You are on page 1of 13

Applied Mathematical Modelling 37 (2013) 1400–1412

Contents lists available at SciVerse ScienceDirect

Applied Mathematical Modelling


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

Dynamics of a delayed predator–prey model with predator migration


Yuming Chen a,b,⇑, Fengqin Zhang a
a
Department of Applied Mathematics, Yuncheng University, Yuncheng, Shanxi 044000, PR China
b
Department of Mathematics, Wilfrid Laurier University, Waterloo, Ontario, Canada N2L 3C5

a r t i c l e i n f o a b s t r a c t

Article history: Based on the availability of prey and a simple predator–prey model, we propose a delayed
Received 26 September 2011 predator–prey model with predator migration to describe biological control. We first study
Received in revised form 15 February 2012 the existence and stability of equilibria. It turns out that backward bifurcation occurs with
Accepted 5 April 2012
the migration rate as bifurcation parameter. The stability of the trivial equilibrium and the
Available online 12 April 2012
boundary equilibrium is delay-independent. However, the stability of the positive equilib-
rium may be delay-dependent. Moreover, delay can switch the stability of the positive
Keywords:
equilibrium. When the positive equilibrium loses stability, Hopf bifurcation can occur.
Delay
Prey–predator model
The direction and stability of Hopf bifurcation is derived by applying the center manifold
Migration method and the normal form theory. The main theoretical results are illustrated with
Equilibrium numerical simulations.
Stability Ó 2012 Elsevier Inc. All rights reserved.
Hopf bifurcation

1. Introduction

A fundamental type of interactions which effects population dynamics of all species is predation. The predator–prey rela-
tionship is ubiquitous in the Nature and hence it has been one of the dominant themes in mathematical biology. Since the
pioneering work of Lotka and Volterra, various predator–prey models have been built by incorporating additional biological
processes into the classical Lotka–Volterra predator–prey equations. Here, we just name a few of the recent advances [1–13].
To take the spatial heterogeneity into account, models are either proposed in a multi-patch environment or described by
reaction–diffusion equations (see, for example, [8–10,14–20] and the references therein). Though diffusions and/or migra-
tions are considered, these models may be inapplicable for biological control.
Biological control is an approach in controlling insects and minimizing losses resulting from insect infestations. It reduces
pest population by other living organisms, often called natural enemies or beneficial species. Virtually all pests have some
natural enemies, and the key to successful pest control is to identify the pest and its natural enemy. For example, grasshop-
pers are a major pest of both cultivated crops and rangeland grasses in the world’s semi-arid regions. There are many strat-
egies to control the grasshopper population. The most important one is to use the grasshopper’s natural enemies such as
flocks of chickens and ducks. In this control, human can play a crucial role in releasing the natural enemies. For example,
the release of natural enemies may depend on the availability of grasshoppers. To the best of our knowledge, so far only
Cheng [21] proposed a discrete model for the rodent-predator system to describe such a phenomenon.
Based on the ideas of Cheng [21] and a basic predator–prey model, we propose the following predator–prey model for
biological control,

⇑ Corresponding author at: Department of Mathematics, Wilfird Laurier University, Waterloo, Ontario, Canada N2L 3C5. Tel.: +1 519 884 0710x2309; fax:
+1 519 884 9738.
E-mail addresses: ychen@wlu.ca (Y. Chen), zhafq@263.net (F. Zhang).

0307-904X/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.apm.2012.04.012
Y. Chen, F. Zhang / Applied Mathematical Modelling 37 (2013) 1400–1412 1401


_
xðtÞ ¼ xðtÞ½r  ayðtÞ;
ð1:1Þ
_yðtÞ ¼ yðtÞ½d þ bxðt  sÞ  cyðtÞ þ m½xðtÞ  pyðtÞ;

where x and y are the biomasses of the prey and predator, respectively. In the absence of predators, the prey species follows
_
the exponential equations, xðtÞ ¼ rxðtÞ, where r is the intrinsic growth rate; ayðtÞ is the hunting term in the presence of pre-
dators; the positive feedback bxðt  sÞ has a positive delay s, which is the time due to converting prey biomass into predator
biomass; p is the consumption rate of the predator on prey per predator per unit time; m is the migration rate of predators; c
is the self-limitation constant of the predator; d is the death rate of the predator (see, for instance, Faria [22] for more detail
about the biological interpretation of the parameters). All parameters are positive except d. The parameter d may be negative
if the predator has alternative food source. The main feature of model (1.1) is that the migration through human control with
stock/harvest is dependent on the availability of prey, xðtÞ  pyðtÞ.
cxðt Þ cyðt Þ
To reduce the number of parameters, we scale the variables by letting uðtÞ ¼ prr and v ðtÞ ¼ r r . It follows from (1.1) that
8  
< uðtÞ
_ ¼ uðtÞ 1  ac v ðtÞ ;
h i
: v_ ðtÞ ¼ v ðtÞ  dr þ bp
c
uðt  r sÞ  v ðtÞ þ pm
r
½u  v :

bp pm
For the simplicity of notations, we still denote ac, dr ; c
; rs and r
respectively by a, d, b, s and m. Then we can rewrite the
above system as

_
uðtÞ ¼ uðtÞ½1  av ðtÞ;
ð1:2Þ
v_ ðtÞ ¼ v ðtÞ½d þ buðt  sÞ  v ðtÞ þ m½uðtÞ  v ðtÞ:
Let C s ¼ Cð½s; 0; R2 Þ be the Banach space of all continuous functions from ½s; 0 into R2 equipped with the sup-norm.
From the point of view of mathematical biology, we only consider (1.2) with initial conditions in C þ s ¼
fu ¼ ðu1 ; u2 Þ 2 C s ju1 ðhÞ P 0 and u2 ðhÞ P 0 for h 2 ½s; 0. For u ¼ ðu1 ; u2 Þ 2 C þ s and t 0 2 R, it is easy to see that (1.2) has
a unique solution ðu; v Þ : ½t0  s; 1Þ ! R2þ with the initial value u, that is, ðuðtÞ; v ðtÞÞ ¼ ðu1 ðt  t 0 Þ; u2 ðt  t0 ÞÞ for
t 2 ½t 0  s; t0  and ðu; v Þ is continuous on ½t0  s; 1Þ and satisfies (1.2) on ðt 0 ; 1Þ. This tells us that C þ
s is invariant and hence
our model is biologically meaningful. Moreover, if u1 ðt 0 Þ > 0 and u2 ðt0 Þ > 0 then for the corresponding solution we have
uðtÞ > 0 and v ðtÞ > 0 for t > t0 . For theory on functional differential equations, we refer the readers to the books by Hale
and Verduyn Lunel [23] and Diekmann et al. [24].
In this paper, we concentrate on the effects of the migration rate and the time delay on the dynamics of (1.2). It turns out
that (1.2) can have very complicated and interesting dynamics. Roughly speaking, there may be backward bifurcation, Hopf
bifurcation can occur and delay can switch the stability of the positive equilibrium.
The remaining of this paper is organized as follows. In Section 2, we study the existence of equilibria and their linear sta-
bility. Then, in Section 3, we investigate the direction and stability of Hopf bifurcation by applying the center manifold meth-
od and the normal form theory. Most of the theoretical results are supported with numerical simulations in Section 4. We
conclude this paper with a brief discussion in Section 5.

2. Existence of equilibria and their linear stability

As mentioned earlier, we are only interested in nonnegative (componentwise) equilibria of (1.2). Let ðu ; v  Þ be such an
equilibrium. Then

u ð1  av  Þ ¼ 0;
v  ðd þ bu  v  Þ þ mðu  v  Þ ¼ 0:
Clearly, O ¼ ð0; 0Þ is always an equilibrium of (1.2). Moreover, if m þ d < 0 then (1.2) has a nontrivial boundary equilibrium
E0 ¼ ð0; ðm þ dÞÞ; if 1 þ aðm þ dÞ > 0 then (1.2) has a positive equilibrium E ¼ ð1þaðmþdÞ ; 1Þ. In summary, we have obtained
aðbþmaÞ a
the following result on the existence of equilibria of (1.2).

Proposition 2.1

(i) If m P d then (1.2) has two equilibria O and E .


(ii) If d  1a < m < d then (1.2) has three equilibria O; E0 and E .
(iii) If m 6 d  1a then (1.2) has two equilibria O and E0 .

By Proposition 2.1, if the predator has alternative food source such that d < 0 then system (1.2) has a backward bifurca-
tion from O at m ¼ d; if the prey is the dominant food source for the predator such that d > 0 then system (1.2) only has
two equilibria O and E . The corresponding bifurcation diagram is shown in Fig. 1, where the stability is for the case where
s ¼ 0 and is from Theorem 2.1. Biologically, ðd þ mÞ is the net growth rate of the predator in the absence of prey. If
ðd þ mÞ > 0 or m < d then the predator can sustain because of sufficient alternative food and small migration rate, that
1402 Y. Chen, F. Zhang / Applied Mathematical Modelling 37 (2013) 1400–1412

Fig. 1. The bifurcation diagram of (1.2) shows a backward bifurcation as m varies from 0 to d. Here we assume that d < 1=a for the purpose of
illustration.

is, E0 exists. When there is alternative food for the predator, d is the intrinsic growth rate of the predator. Note that
aðd þ mÞ is the hunting rate of the predator. If aðd þ mÞ P 1 (the intrinsic growth rate of the prey) or m 6 d  1a then
the prey and predator cannot coexist, that is E does not exist. This explains Proposition 2.1.
Now we study the linear stability of the equilibria one by one.
First, we linearize (1.2) about O to obtain the linear system,

_
uðtÞ ¼ uðtÞ;
v_ ðtÞ ¼ muðtÞ  ðm þ dÞv ðtÞ:
The corresponding eigenvalues are 1 and ðm þ dÞ. Therefore, O is always unstable.
Next, the linearized system of (1.2) about E0 is

_
uðtÞ ¼ ½1 þ aðm þ dÞuðtÞ;
v_ ðtÞ ¼ muðtÞ  bðm þ dÞuðt  sÞ þ ðm þ dÞv ðtÞ:
The corresponding eigenvalues are 1 þ aðm þ dÞ and m þ d. Recall that E0 exists only when m þ d < 0. It follows that E0 is
stable if 1 þ aðm þ dÞ < 0 and unstable if 1 þ aðm þ dÞ > 0, that is, E0 is stable only when E does not exist.
Note that the linear stability of O and E0 is delay-independent. However, as we will see, this may not be true for the linear
stability of E .
The linearized system of (1.2) about E is
8
< uðtÞ
_ ¼  1þðmþdÞa v ðtÞ;
bþma
: v_ ðtÞ ¼ muðtÞ þ b uðt  sÞ  m2 a2 þ2maþbþdma2 v ðtÞ:
a aðbþmaÞ

The corresponding characteristic equations is


 
 1þðmþdÞa  2  
 k bþma  m2 a2 þ 2ma þ b þ dma 1 þ ðm þ dÞa b ks
0 ¼  
2  ¼ k2
þ k þ m þ e ;
 m  ba eks
2 2
k þ m a þ2maþbþdma  aðb þ maÞ b þ ma a
aðbþmaÞ

or

k2 þ Pk þ Q þ Reks ¼ 0; ð2:1Þ
where
2 
m2 a2 þ 2ma þ b þ dma ma½1 þ ðm þ dÞa þ b þ ma
P, ¼ > 0; ð2:2Þ
aðb þ maÞ aðb þ maÞ
m½1 þ ðm þ dÞa
Q, > 0; ð2:3Þ
b þ ma
b½1 þ ðm þ dÞa
R, > 0: ð2:4Þ
aðb þ maÞ
To study the stability of E , we distinguish two cases: s ¼ 0 and s > 0.
When s ¼ 0, the eigenvalues k1 and k2 satisfy
k1 þ k2 ¼ P < 0 and k1 k2 ¼ Q þ R > 0:
It follows that Reðk1 Þ < 0 and Reðk2 Þ < 0. Therefore, E is locally asymptotically stable when s ¼ 0.
Y. Chen, F. Zhang / Applied Mathematical Modelling 37 (2013) 1400–1412 1403

When s – 0, the characteristic equation (2.1) can be rewritten as


ðz þ sPz þ s2 QÞez þ s2 R ¼ 0;
2
ð2:5Þ
where z ¼ ks. Denote by ak (k P 1) the sole root of the equation cot a ¼ ða  s Q Þ=ðsPaÞ which lies on the interval 2 2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ððk  1Þp; kpÞ. Let w ¼ 1 if P2 P 2Q otherwise let w be the odd k for which ak lies closest to s Q  P 2 =2. Suppose that
2
s < p2RP. Then s RsPasinwaw 6 2psPR < 1. By Theorem 13.9 of Bellman and Cooke [25], we know that all roots of (2.5) lie to the left of
the imaginary axis and hence when the delay is small the equilibrium E is locally asymptotically stable.
In the following, we discuss whether the equilibrium E will lose stability when s is large. If the equilibrium E loses sta-
bility, we will verify whether Hopf bifurcation will occur when we choose s as a parameter.
Obviously, k ¼ 0 is not a root of (2.1). If E loses stability, then there must exist s0 such that (2.1) has pure imaginary roots.
Note that if k is a root of (2.1) then so is 
k. Without loss of generality, let k ¼ iw0 with w0 > 0 be a root of (2.1). Substituting it
into (2.1) and separating the real and imaginary parts, we obtain

x20  Q ¼ R cosðsw0 Þ ð2:6Þ


and
Pw0 ¼ R sinðsw0 Þ: ð2:7Þ
Squaring both sides of (2.6) and (2.7) and adding them up, we get

w40 þ ðP 2  2QÞw20 þ ðQ 2  R2 Þ ¼ 0: ð2:8Þ


4 2 2
Let D ¼ P  4P Q þ 4R . Now we distinguish several cases.
Case 1: Q < R. Then (2.8) has a positive solution
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffi
2Q  P 2 þ D
w1 ¼ :
2
Case 2: Q P R and P2 P 2Q . Then (2.8) has no positive solution.
Case 3: Q P R; P2 < 2Q and D < 0. Then (2.8) has no positive solution.
Case 4: Q P R; P2 < 2Q and D ¼ 0. Then (2.8) has a unique positive solution
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2Q  P 2
w2 ¼ :
2
Case 5: Q ¼ R and P 2 < 2Q . Of course, in this case, D > 0 and (2.8) has a unique positive solution
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
w3 ¼ 2Q  P 2 :

Case 6: Q > R; P2 < 2Q and D > 0. Then (2.8) has two positive solutions
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffi
ð2Q  P2 Þ  D
w4 ¼ :
2
The above discussion combined with (2.6) and (2.7) tells us that (2.1) has a root iwj ; j 2 f1; 2; 3; 4g only when

w2 Q
s ¼ skj , x1 2kp þ arccos j R for some k 2 N ¼ f0; 1; 2; . . .g. Note that skj < sjkþ1 (j=1, 2, 3) and sk4 < sk4þ for k 2 N, but
j

kþ1
sk4þ < s4 may not be true for all k 2 N.
Now, we determine whether the roots of (2.1) cross the imaginary axis transversally. Let s0 > 0 such that iw0 is a root of
(2.1). Let wðsÞ be the solution curve of (2.1) passing through iw0 . Then differentiate (2.1) with respect to s to obtain

dkðsÞ dkðsÞ dkðsÞ
2kðsÞ þP  Reks s þ k ¼ 0:
ds ds ds
Evaluating at s ¼ s0 , we get
dkðs0 Þ iRw0 eiw0 s0
¼ :
ds 2iw0 þ P  Rs0 eiw0 s0
This combined with w20 þ Piw0 þ Q þ Reiw0 s0 ¼ 0 or Reiw0 s0 ¼ w20  Q  Piw0 produces

dkðs0 Þ w20 ðP2 þ 2w20  2Q Þ
Re ¼ :
ds ðP  w20 s0 þ Q s0 Þ2 þ w20 ðPs0 þ 2Þ2
1404 Y. Chen, F. Zhang / Applied Mathematical Modelling 37 (2013) 1400–1412

In particular,
 pffiffiffiffi
dkðsk1 Þ Dw21
Re ¼ pffiffiffi 2 2 > 0; ð2:9Þ
ds 2
P þ P 2 D sk1 þ w21 Psk1 þ 2

dkðsk2 Þ
Re ¼ 0;
ds

dkðsk3 Þ ð2Q  P2 Þ2
Re ¼ 2 2 > 0; ð2:10Þ
ds
P  ðQ  P2 Þsk3 þ ð2Q  P2 Þ Psk3 þ 2
 pffiffiffiffi
dkðsk4 Þ  Dw24
Re ¼ pffiffiffi 2 : ð2:11Þ
ds 2 2
P þ P 2 D sk4 þ w24 Psk4 þ 2

From the above discussion, we can deduce the main result of this section.

Theorem 2.1

(i) The trivial equilibrium O ¼ ð0; 0Þ is always unstable.


(ii) The boundary equilibrium E0 ¼ ð0; ðm þ dÞÞ (which exists when m þ d < 0) is stable if 1 þ aðm þ dÞ < 0 and unstable if
1 þ aðm þ dÞ > 0.
(iii) Let D ¼ P4  4P2 Q þ 4R2 , where P; Q , and R be defined by (2.2)–(2.4), respectively. Then the following
statements
are true.
(a) If either (Q P R and P 2 P 2Q ) or (Q P R; P2 < 2Q and D < 0), then the equilibrium E ¼ 1þaðmþdÞ ;
aðbþmaÞ a
1
is stable.
(b) If Q < R, then E is stable for s 2 ½0; s01 Þ and is unstable if s > s01 . Moreover, sk1 ; k 2 N, are Hopf bifurcation values for
(1.2).
(c) If Q P R; P2 < 2Q and D ¼ 0, then E is stable for s 2 ½0; s02 Þ.
(d) If Q ¼ R and P2 < 2Q , then E is stable for s 2 ½0; s03 Þ and is unstable if s > s03 . Moreover, sk3 ; k 2 N, are Hopf bifurcation
values for (1.2).
(e) If Q > R; P2 < 2Q and D > 0, then E is stable for s 2 ½0; s04 Þ. Moreover, sk4 ; k 2 N, are Hopf bifurcation values for
(1.2).

Recall that aðm þ dÞ is the hunting effort of the predator. When there is alternative food for the predator and its intrinsic
growth rate is larger than the migration rate, the prey will die out if the predator’s hunting effort is larger than the intrinsic
growth rate of the prey, which is 1; otherwise the prey will persist. When the predator and the prey coexist, the coexistence
equilibrium is stable for small delay. When delay becomes large, stability may lose and Hopf bifurcation can occur. The exact
dynamics greatly relies on the parameters. For example, Q P R implies that b 6 ma, or the conversion rate is less than the
hunting effort of the migrating predators. The other inequalities involving P2 and 2Q , and D are very difficult to explain bio-
logically because of their complex expressions.
It is also worth to make some observations from Theorem 2.1.

0.45

0.4

0.35

0.3
v

0.25

0.2

0.15

0.1
0 2 4 6 8 10 12 14 16
tau
51 
Fig. 2. The bifurcation diagram of (1.2) at E ¼ ;1
164 4
with a ¼ 4; d ¼ 0:5; b ¼ 2, and m ¼ 21
40
. One can see Hopf bifurcation and stability switch by delay.
Y. Chen, F. Zhang / Applied Mathematical Modelling 37 (2013) 1400–1412 1405

Remark 2.1

(i) The stability of E may be delay-dependent or delay-independent according to the parameter values.
(ii) Since the eigenvalues do not cross the imaginary axis transversally in the case where Q P R; P2 < 2Q and D > 0, we
need further calculations to determine the stability of E when s > s02 . But we will not go further for this critical case.
(iii) With the help of (2.11), we can see that if there exists k0 2 N such that sk4þ
0 k0 þ1
< s4 , then delay can switch the stability
in the case where Q P R; P2 < 2Q and D > 0. This can be seen from the bifurcation diagram (see Fig. 2). Here
a ¼ 4; d ¼ 0:5; b ¼ 2, and m ¼ 21
40
, which are the parameters used in Section 4.

3. Direction and stability of the Hopf bifurcation

In the previous section, we have obtained some conditions which guarantee the occurrence of Hopf bifurcation. In this
section, we shall study the direction and stability of the bifurcating periodic solutions by applying the normal form theory
and center manifold theorem introduced by Hassard et al. [26].
Let xðtÞ ¼ uðstÞ þ 1þmaþda
aðbþmaÞ
and yðtÞ ¼ v ðstÞ þ 1a. Then (1.2) can be rewritten as
  
_
xðtÞ xðtÞ xðt  1Þ
¼ s A1 þ sB þ Fðxt ; yt ; sÞ; ð3:1Þ
_
yðtÞ yðtÞ yðt  1Þ
where
!
0  1þmaþda
bþma
A1 ¼ ;
m d  m  2a  bð1þmaþdaÞ
aðbþmaÞ
!
0 0
B¼ b ;
a
0

saxðtÞyðtÞ
Fðxt ; yt ; sÞ ¼ :
sbxðt  1ÞyðtÞ  sy2 ðtÞ
^ be the critical value of s at which system (1.2) undergoes a Hopf bifurcation at E and the
Without loss of generality, let s
corresponding eigenvalues are iw. ^ Let s ¼ s^ þ h. Then h ¼ 0 is the Hopf bifurcation value of (1.2).
Choose the phase space C ¼ Cð½1; 0; R2 Þ. Define LðhÞ : C ! R2 by
LðhÞ/ ¼ ðs
^ þ hÞA1 /ð0Þ þ ðs
^ þ hÞB/ð1Þ for / 2 C:
By the Riesz representation theorem, there exists a matrix whose components are bounded variation functions
2
gðh; hÞ : ½1; 0 ! R2 in h such that
Z 0
LðhÞ/ ¼ dgðh; hÞ/ðhÞ for / 2 C:
1

In fact, we can choose gðh; hÞ ¼ ðs


^ þ hÞA1 dðhÞ  ðs
^ þ hÞBdðh þ 1Þ, where

1; h ¼ 0;
dðhÞ ¼
0; h – 0:
1
For / 2 C ð½1; 0; R2 Þ, define
(
_
/ðhÞ; h 2 ½1; 0Þ;
AðhÞ/ ¼ R0
1
dgðt; hÞ/ðtÞ; h ¼ 0
and

0; h 2 ½1; 0Þ;
RðhÞ/ ¼
Fð/; s
^ þ hÞ; h ¼ 0:

Then (3.1) can be rewritten as


u_ t ¼ AðhÞut þ RðhÞut : ð3:2Þ
1 2 
For w 2 C ð½0; 1; ðC Þ Þ, define
(
_
wðsÞ; s 2 ð0; 1;

A wðsÞ ¼ R0
1
wðtÞdgðt; 0Þ ¼ swð0ÞA1 þ swð1ÞB; s ¼ 0:
^ ^

For / 2 Cð½1; 0; C2 Þ and w 2 Cð½0; 1; ðC2 Þ Þ, define


1406 Y. Chen, F. Zhang / Applied Mathematical Modelling 37 (2013) 1400–1412

Z 0 Z h

hw; /i ¼ wð0Þ/ð0Þ    hÞdgðh; 0Þ/ðnÞdn:
wðn
1 0

Then A and Að0Þ are adjoint operators. Let qðhÞ and q ðsÞ be the eigenvectors of A and A corresponding to iw
^s ^s
^ and iw ^,
respectively. By direct calculation, we obtain
 T
ið1 þ ma þ daÞ
qðhÞ ¼ ; 1 eiw^ s^h
^ þ maÞ
wðb
and
^s^
!
iw
iðma þ be Þ

q ðsÞ ¼ D ; 1 eiw^ s^s ;
^
wa
w^ 2 aðbþmaÞ
where D ¼ ^s . Note that hq ; qi ¼ 1 and hq ; q
i ¼ 0.
þbeiw^s bÞ
^
^ 2 aðbþmaÞþð1þmaþdaÞðmaþbeiw^
w

Following the algorithm in Hassard et al. [26], we first compute the coordinates to describe the center manifold C0 at
h ¼ 0. Let ut ¼ ðxt ; yt ÞT be the solution of (3.1) when s ¼ s
^, that is, when h ¼ 0. Define
zðtÞ ¼ hq ; ut i; Wðt; hÞ ¼ ut ðhÞ  2RefzðtÞqðhÞg:
On the center manifold C0 we have
Wðt; hÞ ¼ WðzðtÞ; zðtÞ; hÞ;
where
z2 z2 z3
Wðz; z; hÞ ¼ W 20 ðhÞ þ W 11 ðhÞzz þ W 02 ðhÞ þ W 30 ðhÞ þ    ;
2 2 6
z and z are the local coordinates for the center manifold C0 in the direction of q and q
 . Note that W is real if ut is real. We
consider only real solutions.
For the solution ut 2 C0 of (3.1), since h ¼ 0,
^s
z_ ðtÞ ¼ iw ^zðtÞ þ hq ðsÞ; FðWðt; Þ þ 2RefzðtÞqðÞgÞi
^s
¼ iw  ð0ÞFðWðz; z; 0Þ þ 2RefzðtÞqð0ÞgÞ , iw
^zðtÞ þ q ^s  ð0ÞF 0 ðz; zÞ:
^z þ q

We rewrite the above as


^s
z_ ðtÞ ¼ iw ^zðtÞ þ gðzðtÞ; zðtÞÞ; ð3:3Þ
where
z2 z2 z2 z
 ð0ÞFðWðz; z; 0Þ þ 2RefzðtÞqð0ÞgÞ ¼ g 20
gðz; zÞ ¼ q þ g 11 zz þ g 02 þ g 21 þ  ð3:4Þ
2 2 2
By (3.2) and (3.3), we have

 ð0ÞF 0 qðhÞg
AW  2Refq if  1 6 h < 0
_ ¼ u_ t  z_ q  z_ q
W ¼ , AW þ Hðz; z; hÞ;
 ð0ÞF 0 qð0Þg þ F 0
AW  2Refq if h ¼ 0
where
z2 z2
Hðz; z; hÞ ¼ H20 ðhÞ þ H11 ðhÞzz þ H02 ðhÞ þ    ð3:5Þ
2 2
Expanding the above series and comparing the coefficients, we obtain
^s
ðA  2iw ^ÞW 20 ðhÞ ¼ H20 ðhÞ; AW 11 ðhÞ ¼ H11 ðhÞ; ^s
ðA þ 2iw ^ÞW 02 ðhÞ ¼ H02 ðhÞ;    ð3:6Þ
Notice that
^s
iw ^
!
iðma þ be Þ
q ð0Þ ¼ D ;1 ;
^
wa
ið1 þ ma þ daÞ ið1 þ ma þ daÞ z2 z2
xðtÞ ¼ z z þ W ð1Þ
20 ð0Þ
ð1Þ ð1Þ
þ W 11 ð0Þzz þ W 02 ð0Þ þ    ;
^ þ maÞ
wðb ^ þ maÞ
wðb 2 2
2  2
ð2Þ z ð2Þ ð2Þ z
yðtÞ ¼ z þ z þ W 20 ð0Þ þ W 11 ð0Þzz þ W 02 ð0Þ þ    ;
2 2
ið1 þ ma þ daÞ iw^ s^ ið1 þ ma þ daÞ iw^ s^ ð1Þ z2 ð1Þ ð1Þ z2
xðt  1Þ ¼ e z e z þ W 20 ð1Þ þ W 11 ð1Þzz þ W 02 ð1Þ þ   
^ þ maÞ
wðb ^ þ maÞ
wðb 2 2
Y. Chen, F. Zhang / Applied Mathematical Modelling 37 (2013) 1400–1412 1407

and

s
^axðtÞyðtÞ
F0 ¼ :
s^bxðt  1ÞyðtÞ  s^y2 ðtÞ
Direct substitution and comparing the coefficients with (3.4) gives us
" ^s^
#
iw
ð1 þ ma þ daÞðma þ be Þ ibð1 þ ma þ daÞeiw^ s^
g 20 ¼ 2s
^D  þ 1 ;
w^ 2 ðb þ maÞ ^ þ maÞ
wðb

 
^s
bð1 þ ma þ daÞ sinðw ^Þ
g 11 ¼ 2s
^D 1 ;
^ þ maÞ
wðb

" ^s^
#
iw
ð1 þ ma þ daÞðma þ be Þ ibð1 þ ma þ daÞeiw^ s^
g 02 ¼ 2s
^D   1 ;
w^ 2 ðb þ maÞ ^ þ maÞ
wðb

" ^s
iw ^
! iw^s^
!
ð2Þ ð1Þ
ð1 þ ma þ daÞðma þ be Þ W 20 ð0Þ ð2Þ iðma þ be Þ W 20 ð1Þ
g 21 ¼ 2s
^D  W 11 ð0Þ þ ð0Þ þ W 11 ð0Þ
w^ 2 ðb þ maÞ 2 ^
w 2
! ! #
ð2Þ ð1Þ
ibð1 þ ma þ daÞ iw^ s^ ð2Þ ^s
iw ^ W 20 ð0Þ W 20 ð1Þ ð1Þ ð2Þ ð2Þ
þ e W 11 ð0Þ  e þb þ W 11 ð1Þ  2ð2W 11 ð0Þ þ W 20 ð0ÞÞ :
^ þ maÞ
wðb 2 2

We still need to compute W 20 ðhÞ and W 11 ðhÞ. For h 2 ½1; 0, we have
Hðz; z; hÞ ¼ 2Refz ð0ÞF 0 qðhÞg ¼ gqðhÞ  gq
ðhÞ
 
z2 z2 z2 z2
¼  g 20 þ g 11 zz þ g 02 þ    qðhÞ  g20 þ g11 zz þ g02 þ    qðhÞ:
2 2 2 2
Comparing the coefficients with (3.5) yields
ðhÞ
H20 ðhÞ ¼ g 20 qðhÞ  g02 q
and
ðhÞ:
H11 ðhÞ ¼ g 11 qðhÞ  g11 q
It follows from (3.6) that
_ 20 ðhÞ ¼ 2iw
W ^s^W 20 ðhÞ þ g 20 qð0Þeiw^ s^h þ g02 q
ð0Þeiw^ s^h :

Solving for W 20 ðhÞ, we get


ð0Þ iw^ s^h
ig 20 qð0Þ iw^ s^h ig02 q
W 20 ðhÞ ¼ e þ e þ E1 e2iw^ s^h ; ð3:7Þ
^s
w ^ 3w^s ^
similarly, we get
ð0Þ iw^ s^h
ig 11 qð0Þ iw^ s^h ig11 q
W 11 ðhÞ ¼ e þ e þ E2 ; ð3:8Þ
w^s^ ^s
w ^
where E1 and E2 are both two-dimensional vectors and can be determined by setting h ¼ 0 in H. In fact, since
 ð0ÞF 0 qð0Þg þ F 0 ;
Hðz; z; 0Þ ¼ 2Refq
we have
0 1
 iað1þmaþdaÞ
^
wðbþmaÞ
^@
ð0Þ þ 2s
H20 ð0Þ ¼ g 20 qð0Þ  g20 q ^s
A
ibð1þmaþdaÞeiw^
^
wðbþmaÞ
1

and
!
0
ð0Þ þ s
H11 ð0Þ ¼ g 11 qð0Þ  g11 q ^ ibð1þmaþdaÞ :
^
wðbþmaÞ
ðeiw^ s^  eiw^ s^ Þ  2

It follows from (3.6) and the definition of A that


s^A1 W 20 ð0Þ þ s^BW 20 ð1Þ ¼ 2iw
^s^W 20 ð0Þ  H20 ð0Þ
1408 Y. Chen, F. Zhang / Applied Mathematical Modelling 37 (2013) 1400–1412

and
s^A1 W 11 ð0Þ þ s^BW 11 ð1Þ ¼ H11 ð0Þ:
Substituting (3.7) and (3.8) into the above two equations respectively gives
1 1  i i i i

A1 þ Be2iws  2iwI
^^
E1 ¼  ^ H20 ð0Þ þ 4Reðg 20 qð0ÞÞ  g 20 A1 qð0Þ þ ð0Þ þ g 20 eiw^ s^ Bqð0Þ þ
g02 A1 q g02 eiw^ s^ Bq
ð0Þ
s^ ^
w 3w^ ^
w 3w^

and 
1 2 2
E2 ¼  ðA1 þ BÞ1 H11 ð0Þ þ A1 Imðg 11 ð0Þqð0ÞÞ þ BImðg 11 qð0Þeiw^ s^ Þ ;
s^ ^
w ^
w
where I is the 2  2 identity matrix. Consequently, g 21 can be expressed in terms of the parameters and delay s
^. Then we can
compute the following values,
!
i jg j2 g 21
C 1 ð0Þ ¼ g 20 g 11  2jg 11 j2  02 þ ;
^s
2w ^ 3 2
ReðC 1 ð0ÞÞ
l2 ¼  ;
Rek0 ðs^Þ
b2 ¼ 2ReðC 1 ð0ÞÞ:
Using the results of Hassard et al. [26], we obtain the following result.

Theorem 3.1

(i) If l2 > 0 (< 0) then the Hopf bifurcation is supercritical (subcritical).


(ii) If b2 < 0 (> 0) then the bifurcating periodic solutions are stable (unstable).

4. Numerical simulations

From the previous two sections, we know that the stability of the trivial equilibrium O and the boundary equilibrium E0 is
relatively simple. However, the stability of the positive equilibrium E is complex. Roughly speaking, delay can not only
switch the stability but also lead to Hopf bifurcation. As the formulae in Section 3 are quite complicated, in this section,
we use numerical simulations to support the theoretical results.
Recall that we focus on the effects of the migration rate and delay. Let’s consider

_
uðtÞ ¼ uðtÞ½1  4v ðtÞ;
ð4:1Þ
_v ðtÞ ¼ v ðtÞ½0:5 þ 2uðt  sÞ  v ðtÞ þ mðuðtÞ  v ðtÞÞ:

Then (4.1) has two equilibria O ¼ ð0; 0Þ and E ¼ 8ð2mþ1Þ
4mþ3
; 14 . Consequently,
8m2 þ 8m þ 1
P¼ ;
4ð2m þ 1Þ
mð4m þ 3Þ
Q¼ ;
2ð2m þ 1Þ
4m þ 3
R¼ :
4ð2m þ 1Þ
It follows that Q P R when m P 0:5; P 2 P 2Q when m 2 ½0; m1 Þ or m P m3 ; D < 0 when m 2 ðm2 ; m4 Þ, where m1 0:029130,
m2 0:538051, m3 1:277433, and m4 3:241398. By Theorem 2.1, the following results hold

(i) If m P m2 then E is stable.


(ii) If m < 0:5 then E is stable for s 2 ½0; s01 Þ and is unstable if s > s01 .
(iii) If m ¼ 0:5 then E is stable for s 2 ½0; s03 Þ and is unstable for s > s03 .
(iv) If m 2 ð0:5; m2 Þ then E is stable for s 2 ð0; s04 Þ and delay may switch the stability.

Moreover, these skj ’s are bifurcation values. It is interesting to observe that larger migration rate has the better stabilizing
effect. In the following, we do numerical
7 1 simulations for the above four cases.
Firstly, let m ¼ 1. Then E ¼ 24 ; 4 and it is stable. This is supported by Fig. 3, where in (a) the delay is 1 and in (b) the
delay is 2. 
Secondly, let m ¼ 14. Then E ¼ 13 ; 14 , which is stable if s < s01 0:996003 and is unstable if s > s01 . Moreover,
s1 ¼ 7:192343k þ s1 are bifurcation critical values. We first take s ¼ 0:5. Fig. 4 shows that E is stable. However, when we
k 0

increase s to 1:2, Fig. 5 tells us that a stable periodic solution bifurcates from E .
Y. Chen, F. Zhang / Applied Mathematical Modelling 37 (2013) 1400–1412 1409

0.9 0.7

0.8
0.6

0.7
0.5
0.6

0.4
0.5

v
v

0.4 0.3

0.3
0.2
0.2
0.1
0.1

0 0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
u u
(a) tau=1 (b) tau=2

7 
Fig. 3. When m ¼ 1 the positive equilibrium E ¼ ;1
24 4
of (4.1) is stable.

0.9

0.8

0.7

0.6

0.5
v

0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
u
1 
Fig. 4. When m ¼ 14 and s ¼ 0:5 the equilibrium E ¼ ;1
3 4
of (4.1) is stable.

1.4

1.2

0.8
v

0.6

0.4

0.2

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
u

Fig. 5. When m ¼ 14 and s ¼ 1:2, (4.1) has a periodic solution.


1410 Y. Chen, F. Zhang / Applied Mathematical Modelling 37 (2013) 1400–1412

0.9

0.8

0.7

0.6

0.5

v
0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
u
5 
Fig. 6. When m ¼ 12 and s ¼ 2 the equilibrium E ¼ ;1
16 4
of (4.1) is stable.

0.9

0.8

0.7

0.6

0.5
v

0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
u

Fig. 7. When m ¼ 12 and s ¼ 4, (4.1) has a periodic solution.

0.45

0.4

0.35

0.3
v

0.25

0.2

0.15

0.1
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
u
51 
Fig. 8. When m ¼ 21
40
and s ¼ 2, the equilibrium E ¼ ;1
164 4
(4.1) is stable.
Y. Chen, F. Zhang / Applied Mathematical Modelling 37 (2013) 1400–1412 1411

0.45

0.4

0.35

0.3

v
0.25

0.2

0.15

0.1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
u

Fig. 9. When m ¼ 21
40
and s ¼ 5, (4.1) has a periodic solution.

0.45

0.4

0.35

0.3
v

0.25

0.2

0.15

0.1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
u
51 
Fig. 10. When m ¼ 21
40
and s ¼ 10, the equilibrium E ¼ ;1
164 4
is stable.

5 1
Thirdly, we let m ¼ 12. Then E ¼ 16 ; 4 , which is stable if s < s03 2:583068 and is unstable if s > s03 . The bifurcation crit-
ical values are sk3 9:027947k þ s03 . As in the case above, we take s ¼ 2 and s ¼ 4 respectively to support the two situations
(see Figs. 6 and 7).
51 1 k
Finally, we take m ¼ 21 40
. In this case, we have E ¼ 164 ; 4 ; s4 10:079800k þ 3:224614 and sk4þ 19:669331kþ
8:324849. Note that we have s04 < s04þ < s14 < s24 < s14þ . It follows that E is stable for s < s04 , unstable for
  
s 2 s04 ; s04þ , stable for s 2 s04þ ; s14 , unstable for s 2 s14 ; s14þ ; . . ., that is, delay switches the stability. We demonstrate
the above phenomenon in Figs. 8–10 associated with delays s ¼ 2; 5; 10, respectively.

5. Concluding remarks

In this paper, taking into account of the availability of prey, we built a delayed predator–prey model with predator migra-
tion. This model is applicable to the important biological control. Our focus is the effects of the migration rate and delay on
the dynamics of the model.
The model can have at least two but at most three equilibria. When the migration rate is taken as a bifurcation parameter,
one has backward bifurcation. Then we studied the linear stability of the equilibria. It turns out that the stability of the trivial
equilibrium and the boundary equilibrium (if exists) is quite simple. Their stability is delay-independent. However, this may
not be true for the positive equilibrium.
For the positive equilibrium, we obtained sufficient conditions for its stability and instability. In most of the situations, the
stability is delay-dependent. Moreover, delay can switch the stability of the positive equilibrium. When the positive equilib-
1412 Y. Chen, F. Zhang / Applied Mathematical Modelling 37 (2013) 1400–1412

rium loses stability, Hopf bifurcation can occur. Using the center manifold method and the normal form theory, we also de-
rived the direction and stability of the Hopf bifurcation.
Now, we give some comments on biological control. The best result is to eliminate the prey. From Theorem 2.1(ii), this can
be achieved by finding natural enemies which have main alternative food to survive (that is, d < 0) and choosing the migra-
tion rate less than the intrinsic growth rate d of the predator. In this case, the hunting factor a instead of the migration rate
plays an important role. However, eliminating the prey is either impossible or very costly. In practice, we only need to keep
the prey below the economic injury level (EIL), which is the lowest population density of a pest that will cause economic
damage or the amount of pest injury justifying the cost of using controls [27]. From Theorem 2.1(iii), the coexistence equi-
librium E is always stable for small delay. In practice, it is true that the predator will not need a lot of time to convert the
prey biomass. In this case, we can keep the prey under EIL by adjusting the migration rate m. How to adjust m greatly de-
pends on the values of the other parameters. Unfortunately, we cannot give exact suggestions here since it is hard to get
useful information from the complicated expressions of P; Q , and R.
The main theoretical results were demonstrated by numerical simulations with hypothetical parameter values. We hope
to get some real data set to check the validity of our model. For future work, we also want to build some other models by
considering other functional responses or the effect of pulses.

Acknowledgments

The authors greatly appreciate the anonymous reviewers’ constructive comments on improving the presentation of this
work. The work is supported partially by the Natural Science and Engineering Research Council of Canada (NSERC), by the
Early Researcher Award (ERA) program of Ontario, by the National Natural Science Foundation of China (No. 10771139, No.
11071283), by the One Hundred Talents Project of Shanxi Province, by the Natural Science Foundation of Shanxi Province
(2009011005-3), and by the Program of Key Disciplines in Shanxi Province (20111028).

References

[1] N. Bairagiand, D. Jana, On the stability and Hopf bifurcation of a delay-induced predator–prey system with habitat complexity, Appl. Math. Model. 35
(2011) 3255–3267.
[2] S. Banerjee, B. Mukhopadhyay, R. Bhattacharyya, Effect of maturation and gestation delays in a stage structure predator prey model, J. Appl. Math.
Inform. 28 (2010) 1379–1393.
[3] S. Chakraborty, J. Chattopadhyay, Effect of cannibalism on a predator–prey system with nutritional value: a model based study, Dynam. Syst. 26 (2011)
13–22.
[4] T.K. Kar, K. Chakraborty, U.K. Pahari, A prey–predator model with alternative prey: mathematical model and analysis, Can. Appl. Math. Q. 18 (2010)
137–168.
[5] S. Kovács, Bifurcation in a predator–prey model with diffusion: rotating waves, Differ. Equat. Dynam. Syst. 17 (2009) 135–146.
[6] B. Leard, J. Rebaza, Analysis of predator–prey models with continuous threshold harvesting, Appl. Math. Comput. 217 (2011) 5265–5278.
[7] H. Li, Y. Takeuchi, Dynamics of the density dependent predator–prey system with Beddington–DeAngelis functional response, J. Math. Anal. Appl. 374
(2011) 644–654.
[8] P.-P. Liu, An analysis of a predator–prey model with both diffusion and migration, Math. Comput. Model. 51 (2010) 1064–1070.
[9] X. Liu, M. Han, Chaos and Hopf bifurcation analysis for a two species predator–prey system with prey refuge and diffusion, Nonlinear Anal. Real World
Appl. 12 (2011) 1047–1061.
[10] K. Oeda, Effect of cross-diffusion on the stationary problem of a prey–predator model with a protection zone, J. Differ. Equat. 250 (2011) 3988–4009.
[11] A. Sengupta, T. Kruppa, H. Lowen, Chemotactic predator–prey dynamics, Phys. Rev. E 83 (3) (2011) 031914 (6 pp).
[12] J. Sugie, Y. Saito, M. Fan, Global asymptotic stability for predator–prey systems whose prey receives time-variation of the environment, Proc. Amer.
Math. Soc. 139 (2011) 3475–3483.
[13] J. Zu, M. Mimura, The impact of Allee effect on a predator–prey system with Holling type II functional response, Appl. Math. Comput. 217 (2010) 3542–
3556.
[14] R.S. Banzon, A two-diemnsional predator–prey model, Sci. Diliman 11 (1991) 21–24.
[15] A. El Abdllaoui, P. Auger, B.W. Kooi, R. Bravo de la Parra, R. Mchich, Effects of density-dependent migrations on stability of a two-patch predator–prey
model, Math. Biosci. 210 (2007) 335–354.
[16] Y. Huang, O. Diekmann, Predator migration in response to prey density: what are the consequences?, J Math. Biol. 43 (2001) 561–581.
[17] R. Mchich, P. Auger, J.-C. Poggiale, Effect of predator density dependent dispersal of prey on stability of a predator–prey system, Math. Biosci. 206
(2007) 343–356.
[18] S.M. Merchant, W. Nagata, Wave train selection behind invasion fronts in reaction–diffusion predator–prey models, Physica D 239 (2010) 1670–1680.
[19] G. Sun, S. Sarwardi, P.J. Pal, Md. S. Rahman, The spatial patterns through diffusion driven instability in modified Leslie-Gower and Holling-type II
predator–prey model, J. Biol. Syst. 18 (2010) 593–603.
[20] G.-Q. Sun, Z. Jin, Q.-X. Liu, L. Li, Dynamical complexity of a spatial predator–prey model with migration, Ecol. Mod. 219 (2008) 248–255.
[21] D.-P. Cheng, Mathematical analysis on asymptotic stability of rodent–predator system, J. Biomath. 18 (2003) 283–286.
[22] T. Faria, Stability and bifurcation for a delayed predator–prey model and the effect of diffusion, J. Math. Anal. Appl. 254 (2001) 433–463.
[23] J.K. Hale, S.M. Verduyn Lunel, Introduction to Functional Differential equations, Springer-Verlag, New York, 1993.
[24] O. Diekmann, S.A. van Gils, S.M. Verduyn Lunel, H.-O. Walther, Delay Equations: Functional-, Complex-, and Nonlinear Analysis, Springer-Verlag, New
York, 1995.
[25] R. Bellman, K.L. Cooke, Differential–Difference Equations, Academic Press Inc., New York, 1963.
[26] B.D. Hassard, N.D. Kazarinoff, Y.H. Wan, Theory and Applications of Hopf Bifurcations, Cambridge University Press, Cambridge, 1981.
[27] V.M. Stern, R.F. Smith, R. van den Bosch, K.S. Hagen, The integrated control concept, Hilgardia 29 (1959) 81–101.

You might also like