Computers and Mathematics With Applications: Anal Chatterjee, Samares Pal

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Computers and Mathematics with Applications ( ) –

Contents lists available at ScienceDirect

Computers and Mathematics with Applications


journal homepage: www.elsevier.com/locate/camwa

Interspecies competition between prey and two different


predators with Holling IV functional response in diffusive
system
Anal Chatterjee a , Samares Pal b,∗
a
Department of Mathematics, Sheikhpara A.R.M. Polytechnic, Sheikhpara 742409, India
b
Department of Mathematics, University of Kalyani, Kalyani 741235, India

article info abstract


Article history: This paper deals with a prey–middle predator–top predator ecosystem model with Holling
Received 9 July 2015 type IV predator response in the unreserved zone. The model system is studied analytically
Received in revised form 7 November 2015 and the threshold values for the existence and stability of various steady states are
Accepted 12 December 2015
worked out. The global stability analysis is carried out. It is observed that if the intrinsic
Available online xxxx
growth rate of prey population crosses a certain critical value, the system enters into
Hopf bifurcation. The existence of bionomic equilibrium of the system has been discussed.
Keywords:
Equilibria
Further, we study a path of optimal harvesting policy by introducing the Pontryagins
Harvesting efforts maximum principle. Moreover we have found out a condition for diffusive instability of
Global stability a locally stable equilibrium. Finally, some numerical simulations are performed to justify
Optimal harvesting policy analytical findings.
Diffusion © 2015 Elsevier Ltd. All rights reserved.

1. Introduction

In the natural world, dynamical behavior exhibited by various ecological systems are very complicated. Prey–predator
communities are embedded in complex food webs. These webs are related by trophic (predation) as well as non-trophic
interactions (e.g., competition, mutualism and interference). The top predator which appears at the bottom of the food
chain. Predator mainly depends on prey species. They provide an important food source for predator and therefore form an
integral part of the marine food web.
Predator–prey systems model some biological phenomenons and relationship between predator and prey in the real
world that play a crucial role in mathematics and have been extensively considered in many ways by many researchers
[1–9]. Recently, The researcher in [10] studied the existence and the non existence of non constant positive steady states of
prey–predator reaction–diffusion system, which indicates the effect of large diffusivity. The diffusive Lotka–Volterra preda-
tor–prey system with two delays is analyzed in [11]. Also, a spatio-temporal stage-structured prey–predator system with
stage-structure of predator only given by reaction–diffusion equations based on the Holling type III functional response has
been discussed in [12].

∗ Corresponding author. Tel.: +91 33 25666571; fax: +91 33 25828282.


E-mail addresses: chatterjeeanal172@gmail.com (A. Chatterjee), samaresp@yahoo.co.in (S. Pal).

http://dx.doi.org/10.1016/j.camwa.2015.12.022
0898-1221/© 2015 Elsevier Ltd. All rights reserved.
2 A. Chatterjee, S. Pal / Computers and Mathematics with Applications ( ) –

In the history of population ecology, both mathematicians and ecologists have a great interest in the Holling type
prey–predator models [13–15] including Holling type I–III, originally due to Holling [16]. Some experiments and obser-
vations indicate that a non-monotonic response occurs when the nutrient concentration reaches a high level [17–19]. At
this level, an inhibitory effect on the specific growth rate may occur. To model such an inhibitory effect, [20] proposed the
Monod–Haldane function, and also called a Holling type-IV function. Ecologists have used Holling type-IV function phe-
nomenologically to describe situations where predator interference or induced prey defenses lead to decreased predator
success at high predator density [21,22].
The authors in [23] obtained condition for the stability analysis of a food chain system Holling type IV functional response.
In [24,25], the scientist used the response function in a mite prey–predator interaction model and called it a Holling type IV
function.
The authors in [26] have studied three species food chain system with Holling type IV functional response and periodic
constant impulsive perturbation of top predator. Recently, prey–predator model with Holling IV functional response has
been studied in [27]. The authors studied chaotic dynamics. They also discussed preys and predators can tend to stable
equilibria when the preys and predators are in a chaotic dynamic by using chaos control. A discrete-time prey–predator
system with Holling type-IV function response in the presence of some alternative food to predator and harvesting of prey
species have been studied in [28]. The authors analyzed that the suitable price of resources can control the excessive harvest
to promote the sustainable development of species. The existence of bionomic equilibrium as well as optimal harvest policy
of prey–predator system have been studied in [29]. Different bifurcations in ratio-dependent predator–prey model have
been discussed in [30]. In recent past, an optimal control problem in a diffusive predator–prey system with time delay and
prey harvesting has been discussed in [31]. The scientists in [32,33], formulated and analyzed an optimal control problem
with drug treatment on the host system and multi-group coupled within-host system. A multi-delayed predator–prey model
in which the prey species is subject to harvesting under stochastic environment has been analyzed in [34]. In [35], the
author studied Kolmogorov-type predator–prey models, generalized Gause-type predator–prey models with harvesting. In
recent past, the scientists in [36] analyzed the stability and direction of periodic solutions of functional differential equations
and partial differential equations in a delayed prey–predator model with harvesting. Recently, phytoplankton–zooplankton
model with harvesting is proposed and investigated for the existence of bionomic equilibria and the optimal harvesting
policy [37–39]. In our earlier work, we have used a Holling type-II harvest function to model density dependent plankton
population and studied the effect of harvesting in marine ecosystem [40].
To exploit biological resources for the necessity of human beings and the society, harvesting is a very frequently used
process in our ecosystem. In general, harvesting terms have been used in the ecosystem in different ways. The most simple
and common way to harvest the ecological resources is when the resource population is harvested at a constant rate and
mathematically it is represented by h(t ) = h, where h being a constant. The disadvantage of the constant rate harvesting
is that it is independent of the density of the harvesting stock. On the other hand another harvesting strategy is based on
the catch-per-unit-effort (CPUE) hypothesis and mathematically governed by the equation as h(t ) = qEx(t ), where q is the
catchability coefficient, E is the constant external effort and x(t ) is the density of the harvested species at time t.
In this paper, one prey and two predator interaction model in presence of harvesting is described in the unreserved zone.
The stability of equilibrium point is analyzed. Conditions for global asymptotic stability of positive interior equilibrium
have been studied. We have also derived the conditions for instability of the system around the interior equilibrium and
Hopf bifurcation. Further, we have studied existence of bionomic equilibrium and a path of optimal harvesting policy by
introducing the Pontryagins maximum principle. Finally, we have analogized a condition for diffusive instability of a locally
stable equilibrium. Numerical simulations under a set of parametric values have been performed to support our analytical
result. We end the paper with a brief conclusion.

2. The mathematical model

Let P (t ) be the concentration of the prey population in the unreserved zone at time t with carrying capacity K and con-
stant intrinsic growth rate r. Let Z (t ) and F (t ) denote the middle population and top predator in the unreserved zone at
time t.
Let α1 be the maximal ingestion rate for middle predator and α2 be the maximal conversion rate for growth of middle
predator respectively (α1 ≤ α2 ). Let β1 be the maximal ingestion rate for top predator and β2 be the maximal conversion
rate for growth of top predator respectively (β1 ≤ β2 ). Here a1 and c1 are positive constant parameters and b1 and d1 are half
saturation constant for middle predator and top predator respectively. Let µ1 , µ2 be the mortality rate of the middle predator
and top predator respectively. h1 , h2 and h3 are the harvesting rates of prey, middle predator and top predator respectively.
Thus the following system of equation is given by:
α1 PZ
 
dP P
= rP 1 − − − h1 P ≡ G1 (P , Z , F )
dt K 1 + a1 P + b 1 P 2
dZ α2 PZ β1 ZF
= − − (h2 + µ1 )Z ≡ G2 (P , Z , F )
dt 1 + a1 P + b1 P 2 1 + c1 Z + d1 Z 2
dF β2 ZF
= − (h3 + µ2 )F ≡ G3 (P , Z , F ). (1)
dt 1 + c1 Z + d1 Z 2
A. Chatterjee, S. Pal / Computers and Mathematics with Applications ( ) – 3

The system (1) will be analyzed with the following initial conditions,

P (0) > 0, Z (0) > 0, F (0) > 0. (2)

3. Some preliminary results

3.1. Positive invariance

By setting X = (P , Z , F )T ∈ R3 and G(X ) = [G1 (X ), G2 (X ), G3 (X )]T , with G : R+ 3 → R3 and G ∈ C ∞ (R3 ), Eq. (1)
becomes

Ẋ = G(X ), (3)

together with X (0) = X0 ∈ R+ . It is easy to check that whenever X (0) ∈ R+ with Xi = 0, for i = 1, 2, 3, then
3 3

Gi (X ) |Xi =0 ≥ 0. Then any solution of Eq. (3) with X0 ∈ R+ 3 , say X (t ) = X (t ; X0 ), is such that X (t ) ∈ R+ 3 for all t > 0.

3.2. Boundedness of the system

Theorem 1. The model system (1) is dissipative.

Proof. See Appendix.

3.3. Equilibria

The system (1) possesses the following equilibria:

(i) The prey free equilibrium S0 = (0, 0, 0).


K (r −h )
(ii) The middle predator free equilibrium S1 = (P1 , 0, 0) = ( r 1 , 0, 0), which exists if r > h1 .
√ equilibrium is S2 = (P2 , Z2 , 0) where P2 and Z2 are given as follows:
(iii) The top predator free
α2 −a1 (h2 +µ1 )± {α2 −a1 (h2 +µ1 )}2 −4b1 (h2 +µ1 ) (rK −h1 K −rP2 )(1+a1 P2 +b1 P22 )
P2 = 2b1 (h2 +µ1 )
and Z2 = α1 K
, which exists if α2 > a1 (h2 + µ1 ) and
(r −h1 )K
0 < P2 < r
.

{β2 −c1 (h3 +µ2 )}± (β2 −c1 (h3 +µ2 ))2 −4(h3 +µ2 )2 d1
(iv) The positive interior equilibrium is S ∗ (P ∗ , Z ∗ , F ∗ ) where Z ∗ = 2(h3 +µ2 )d1
, F∗ =
[−(h2 +µ1 )b1 P ∗2 +{α2 −a1 (h2 +µ1 )}P ∗ −(h2 +µ1 )] a1 (rK −h1 K ) ∗2
(1+a1 P ∗ +b1 P ∗2 )β1
and P ∗ is given from cubic equation P ∗3 + { b1
− r
P } + { b11 −
a1 (rK −h1 K ) α1 KZ ∗
rb1
}P + (

rb1
rK −h1 K
− rb1
) = 0 and this equation can be written as

f (P ∗ ) = AP ∗3 + BP ∗2 + CP ∗ + G1 = 0. (4)

K (r −h1 ) α KZ ∗ K (r −h1 ) α1 KZ ∗
We note that 0 < P ∗ < . We have f (0) = G1 = 1rb − < 0 if Z ∗ < and f ( )= > 0. So the
rK −h1 K r −h 1
r  1  rb1 α1 r rb1
K (r −h1 )
positive root of P of the above equation lies in 0,

r
. Also Z is positive if β2 > c1 (h3 + µ2 ).

3.4. Eigenvalue analysis and Hopf-bifurcation

In this section, local stability analysis of the system around the biologically feasible equilibria is performed. It is clear that
S0 (0, 0, 0) is always unstable.

α2 P1
Lemma 1. If R0 =
(h2 +µ1 )(1+a1 P1 +b1 P12 )
> 1 then the middle predator free steady state S1 of the system (1) is unstable.

Proof. See Appendix.

Lemma 1 demonstrates that if the ratio of maximal conversion rate for the growth of middle predator and the loss rate
of the middle predator is greater than to unity then the middle predator free equilibrium S1 is unstable.

Lemma 2. There exists a feasible top predator free steady state S2 of the system (1) which is unstable if at least one of the conditions
holds
4 A. Chatterjee, S. Pal / Computers and Mathematics with Applications ( ) –

1
P2 > , (5)
b
2rP2 α1 Z2 − b1 α1 P22 Z2
+ + h1 < r , (6)
K (1 + a1 P2 + b1 P22 )2
β2 Z2
R1 = > 1. (7)
(µ2 + h2 )(1 + c1 Z2 + d1 Z22 )

Proof. See Appendix.

Eq. (7) indicates that the ratio of maximal conversion rate for the growth of top predator and the loss rate of the top predator
is greater than to unity then the top predator free equilibrium S2 is unstable.
Stability analysis of the positive interior equilibrium of the system (1)
The jacobian matrix of system (1) around the positive equilibrium is given as follows: S ∗ = (P ∗ , Z ∗ , F ∗ ) is

2rP ∗ α1 Z ∗ − b1 α1 P ∗2 Z ∗ α1 P ∗
 
r− − − h1 − 0

 K (1 + a1 P ∗ + b1 P ∗2 ) 1 + a 1 P ∗ + b 1 P ∗2 

α2 Z ∗ − b1 α2 P ∗2 Z ∗ β1 c1 Z ∗ F ∗ + 2β1 d1 Z ∗2 F ∗ β1 Z ∗
 

 
V =  − 

 (1 + a1 P ∗ + b1 P ∗2 ) (1 + c1 Z ∗ + d1 Z ∗2 )2 ∗ ∗
1 + c1 Z + d1 Z 2

 
 β2 F ∗ − β2 d1 Z ∗2 F ∗ 
0 0
(1 + c1 Z ∗ + d1 Z ∗2 )2
n11 n12 0
 
= n21 n22 n23 .
0 n32 0

The characteristic equation is

Q 3 + A1 Q 2 + A2 Q + A3 = 0,

where A1 = −(n11 + n22 ), A2 = n11 n22 − n32 n23 − n12 n21 , A3 = n11 n32 n23 .
If A1 > 0, A3 > 0 and A1 A2 − A3 > 0. Then according the Routh–Hurwitz criteria, all roots of above equation have
negative real parts. Thus S ∗ is locally asymptotically stable.

Theorem 2. When the intrinsic growth rate of prey, r, crosses a critical value, say r ∗ , the system (1) enters into a Hopf-bifurcation
around the positive equilibrium, which induces oscillations of the populations.

Proof. See Appendix.

3.5. Global stability

Local stability in a model ecosystem establish through eigenvalue analysis may be of little practical significance for a
real ecosystem because it guarantees only stability relative to small perturbations of the initial state from the equilibrium.
The Lyapunov function is proportional to the energy embodied in the species standing biomass. The condition that dV dt
is
negative means energy is continuously dissipated at relatively large population densities. Also the condition that dV
dt
is
negative indicates that, at relatively low population densities the ecosystem is continuously absorbing energy from an
external source [41]. In multi species interactions, the Lyapunov function is simply the weighted sum of the Lyapunov
function for each member species.

Theorem 3. If the non negative equilibrium S ∗ exists, then (P ∗ , Z ∗ , F ∗ ) is globally asymptotically stable in the P − Z − F plane.

Proof. See Appendix.

4. Bionomic equilibrium

In this section, we study the bionomic equilibrium of the model system (1). We replace h1 = q1 E1 , h2 = q2 E2 , and
h3 = q3 E3 of system (1). q1 , q2 and q3 represent catchability coefficients of prey, middle predator and top predator
respectively in unreserved zone. E1 , E2 and E3 represent the efforts applied to harvest the prey, middle predator and top
A. Chatterjee, S. Pal / Computers and Mathematics with Applications ( ) – 5

predator species respectively in unreserved zone. s1 , s2 and s3 be the harvesting cost per unit effort for prey species, middle
predator and top predator species respectively. p1 , p2 and p3 be the price per unit biomass of the prey, middle predator and
top predator species respectively. Therefore, the net economic revenue at any time t is given by
π = (p1 q1 PE1 − s1 E1 ) + (p2 q2 ZE2 − s2 E2 ) + (p3 q3 FE3 − s3 E3 )
= π1 + π2 + π3 . (8)
Here π1 = (p1 q1 P − s1 )E1 , π2 = (p2 q2 Z − s2 )E2 , π3 = (p3 q3 F − s3 )E3 ; that is, π1 , π2 and π3 represent the net revenues for
the prey, middle predator and top predator species respectively.
The bionomic equilibrium (P∞ , Z∞ , F∞ , E1∞ , E2∞ , E3∞ ) given by the following simultaneous equations:

α1 PZ
 
P
rP 1− − − q1 E1 P = 0, (9)
K 1 + a1 P + b 1 P 2
α2 PZ β1 ZF
− − µ1 Z − q2 E2 Z = 0, (10)
1 + a1 P + b 1 P 2 1 + c1 Z + d1 Z 2
β2 ZF
− µ2 F − q3 E3 F = 0, (11)
1 + c1 Z + d1 Z 2
(p1 q1 PE1 − s1 E1 ) + (p2 q2 ZE2 − s2 E2 ) + (p3 q3 FE3 − s3 E3 ) = 0. (12)
In order to determine the bionomic equilibrium, we now consider only for non-trivial bionomic equilibrium point
(P∞ , Z∞ , F∞ , E1∞ , E2∞ , E3∞ ): If s1 < p1 q1 P , s2 < p2 q2 Z and s3 < p3 q3 F , then the revenues for all the species are being
positive then the whole fishery will be in operation. In this case P∞ = p 1q , Z∞ = p 2q and F∞ = p 3q . Now substituting
s s s
1 1 2 2 3 3
P∞ , Z∞ and F∞ into (9)–(11), we get
α1 Z∞
   
1 P∞
E1∞ = r 1− − 2
q1 K 1 + a1 P∞ + b1 P∞
α 2 P∞ β1 F∞
 
1
E2∞ = 2
− 2
− µ1
q2 1 + a1 P∞ + b1 P∞ 1 + c1 Z∞ + d1 Z∞
β2 Z∞
 
1
E3∞ = 2
− µ 2 . (13)
q3 1 + c1 Z∞ + d1 Z∞
Now E1∞ > 0 if

α1 Z∞
 
P∞
r 1− > 2
, (14)
K 1 + a1 P ∞ + b 1 P ∞
E2∞ > 0 if
α2 P∞ β 1 F∞
2
> 2
+ µ1 , (15)
1 + a1 P ∞ + b 1 P ∞ 1 + c1 Z∞ + d1 Z∞
E3∞ > 0 if
β2 Z∞
2
> µ2 . (16)
1 + c1 Z∞ + d1 Z∞
Thus the nontrivial bionomic equilibrium point (P∞ , Z∞ , F∞ , E1∞ , E2∞ , E3∞ ) exists if conditions (14)–(16) hold.

5. Optimal harvesting policy

In this section, we study optimal harvesting policy of the system (1) in unreserved zone. Here we introduce the
Pontryagin’s maximum principle to obtain a path of optimal harvesting policy. We consider the following present value
J of a continuous time-stream [42].
 ∞
J = L(P , Z , F , E1 , E2 , E3 , t )e−δ t dt , (17)
0

where L is the net revenue given by


L(P , Z , F , E1 , E2 , E3 , t ) = (p1 q1 PE1 − s1 E1 ) + (p2 q2 ZE2 − s2 E2 ) + (p3 q3 FE3 − s3 E3 ), (18)
δ denotes the instantaneous annual rate of discount, the aim of this section is to maximize J subjected to the three equations
of the system (1).
6 A. Chatterjee, S. Pal / Computers and Mathematics with Applications ( ) –

Firstly, we construct the following Hamiltonian function

H = e−δ t [[p1 q1 P − s1 ] E1 + [p2 q2 Z − s2 ] E2 + [p3 q3 F − s3 ] E3 ]


α1 PZ
   
P
+ λ1 rP 1 − − − q1 E1 P
K 1 + a1 P + b 1 P 2
α2 PZ β1 ZF β2 ZF
   
+ λ2 − − µ1 Z − q2 E2 Z + λ3 − µ2 F − q3 E3 F , (19)
1 + a1 P + b1 P 2 1 + c1 Z + d1 Z 2 1 + c1 Z + d1 Z 2

where λ1 , λ2 and λ3 are additional unknown functions called the adjoint variables, E1 , E2 and E3 are the control variables
satisfying the constraints 0 ≤ E1 ≤ (E1 )max , 0 ≤ E2 ≤ (E2 )max , 0 ≤ E3 ≤ (E3 )max , and φ1 (t ) = e−δ t (p1 q1 P −s1 )−λ1 q1 P , φ2 =
e−δ t (p2 q2 Z − s2 ) − λ2 q2 Z , φ3 = e−δ t (p3 q3 F − s3 ) − λ3 q3 F are called the switching functions. Our aim is to find an optimal
equilibrium ((P )δ , (Z )δ , (F )δ , (E1 )δ , (E2 )δ , (E3 )δ , ) to maximize Hamiltonian H.
Since Hamiltonian H is linear in the control variables E1 , E2 and E3 , the optimal control can be extreme controls or the
singular controls, thus we have E1 = (E1 )max , when φ1 (t ) > 0 i.e., when λ1 (t )eδ t < p1 − q 1P ; E1 = 0, when φ1 (t ) < 0
s
1
i.e., when λ1 (t )eδ t > p1 − =( ) , φ (t ) > 0 i.e., when λ2 (t )eδt < p2 − qs22Z ; E2 = 0, when φ2 (t ) < 0
s1
q1 P
; E2 E2 max when 2
i.e., when λ2 (t )eδ t > = ( ) , when φ3 (t ) > 0 i.e., when λ3 (t )eδt < p3 − qs33F ; E3 = 0, when
s2
− p2 q2 Z
; similarly E3 E3 max
δt
φ3 (t ) < 0 i.e., when λ ( ) >3 t e p3
s3

q3 F
; when 1 t φ ( ) = 0, λ1 (t )eδt = p1 − qs11P or

∂H
= 0, (20)
∂ E1
when φ2 (t ) = 0, λ2 (t )eδ t = p2 −
s2
q2 Z
or

∂H
= 0, (21)
∂ E2
when φ3 (t ) = 0, λ3 (t )eδ t = p3 −
s3
q3 F
or

∂H
= 0. (22)
∂ E3
In this case, the optimal control is called the singular control, and (20), (21), (22) are the necessary conditions for the
maximization of Hamiltonian H. By Pontryagin’s maximal principle, the adjoint equations are

dλ1 ∂H
= −
dt ∂P
α1 Z − b1 α1 P 2 Z α2 Z − b1 α2 P 2 Z
 
2rP
= − e−δt p1 q1 E1 + λ1 (r − − − q1 E 1 ) + λ2 ( ) , (23)
K (1 + a1 P + b1 P 2 )2 (1 + a1 P + b1 P 2 )2
dλ2 ∂H α1 P
  
= − = − e−δt p2 q2 E2 + λ1 −
dt ∂Z  1 + a1 P + b 1 P 2
α2 P β1 F − β1 d1 Z 2 F β1 F − β1 d1 Z 2 F
  
+ λ2 − − µ 1 − q 2 E 2 + λ 3 − µ 2 − q3 E 3 ,
1 + a1 P + b 1 P 2 (1 + c1 Z + d1 Z 2 )2 (1 + c1 Z + d1 Z 2 )2
(24)
dλ3 ∂H
= −
dt ∂F
β1 Z β2 Z
    
= − e −δ t
p3 q3 E3 − λ2 + λ3 − µ2 − q3 E3 . (25)
1 + c1 Z + d1 Z 2 1 + c1 Z + d1 Z 2

Considering the interior equilibrium S ∗ (P ∗ , Z ∗ , F ∗ ) and substituting the value of λ2 from (21) in (23) we get
dλ1
− λ1 B1 = −B2 e−δt
dt
a1 α1 P ∗ Z ∗ +2b1 α1 P ∗2 Z ∗
)( α12+Z a−Pb∗1+α2bP P ∗Z2 ).
∗ ∗2 ∗
rP ∗
where B1 = K
− (1+a1 P ∗ +b1 P ∗2 )
, B2 = p1 q1 E1 + (p2 − s2
q2 Z ∗ 1 1
We can calculate that
B2
λ1 = e−δ t . (26)
B1 + δ
A. Chatterjee, S. Pal / Computers and Mathematics with Applications ( ) – 7

Again considering the interior equilibrium S ∗ (P ∗ , Z ∗ , F ∗ ) and substituting the value of λ1 and λ3 from (22) in (24) we get
dλ2
− λ2 B3 = −B4 e−δt
dt
β1 c1 F ∗ Z ∗ +2β1 d1 Z ∗2 F ∗ α1 P ∗
)( β11+F c−β
∗ ∗2 ∗
where B3 = (1+c1 Z ∗ +d1 Z ∗2 )2
, B4 = p2 q2 E2 + B2
B1 +δ
(− 1+a ∗ ∗2 ) + (p3 − s3
q3 F ∗
1 d1 Z F
Z ∗ +d Z ∗2
− µ2 − q3 E3 ).
1 P +b 1 P 1 1
We can calculate that
B4
λ2 = e−δ t . (27)
B3 + δ
Similarly, by considering the interior equilibrium S ∗ (P ∗ , Z ∗ , F ∗ ) and substituting the value of λ2 from (27) in (25) we get
dλ3
= −B5 e−δt
dt
β1 Z ∗
where B5 = p3 q3 E3 −
B4
B3 +δ 1+c1 Z ∗ +d1 Z ∗2
. We can calculate that

B5 −δ t
λ3 = e . (28)
δ
It is obviously that λ1 (t ), λ2 (t ) and λ3 (t ) are bounded as t −→ ∞.
Substituting (20) and (26) we subsist a singular path

s1 B2
p1 − = . (29)
q1 P ∗ B1 + δ
Substituting (21) and (27) we get a singular path

s2 B4
p2 − = . (30)
q2 Z ∗ B3 + δ
Substituting (22) and (28) we obtain a singular path

s3 B5
p3 − = . (31)
q3 F∗ δ
Using P ∗ , Z ∗ and F ∗ we get the value of B1 , B2 , B3 , B4 and B5 . Thus from (29)–(31) it can be written as
 
s1 B2
F1 (P ) =

p1 − − ,
q1 P ∗ B1 + δ
 
s2 B4
G1 (Z ∗ ) = p2 − ,
q2 Z ∗ B3 + δ
 
s3 B5
H1 ( F ∗ ) = p3 − − .
q3 F ∗ δ
There exists a unique positive root P ∗ = (P ∗ )δ of F1 (P ∗ ) = 0 in the interval 0 < P ∗ < x1 if the following holds

F1 (0) < 0, F1 (x1 ) > 0, F1 (P ∗ ) > 0 for P ∗ > 0 and there exists a unique positive root Z ∗ = (Z ∗ )δ of G1 (Z ∗ ) = 0 in the

interval 0 < Z ∗ < y1 if the following holds G1 (0) < 0, G1 (y1 ) > 0, G1 (Z ∗ ) > 0 for Z ∗ > 0. Similarly, there exists a unique

positive root F ∗ = (F ∗ )δ of H1 (F ∗ ) = 0 in the interval 0 < F ∗ < z1 if the following holds H1 (0) < 0, H1 (z1 ) > 0, H1 (F ∗ ) > 0
for F ∗ > 0. Thus we get P ∗ = (P ∗ )δ , Z ∗ = (Z ∗ )δ , F ∗ = (F ∗ )δ ,
(P )δ α1 (Z )δ
 
1
(E1 )δ = r (1 − )− ,
q1 K 1 + a1 (P )δ + b1 (P )2δ
α2 (P )δ β1 (F )δ
 
1
(E2 )δ = − − µ1 ,
q2 1 + a1 (P )δ + b1 (P )2δ 1 + c1 (Z )δ + d1 (Z )2δ
β2 (Z )δ
 
1
(E3 )δ = − µ2 .
q3 1 + c1 (Z )δ + d1 (Z )2δ

Hence once the optimal equilibrium, ((P )δ , (Z )δ , (F )δ ) is determined, the optimal harvesting effort (E1 )δ , (E2 )δ , (E3 )δ can
be determined. From (20)–(22) and (26)–(28), we find that λi (t ) where i = 1, 2, 3 do not change with time in optimal
equilibrium. Hence they remain bounded as t → ∞.
8 A. Chatterjee, S. Pal / Computers and Mathematics with Applications ( ) –

6. Stability analysis in the presence of diffusion

We have also incorporated the spatial effects on the system (1) via simple diffusions. If dP , dZ , dF are the self diffusion
coefficients of prey, middle predator and top predator population respectively then the model system (1) in the presence of
one-dimensional diffusion has the following form:

∂P α1 PZ
 
P
= rP 1 − − − h1 P + d P ∇ 2 P
∂t K 1 + a1 P + b1 P 2
∂Z α2 PZ β1 ZF
= − − (h2 + µ1 )Z + dZ ∇ 2 Z
∂t 1 + a1 P + b 1 P 2 1 + c1 Z + d1 Z 2
∂F β2 ZF
= − (h3 + µ2 )F + dF ∇ 2 F . (32)
∂t 1 + c1 Z + d1 Z 2

To study the effect of diffusion on the model system, we have considered the linearized form of the system (32) about
(P̂ , Ẑ , F̂ ) is:
∂U
= n̂11 U + n̂12 V + n̂13 W + dP ∇ 2 P
∂t
∂V
= n̂21 U + nˆ22 V + n̂23 W + dZ ∇ 2 Z
∂t
∂W
= n̂31 U + n̂32 V + n̂33 W + dF ∇ 2 F , (33)
∂t
α1 Ẑ −b1 α1 P̂ 2 Ẑ
where P = P̂ + U , Z = Ẑ + V , F = F̂ + W and n̂11 = r − 2r P̂
K

(1+a1 P̂ +b1 P̂ 2 )
− h1 , n̂12 = − 1+a αP̂1+P̂ b P̂ 2 , n̂13 = 0, n̂21 =
1 1
α2 Ẑ −b1 α2 P̂ 2 Ẑ α2 P̂ β1 F̂ −β1 d1 Ẑ 2 F̂
µ1 ), n̂23 = − 1+c βẐ1+Ẑd Ẑ 2 , n̂31 = 0, n̂32 = (1β+2 F̂c−β
2

(1+a1 P̂ +b1 P̂ 2 )
, n̂22 = − (1+c1 Ẑ +d1 Ẑ 2 )
− ( h2 + 2 d1 Ẑ F̂
, n̂33 =
1+a1 P̂ +b1 P̂ 2 1 Ẑ +d1 Ẑ )
2 2
1 1
β2 Ẑ
1+c1 Ẑ +d1 Ẑ 2
− (h3 + µ2 ).
It may be noted that (U , V , W
 )  perturbations of (P , Z , F ) about the equilibrium point (P̂ , Ẑ , F̂ ). We start by
are small
U v1
assuming solutions of the form V = v2 eλt +i(kx x+ky y) where λ > 0 is the frequency, vi > 0 represents the amplitude
W v3
(i = 1, 2, 3) and kx , ky > 0 are wave number of the perturbations in time t. Substituting k2 = k2x + k2y , the system (33)
becomes

∂U
= (n̂11 − dP k2 )U + n̂12 V + n̂13 W
∂t
∂V
= n̂21 U + (nˆ22 − dZ k2 )V + n̂23 W
∂t
∂W
= n̂31 U + n̂32 V + (n̂33 − dF k2 )W . (34)
∂t
The eigenvalues for the steady state S0 (0, 0, 0), obtained from the characteristic equation of the linearized system (34)
are r − h1 − dP k2 , −(h2 + µ1 ) − dZ k2 and −(h3 + µ2 ) − dF k2 . In absence of diffusion, S0 is always unstable. Under the
same conditions S0 (in presence of diffusion) is also unstable. Again the eigenvalues of the jacobian matrix for the steady state
α2 P 1
S1 (P1 , 0, 0) are −(r − h1 )− dP k2 , −(h3 +µ2 )− dF k2 and (h2 +µ1 )[ − 1]− dZ k2 = (h2 +µ1 )[R0 − 1]− dZ k2 .
2 (1+a1 P1 +b1 P1 )(h2 +µ1 )
In absence of diffusion, S1 is locally asymptotically stable for R0 < 1. Under the same conditions S1 (in presence of diffusion)
is also spatially stable.
Again the eigenvalues of the jacobian matrix for the steady state S2 = (P2 , Z2 , 0) in presence of diffusion are λ1 ′ , λ2 ′
which are the roots of the equation

α1 P2 (α2 Z2 − b1 α2 P22 Z2 )
λ2 + λT + TdZ k2 + = 0,
(1 + a1 P2 + b1 P22 )3

α1 Z2 −b1 α1 P22 Z2
 
β2 Z 2
where T = −r + 2rP2
K
+ (1+a1 P2 +b1 P22 )2
+ h1 + dP k2 + dZ k2 and λ3 ′ = 1+c1 Z2 +d1 Z 2
− (h3 + µ2 ) − d2F k2 . It is clear that S2
is unstable i.e., one of eigenvalue is positive in presence of diffusion if any of (5)–(7) conditions hold. Thus under the same
conditions S2 (in presence of diffusion) is also spatially stable.
A. Chatterjee, S. Pal / Computers and Mathematics with Applications ( ) – 9

Fig. 1. (a) The equilibrium point S ∗ is stable for the parametric values as given in this paper. (b) The figure depicts oscillatory behavior around the positive
interior equilibrium point S ∗ of system (1) for increasing r, from 3.7 to 4 with other parametric values are unaltered. (c) The figure depicts oscillatory
behavior around the positive interior equilibrium point S ∗ of system (1) for increasing K , from 0.5 to 0.58 with same set of other parametric values. (d) The
figure depicts top predator free equilibrium point S2 of system (1) for decreasing β2 , from 0.2 to 0.15 with other parametric values are unaltered.

At S ∗ , the characteristic equation of the linearized system (34) is λ3 + A∗1 λ2 + A∗2 λ + A∗3 = 0, where A∗1 =
A1 + (dP + dZ + dF )k2 , A∗2 = A2 − [n22 dP + n11 dZ + (n11 + n22 )dF ] k2 + (dP dZ + dZ dF + dF dP )k4 and A∗3 = A3 +
∗ α Z ∗ −b α P ∗2 Z ∗
[−n23 n32 dP + (n11 n22 − n12 n21 )dF ] k2 −(n11 dZ dF + n22 dP dZ )k4 + dP dZ dF k6 . Here n11 = r − 2rP
K
− (11+a P ∗1+1b P ∗2 ) − h1 , n12 =
1 1
α1 P ∗ α Z ∗ −b α P ∗ 2 Z ∗ β1 c1 Z ∗ F ∗ +2β1 d1 Z ∗2 F ∗ β Z∗
− 1 +a ∗ +b P ∗ 2 , n13 = 0, n21 = (12+a P ∗1+2b P ∗2 ) , n22 = (1+c1 Z ∗ +d1 Z ∗2 )2
, n23 = − 1+c Z ∗1 +d Z ∗2 , n31 = 0, n32 =
1P 1 1 1 1 1
β2 F ∗ −β2 d1 Z ∗2 F ∗
(1+c1 Z +d1 Z ∗2 )2
∗ , n33 = 0.

Lemma 3. If (3A2 + 2n211 + 2n222 )d > − (3n11 n22 + n23 n32 + n12 n21 ) D, the stability of the non-spatial system (1) at S ∗ implies
the stability of the diffusive system (32) at S ∗ .

Proof. See Appendix.

√ for diffusive instability of the system at the positive equilibrium S is given by k H1 − k H2 + k H3 >
∗ 6 4 2
Lemma 4. The condition
A3 − A1 A2 and H2 ≥ 2 H3 H1 .

Proof. See Appendix.

7. Numerical simulations

In this section, we focus our attention on the dynamic behaviors of the prey–predator population by using some numerical
simulations. We begin with a parameter set (see Table 1, Ref. [43]) for which the existence condition of the coexistence
equilibrium point S ∗ = (2.81, 0.83, 1.44) is satisfied and the coexistence equilibrium point S ∗ is locally asymptotically
stable (cf. Fig. 1(a)).
10 A. Chatterjee, S. Pal / Computers and Mathematics with Applications ( ) –

Fig. 2. (a) The figure depicts oscillatory behavior around the positive interior equilibrium point S ∗ of system (1) for increasing β2 , from 0.2 to 0.25 with other
parametric values are unaltered. (b) The figure depicts oscillatory behavior around the positive interior equilibrium point S ∗ of system (1) for decreasing
µ2 , from 0.09 to 0.08 with other parametric values are unaltered. (c) The figure depicts oscillatory behavior around the positive interior equilibrium point
S ∗ of system (1) for decreasing h3 , from 0.01 to 0.003 with other parametric values are unaltered. (d) Bifurcation diagrams in terms of r with the other
parameters at the reference values given in Table 1.

7.1. Effect of r

Increasing the value of intrinsic growth rate of prey population r from 3.7 to 4 the system goes to oscillatory behavior of
the system (cf. Fig. 1(b)).

7.2. Effect of K

For K = 0.58, leaving all other parameters unaltered, the system exhibits persistent oscillations around the positive
interior equilibrium S ∗ (cf. Fig. 1(c)).

7.3. Effect of β2

Decreasing the value of β2 from 0.2 to 0.15, the system exhibits top predator free equilibrium (cf. Fig. 1(d)). But increasing
the value of β2 = 0.25, the system exhibits persistent oscillations (cf. Fig. 2(a)).
A. Chatterjee, S. Pal / Computers and Mathematics with Applications ( ) – 11

Fig. 3. (a) Bifurcation diagrams in terms of K with the other parameters at the reference values given in Table 1. (b) Biomass distribution of prey over time
and space of the model (32) for parameter values as given in this paper with diffusion coefficients dP = 0.7, dZ = 0.7 and dF = 0.7. (c) Biomass distribution
of middle predator over time and space of the model (32) for parameter values as given in this paper with diffusion coefficients dP = 0.7, dZ = 0.7 and
dF = 0.7. (d) Biomass distribution of top predator over time and space of the model (32) for parameter values as given in this paper with diffusion
coefficients dP = 0.7, dZ = 0.7 and dF = 0.7.

7.4. Effect of µ2

Decreasing the value of µ2 from 0.09 to 0.08 once again the system exhibits oscillatory behavior (cf. Fig. 2(b)).

7.5. Effect of h3

For h3 = 0.003, the system exhibits oscillatory behavior around the positive interior equilibrium S ∗ (cf. Fig. 2(c)).

7.6. Hopf-bifurcation

For a clear understanding of the dynamical implications of some parameter changes we plot suitable bifurcation
diagrams. We plot a bifurcation diagram for r, keeping the remaining parameters always at their reference values (cf.
Fig. 2(d)). Similarly, we plot another supercritical bifurcation diagram for K , with the threshold value for the bifurcation
given by K = 0.5744 with first Lyapunov coefficient (l1 = −5.562011e−001 ) (cf. Fig. 3(a)).
Now keeping all parameters fixed and with diffusion coefficients dP = 0.7, dZ = 0.7 and dF = 0.7 we have plotted
Fig. 3(b)–(d) which depict that all the species show stable biomass distribution and the interior positive equilibrium point
12 A. Chatterjee, S. Pal / Computers and Mathematics with Applications ( ) –

Fig. 4. (a) Biomass distribution of prey over time and space of the model (32) for r = 4 and other parameter values as given in this paper with diffusion
coefficients dP = 0.7, dZ = 0.7 and dF = 0.7, the system is oscillatory about S ∗ . (b) Biomass distribution of middle predator over time and space of the
model (32) for r = 4 and other parameter values as given in this paper with diffusion coefficients dP = 0.7, dZ = 0.7 and dF = 0.7, the system is oscillatory
about S ∗ . (c) Biomass distribution of top predator over time and space of the model (32) for r = 4 and other parameter values as given in this paper with
diffusion coefficients dP = 0.7, dZ = 0.7 and dF = 0.7, the system is oscillatory about S ∗ .

Table 1
A set of parametric values.
Parameter Definition Value

r Constant intrinsic growth rate for prey 3.7


K Carrying capacity for prey 0.5
α1 Maximal ingestion rate for middle predator 2
α2 Maximal conversion rate for growth of middle predator 1.8
β1 Maximal ingestion rate for top predator 0.8
β2 Maximal conversion rate for growth of middle predator 0.2
a1 Constant parameter 0.5
b1 Half saturation constant for middle predator 0.5
c1 Constant parameter 0.4
d1 Half saturation constant for top predator 0.4
µ1 Mortality rate of the middle predator 0.06
µ2 Mortality rate of the top predator 0.09
h1 Harvesting rates of prey 0.03
h2 Harvesting rates of middle predator 0.03
h3 Harvesting rates of top predator 0.01

S ∗ is spatially stable. The oscillatory behavior of the diffusive system is observed for high value of intrinsic growth rate for
r = 4 (cf. Fig. 4(a)–(c)). Increasing the value of K from 0.5 to 0.65, once again the system exhibits oscillatory behavior in
diffusive system (cf. Fig. 5(a)–(c)).
A. Chatterjee, S. Pal / Computers and Mathematics with Applications ( ) – 13

Fig. 5. (a) Biomass distribution of prey over time and space of the model (32) for K = 0.65 and other parameter values as given in this paper with diffusion
coefficients dP = 0.7, dZ = 0.7 and dF = 0.7, the system is oscillatory about S ∗ . (b) Biomass distribution of middle predator over time and space of the
model (32) for K = 0.65 and other parameter values as given in this paper with diffusion coefficients dP = 0.7, dZ = 0.7 and dF = 0.7, the system is
oscillatory about S ∗ . (c) Biomass distribution of top predator over time and space of the model (32) for K = 0.65 and other parameter values as given in
this paper with diffusion coefficients dP = 0.7, dZ = 0.7 and dF = 0.7, the system is oscillatory about S ∗ .

8. Conclusion

In this paper, we have studied a prey–predator interaction model, in which top predator predate middle predator and
middle predator captures prey. The model is realistic because we introduce the harvesting effort to remain under control to
keep the ecological balance. A fully dynamic interaction between prey–middle predator–top predator is one of the important
feature of this model.
Firstly, the prey–predator model is studied analytically and the threshold values for the feasibility and stability of the
four possible steady states are assessed. We have also studied global stability at positive interior equilibrium. We introduce
the Pontryagin’s maximum principle to obtain a path of optimal harvesting policy.
In what follows we further outline the investigations performed on the roles that varying the model parameters have,
considering changes in one of them.
Furthermore, analytical results for a Hopf bifurcation at the coexistence equilibrium are determined. Numerical evidence
supports this finding. The persistent oscillations are shown to occur for changes in various parameters.
The bifurcation diagrams of Figs. 2(d) and 3(a) show that increasing values of the intrinsic growth rate of prey as well
as of its carrying capacity leads to sustained population oscillations. The same outcome occurs if instead we decrease the
14 A. Chatterjee, S. Pal / Computers and Mathematics with Applications ( ) –

harvesting rate and mortality rate of top predator respectively. Also our investigation indicates that maximal conversion
rate for growth of top predator plays an important role to change different steady state behaviors.
In addition, We have also found that diffusive system is stable in suitable conditions involving both the model parameters
as well as self diffusion coefficients. Finally, we compare with the two model by changing the different parametric
values.

Acknowledgments

The paper has improved considerably, both in presentation and scientific content, by the critical comments made by the
reviewers. The authors are thankful to them. The research is partially supported by University Grants Commission, New
Delhi [grant number MRP-MAJ-MATH-2013-609].

Appendix

Proof of Theorem 1. From first equation of system (1) we have dP


dt
≤ rP (1 −
) ∀t ≥ 0 =⇒ P (t ) ≤ P
K
K
1+C1 Ke− rt
with
C1 = [ 1
P (0)
− K ] be the constant of integration. Hence for the large value of time we get P (t ) ≤ K ∀t ≥ 0.
1

α β α2 β1 α2
Let us define a function m(t ) = α2 P + Z + β1 F , dm
dt
+ h P
α1 1
+ (h2 + µ1 )Z + (h
β2 3
+ µ2 ) F = α1
rP (1 − KP ),
1 2

dm α2 rK
+ D0 m ≤ (35)
dt 4α1
where D0 = min{h1 , h2 + µ1 , h3 + µ2 }. It is clear that the right hand side of (35) is bounded.
Using the variation of constants formula, this inequality is transformed into
rK α2
m(t ) ≤ (1 − e−D0 t ) + m(0)e−D0 t ,
4α1 D0

and then all species are uniformly bounded for any initial value in R3+ .
Therefore for a biologically realistic model (1) has to be dissipative (i.e., all population are uniformly limited in time
by their environments). So according to the above theorem we assume that there exists (η1 , η2 , η3 ) > 0 such that
Ω (P (0), Z (0), F (0)) ⊂ R3+ = {(P , Z , F ) : 0 ≤ P ≤ η1 , 0 ≤ Z ≤ η2 , 0 ≤ F ≤ η3 , } for all (P (0), Z (0), F (0)) ≥ 0,
where Ω (P (0), Z (0), F (0)) is the omega limit set of the orbit initiating at (P(0), Z(0), F(0)). Hence the model system (1) is
dissipative. 

Proof of Theorem 2. The necessary and sufficient conditions for the existence of the Hopf-bifurcation for r = r ∗ if it exists
are (i) Ai (r ∗ ) > 0, i = 1, 2, 3, (ii) A1 (r ∗ )A2 (r ∗ ) − A3 (r ∗ ) = 0 and (iii) the eigenvalues of above characteristic equation should
be of the form λi = ui + ivi , and dri ̸= 0, i = 1, 2, 3.
du

We will now verify the Hopf-bifurcation condition (iii), putting λ = u + iv in the above equation, we get

(u + iv)3 + A1 (u + iv)2 + A2 (u + iv) + A3 = 0. (36)

On separating the real and imaginary parts and eliminating v , we get

8u3 + 8A1 u2 + 2u(A21 + A2 ) + A1 A2 − A3 = 0. (37)

It is clear from the above that u(r ∗ ) = 0 iff A1 (r ∗ )A2 (r ∗ ) − A3 (r ∗ ) = 0. Further, at r = r ∗ , u(r ∗ ) is the only root, since the
discriminate 8u2 + 8A1 u + 2(A21 + A2 ) = 0 if 64A21 − 64(A21 + A2 ) < 0. Further, differentiating (37) with respect to r, we
have 24u2 du
dr
+ 16A1 u du
dr
+ 2(A21 + A2 ) du
dr
+ 2u[2A1 dA1
dr
+ dA2
dr
]+ dS
dr
= 0 where S = A1 A2 − A3 .
− dS
Since at r = r ∗ , u(r ∗ ) = 0 we get ̸= 0.
 du  dr
dr r =r ∗
= 2(A21 +A2 )
This ensures that the above system has a Hopf-bifurcation around the positive interior equilibrium E ∗ . 

Proof of Theorem 3. The proof is based on a lyapunov direct method. Consider the following
  positiveC definite function
V (P , Z , F ) = V1 (P , Z , F )+V2 (P , Z , F )+V3 (P , Z , F ), where V1 (P , Z , F ) = C1 P − P ∗ − P ∗ ln PP∗ , V2 = 22 (Z − Z ∗ )2 , V3 =

C3
2
(F − F ∗ )2 .
Obviously, V is a continuous function on Int R3+ , Cj (j = 1, 2, 3) are the positive constants to be determined. Now we have
to investigate the global dynamics of the non negative equilibrium point S ∗ of the model system (1). Therefore the derivative
of V with respect to time along the solution of system (1) is computed as

dV dV1 dV2 dV3


= + + . (38)
dt dt dt dt
A. Chatterjee, S. Pal / Computers and Mathematics with Applications ( ) – 15

Since
α1 b1 Z ∗ (P + P ∗ + a1 ) α1 c1 (P − P ∗ )(Z − Z ∗ )
 
dV1 r
= −C1 − ( P − P )
∗ 2
− , (39)
dt K (1 + a1 P ∗ + b1 P ∗2 )(1 + a1 P + b1 P 2 ) 1 + a 1 P ∗ + b 1 P ∗2

α2 P β2 F ∗ (1 − ZZ ∗ d1 )
 
dV2
= −C2 (h2 + µ1 ) − + (Z − Z ∗ )2
dt 1 + a1 P + b 1 P 2 (1 + c1 Z + d1 Z 2 )(1 + c1 Z ∗ + d1 Z ∗2 )
C2 α2 Z ∗ (1 − b1 PP ∗ )(P − P ∗ )(Z − Z ∗ ) C2 β2 Z (Z − Z ∗ )(F − F ∗ )
+ − , (40)
(1 + a1 P + b1 P 2 )(1 + a1 P ∗ + b1 P ∗2 ) 1 + c1 Z + d1 Z 2

β2 Z C3 β2 F ∗ (1 − ZZ ∗ d1 )(Z − Z ∗ )(F − F ∗ )
 
dV3
= −C3 (h3 + µ2 ) − ( F − F )
∗ 2
+ . (41)
dt (1 + c1 Z + d1 Z 2 ) (1 + c1 Z + d1 Z 2 )(1 + c1 Z ∗ + d1 Z ∗2 )
Substituting (39)–(41) in (38) we obtain
α1 b1 Z ∗ (P + P ∗ + a)
 
dV r
= −C1 − (P − P ∗ )2
dt K ( 1 + a 1 P ∗ + b 1 P ∗2 )
α2 P ∗ β1 F (1 − ZZ ∗ d1 )
 
− C2 (h2 + µ1 ) − + (Z − Z ∗ )2
1 + a 1 P ∗ + b 1 P ∗2 (1 + c1 Z + d1 Z 2 )(1 + c1 Z ∗ + d1 Z ∗2 )
β2 Z
 
− C3 (h3 + µ2 ) − (F − F ∗ )2
1 + c1 Z + d1 Z 2

α1 C1 C2 α2 Z (1 − b1 PP ∗ )
 
+ (P − P ∗ )(Z − Z ∗ ) − +
1 + a1 P + b 1 P 2 (1 + a1 P + b1 P 2 )(1 + a1 P ∗ + b1 P ∗2 )
C3 β2 F ∗ (1 − ZZ ∗ d1 ) C2 β1 Z ∗
 
+ (Z − Z ∗ )(F − F ∗ ) − .
(1 + c1 Z + d1 Z 2 )(1 + c1 Z ∗ + d1 Z ∗2 ) 1 + C1 Z + d1 Z 2
The above equation can be written as sum of the quadratics
 
dV 1 1
= − a11 (P − P ∗ )2 + a12 (P − P ∗ )(Z − Z ∗ ) − a22 (Z − Z ∗ )2
dt 2 2
 
1 1
+ − a22 (Z − Z ) + a23 (Z − Z )(F − F ) − a33 (F − F )
∗ 2 ∗ ∗ ∗ 2
2 2
 
1 1
+ − a11 (P − P ∗ )2 + a13 (P − P ∗ )(F − F ∗ ) − a33 (F − F ∗ )2 ,
2 2
where
α1 b1 Z ∗ (P + P ∗ + a1 )
 
r
a11 = C1 − ,
K (1 + a1 P ∗ + b1 P ∗2 )
α2 P ∗ β1 F (1 − ZZ ∗ d1 )
 
a22 = C2 (h2 + µ1 ) − + ,
1 + a 1 P ∗ + b 1 P ∗2 (1 + c1 Z + d1 Z 2 )(1 + c1 Z ∗ + d1 Z ∗2 )
β2 Z
 
a33 = C3 (h3 + µ2 ) − ,
1 + c1 Z + d1 Z 2

α1 C1 C2 α2 Z (1 − b1 PP ∗ )
 
a12 = − + ,
1 + a1 P + b 1 P 2 (1 + a1 P + b1 P 2 )(1 + a1 P ∗ + b1 P ∗2 )
C3 β2 F ∗ (1 − ZZ ∗ d1 ) C2 β1 Z ∗
 
a23 = − ,
(1 + c1 Z + d1 Z 2 )(1 + c1 Z ∗ + d1 Z ∗2 ) 1 + C1 Z + d1 Z 2
a13 = 0.
dV
Sufficient conditions for dt
to be negative definite are that the following inequalities hold:

a11 > 0, a22 > 0, a33 > 0, (42)

a212 < a11 a22 , (43)


16 A. Chatterjee, S. Pal / Computers and Mathematics with Applications ( ) –

a223 < a22 a33 , (44)

a213 < a11 a33 . (45)


r (1+a1 P ∗ +b1 P ∗2 )(1+a1 K +b1 K 2 )
Now a11 > 0 if Z ∗ < K α1 b1 (P ∗ +a1 )
.
 
α2 P ∗ β1 d1 Z ∗ ζ η3
Further, a22 > 0 if h2 ( +µ )− 1 1+a P ∗ +b P ∗2
> 1+c1 Z ∗ +d1 Z ∗2
.
1 1

Also, a33 > 0 if h3 + µ2 > β2 ζ where ζ = √ 1


2 d1 +c1
is the maximum value of 1+c ZZ+d Z 2 . Now a11 a22 > a212 if
1 1

α1 b1 Z ∗ (η1 + P ∗ + a1 ) α2 P ∗ β1 η3 (1 − η2 Z ∗ d1 )
  
r
− ( h 2 + µ1 ) − +
K (1 + a1 P ∗ + b1 P ∗2 )M ∗
1 + a1 P + b 1 P ∗ 2 M1 (1 + c1 Z ∗ + d1 Z ∗2 )

(1 − b1 η1 P ∗2 )2
 
>
M (1 + a1 P ∗ + b1 P ∗2 )2
2

a21 c12
where M = 1 − 4b1
, M1 = 1 − 4d1
and taking C1 = C2 α22 η22 . Choosing C2 = C3 β22 F ∗2 , then a22 a33 > a223 if

α2 P ∗ β1 η3 (1 − η2 Z ∗ d1 ) β2 η2 (1 − η2 Z ∗ d1 )2
  
( h 2 + µ1 ) − + (h3 + µ2 ) − > .

1 + a1 P + b 1 P ∗ 2 (1 + c1 Z ∗ + d1 Z ∗2 )M M1 M12 (1 + c1 Z ∗ + d1 Z ∗2 )2
Clearly < a11 a33 since a13 = 0. Therefore S (P , Z , F ) is globally asymptotically stable provided satisfying all the
a213 ∗ ∗ ∗ ∗

above conditions. 

Stability analysis of the system (1).


Let S = (P , Z , F ) be any arbitrary equilibrium. Then the jacobian matrix about S is given by
 
2
2rP α1 Z − b1 α1 P Z α1 P
r − − − h1 − 0
 
2 2 
 K (1 + a1 P + b1 P ) 1 + a1 P + b1 P 
 
 
2 2
α2 Z − b1 α2 P Z α2 P β1 F − β1 d1 Z F β1 Z .
 
V =

2 2
− 2
− (h2 + µ1 ) − 2

(1 + a1 P + b1 P ) 1 + a1 P + b 1 P (1 + c1 Z + d1 Z ) 1 + c1 Z + d1 Z
 
 
 
2
β2 F − β2 d1 Z F β2 Z
 
( µ )
 
0 2 2
− h3 + 2
(1 + c1 Z + d1 Z ) 1 + c1 Z + d1 Z

By computing the jacobian matrix for the equilibrium S0 of the system (1) we find that the eigenvalues of the jacobian matrix
V0 are λ1 = r − h1 > 0, λ2 = −(h2 + µ1 ), λ3 = −(h3 + µ2 ) < 0. Clearly S0 is always unstable.

Proof of Lemma 1. Now again computing the jacobian matrix for the equilibrium S1 of the system (1) we find that the
α2 P 1
eigenvalues of the jacobian matrix V1 are λ11 = −(r − h1 ) < 0, λ12 = −(h3 + µ2 ) < 0 and λ13 = 2 − (h2 + µ1 ) = 1+a1 P1 +b1 P1
α2 P 1 α2 P1
(h2 + µ1 )[ (1+a − 1] = (h2 + µ1 )[R0 − 1]. It is clear that λ13 < 0 if < 1 i.e. R0 < 1 where
1 P1 +b1 P1 )(h2 +µ1 ) (1+a1 P1 +b1 P12 )(h2 +µ1 )
2

α2 P1
R0 = .
(1+a1 P1 +b1 P12 )(h2 +µ1 )
So S1 is asymptotically stable if and only if R0 < 1. Clearly if R0 > 1, then predator free steady state S1 is unstable which
indicates the proof of Lemma 1. 

Proof of Lemma 2. Now again computing the jacobian matrix for the equilibrium S2 of the system (1) we find three
α Z −b1 α1 P22 Z2
eigenvalues of the jacobian matrix V2 in which two of them are given from quadratic equation χ 2 + (
2rP2
+ (11+a2 +
1 P2 +b1 P2 )
K 2 2

α1 P2 (α2 Z2 −b1 α2 P22 Z2 ) α1 Z2 −b1 α1 P22 Z2


h1 − r )χ + = 0. So two roots (χ1 , χ2 ) are negative if + h1 > r and P22 < .
2rP2 1
(1+a1 P2 +b1 P22 )3 K
+ (1+a1 P2 +b1 P22 )2 b1
β2 Z 2 β2 Z2
χ3 = (1+c1 Z2 +d1 Z22 )
− (µ2 + h2 ) = (µ2 + h2 )[
(1+c1 Z2 +d1 Z22 )(µ2 +h2 )
− 1] = (µ2 + h2 )[R1 − 1]. Thus χ3 is negative if R1 < 1
β2 Z 2
where R1 = . Thus S2 is unstable if at least one of the conditions (5)–(7) holds. 
(1+c1 Z2 +d1 Z22 )(µ2 +h2 )

Proof of Lemma 3. If the non-spatial system (1) is locally asymptotically stable at S ∗ , then A1 > 0, A2 > 0, A3 > 0 and
A1 A2 − A3 > 0.
Here, A1 > 0 implies A∗1 > 0.
Now A∗1 A∗2 − A∗3 = A1 A2 − A3 + k6 H1 − k4 H2 + k2 H3 where

H1 = (dP + dF )(dP dZ + dZ dF + dF dP ) + d2Z (dP + dF ) > 0.


A. Chatterjee, S. Pal / Computers and Mathematics with Applications ( ) – 17

H2 = n11 dP (dZ + dF ) + n22 dZ (dP + dF ) + (dP + dZ + dF ){n22 dP + n11 dZ + (n11 + n22 )dF }
and
H3 = A2 (dP + dZ + dF ) + n211 (dZ + dF ) + n222 (dP + dF ) + n11 n22 (dP + dZ + dF ) + n23 n32 dP + n12 n21 dF .
Since A1 > 0, it follows that
H2 ≤ 2D2 (n11 + n22 ) + 3D.2D(n11 + n22 ) = 8D2 (n11 + n22 ) = −8D2 A < 0, where D = max{dP , dZ , dF }.
Also A2 > 0 implies H3 > 0 if (3A2 + + )d > −{3n11 n22 + n23 n32 + n12 n21 }D where d = min{dP , dZ , dF }. Therefore
n211 n222
k6 H1 − k4 H2 + k2 H3 > 0 and consequently A∗1 A∗2 > A∗3 (since A1 A2 > A3 ).
Hence, it follows that whenever (3A2 + n211 + n222 )d > − (3n11 + n23 n32 + n12 n21 ) D, the stability of the non-spatial
system (1) at S ∗ implies the stability of the system (32) at S ∗ . 
Proof of Lemma 4. Diffusive instability occurs if A∗1 A∗2 ≤ A∗3 . This happens if k4 H1 − k2 H2 + H3 ≤ 0. Since H1 > 0 and
k4 H1 +H3
H3 > 0, it follows that k4 H1 − k2 H2 + H3 ≤ 0 if H2 ≥ > 0. Let H (ξ ) = ξ 2 H1 − ηH2 + H3 , ξ = k2 . Since
k2
H = 2H1 = 2(dP + dZ )(dP dZ + dZ dF + dF dP ) + (dP + dF ) > 0. It follows that H has a minimum at ξ = k2c , where
′′
d2Z
H √
H22
k2c = 2H2 . Thus, Hmin = k4c H1 − k2c H2 + H3 = H3 − 4H ≤ 0 if H2 ≥ 2 H1 H3 . 
1 1

References

[1] G. Hu, W. Li, Hopf bifurcation analysis for a delayed predator–prey system with diffusion effects, Nonlinear Anal. RWA 11 (2010) 819–826.
[2] Y. Song, Y. Peng, J. Wei, Bifurcations for a predator–prey system with two delays, J. Math. Anal. Appl. 377 (1) (2008) 466–479.
[3] Y. Qu, J. Wei, Bifurcation analysis in a time-delay model for prey–predator growth with stage structure, Nonlinear Dynam. 49 (2007) 285–294.
[4] D. Xiao, S. Ruan, Multiple bifurcations in a delayed predator–prey system with nonmonotonic functional response, J. Differential Equations 176 (2)
(2001) 494–510.
[5] R.M. May, Limit cycles in predator–prey communities, Science 177 (4052) (1972) 900–902.
[6] A.B. Medvinsky, S.V. Petrovskii, I.A. Tikhonova, H. Malchow, B.L. Li, Spatiotemporal complexity of plankton and fish dynamics, SIAM Rev. 44 (3) (2002)
311–370.
[7] S. Bhattacharya, M. Martcheva, Oscillations in a size structured prey predator model, Math. Biosci. 228 (1) (2010) 31–44.
[8] S. Bhattacharya, M. Martcheva, X.Z. Li, A predator prey model with immune response in infected prey, J. Math. Anal. Appl. 411 (2014) 297–313.
[9] X. Wu, J. Li, Z. Wang, Existence of positive periodic solutions for a generalized prey–predator model with harvesting term, Comput. Math. Appl. 55
(2008) 1895–1905.
[10] S. Xu, Dynamics of a general prey–predator model with prey-stage structure and diffusive effects, Comput. Math. Appl. 68 (2014) 405–423.
[11] S. Chen, J. Shi, J. Wei, A note on Hopf bifurcations in a delayed diffusive Lotka–Volterra predator–prey system, Comput. Math. Appl. 62 (2011)
2240–2245.
[12] J. Bhattacharyya, S. Pal, The role of space in stage-structured cannibalism with harvesting of an adult predator, Comput. Math. Appl. 66 (2013) 339–355.
[13] R.M. May, Stability and Complexity in Model Ecosystems, Princeon University Press, 2001.
[14] V. Rai, R.K. Upadhyay, Evolving to the edge of chaos: Chance or necessity? Chaos Solutions Fractals 30 (5) (2006) 1074–1087.
[15] S. Rinaldi, S. Muratori, Y. Kuznetsov, Multiple attactors, catastrophes and chaos in seasonally perturbed predator–prey communitie, Bull. Math. Biol.
55 (1) (1993) 15–35.
[16] C.S. Holling, The functional response of invertebrate predators to prey density, Mem. Entomol. Soc. Can. 98 (1965) 5–86.
[17] B. Boon, H. Landelout, Kinetics of nitrite oxidation by nitrobacter winogradski, Biochem. J. 85 (1962) 440–447.
[18] V.H. Edwards, Influence of high substrate concentrations on microbial kinetics, Biotech. Bioengrg. 12 (1970) 679–712.
[19] R.D. Yang, A.E. Humphrey, Dynamics and steady state studies of phenol biodegeneration in pure and mixed cultures, Biotech. Bioengrg. 17 (1975)
1211–1235.
[20] J.F. Andrews, A mathematical model for the continuous culture of microorganisms utilizing inhibitory substrates, Biotech. Bioengrg. 10 (1968)
707–723.
[21] R.D. Holt, Optimal foraging and the form of the predator isocline, Amer. Nat. 122 (1983) 521–541.
[22] Y. Chen, Multiple periodic solutions of delayed predator–prey systems with type IV functional responses, Nonlinear Anal. RWA 5 (2004) 45–53.
[23] H. Baek, A food chain system with Holling type IV functional response and impulsive perturbations, Comput. Math. Appl. 60 (5) (2010) 1152–1163.
[24] J.B. Collings, The effects of the functional response on the bifurcation behavior of a mite predator–prey interaction model, J. Math. Biol. 36 (2) (1997)
149–168.
[25] S. Ruan, D. Xiao, Global analysis in a predator–prey system with nonmonotonic functional response, SIAM J. Appl. Math. 61 (4) (2001) 1445–1472.
[26] S. Zhang, F. Wang, L. Chen, A food chain model with impulsive perturbations and Holling IV functional response, Chaos Solitons Fractals 26 (2005)
855–866.
[27] Q. Chen, Z. Teng, Z. Hu, Bifurcation and control for a discrete-time prey–predator model with holling-iv functional response, Int. J. Appl. Math. Comput.
Sci. 23 (2) (2013) 247–261.
[28] R.L. Zhang, W.X. Wang, L.J. Qin, Optimal harvesting policy of discrete-time predator–prey dynamic system with holling type-iv functional response
and its simulation, Appl. Comput. Math. 4 (1) (2015) 20–29.
[29] D. Pal, G.S. Mahaptrab, G.P. Samanta, Optimal harvesting of prey–predator system with interval biological parameters: A bioeconomic model, Math.
Biosci. 241 (2) (2013) 181–187.
[30] D. Xiao, W. Li, M. Han, Dynamics in a ratio-dependent predator–prey model with predator harvesting, J. Math. Anal. Appl. 324 (1) (2006) 14–29.
[31] X. Chang, J. Weia, Hopf bifurcation and optimal control in a diffusive predator–prey system with time delay and prey harvesting, Nonlinear Anal.
Model. Control 17 (4) (2012) 379–409.
[32] E. Numfor, S. Bhattacharya, S. Lenhart, M. Martcheva, Optimal control in coupled within host and between host models, Math. Model. Nat. Phenom. 7
(2) (2014) 32–63.
[33] E. Numfor, S. Bhattacharya, S. Lenhart, M. Martcheva, Optimal Control in multi-group coupled within-host and between host models, in: Tenth
Mississippi State Conference on Differential Equations and Computational Simulations, Electron. J. Differential Equations (2015) in press.
[34] D. Jana, G.P. Samanta, Role of multiple delays in ratio-dependent prey–predator system with prey harvesting under stochastic environment, Neural
Parallel Sci. Comput. 22 (2014) 205–222.
[35] S. Ruan, On nonlinear dynamics of predator–prey models with discrete delay, Math. Model. Nat. Phenom. 4 (2) (2009) 140–188.
[36] Y. Li, M. Wang, Hopf bifurcation and global stability of a delayed predator–prey model with prey harvesting, Comput. Math. Appl. 69 (2015) 398–410.
[37] L. Yunfei, P. Yongzhena, G. Shujing, L. Changguo, Harvesting of a phytoplankton-zooplankton model, Nonlinear Anal. RWA 11 (2010) 3608–3619.
18 A. Chatterjee, S. Pal / Computers and Mathematics with Applications ( ) –

[38] P. Yongzhen, L. Yunfei, L. Changguo, Evolutionary consequences of harvesting for a two-zooplankton one-phytoplankton system, Appl. Math. Model.
36 (4) (2012) 1752–1765.
[39] Y. Lv, R. Yuan, Y. Pei, Stable coexistence mediated by specialist harvesting in a two zooplankton-phytoplankton system, Appl. Math. Model. 37 (2013)
9012–9030.
[40] S. Pal, A. Chatterjee, Plankton nutrient interaction model with harvesting under constant environment, in: Proceedings of International Symposium on
Mathematical and Computational Biology, World Scientific, Fields Institute for Research in Mathematical Sciences, Toronto, Canada, 2013, pp. 28–42.
[41] S.B. Goh, Global stability in many-species systems, Amer. Nat. 111 (977) (1977) 135–143.
[42] S. Toaha, J. Kusuma, Khaeruddin, M. Bahri, Stability analysis and optimal harvesting policy of prey–predator model with stage structure for predator,
Appl. Math. Sci. 8 (159) (2014) 7923–7934.
[43] R.K. Upadhyay, S.N. Raw, Complex dynamics of a three species food-chain model with Holling type IV functional response, Nonlinear Anal. Model.
Control 16 (3) (2011) 353–374.

You might also like