DISCRETE AND CONTINUOUS doi:10.3934/dcdsb.2016.21.

737
DYNAMICAL SYSTEMS SERIES B
Volume 21, Number 3, May 2016 pp. 737–761

A PREY-PREDATOR MODEL WITH MIGRATIONS

AND DELAYS?

Isam Al-Darabsaha , Xianhua Tangb and Yuan Yuana,b,†

a Department of Mathematics and Statistics
Memorial University of Newfoundland
St. John’s NL, A1C 5S7, Canada
b School of Mathematics and Statistics

Central South University

Changsha, Hunan 410083, China

(Communicated by Yang Kuang)

Abstract. In this paper we propose a prey-predator model in multiple patches

through the stage structured maturation time delay with migrations among
patches. Focus on the case with two patches, we discuss the existence of equi-
librium points and the uniform persistence. In particular, when the maturation
times are the same in the patches, we study the local and global attractivity
of boundary equilibrium point with general migration function and the local
stability of the positive equilibrium with constant migration rate. Numerical
simulations are provided to demonstrate the theoretical results, to illustrate the
effect of the maturation time, the migration rate on the dynamical behavior of
the system.

1. Introduction. The prey-predator relationship is prevalent in the nature and

hence it is one of the major themes in mathematical models. In the 1920s A.
Lotka and V. Volterra formulated the original model to describe the prey-predator
interactions. Since then, different prey-predator models have been built by in-
corporating additional biological processes into the classical LotkaVolterra model
[2, 6, 8–10, 15, 17, 18, 20, 22, 23, 32, 36, 37]. For instance, in [8], the authors stud-
ied the dynamics of a diffusive prey-predator system with Allee effects in the prey
growth. In [15], a prey-predator model with a consideration of the prey refuge has
been studied. It is well-known that some kind of time delays are unavoidable in
prey-predator population and they can affect the stability of equilibrium points [19].
In [6], the authors presented a stage-structured predator prey model with gestation
delay. A delay prey-predator system without instantaneous negative feedback has
been studied in [32]. The authors in [37] considered a class of delayed Lokta-Volterra
prey-predator model with two delays. Time delay due to maturation of the preda-
tor is inescapable since the juvenile predators usually live on their parents until
they become mature, or on a resource that is not wanted by adults and available

2010 Mathematics Subject Classification. Primary: 34D20, 92D25; Secondary: 92D40.

Key words and phrases. Prey-predator, time delay, prey/predator migration, equilibrium
points, uniform persistence, stability.
? The research is supported in part by the NSERC of Canada.
† Corresponding author: yyuan@mun.ca.

737
738 ISAM AL-DARABSAH, XIANHUA TANG AND YUAN YUAN

in abundance. The effect of maturation delay on the behavior of populations has

been studied in [12].
In prey-predator system with multi-patch environment, migration occurs when
a species moves from its patch to another due to different factors like competition,
age, sex, lack of food, climate and the season which make it one of the prevalent
phenomena in nature [26]. For example, in aquatic environments, many zooplankton
species exhibit vertical movements each day due to light and food. During the
day time, some species migrate downwards into the darkness to reduce the risk of
predation by fish, while at night time, these species move upward to consume the
phytoplankton [4, 24]. Empirical studies have shown that the interaction between
patches has important consequences for population stability and persistence e.g.
[11,31,33]. The influence of such interaction, migration, capture scientists’ attention
in the communities. Some researchers considered different migration rates for both
predator and prey species. For instance, the authors constructed a two patches
model with migration of both the predator and the prey in [10], and assumed the
migration of the predator depend on the population of the prey in each patch,
while the prey migration is in a constant rate. They found upper and lower bounds
for the populations and discussed the stability and instability of the positive and
boundary equilibrium points. In [17], a model was provided with two patches where
the prey does not migrate and the predator in the patch with a higher density will
migrate to the patch with a lower density. In [22], it was assumed that the predator
migrates with a constant migration rate, while the prey migration depends on the
predator density, the authors proved the existence of a unique strictly positive
equilibrium. In [2], it was assumed that the prey (predator) migration rate depend
on the predator (prey) density and showed that for a large class of density-dependent
migration rules there exists a unique and stable equilibrium for migration.
To the best of our knowledge, there is not so many result in the literature in the
research of joined consideration of migration and time delay in biological models.
Introducing time delays to prey-predator models usually brings challenges to the
mathematical analysis and may change the dynamics. In [36], a model with two
prey and one predator population, time delay due to gestation and constant prey
migration rate was investigated. The authors discussed the persistence, derived
conditions for the existence and local/global stability of the positive equilibrium.
All the models studied in [2, 10, 17, 22] have not considered the influence of time
delay in any evolution stage although migrations were collaborated. In this pa-
per, we propose a general model with multiple parallel patches which combined
the migration interactions among all patches and constant maturation time delay
through stage structure. Following the biological feasibility, we take general func-
tional responses with respect to migration rates in both prey and predator classes
to allow a unified treatment for all biological important cases. From the viewpoint
of dynamics, we analyze the model with proving the positivity and boundedness of
the solutions, local stability of the steady states, global attractivity of boundary
equilibrium point and uniform persistence of the system. Then, we give numerical
simulations to show the theoretical results and to see how the dynamical behavior
is sensitive to the maturation time of predator, the migration rates in prey and
predator.
The paper is organized as follows: in Section 2, a general model of multiple paral-
lel prey-predator patches with migration of both species is studied. When the model
contains two patches, we provide conditions for the uniform persistence, discuss the
 A PREY-PREDATOR MODEL WITH MIGRATIONS AND DELAYS 739

existence of boundary equilibrium points, and further investigate their stabilities

with the same time delays. In Section 3, we explore the existence and stability of
positive equilibrium point for a special case when the two patches are symmetric
and the prey and predator migrations are constants. Numerical simulations are
given to compliment the theoretical analysis for such a general system in Section 4.
Conclusion is drawn in Section 5.

2. A model of multiple parallel prey-predator patches with migration

and maturation delays. In previous work [3], we studied a model including mul-
tiple parallel food chains, each consisting of a prey species Pi and its dedicated
predator Zi , with maturation delay and without interaction between patches. In
this paper we consider the interaction between patches, assume all Pi compete for
limiting nutrient, only adult predators are capable of preying on the prey species
and propose a more complex model for the multi-patch prey-predator system, with
prey (predator) migration rates depending on the corresponding predator (prey)
density:
n n
dPi X X
= Pi [µi (N ) − Zi hi (Pi )] + Pk fki (Zk ) − Pi fik (Zi ),
dt
k=1 k=1
k6=i k6=i

n
!
dZi −di τi
X
= bi e Pi (t − τi )Zi (t − τi )hi (Pi (t − τi )) − Zi i Zk (1)
dt
k=1
n
X n
X
+ Zk gki (Pk ) − Zi gik (Pi ),
k=1 k=1
k6=i k6=i

where i = 1, . . . , n,Pµi (N ) is the P

growth rate of Pi as a function of nutrient concen-
n n
tration N = T − k=1 αk Pk − k=1 βk Zk ≥ 0, with the coefficients αk , βk , (k =
1, · · · , n) related to the efficiency ofPnutrient consuming for each species. The au-
n Pn
thor in [5] used the form N = T − i=1 Pi − i=1 Zi which treats the consuming
abilities for all the species equally. As we know that in natural world, different
species may have different strength in taking nutrient, so to be more realistic, we
give different weight on nutrient consumption for the species. The constant param-
eter T is the internal nutrient !supply, hi (Pi ) is the per-prey-per-predator harvest
Pn
rate of Pi by Zi and i Zj is predation on the ith predator by higher trophic
j=1
n
P
levels which is assumed to be proportional to the total predator Zj (see [5]); bi
j=1
denotes the adult predator’s birth rate, di is the mortality death rate of the juvenile
(through-stage death rate), τi is time to mature and fij (Zi ) (gij (Pi )), j = 1, . . . , n
and i 6= j, represents the migration rate for prey (predator) from patch i to patch
j. All the parameters are positive constants. An architecture of the model (1) with
3 patches is given in Fig. 1.
From the biological point of view, we assume all the functions in (1) are con-
tinuous and differentiable over R+ , and satisfy the following hypotheses for i, j =
1, . . . , n with i 6= j:
dµi
(C1 ) µi (0) = 0, dN > 0 when N ∈ (0, T );
d(Pi hi (Pi ))
(C2 ) hi (Pi ) ≥ 0, dPi > 0, 0 ≤ Pi hi (Pi ) ≤ Mi , Mi is a constant;
740 ISAM AL-DARABSAH, XIANHUA TANG AND YUAN YUAN

Figure 1. Three parallel food chains with migration between patches.

∂i (Z) Pn
(C3 ) ∂Zk ≥ 0 for each Zk in Z = k=1 Zk ;
dfij
(C4 ) fij (Zi ) > 0 with fij (0) = γ0 and dZi ≥ 0 for Zi ∈ (0, βTi );
dgij
(C5 ) gij (Pi ) > 0 and dPi ≤ 0 for Pi ∈ (0, αTi ).
The assumptions of (C1 ), (C2 ) and (C3 ) are biologically obvious, (C4 ) and (C5 )
indicate that the preys escape from a patch more actively when more predators
are in the same patch, while if more preys stay in a patch, the movement of the
predator to other patch is inactive .
The general model (1) can cover several general, partial general and specific
models in literature with some particular choice of the functions. To name only
a few models with two patches without time delay, the system in [20] is a special
model of (1) with µ(N ) = µ(P ), h(P ) = p(P )
P , same , f (Z) = constant and
P b d
g(P ) = 0; when we take µ(N ) = r(1 − K ), h(P ) = b+P , g(P ) = 1+P , (Z) and
f (Z) are constant, the model (1) is the same as that in [10]; When one choose
fij (Z) = βi Z + β0 and gij as constant, the migration terms in (1) become the fast
df dg
model in [22]; while we restrict f and g with dZ > 0 and dP < 0, then the migration
terms in (1) become the fast model in [2].
Let τ = max{τi , i = 1, . . . , n} and define C := C([−τ, 0), R2n ). For φ =
2n
P
(φ1 , . . . , φ2n ) ∈ C, define kφk = kφi k∞ where
i=1

kφi k∞ = max |φi (θ)| .

θ∈[−τ,0]

Then C is a Banach space and C + = C([−τ, 0), R2n + ) is a normal cone of C with
nonempty interior in C. We have the following result for the positivity and bound-
edness properties of the system (1).
Theorem 2.1. If the initial condition (P1 (0), Z1 (0), . . . , Pn (0), Zn (0)) ∈ C + , then
the solutions of (1) are nonnegative in C. In addition, if i (0) > 0 for each i =
1, . . . , n then all the solutions are ultimately bounded in C.
 A PREY-PREDATOR MODEL WITH MIGRATIONS AND DELAYS 741

Proof. When n = 2 the system (1) becomes

dP1
= P1 [µ1 (N ) − Z1 h1 (P1 )] + P2 f21 (Z2 ) − P1 f12 (Z1 ),
dt
dZ1
= b1 e−d1 τ1 P1 (t − τ1 )Z1 (t − τ1 )h1 (P1 (t − τ1 )) − Z1 1 (Z1 + Z2 )
dt
+Z2 g21 (P2 ) − Z1 g12 (P1 ), (2)
dP2
= P2 [µ2 (N ) − Z2 h2 (P2 )] + P1 f12 (Z1 ) − P2 f21 (Z2 ),
dt
dZ2
= b2 e−d2 τ2 P2 (t − τ2 )Z2 (t − τ2 )h2 (P2 (t − τ2 )) − Z2 1 (Z1 + Z2 )
dt
+Z1 g12 (P1 ) − Z2 g21 (P2 ),

with N = T − α1 P1 − α2 P2 − β1 Z1 − β2 Z2 . It is straight forward that the solution

of (2) exists and is unique with initial value in C + and Pi (t) ≥ 0 and Zi (t) ≥ 0,
i = 1, 2 for all t > 0 (Theorem 5.2.1 in [29]).
Further, from the first and third equations in (2), and the fact that µi (N ) is
increasing, we have
P10 + P20 = P1 [µ1 (N ) − Z1 h1 (P1 )] + P2 [µ2 (N ) − Z2 h2 (P2 )]
≤ (P1 + P2 ) µ̆ (T − α̂ (P1 + P2 )) ,
where µ̆ (· · · ) = max{µ1 (· · · ) , µ2 (· · · )} and α̂ = min{α1 , α2 }. In the remain of
the paper, we denote ˆ·, ˘· as min and max, respectively. Since µ̆(0) = 0, we know
lim sup (P1 + P2 ) ≤ T /α̂, implying P1 (t) + P1 (t) is ultimately bounded.
t→∞
Let P (t) = P1 (t) + P2 (t), Z(t) = Z1 (t) + Z2 (t) and
ˆ ˆ
S(t) = b̆2 e−2dτ̂ (P (t − τ1 ) + P (t − τ2 )) + b̆e−dτ̂ Z(t),
we can obtain
ˆ ˆ
S 0 (t) ≤ b̆2 e−2dτ̂ P1 (t − τ1 )µ1 (N1 ) + b̆2 e−2dτ̂ P2 (t − τ1 )µ2 (N1 )
ˆ ˆ
+b̆2 e−2dτ̂ P1 (t − τ2 )µ1 (N2 ) + b̆2 e−2dτ̂ P2 (t − τ2 )µ2 (N2 )
ˆ ˆ
−b̆e−dτ̂ Z1 (t)1 (Z1 (t) + Z2 (t)) − b̆e−dτ̂ Z2 (t)2 (Z1 (t) + Z2 (t)) ,
where Ni = T − α1 P1 (t − τi ) − α2 P2 (t − τi ) − β1 Z1 (t − τi ) − β2 Z2 (t − τi ), i = 1, 2.
From the hypotheses (C1 ),(C3 ), we have µi (Nj ) ≤ µi (T ) and i (Z1 + Z2 ) ≥ i (0)
for i, j = 1, 2. Hence,
ˆ
S 0 (t) ≤ b̆2 e−2dτ̂ µ̆(T )(P (t − τ1 ) + P (t − τ2 ))
ˆ
+b̆2 e−2dτ̂ ˆ (0) (P (t − τ1 ) + P (t − τ2 ))
 
ˆ ˆ
 (0) b̆2 e−2dτ̂ (P (t − τ1 ) + P (t − τ2 )) + b̆e−dτ̂ Z(t)
−ˆ
ˆ
T b̆2 e−2dτ̂
≤ (µ̆(T ) + ˆ (0)) − ˆ (0) S(t).
α̂
T b̆2 e−2d̂τ̂
If ˆ(0) > 0 then lim sup S(t) ≤ α̂ˆ
(0) (µ̆(T ) + ˆ (0)). Therefore, P1 (t), P2 (t),
t→∞
Z1 (t) and Z2 (t) are ultimately bounded in C.
With general n, the idea is similar by using mathematical induction.

Remark 1. Theorem 2.1 implies that there exists t` > 0, LP > 0 and LZ > 0 such
n
P n
P
that 0 ≤ Pi (t) ≤ LP and 0 ≤ Zi (t) ≤ LZ for t > t` .
i=1 i=1
742 ISAM AL-DARABSAH, XIANHUA TANG AND YUAN YUAN

Obviously the system (1) has a trivial equilibrium point E0 = (0, 0, . . . , 0, 0). In
| {z }
2n
addition, we have

Proposition 1. Given the hypothesis (C1 − C5 ), then a unique predator-free equi-

librium point
 
 T T 
Ẽ = 
Pn , 0, . . . , P
n , 0

αi αi
i=1 i=1
| {z }
2n

exists in (1).

Proof. When n = 2 and Z1 = Z2 = 0. A predator-free equilibrium point must

satisfy

P1 µ1 (T − α1 P1 − α2 P2 ) + f21 (0)P2 − f12 (0)P1 = 0,

P2 µ2 (T − α1 P1 − α2 P2 ) + f12 (0)P1 − f21 (0)P2 = 0. (3)

Since f12 (0) = f21 (0) = γ0 , it is easy to check that Ẽ satisfies (3).
To prove the uniqueness of such Ẽ, it’s easy to see that when one of Pi , (i = 1, 2)
is zero, another one must be zero as well, which excludes the case with one prey
species can survive and another cannot.
Further, adding two equations in (3) yields

P1 µ1 (T − α1 P1 − α2 P2 ) + P2 µ2 (T − α1 P1 − α2 P2 ) = 0.

When Pi > 0, due to µi ≥ 0, (i = 1, 2), we have α1 P1 + α2 P2 = T . Substituting

T
into (3), we obtain P1 = P2 = α1 +α2
:= T . Consequentially, we have N = 0 at Ẽ.
Again, by mathematical induction method the result hold for a general n.

Remark 2. With migrations between patches, only one positive predator-free equi-
librium point exists for system (1). But without migration, the system has infinite
number of predator-free equilibrium points with the form
 
Ẽ = α1 P˜1 , 0, α2 P˜2 , 0, . . . , αn P˜n , 0
| {z }
2n

n
αk P˜k = T in [3]. The result in Proposition 1 implies the effect of migra-
P
with
k=1
tions.

To study the stability of each equilibrium point (x1 ,y1 ,x2 ,y2 , . . . ,xn ,yn ), we can
obtain the general form of the characteristic equation, which is,
n
!
X
0 k −λτk
∆ (λ) = det λI − J − J e =0
k=1


where J 0 = (aij ) and J k = ckij are 2n × 2n matrices with
 A PREY-PREDATOR MODEL WITH MIGRATIONS AND DELAYS 743

   
−x i+1 α i+1 µ0i+1 N̄ + y i+1 h0i+1 x i+1





 2 2 2 2 2 2

 h
 i Pn
+ µ N̄ − y h x − f (Z i+1 ) , i odd, j = i;

i+1 i+1 i+1 i+1 i+1

k



 2 2 2 2 k=1 2 2

 k6= i+1


  2 
  
−x i+1 β i+1 µ0i+1 N̄ + h i+1 x i+1




 , i odd, j = i + 1;
2 2 2 2

 2
−α j+1 x i+1 µ0i+1 N̄ + f j+1 i+1 (Z j+1 )


, i, j odd, i 6= j;

aij = 2 2 2 2
2 2


 −β j x i+1 µ0i+1 N̄ , i odd, j even j 6= i + 1;
2 2



  n   n2  n
yk − y i 0i
 P P P


 − i yk − g i k (P i ) , i even, j = i;

 2 k=1 2 2 k=1 k=1 2 2

k6= i



 2
n
  

−y i 0i
P
yk + g j i (P j ) , i, j even, i 6= j;



2
 2 2 2


 2 k=1
0 , i even, j odd,
n
P n
P
where N̄ = T − αk xk − βk yk and
k=1 k=1

 0 , i odd, any j; or i even, j 6= 2k, j 6= 2k − 1;
k
bk e−dk τk yk hk (xk )+xk h0k (xk )


cij = , i = 2k, j = 2k − 1;
−dk τk
bk e xk hk (xk ) , i = 2k, j = 2k;


Due to the complexity of the construction in the characteristic equation with

delay, we only discuss the stability of boundary equilibrium points when n = 2.
Theorem 2.2. E0 is always an unstable steady state.
Proof. The linearization of (2) at E0 is
x01 (t) = (µ1 (T ) − γ0 )x1 (t) + γ0 x2 (t),
y10 (t) = −(1 (0) + g12 (0))y1 (t) + g21 (0)y2 (t),
x02 (t) = γ0 x1 (t) + (µ2 (T ) − γ0 )x2 (t),
y20 (t) = g12 (0)y1 (t) − (2 (0) + g21 (0))y2 (t). (4)
The characteristic equation of (4) is
λ2 + (2γ0 − µ1 (T ) − µ2 (T ))λ + µ1 (T )µ2 (T ) − γ0 (µ1 (T ) + µ2 (T ))


∆(λ) =
× λ2 + (1 (0) + 2 (0) + g12 (0) + g21 (0))λ + 1 (0)2 (0) + g12 (0)2 (0) + g21 (0)1 (0) .

Let
h(λ) = λ2 + (2γ0 − µ1 (T ) − µ2 (T ))λ + µ1 (T )µ2 (T ) − γ0 (µ1 (T ) + µ2 (T )).
2
h(λ) = 0 has real roots since the discriminant of h(λ) is 4γ02 +(µ1 (T ) − µ2 (T )) > 0.
Assume all the roots in h(λ) = 0 have negative real parts, then
2γ0 − µ1 (T ) − µ2 (T ) > 0 and µ1 (T )µ2 (T ) − γ0 (µ1 (T ) + µ2 (T )) > 0,
which is equivalent to
µ1 (T ) + µ2 (T ) µ1 (T )µ2 (T )
γ0 > and > γ0 .
2 µ1 (T ) + µ2 (T )

Thus (µ2µ(T1 (T )µ2 (T )

)+µ2 (T ))2
> 1 leads to a contradiction. Therefore h(λ) = 0 has at least
1
one positive real root, implying E0 is unstable steady state.
744 ISAM AL-DARABSAH, XIANHUA TANG AND YUAN YUAN

Remark 3. We can conjecture that E0 = (0, 0, . . . , 0, 0) is unstable when n ≥

| {z }
2n
3, although we can only prove the result when all the functions µi (N ) are the
same, µi (N ):=µ(N ), and all the prey (predictor) migration rates are same constant
fij (Zi ) = m (gij (Pi ) = s). In this case the characteristic equation at E0 is ∆(λ) =
det(A)det(B) = 0 where
 
 λ − µ(T ) + (n − 1)m −m ··· −m 
 
 −m λ − µ(T ) + (n − 1)m · · · −m 
A=
 
.. .. . . .
. 

 . . . . 

 −m −m · · · λ − µ(T ) + (n − 1)m 
and
 
 λ + 1 (0) + (n − 1)s −s ··· −s 
 
 −s λ +  2 (0) + (n − 1)s ··· −s 
B= .
 
.. .. .. ..

 . . . . 

 −s −s ··· λ + n (0) + (n − 1)s 
It is easy to show that all the roots of det(B) = 0 are negative real roots by
using Gershgorin circle theorem in [27] and the symmetry of B. While det(A) =
(λ − µ(T ))(λ − µ(T ) + nm)n−1 = 0 has a positive real root λ = µ(T ), which means
E0 is unstable. When µ(T ) < nm, E0 is a saddle point; and if µ(T ) > nm, it is a
source.
To discuss the stability of the predator-free equilibrium point Ẽ = (T , 0, T , 0),
we introduce the following result given in [19] to discuss the global attractivity of
Ẽ.
Lemma 2.3. In a delay system
u0 (t) = au(t − τ ) − bu(t),

u(t) = φ(t) > 0, −τ ≤ t ≤ 0,
where a, b, τ > 0, the following results hold
(i) if a < b, then lim u(t) = 0.
t→∞
(ii) if a > b, then lim u(t) = ∞.
t→∞

For the special case with τ1 = τ2 := τ , we know Ẽ is globally attractive under

certain condition, which is given in the following theorem.
ˆ
Theorem 2.4. Assume τ1 = τ2 := τ > 0. When b̆T h̆(T )e−dτ < ˆ(0) then the
solutions of (2) satisfy Pi → T and Zi → 0, i = 1, 2.
Proof. Adding the second and the forth equations in (2), we have
Z10 + Z20 = b1 e−d1 τ P1 (t − τ )h1 (P1 (t − τ ))Z1 (t − τ ) − 1 (Z1 + Z2 )
+b2 e−d2 τ P2 (t − τ )h2 (P2 (t − τ ))Z2 (t − τ ) − 2 (Z1 + Z2 ).
Let δ > 0. Following the hypotheses (C1 ) and (C3 ), we have i (Z1 + Z2 ) ≥ ˆ(0) and
Pi (t − τ )hi (Pi (t − τ )) ≤ T hi (T ) for any t > tδ + τ . Thus
ˆ
Z10 + Z20 ≤ b̆T h̆(T )e−dτ (Z1 (t − τ ) + Z2 (t − τ )) − ˆ(0)(Z1 (t) + Z2 (t)).
 A PREY-PREDATOR MODEL WITH MIGRATIONS AND DELAYS 745

The comparison theory implies that Z1 (t) + Z2 (t) is bounded above by the solu-
tion of
ˆ
u0 (t) = b̆T h̆(T )e−dτ u(t − τ ) − ˆ(0)u(t), t > tδ + τ
such that u(t) = Z1 (t) + Z2 (t) for t ∈ [tδ , tδ + τ ]. When
ˆ
b̆T h̆(T )e−dτ < ˆ(0),
from Lemma 2.3, we have lim (Z1 (t) + Z2 (t)) = 0. Since both Z1 (t) and Z2 (t) are
t→∞
nonnegative from Theorem 2.1, we have lim Zi (t) = 0, i = 1, 2.
t→∞
By the existences and uniqueness of the predator-free equilibrium point Ẽ in
Proposition 1 and Remark 2, we obtain lim Pi (t) = T , i = 1, 2.
t→∞

ˆ
In general, beyond the strong condition b̆T h̆(T )e−dτ < ˆ(0) and/or with two
different delays τ1 , τ2 , we can discuss the local stability of Ẽ by setting x1 = P1 − T ,
y1 = Z1 ,x2 = P2 − T and y2 = Z2 . Then the linearization of (2) at Ẽ is
x01 (t) = a11 x1 (t) + a12 y1 (t) + a13 x2 (t) + a14 y2 (t),
y 01 (t) = a22 y1 (t) + a24 y2 (t) + b22 y1 (t − τ1 ),
x02 (t) = a31 x1 (t) + a32 y1 (t) + a33 x2 (t) + a34 y2 (t),
y 02 (t) = a42 y1 (t) + a44 y2 (t) + c44 y2 (t − τ2 ), (5)
where
a11 = −α1 T µ1 0 (0) − γ0 < 0, a12 = −β1 T µ1 0 (0) − T h1 (T ) − T f12
0
(0) < 0,
a13 = −α2 T µ1 0 (0) + γ0 , a14 = −β2 T µ1 0 (0) + T f21
0
(0),
a22 = −1 (0) − g12 (T ) < 0, a24 = g21 (T ) > 0, b22 = b1 e−τ1 d1 T h1 (T ) > 0,
a31 = −α1 T µ2 0 (0) + γ0 , a32 = −β1 T µ2 0 (0) + T f12
0
(0),
a33 = −α2 T µ2 0 (0) − γ0 < 0, a34 = −β2 T µ2 0 (0) − T h2 (T ) − T f21
0
(0) < 0,
a42 = g12 (T ) > 0, a44 = −2 (0) − g21 (T ) < 0, c44 = b2 e−τ2 d2 T h2 (T ) > 0.
The characteristic equation of (5) is
∆(λ, τ1 , τ2 ) = G(λ)K(λ, τ1 , τ2 ) = 0, (6)
with
G(λ) = λ2 − (a11 + a33 )λ + a11 a33 − a13 a31 ,
K(λ, τ1 , τ2 ) = λ − a22 − b22 e−λτ1 λ − a44 − c44 e−λτ2 − a24 a42 .

It is easy to check that G(λ) = 0 has two roots with negative real part, since
both −(a11 + a33 ) > 0 and a11 a33 − a13 a31 = γ0 T (α1 + α2 )(µ01 (0) + µ02 (0)) > 0. So
we only need to discuss the distribution of roots in K(λ, τ1 , τ2 ) = 0.
In the stability analysis, we are more interested in the influence of the key pa-
rameters, functions including the migration rates fij , gij and the time delays τi
(i = 1, 2). We notice that K(λ, τ1 , τ2 ) is independent of the prey migration rates
fij , so the prey migrations do not affect the stability of the predator-free equilib-
rium point Ẽ. On the other hand, if there is no predator migration, i.e. gij = 0,
Ẽ is locally asymptotically stable if bk T hk (T )e−dk τk < k (0)( k = 1, 2) in [3]; but
with general predator migration, the stability at Ẽ becomes complicated. In the
following, we focus on the stability with respect to time delay only and we will
investigate the influence of migration rates numerically.
746 ISAM AL-DARABSAH, XIANHUA TANG AND YUAN YUAN

More specifically, K(λ, τ1 , τ2 ) can be written as

K(λ, τ1 , τ2 ) = λ2 + k1 λ + k2 + e−λτ1 (k3 λ + k4 ) + e−λτ2 (k5 λ + k6 ) + k7 e−λ(τ1 +τ2 ) , (7)
where
k1 = −(a22 + a44 ) > 0, k2 = a22 a44 − a24 a42 > 0, k3 = −b22 < 0,
k4 = b22 a44 < 0, k5 = −c44 < 0, k6 = c44 a22 < 0, k7 = b22 c44 > 0. (8)
From
K(λ, 0, 0) = λ2 + (k1 + k3 + k5 ) λ + (k2 + k4 + k6 + k7 ) ,
following the Routh-Hurwitz stability criterion we have
Proposition 2. At τ1 = τ2 = 0, the predator-free equilibrium point Ẽ is locally
asymptotically stable if and only if
k1 + k3 + k5 > 0 & k2 + k7 + k4 + k6 > 0. (H1 )
With τ1 , τ2 > 0, first, we consider a simple case τ1 = τ2 := τ > 0, Eq. (7)
becomes
K(λ, τ ) = λ2 + k1 λ + k2 + (k3 + k5 ) λe−λτ + (k4 + k6 ) e−λτ + k7 e−2λτ . (9)
Then, for a pure imaginary root of (9), let λ = iω, ω > 0, then separate the real
and imaginary parts to obtain
ω 2 − k2 = (k3 + k5 ) ω sin ωτ + (k4 + k6 ) cos ωτ + k7 cos 2ωτ
k1 ω = − (k3 + k5 ) ω cos ωτ + (k4 + k6 ) sin ωτ + k7 sin 2ωτ. (10)
Squaring and adding both equations of (10) lead to
h i h i
2 2
ω 4 − 2k2 + (k3 + k5 ) − k12 ω 2 + k22 − (k4 + k6 ) − k72
= 2 (k4 + k6 ) k7 cos ωτ − 2ω (k3 + k5 ) k7 sin ωτ,
≤ −2 (k4 + k6 ) k7 − 2τ (k3 + k5 ) k7 ω 2 , (11)
since k7 > 0, (k4 + k6 ) < 0, (k3 + k5 ) < 0, ω > 0. (11) is equivalent to
h i h i
2 2
ω 4 − 2k2 + (k3 + k5 ) − k12 − 2τ (k3 + k5 ) k7 ω 2 + k22 − (k4 + k6 − k7 ) ≤ 0.
From the latter inequality we know that if
2
2k2 + (k3 + k5 ) − k12 − 2τ (k3 + k5 ) k7 < 0 (12)
and
2
k22 − (k4 + k6 − k7 ) > 0, (13)
then there is no such ω > 0. In addition, if K(0, τ ) = k2 + k7 + k4 + k6 6= 0, then
(9) has no roots with zero real part. Related to the time delay, we have,
Theorem 2.5. Assume τ1 = τ2 := τ > 0 and
(2 (0) + g21 (T )) b1 T h1 (T ) + (1 (0) + g12 (T )) b2 T h2 (T ) + b1 b2 T 2 h1 (T )h2 (T )
A= .
1 (0)2 (0) + g21 (T )1 (0) + g12 (T )2 (0)
Let
2 2
(1 (0) + 2 (0) + g12 (T ) + g21 (T )) − 2(b1 T h1 (T ) + b2 T h2 (T ))
τ̄ = > 0,
4b2 b2 T 2 (b1 T h1 (T ) + b2 T h2 (T )) h1 (T )h2 (T )
 1
dˆ
ln A , when A > 1
¯
τ̄ =
0 , when 0 < A < 1,
 A PREY-PREDATOR MODEL WITH MIGRATIONS AND DELAYS 747

and τ̄ > τ̄¯. If the predator-free equilibrium point Ẽ is locally asymptotically stable
at τ ∗ ∈ (τ̄¯, τ̄ ) then Ẽ is locally asymptotically stable for all τ ∈ (τ̄¯, τ̄ ).
Proof. Firstly, we consider the left side of (12). Substituting the expression k1 , . . . ,
k7 from (8), we have
2k2 + (k3 + k5 )2 − k12 − 2τ (k3 + k5 ) k7
< (b1 T h1 (T ) + b2 T h2 (T ))2 + 2τ (b1 T h1 (T ) + b2 T h2 (T )) b1 b2 T 2 h1 (T )h2 (T )
− (1 (0) + 2 (0))2 + (g12 (T ) + g21 (T ))2

< (b1 T h1 (T ) + b2 T h2 (T ))2 + 2τ (b1 T h1 (T ) + b2 T h2 (T )) b1 b2 T 2 h1 (T )h2 (T )

1
− (1 (0) + 2 (0) + g12 (T ) + g21 (T ))2 .
2
So if
2 2
(1 (0) + 2 (0) + g12 (T ) + g21 (T )) − 2(b1 T h1 (T ) + b2 T h2 (T ))
τ < τ̄ :=
4b2 b2 T 2 (b1 T h1 (T ) + b2 T h2 (T )) h1 (T )h2 (T )
then (12) holds.
Secondly, since k2 − k4 − k6 + k7 is always positive, we can see (13) is true if and
only if k2 + k4 + k6 − k7 > 0. From
k2 + k4 + k6 − k7
= 1 (0)2 (0) + g21 (T )1 (0) + g12 (T )2 (0) − (2 (0) + g21 (T )) b1 T h1 (T )e−d1 τ
− (1 (0) + g12 (T )) b2 T h2 (T )e−d2 τ − b1 b2 T 2 h1 (T )h2 (T )e−(d2 +d1 )τ > B,
where
B := 1 (0)2 (0) + g21 (T )1 (0) + g12 (T )2 (0)
ˆ
− (2 (0) + g21 (T )) b1 T h1 (T ) + (1 (0) + g12 (T )) b2 T h2 (T ) + b1 b2 T 2 h1 (T )h2 (T ) e−dτ

So if A > 1, then B > 0 for τ > τ̄¯ := d1ˆ ln A and when 0 < A < 1, B > 0 for all
τ > 0. Therefore, when τ ∈ (τ̄¯, τ̄ ), both (12) and (13) hold. Thus, (9) has no pure
imaginary roots.
Under the assumption that Ẽ is locally asymptotically stable at τ ∗ , we have that
Ẽ is locally asymptotically stable on this interval, since there are no sign changes
in roots of K(λ, τ ) = 0 for τ ∈ (τ̄¯, τ̄ ).
Adapting ideas from [35, 38], we can study the persistence of the system (2),
when one of the following assumptions holds:
b1 e−d1 τ1 T h1 (T ) > 1 (LZ ) + g12 (0) (R1 )
or
b2 e−d2 τ2 T h2 (T ) > 2 (LZ ) + g21 (0). (R2 )
Let
X = {φ = (φ1 , φ2 , φ3 , φ4 ) ∈ C + }
and the solution semiflow Φ(t)φ = ut (φ), φ ∈ X, t ≥ 0.
The following result shows that system (2) is uniformly persistent if either (R1 )
or (R2 ) holds.
Theorem 2.6. When either (R1 ) or (R2 ) holds, there is a positive number η > 0
such that every solution (P1 (t), Z1 (t), P2 (t), Z2 (t)) in system (2) with Pi (0) > 0 and
Zi (0) > 0 ,i = 1, 2, satisfies
lim inf (P1 (t), Z1 (t), P2 (t), Z2 (t)) ≥ (η, η, η, η).
t→∞
748 ISAM AL-DARABSAH, XIANHUA TANG AND YUAN YUAN

Proof. We prove the theorem under the first assumption of (R1 ). For the second
case, a similar proof works.
Define
X1 = {φ ∈ X : φi (0) = 0, i = 1, 3},
X2 = {φ ∈ X : φi (0) = 0, i = 2, 4},
∂X0 = X1 ∪ X2 ,
X0 = X \ ∂X0 = {φ ∈ X : φi (0) > 0, i = 1, 2, 3, 4},
and
M∂ = {φ ∈ ∂X0 : Φ(t)φ ∈ ∂X0 , t ≥ 0}, .
Claim 1. There exists a δ1 > 0, such that for any φ ∈ X0 ,
lim sup kΦ(t)φ − E0 k ≥ δ1 .
t→∞

By contradiction, suppose that lim sup kΦ(t)ψ − E0 k < δ1 fore some ψ ∈ X0 .

t→∞
Thus, there exists t0 > 0 such that 0 ≤ Pi (t, ψ) < δ1 and 0 ≤ Zi (t, ψ) < δ1 for
t > τ + t0 , i = 1, 2.
Let Fi (P1 , Z1 , P2 , Z2 ) = µi (N ) − Zi hi (Pi ) − fij (Zi ), i = 1, 2, j = 3 − i. Since
Fi is continuous, Fi (P1 , Z1 , P2 , Z2 ) → µi (T ) − γ0 as (P1 , Z1 , P2 , Z2 ) → (0, 0, 0, 0),
there exists σ1 = σ1 (δ1 ) such that
|Fi (P1 , Z1 , P2 , Z2 ) − (µi (T ) − γ0 )| < σ1 .
By (C4 ), if φ1 (0) > 0 and φ3 (0) > 0, then
Pi0 (t) = Pi Fi (P1 , Z1 , P2 , Z2 ) + Pj fji (Zj ) ≥ γ0 Pj (t) + (µi (T ) − γ0 − σ1 ) Pi (t).
Consider the linear system
u01 (t) = γ0 u2 (t) + (µ1 (T ) − γ0 − σ1 ) u1 (t),
u02 (t) = γ0 u1 (t) + (µ2 (T ) − γ0 − σ1 ) u2 (t). (14)
From the linearization of (2) at E0 in (4), we know that the principle eigenvalue λ1
and the corresponding eigenfunction are positive. Due to the continuity of λ, there
exists a sufficiently small positive number such that λ1 (σ1 ) > 0, and a solution
of (14) is ~u(t) = eλ1 (σ1 )t φ0 , where φ0 is the positive eigenfunction associated with
λ1 (σ1 ), ~u and φ0 are vectors with two components. When Pi (t, φ0 )  0 for all
t > 0, i = 1, 2, the comparison theory implies that there exists a small ξ > 0 such
that (P1 (t, φ0 ), P2 (t, φ0 )) ≥ ξeλ1 (σ1 )t φ0 for all t ≥ t0 in (14). Since λ1 (σ1 ) > 0, we
have lim (P1 (t, φ0 ), P2 (t, φ0 ))T = ∞, which is a contradiction.
t→∞
Claim 2. There exists a δ2 > 0, such that for any φ ∈ X0 ,
lim sup kΦ(t)φ − Ẽk ≥ δ2 .
t→∞

Assume the contrary, that lim sup kΦ(t)φ − Ẽk < δ2 fore some φ ∈ X0 . Then
t→∞
there exists a large number t1 > 0 such that kΦ(t)φ − (T , 0, T , 0)k < δ2 for all
t ≥ t1 + τ .
Without loss of generality, we assume that (R1 ) holds. Choose ζ > 0 small
enough such that
b1 e−d1 τ1 (T − ζ)h1 (T − ζ) > 1 (LZ ) + g12 (0).
 A PREY-PREDATOR MODEL WITH MIGRATIONS AND DELAYS 749

Let tζ > t1 + τ be sufficiently large such that

T − ζ < P1 (t) < T + ζ, for t > tζ .
Thus, by (C3 ) and (C5 ) for t > tζ
Z10 (t) > b1 e−d1 τ1 (T − ζ)h1 (T − ζ)Z1 (t − τ1 ) − (1 (LZ ) + g12 (0))Z1 (t).
By the comparison principle and Lemma 2.3, we get Z1 (t) → ∞ as t → ∞, which
is a contradiction.
S
Claim 3. ω(φ) = E0 ∪ Ẽ.
φ∈M∂
Let ω(φ) be the omega limit set of the orbit Φ(t) through φ ∈ X. For any φ ∈ M∂ ,
we have Φ(t)φ ∈ X1 or Φ(t)φ ∈ X2 . If Φ(t)φ ∈ X1 , i.e. P1 (t, φ) = P2 (t, φ) ≡ 0,
then
(Z1 + Z2 )0 ≤ − min{1 (0), 2 (0)}(Z1 + Z2 ).
By the comparison theory, we get Z1 (t, φ)+Z2 (t, φ) → 0 as t → ∞. Thus, Zi (t, φ) →
0 as t → ∞, i = 1, 2.
If Φ(t)φ ∈ X2 , i.e. Z1 (t, φ) = Z2 (t, φ) ≡ 0, then
Pi0 = Pi µi (T − α1 P1 − α2 P2 ) + γ0 (Pj − Pi ) := Gi (P1 , P2 ). (15)
From the proof Proposition 1, the system (15) has only two equilibrium points
T
(0, 0) and ( α1 +α , T ). It is easy to check that (0, 0) is always unstable. Denote
2 α1 +α2
T
G(P1 , P2 ) = (G1 (P1 , P2 ), G2 (P1 , P2 )) , E = {(P1 , P2 ) : 0 < α1 P1 + α2 P2 < T } and
ρ = (P1 P2 )−1 . Then
−α1 0 γ0 α2 0 γ0
∇ · (ρG) = µ 1 (T − α1 P1 − α2 P2 ) − 2 − µ (T − α1 P1 − α2 P2 ) − 2 < 0,
P2 P1 P1 2 P2
for any P1 , P2 ∈ E. Thus by Dulacs negative criterion, there is no periodic orbits
T
within E. Then we have, Pi (t, φ) → α1 +α2
as t → ∞, i = 1, 2. Consequently, we
S
have ω(φ) = E0 ∪ Ẽ. Thus the claim holds.
φ∈M∂
Define a continuous function p : X → R+ by
p(φ) = min{φ1 (0), φ2 (0), φ3 (0), φ4 (0)}, ∀φ ∈ X.
It is clear that p−1 (0, ∞) ⊂ X0 and if p(φ) > 0 then p(Φ(t)φ) > 0 for all t > 0.
By Claim 3, we get that for any forward orbit of Φ(t) in M∂ converges to E0 or Ẽ,
by Claim 1 and 2, we conclude that E0 and Ẽ are two isolated invariant in X, and
(W s (E0 ) ∪ W s (Ẽ)) ∩ X0 = ∅ and no subset of {E0 , Ẽ} form a cycle in ∂X. By [30],
it then follows that there exists η > 0 such that lim inf p(Φ(t)φ) ≥ η for all φ ∈ X0 ,
t→∞
which implies the uniform persistence. This completes the proof.

3. With constant migration rates. After we obtain the conditions for the ex-
istence and stability of the boundary equilibrium points E0 and Ẽ, we simplify the
model (1) and assume that all the prey (predictor) migration rates are the same
constant fij (Zi ) = m (gij (Pi ) = s) to explore the existence and local stability of
positive equilibrium solutions for i = 1, 2, j = 3 − i. With this simplification, the
system (2) becomes
dP1
= P1 [µ1 (N ) − Z1 h1 (P1 )] + m (P2 − P1 ) ,
dt
dZ1
= b1 e−d1 τ1 P1 (t − τ1 )Z1 (t − τ1 )h1 (P1 (t − τ1 )) − Z1 1 (Z1 + Z2 ) + s (Z2 − Z1 ) ,
dt
750 ISAM AL-DARABSAH, XIANHUA TANG AND YUAN YUAN

dP2
= P2 [µ2 (N ) − Z2 h2 (P2 )] + m (P1 − P2 ) , (16)
dt
dZ2
= b2 e−d2 τ2 P2 (t − τ2 )Z2 (t − τ2 )h2 (P2 (t − τ2 )) − Z2 2 (Z1 + Z2 ) + s (Z1 − Z2 ) .
dt
When the system (16) exists a positive equilibrium point E ∗ = (P1∗ , Z1∗ , P2∗ , Z2∗ ),
then Pi∗ , Zi∗ , i = 1, 2 must satisfy
P1∗ [µ1 (N ∗ ) − Z1∗ h1 (P1∗ )] + m (P2∗ − P1∗ ) = 0,
−d1 τ1
b1 e P1∗ Z1∗ h1 (P1∗ ) − Z1∗ 1 (Z1∗ + Z2∗ ) + s (Z2∗ − Z1∗ ) = 0,
P2∗ [µ2 (N ∗ ) − Z2∗ h2 (P2∗ )] + m (P1∗ − P2∗ ) = 0,
b2 e−d2 τ2 P2∗ Z2∗ h2 (P2∗ ) − Z2∗ 2 (Z1∗ + Z2∗ ) + s (Z1∗ − Z2∗ ) = 0, (17)
∗
with N = T − −α1 P1∗ − α2 P2∗
− β1 Z1∗ β2 Z2∗
> 0. Otherwise, it leads to a con-
tradiction with the summation of the first and third equations in (17). From the
second and forth equations of (17), we have
Z1∗ 1 (Z1∗ + Z2∗ ) + s (Z2∗ − Z1∗ )
= b1 e−d1 τ1 P1∗ h1 (P1∗ ) ≤ b1 e−d1 τ1 T h1 (T ),
Z1∗
Z2 2 (Z1 + Z2 ) + s (Z1∗ − Z2∗ )
∗ ∗ ∗
= b2 e−d2 τ2 P2∗ h2 (P2∗ ) ≤ b2 e−d2 τ2 T h2 (T ), (18)
Z2∗
respectively. Since i (0) ≤ i (Z1∗ + Z2∗ ) for i = 1, 2, we have
Z2∗ b1 e−d1 τ1 T h1 (T ) + s − 1 (0)
≤
Z1∗ s
Z1∗ b2 e−d2 τ2 T h2 (T ) + s − 2 (0)
∗ ≤ ,
Z2 s
provided that bi e−di τi T hi (T ) + s > i (0), i = 1, 2. Thus
s b2 e−d2 τ2 T h2 (T ) + s − 2 (0)
≤ ,
b1 e−d1 τ1 T h1 (T ) + s − 1 (0) s
which leads to
(b1 e−d1 τ1 T h1 (T ) − 1 (0))(b2 e−d2 τ2 T h2 (T ) − 2 (0))
s≤ . (H2 )
(1 (0) − b1 e−d1 τ1 T h1 (T )) + (2 (0) − b2 e−d2 τ2 T h2 (T ))
Therefore,
Lemma 3.1. When (H2 ) does not hold, then E ∗ does not exist. In particular, if
τ1 = τ2 := τ and
(b1 T h1 (T ) − 1 (0)) (b2 T h2 (T ) − 2 (0)) + s(b1 T h1 (T ) + b2 T h2 (T )) < 1 (0) + 2 (0),
E ∗ does not exist for any feasible τ > 0.
The condition (H2 ) is necessary for the existence of a positive equilibrium point
E ∗ = (P1∗ , Z1∗ , P2∗ , Z2∗ ) in (17), but is not a sufficient condition in general.
Define Hi (Pi ) = Pi hi (Pi ) for i = 1, 2. From the hypothesis (C2 ), Hi (Pi ) is
T
positive and increases in (0, α1 +α 2
). We can obtain
 ∗
Z1 1 (Z1∗ + Z2∗ ) + s (Z1∗ − Z2∗ )

∗ −1
P1 = H 1 ;
b1 e−d1 τ1 Z1∗
 ∗
Z2 2 (Z1∗ + Z2∗ ) + s (Z2∗ − Z1∗ )

P2∗ = H2−1 ,
b2 e−d2 τ2 Z2∗
 A PREY-PREDATOR MODEL WITH MIGRATIONS AND DELAYS 751

from the second and forth equations of (17), respectively, where Z1∗ and Z2∗ are the
implicit positive solutions of the following equations:
Z1∗ 1 Z1∗ + Z2∗ + s Z1∗ − Z2∗ Z ∗ 1 Z1∗ + Z2∗ + s Z1∗ − Z2∗


!


−1 ∗

H1 µ1 N − 1
b1 e−d1 τ1 Z1∗ b1 e−d1 τ1
Z2∗ 2 Z1∗ + Z2∗ + s Z2∗ − Z1∗ Z1∗ 1 Z1∗ + Z2∗ + s Z1∗ − Z2∗


!

!!
−1 −1
+m H2 − H1 = 0;
b2 e −d τ
2 2Z ∗ b1 e−d1 τ1 Z1∗
2
Z2∗ 2 Z1∗ + Z2∗ + s Z2∗ − Z1∗ Z ∗ 2 Z1∗ + Z2∗ + s Z2∗ − Z1∗


!


−1 ∗

H2 µ2 N − 2
b2 e−d2 τ2 Z2∗ b2 e−d2 τ2
Z1∗ 1 Z1∗ + Z2∗ + s Z1∗ − Z2∗ Z2∗ 2 Z1∗ + Z2∗ + s Z2∗ − Z1∗


!

!!
−1 −1
+m H1 − H2 = 0,
b1 e −d τ
1 1Z ∗ b2 e−d2 τ2 Z2∗
1

with
Z1∗ 1 Z1∗ + Z2∗ + s Z1∗ − Z2∗ Z2∗ 2 Z1∗ + Z2∗ + s Z2∗ − Z1∗ 2


!

!
∗ −1 −1 ∗
X
N = T − α1 H 1 − α2 H2 − βi Zi ,
b1 e −d1 τ1 ∗
Z1 b2 e−d 2 τ2 ∗
Z2 i=1

which is impossible to find an analytical solution in general.

With fixed parameters, feasible values of delays and certain functions, for in-
stance, choosing the functions given in Table (1) (see [10]), we have E1∗ = (0.093,
0.57, 0.048, .56) when τ1 = 0.4 and τ2 = 0.7.

h(P ) µ(N ) (Z)

1 N
Patch 1 P +1 T
0.02
2 N
Patch 2 P +2 T
0.03
Table 1. : Existence of E1∗
with αi = βi = 1, bi = 0.25, di = 0.01,
s = 0.1, m = 0.6 and T = 7 for i = 1, 2.

When the two parallel food chains are symmetric, that is, the growth for all
patches have the same functional forms, parameter values and the maturation delay,
the system (16) becomes
dPi
= Pi [µ (N ) − Zi h (Pi )] + m (Pj − Pi ) , (19)
dt
dZi
= be−dτ Pi (t − τ ) Zi (t − τ ) h (Pi (t − τ )) − Zi  (Zi + Zj ) + s (Zj − Zi ) ,
dt
i = 1, 2, j = 3 − i and N = T − α(P1 + P2 ) − β(Z1 + Z2 ). Then (H2 ) becomes

(be−dτ T h(T ) + 2s − (0))(be−dτ T h(T ) − (0)) ≥ 0, (H02 )

 
which implies τ ≤ τmax = d1 ln bT(0)
h(T )
.
Let
τs = sup {τ ∈ (0, τmax ) |system (19) has at least a positive equilibrium point } .
When the per-prey-per-predator harvest rate h(P ) is increasing, we know the pos-
itive equilibrium point E ∗ exists with the form of (P ∗ , Z ∗ , P ∗ , Z ∗ ).
T
Proposition 3. Assume h(P ) is an increasing function on the interval (0, 2α ). If
the positive equilibrium point E exists, then it must have the form of (P , Z ∗ , P ∗ ,
∗ ∗

Z ∗ ).
752 ISAM AL-DARABSAH, XIANHUA TANG AND YUAN YUAN

Proof. Assume (P1∗ , Z1∗ , P2∗ , Z2∗ ) is a positive equilibrium point. We can show P1∗ =
P2∗ and Z1∗ = Z2∗ . Otherwise, we have,
2 2
m(P1∗ − P2∗ ) = Z2∗ P1∗ H(P2∗ ) − Z1∗ P2∗ H(P1∗ ), (20)
2 2
s(Z1∗ − Z2∗ ) = be−dτ Z1∗ Z2∗ (H(P1∗ ) − H(P2∗ )). (21)
Suppose P1∗
> P2∗ ,
then by (21) and the hypothesis (C2 ), we obtain Z1∗ > Z2∗ . Form
the right hand side of (20), we have
Z2∗ P1∗ H(P2∗ ) − Z1∗ P2∗ H(P1∗ ) < Z1∗ P1∗ P2∗ (h(P2∗ ) − h(P1∗ )) < 0 (22)
when h(P ) is increasing. Then we arrive to a contradiction. Similar contradiction
arrives if P1∗ < P2∗ . Therefore, P1∗ = P2∗ and Z1∗ = Z2∗ .
T
Remark 4. When h(P ) is not necessary increasing on the interval (0, 2α ), beside
∗ ∗ ∗ ∗
the positive equilibrium point (P , Z , P , Z ), without additional condition, we
cannot exclude any other form in E ∗ . In other words, it is possible to have other
positive equilibrium point with different Pi∗ and Zi∗ , i = 1, 2.
When the positive equilibrium point E ∗ = (P ∗ , Z ∗ , P ∗ , Z ∗ ) exists, the lineariza-
tion of (19) at E ∗ is
dx1
= a11 x1 (t) + a12 y1 (t) + a13 x2 (t) + a14 y2 (t),
dt
dy1
= a22 y1 (t) + a24 y2 (t) + b21 x1 (t − τ ) + b22 y1 (t − τ ), (23)
dt
dx2
= a13 x1 (t) + a14 y1 (t) + a11 x2 (t) + a12 y2 (t),
dt
dy2
= a24 y1 (t) + a22 y2 (t) + b21 x2 (t − τ ) + b22 y2 (t − τ ),
dt
with
a11 = −αP ∗ µ0∗ − P ∗ Z ∗ h0∗ − m, a12 = −βP ∗ µ0∗ − P ∗ h∗ < 0,
a13 = −αP ∗ µ0∗ + m, a14 = −βP ∗ µ0∗ < 0, a22 = −Z ∗ 0∗ − ∗ − s < 0
a24 = −Z ∗ 0∗ + s, b21 = be−dτ Z ∗ (P ∗ h0∗ + h∗ ) > 0, b22 = be−dτ P ∗ h∗ > 0,
where µ∗ , h∗ , ∗ denotes the value of the function µ, h,  and µ0∗ , h0∗ , 0∗ denotes the
value of the derivative of the associated function at (P ∗ , Z ∗ , P ∗ , Z ∗ ) respectively.
Lemma 3.2. ( [1, 28]) When the block matrices A, B are square matrices of the
same order, then
 
A B
det = det(A − B) det(A + B).
B A
   
λ − a11 −a12 −a13 −a14
In our case, A = and B = .
−b21 e−dτ λ − a22 − b22 e−dτ 0 −a24
Thus the characteristic equation of (23) is
∆(λ, τ ) = Φ(λ, τ )Ψ(λ, τ ) = 0, (24)
where
Φ(λ, τ ) = λ2 + φ1 λ + φ2 + φ3 e−λτ + ζλe−λτ , (25)
2 −λτ −λτ
Ψ(λ, τ ) = λ + ψ1 λ + ψ2 + ψ3 e + ζλe , (26)
 A PREY-PREDATOR MODEL WITH MIGRATIONS AND DELAYS 753

φ1 = a24 + a13 − a22 − a11 , φ2 = (a24 − a22 )(a13 − a11 ), ζ = −b22 < 0.
φ3 = b21 (a14 − a12 ) − b22 (a13 − a11 ), ψ1 = −(a24 + a13 + a22 + a11 ),
ψ2 = (a24 + a22 )(a13 + a11 ), ψ3 = b21 (a14 + a12 ) + b22 (a13 − a11 ).
As we know that all the roots of Φ(λ, τ ) = 0 and Ψ(λ, τ ) = 0 with Re(λ) ≥ 0 lie
in a bounded domain [34]. When τ = 0,
∆(λ, 0) = Φ(λ, 0)Ψ(λ, 0)
= (λ2 + (φ1 + ζ)λ + (φ2 + φ3 ))(λ2 + (ψ1 + ζ)λ + (ψ2 + ψ3 )) = 0.
Parallel to the Proposition 2, we have
Proposition 4. At τ = 0, the positive equilibrium point E ∗ is locally asymptotically
stable if and only if
φ1 > −ζ , φ2 + φ3 > 0, ψ1 > −ζ and ψ2 + ψ3 > 0. (HΦΨ
0 )

Let τ > 0 and suppose λ = iω (ω > 0) is a purely imaginary root of Φ(λ, τ ) = 0

in (25). Substituting it into Φ(λ, τ ) = 0 and separating the real and imaginary
parts, we obtain:
ω 2 − φ2 = φ3 cos ωτ + ζω sin ωτ,
φ1 ω = φ3 sin ωτ − ζω cos ωτ. (27)
Squaring and adding both equations of (27) lead to
FΦ (ω, τ ) = ω 4 + qΦ (τ )ω 2 + `Φ (τ ) = 0
where
qΦ (τ ) = φ21 − 2φ2 − ζ 2 , `Φ (τ ) = φ22 − φ23 .
Let ω = ω(τ ) be a positive root of FΦ (ω, τ ) = 0. From (27), for iω(τ ) to be a
solution of Φ(λ, τ ) = 0, ω(τ ) needs to satisfy the system
ζω 3 + (φ1 φ3 − ζφ2 )ω
sin ω(τ )τ = ,
ζ 2 ω 2 + φ23
(φ3 − ζφ1 )ω 2 − φ2 φ3
cos ω(τ )τ = . (28)
ζ 2 ω 2 + φ23
Define the function θΦ (τ ) ∈ [0, 2π) such that sin θΦ (τ ) and cos θΦ (τ ) are given by
(28), and
θΦ (τ ) + 2nπ
SnΦ (τ ) = τ − , n ∈ N. (29)
ω(τ )
Following [7, 21], iω(τ ∗ ) is a root in Φ(λ, τ ) = 0 if and only if τ ∗ is a zero of a
function SnΦ for some n ∈ N.
Proposition 5. If τ ∗ ∈ (0, τmax ) is a positive root of SnΦ given in (29) for some
n ∈ N, then a pair of purely imaginary roots iω(τ ∗ ) of Φ(λ, τ ) = 0 exist which
Φ
dSn (τ ∗ )
crosses the imaginary axis from left to right if τ > 0 and crosses the imaginary
Φ
dSn (τ ∗ )
axis from right to left if τ < 0, and
(  )  Φ  
dRe (λ)  dSn (τ ) 
Sign = Sign .
dτ λ=iω(τ ∗ ) dτ τ =τ ∗
754 ISAM AL-DARABSAH, XIANHUA TANG AND YUAN YUAN

Similarly, we can obtain FΨ (ω, τ ) = 0 and SnΨ (τ ) to find a pure imaginary roots
iω(τ ∗ ) for Ψ(λ, τ ) = 0 in (26). Then the stability of E ∗ = (P ∗ , Z ∗ , P ∗ , Z ∗ ) is just
followed:
Proposition 6. (i) When (HΦΨ 0 ) holds and both FΦ (ω, τ ) = 0, FΨ (ω, τ ) = 0
have no positive roots, E ∗ is locally asymptotically stable for any feasible time
delay τ ∈ (0, τs );
∗
(ii) When (HΦΨ 0 ) holds and FΦ (ω, τ ) = 0\FΨ (ω, τ ) = 0 has a positive root, E is
∗
locally asymptotically stable for small time delay τ , say τ < min{τn } where
τn∗ is a zero of a function SnΦ \SnΨ for some n ∈ N;
(iii) When (HΦΨ 0 ) does not hold and FΦ (ω, τ ) = 0\FΨ (ω, τ ) = 0 has a positive root,
Φ
dSn (τ ∗ ) Ψ
dSn (τ ∗ )
say τ ∗ , E ∗ is unstable for small time delay τ . If τ < 0\ τ < 0, E ∗
can become stable with τ > τ ∗ .
Remark 5. We omit the implicit conditions for the existence of positive roots in
FΦ (ω, τ ) = 0 or FΨ (ω, τ ) = 0, which can be found in [3].

4. Numerical simulations. In this section, we fix the parameters

di = 0.01, bi = 0.25, αi = 1 βi = 1, (30)
for each i and choose some different functions from literature, see Table (2), to
compliment the theoretical results given previously.

µ(N ) P h(P ) (Z) f (Z) g(P )

µmax N
K+N rP aP n o m con- s con-
Monod a+bP Holling b+P n , n > 1 constant stant stant
type II [14] sigmoid type [13]
function [25] [20] [10] [22]
d
N
rP 2
a+bP 2
Holling r(1 − e−aP ) ϑ1 Z + 1+P
T [10] aZ + b
type III [14] Ivlev type [16] ϑ2 [22] [10]
Table 2. Functional responses.

In the following example, we consider a symmetric system with constant migra-

tion rates. First, we fix the value of the migration rates and study the dynamic
behavior of system (19) as the time delay varies. Then, we fix the time delay and
one of the migration rates to explore the affection of another migration rate on the
stability of the positive steady state.
Example 1. A symmetric system with constant migration rates. Choose
N P
µ(N ) = , h(P ) = 2 , (Z) = 0.15 and T = 8. (31)
1+N P +1
With constant migration rates, the system has a coexistence equilibrium point
E ∗ (τ ) = (P ∗ (τ ), Z ∗ (τ ), P ∗ (τ ), Z ∗ (τ )), where
r
∗ 3
P (τ ) = −0.01τ
;
5e −3
√
9P ∗ + 2 − 16P ∗4 − 64P ∗3 + 97P ∗2 − 28P ∗ + 4
Z ∗ (τ ) = ,
4P ∗
for τ < 45.02, which is independent of the constant migration rates.
 A PREY-PREDATOR MODEL WITH MIGRATIONS AND DELAYS 755

At τ = 0, E ∗ (0) = (1.22,1.47,1.22,1.47) is stable by checking the condition

(HΦΨ0 ). With τ >0 and fixed values g12 (P1 ) = g21 (P2 ) = 0.1 and f12 (Z1 ) =
f21 (Z2 ) = 0.6, we plot S0Ψ (τ ), S1Ψ (τ ) and S2Ψ (τ ) defined in Eq. (24) in Fig. 2. We
can see that the curve of S1Ψ (τ ) is negative when τ < τ1∗ = 0.41 and then becomes
positive as τ increases up to τ2∗ = 24.11; and return to negative after that, implying
for small delay τ < τ1∗ , the positive equilibrium point E ∗ (τ ) is stable and loses the
stability when τ ∈ (τ1∗ , τ2∗ ) and back to stable for large time delay, i.e τ > τ2∗ , before
it disappears at τ = 45.02. These tell us the occurrence of stability switch, which
is confirmed in Fig. 3. Biologically, when the maturation process is too short, both
of the prey and the predator move to a certain level with constant concentration;
while the maturation time increases, all the species oscillate regularly. Nevertheless,
when the predator is mature enough, it behaviors less active which gives a chance
to prey to increase. Therefore, the state of coexistence of the predator and the prey
is stable, rather than oscillatory.
We notice that the chosen function h(P ) in (31) is not a increasing function,
from Remark 4 it is possible to have other positive equilibrium point beside E ∗ (τ )
but we failed to find it although we have tried different choices of the functions and
parameters.
SnΨ (τ)
100

S0Ψ (τ)

50

S1Ψ (τ)
τ1* τ2*
τ
5 10 15 20 25 30
S2Ψ (τ)
- 50

- 100

Figure 2. Functions S0Ψ (τ ), S1Ψ (τ ) and S2Ψ (τ ) for τ ∈ (0, 30)

corresponding to the choice in (31).

4 4 4 4
P2(t) P2(t) P2(t) P2(t)
Z2(t) Z2(t) Z2(t) Z2(t)
3 3 3 3
P2(t) & Z2(t)

P2(t) & Z2(t)

P2(t) & Z2(t)

P2(t) & Z2(t)

2 2 2 2

1 1 1 1

0 0 0 0
0 50 100 150 200 250 300 350 0 100 200 300 400 500 600 700 800 0 500 1000 1500 2000 0 500 1000 1500
time t time t time t time t

(a) τ = 0.1. (b) τ = 7. (c) τ = 20. (d) τ = 25.

Figure 3. Time series P2 (t) and Z2 (t) of system (19) with pa-
rameters and functions given in (30), (31), g12 (P1 ) = g21 (P2 ) = 0.1
and f12 (Z1 ) = f21 (Z2 ) = 0.6.

Next, we discuss how the variation of prey and predator migration rates effect
the dynamical system. Since the system is symmetric, let f12 (Z1 ) = f21 (Z2 ) = m,
g12 (P1 ) = g21 (P2 ) = s. In general, the stability is corresponded to the roots in
Φ(λ, τ ) = 0 given in (25), where Φ depends on both migration rates m, s, and
Ψ(λ, τ ) = 0 given in (26) which depends only on m. First, we fixed the time
delay at τ = 0.1. With the variation of m, when s = 0.1, we can plot Re(λ) v.s
m numerically from Φ(λ, 0.1) = Ψ(λ, 0.1) = 0, see Fig. 4a. Similarly, with the
756 ISAM AL-DARABSAH, XIANHUA TANG AND YUAN YUAN

variation of s, we plot Re(λ) v.s s in Fig. 4b when m = 0.6. We can observe, from
Figs 4a and 4b, that Re(λ) < 0, so the variation of m or s does not change the
stability of E ∗ (0.1). This gives a good agreement with Fig. 3a. Parallely, with the
large τ = 7, we can see, from Fig. 4c, that Re(λ) ≥ 0, implying the instability,
which is consistent with those given in Fig. 3b.
Re(λ)
Re(λ) 0.05
Re(λ)
m
● 5 10 ● ● ● ● ● ●15
● ● ● ● ● ● ● ● ● ●20
●
●
●
●●
● ●●●●●●●●●●●●●●● ● ● ● ● ● ● ● ● ● ● ● ● ●
●●
●
●
●
● s 0.4 ●●●
●
● 5 10 15 20 ●●●●●●●●
●●●
●●●
●●
●●●●●
●●
-0.1 -0.05● ●●●
●●●●●●●●●●●●●
●●●● ●●●●●●●● ●●
●
● ●●●●●●●● ● ● ● ● ● ●
●● ●●
●●●
●● ●
● 0.2 ●
●● ● ●●●●
●
-0.10 ● ● ● ●●●●
● ●
● ●
● ●
-0.2 ● ●●●●●● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ● ● ● ●
● ●● ●●
● ●
● ●● ●
●● ● ●
-0.15 ●●●
●●
●
● ● ●●●● m
5 10 15 20
● ●
● ●●●●
●●●●●●●●●●●●●●●
-0.3 ● ●●●● ●●●●●●●●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
-0.20 ● ●● ●
●
●●●●
●●
●
● ●
● ●
●
●
-0.2
●
-0.25 ●

-0.4

(b) Re(λ) in Φ(λ, τ ) = 0

(a) Re(λ) v.s m for m ∈ (c) Re(λ) v.s m for m ∈
v.s s for s ∈ [0, 20] with m =
[0, 20] with s = 0.1, τ = 0.1. [0, 20] with s = 0.1, τ = 7.
0.6, τ = 0.1.

Figure 4. Re(λ) v.s m and s for τ = 0.1, 7.

In the second example, we consider a non-symmetric system with constant and

non-constant migration rates.
Example 2. A non-symmetric system. Choose
N N 2(1 − e−5P ) 2P 2
µ1 (N ) = , µ2 (N ) = , h1 (P ) = , h2 (P ) =
T 1+N P 5P 3 + 1
1 (Z) = 0.01, 2 (Z) = 0.1Z + 0.05, T = 10. (32)

First, we take a constant migration rates f12 (Z1 ) = f21 (Z2 ) = m and g12 (P1 ) =
g21 (P2 ) = s. With fixed m = 0.6 and s = 0.8, in Fig. 5, we can observe interesting
dynamical behavior as τ varies. When the maturation time is τ = 5, a stable
steady state exists (Fig. 5a); with the increasing of the maturation time, the steady
state loses the stability and the system becomes oscillatory (Fig. 5b); With further
increasing of τ , it is interested to observe the occurrence of period-doubling (Fig.
5c), and then regain the stability (Fig. 5d) when τ is even larger.
To see the influence of the prey migration rate m, we fix s = 0.8 and τ = 5. Fig.
6 shows that the increasing of m can destabilize the steady state. Corresponding to
the remarkable behavior in Fig. 5c, where m = 0.6, s = 0.8, τ = 20, we keep all the
parameters and delay the same, just slightly change the value of m around m = 0.6,

6 6 6 6
P2(t) P2(t) P2(t) P2(t)
5 Z2(t) 5 Z2(t) 5 Z2(t) 5 Z2(t)

4 4 4 4
P2(t) & Z2(t)

P2(t) & Z2(t)

P2(t) & Z2(t)

P2(t) & Z2(t)

3 3 3 3

2 2 2 2

1 1 1 1

0 0 0 0
0 100 200 300 400 500 0 200 400 600 800 1000 1200 0 500 1000 1500 2000 0 500 1000 1500
time t time t time t time t

(a) τ = 5. (b) τ = 18. (c) τ = 20. (d) τ = 21.

Figure 5. Time series P2 (t) and Z2 (t) of system (2) with param-
eters and functions given in (30) and (32).
 A PREY-PREDATOR MODEL WITH MIGRATIONS AND DELAYS 757

6 6 6
P2(t) P2(t)
5 5 Z2(t) 5 Z2(t)

4 4 4
P2(t) & Z2(t)

P2(t) & Z2(t)

P2(t) & Z2(t)

P2(t)
3 Z2(t) 3 3

2 2 2

1 1 1

0 0 0
0 50 100 150 200 250 300 0 100 200 300 400 500 0 100 200 300 400 500 600 700 800
time t time t time t

(a) m = 0.5. (b) m = 1. (c) m = 1.5.

Figure 6. The affection of the prey migration rate when τ = 5, s = 0.8.

6 6 6
P2(t) P2(t) P2(t)
5 Z2(t) 5 Z2(t) 5 Z2(t)

4 4 4
P2(t) & Z2(t)

P2(t) & Z2(t)

P2(t) & Z2(t)

3 3 3

2 2 2

1 1 1

0 0 0
0 50 100 150 200 250 300 350 400 0 500 1000 1500 2000 0 200 400 600 800 1000
time t time t time t

(a) m = 0.5. (b) m = 0.6. (c) m = 0.7.

1.6 1.6 1.6

1.4 1.4 1.4

1.2
1.2 1.2
1
Z2(t)

Z2(t)
Z2(t)

1 1
0.8
0.8 0.8
0.6

0.4 0.6 0.6

0.2 0.4 0.4

0 1 2 3 4 5 6 0 1 2 3 4 5 6 0 1 2 3 4 5 6
P2(t) P2(t) P2(t)

(d) m = 0.5. (e) m = 0.6. (f) m = 0.7.

Figure 7. The affection of the prey migration rate when τ = 20, s = 0.8.

Fig. 7 shows the sensitivity of the dynamical behavior with the variation of the prey
migration rate m. It is observed that the double-periodic behavior disappears and
one period orbit exists at m = 0.7 and a stable steady state appears at m = 0.5.
More clearly, Figs. 7a-7c show the solution curves with different m, and Figs. 7d-7f
depict the associated phase portraits. Same in the following Figs. 8 and 10.
Similarly, we fix m = 0.6 and vary the value of s. When s = 0.6, there exists
periodic oscillation. Increasing s from 0.6 to 0.7 or 0.8, another oscillatory behavior
appears to form a periodic-doubling. It is very interesting to observe that the
stability switch occurs with the further increasing of s, see Fig. 8.
If we choose the migration rates in (31) as non-constant functions,
0.1
g12 (P ) = g21 (P ) = , f12 (Z) = f21 (Z) = ϑZ + 0.2; (33)
1+P
where ϑ is a positive real number. First, we fix ϑ = 0.5. Fig. 9 shows the occurrence
of stability switch and a periodic-doubling when the time delay varies.
Then we pick the moment shown in Fig. 9c, change the value of ϑ slightly around
ϑ = 0.5. It is surprising that the periodic-doubling is broken when we increase or
decrease ϑ, see Fig. 10, implying that the system may have rich dynamics with the
change of migration rate.
In the last example, we consider a non-symmetric system having three patches
and with constant migration rates. We compare this system with the model without
migration rates which is studied in [3].
758 ISAM AL-DARABSAH, XIANHUA TANG AND YUAN YUAN

6 6
6 P2(t) P2(t)
P2(t) 5 5
Z2(t) Z2(t)
Z2(t)
4

P2(t) & Z2(t)

P2(t) & Z2(t)

4

P2(t) & Z2(t)

4
3 3

2 2 2

1 1

0 0 0
0 500 1000 1500 0 500 1000 1500 0 500 1000 1500
time t time t time t

1.6 1.6

1.5 1.4 1.4

1.2 1.2
Z2(t)

Z2(t)

Z2(t)
1 1
1
0.8 0.8

0.6 0.6
0.5
0 1 2 3 4 5 6 0.4 0.4
P2(t) 0 1 2 3 4 5 6 0 1 2 3 4 5 6
P2(t) P2(t)

(a) s = 0.6. (b) s = 0.7. (c) s = 0.9.

6 6 6
P2(t) P2(t) P2(t)
5 5 5
Z2(t) Z2(t) Z2(t)
P2(t) & Z2(t)

P2(t) & Z2(t)

P2(t) & Z2(t)

4 4 4

3 3 3

2 2 2

1 1 1

0 0 0
0 500 1000 1500 0 500 1000 1500 0 500 1000 1500
time t time t time t

1.6 1.6

1.4 1.4 1.6

1.4
1.2 1.2
1.2
Z2(t)

Z2(t)

Z2(t)
1 1
1
0.8 0.8
0.8
0.6 0.6 0.6
0.4 0.4 0.4
0 1 2 3 4 5 6 0 1 2 3 4 5 6 0 1 2 3 4 5 6
P2(t) P2(t) P2(t)

(d) s = 1.1. (e) s = 1.2. (f) s = 5.

Figure 8. The affection of the predator migration rate when τ =

20, m = 0.6.

7 P2(t) 7 P2(t) 7 P2(t) 7 P2(t)

6 Z2(t) 6 Z2(t) 6 Z2(t) 6 Z2(t)

5 5 5 5
P2(t) & Z2(t)

P2(t) & Z2(t)

P2(t) & Z2(t)

P2(t) & Z2(t)

4 4 4 4

3 3 3 3

2 2 2 2

1 1 1 1

0 0 0 0
0 50 100 150 200 250 300 350 400 0 200 400 600 800 1000 0 500 1000 1500 2000 2500 3000 0 500 1000 1500 2000 2500 3000 3500 4000
time t time t time t time t

(a) τ = 5. (b) τ = 19. (c) τ = 32. (d) τ = 33.

Figure 9. Time series P2 (t) and Z2 (t) of system (2) with param-
eters and functions given in (30), (32) and (33).

Example 3. When n = 3, choose

N 4N P
µ1 (N ) = , µ2 (N ) = µ3 (N ) = , h1 (P ) = 2 ,
15 + N 10 + N P + 0.1
1
h2 (P ) = h3 (P ) = , 1 (Z) = 0.05 + 0.31Z, 2 (Z) = 3 (Z) = 0.15,
P +1
gij (Pi ) = s, fij (Zi ) = m, i, j = 1, 2, 3 i 6= j, T = 9. (34)

If there is no migration, i.e. m = s = 0, when the maturation time is very small,

a doubly periodic solution exists [3], see Figs. 11. While when m and s are very
small, then the doubly periodic orbit is disappeared, although a periodic solution
still exists, see Fig. 12a. However, when either m or s is relative large, the periodic
solution is vanished and the system becomes stable, see Fig. 12b-12d. Therefore
the dynamical behavior is sensitive to prey and predator migration rates. Seems
the active migration for prey or predator can stabilize the ecosystem.

5. Conclusion. In this paper, we propose a general model with multiple paral-

lel food chains through the stage structured maturation time delay, and consider
 A PREY-PREDATOR MODEL WITH MIGRATIONS AND DELAYS 759

7 P2(t) 7 P2(t) 7 P2(t)

6 Z2(t) 6 Z2(t) 6 Z2(t)

5 5 5
P2(t) & Z2(t)

P2(t) & Z2(t)

P2(t) & Z2(t)

4 4 4

3 3 3

2 2 2

1 1 1

0 0 0
0 100 200 300 400 500 600 700 800 0 500 1000 1500 2000 2500 3000 0 500 1000 1500
time t time t time t

(a) ϑ = 0.4. (b) ϑ = 0.5. (c) ϑ = 0.6.

1

0.8 0.8
0.9
0.75
0.75

0.8 0.7
0.7
Z2(t)

Z2(t)

Z2(t)
0.65
0.65
0.7 0.6
0.6
0.55
0.6
0.55 0.5

0.5 0.5 0.45

0 1 2 3 4 5 6 7 1 2 3 4 5 6 7 1 2 3 4 5 6 7
P2(t) P2(t) P2(t)

(d) ϑ = 0.4. (e) ϑ = 0.5. (f) ϑ = 0.6.

Figure 10. The affection of the prey migration rate when τ = 32.

Z2,3 (t )

P 2,3 (t )
0 1 2 3 4

Figure 11. Phase portrait of (16) with parameters and functions

given in (30) and (34), m = s = 0.
Z2,3 (t ) Z2,3 (t ) Z2,3 (t ) Z2,3 (t )
2.5
1.5

2.5 1.4
1.5

2.0
1.3

2.0
1.2
1.0
1.5
1.1

1.5
1.0
0.5
1.0
0.9
1.0

P 2,3 (t ) P 2,3 (t ) P 2,3 (t ) P 2,3 (t )

0.5 1.0 1.5 0.5 1.0 1.5 2.0 2.5 3.0 3.5 0.2 0.4 0.6 0.8 0.3 0.4 0.5 0.6 0.7

(a) m = 0.1 and s = (b) m = 0.1 and s = (c) m = 2.8 and s = (d) m = 2.8 and s =
0.1. 0.9. 0.1. 0.9

Figure 12. Phase portrait of (16) with parameters and functions

given in (30) and (34).

predator and prey migrations among all patches. We have carried out mathematical
analysis to discuss the existence of the steady states and their stabilities. Taking
the system with two patches as an example, we have proved the existence and
uniqueness of two boundary equilibrium point E0 and the predator-free equilibrium
point Ẽ. We have shown that E0 is always unstable, provided conditions for the
uniform persistence, obtained sufficient condition for the global attractivity of Ẽ
760 ISAM AL-DARABSAH, XIANHUA TANG AND YUAN YUAN

and discussed its local stability with the same time delays. When both patches are
symmetric with constant migration rates, we have discussed the existences and local
stability of positive equilibrium point; obtained that the time delay can not only
destroy the existence, but also destabilize the positive equilibrium even it exists. To
complement the analytical results, we have used numerical techniques to examine
how the maturation time of predator, the prey and predator migrations affect on the
dynamical behavior. The numerical simulations show the rich dynamics including
stability switch and periodic-doubling.

Acknowledgments. The authors would like to thank the two anonymous referees
for their careful reading and helpful suggestions.

REFERENCES

[1] K. M. Abadir and J. R. Magnus, Matrix Algebra, Cambridge University Press, 2005.
[2] A. El Abdllaoui, P. Auger, B. W. Kooi, R. B. Parra and R. Mchich, Effects of density-
dependent migrations on stability of a two-patch predator-prey model, Mathematical Bio-
sciences, 210 (2007), 335–354.
[3] I. Al-Darabsah and Y. Yuan, Dynamics of a general stage structured n parallel food chains,
Proceedings of the International Symposium on Mathematical and Computational Biology,
World Scientific, (2014), 63–84.
[4] V. Andersen, A. Gubanova, P. Nival and T. Ruellet, Zooplankton community during the
transition from spring bloom to oligotrophy in the open NW mediterranean and effects of
wind events. 2. vertical distributions and migrations, Journal of Plankton Research, 23 (2001),
243–261.
[5] R. A. Armstrong, Stable model structures for representing biogeochemical diversity and size
spectra in plankton communities, Journal of Plankton research, 21 (1999), 445–464.
[6] M. Bandyopadhyay and S. Banerjee, A stage-structured prey-predator model with discrete
time delay, Applied Mathematics and Computation, 182 (2006), 1385–1398.
[7] E. Beretta and Y. Kuang, Geometric stability switch criteria in delay differential systems
with delay dependent parameters, SIAM J Math Anal., 33 (2002), 1144–1165.
[8] Y. Cai, M. Banerjee, Y. Kang and W. Wang, Spatiotemporal complexity in a predator-prey
model with weak Allee effects, Mathematical Biosciences and Engineering, 11 (2014), 1247–
1274.
[9] B. Daia, H. Su and D. Hu, Periodic solution of a delayed ratio-dependent predator-prey model
with monotonic functional response and impulse, Nonlinear Analysis, 70 (2009), 126–134.
[10] W. Feng, B. Rock and J. Hinson, On a new model of two-patch predator-prey system with
migration of both Species, J. Applied Analysis and Computation, 1 (2011), 193–203.
[11] D. R. French and J. M. Travis, Density-dependent dispersal in host-parasitoid assemblages,
Oikos, 95 (2001), 125–135.
[12] S. A. Gourley and Y. Kuang, A stage structured predator-prey model and its dependence on
maturation delay and death rate, Journal of Mathematical Biology, 49 (2004), 188–200.
[13] A. Hastings and L. Gross, editors, Encyclopedia of Theoretical Ecology, University of Califor-
nia Press, 2012.
[14] C. S. Holling, The functional response of predator to prey density and its role in mimicry and
population regulation, Memoirs of the Entomological Society of Canada, 97 (1965), 5–60.
[15] Y. Huang, F. Chen and L. Zhong, Stability analysis of a prey-predator model with Holling type
III response function incorporating a prey refuge, Applied Mathematics and Computation,
182 (2006), 672–683.
[16] V. S. Ivlev, Experimental Ecology of the Feeding of Fishes, Yale University Press, 1961.
[17] V. Jansen, Theoretical Aspects of Metapopulation Dynamics, Ph.D. thesis Leiden University
The Netherlands, 1994.
[18] Y. Kang, S. K. Sasmal, A. R. Bhowmick and J. Chattopadhyay, Dynamics of a predator-
prey system with prey subject to Allee effects and disease, Mathematical Biosciences and
Engineering, 11 (2014), 877–918.
[19] Y. Kuang, Delay Differential Equation with Application in Population Dynamics, New York:
Academic Press, 1993.
 A PREY-PREDATOR MODEL WITH MIGRATIONS AND DELAYS 761

[20] Y. Kuang and Y. Takeuchi, Predator-prey dynamics in models of prey dispersal in two-patch
environments, Mathematical Biosciences, 120 (1994), 77–98.
[21] M. Y. Li and H. Shu, Impact of intracellular delays and target-cell dynamics on in vivo viral
infections, SIAM J. Appl. Math., 70 (2010), 2434–2448.
[22] R. Mchich, P. Auger and J. C. Poggiale, Effect of predator density dependent dispersal of
prey on stability of a predator-prey system, Mathematical Biosciences, 206 (2007), 343–356.
[23] J. D. Murray, Mathematical Biology II: Spatial Models and Biomedical Applications, Third
edition. Interdisciplinary Applied Mathematics, 18. Springer-Verlag, New York, 2003.
[24] K. Olli, Diel vertical migration of phytoplankton and heterotrophic flagellates in the Gulf of
Riga, Journal of Marine Systems, 23 (1999), 145–163.
[25] I. L. Pepper, C. P. Gerba and T. J. Gentry, editors, Environmental Microbiology, Academic
Press, 2014.
[26] P. Pillai, A. Gonzalez and M. Loreau, Evolution of dispersal in a predator-prey metacommu-
nity, The American Naturalist, 179 (2012), 204–216.
[27] T. S. Shores, Applied Linear Algebra and Matrix Analysis, Springer, 2007.
[28] J. R. Silvester, Determinants of block matrices, Maths Gazette, 84 (2000), 460–467.
[29] H. L. Smith, Monotone Dynamical Systems: An Introduction to the Theory of Competitive
and Cooperative Systems, American Mathematical Society, 1995.
[30] H. L. Smith and X. Q. Zhao, Robust persistence for semidynamical systems, Nonlinear Anal-
ysis, 47 (2001), 6169–6179.
[31] P. D. Srinivasu and I. L. Gayatri, Influence of prey reserve capacity on predator-prey dynamics,
Ecological Modelling, 181 (2005), 191–202.
[32] X. H. Tang and X. Zou, Global attractivity in a predator-prey system with pure delays,
Proceedings of the Edinburgh Mathematical Society, 51 (2008), 495–508.
[33] D. Tilman and P. Kareiva, Spatial Ecology, Princenton University Press, 1997.
[34] F. Wang, Y. Kuang, C. Ding and S. Zhang, Stability and bifurcation of a stage-structured
predator-prey model with both discrete and distributed delays, Chaos, Solitons & Fractals,
46 (2013), 19–27.
[35] Z. Wang and X. Q. Zhao, Global dynamics of a time-delayed dengue transmission model,
Canadian Appl. Math. Quarterly, 20 (2012), 89–113.
[36] R. Xu and L. Chen, Persistence and stability for a two-species ratio-dependent predator-
prey system with time delay in a two-patch environment, Computers & Mathematics with
Applications, 40 (2000), 577–588.
[37] C. Xu, X. Tang, M. Liao and X. He, Bifurcation analysis in a delayed Lokta-Volterra predator-
prey model with two delays, Nonlinear Dynamics, 66 (2011), 169–183.
[38] Y. Yuan and J. Bélair, Threshold dynamics in an SEIRS model with latency and temporary
immunity, Journal of Mathematical Biology, 69 (2014), 875–904.

Received January 2015; revised September 2015.

E-mail address: imaad6@mun.ca
E-mail address: tangxh@mail.csu.edu.cn
E-mail address: yyuan@mun.ca