspiration (PET, mm), the range of mean annual
temperature within a province (TEMrange, C), the range of annual precipitation within a province
(PRECrange, mm), and the range of elevation within a province (ELEVrange, m). These nine environmental
variables represent energy availability, water availability, seasonality, and habitat heterogeneity (Qian, 2013).
All these covariates are standardized before performing SAR modeling.
3 Results
As showed in Fig. 1, the woody plant diversity is highest in southern part of China, which was identified as
one of the global biodiversity hotspots of the world (Myers et al., 2000; Chen and Bi, 2007). Yunnan has the
highest woody plant richness, followed by Guangxi Province.
80 90 100 110 120 130
0
1
0
2
0
3
0
4
0
5
0
Fig. 1 Observed woody plant diversity at provincial level of China. Colors from light to dark grey indicate diversity from low to
high.
80 90 100 110 120 130
0
1
0
2
0
3
0
4
0
5
0
Fig. 2 Observed alien plant diversity at provincial level of China. Colors from light to dark grey indicate diversity from low to
high.
84
Computational Ecology and Software, 2014, 4(2): 8288
IAEES www.iaees.org
As shown in Fig. 2, the alien plant diversity is highest in southern and eastern part of China. Interestingly,
the biodiversity hotspot, Yunnan Province, also has very high alien plant diversity following Jiangsu Province
(Wang et al., 2009).
80 90 100 110 120 130
1
0
2
0
3
0
4
0
5
0
Fig. 3 Relative risk of plant invasiveness at provincial level of China using empirical Bayesian smoothing method. Provinces
which have red boundaries indicate that their relative risks of plant invasion are significantly higher than 1. Colors from light to
dark grey indicate invasive risks from low to high.
80 90 100 110 120 130
1
0
2
0
3
0
4
0
5
0
Fig. 4 Relative risk of plant invasiveness at provincial level of China using PoissonGamma model. Provinces which have red
boundaries indicate that their relative risks of plant invasion are significantly higher than 1. Colors from light to dark grey
indicate invasive risks from low to high.
85
Computational Ecology and Software, 2014, 4(2): 8288
IAEES www.iaees.org
The relative risk of plant invasion based on empirical Bayesian smoothing and PoissonGamma models are
highly similar (Figs. 34). As seen, northern and western parts of China have high risk of plant invasion (dark
grey colors). In contrast, southern part of China is less sensitive to plant invasion (light grey colors). Using
Poisson exact test, it is identified that 13 provinces from northern part of China have significant relative risk
values higher 1 (those provinces with red boundaries in Figs. 34).
As indicated by Table 1, the SAR model predicted that PRECrange was the most significant covariate
(P<0.05) predicting the relative risk of plant invasion across different provinces of China. The next two
important variables are TEMrange and PET respectively, but their coefficients are marginally significant
(P<0.1). The coefficient for quantifying the strength of spatial autocorrelation is equal to 0.7903, which is a
quite large value given that the interval for is [0, 1].
Before applying SAR model, there is significant spatial autocorrelation for the relative risk of plant
invasion with Morans I=0.639 (P<0.001) (Table 1 and Figs. 34). However, after SAR model has been done,
the spatial autocorrelation has been removed to a great extent as indicated by the nonsignificant Morans I
value for the residuals of the model (Morans I=0.131, P>0.05).
Table 1 Coefficient estimation for the covariates in the SAR model used for predicting relative risk of plant invasion across
different provinces of China. Here the relative risk values are from empirical Bayesian smoothing method, the result for Poisson
Gamma method is almost identical and thus not presented here.
Covariates Estimate Std. Error z value Pr(>z)
(Intercept) 0.083 0.049 1.695 0.09
Area 0.119 0.102 1.171 0.242
Elev 0.527 0.883 0.597 0.55
TEM 0.496 1.219 0.407 0.684
TEMmin 0.721 0.755 0.955 0.339
TEMmax 0.505 0.669 0.755 0.45
PET 0.81 0.486 1.668 0.095
PREC 0.384 0.375 1.026 0.305
AET 0.341 0.376 0.907 0.364
ElevRange 0.21 0.789 0.267 0.79
TEMrange 0.267 0.151 1.768 0.077
PRECrange 0.214 0.076 2.827 0.005
Morans I P value
Before SAR 0.639 0.001
After SAR 0.131 0.717
4 Discussion
Different from previous studies, which showed that areas with high relative outbreak risk of alien plants are
typically those with high richness of alien plants, the present study found that those areas with low native plant
diversity would have higher risk of alien plant occurrence. The remarkable difference on the risk assessment is
due to the inclusion of biodiversity information from native woody plant species. As mentioned above, all the
previous studies (Bai et al., 2013; Liu et al., 2005; Wu et al., 2006; Feng and Zhu, 2010) did not explicitly
consider the interaction between native and alien plant diversity. The biodiversity resistance hypothesis (Law
86
Computational Ecology and Software, 2014, 4(2): 8288
IAEES www.iaees.org
and Morton, 1996; Stachowicz et al., 2002) provides a way to effectively incorporate diversity information
from native plants for quantifying invasive risk of alien plants over provinces.
Based on the comparative results (Figs 34), both PoissonGamma and empirical Bayesian smoothing
methods consistently identified that provinces from northern part of China are remarkably higher than those
provinces in southern part of China. This is expected, since we applied biodiversity resistance hypothesis when
modeling the invasive risk of alien plants. In southern and southwestern part of China, the provinces there
have very high species richness in comparison to those provinces in northern part of China. In particular,
Yunnan province constitutes a well recognized biodiversity hotspot of the world (Chen and Bi, 2007; Myers et
al., 2000; Zhang et al., 2010; Hua, 2008; Zhang, 2012b). Thus, even though this province is found to have a
very high number of alien plants (Fig. 2), its relative risk of alien plant invasion is still comparatively low
(Figs. 34). The very reason is that the calculation of relative risk of alien plant invasion requires the
population at risk in the denominator, for which we use the native woody plant richness as the surrogate.
Yunnan Province has the highest woody plant diversity (Fig. 1), thus its relative risk is predicted to be low
accordingly.
Acknowledgements
This work is supported by China Scholarship Council.
References
Bai F, Chisholm R, Sang W, Dong M. 2013. Spatial risk assessment of alien invasive plants in China.
Environmental Science & Technology, 47: 76247632
Bivand R, Altman M, Anselin L . 2013. spdep: spatial dependence: weighting schemes, statistics and models.
http://CRAN.Rproject.org/package=spdep
Chen Y. 2013. Distributional patterns of alien plants in China: the relative importance of phylogenetic history
and functional attributes. ISRN Ecology, 527052
Chen Y, Bi J. 2007. Biogeography and hotspots of amphibian species of China: Implications to reserve
selection and conservation. Current Science, 92: 480489
Dormann C. 2007. Effects of incorporating spatial autocoorrelation into the analysis of species distribution
data. Global Ecology and Biogeography, 16: 129138
Dormann C, McPherson J , Araujo M, et al. 2007. Methods to account for spatial autocorrelation in the analysis
of species distirbutional data: a review. Ecography (Cop), 30: 609628
Feng J , Zhu Y. 2010. Alien invasive plants in China: risk assessment and spatial patterns. Biodiversity and
Conservation, 19: 34893497
Hua Z. 2008. Advances in biogeography of the tropical rain forest in southern Yunnan, southwestern China.
Tropical Conservation Science, 1: 3442
Law R, Morton R. 1996. Permanence and the assembly of ecological communities. Ecology, 77: 762775
Liu J , Liang S, Liu F, et al. 2005. Invasive alien plant species in China: regional distribution patterns. Diversity
and Distributions, 11: 341347
Myers N, Mittermeier R, Mittermeier C, et al. 2000. Biodiversity hotspots for conservation priorities. Nature,
403: 853858
Qian H. 2013. Environmental determinants of woody plant diversity at a regional scale in China. PLoS One, 8:
e75832
87
Computational Ecology and Software, 2014, 4(2): 8288
IAEES www.iaees.org
R Development Core Team. 2011. R: A Language and Environment for Statistical Computing. Vienna, Austria
Stachowicz J , Fried H, Osman R, Whitlatch R. 2002. Biodiversity, invasion resistance, and marine ecosystem
function: reconciling pattern and process. Ecology, 83: 25752590
Wang Z, Chen Y, Chen Y. 2009. Functional grouping and establishment of distribution patterns of invasive
plants in China using selforganizing maps and indicator species analysis. Archives of Biological Science,
61: 7178
Wu X, Luo J , Chen J , Li B. 2006. Spatial patterns of invasive alien plants in China and its relationship with
environmental and anthropological factors. J ournal of Plant Ecology, 30: 576584
Yan X, Shou H, Ma J . 2012. The problem and status of the alien invasive plants in China. Plant Diversity and
Resources, 34: 287313
Zhang D, Chen M, Murphy R, et al. 2010. Genealogy and palaeodrainage basins in Yunnan Province:
phylogeography of the Yunnan spiny frog, Nanorana yunnanensis (Dicroglossidae). Molecular Ecology,
19: 34063420
Zhang WJ . 2011. Constructing ecological interaction networks by correlation analysis: hints from community
sampling. Network Biology, 1(2): 8198
Zhang WJ . 2012a. Computational Ecology: Graphs, Networks and Agentbased Modeling. World Scientific,
Singapore
Zhang WJ . 2012b. Did Eucalyptus contribute to environment degradation? Implications from a dispute on
causes of severe drought in Yunnan and Guizhou, China. Environmental Skeptics and Critics, 1(2): 3438
Zhang WJ , Chen B. 2011. Environment patterns and influential factors of biological invasions: a worldwide
survey. Proceedings of the International Academy of Ecology and Environmental Sciences, 1(1): 114
88
Computational Ecology and Software, 2014, 4(2): 89103
IAEES www.iaees.org
Article
Stability analysis of a discrete ecological model
Q. Din
1
, E. M. Elsayed
2,3
1
Department of Mathematics, Faculty of Basic & Applied Sciences, University of PoonchRawalakot, Pakistan
2
Department of Mathematics, Faculty of Science,King Abdulaziz University, P.O. Box 80203, Jeddah 21589, Saudi Arabia
3
Department of Mathematics, Faculty of Science, Mansoura University, Mansoura 35516, Egypt
Email: qamar.sms@gmail.com,emmelsayed@yahoo.com
Received 14 January 2014; Accepted 20 February 2014; Published online 1 June 2014
Abstract
In this paper, we study the qualitative behavior of following discretetime population model:
x
n+1
= o + [x
n
+ yx
n1
c

n
, y
n1
= o + ey
n
+ y
n1
c
x
n
,
where parameters o, [, y, o, e, and initial conditions x
0
, x
1
, y
0
, y
1
are positive real numbers. More
precisely, we investigate the existence and uniqueness of positive equilibrium point,boundedness character,
persistence, local asymptotic stability, global behavior and rate of convergence of unique positive equilibrium
point of this model. Some numerical examples are given to verify our theoretical results.
Keywords population models; difference equations; steadystates; boundedness; local and global character.
1 Introduction
In population dynamics, difference and differential equations are used in many models (Nedorezov, 2012;
Nedorezov and Sadykov, 2012; Elena et al., 2013). Exponential difference equations have many applications
in population dynamics. One can see the references (ElMetwally et al., 2001; Papaschinopoulos et al., 2011;
Papaschinopoulos et al., 2012) for some interesting results related to qualitative behavior of population models.
From Zhou and Zou (2003) and Liu (2010) it is clear that difference equations are much better as compared to
differential equations, when the populations are of nonoverlapping generations. In literature, there are many
papers in which discrete dynamical systems are used to study the qualitative behavior of population models
(Ahmad, 1993; Zhou and Zou, 2003; Tang andZou, 2006; Din, 2013; Din and. Donchev, 2013; Din et al., 2013;
Din, 2014). In case of discretetime models one can compute mathematical results more efficiently. There are
many research papers which are related to mathematical models in population dynamics. During the last few
decades, many researchers are attracted by such computational models. For detail of biological models see
(EdelsteinKeshet, 1988; Brauer and CastilloChavez, 2000; Allen, 2007). Nonlinear difference equations are
ComputationalEcologyandSoftware
ISSN2220721X
URL:http://www.iaees.org/publications/journals/ces/onlineversion.asp
RSS:http://www.iaees.org/publications/journals/ces/rss.xml
Email:ces@iaees.org
EditorinChief:WenJunZhang
Publisher:InternationalAcademyofEcologyandEnvironmentalSciences
Computational Ecology and Software, 2014, 4(2): 89103
IAEES www.iaees.org
very important in application. Naturally, such equations appear as discrete approximation of differential and
delay differential equations and these models are related to applied sciences such as ecology,biology, physics,
chemistry, physiology, engineering, biochemistry and economics. We cannot find the solutions of nonlinear
difference equations in all cases. So, one can study the behavior of solutions by local asymptotic stability of
equilibrium points. ElMetwally et al. (2001) studied the qualitative behavior of following population
model:x
n+1
= o + [x
n1
c
x
n
. Papachinopoulous et al. (2011) investigated the dynamics of following
population model:x
n+1
= o + bx
n1
c

n
, y
n+1
= c +Jy
n1
c
x
n
. Recently, Papachinopoulous et al. (2012)
study following two exponential nonlinear difference equations: u
n+1
= o
1
+[
1
u
n1
c

n
,:
n+1
= y
1
+
o
1
u
n1
c
u
n
, andu
n+1
= o
1
+[
1
u
n1
c
u
n
, u
n+1
= y
1
+ o
1
u
n1
c

n
.
Our aim in this paper is to investigate existence and uniqueness of positive equilibrium point, boundedness
character, persistence, local asymptotic stability, global behavior of unique positive equilibrium point of
following fourdimensional discretetime population model:
x
n+1
= o + [x
n
+ yx
n1
c

n
, y
n+1
= o + ey
n
+ y
n1
c
x
n
, (1)
where parameters o, [, y, o, e, and initial conditions x
0
, x
1
, y
0
, y
1
are positive real numbers. System (1) is
an extension of (Papachinopoulouset al., 2011).
2 Boundedness and Persistence
The following Theorem shows that every positive solution of (1) is bounded and persists.
Theorem 2.1 Assume that c
6
[ + c
6 2
c
6
[
2
+ 4y < 2c
6
, c
u
e + c
u 2
c
u
e
2
+ 4 < 2c
u
, then every
positive solution {(x
n
, y
n
)} of system (1) is bounded and persists.
Proof. Consider the following secondorder difference equations:
x
n+1
= o + [x
n
+ yx
n1
c
6
, y
n+1
= o + ey
n
+y
n1
c
u
. (2.1)
The solutions of (2.1) are given by
x
n
=
o
1 [ c
6
y
+2
n
_c
6
_c
6
[  c
6 2
_c
6
[
2
+ 4y__
n
c
1
+ 2
n
_c
6
_c
6
[ + c
6 2
_c
6
[
2
+4y__
n
c
2
,
y
n
=
6
1cc
o
+ 2
n
_c
u
[c
u
e c
u 2
c
u
e
2
+4]
n
c
3
+ 2
n
_c
u
[c
u
e + c
u 2
c
u
e
2
+ 4]
n
c
4
,
wherec
1
, c
2
, c
3
, c
4
depend on initial values x
1
, x
0
, y
1
, y
0
.
Assume that c
6
[ +c
6 2
c
6
[
2
+4y < 2c
6
, c
u
e + c
u 2
c
u
e
2
+ 4 < 2c
u
. Then it follows that:
x
n
u
1[c
6
y
+c
1
+ c
2
and y
n
6
1cc
o
+ c
3
+ c
4
for all n = 1,2, . Evidently, the sequences {x
n
]
and {y
n
] are bounded for all n = 1,2, . Then, by comparison we obtain
x
n
u
1[c
6
y
andy
n
6
1cc
o
0
o +[
o
1 [ c
6
y
+y
o
1 [ c
6
y
c
6
=
o
1  [  c
6
y
,
and
y
1
= o + ey
0
+ y
1
c
x
0
o + e
o
1  e  c
u
+
o
1  e  c
u
=
o
1  e c
u
.
Hence, x
1
e I and y
1
e [. Suppose that result is true for n = k > 1, i.e., x
k
e I and y
k
e [. Then, for
n = k + 1 from system (1) one has:
x
k+1
= o + [x
k
+ yx
k1
c

k
o +[
o
1  [  c
6
y
+ y
o
1 [  c
6
y
c
6
=
o
1  [  c
6
y
,
y
k+1
= o + ey
k
+ y
k1
c
x
k
o +e
o
1  e  c
u
+
o
1  e  c
u
=
o
1  e c
u
s
Hence, the proof follows by induction.
4 Existence and Uniqueness of Positive Equilibrium and Local Stability
Theorem 4.1 Assume that c
6
[ + c
6 2
c
6
[
2
+ 4y < 2c
6
, c
u
e + c
u 2
c
u
e
2
+ 4 < 2c
u
, u < [, e< 1 and
exp_
s66 cxp_
o
1c
6
y
]
1s cxp_
o
1c
6
y
]
_ < y < (1  [) exp[
6
1sc
o
.
Then the system (1) has a unique positive equilibrium point (x , y) such that x e jo,
u
1[yc
6
[ = Iand
y e jo,
6
1sc
o
[ = [.
Proof. Consider the following system of equations:
x = o + [x + yxc

, y = o + ey + yyc
x
Then, one has x =
u
1[yc
j
and y =
6
1sc
x
. Taking F(x) =
u
1[yc
](x)
x, where (x) =
6
1sc
x
and
x e I. Then, it follows that
F(o) =
u([c
](o)+{
)
(1[)c
](o)

. Now, F(o) > u if and only if (1 [)c
](u)
 > u, i. c., < (1  [) exp[
6
1sc
o
.
Hence, it follows that F(o) > u if and only if < (1  [) exp [
6
1sc
o
.
Furthermore, we have
F[
u
1[yc
6
= o _
1
1[c
]_
o
1yc
6
_

1
1[yc
6
_,
91
Computational Ecology and Software, 2014, 4(2): 89103
IAEES www.iaees.org
where [
u
1[c
6
y
=
6
1sc

o
1yc
6
. It is easy to see that F [
u
1[yc
6
< u if and only if
c
]_
o
1yc
6
]
< yc
6
, i.e., exp_
s66 cxp_
o
1yc
6
]
1s cxp_
o
1yc
6
]
_ < y.
It follows that F [
u
1[c
6
y
< u, if and only if exp _
s66 cxp_
o
1yc
6
]
1s cxp_
o
1yc
6
]
_ < y.
Hence, F(x) has positive solution in I.
Moreover, we have that
JF
Jx
(x) = 1 
c
](x)
o'(x)
(c
](x)
(1 + [) + )
2
= 1 +
c
x+
c
x
6
c
x
(1+s)+{
u6
2
(c
x
(1+s)+)
2
(1+[+c
c
x
6
c
x(1+s)+{
)
2
< u.
Hence,F(x) has a unique positive solution in I. The proof is therefore completed.
Consider a fourthdimensional discrete dynamical system of the form:
x
n+1
= (x
n
, y
n
, x
n1
, y
n1
)
y
n+1
= g(x
n
, y
n
, x
n1
, y
n1
), n = u,1, ,
_ (4.1)
where: I
2
[
2
 I and g: I
2
[
2
 [ are continuously differentiable functions, and I, [ are real intervals.
Furthermore, a solution {(x
n
, y
n
)] of system (4.1) is uniquely determined by initial conditions (x
, y
) e I
[ for i e {1,u]. Along with system (4.1) we consider a vector map of the formF = (, g, x
n
, y
n
). A constant
solution or equilibrium point (fixedpoint) of (4.1) is point that satisfies:
x = (x , y, x , y),
y = g(x , y, x , y).
The point (x , y) is also point called a fixed point of the vector map F.
Definition 4.1 Assume that (x , y) be aconstant solution (steadystate) of (4.1).
1. A constant solution(x , y) is called stable, if for an arbitrary e > u there exists a positiveo such that
for any initial condition (x
, y
), i e {1, u] if  (x
0
=1
, y
, y
0
=1
)  (x , y) < p and (x
n
, y
n
)  (x , y) as n  .
4. A constant solution(x , y) is said to be global attracter if (x
n
, y
n
)  (x , y) as n  .
5. A constant solution (x , y) is said to be asymptotically global attractor if it is global attractor and stable.
Assume that(x , y) be a fixedpoint of map defined by F=(, x
n
, g, y
n
), where and g are continuously
differentiable functions about (x , y). The linearized system of (4.1)about the fixedpoint (x , y) is given by:
X
n+1
= F(X
n
) = F
]
X
n
,
whereX
n
= _
x
n
y
n
x
n1
y
n1
_ and F
]
is a J acobain matrix of (4.1) about the fixedpoint (x , y).
Lemma 4.1 (Sedaghat, 2003) Consider the discrete dynamical system of the form X
n+1
= F(X
n
), n = u,1, ,
such that X
lie inside an open unit disc z < 1, then the fixedpoint X
is
92
Computational Ecology and Software, 2014, 4(2): 89103
IAEES www.iaees.org
locally asymptotically stable.
(ii) If one of the eigenvalue of J acobian matrix [
P
has an absolute value greater than one, then X
is unstable.
To construct corresponding linearized form of system (1) we consider the following transformation:
(x
n
, y
n
, x
n1,
, y
n1
)  (, g,
1
, g
1
), (4.2)
where = x
n+1
, g = y
n+1
,
1
= x
n
and g
1
= y
n
. The J acobian matrix about the fixed point (x , y) of (1)
under the transformation (4.2) is given by
F
]
(x , y) = _
[ yx c

yc

u
yc
x
e u c
x
1
u
u
1
u
u
u
u
_.
Theorem 4.2The unique positive equilibrium point (x , y) of system (1) is locally asymptotically stable if the
following condition is satisfied:
[ + e +[e + (1 + [)c
u
+ y(1 +e)c
6
+yc
u6
_1 +
u6
(1[c
6
y)(1sc
o
)
] < 1. (4.3)
Proof. The characteristic polynomial of J acobain matrix F
]
(x , y) about (x , y) is given by
P(z) = z
4
 ([ +e)z
3
 Az
2
+ Bz +yc
x 
, (4.4)
where A = c
x
+ yc

+ yx yc
x 
[e and B = [c
x
+yec

.
Let (z) = z
4
and (z) = ([ + e)z
3
+Az
2
 Bz  yc
x 
.
Assume that condition (4.3) is satisfied and z = 1, then one has
(z) < [ +e + A + B + yc
x 
= [ +e + c
x
+yc

+ yx yc
x 
[e
+ [c
x
+ yec

+ yc
x 
< [ + e +c
x
+yc

+yx yc
x 
+ [e
+ [c
x
+ yec

+ yc
x 
< [ + e +[e + (1 + [)c
u
+ y(1 +e)c
6
+yc
u6
_1 +
u6
(1[c
6
y)(1sc
o
)
] < 1.
Then, by Rouches theorem (z) and (z)  (z) have same number of zeroes in an open unit disc
z < 1. Hence, all the roots of (4.3) satisfies z < 1, and it follows from Lemma 4.1 that the unique positive
equilibrium point (x , y) of system (1) is locally asymptotically stable.
5 Global Stability
Theorem 5.1 The unique positive equilibrium point (x , y) of system (1) is globally asymptotically stable.
Proof. Let{(x
n
, y
n
)] be an arbitrary positive solution of system (1) and let u <
1
= lim
n
in x
n
, u <
2
=
lim
n
in y
n
, S
1
= lim
n
sup x
n
< , S
2
= lim
n
sup y
n
< . Then, from (1) one has:
1
o +[
1
+y
1
c
S
2
2
o +[
2
+y
2
c
S
1
S
1
o +[S
1
+yS
1
c

2
S
2
o +[S
2
+yS
2
c

1
Furthermore,
1
o
1  [  yc
S
2
,
2
o
1 e  c
S
1
,
93
Computational Ecology and Software, 2014, 4(2): 89103
IAEES www.iaees.org
S
1
o
1  [  yc

2
,
S
2
o
1  e  c

1
.
Then, it follows that:
1
S
2
oS
2
1  [  yc
S
2
,
2
S
1
oS
1
1  e  c
S
1
,
2
S
1
o
2
1  [  yc

2
,
1
S
2
o
1
1  e  c

1
.
Hence one has
uS
2
1[yc
S
2
6
1
1cc

1
and
6S
1
1cc
S
1
u
2
1[yc

2
. Now consider the following functions:
(x) =
ox
1  e c
x
onJg(y) =
oy
1  [  yc

, xe, y }.
Moreover, for every (x, y)e [:
J
Jx
(x) = 
c
x
xo
(1  e c
x
)
2
+
o
(1  e  c
x
)
> u,
Jg
Jy
(y) = 
c

yoy
(1 [ c

y)
2
+
o
(1 [ c

y)
> u.
Hence,
1
= S
1
and
2
= S
2
.
6 Existence of Unbounded Solutions
Theorem 6.1 Let [, e e (u,1) and {(x
n
, y
n
)] be a positive solution of system (3.1) Then following are true:
(i) If y > c
6
1e
, tbcn x
n
 , y
n
 o +eIn(y)os n  .
(ii) If > c
o
1y
, tbcn x
n
 o +[In(), y
n
 os n  .
Proof. (i) Let y > c
6
1e
. Choosing the initial conditions x
0
, x
1
, y
0
, y
1
for the system (1) such that
x
0
, x
1
> o and y
0
, y
1
< b with = In
b
b6cb
, b = Iny. Then from system (1) one has:
y
1
= o + y
1
ey
0
+ y
1
c
x
0
< o +eb +bc
u
= b,
and
x
1
= o + [x
0
+ yx
1
c
0
> o +[o + yoc
b
= o + o[ +o > o.
Suppose that result is true for n = k > 1, i.e., x
k
> oand y
k
< b. Then for n = k + 1 from system (1)
one has:
x
k+1
= o +[x
k
+yx
k1
c

k
> o +[o +yoc
b
= o + o[ + o > o,
and
y
k+1
= o + ey
k
+y
k1
c
x
k
< o +eb +bc
u
= b.
Hence, x
n
> o, y
n
< b for all n = 1,2, . Furthermore, x
n+1
= o + [x
n
+ yx
n1
c

n
> o + [x
n
+
yx
n1
c
b
= o +[x
n
+ x
n1
. Let z
n+1
= o +[z
n
+z
n1
with initial conditions z
1
, z
0
. Then,z
n
= 
u
[
+
94
Computational Ecology and Software, 2014, 4(2): 89103
IAEES www.iaees.org
2
n
([  4 +[
2
)
n
z
0
+ 2
n
([ + 4 +[
2
)
n
z
1
. Hence by comparison x
n
> z
n
. But z
n
 osn  , so
lim
n
x
n
= . Moreover, o y
n+1
= o + ey
n
+y
n1
c
x
n
< o + eb + bc
x
n
. Hence, lim
n
y
n
= o +
eIn(y).
(ii) The proof of (ii) is similar, therefore it is omitted.
7 Rate of Convergence
To discuss the rate of convergence of discrete dynamical system, we consider the following system:
Z
n+1
= (K + I(n))Z
n
, (7.1)
whereZ
n
is a kdimensional vector, K e C
kk
is a constant matrix, and I: L
+
 C
kk
is a matrix function
such that:
[I(n)[  u, (7.2)
asn  , where [. [ denotes a matrix norm which associated with the vector norm of the form:
[(u, :)[ = u
2
+:
2
.
Lemma 7.1 (Perrons theorem) (Pituk, 2002) Assume that the condition (7.2) holds true. If Z
n
be a solution
of (7.1) then either Z
n
= u for alln  , or
= lim
n
([Z
n
[)
1
n
,
is defined and it is equal to the absolute value of one of the eigenvalues of matrix K.
Lemma 7.2 (Pituk, 2002) Assume that condition (7.2) holds. If Z
n
is a solution of (7.1), then either Z
n
= u
for alln  , or
= lim
n
[z
n+1
[
[z
n
[
,
is defined and is equal to the absolute value of one the eigenvalues of matrix K.
Let {(x
n
, y
n
)] be any solution of the system (1) such that lim
n
x
n
= x and lim
n
y
n
= y, where
x e jo,
u
1[yc
6
[ and y e jo,
6
1[yc
6
[. To find the error terms, one has from the system (1)
x
n+1
x = [x
n
+ yx
n1
c

n
[x  yx c

= [(x
n
 x ) +
yx (c

n
 c

)
y
n
y
(y
n
 y) + yc

n
(x
n1
 x ),
and
y
n+1
y = ey
n
+ y
n1
c
x
n
 ey yc
x
=
(c
x
n
c
x
)
x
n
x
(x
n
 x ) +e(y
n
 y) + c
x
n
(y
n1
y).
Let c
n
1
= x
n
x , and c
n
2
= y
n
 y, then one has
c
n+1
1
= o
n
c
n
1
+b
n
c
n
2
+ c
n
c
n1
1
+ J
n
c
n1
2
,
and
c
n+1
2
=
n
c
n
1
+ g
n
c
n
2
+ b
n
c
n1
1
+ k
n
c
n1
2
,
where
95
Computational Ecology and Software, 2014, 4(2): 89103
IAEES www.iaees.org
o
n
= [, b
n
=
yx (c

n
c

)
y
n
y
, c
n
= yc

n
, J
n
= u,
n
=
y(c
x
n
 c
x
)
x
n
x
, g
n
= c
x
n
,
b
n
= u, k
n
= e.
Moreover,
lim
n
o
n
= [ , lim
n
b
n
= yx c

, lim
n
c
n
= yc

, lim
n
J
n
= u, lim
n
n
= yc
x
,
lim
n
g
n
= e, lim
n
b
n
= u, lim
n
k
n
= c
x
.
Now the limiting system of error terms can be written as
l
l
l
l
l
c
n+1
1
c
n+1
2
c
n
1
c
n
2
1
1
1
1
1
= _
[ yx c

yc

u
yc
x
e u c
x
1
u
u
1
u
u
u
u
_
l
l
l
l
l
c
n
1
c
n
2
c
n1
1
c
n1
2
1
1
1
1
1
,
which is similar to linearized system of (1) about the equilibrium point (x , y).
Using Lemma 7.1 one has the following result.
Theorem 7.1 Assume that {(x
n
, y
n
)] be a solution of (1) such that lim
n
x
n
= x , lim
n
y
n
= y, where
(x , y) be unique equilibrium point of (1), then the error vector E
n
=
l
l
l
l
l
c
n+1
1
c
n
1
c
n+1
2
c
n
2
1
1
1
1
1
of every solution of system (1)
satisfies both of the following asymptotic relations:
lim
n
([E
n
[)
1
n
,
= z
1,2,3,4
F
]
(x , y),
lim
n
[E
n+1
[
[E
n
[
= z
1,2,3,4
F
]
(x , y),
wherez
1,2,3,4
F
]
(x , y) are the characteristic roots of the J acobian matrixF
]
(x , y).
8 Examples
In this section, we consider the following numerical examples.
Example 8.1 Let o = 1.1, [ = u.u1, y = 2.9, o = 1.2, e = u.u2, = 2.Su4.Then, system (1) can be written
as
x
n+1
= 1.1 +u.u1x
n
+ 2.9x
n1
c

n
, y
n+1
= 1.2 + u.u2y
n
+ 2.9y
n1
c
x
n
(8.1)
with initial conditions x
1
= 1.S, x
0
= 1.2, y
1
= 1.S, y
0
= 1.S.
In this case the unique positive equilibrium point of the system (8.1)is given by
(x , y) = (7.9S981,1.22SS2). Moreover, in Fig. 8.1 the plot of x
n
is shown in Fig. 8.1a, the plot of y
n
is
shown in Fig. 8.1b, and an attractor of the system is shown in Fig. 8.1c.
96
Computational Ecology and Software, 2014, 4(2): 89103
IAEES www.iaees.org
(8.1a) Plot of x
n
for the system (8.1)
(8.1b) Plot of y
n
for the system (8.1)
97
Computational Ecology and Software, 2014, 4(2): 89103
IAEES www.iaees.org
(8.1c) An attractor of the system (8.1)
Fig. 8.1 Plots for the system (8.1).
Example 8.2 Let o = 1.S, [ = u.u26, y = S.14, o = 1.67, e = u.uu96, = 2.6. Then, system (1) can be
written as
x
n+1
= 1.S +u.u26x
n
+S.14x
n1
c

n
,
y
n+1
= 1.67 + u.uu96y
n
+ 2.6y
n1
c
x
n
(8.2)
with initial conditions x
1
= 2.2, x
0
= 2.1, y
1
= 2.SS, y
0
= 2.2S.
In this case the unique positive equilibrium point of the system (8.2) is given by
(x , y) = (2.164SS,2.41S79). Moreover, in Fig. 8.2 the plot of x
n
is shown in Fig. 8.2a, the plot of y
n
is
shown in Fig. 8.2b, and an attractor of the system is shown in Fig. 8.2c.
(8.2a) Plot of x
n
for the system (8.2)
98
Computational Ecology and Software, 2014, 4(2): 89103
IAEES www.iaees.org
(8.2b) Plot of y
n
for the system (8.2)
(8.2c) An attractor of the system (8.2)
Fig.8.2 Plots for the system (8.2)
Example 8.3 Let o = 1.S, [ = u.uuu1, y = S.S, o = 1.67, e = u.uuuu2, = 2.77. Then, system (1) can be
written as
x
n+1
= 2.S + u.uuu1x
n
+ S.Sx
n1
c

n
y
n+1
= 1.67 + u.uuuu2y
n
+ 2.77y
n1
c
x
n
(8.3)
with initial conditions x
1
= 2.2S, x
0
= 2.1S, y
1
= 2.S6, y
0
= 2.27. In this case the unique positive
equilibrium point of the system (8.3) is given by (x , y) = (1.71494,S.SSuSS).
99
Computational Ecology and Software, 2014, 4(2): 89103
IAEES www.iaees.org
Moreover, in Fig. 8.3 the plot of x
n
is shown in Fig. 8.3a, the plot of y
n
is shown in Fig. 8.3b, and an
attractor of the system is shown in Fig. 8.3c.
(8.3a) Plot of x
n
for the system (8.3)
(8.3b) Plot of y
n
for the system (8.3)
100
Computational Ecology and Software, 2014, 4(2): 89103
IAEES www.iaees.org
(8.3c) An attractor of the system (8.3)
Fig. 8.3 Plots for the system (8.3).
Example 8.4 Let o = 2.S, [ = u.uuu9, y = 4.S, o = 2.u9, e = u.uuu8, = 6.SS. Then system (1) can be
written as:
x
n+1
= 2.S + u.uuu9x
n
+ 4.Sx
n1
c

n
,
y
n+1
= 2.u9 + u.uuu8y
n
+ 6.SSy
n1
c
x
n
(8.4)
with initial conditions x
1
= 2.1, x
0
= 2.2, y
1
= 1.1, y
0
= 1.S. In this case the unique positive equilibrium
point of the system (8.4) is given by (x , y) = (2.SS81,S.24482). Moreover, in Fig. 8.4 the plot of x
n
is
shown in Fig. 8.4a,the plot of y
n
is shown in Fig. 8.4b,and an attractor of the system (8.4) is shown in Fig.
8.4c.
(8.4a) Plot of x
n
for the system (8.4)
101
Computational Ecology and Software, 2014, 4(2): 89103
IAEES www.iaees.org
(8.4b) Plot of y
n
for the system (8.4)
(8.4c) An attractor of the system (8.4)
Fig. 8.4 Plots for the system (8.4).
9 Conclusion
This work is related to qualitative behavior of a fourdimensional exponential discrete population model. We
prove that system (1) has a unique positive equilibrium point which is globally asymptotically stable. Usually,
biologists believe that a globally asymptotically stable equilibrium point is very important in ecological point
of view. Method of linearization is used for local asymptotic stability of steadystate of (1). Furthermore, we
prove the boundedness and persistence of positive solutions of (1), and an invariant set for its solutions is
102
Computational Ecology and Software, 2014, 4(2): 89103
IAEES www.iaees.org
investigated. We prove the rate of convergence of positive solutions of system (1) which converge to its
unique positive equilibrium point.
References
Ahmad S. 1993. On the nonautonomous LotkaVolterra competition equation. Proceedings of the American
Mathematical Society, 117: 199204
Allen LJ S. 2007. An Introduction to Mathematical Biology. Prentice Hall, USA
Brauer F, CastilloChavez C. 2000.Mathematical Models in Population Biology and Epidemiology. Springer
Din Q. 2013.Dynamics of a discrete LotkaVolterra model. Advances in DifferenceEquations, 1: 113
Din Q. 2014.Global stability of a population model. Chaos, Soliton and Fractals, 59: 119128.
Din Q, Donchev T. 2013.Global character of a hostparasite model. Chaos, Soliton and Fractals, 54: 17
Din Q, Khan AQ, Qureshi MN. 2013. Qualitative behavior of a hostpathogen model. Advances in Difference
Equations, 1: 263
EdelsteinKeshet L. 1988. Mathematical Models in Biology. McGrawHill, USA
Elaydi S. 2005. An Introduction to Difference Equations (3rd ed). SpringerVerlag, New York, USA
Elena E, Grammauro M, Venturino E. 2013. Predators alternative food sources do not support ecoepidemics
with two strainsdiseased prey. Network Biology, 3(1): 2944
ElMetwally E, Grove EA, Ladas G, et al. 2001.On the difference equation x
n+1
= o + [x
n1
c
x
n
.
Nonlinear Analysis, 47: 46234634
Grove EA, Ladas G. 2004. Periodicities in Nonlinear Difference Equations. Chapman and Hall/CRC Press,
Boca Raton, USA
Liu X. 2010. A note on the existence of periodic solution in discrete predatorprey models. Applied
Mathematical Modelling, 34: 24772483
Nedorezov LV. 2012. Continuousdiscrete model of population dynamics with time lag in a reaction of intra
population selfregulative mechanisms. Network Biology, 2(4):139147
Nedorezov LV, Sadykov AM. 2012. About a modification of Rogers model of parasitehost system dynamics.
Proceedings of the International Academy of Ecology and Environmental Sciences, 2(1): 4149
Papaschinopoulos G, Radin MA, Schinas CJ . 2011. On the system of two difference equations of exponential
form: x
n+1
= o +bx
n1
c

n
, y
n+1
= c + Jy
n1
c
x
n
. Mathematical and Computer Modelling, 54:
29692977
Papaschinopoulos G, Schinas CJ . 2012. On the dynamics of two exponential type systems of difference
equations. Computers & Mathematics with Applications, 64(7): 23262334
Pituk M. 2002.More on Poincares and Perrons theorems for difference equations. J ournal of Difference
Equations and Applications, 8: 201216
Sedaghat H. 2003. Nonlinear difference equations: Theory with applications to social science models. Kluwer
Academic Publishers, Dordrecht, Netherlands
Tang X, Zou X. 2006.On positive periodic solutions of LotkaVolterra competition systems withdeviating
arguments. Proceedings of the American Mathematical Society, 134: 29672974
Zhou Z, Zou X. 2003. Stable periodic solutions in a discrete periodic logistic equation. Applied Mathematics
Letters, 16(2): 165171
103
Computational Ecology and Software, 2014, 4(2): 104115
IAEES www.iaees.org
Article
Qualitative behavior of an anticompetitive system of thirdorder
rational difference equations
Q. Din
1
, M. N. Qureshi
2
, A. Q. Khan
2
1
Department of Mathematics, Faculty of Basic & Applied Sciences, University of PoonchRawalakot, Pakistan
2
Department of Mathematics, University of Azad J ammu & Kashmir, Muzaffarabad, Pakistan
Email: qamar.sms@gmail.com,nqureshi@ajku.edu.pk,abdulqadeerkhan1@gmail.com
Received 6 December 2013; Accepted 10 January 2014; Published online 1 June 2014
Abstract
In this paper, our aim is to study the equilibrium points, local asymptotic stability, global behavior of an
equilibrium points and rate of convergence of an anticompetitive system of thirdorder rational difference
equations of the form:
x
n+1
=
u
n2
[+yx
n
x
n1
x
n2
, y
n+1
=
u
1
x
n2
[
1
+y
1
n1
n2
, n = u,1, ,
where the parameters o, [, y, o
1
, [
1
, y
1
and initial conditions x
0
, x
1
, x
2
, y
0
, y
1
, y
2
are positive real
numbers. Some numerical examples are given to verify our theoretical results.
Keywords rational difference equations; stability; global character; rate of convergence.
1 Introduction
Systems of rational difference equations have been studied by several authors. Especially there has been a
great interest in the study of the attractivity of the solutions of such systems.Cinar (2004) investigated the
periodicity of the positive solutions of the system of rational difference equations:
x
n+1
=
1
n
, y
n+1
=
n
x
n1
n1
.
Stevic (2012) studied the system of two nonlinear difference equations:
x
n+1
=
u
n
1+
n
, y
n+1
=
w
n
1+s
n
,
whereu
n
, :
n
, w
n
, s
n
are some sequencesx
n
ory
n
. Stevic(2012) studied the system of three nonlinear difference
equations:
x
n+1
=
u
1
x
n2
b
1
n
z
n1
x
n2
+c
1
, y
n+1
=
u
2
n2
b
2
z
n
x
n1
n2
+c
2
, z
n+1
=
u
3
z
n2
b
3
x
n
n1
z
n2
+c
3
,
ComputationalEcologyandSoftware
ISSN2220721X
URL:http://www.iaees.org/publications/journals/ces/onlineversion.asp
RSS:http://www.iaees.org/publications/journals/ces/rss.xml
Email:ces@iaees.org
EditorinChief:WenJunZhang
Publisher:InternationalAcademyofEcologyandEnvironmentalSciences
Computational Ecology and Software, 2014, 4(2): 104115
IAEES www.iaees.org
where the parameters o
, b
, c
, i e {1,2,S] are real numbers.Bajo and Liz (2011) investigated the global
behavior of difference equation:
x
n+1
=
x
n1
u+bx
n1
x
n
,
for all values of real parameterso, b.Touafek and Essayed(2012) studied the periodic nature and got the form of
the solutions of the following systems of rational difference equations:
x
n+1
=
x
n3
_1_x
n3
n1
, y
n+1
=
n3
_1_
n3
x
n1
.
Recently, Zhang et al. (2012) studied the dynamics of a system of rational thirdorder difference equation:
x
n+1
=
x
n2
B +y
n
y
n1
y
n2
, y
n+1
=
y
n2
B +x
n
x
n1
x
n2
, n = u,1, .
Our aim in this paper is to investigate the qualitative behavior of positive solution for an anticompetitive
system of thirdorder rational difference equations:
x
n+1
=
u
n2
[+yx
n
x
n1
x
n2
, y
n+1
=
u
1
x
n2
[
1
+y
1
n1
n2
, n = u,1, (1)
whereo, [, y, o
1
, [
1
, y
1
and initial conditions x
0
, x
1
, x
2
, y
0
, y
1
, y
2
are positive real numbers.
Let us consider sixdimensional discrete dynamical system of the form:
x
n+1
= (x
n
, x
n1
, x
n2
, y
n
, y
n1
, y
n2
)
y
n+1
= g(x
n
, x
n1
, x
n2
, y
n
, y
n1
, y
n2
)
_ (2)
n = u,1, , where: I
3
[
3
 I and g: I
3
[
3
 [are continuously differentiable functions and I, [ are
some intervals of real numbers. Furthermore, a solution {(x
n
, y
n
)]
n=2
, y
) e I [ fori e {2, 1,u] . Along with the system (2), we consider the
corresponding vector map F = (, x
n
, x
n1
, x
n2
, g, y
n
, y
n1
, y
n2
). An equilibrium point of system (2) is a
point (x , y) that satisfies
x = (x , x , x , y, y, y)
y = g(x , x , x , y, y, y).
The point (x , y) is also called a fixed point of the vector map F.
Definition 1. Let (x , y) be an equilibrium point of the system (2).
(i) An equilibrium point (x , y) is said to be stable if for every e > u there exists o > u such thatfor
every initial conditions (x
, y
), i e {2, 1,u] if [ (x
, y
)
0
=2
(x , y)[ < o implies [(x
n
, y
n
) 
(x , y)[ < e for all n > u, where[[ is usual Euclidian norm in R
2
.
(ii) An equilibrium point (x , y) is said to be unstable if it is not stable.
(iii) An equilibrium point (x , y) is said to be asymptotically stable if there exists p > u such that
[ (x
, y
)
0
=2
(x , y)[ < p and(x
n
, y
n
)  (x , y) as n  .
(iv) An equilibrium point (x , y) is called global attractor if (x
n
, y
n
)  (x , y) as n  .
(v) An equilibrium point (x , y) is called asymptotic global attractor if it is a global attractor andstable.
Definition 2. Let(x , y) be an equilibrium point of a map
F = (, x
n
, x
n1
, x
n2
, g, y
n
, y
n1
, y
n2
)
where and gare continuously differentiable functions at(x , y). The linearized system of (2) about the
equilibrium point (x , y) is given by
X
n+1
= F(X
n
) = F
]
X
n
,
105
Computational Ecology and Software, 2014, 4(2): 104115
IAEES www.iaees.org
whereX
n
=
`
x
n
x
n1
x
n2
y
n
y
n1
y
n2
/
and F
]
is J acobean matrix of the system (2) about the equilibrium point (x , y).
To construct corresponding linearized form of the system (1) we consider the following transformation:
(x
n
, x
n1
, x
n2
, y
n
, y
n1
, y
n2
)  (,
1
,
2
, g, g
1
, g
2
) (3)
where =
u
n2
[+yx
n
x
n1
x
n2
, g =
u
1
x
n2
[
1
+y
1
n1
n2
,
1
= x
n
,
2
= x
n1,
g
1
= y
n
, g
2
= y
n1
. The J acobian matrix
about the fixed point (x , y) under the transformation (3) is given by
F
]
(x , y) =
`
A
1
u
A
u
1
A u u B
u u u u
u u u u
u u C
u
u
u
u
u 1 u u
u u 1 u
/
.
whereA = 
uyx
2
([+yx
3
)
2
, B =
u
[+yx
3
, C =
u
1
[
1
+y
1
3
and = 
u
1
y
1
x
2
([
1
+y
1
3
)
2
.
Theorem 1. (Sedaghat, 2003). For the system X
n+1
= F(X
n
), n = u,1, , of difference equations such
that X
is locally asymptotically stable. If one of them has a modulus greater than one, then X
is
unstable.
2 Main Results
Let (x , y)be an equilibrium point of the system (1), then system (1) has only(u,u)equilibrium point.
Theorem 2. Let (x
n
, y
n
) be positive solution of system (1), then for every m u the following results hold.
(i) u x
n
`
1
1
1
1
_
o
[
o
1
[
1
]
m+1
x
p
, n = 6m +(6 p),p = u,1,2,
_
o
[
]
m+1
_
o
1
[
1
]
m
y
3p
, n = 6m +(6 p), p = S,4,S.
(ii) u y
n
`
1
1
1
1
_
o
[
o
1
[
1
]
m+1
y
p
, n = 6m +(6 p),p = u,1,2,
_
o
[
]
m
_
o
1
[
1
]
m+1
x
3p
, n = 6m + (6 p), p = S,4,S.
Proof. (i)The result is obviously true for m = u. Suppose that results are true for m = k 1, i. c.,
u x
n
`
1
1
1
1
_
o
[
o
1
[
1
]
k+1
x
p
, n = 6k +(6  p),p = u,1,2,
_
o
[
]
k+1
_
o
1
[
1
]
k
y
3p
, n = 6k +(6  p), p = S,4,S.
and
u y
n
`
1
1
1
1
_
o
[
o
1
[
1
]
k+1
y
p
, n = 6k +(6 p),p = u,1,2,
_
o
[
]
k
_
o
1
[
1
]
k+1
x
3p
, n = 6k +(6  p), p = S,4,S.
Now, for m = k +1 one has
106
Computational Ecology and Software, 2014, 4(2): 104115
IAEES www.iaees.org
u x
6k+(12p)
=
u
6k+(9p)
[+y x
10r+1
11p
i=9p
u
6k+(9p)
[
[
u
[
k+2
[
u
1
[
1
k+1
y
3p
, p = S,4,S,
and
u x
6k+(12p)
[
u
[
u
1
[
1
k+2
x
p
, p = u,1,2.
So, the induction is complete. Similarly, inequalities (ii) hold by induction.
Theorem 3. For the equilibrium point(u,u) of the system (1) the following results hold:
(i)If o < [ and o
1
< [
1
, then equilibrium point (u,u) of the system (1) is locally asymptotically stable.
(ii)If o > [oro
1
> [
1
, then equilibrium point (u,u) is unstable.
Proof. (i).The linearized system of (1) about the equilibrium point (u,u) is given by
X
n+1
= F
]
(u,u)X
n
,
whereX
n
=
`
x
n
x
n1
x
n2
y
n
y
n1
y
n2
/
and F
]
(u,u) =
`
u
1
u
u
u
1
u u
u
u
[
u u u u
u u u u
u u
u
1
[
1
u u u
u
u
u
u
u 1 u u
u u 1 u
/
.
The characteristic polynomial of F
]
(P
0
) is given by
P(z) = z
6

uu
1
[[
1
. (4)
The roots of P(z) are
z
k
= [
uu
1
[[
1
1
6
exp [
kn
3
,
for k = u,1, ,S. Now, it is easy to see that z
K
 < 1 for all k = u,1, ,S. Since all eigenvalues of J acobian
matrix F
]
(u, u) about (0, 0) lie in open unit disk z < 1. Hence, the equilibrium point (0, 0) is locally
asymptotically stable.
(ii). It is easy to see that if o > [ or o
1
> [
1
, then there exist at least one root z of equation (4) such that
z > 1. Hence by Theorem 1 if o > [ or o
1
> [
1
, then (0, 0) is unstable.
Theorem 4. The system (1) has no prime periodtwo solutions.
Proof. Assume that (p
1
, q
1
), (p
2
, q
2
), (p
1
, q
1
), , be prime period two solution of system (1) such that
p
, q
u
[
y
n2
< y
n2
.
This implies that x
6n+1
< y
6n2
and x
6n+7
< y
6n+4
. Also
y
n+1
=
o
1
x
n2
[
1
+y
1
y
n
y
n1
y
n2
u
1
[
1
x
n2
< x
n2
.
This implies that y
6n+1
< x
6n2
and y
6n+7
< x
6n+4
. So x
6n+7
< y
6n+4
< x
6n+1
and y
6n+7
< x
6n+4
<
y
6n+1.
Hence, the subsequences {x
6n+1
], {x
6n+2
], {x
6n+3
], {x
6n+4
], {x
6n+5
], {x
6n+6
] and
{y
6n+1
], {y
6n+2
], {y
6n+3
], {y
6n+4
], {y
6n+5
], {y
6n+6
] are decreasing. Therefore the sequences {x
n
] and {y
n
]
are decreasing. Hencelim
n
x
n
= lim
n
y
n
= u.
3 Rate of Convergence
In this section we will determine the rate of convergence of a solution that converges to the unique equilibrium
point of the system (1). The following results give the rate of convergence of solutions of a system of
difference equations
X
n+1
= (A + B(n))X
n
, (7)
whereX
n
is an m dimensional vector, A e C
mm
is a constant matrix, and B: Z
+
 C
mm
is a matrix
function satisfying
[B(n)[  u (8)
as n  , where [[ denotes any matrix norm which is associated with the vector norm
[B(n)[ = x
2
+ y
2
.
Proposition 1 (Perrons theorem) (Pituk, 2002) Suppose that condition (8) holds. If X
n
is a solution of (7),
then either X
n
= u for all large n or
p = lim
n
([X
n
[)
1
n
(9)
or
p = lim
n
[X
n+1
[
[X
n
[
(10)
exist and is equal to the modulus of one the eigenvalues of matrix A.
Assume that lim
n
x
n
= x , lim
n
y
n
= y.Then error terms for the system (1) are given by
x
n+1
 x = A
(x
n
 x )
2
=0
+ B
(y
n
 y)
2
=0
,
y
n+1
 y = C
(x
n
 x )
2
=0
+
(y
n
 y)
2
=0
.
Set c
n
1
= x
n
 x andc
n
2
= y
n
 y, one has
c
n+1
1
= A
2
=0
c
n
1
+ B
2
=0
c
n
2
, c
n+1
2
= C
2
=0
c
n
1
+
2
=0
c
n
2
.
where
A
0
= 
uy x
ni
2
i=1
([+y x
ni
2
i=0
)([+yx
3
)
, A
1
= 
uyx x
n2
([+y x
ni
2
i=0
)([+yx
3
)
, A
2
= 
uyx
2
([+y x
ni
2
i=0
)([+yx
3
)
,
108
Computational Ecology and Software, 2014, 4(2): 104115
IAEES www.iaees.org
B
= u for i e {u,1],B
2
=
u
[+y x
ni
2
i=0
,C
= ufor i e {u,1], C
2
=
u
1
[
1
+y
1
ni
2
i=0
,
0
= 
u
1
y
1
x
ni
2
i=1
([
1
+y
1
ni
2
i=0
)([
1
+y
1
3
)
,
1
= 
u
1
y
1
x
n2
([
1
+y
1
ni
2
i=0
)([
1
+y
1
3
)
,
2
=
u
1
y
1
x
2
([
1
+y
1
ni
2
i=0
)([
1
+y
1
3
)
.
Taking the limit, we obtain lim
n
A
== 
uyx
2
([+yx
3
)
2
for i e {u,1,2] , lim
n
B
= u for i e {u,1] ,
lim
n
B
2
=
u
[+yx
3
, lim
n
C
3
and lim
n
== 
u
1
y
1
x
3
([
1
+y
1
3
)
2
for
i e {u,1,2]. So, the limiting system of error terms can be written as E
n+1
= F
]
(P
0
)E
n
,whereE
n
=
`
c
n
1
c
n1
1
c
n2
1
c
n
2
c
n1
2
c
n2
2
/
.
Using the Proposition (1), one has following result.
Theorem 6. Assume that {(x
n
, y
n
)] be a positive solution of the system (1) such that lim
n
x
n
= x and
lim
n
y
n
= y where (x , y) = (u,u). Then, the error term E
n
of every solution of (1) satisfies both of the
following asymptotic relations
lim
n
([c
n
[)
1
n
= zF
]
(x , y), lim
n
[c
n+1
[
[c
n
[
= zF
]
(x , y),
wherezF
]
(x , y) are the characteristic roots of J acobian matrix F
]
(x , y)about(u,u) .
4 Examples
In order to verify our theoretical results we consider several interesting numerical examples in this section.
These examples represent different types of qualitative behavior of solutions to the system of nonlinear
difference equations (1). All plots in this section are drawn with Mathematica.
Example 1. Consider the system (1) with initial conditionsx
2
= 1.9 , x
1
= u.9, x
0
= u.8, y
2
= 1.9 , y
1
=
1.S, y
0
= 1.9 . Moreover, choosing the parameters o = S1, [ = S2, y = 1.9, o
1
= S21, [
1
= S22, y
1
=
1.8.Then , the system (1) can be written as
x
n+1
=
31
n2
32+1.9x
n
x
n1
x
n2
, y
n+1
=
521x
n2
522+1.8
n
n1
n2
, n = u,1, , (11)
and with initial condition x
2
= 1.9 , x
1
= u.9, x
0
= u.8, y
2
= 1.9 , y
1
= 1.S, y
0
= 1.9. . Moreover, in
Fig. 1 the plot of x
n
is shown in Fig. 1a, the plot of y
n
is shown in Fig. 1b and an attractor of the system (9)
is shown in Fig. 1c.
Example 2. Consider the system (1) with initial conditionsx
2
= S.S , x
1
= 2.6, x
0
= 2.1, y
2
= 1.2 , y
1
=
2.8, y
0
= 2.9. Moreover, choosing the parameters o = 1.1, [ = 1.12, y = u.uu4, o
1
= u.8, [
1
= u.81, y
1
=
u.uu7. Then , the system (1) can be written as
x
n+1
=
1.1
n2
1.12+0.004x
n
x
n1
x
n2
, y
n+1
=
0.8x
n2
0.81+0.007
n
n1
n2
, n = u,1, , (12)
and with initial condition x
2
= S.S , x
1
= 2.6, x
0
= 2.1, y
2
= 1.2 , y
1
= 2.8, y
0
= 2.9. Moreover, in Fig.
2 the plot of x
n
is shown in Fig. 2a, the plot of y
n
is shown in Fig. 2b and an attractor of the system (12) is
shown in Fig. 2c.
Example 3. Consider the system (1) with initial conditionsx
2
= S.4 , x
1
= 2.6, x
0
= 2.1, y
2
= 1.2 , y
1
=
109
Computational Ecology and Software, 2014, 4(2): 104115
IAEES www.iaees.org
2.8, y
0
= 2.7. Moreover, choosing the parameterso = 1Suu, [ = 1SSS, y = 1.9, o
1
= 14uu, [
1
= 14Su, y
1
=
SSu. Then, the system (1) can be written as
x
n+1
=
1500
n2
1535+1.9x
n
x
n1
x
n2
, y
n+1
=
1400x
n2
1450+350
n
n1
n2
, n = u,1, , (13)
and with initial condition x
2
= S.4 , x
1
= 2.6, x
0
= 2.1, y
2
= 1.2 , y
1
= 2.8, y
0
= 2.7. Moreover, in Fig.
3 the plot of x
n
is shown in Fig. 3a, the plot of y
n
is shown in Fig. 3b and an attractor of the system (13) is
shown in Fig. 3c.
(a) Plot ofx
n
for the system (11)
(b) Plot of y
n
for the system (11)
110
Computational Ecology and Software, 2014, 4(2): 104115
IAEES www.iaees.org
(c) An attractor of the system (11)
Fig. 1 Plots for the system (11).
(a) Plot of x
n
for the system (12)
111
Computational Ecology and Software, 2014, 4(2): 104115
IAEES www.iaees.org
(b) Plot of y
n
for the system (12)
(c) An attractor of the system (12)
Fig. 2 Plots for the system (12).
112
Computational Ecology and Software, 2014, 4(2): 104115
IAEES www.iaees.org
(a) Plot of x
n
for the system (13)
(b) Plot of y
n
for the system (13)
113
Computational Ecology and Software, 2014, 4(2): 104115
IAEES www.iaees.org
(c) An attractor of the system (13)
Fig.3 Plots for the system (13).
5 Conclusions
In the paper, we investigate some dynamics of a sixdimensional discrete system. The system has only one
equilibrium point (u,u). The linearization method is used to show that equilibrium point (u,u)is locally
asymptotically stable. Also system has no prime periodtwo solutions. The main objective of dynamical
systems theory is to predict the global behavior and rate of convergence of the system. In the paper, we use
simple technique to prove the global asymptotic stability of equilibrium point(u,u). Some numerical examples
are provided to support our theoretical results.
Acknowledgements
This work was supported by the Higher Education Commission of Pakistan.
References
Agarwal RP. 2000. Difference Equations and Inequalities (Second Edition). Marcel Dekker, New York, USA
Bajo I, Liz E.2011.Global behaviour of a secondorder nonlinear difference equation.J ournal of Difference
Equations and Applications, 17(10): 14711486
Cinar C.2004. On the positive solutions of the difference equation system x
n+1
=
1
n
, y
n+1
=
n
x
n1
n1
.
Applied Mathematics and Computation, 158: 303305
Din Q. 2013.Dynamics of a discrete LotkaVolterra model. Advances in Difference Equations, 1: 113
114
Computational Ecology and Software, 2014, 4(2): 104115
IAEES www.iaees.org
Din Q. Global character of a rational difference equation, Thai J ournal of Mathematics (in press)
Din Q. 2014.Global stability of a population model. Chaos, Soliton and Fractals, 59: 119128.
Din Q, Donchev T. 2013.Global character of a hostparasite model. Chaos, Soliton and Fractals, 54: 17
Din Q. On a system of rational difference equation. Demonstratio Mathematica (in press)
Din Q, Khan AQ, Qureshi MN. 2013. Qualitative behavior of a hostpathogen model. Advances in Difference
Equations, 1: 263
Din Q, Qureshi MN, Khan AQ, Dynamics of a fourthorder system of rational difference equations, Advances
in Difference Equations, doi:10.1186/168718472012216
Grove EA, Ladas G. 2004.Periodicities in Nonlinear Difference Equations. Chapman and Hall/CRC Press,
Boca Raton, USA
Kalabusic S, Kulenovic MRS, Pilav E. 2009. Global dynamics of a competitive system ofrational difference
equations in the plane. Advances in Difference Equations, Article ID 132802
Kalabusic S, Kulenovic MRS, Pilav E. 2011. Dynamics of a twodimensional system of rational difference
equations of LeslieGower type. Advances in Difference Equations. doi: 10.1186/16871847201129
Khan AQ, Qureshi MN, Din Q. 2013.Global dynamics of some systems of higherorder rational difference
equations.Advances in Differential Equations. doi: 10.1186/168718472013354
Kocic VL, Ladas G: 1993. Global Behavior of Nonlinear Difference Equations of Higher Orderwith
Applications. Kluwer Academic, Dordrecht, Netherlands
Kurbanli AS. 2011. On the behavaior of positive solutions of the system of rational differencex
n+1
=
x
n1
n
x
n1
1
, y
n+1
=
n1
x
n
n1
1
, z
n+1
=
1
n
z
n
. Advances in Difference Equations, 40
Kurbanli AS, Cinar C, Yalcinkaya I. 2011. On the behavior of positive solutions of the system ofrational
difference equations x
n+1
=
x
n1
n
x
n1
+1
, y
n+1
=
n1
x
n
n1
+1
. Mathematical and Computer Modelling, 53:
12611267
Pituk M. 2002. More on Poincare's and Perron's theorems for di_erence equations.J ournal of Difference
Equations and Applications, 8: 201216
Sedaghat H. 2003.Nonlinear difference equations: Theory with applications to social science models. Kluwer
Academic Publishers, Dordrecht, Netherlands
Shojaei M, Saadati R, Adibi H. 2009. Stability and periodic character of a rational third orderdifference
equation. Chaos, Solitons and Fractals, 39: 12031209
Stevic S. 2012.On a thirdorder system of difference equations. Applied Mathematics and Computation, 218:
76497654
Touafek N, Elsayed EM. 2012. On the solutions of systems of rational difference equations. Mathematical and
Computer Modelling, 55: 19871997
Zhang Q, Yang L, Liu J .2012. Dynamics of a system of rational third order difference equation. Advances in
Difference Equations. doi:10.1186/168718472012136
115
Computational Ecology and Software, 2014, 4(2): 116128
IAEES www.iaees.org
Article
Complex dynamics of a stochastic discrete modified LeslieGower
predatorprey model with MichaelisMenten type prey harvesting
A. Elhassanein
Dept. of Math., Faculty of Science, Damanhour University, Damanhour, Egypt
Email: el_hassanein@yahoo.com
Received 30 January 2014; Accepted 5 March 2014; Published online 1 June 2014
Abstract
This paper introduced a stochastic discretized version of the modified LeslieGower predatorprey model with
MichaelisMenten type prey harvesting. The dynamical behavior of the proposed model was investigated. The
existence and stability of the equilibria of the skeleton were studied. Numerical simulations were employed to
show the model's complex dynamics by means of the largest Lyapunov exponents, bifurcations, time series
diagrams and phase portraits. The effects of noise intensity on its dynamics and the intermittency phenomenon
were also discussed via simulation.
Keywords ecosystems; LeslieGower predatorprey model; skeleton; bifurcations; chaos.
1 Introduction
Deterministic nonlinear predatorprey models (ODE models) are widely used to understand the dynamics of
the ecosystems (Rosenzweig and MacArthur, 1963; Clark, 1976; Steven, 1994; Srinivasu, 2001; Kondoh, 2003;
Murdoch et al., 2003; Nedorezov and Sadykov, 2012). In recent years, there has been an increasing interest in
LeslieGower type predatorprey model (Hsu and Huang, 1995; Li and Xiao, 2007; Liang and Pan, 2007; Song
et al., 2009; Tian and Weng, 2011; Tian and Zhu, 2012). The local and global stability for a predatorprey
model of modified LeslieGower and Hollingtype II with timedelay has been considered by Lin and Ho
(2006). Bifurcation analysis of a LeslieGower preypredator model with Hollingtype III functional response
has been studied by Li and Xiao (2007). The global stability of a LeslieGower preypredator model with
proportional harvesting in both prey and predator has been studied by Zhang et al. (2011). By defining a
suitable Lyapunov function, the global stability of the unique interior equilibrium of the system was shown,
which means that suitable harvesting has no influence on the persistent property of the harvesting system.
MenaLorca et al. (2007) studied the dynamics of the LeslieGower model subjected to the Allee effect with
proportionate harvesting. The dynamical behavior of the periodic preypredator model with a modified Leslie
Gower Hollingtype II scheme and impulsive effect has been studied by Song and Li (2008). Phase portraits
ComputationalEcologyandSoftware
ISSN2220721X
URL:http://www.iaees.org/publications/journals/ces/onlineversion.asp
RSS:http://www.iaees.org/publications/journals/ces/rss.xml
Email:ces@iaees.org
EditorinChief:WenJunZhang
Publisher:InternationalAcademyofEcologyandEnvironmentalSciences
Computational Ecology and Software, 2014, 4(2): 116128
IAEES www.iaees.org
near the interior equilibria of a LeslieGower model with constant harvesting in prey has been considered by
Zhu and Lan (2010). Gupta and Chandra (2013) introduced a modified version of the LeslieGower prey
predator model with Hollingtype II functional response in the presence of nonlinear harvesting in prey. Many
authors (Agarwal, 2000; Agarwal, 1997; Freedman, 1980; Murray, 1989) have argued that the discrete time
models governed by difference equations are more appropriate than the continuous ones when the populations
have nonoverlapping generations. Discrete time models can also provide efficient computational models of
continuous models for numerical simulations. Discrete LeslieGower predatorprey model has also been
studied by many authors. Huo and Li (2004) considered a discrete LeslieGower predatorprey model. They
obtained sufficient conditions which guarantee the permanence of the model. Under the assumption that the
coefficients in the model are periodic, the existence of periodic solution is also obtained. Agiza et al (2009)
investigated discretetime preypredator model with Holling type II. The existence and stability of three fixed
points have been analyzed. The bifurcation diagrams, phase portraits and Lyapunov exponents have been
obtained for different parameters of the model. The fractal dimension of a strange attractor of the model has
been also calculated. They showed that the discrete model exhibits rich dynamics compared with the
continuous model. It has been shown that for the discretetime preypredator models the dynamics can produce
a much richer set of patterns than those observed in continuoustime models, see Danca (1997), J ing and Yang
(2006), Liu and Xiao (2007). The main objective of this paper is to propose a stochastic discrete version of the
modified LeslieGower predatorprey model with MichaelisMenten type prey harvesting, and analyze its
chaotic behavior.
The organization of this paper is as follows. In section 2, a stochastic discrete version of the modified
LeslieGower predatorprey model with MichaelisMenten type prey harvesting is formulated. In section 3, the
stability condition of the system are derived. The simulation is used in section 4, to discuss the analytical
results and to show the effects of noise intensity on the dynamics of the system.
2 The Stochastic Discrete Model
In the case of continuous time the modified LeslieGower predatorprey model with MichaelisMenten type
prey harvesting has the following form
_
dx
1
dt
= rx
1
[1 
x
1
K

u
1
x
1
x
2
n+x
1

qLx
1
m
1
L+m
2
x
1
,
dx
2
dt
= sx
2
[1 
u
2
x
2
n+x
1
(2.1)
here, x
1
(t) and x
2
(t) are the prey and predator population densities respectively r and K are intrinsic
growth rate and environmental carrying capacity for the prey species respectively. o
1
is the maximum value
of the per capita reduction rate of prey, n measures the extent to which the environment provides protection
to prey and predator (under the assumption that the extent to which the environment provides protection to
both the predator and prey is the same (J i et al., 2009, 2011), s measures the growth rate of the predator
species, so
2
is the maximum value of the per capita reduction rate of predator, q is the catchability coefficient,
E is the effort applied to harvest the prey species, and m
1
, and m
2
are suitable constants. All the parameters
are assumed to be positive due to biological considerations.
Using a suitable nondimensional scheme, the system (2.1) can be transformed into the following system,
Gupta and Chandra (2013).
_
dx
dt
= x [1 
x
1
K
x 
u
m+x

h
c+x
,
d
dt
= py [1 
[
m+x
(2.2)
117
Computational Ecology and Software, 2014, 4(2): 116128
IAEES www.iaees.org
with the initial conditions, x(u) = x
0
> u, y(u) = y
0
> u where o, [, m, , c, andp are all positive.
Applying the forward Euler scheme to model (2.2) we obtain the following stochastic discrete modified
LeslieGower predatorprey model with MichaelisMenten type prey harvesting model
_
x
t
= x
t1
+ lx
t1
[1 x
t1

u
t1
m+x
t1

h
c+x
t1
+ ke
t
,
y
t
= y
t1
+ply
t1
[1 
[
t1
m+x
t1
+ kp
t
(2.3)
where l is the step size o, [, m, , c, andp are defined as model (2.2) and (e
t
, p
t
) are assumed to be an
i.i.d. white noise sequence conditional upon the history of the time series, which is denoted
t1
=
{(x
t1
, y
t1
)] that is, E{(e
t
, p
t
)\
t1
] = u andE{e
t
2
\
t1
] = o
2
E{p
t
2
\
t1
] = o
2
, and k is a scalar
parameter of the noise intensity.
3 The Skeleton (Free Noise System)
In this section we study the chaotic behaviour of the free noise system (2.3) caused by the change of time step.
Where k = u, the system (2.3) becomes
_
x
t
= x
t1
+ lx
t1
[1  x
t1

u
t1
m+x
t1

h
c+x
t1
,
y
t
= y
t1
+ ply
t1
[1 
[
t1
m+x
t1
(3.1)
Fixed points of the system (3.1) are derived in the following.
Lemma 3.1. The fixed points of the system (3.1), are
(i) E
0
= (u,u);
(ii) E
1
= (x

1
, y
1

), where
x

1
=
1
2
[c 1 
u
[
+_
1
4
[c  1 
u
[
2
[b  c +
uc
[
, y
1

=
m+x

1
[
and
(iii) E
2
= (x

2
, y
2

), where x

2
=
1
2
[c 1 
u
[
 _
1
4
[c  1 
u
[
2
 [b  c +
uc
[
, y
2

=
m+x

2
[
Proof. The fixed points of the system (3.1) are obtained as the solution of the algebraic system:
_
x = x +lx [1  x
t1

u
m+x

h
c+x
,
y = y +ply [1 
[
m+x
which is obtained by setting x
t
= x
t1
= x and y
t
= y
t1
= y in (3.1), it is easy to complete the proof.
The local stability analysis of the system (3.1) can be studied by computing the variation matrix
corresponding to each fixed point. The variation matrix of the system at state variable is given by
[(x, y) = _
1 +l(1 2x 
um
(m+x)
2

h
(c+x)
2
)
Iux
m+x
Ip[
2
(m+x)
2
1 + lp(1 
2[
m+x
)
_. (3.2)
Theorem 3.1. The fixed point E
0
is unstable fixed point for all parameters values.
Proof. In order to prove this result, we estimate the eigenvalues of J acobian matrix at E
0
. The J acobian
matrix for E
0
is
[(E
0
) = _
1 + l(1 
h
c
)
Iux
m+x
u 1 + lp
_.
Hence the eigenvalues of [(E
0
) are
1
= 1 + l(1 
h
c
),
2
= 1 +lp, and since all parameters are positive
118
Computational Ecology and Software, 2014, 4(2): 116128
IAEES www.iaees.org
the proof is completed.
Theorem 3.2. The fixed point E
, max {b
, J
] , max {c
]] < l < g
;
(A2) max{o
, max {b
, J
, min {c
];
(AS) max{o
, max {c
, min {b
, J
];
(A4) o
, min {b
, J
] , min {c
]]
otherwise it is unstable fixed point, where o
=
2y
i
q
i
, b
=
p+p
2
4q
i
q
i
, J
=
pp
2
4q
i
q
i
, c
=
p+y
i
+(p+y
i
)
2
8q
i
2q
i
,
=
p+y
i
(p+y
i
)
2
8q
i
2q
i
,
=
p+y
i
q
i
,
= p(y
+
pI
i
[(2m+I
i
)
), y
= 1  I

2um
[
1
2m+I
i

4hc
(2c+I
i
)
2
, and I
= 2x

.
Proof. The J acobian matrix (3.2) at E
) =
`
1 +l y
lp I
m +x
lp
[
1 lp
/
.
where y
= 1  I

2um
[
1
2m+I
i

4hc
(2c+I
i
)
2
,
= p(y
+
pI
i
[(2m+I
i
)
) and I
= 2x

.
The corresponding characteristic equation of matrix [(E
), i = 1,2 is
P
() =
2
 tr[(E
) + Jct[(E
) = u,
where tr[(E
) = 2 +l(y
 p) and Jct[(E
) = 1 (y
+p) +l
2
If the eigenvalues of the J acobian matrix of a fixed point are inside the unit circle of the complex plan,
interior fixed point is local stable. Using J ury's conditions (Chatterjee and Yilmaz, 1992; Zhu and Lan, 2010),
we have necessary and sufficient condition for local stability of interior fixed point which are the necessary
and sufficient condition for 
1,2
 < 1, which completes the proof.
4 Numerical Simulation
4.1 Deterministic system
The main purpose of this subsection is to investigate the qualitative behavior of the solution of the nonlinear
system (3.1). To provide some numerical evidence for its chaotic behavior, we present various numerical
results here to show the chaoticity including its bifurcation diagrams, Lyapunov exponents, and fractal
dimension. In Fig. 1(a) bifurcation diagram of the system (3.1) is plotted on the interval u.2 < l < u.9 for
initial point (x
0
, y
0
) = (u.u1,u.u2) with (o, [, m, b, c, p) = (1,2,2,1,2.1,S). In Fig. 2, a plot of the Lyapunov
exponent for attractors of the system (3.1) according to Fig. 1 is presented. A positiveness of this exponent for
l

> u.6689 confirms the chaotic character of attractors in this parametrical zone (here, the value l

=
u.6689 is a tangent bifurcation point). In Fig. 3 phase portraits are given for different values of l to show the
chaotic behavior of the system. For the given parameters the only positive equilibrium point is E

1
=
119
Computational Ecology and Software, 2014, 4(2): 116128
IAEES www.iaees.org
(u.uSu662,1.u1SSS) which is attracting point for u.2 < l < u.66890 as we can see in Fig. 3. (a). Fig. 3.(b)
show the periodtwo orbit in the parameter zone 0.6689<l <0.8195. The periodfour orbit in the parameter
zone u.819S < l < u.8S27 is clear in Fig. 3(c). The chaotic attractor for u.8S27 < l < u.9 is clear in Fig.
3(d). Which means that the system (3.1) undergoes a discrete Hopf bifurcation. One of the commonly used
characteristics for classifying and quantifying the chaoticity of a dynamical system is fractal dimensions,
(Cartwright, 1999; Zhu and Lan, 2010). Via simulation we get two Lyapunov exponents z
1
= u.u8u26 > u >
z
2
= u.SS26S for l = u.871, which means that J
L
= 1 +
0.08026
.35263
= 1.2276. There for the system (3.1)
exhibits a fractal structure and its attractor has a fractal dimension which is chaotic behavior.
Fig. 1 Bifurcation diagram of (3.1), for u.2 < l < u.9 and initial point (x
0
, y
0
) = (u.u1,u.u2) with
(o, [, m, b, c, p) = (1,2,2,1,2.1,S).
Fig. 2 Maximum Lyapunov exponents corresponding to Fig. 1.
120
Computational Ecology and Software, 2014, 4(2): 116128
IAEES www.iaees.org
Fig. 3 Phase portrait of the system (3.1) for: (a) l = u.6 ; (b) l = u.6679 (c) l = u.671 (d) l = u.87 with
(o, [, m, b, c, p) = (1,2,2,1,2.1,S).
4.2 Stochastic model
Along with the deterministic systems (3.1) we consider a stochastic system forced by additive noise (2.3),
wheree
t
and p
t
are uncorrelated Gaussian random processes with parameters E(e
t
, p
t
) = u andEe
t
2
= 1
Ep
t
2
= 1, and k is a scalar parameter of the noise intensity. We study a behavior of this stochastic model for
different sets of parameters l and k. In Figs. 4, 5 and 6, stochastic attractors of the system (2.3) are plotted on
the interval u.2 < l < u.9 for three values of the noise intensity k = u.uuu2, u.uuu4, anu u.uuu6.
Lyapunov exponents corresponding to each value of noise intensity are given in Fig. (7) (a), (b) and (c)
respectively. As we can see, noise deforms the deterministic attractor (compare Figs. 2 and 7). As noise
intensity increases, a border between order and chaos moves to the left. The changes of the arrangement of
attractors are accompanied by the changes in dynamical characteristics (compare Lyapunov exponents in Figs.
2 and 8). The most essential difference between stochastic and deterministic attractors is observed near the
bifurcation point l

= u.6689. The underlying reason is that in the vicinity of this bifurcation point attractors
are highly sensitive to random disturbances. Consider in detail a behavior of stochastic system (2.3)
near l

= u.6689. Compare the stochastic response of this system for two fixed values l = u.668 and
l = u.671. Consider l = u.668. For low noise k = u.uuu2, random states are concentrated near the stable
deterministic equilibrium E

1
(see Figs. 8 and 12(a)), they have distribution with mean equal approximately
E

1
= (u.uSu6,1.u1SS) and covariance matrix
121
Computational Ecology and Software, 2014, 4(2): 116128
IAEES www.iaees.org
= [
u.uuuuuuS7 u.uuuuuuS8
u.uuuuuuS8 u.uuuu2u28
.
For k = u.uuu6 one can see stochastic oscillations of large amplitude. Indeed, as the noise intensity increases,
the dispersion of random states near E

1
growsc see Table 1. After these oscillations, iterations come to the
vicinity of the point E

1
again and so on (see Figs. 9 and 12(b)). In this case, the stochastic model (2.3)
exhibits a coexistence of two different dynamical regimes even if the system (3.1) has a stable equilibrium
only. This type of dynamics of the system (2.3) can be determined as a noiseinduced intermittency
(Bashkirtseva and Ryashko 2013). In Figs. 10 and 11, time series of the stochastic system (2.3) with
l = u.671 for k = u.uuu2 and k = u.uuu2 are plotted. As can be seen, noiseinduced intermittency for this
l = u.671 is observed for the lower noise intensity, see also Fig. 13. For stochastic attractors and their
dynamic characteristics, a dependence on noise level is illustrated in Figs. 14 and 15 for l = u.668 and
l = u.671. In Fig. 14, one can see a sharp growth of the size of the attractor as noise intensity exceeds some
critical value. A change of the sign of Lyapunov exponent from minus to plus can be interpreted as a transition
from regular to noiseinduced chaotic regime (see Fig. 16). Thus, the results presented here give us a
qualitative description of noiseinduced transitions from the regular regime to intermittency.
Fig. 4 Bifurcation diagrams of the stochastic model (2.3) with k = u.uuu2, u.2 < l < u.9, and (o, [, m, b, c, p) =
(1,2,2,1,2.1,S).
Fig. 5 Bifurcation diagrams of the stochastic model (2.3) with k = u.uuu4, u.2 < l < u.9, and (o, [, m, b, c, p) =
(1,2,2,1,2.1,S).
122
Computational Ecology and Software, 2014, 4(2): 116128
IAEES www.iaees.org
Fig. 6 Bifurcation diagrams of the stochastic model (2.3) with k = u.uuu6, u.2 < l < u.9, and (o, [, m, b, c, p) =
(1,2,2,1,2.1,S).
Fig. 7 Lyapunov exponents of the stochastic model (2.3) with u.2 < l < u.9: for (a) k = u.uuu2; (b) k = u.uuu4; (c)
k = u.uuu6.
Fig. 8 Time series of the stochastic model (2.3) withl =0.668, k = u.uuu2 and (o, [, m, b, c, p) = (1,2,2,1,2.1,S).
123
Computational Ecology and Software, 2014, 4(2): 116128
IAEES www.iaees.org
Fig. 9 Time series of the stochastic model (2.3) with l = .668, k = u.uuu6 (o, [, m, b, c, p) = (1,2,2,1,2.1,S).
Fig. 10 Time series of the stochastic model (2.3) with l = .671, k = u.uuu2 and (o, [, m, b, c, p) = (1,2,2,1,2.1,S).
Fig. 11 Time series of the stochastic model (2.3) with l = .671, k = u.uuu6 and (o, [, m, b, c, p) = (1,2,2,1,2.1,S).
Fig. 12 Phase portrait of the system (2.3) with l = .668, and (o, [, m, b, c, p) = (1,2,2,1,2.1,S) for: (a) k = u.uuu2;
(b) k = u.uuu6.
124
Computational Ecology and Software, 2014, 4(2): 116128
IAEES www.iaees.org
Fig. 13 Phase portrait of the system (2.3) with l = .671 and (o, [, m, b, c, p) = (1,2,2,1,2.1,S) for: (a) k = u.uuu2;
(b) k = u.uuu6.
Fig. 14 Attractors of the stochastic system (2.3) for u k u.uuu6 ; l = u.668 and (o, [, m, b, c, p) =
(1,2,2,1,2.1,S).
Fig. 15 Attractors of the stochastic system (2.3) for u k u.uuu6 ; l = u.671 , and (o, [, m, b, c, p) =
(1,2,2,1,2.1,S).
Fig. 16 Lyapunov exponents of the stochastic system (2.3) with u k u.uuu6 and
(o, [, m, b, c, p) = (1,2,2,1,2.1,S) for: (a) l = u.668; (b) l = u.671.
125
Computational Ecology and Software, 2014, 4(2): 116128
IAEES www.iaees.org
Table 1 Mean of the stochastic states (x
t
, y
t
) of the system (2.3) at different values of l , k and (o, [, m, b, c, p) =
(1,2,2,1,2.1,S).
k l 0.668 0. 671
0.0002 (0.030572,1.0153) (0.032278,1.0142)
u.uuu6 (0.029719,1.0147) (0.033094,1.0145)
Table 2 Covariance matrices of the stochastic states(x
t
, y
t
) of the system (2.3) at different values of l , k and
(o, [, m, b, c, p) = (1,2,2,1,2.1,S).
k l 0.668 0. 671
0.0002
[
u.uuuuuuS7 u.uuuuuuS8
u.uuuuuuS8 u.uuuu2u28
[
u.uuuuu14u u.uuuu1124
u.uuuu1124 u.uu19S619
u.uuu6
[
u.uuuu1u82 u.uuuuu6u6
u.uuuuu6u6 u.uuu1678S
[
u.uuuu1211 u.uuuu17S2
u.uuuu17S2 u.uu2uS988
5 Conclusion
From a mathematical as well as biological point of view the predatorprey models can be formulated as
systems of differential or difference equations (Nedorezov and Sadykov, 2012). The current paper have
proposed a stochastic discrete modified LeslieGower predatorprey model with MichaelisMenten type prey
harvesting, where the protection provided by the environment for both the prey and predator is the same. The
model shows rich and varied dynamics. The local stability of fixed points have been discussed. The results
show that the origin is unstable equilibrium point of the system. There is a unique interior equilibrium point
which is locally stable for certain parametric restrictions. The effectiveness of the time step on the dynamics of
the system has been shown. The chaotic behaviour of the system has been proved. We focus on the study of
the noiseinduced typeI intermittency phenomenon and chaotization observed near tangent bifurcation. The
remarkable feature of the dynamics of the model considered here is that small noises generate largeamplitude
chaotic oscillation.
References
Agarwal RP. 2000. Difference Equations and Inequalities: Theory, Method and Applications. Monographs and
Textbooks in Pure and Applied Mathematics (Volume 2). Marcel Dekker, New York, USA
Agarwal RP, Wong PJ Y. 1997. Advance Topics in Difference Equations. Kluwer, Dordrech. Germany
Agiza HN, ELabbasy EM, ELMetwally H, Elsadany AA 2009. Chaotic dynamics of a discrete preypredator
model with Holling type II. Nonlinear Analysis, RWA 10: 116129
Bashkirtseva I, Ryashko L. 2013: Stochastic sensitivity analysis of noiseinduced intermittency and transition
to chaos in onedimensional discretetime systems. Physica A, 392: 295306
Cartwright J HE 1999. Nonlinear stifiness, Lyapunov exponents, and attractor dimension. Physical Letters A,
264: 298304
Chatterjee S, Yilmaz M. 1992. Chaos, fractals and statistics. Statistical Science, 7: 4968
Clark CW 1976. Mathematical Bioeconomics the Optimal Management of Renewable Resources. Wiley, New
York
Danca M, Codreanu S, Bako B 1997. Detailed analysis of a nonlinear preypredator model. Journal of
126
Computational Ecology and Software, 2014, 4(2): 116128
IAEES www.iaees.org
Biological Physics, 23: 1120
Freedman HI 1980. Deterministic Mathematics Models in Population Ecology. Marcel Dekker, New York,
USA
Gupta RP, Chandra P. 2013. Bifurcation analysis of modified LeslieGower predatorprey model with
MichaelisMenten type prey harvesting. Journal of Mathematical Analysis and Applications, 398: 278295
Hsu SB, Huang TW. 1995. Global stability for a class of predatorprey systems. SIAM J ournal of Applied
Mathematics, 55: 763783
Huo H, and Li WT. 2004. Stable periodic solution of the discrete periodic LeslieNewer predatorprey model.
Mathematical and Computer Modelling 40: 261269
J i C, J iang D, Shi N. 2009. Analysis of a predatorprey model with modified LeslieGower and Hollingtype
II schemes with stochastic perturbation. J ournal of Mathematical Analysis and Applications, 359: 482498
J i C, J iang D, Shi N. 2011. A note on a predatorprey model with modified LeslieGower and Hollingtype II
schemes with stochastic perturbation. Journal of Mathematical Analysis and Applications, 377(1): 435440
J ing ZJ , Yang J . 2006. Bifurcation and chaos discretetime predatorprey system. Chaos, Solitons and Fractals,
27, 259277
Kondoh M 2003. Foraging adaptation and the relationship between foodweb complexity and stability. Science,
299: 13881391
Li Y, Xiao D. 2007. Bifurcations of a predatorprey system of Holling and Leslie types. Chaos, Solitons and
Fractals, 34, 606620
Liang Z, Pan H. 2007. Qualitative analysis of a ratiodependent HollingTanner model. J . Math. Anal. Appl.
334, 954964
Lin CM, Ho CP. 2006. Local and global stability for a predatorprey model of modified LeslieGower and
Hollingtype II with timedelay. Tunghai Science, 8: 3361
Liu X, Xiao D 2007. Complex dynamic behaviors of a discretetime predatorprey system. Chaos, Solitons and
Fractals, 32: 8094
MenaLorca J , GonzalezOlivares E, GonzalezYanz B. 2007. The LeslieGower predatorprey model with
Allee effect on prey: a simple model with a rich and interesting dynamics. In: Proceedings of the
International Symposium on Mathematical and Computational Biology BIOMAT 2006 (Mondaini R, ed).
105, 132, Epapers Servios Editoriais Ltda, R' io de J aneiro, Brazil
Murdoch W, Briggs C, Nisbet R. 2003. ConsumerResource Dynamics. Princeton University Press, New York,
USA
Murray J D 1989. Mathematical Biology. SpringerVerlag, New York, USA
Nedorezov LV, Sadykov AM. 2012. About a modification of Rogers model of parasitehost system dynamics.
Proceedings of the International Academy of Ecology and Environmental Sciences, 2(1): 4149
Song Y, Yuan S, Zhang J . 2009. Bifurcation analysis in the delayed LeslieGower predatorprey system.
Applied Mathematical Modelling, 33: 40494061
Rosenzweig ML, MacArthur RH. 1963. Graphical representation and stability conditions of predatorprey
interactions. American Naturalist, 97: 209
Song XY, Li YF. 2008. Dynamic behaviors of the periodic predatorprey model with modified LeslieGower
Hollingtype II schemes and impulsive effect. Nonlinear Analysis, RWA 9: 6479
Srinivasu PDN. 2001. Bioeconomics of a renewable resource in presence of a predator. Nonlinear Analysis,
RWA 2: 497506
Steven HS. 1994. Nonlinear dynamics and chaos: with applications to physics, biology, chemistry and
engineering. Perseus Books Publishing, USA
127
Computational Ecology and Software, 2014, 4(2): 116128
IAEES www.iaees.org
Tian C, Zhu P. 2012. Existence and asymptotic behavior of solutions for quasilinear parabolic systems. Acta
Applicandae Mathematicae. DOI 10.1007/s 1044001297017
Tian Y, Weng P. 2011. Stability analysis of diffusive predatorprey model with Modified LeslieGower and
HollingType II schemes. Acta Applicandae Mathematicae, 114(3): 173192
Zhang N, Chen F, Su Q, Wu T. 2011. Dynamic behaviors of a harvesting LeslieGower predatorprey model.
Discrete Dynamics in Nature and Society. http://dx.doi.org/10.1155/2011/47394.
Zhu CR, Lan KQ. 2010. Phase portraits, Hopfbifurcations and limit cycles of LeslieGower predatorprey
systems with harve sting rates. Discrete and Continuous Dynamical SystemsSeries B, 14: 289306
128
Computational Ecology and Software, 2014, 4(2): 129134
IAEES www.iaees.org
Article
Synthetic Aperture Radar (SAR) image segmentation by fuzzy c
means clustering technique with thresholding for iceberg images
Usman Seljuq
1
, Rashid Hussain
2
1
Sir Syed University of Engineering & Technology Karachi75300, Pakistan
2
Faculty of Engineering Science and Technology, Hamdard University, Karachi 74600, Pakistan
Email: rashid.hussain@hamdard.edu.pk
Received 27 December 2013; Accepted 5 February 2014; Published online 1 June 2014
Abstract
Fuzzy cmeans (FCM) clustering algorithm is widely used for image segmentation. The purpose of clustering
is to identify natural groupings of data from a large data set, which results in concise representation of
systems behavior. It can be used to detect icebergs regardless of ambient conditions like rain, darkness and
fog. As a result SAR images can be used for iceberg surveillance. In this paper we have investigate FCM with
thresholding for iceberg image segmentation for Synthetic Aperture Radar (SAR) images. The results showed
that the assessment parameters; mean and entropy have lower values for efficient segmentation.
Keywords Synthetic Aperture Radar (SAR); fuzzy cmeans clustering; thresholding; image segmentation.
1 Introduction
Image segmentation is a preliminary step in pattern recognition and scene analysis understanding. It partitions
the image into its constituent regions and objects. Image processing is effected by illumination conditions,
environmental disturbances, noise and temperature fluctuation. Under some severe conditions of unexpected
and improper illumination, the blurring of images makes it very difficult for target recognition, so under such
conditions segmentation is necessary. The SAR system is used in investigating environmental and ecological
activities (Hussain, 2012). SAR images are not affected by clouds, sunlight, day and night effects. The
satellites and airplanes equipped with SAR are used to monitor icebergs. SAR is an active microwave that
captures images (Topouzelis, 2008). One of the most important applications of SAR system is to collect the
data from the ground surface through image reconstruction. The ground surface has different areas such as
roads, deserts, pounds, buildings, grassland and cultivated plants, so those areas have to be segmented
according to the applications. Thus segmentation problem arises and because of this various segmentation
methods have been proposed. Correct segmentation is an important issue in SAR image segmentation. Several
methods are used for SAR image segmentation such as clustering algorithms, threshold methods and
ComputationalEcologyandSoftware
ISSN2220721X
URL:http://www.iaees.org/publications/journals/ces/onlineversion.asp
RSS:http://www.iaees.org/publications/journals/ces/rss.xml
Email:ces@iaees.org
EditorinChief:WenJunZhang
Publisher:InternationalAcademyofEcologyandEnvironmentalSciences
Computational Ecology and Software, 2014, 4(2): 129134
IAEES www.iaees.org
morphological methods. This paper reexamines the approach of SAR clustering based image segmentation.
Fuzzy clustering has a potential application in the fields of image processing and pattern recognition. Some of
the recent research includes an Improved FCM Algorithm Based on the SVDD for Unsupervised
Hyperspectral Data Classification (Niazmardi et al., 2013), Fuzzy CMeans Algorithms for Very Large Data
(Havens et al., 2012), Weighted Fuzzy CMeans Clustering Algorithm for Remotely Sensed Image
Classification (Chen et al., 2011), Unsupervised Subpixel Mapping of Remotely Sensed Imagery Based on
Fuzzy CMeans Clustering Approach (Zhang et al., 2014), A MultipleKernel Fuzzy CMeans Algorithm for
Image Segmentation (Chen et al., 2011), and Change Detection in Synthetic Aperture Radar Images based on
Image Fusion and Fuzzy Clustering (Gong et al., 2012).
Clustering is commonly used in pattern recognition, machine learning, biomedical and image analysis. In
fuzzy logic FCM is a technique of clustering that allows one piece of data belongs to two or more clusters (Pal
and Bezdek, 1995).
This paper is organized as follows: section 2 represents the need for image segmentation. Section 3
discusses the proposed FCM algorithm for the segmentation of SAR iceberg images with thresholding. The
simulation results and evaluation of the reconstructed image are represented in section 4 and section 5. Finally,
in section 6 conclusions are given.
2 Iceberg Image Segmentation
Image segmentation is used to locate the boundaries and objects in images. One of the most practical
applications of image segmentation is to locate the objects in SAR images (roads, buildings, forests etc.) image
processing is effected by random noise, environmental disturbances due to atmospheric pressure or
temperature fluctuations and illumination conditions, so image segmentation is necessary that divides an image
into regions of different characteristics.
SAR image segmentation falls into two categories, a) Segmentation based on grey levels b) Segmentation
based on texture. Image segmentation thresholding techniques can be found in (Sahoo et al., 1988). This paper
deals with SAR image segmentation based on grey levels with global thresholding technique.
3 Investigating Fuzzy CMean Clustering With Thresholding
The investigation is based upon the combination of FCM clustering with thresholding algorithm. Clustering of
data is a technique by which large sets of data are grouped into smaller ones.
3.1 Fuzzy clustering sets of clusters of similar data
The partition should have two properties, (a) homogeneity inside clusters: the data, which belong to one cluster,
should be as similar as possible; (b) heterogeneity between clusters: The data, which belongs to different
clusters, should be as different as possible. Many techniques have been used for clustering data; in this paper
fuzzy cmeans clustering is used (Sahoot et al., 1988). FCM was developed by Dunn in 1973 (Dunn, 1973),
improved by Bezdek in 1981 (Bezdek, 1981), and further developed by many researchers (Cannon et al., 1986;
Lyer et al., 2002; Wang et al., 2006; Hussain, 2012).
In 2010, Krinidis and Chatzis proposed research incorporating local spatial information into FCM
algorithm to get improved clustering performance (Krinidis, 2010).
3.2 FCM algorithm
FCM divides a set of Sdimensional vectors X =[X
1
,..,X
n
] into c clusters. Each cluster is a fuzzy set
defined by the sample space X =[X
1
,..,X
n
]. The i
th
cluster is presumed to have the center vector v
i
=
[v
i1
, . . . . , v
is
], where X
j
=[x
j1
, . . . . , x
js
] represents thejth sample for ( j=1, . . . , n ). FCM determines the
fuzzy partition matrix U and cluster centers v
i
( 1,2,..,c) by minimizing the objective function J which is
130
Computational Ecology and Software, 2014, 4(2): 129134
IAEES www.iaees.org
defined by (1).
c n
J(U, v
1
,v
2
,. . . . ,v
c
; X) = u
m
ij
d
2
ij
(1)
i=1 j=1
m in equation (1) adjusts the weighting effect. Large values of m increases the fuzziness
where d
ij
is the Euclidean distance and is defined
(2)
v
i
and u
ij
is given by
(3)
(4)
where m1
3.3 Thresholding
In fuzzy cmeans clustering the segmented part of SAR image is not visible clearly. Thusthresholding is
applied on the segmented image (Otsu, 1979; Gonzalez, 2002). There are two techniques of thresholding: local
and global thresholding. In local thresholding the image is divided into no of subimages, the threshold for
each subimages depends upon the local properties of the point. In global thresholding the entire image is
segmented with one or more values.
We have used global thresholding method. The thresholded image g(x,y) is defines as
g(x,y )={ 1 if f(x,y) T
{ 0 if f(x,y) <T (5)
where T is constant, this approach is called global thresholding (Dunn., 1973).
4 Evaluation of the Reconstructed Images
To evaluate the accuracy of the segmented image following steps are followed.
(1) Iceberg image is taken as input.
(2) Segmentation algorithm is applied on the image.
(3) Assessment parameters are applied for fuzzy cmeans clustering algorithm.
The proposed algorithm is applied by using MATLAB .The images are resized to 256x256, performance of
the algorithm is evaluated by following assessment parameters. 1) Mean 2) Entropy. Mean and entropy must
be a less values for a better segmentation algorithm.
Entropy is the amount of information based on the probabilistic occurrence of the picture elements. These
probabilities can be found from the image histogram. Entropy describes the average information of every pixel.
The more gray levels are present in the image, the higher the entropy. Suppose that two images have
S
K
jk ik ij x v d
1
2
) (
n
j
m
ij
n
j
ij
i
u
X u
v
m
1
1
c
k
m
kj
ij
ij
d
d
u
1
) 1 /( 2
) (
1
131
IAEES
probabili
5 Result
From the
images c
fuzzy cm
d
ity functions p
s and Discus
e figure 1 the
can be observ
means clutter
(
S.No.
1
2
log ) (
1
i p
k
i
Com
p and q, the e
ssion
difference be
ve. The fuzzy
ring.
(a)
Fig. 1
(a)
Fig. 2
. STAST
PARAME
MEAN
ENTROPY
) (
) (
g
2
i q
i p
mputational Ecol
entropy of p w
etween the fu
y cmeans th
(a) original im
Table 1 FC
(a) original im
TICAL
ETERS
Y
logy and Softw
with respect t
uzzy cmeans
hreshold gives
(b)
age (b) FCM (c
CM with hresho
(b
age (b) FCM (c
ORIGINAL
IMAGE
98.4815
7.3351
ware, 2014, 4(2)
to q is given b
clustering an
s better repre
c) FCM with th
olding for Fig. 1
b)
c) FCM with th
L
FCM
0.4413
4.0313
: 129134
by (6).
nd fuzzy cme
esentation of
resholding.
1.
resholding.
FCM W
THRESH
0.2169
0.7545
w
(6)
eans threshol
the image as
(c)
(c)
WITH
OLD
www.iaees.org
ld for iceberg
s compare to
g
o
132
Computational Ecology and Software, 2014, 4(2): 129134
IAEES www.iaees.org
Table 2 FCM with thresholding for Fig. 2.
From Table 1 and 2, the less values of assessment parameters like mean and entropy shows that the fuzzy
cmeans threshold method is better for iceberg images when compare with fuzzy cMeans clustering. With the
overall performance evaluation we can say that fuzzy cmeans threshold method gives desirable results when
compare to fuzzy cmeans method.
6 Conclusion
Fuzzy cmeans thresholding method has been tested on various SAR images. The scheme implemented here
makes use of thresholding and fuzzy cmeans clustering algorithm for image segmentations. This approach
made it possible to segment patterns in SAR images, which could go undetected by conventional image
clustering methods. This analysis demonstrates how new computational techniques help in satellite imaging
for environmental and ecological monitoring.
Acknowledgments
We wish to thank Matthew R. Lopez from Sandia National Laboratories Albuquerque, NM, USA, Ellen O'
Leary from Radar Data Center, J et Propulsion Laboratory Pasadena, CA and J .C Bezdek for facilitating
theoretical concepts, and to the Graduate School of Engineering Sciences and Information Technology,
Hamdard University for their logistic support and services.
References
Bezdek J C. 1981. Pattern Recognition with Fuzzy Objective Function Algorithms. Plenum Press, New York,
USA
Cannon RL, Dave J V, Bezdek, J C. 1986. Efficient implementation of the fuzzy cmeans clustering algorithms.
IEEE Transaction on Pattern Analysis and Machine Intelligence, 248255
Chen L, Chen CLP, Lu M. 2011. A multiplekernel fuzzy cmeans algorithm for image segmentation. IEEE
Transactions on Systems, Man, and CyberneticsPart B, 41: 12631274
Dunn JC. 1973. A fuzzy relative of the ISODATA process and its use in detecting compact well separated
clusters. J ournal of Cybernetics, 3: 3257
Gong MG, Zhou ZQ, Ma J J . 2012. Change detection in synthetic aperture radar images based on image fusion
and fuzzy clustering. IEEE Transactions on Image Processing, 21(4): 21412151
Gonzalez R, Woods R. 2002. Digital Image Processing (2nd Edition). PrenticeHall, USA
Havens TC, Bezdek J C, Leckie C, et al. 2012. Fuzzy cmeans algorithms for very large data. IEEE
Transactions on Fuzzy Systems, 20(6): 11301146
Hung CC, Kulkarni S, Kuo BC. 2011. A new weighted fuzzy cmeans clustering algorithm for remotely
sensed image classification. IEEE J ournal on Signal Processing, 5(3): 543553
Hussain R. 2012. Synthetic Aperture Radar (SAR) images features clustering using fuzzy cmeans (FCM)
SNO STASTICAL
PARAMETERS
ORIGINAL
IMAGE
FCM FCM WITH
THRESHOLD
1 MEAN 97.4866 0.4707 0.1869
2 ENTROPY 7.4895 7.0259 0.695
133
Computational Ecology and Software, 2014, 4(2): 129134
IAEES www.iaees.org
clustering algorithm. Computational Ecology and Software, 2(4): 220225
Krinidis S, Chatzis V. 2010. A robust fuzzy local information cmeans clustering algorithm. IEEE
Transactions on Image Processing, 19(5): 13281337
Lyer NS, Kandel A, Schneider M. 2000. Featurebased fuzzy classification for interpretation of mammograms.
Fuzzy Sets and Systems, 114(2): 271280
Niazmardi S, Homayouni S, Safari A. 2013. An improved FCM algorithm based on the SVDD for
unsupervised hyperspectral data classification. IEEE J ournal of Remote Sensing, 6(2): 831839
Otsu N. 1979. A threshold selection method from graylevel histograms. IEEE Transactions on Systems, Man
and Cybernetics, 9(1): 6266
Pal NR, Bezdek J C. 1995. On cluster validity for the fuzzy cmeans model. IEEE Trans Fuzzy Systems, 3(3):
370379
Sahoo PK, Soltani S, Wong AKC, et al. 1988. A survey of thresholding technique. Computer Vision Graphics
Image Process, 41: 233260
Topouzelis KN. 2008. Oil spill detection by SAR images: dark formation detection, feature, extraction and
classification algorithms. Sensors, 8: 8: 66426659
Wang WN, Zhang YJ , Li Y, et al. 2006. The global fuzzy cmeans algorithm. IEEE Proceedings of the 6th
World Congress on Intelligence. Control and Automation Clustering, 1: 36043607
Zhang Y, Du Y, Li X, Fang SM. 2014. Unsupervised subpixel mapping of remotely sensed imagery based on
fuzzy cmeans clustering approach. IEEE Geoscience and Remote Sensing Letters, 11(5): 10241028
134
Computational Ecology and Software
The Computational Ecology and Software (CES) is an open access, peerreviewed online journal that considers
scientific articles in all different areas of computational ecology. It is the transactions of the International
Society of Computational Ecology. The journal is concerned with the ecological researches, constructions and
applications of theories and methods of computational sciences including computational mathematics,
computational statistics and computer science. It features the simulation, approximation, prediction, recognition,
and classification of ecological issues. Intensive computation is one of the major stresses of the journal. The
journal welcomes research articles, short communications, review articles, perspectives, and book reviews. The
journal also supports the activities of the International Society of Computational Ecology. The topics to be
covered by CES include, but are not limited to:
Computation intensive methods, numerical and optimization methods, differential and difference
equation modeling and simulation, prediction, recognition, classification, statistical computation
(Bayesian computing, randomization, bootstrapping, Monte Carlo techniques, stochastic process, etc.),
agentbased modeling, individualbased modeling, artificial neural networks, knowledge based systems,
machine learning, genetic algorithms, data exploration, network analysis and computation, databases,
ecological modeling and computation using Geographical Information Systems, satellite imagery, and
other computation intensive theories and methods.
Artificial ecosystems, artificial life, complexity of ecosystems and virtual reality.
The development, evaluation and validation of software and algorithms for computational ecology. The
development and evaluation of apparatus, instruments and machines for ecological and environmental
analysis, investigation and monitoring based on the software of computational ecology.
Methodological applications of computational ecology in the researches of ecology and environmental
sciences.
Authors can submit their works to the email box of this journal, ces@iaees.org and (or) wjzhang@iaees.org. All
manuscripts submitted to CES must be previously unpublished and may not be considered for publication
elsewhere at any time during review period of this journal.
In addition to free submissions from authors around the world, special issues are also accepted. The organizer of
a special issue can collect submissions (yielded from a research project, a research group, etc.) on a specific
topic, or submissions of a conference for publication of special issue.
Editorial Office: ces@iaees.org
Publisher: International Academy of Ecology and Environmental Sciences
Address: Flat C, 23/F, Lucky Plaza, 315321 Lockhart Road, Wanchai, Hong Kong
Tel: 008526555 7188
Fax: 008523177 9906
Email: office@iaees.org
Computational Ecology and Software
ISSN 2220721X
Volume 4, Number 2, 1 J une 2014
Articles
Spatial risk assessment of alien plants in China using biodiversity resistance
theory
YouHua Chen 8288
Stability analysis of a discrete ecological model
Q. Din, E. M. Elsayed 89103
Qualitative behavior of an anticompetitive system of thirdorder rational
difference equations
Q. Din, M. N. Qureshi, A. Q. Khan 104115
Complex dynamics of a stochastic discrete modified LeslieGower
predatorprey model with MichaelisMenten type prey harvesting
A. Elhassanein 116128
Synthetic Aperture Radar (SAR) image segmentation by fuzzy cmeans
clustering technique with thresholding for iceberg images
Usman Seljuq, Rashid Hussain 129134
The International Academy of Ecology and Environmental Sciences (IAEES) is a nonprofit and registered
international organization. It devotes to promote global ecology and environmental sciences and protect global
ecological environments, by publishing scientific publications, conducting research activities, launching
environmental programs, disseminating knowledge and technologies, sponsoring conferences, and providing
information and discussion spaces, etc. The responsibility for these publications rests with the International
Academy of Ecology and Environmental Sciences.
IAEES
http://www.iaees.org/