Nonlinear Analysis: Real World Applications 16 (2014) 103–119

Contents lists available at ScienceDirect

Nonlinear Analysis: Real World Applications

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

Dynamical complexity induced by Allee effect in a

predator–prey model
Weiming Wang a,∗ , Ya-nuo Zhu a , Yongli Cai b , Wenjuan Wang c
a
College of Mathematics and Information Science, Wenzhou University, Wenzhou, 325035, PR China
b
School of Mathematics and Computational Science, Sun Yat-sen University, Guangzhou, 510275, PR China
c
Department of Biostatistics and Computational Biology, University of Rochester Medical Center, 601 Elmwood Ave, Box 630, Rochester,
NY 14642, USA

article info abstract

Article history: In this paper, we investigate the complex dynamics induced by Allee effect in a
Received 6 February 2013 predator–prey model. For the non-spatial model, Allee effect remains the boundedness of
Accepted 21 September 2013 positive solutions, and it also induces the model to exhibit one or two positive equilibria.
Especially, in the case with strong Allee effect, the model is bistable. For the spatial model,
without Allee effect, there is the nonexistence of diffusion-driven instability. And in the
case with Allee effect, the positive equilibrium can be unstable under certain conditions.
This instability is induced by Allee effect and diffusion together. Furthermore, via numerical
simulations, the model dynamics exhibits both Allee effect and diffusion controlled pattern
formation growth to holes, stripe–hole mixtures, stripes, stripe–spot mixtures, and spots
replication. That is to say, the dynamics of the model with Allee effect is not simple, but
rich and complex.
© 2013 Elsevier Ltd. All rights reserved.

1. Introduction

The understanding of patterns and mechanisms of spatial dispersal of interacting species is an issue of significant current
interest in conservation biology and ecology, and biochemical reactions [1]. Interactions among species are multiform
in communities [2,3]. Among these, a typical type of interactions is the one between a pair of predator and prey. The
predator–prey model plays a major role in the studies of biological invasion of foreign species, epidemic spreading,
extinction/spread of flame balls in combustion or autocatalytic chemical resection.
A prototypical predator–prey interaction model is of form [4,5]:
du


 = a(u) − f (u)g (v),
dt (1)
 dv = σ f (u)g (v) − z (v),

dt
where u(t ) and v(t ) are the densities of the prey and the predator at time t > 0, respectively.
In model (1), a(u) is the growth rate of the prey in the absence of predation, which can be given by
 
K −u
a(u) = α u min 1, , α > 0, ε ≥ 0, K > 0,
K −ε

∗ Corresponding author. Tel.: +86 0577 86689222.

E-mail address: [email protected] (W. Wang).

1468-1218/$ – see front matter © 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.nonrwa.2013.09.010
104 W. Wang et al. / Nonlinear Analysis: Real World Applications 16 (2014) 103–119

where, the parameter α stands for the specific growth

 rate  of prey u, and K for the carrying capacity of the prey. If ε = 0, it
reduces to the traditional logistic form a(u) = α u 1 − Ku .
The product f (u)g (v) gives the rate at which prey is consumed. The prey consumed per predator, f (u)g (v)/v , was termed
the functional response by Solomon [6]. These functions can be defined in different ways. In Refs. [4,5], Harrison defined
ωu
f (u) = ,
φu + 1
which is known as a Holling Type II functional response, and ω is the maximum rate of prey consumption by a single predator
and φ is the half-saturation constant, the level of prey at which half the maximum consumption rate occurs. In this paper,
following Lotka [7], we adopt

f (u) = cu, (2)

which is a linear functional response without saturation, where c > 0 denotes the capture rate [8]. And following Harrison
[4,5], we set:

v
g (v) = , (3)
mv + 1
where m > 0 represents a reduction in the predation rate at high predator densities due to mutual interference among
the predators while searching for food. If m = 0, g (v) reduces to the traditional form g (v) = v , indicating that the
prey consumed is proportional to the number of predators, but there is evidence that there is mutual interference among
predators searching for food, resulting in decreased consumption per predator as the predator density increases. This
requires that g (v)/v be decreasing, which can be modeled by using β > 0 [4,5]. Freedman [9] used g (v) = v m , 0 < m < 1
to model predator mutual feeding interference, but Eq. (3) has the advantage that v/(1 + mv) ≈ v for small v , whereas
v m /v → ∞ as v → 0 [4,5].
The proportionality constant σ is the rate of prey consumption. And the function z (v) is given by

z (v) = γ v + lv 2 , γ > 0, l ≥ 0,
where γ denotes the natural death rate of the predator, and l > 0 can be used to model predator intraspecific competition
that is not direct competition for food, such as some type of territoriality [4]. In this paper, we will discuss the case of l = 0,
which is used in a much more traditional case.
Based on the above discussions, we can obtain the following model:

cuv

du  u
 = αu 1 − − ,
mv + 1

dt K

(4)
dv cσ u
 
= v −γ + ,


mv + 1

dt

where α, K , γ have the same definitions as the above.

Let β = Kα , s = c σ , model (4) leads to the following dimensionless equation:

cuv

du
 = u (α − β u) − ,
mv + 1

dt

(5)
dv
 
su
= v −γ + ,


mv + 1

dt

where s denotes the conversion rate.

On the other hand, Allee effect in the population growth has been studied extensively. Allee effect, named after ecologist
Warder Clyde Allee [10], is a phenomenon in biology characterized by a positive correlation between the population size
or density and the mean individual fitness (oftentimes measured as per capita population growth rate) of a population or
species [11], and may occur under several mechanisms, such as difficulties in finding mates when the population density is
low, social dysfunction at small population sizes and increased predation risk due to failing flocking or schooling behavior,
see Refs. [12–14]. In an ecological point of view, Allee effect has been divided into strong and weak cases. The strong Allee
effect introduces a population threshold, and the population must surpass this threshold to grow. In contrast, a population
with a weak Allee effect does not have a threshold. It has been attracting much more attention recently owing to its strong
potential impact on the population dynamics of many plant and animal species [15]. Detailed investigations relating to Allee
effect may be found in Refs. [16–25].
Allee effect has been modeled in different ways using various mathematical tools, and in most predation models it has
been considered that Allee effect influences only the prey population. For instance, if u = u(t ) indicates the population size,
 W. Wang et al. / Nonlinear Analysis: Real World Applications 16 (2014) 103–119 105

the most usual continuous growth the equation to express Allee effect is given by:
du
= u(α − β u)(u − q), (6)
dt
which is called multiplicative Allee effect [26], where an Allee limit q is incorporated such that per capita population growth
is negative below q. While u > q, the per capita growth rate is positive. Clearly, if q ≤ 0, Eq. (6) represents weak Allee effect
and if q > 0 represents strong Allee effect. Moreover, the equation
 
du q
= u α − βu − (7)
dt u+b
qu
is called additive Allee effect, which was first deduced in [12] and applied in [27,28]. In Eq. (7), u+b is the term of additive
Allee effect, b, q > 0 are Allee effect constants. If q < bα , it is called weak Allee effect, while q > bα , strong Allee effect.
Based on the above discussions, we can establish the following predation model with additive Allee effect on prey:

cuv
 
du q

 = u α − β u − − , f (u, v),
mv + 1

dt u+b

(8)
dv
 
su
= v −γ + , g (u, v).


mv + 1

dt
In addition, assuming that prey u and predator v move randomly—described as Brownian random motion, we propose a
simple reaction–diffusion model corresponding to model (8) as follows:

∂u cuv
 
q

 = u α − β u − − + d1 1u,
∂t mv + 1

u+b

(9)
∂v
 
su
= v −γ + + d2 1v,


∂t mv + 1


with the positive initial conditions:

u(x, y, 0) > 0, v(x, y, 0) > 0, (x, y) ∈ Ω = (0, L) × (0, L), (10)
and the zero-flux boundary conditions:
∂u ∂v
= = 0, (x, y) ∈ ∂ Ω , (11)
∂ν ∂ν
where Ω is a bounded open domain in R2+ with boundary ∂ Ω , and ∆ = ∂∂x2 + ∂∂y2 is the Laplacian operator in the two-
2 2

dimensional space. d1 > 0 and d2 > 0 are the diffusion coefficients of prey u and predator v , respectively. ν is the outward
unit normal vector on ∂ Ω , and zero-flux conditions reflect the situation where the population cannot move across the
boundary of the domain.
For the sake of simplicity, we call model (8) as the non-spatial model, and model (9) the spatial model.
The main purpose of this paper is to present Allee effect on the dynamical behavior of non-spatial model (8) and spatial
model (9).
The rest of the paper are structured in the following way. In the next section, we give some mathematical analysis of the
non-spatial model, such as boundedness, existence and stability of the equilibria. In Section 3, we discuss the stability of the
spatial model with/without Allee effect and derive the conditions for the occurrence of Allee–diffusion-driven instability
of the case with Allee effect, and illustrate typical patterns via numerical simulations. Finally, conclusions and remarks are
presented in Section 4.

2. Mathematical analysis of the non-spatial model

In this section, we restrict ourselves to the stability analysis of non-spatial model (8), in the absence of diffusion, where
only the interaction part of the model system is taken into account. We find the non-negative equilibrium states of the
model system and discuss their stability properties with respect to variation of several parameters. The first one concerns
the boundedness of positive solutions.

2.1. Boundedness of positive solutions

It is worthy to note that the solutions of model (8) always exist and stay positive. In fact, from the first equation of (8),
we can get:
du
≤ u(α − β u).
dt
106 W. Wang et al. / Nonlinear Analysis: Real World Applications 16 (2014) 103–119

A standard comparison theorem shows that

α
lim sup u(t ) ≤ .
t →∞ β
As a result, for any ε > 0, there exists a T > 0, such that u(t ) ≤ βα + ε for t > T . Then from the second equation of (8),
we obtain:
 
s βα + ε

dv
≤ v −γ + 
dt mv + 1
 
α

−γ (mv + 1) + s β
+ε
= v 
mv + 1

α
  
≤ v −γ (mv + 1) + s +ε .
β
Since ε is arbitrary, if sα ≥ βγ , we have:
sα − βγ
lim sup v(t ) ≤ .
t →∞ mβγ

Further, we can get the following theorem.

Theorem 2.1. All non-negative solutions of model (8) that start in R2+ are uniformly bounded.

Proof. Define the function

W (t ) = u(t ) + v(t ),

and the time derivative of W along the solutions of model (8) is:

dv cuv
   
dW du q su
= + = u α − βu − − + v −r + .
dt dt dt u+b mv + 1 mv + 1

For each η > 0, and 0 ≤ u ≤ βα , the following inequality holds:

sα
 
dW
+ ηW ≤ u(α − β u) + v −γ + + η(u + v)
dt mβv
sα
≤ u(α + η − β u) + (η − γ )v +
mβ
4sα + m(α + η)2
≤ + (η − γ )v.
4mβ

Let 0 < η < γ , then:

dW 4sα + m(α + η)2

+ ηW ≤ , Ψ.
dt 4mβ

By using the theory of differential inequality [29], for all t ≥ T ≥ 0, we have:

0 ≤ W (t ) ≤ Ψ − (Ψ − W (T ))e−(t −T ) .

Hence,

lim sup(u(t ) + v(t )) ≤ Ψ .

t →∞

This completes the proof. 

Remark. It needs to note that in the case of q = 0, i.e., model (8) without Allee effect, or model (5), the boundedness of the
solution is similar to the results in Theorem 2.1. That is to say, the solutions of the model without Allee effect always exist
and stay positive, and all the non-negative solutions that start in R2+ are uniformly bounded.
 W. Wang et al. / Nonlinear Analysis: Real World Applications 16 (2014) 103–119 107

2.2. Existence of positive equilibria

In this subsection, we mainly focus on the existence of positive equilibria of model (8).
Any positive equilibrium E = (u, v) of model (8) satisfies

q cv

α − β u −
 − = 0,
u+b mv + 1
su (12)
−γ +
 = 0,
mv + 1
which yields

msβ u3 + (cs + bmsβ − msα)u2 + (bcs + msq − bmsα − c γ )u − bc γ = 0. (13)

Let u be a real positive root of the following cubic equation:

a0 x3 + 3a1 x2 + 3a2 x + a3 = 0, (14)

where a0 = msβ, 3a1 = cs + bmsβ − msα , 3a2 = bcs + msq − bmsα − c γ , a3 = −bc γ .
Set z = a0 x + a1 , then Eq. (14) can be rewritten as:

h(z ) = z 3 + 3b1 z + b2 = 0, (15)

where b1 = a0 a2 − b2 = a21 ,
− 3a0 a1 a2 + a20 a3 2a31 .
Next, we will discuss the existence of the positive root of Eq. (15).

Lemma 2.2. (a) If b2 < 0, Eq. (15) has one positive root.
(b) Suppose that b2 > 0 and b1 < 0 hold.
(b1) If b22 + 4b31 < 0, Eq. (15) has two positive roots.
(b2) If b22 + 4b31 = 0, Eq. (15) has a positive root of multiplicity two;
(c) If b2 = 0 and b1 < 0, Eq. (15) has one positive root.

Proof. (a) When b2 < 0, it can be easily shown that the following holds.
If b1 ≥ 0, h(z ) is strictly increasing and continuous in [0, +∞) since h′ (z ) = 3(z 2 + b1 ) ≥ 0, which yields
h(z ) ≥ h(0) = b2 . Consequently, h(z ) has a positive root; if b1 ≤ 0, it is easy to see that Eq. (15) has one positive root.
Hence, Eq. (15) has a positive root when b2 < 0.
(b) If b2 > 0, then b1 < 0, otherwise, h(z ) = z 3 + 3b1 z + b2 may not be equal to zero.
  2
3
−4b2 +4 4b31 +b22 −4b1

z13 +4b2
(b1) If b22 + 4b31 < 0, Eq. (15) has two positive roots, z1 =   and z2 = −
z1
2
+ √
2 z1
.
3
2 −4b2 +4 4b31 +b22

(b2) If b22 + 4b31 = 0, Eq. (15) has a positive root of multiplicity two
 2

3
−4b2 + 4 4b31 + b22 − 4b1
z1 =   .
3
2 −4b2 + 4 4b31 + b22

(c) When b2 = 0, we know b1 < 0, otherwise, h(z ) = z 3 + 3b1 z + b2 may not be equal to zero. And Eq. (15) has one
positive root. 

In addition, we get:
1 1
b1 = a0 a2 − a21 = msβ(msq + bcs − bmsα − c γ ) − (cs + bmsβ − msα)2 ,
3 9
b2 = a20 a3 − 3a0 a1 a2 + 2a31
1 1
= m2 s3 qβ(mα − bmβ − c ) − ms2 β(c + bmβ − mα)(bcs − bmsα − c γ )
3 3
2
+ (cs + bmsβ − msα)3 − bcm2 s2 β 2 γ .
27
108 W. Wang et al. / Nonlinear Analysis: Real World Applications 16 (2014) 103–119

From b1 = 0, b2 = 0, we can determine that q = f1 (b) and q = f2 (b), respectively, where

b2 m2 sβ 2 − bcmsβ + bm2 sβα + 3cmβγ + s(mα − c )2
f1 (b) , ,
3sm2 β
9ms2 β(c + bmβ − mα)(bcs − bmsα − c γ ) − 2(cs + bmsβ − msα)3 + 27bcm2 s2 β 2 γ
f2 (b) , .
9m2 s3 β(c + bmβ − mα)
From Lemma 2.2, we can obtain the existence of positive equilibria of model (8) with weak (i.e., 0 < q < bα ) and strong
(i.e., q > bα ) Allee effects, respectively.

Theorem 2.3 (Existence of Positive Equilibria of Model (8) with Weak Allee Effect).
(a) If either of the following inequalities holds:
mα − bmβ − c > 0, 0 < q < min{bα, f2 (b)};
mα − bmβ − c < 0, max{0, f2 (b)} < q < bα,
 
model (8) has a unique positive equilibrium Ew = (uw , vw ) =
z1 −a1
a0
, s(az10 −
mγ
a1 )
− 1
m
.
(b) If either of the following inequalities holds:
mα − bmβ − c > 0, max{0, f2 (b)} < q < min{bα, f1 (b)};
mα − bmβ − c < 0, 0 < q < min{bα, f1 (b), f2 (b)},
then:    
s(z −a )
(b1) if b22 + 4b31 < 0, model (8) has two positive equilibria Ew3 = 1a 1 , a1 mγ1 −
z −a
0 0
1
m
and Ew4 =
z2 −a1
a0
, s(az20 −
mγ
a1 )
− 1
m
;
√ √ 
(b2) if b22 + 4b31 = 0, model (8) has a unique positive equilibrium Êw = (ûw , v̂w ) = |b1 |, s |mb1γ|−γ .
(c) If 0 < q < min{bα, f1 (b)} and b2 = 0 hold, model (8) has a unique positive equilibrium Ẽw = ũw , ṽw = −3b1 ,
  √
√
s −3b1 −γ

mγ
.

Theorem 2.4 (Existence of Positive Equilibria of Model (8) with Strong Allee Effect).
(a) If either of the following inequalities holds:
mα − bmβ − c > 0, bα < q < f2 (b);
mα − bmβ − c < 0, max{f2 (b), bα} < q,
 
model (8) has a unique positive equilibrium Es = (us , vs ) =
z1 −a1
a0
, s(az10 −
mγ
a1 )
− 1
m
.
(b) If either of the following inequalities holds:
mα − bmβ − c > 0, max{f2 (b), bα} < q < f1 (b);
mα − bmβ − c < 0, bα < q < min{f1 (b), f2 (b)},
 
(b1) if b22 + 4b31 < 0, model (8) has two positive equilibria Es3 = (us3 , vs3 ) =
z1 −a1
a0
, s(az10 −
mγ
a1 )
− 1
m
and Es4 = (us4 , vs4 ) =
 
s(z2 −a1 )
z2 −a1
a0
, a0 mγ
− 1
m
;
√ √ 
(b2) if b22 + 4b31 = 0, model (8) has a unique positive equilibrium Ês = (ûs , v̂s ) =
|b1 |, s |mb1γ|−γ ;
√ √ 
(c) If bα < q < f1 (b) and b2 = 0 hold, model (8) has a unique positive equilibrium Ẽs = (ũs , ṽs ) = −3b1 , s −m3bγ1 −γ .

Remark. It is worthy to note that the conditions of the existence of the positive equilibria in model (8) are very complex,
due to the Allee effect. In fact, in the case of model (8) without Allee effect (i.e., q = 0), or model (5), there are two boundary
equilibria E0 = (0, 0), E1 = ( βα , 0) and a unique positive equilibrium E ∗ = (u∗ , v ∗ ) with sα > βγ , where

msα − cs + s2 (mα − c )2 + 4cmsβγ su∗ − γ


∗
u = , v∗ = .
2msβ mγ

2.3. Stability of the equilibria

In this subsection, we are dedicated to focus on the stability of the equilibria of the non-spatial model. For the sake of
studying Allee effect on the stability of the model, we first present the stability analysis in the case without Allee effect.
 W. Wang et al. / Nonlinear Analysis: Real World Applications 16 (2014) 103–119 109

2.3.1. The case without Allee effect

In the case without Allee effect, model (8) reduced
 to 
model (5). From the discussions in Section 2.2, we know that model
α
(5) has two boundary equilibria E0 = (0, 0), E1 = β
,0 and a unique positive equilibrium E ∗ = (u∗ , v ∗ ) with sα > βγ .
The Jacobian matrices around E0 , E1 and E are respectively: ∗

α
 
0
JE0 = ,
0 −γ
 cα 
−α −
β
JE0 =  sα  ,
 
0 −γ +
β
cu∗
 
∗
 −β u −
(mv ∗ + 1)2 
JE ∗ =
sv ∗ msu∗ v ∗ 
 
−
mv ∗ + 1 (mv ∗ + 1)2
and the trace and the determinant of JE ∗ are respectively
msu∗ v ∗
tr(JE ∗ ) = −β u∗ − < 0,
(mv ∗ + 1)2
bmβ(u∗ )2 v ∗ bcu∗ v ∗
det(JE ∗ ) = + > 0.
(mv + 1)
∗ 2 (mv ∗ + 1)3
Then, one can easily obtain the following results.

Theorem 2.5. For model (5),

(a) E0 is a saddle point;
(b) if sα > βγ , E1 is a saddle point, while sα < βγ , a stable node;
(c) if sα > βγ , E ∗ is locally asymptotically stable.

Furthermore, we can get the global stability of E ∗ .

Theorem 2.6. If sα > βγ , the positive equilibrium E ∗ = (u∗ , v ∗ ) of model (5) is globally asymptotically stable.

For simplicity, we omit the proof, and more details can be found in the proof of Theorem 2.9 in the case of q ̸= 0.

2.3.2. The case with weak Allee effect

 (8) with weak Allee effect,i.e., 0 < q < bα . In this
In this subsection, we deal with the stability of equilibria of model
√
α−bβ+ (α−bβ)2 +4β(bα−q)
case, model (8) has two boundary equilibria Ew0 = (0, 0) and Ew1 = 2β
,0 .

The Jacobian matrix of model (8) takes the form

 q qu cv cu 
α − 2β u − + − −
u+b (u + b)2 mv + 1 (mv + 1)2
J = ;
 
sv su msuv
−γ + −
mv + 1 mv + 1 (mv + 1)2
then for the two boundary equilibria Ew0 and Ew1 , we have the following theorem.

Theorem 2.7. Suppose that 0 < q < bα , then

(a) Ew0 is a saddle point;
(b) Ew1 is a stable node point if α > bβ and γ > γ̄ , or a saddle point if α > bβ and γ < γ̄ , where
 
s α − bβ + (α + bβ)2 − 4qβ

γ̄ = . (16)
2β

Proof. (a) The eigenvalues of the Jacobian matrix at Ew0 = (0, 0) are α − > 0 and −γ . Therefore, model (8) is always
q
b
unstable around Ew0 which is, in fact, a saddle point.
110 W. Wang et al. / Nonlinear Analysis: Real World Applications 16 (2014) 103–119
 √  √
α−bβ+ (α−bβ)2 +4β(bα−q) s(α−bβ+ (α+bβ)2 −4qβ)
(b) The eigenvalues of the Jacobian matrix at Ew1 = 2β
,0 are −γ + 2β
and
√  √ 
2 (α+bβ)2 −4qβ α (α+bβ)2 −4qβ+α 2 +β(bα−2q)
− √ . It is noted that if α > bβ then α 2 + β(bα − 2q) > α(α − bβ) > 0.
α+bβ+ (α+bβ) −4qβ
2

Hence, if α > bβ and γ > γ̄ , Ew1 is a stable node point, while Ew1 is a saddle point when α < bβ and γ < γ̄ hold. 

In the following, we shall discuss the stability of the positive equilibria in two situations. One is the case of unique positive
equilibrium Ew = (uw , vw ), the other
 is the case of two positive equilibria Ew3 = (uw3 , vw3 ) and Ew4 = (uw4 , vw4 ).
u,
u−γ
s
E = 
For model (8), let  mγ
be the arbitrary positive equilibrium point in the above two cases. Then the Jacobian
matrix of (8) evaluated in the positive equilibrium 
E takes the form:

cγ 2
 
q
u
−β
u+
(
u + b)2
−
s2
u  .
E = 
J 
su − γ
 (γ − su)γ 

mu s
u
The trace and the determinant of J
E are respectively given by

u2 + (
qs u + b)2 (γ 2 − sγ
u − sβ
u2 )
,
 
E =
tr J
u(
s u + b)2
 
γ (s
u − γ ) (
u + b)2 (c γ + msβ
u2 ) − mqs
u2
,
 
E =
det J
u2 (
ms2 u + b) 2

E is determined by the sign of ϕ(

u) :
 
and the sign of det J

ϕ(
u) = (
u + b)2 (c γ + msβ
u2 ) − mqs
u2 .

Then we present the following results of the stability of the positive equilibria of model (8).

Theorem 2.8. Suppose that (uw + b)2 (c γ + msβ u2w ) − mqsu2w > 0, and define

(γ − suw )γ (uw + b)2

 
q[uw ] = β uw − .
suw uw

 
(a) If q < min{q[uw ] , bα}, the positive equilibrium Ew = uw ,
suw −γ
mγ
is locally asymptotically stable. In addition,

(a1) if (q − q ) < 4 det(JEw ), Ew is a stable focus;

[uw ] 2

(a2) if (q − q[uw ] )2 > 4 det(JEw ), Ew is a stable node point.

 
(b) If q[uw ] < q < bα , the positive equilibrium Ew = uw ,
suw −γ
mγ
is unstable. Moreover,

(b1) if (q − q ) < 4 det(JEw ), Ee is an unstable focus surrounded by a limit cycle;

[uw ] 2

(b2) if (q − q[uw ] )2 > 4 det(JEw ), Ew is an unstable node point and the limit cycle disappears.
(c) If β(uw + b)2 < q < bα and γ < suw , model (8) enters to a Hopf-bifurcation around Ew at q = q[uw ] .

Proof. (a) The trace of the Jacobian matrix at Ew is

qsu2w + (uw + b)2 (γ 2 − sγ uw − sβ u2w )

tr(JEw ) = .
su(uw + b)2
 
When q < q[uw ] and (uw + b)2 (c γ + msβ u2w )− mqsu2w > 0 hold, we have tr(JEw ) < 0, det(JEw ) > 0. Hence Ew = uw ,
suw −γ
mγ
is locally asymptotically stable.  
(b) If q[uw ] < q < bα , we have tr(JEw ) > 0. Hence, Ee = uw ,
suw −γ
mγ
is unstable. More precisely, by the Poincaré–
Bendixson theorem [30], there is at least one limit cycle surrounding this positive equilibrium.
 (c) The eigenvalues
 of the Jacobian JEw satisfy λ2 − tr(JEw )λ + det(JEw ) = 0 at the positive equilibrium point Ew =
uw ,
suw −γ
mγ
. It is clear to see that:
 W. Wang et al. / Nonlinear Analysis: Real World Applications 16 (2014) 103–119 111

(c1) tr(JEw ) |q=q[uw ] = 0;

(c2) det(JEw ) |q=q[uw ] > γ (suw − γ ) (uw + b)2 (c γ + msβ u2w ) − mqsu2w > 0;
 

(c3) The characteristic equation of JEw at Ew becomes λ2 + det(JEw ) |q=q[uw ] = 0, and the eigenvalues will be purely imaginary
and there is no other eigenvalue with negative real parts. We also have
(c4)
(
d tr(JEw ) ) = uw
̸= 0.
dq
q=q[uw ] (uw +b)2


Therefore, from the Poincaré–Andronov–Hopf bifurcation theorem [30], we know that model (8) undergoes a Hopf-
bifurcation at Ew as q passes through value q[uw ] . 

Theorem 2.9. Suppose q < min{α, bβ(uw + b)}, then the positive equilibrium Ew of model (8) is globally asymptotically stable.
Proof. Construct a Lyapunov function:
v
u
ξ − uw η − vw
 
c
V (u, v) = dξ + dη. (17)
uw ξ s(1 + mvw ) vw η
Set
 ∂V uw
 =1− = 0,
∂u

u
 ∂V = c vw 

1− = 0,

∂v s(1 + mvw ) v
then we can obtain (u, v) = (uw , vw ). And the Hessian matrix at (uw , vw ) is given by:
 1 
0
H (E )|(uw ,vw ) =  uw ,
 
c
0
svw (1 + mvw )
and it is obvious that H (E )|(uw ,vw ) is positive definite. Hence, we have
min(V (u, v)) = V (uw , vw ) = 0.
Therefore V (u, v) > 0 for all (u, v) ̸= (uw , vw ).
The derivative of V with respect to the time t along the solutions of (8) is
cv
   
dV q su
= (u − uw ) α − β u − − + (v − vw ) −γ + ,
dt u+b mv + 1 mv + 1
and using the facts
 q c vw
 α − β uw − − = 0,
mvw + 1

uw + b
suw
−γ +
 = 0,
mvw + 1
we get
 
dV q cmuw
= −(u − uw )2 β − − (v − vw )2 . (18)
dt (uw + b)(u + b) (mv + 1)(mvw + 1)2
Since u > 0 and q < bβ(uw + b), we have
q q
β− >β− > 0.
(uw + b)(u + b) b(uw + b)

It follows that < 0. Hence Ew = (uw , vw ) is globally asymptotically stable. The proof is completed. 
dV
dt
   
Similar to Theorem 2.8, we have the following results of the stability of Ew3 = uw3 , wm3γ and Ew4 = uw4 , wm4γ
su −γ su −γ
.

Theorem 2.10. Suppose that (uw3 + b)2 (c γ + msβ u2w3 ) − mqsu2w3 > 0 and (uw4 + b)2 (c γ + msβ u2w4 ) − mqsu2w4 < 0. And
define

(γ − suw3 )γ (uw3 + b)2

 
q[uw3 ] = β uw 3 − .
suw3 uw 3
112 W. Wang et al. / Nonlinear Analysis: Real World Applications 16 (2014) 103–119

Fig. 1. Dynamics of model (8) with weak Allee effect. The parameters are taken as α = 1, β = 0.3, γ = 0.3, b = 0.5, c = 0.6, m = 0.6, q = 0.35, s = 2.
Ew0 = (0, 0) and Ew1 = (3.0102, 0) are saddle points, Ew = (0.31441, 1.8268) is globally asymptotically stable. The dashed curve is the u-nullcline
f (u, v) = 0, and the dotted curve is the v -nullcline g (u, v) = 0.

 
(a) If q < min{q[uw3 ] , bα}, then Ew3 = uw3 ,
suw3 −γ
mγ
is locally asymptotically stable;
 
(b) Ew4 = uw4 ,
suw4 −γ
mγ
is a saddle point.

In Fig. 1, we show the global stability of the unique positive equilibrium Ew with weak Allee effect.

2.3.3. The case with strong Allee effect

In this subsection, we discuss the stability of equilibria of model (8) with strong Allee effect, i.e., q > bα .
Model
 (8) with strong Allee effect has the following three boundary equilibria Es0 = (0, 0),
√   √
α−bβ− (α−bβ)2 +4β(bα−q) α−bβ+ (α−bβ)2 +4β(bα−q) (α+bβ)2
Es1 = 2β
,0 and Es2 = 2β
,0 with α > bβ and bα < q < 4β
.

Theorem 2.11. (a) Es0 is a stable node point.

(b) Suppose that Es1 and Es2 exist, then
(b1) Es1 is an unstable point;
(b2) Es1 is a stable node point if γ > γ̄ , or a saddle point if γ < γ̄ , where γ̄ has the same definition as in (16).

And in the similar way to Theorem 2.9, we can obtain the following results. For simplicity, we omit the proof.

Theorem 2.12. Suppose that (us + b)2 (c γ + msβ u2s ) − mqsu2s > 0, and define

(γ − sus )γ (us + b)2

 
q[us ]
= β us − .
sus us

 
(a) If bα < q < q[us ] , Es = us ,
sus −γ
mγ
is locally asymptotically stable; moreover,
(a1) if (q − q[us ] )2 < 4 det(JEs ), Es is a stable focus;
(a2) if (q − q[us ] )2 > 4 det(JEs ), Es is a stable node point.
 
(b) If q > max{q[us ] , bα}, Es = us ,
sus −γ
mγ
is unstable; moreover,
(b1) if (q − q ) < 4 det(JEs ), Es is an unstable focus surrounded by a limit cycle;
[us ] 2

(b2) if (q − q[us ] )2 > 4 det(JEs ), Es is an unstable node point and the limit cycle disappears.
(c) If q > max{β(us + b)2 , bα} and γ < sus , the model enters into a Hopf-bifurcation around Es at q = q[us ] .
 W. Wang et al. / Nonlinear Analysis: Real World Applications 16 (2014) 103–119 113

Fig. 2. The existence of the limit cycle in model (8) with strong Allee effect. The parameters are taken as α = 1, β = 0.75, γ = 0.4, b = 0.468, c =
0.4, m = 0.6, q = 0.49, s = 2.5. Es0 = (0, 0) is a stable node point, Es1 = (0.035342, 0) and Es2 = (0.82999, 0) are saddle points. Es = (0.19298, 0.34358)
is an unstable focus surrounded by a stable limit cycle. The dashed curve is the u-nullcline f (u, v) = 0, and the dotted curve is the v -nullcline g (u, v) = 0.

Fig. 2 shows the existence of the stable limit cycle around Es = (0.19298, 0.34358), which is an unstable
 focus. 
us3 , m
sus3 −γ
Similar to Theorem 2.8, we can get the following results of the stability of the positive Es3 = γ
and
 
Es4 = us4 ,
sus4 −γ
mγ
.

Theorem 2.13. (i) Suppose that (us3 + b)2 (c γ + msβ u2s3 ) − mqsu2s3 > 0, and define

(γ − sus3 )γ (us3 + b)2

 
q[us3 ] = β us3 − .
sus3 us3
 
(a) If bα < q < q[us3 ] , Es3 = us3 ,
sus3 −γ
mγ
is locally asymptotically stable; moreover,
(a1) if (q − q ) < 4 det(JEs3 ), Es3 is a stable focus;
[us3 ] 2

(a2) if (q − q[us3 ] )2 > 4 det(JEs3 ), Es3 is a stable node point.

 
(b) If q > max{q[hb] , bα}, Es3 = us3 ,
sus3 −γ
mγ
is unstable; moreover,
(b1) if (q − q ) < 4 det(JEs3 ), Es3 is an unstable focus surrounded by a limit cycle;
[us3 ] 2

(b2) if (q − q[us3 ] )2 > 4 det(JEs3 ), Es3 is an unstable node point.

 
(ii) If (us4 + b)2 (c γ + msβ u2s4 ) − mqsu2s4 < 0, then Es4 = us4 ,
sus4 −γ
mγ
is a saddle point.

Theorems 2.11(a) and 2.12 or 2.13 show that when bα < q < q[us ] or bα < q < q[us3 ] holds, both Es0 and Es or
Es3 are locally asymptotically stable simultaneously, that is to say, in this case, model (8) is bistable. In Fig. 3, we show the
bistable dynamics that there is a separatrix curve generated by the stable manifold of the positive equilibrium Es1 . The orbits
initiating the right of the separatrix curve tend to Es that represents the coexistence of the prey u and predator v , while the
orbits initiating the left of the separatrix curve tend to Es0 that represents the extinction of the prey u and predator v . This
shows that the model is highly sensitive to the initial conditions.
In Figs. 2 and 3, one can see that Es1 and Es2 are all saddle points. But there are some differences between them. In fact,
u-axis is the unstable manifold of Es1 , while for Es2 , u-axis is the stable manifold.

3. Stability/instability analysis of the spatial model

In 1952, Alan Turing published one paper on the subject called ‘‘The Chemical Basis of Morphogenesis’’ [31], putting forth
the Turing hypothesis of diffusion-driven instability. An equilibrium is called diffusion-driven instability, which means that
the equilibrium is asymptotically stable in the non-spatial model but unstable with respect to solutions in the spatial model.
Turing’s revolutionary idea was that passive diffusion could interact with the chemical reaction in such a way that even if
114 W. Wang et al. / Nonlinear Analysis: Real World Applications 16 (2014) 103–119

Fig. 3. Bistability of model (8) with strong Allee effect. The parameters are taken as α = 1, β = 0.75, γ = 0.4, b = 0.475, c = 0.4, m = 0.6, q =
0.49, s = 2.5. Es0 = (0, 0) is a stable node point, Es1 = (0.02397, 0) and Es2 = (0.83436, 0) are saddle points. Es∗ = (0.19612, 0.37627) is locally stable.
The dashed curve is the u-nullcline f (u, v) = 0, and the dotted curve is the v -nullcline g (u, v) = 0.

the reaction by itself has no symmetry-breaking capabilities, diffusion can destabilize the symmetry so that the system with
diffusion can have them [32].
In this section, we will focus on the effect of diffusion or/and Allee effect on the model system about the positive
equilibrium point.

3.1. The case without Allee effect

We first consider the stability of the positive equilibria of model (9) without Allee effect, i.e., q = 0 and the model is
given by:
∂u cuv

 = u(α − β u) − + d1 1u,
∂t mv + 1


(19)
∂v
 
su
= v −γ + + d2 1v.


∂t mv + 1


In the previous section, we show that model (19) has a unique positive equilibrium E ∗ = (u∗ , v ∗ ) when sα > βγ , which
is locally asymptotically stable (cf., Theorem 2.5). Next, we will discuss the effect of diffusion on E ∗ .

Theorem 3.1. If the positive equilibrium E ∗ of model (19) exists, then it is uniformly asymptotically stable.
Proof. Set U1 = u − u∗ , V1 = v − v ∗ , the linearized system (19) around E ∗ = (u∗ , v ∗ ) is as follows:
 ∂U cu∗
1
 = d1 1U1 − β u∗ U1 − U2 ,
∂t (mv ∗ + 1)2




 ∂U

sv ∗ msu∗ v ∗
2
= d2 1U2 + U 1 − U2 , (20)

 ∂t   mv + 1
∗ (mv ∗ + 1)2
∂U  ∂ U2 


 1  = = 0.


∂ν ∂ Ω ∂ν ∂ Ω
Following Malchow et al. [33], we know that any solution of system (20) can be expanded into a Fourier series:
∞ ∞

 
( , ) ( , ) αnm (t ) sin kr,



 U1 r t = u nm r t =
n,m=0 n,m=0

∞ ∞ (21)
 
U2 (r, t ) = vnm (r, t ) = βnm (t ) sin kr,




n,m=0 n,m=0

where r = (x, y), and 0 < x < L, 0 < y < L. k = (kn , km ) and kn = nπ /L, km = mπ /L are the corresponding wavenumbers.
 W. Wang et al. / Nonlinear Analysis: Real World Applications 16 (2014) 103–119 115

Having substituted unm and vnm into (20), we obtain

dαnm cu∗

 = (−β u∗ − d1 k2 )αnm + − βnm ,
(mv ∗ + 1)2

dt

(22)
dβnm sv ∗ msu∗ v ∗
 
αnm + − − d2 k βnm ,
2

 =
mv ∗ + 1 (mv ∗ + 1)2

dt

where k2 = k2n + k2m .

A general solution of (22) has the form C1 exp(λ1 t ) + C2 exp(λ2 t ), where the constants C1 and C2 are determined by the
initial conditions (10) and the exponents λ1 , λ2 are the eigenvalues of the following matrix

cu∗
 
∗ 2
−β u − d1 k −
(mv ∗ + 1)2 .

JE ∗ =  (23)
sv ∗ msu v
 ∗ ∗
 2

− − d 2 k
mv ∗ + 1 (mv + 1)
∗ 2

Correspondingly, λi (i = 1, 2) is the solution of the following equation:

λ2i − tr(
JE ∗ )λi + det(
JE ∗ ) = 0,

where the trace and the determinant of JE ∗ are respectively

msu∗ v ∗
tr(
JE ∗ ) = −(d1 + d2 )k2 − β u∗ − ,
(mv ∗ + 1)2
d1 msu∗ v ∗ bmβ u∗ 2 v ∗ bcu∗ v ∗
 
det(
J ) = d1 d2 k4 + d2 β u∗ + k2
+ + .
(mv ∗ + 1)2 (mv ∗ + 1)2 (mv ∗ + 1)3

It is easy to know that tr(

JE ∗ ) < 0 and det(
JE ∗ ) > 0. Hence, the positive equilibrium E ∗ of model (19) is uniformly
asymptotically stable. 

From Theorem 3.1, we can see that the positive equilibrium E ∗ of model (19) is uniformly asymptotically stable.
Considering the results in Theorem 2.5 together, one can see that there is no effect on the stability of the positive equilibrium
whether with diffusion or not. That is to say, there is the nonexistence of diffusion-driven instability in model (19), which
is the special case of model (9) without Allee effect.

3.2. The case with Allee effect

In this subsection, we restrict ourselves to the stability analysis of spatial model (9), which is in the presence of Allee
effect.
For the sake of learning the effect of Allee effect on the positive equilibrium of model (9), we first give a definition called
Allee–diffusion-driven instability as follows.

Definition 3.2. If a positive equilibrium is uniformly asymptotically stable in the reaction–diffusion model without Allee
effect (e.g., model (19)), but unstable with respect to solutions of the reaction–diffusion model with Allee effect (e.g., model
(9)), then this instability is called Allee–diffusion-driven instability.

Next, we will only investigate the stability of the unique positive equilibrium Ew = (uw , vw ) of the reaction–diffusion
model (9). For simplicity, we take the weak Allee effect case (0 < q < bα) as an example.
The Jacobian matrix of model (9) at Ew = (uw , vw ) is given by:

cγ 2
  
q
−β + uw − d 1 k 2 −
 (uw + b)2 s 2 uw 
JEw = 
  
suw − γ γ (suw − γ ) 
− − d2 k 2
muw suw

and the characteristic equation of

JEw at Ew is

λ2 − tr(
JEw )λ + det(
JEw ) = 0, (24)
116 W. Wang et al. / Nonlinear Analysis: Real World Applications 16 (2014) 103–119

where

tr(
JEw ) = −(d1 + d2 )k2 + tr(JEw ), (25)
d1 γ (suw − γ )
   
q
det(
JEw ) = d1 d2 k4 + + β− d 2 uw k2 + det(JEw ). (26)
suw (uw + b)2

The conditions of the instability are that at least tr(

JEw ) < 0 or det(
JEw ) > 0 is violated.
From Theorem 2.8, we know that tr(JEw ) < 0, so tr( JEw ) = tr(JEw ) − (d1 + d2 )k2 < 0 is always true. Hence, only violation
of det(
JEw ) > 0 gives rise to Allee–diffusion-driven-instability.
A necessary condition for the instability of model (9) is

d1 γ (suw − γ )
 
q
+ β− d2 uw , Θ < 0,
suw (uw + b)2

otherwise det( JEw ) > 0 for all k since det(JEw ) > 0. For the instability, we must have det(
JEw ) < 0 for some k. And we notice
that det(
JEw ) achieves its minimum

4d1 d2 det(JEw ) − Θ 2
min det(
JEw ) = (27)
k 4d1 d2

at the critical value k∗ 2 > 0 where

Θ
2
k∗ = − .
2d1 d2
And Θ < 0 is equivalent to

d1 γ (suw − γ )
 
+ β (uw + b)2 < q < bα.
d2 su2w

mink det(
JEw ) < 0 is equivalent to 4d1 d2 det(JEw ) − Θ 2 < 0, which is equivalent to
 
d1 γ (suw − γ ) 2 d1 d2 det(JEw )

q > (uw + b) 2
β+ + .
d2 su2w d2 uw

And from det(

JEe ) = 0, we can determine k1 and k2 as:

−Θ − Θ 2 − 4d1 d1 det(JEw )

k21 = ,
2d1 d2
(28)
−Θ + Θ 2 − 4d1 d1 det(JEw )

k22 = .
2d1 d2

In conclusion, when k21 < k2 < k22 , det

(JEw ) < 0, and Ew of model (9) is unstable.
Summarizing the above discussions, we can obtain the following theorem immediately.

Theorem 3.3. Assume that the following conditions hold:

 
γ (suw −γ )
(a) q < + β (uw + b)2 ;
su2
w
(b) (uw + b)2 (c γ + msβ u2w ) − mqsu2w > 0;
 
d1 γ (suw −γ )
(c) + β (uw + b)2 < q < bα ;
d2 su2w
 √ 
d γ (suw −γ ) 2 d1 d2 det(JEw )
(d) q > (uw + b)2 β + 1 + ,
d su2 d2 uw 2 w

then the positive equilibrium Ew of model (9) is Allee–diffusion-driven-instability when 0 < k21 < k2 < k2 .

In the same way, one can obtain similar results of Theorem 3.3 in the case with strong Allee effect, i.e., q > bα . That
is to say, Allee–diffusion-driven-instability can also occur at Es of model (9) with strong Allee effect. And there are similar
instability results of the positive equilibria Ew3 and Es3 .
 W. Wang et al. / Nonlinear Analysis: Real World Applications 16 (2014) 103–119 117

3.3. Pattern formation

Diffusion-driven instability in reaction–diffusion systems has been proposed as a mechanism for pattern formation in
numerous embryological and ecological contexts. Mathematically speaking, an equilibrium is a diffusion-driven instability
(or Turing instability) meaning that it is an asymptotically stable equilibrium of model (8) but is unstable with respect to
the solutions of spatial model (9) [34].
In this subsection, in the two-dimensional space, we perform extensive numerical simulations of the spatially extended
model (9) in the case with weak Allee effect, and show qualitative results. All of the numerical simulations employ the zero-
flux boundary conditions (11) with a system size of 200 × 200. Other parameters are fixed as α = 1, β = 0.3, γ = 0.3, b =
0.5, c = 0.6, m = 0.6, q = 0.35, d1 = 0.015, d2 = 1.
The numerical integration of model (9) is performed by using an explicit Euler method for the time integration [35] with
a time step size 1t = 100 1
, and the standard five-point approximation [36] for the 2D Laplacian with the zero-flux boundary
conditions. The initial conditions are always a small amplitude random perturbation around the positive constant steady
state solution Ew . After the initial period during which the perturbation spreads, the model goes into either a time-dependent
state, or an essentially steady state solution (time-independent).
In the numerical simulations, different types of dynamics can be observed and it is found that the distributions of predator
and prey are always of the same type. Consequently, we can restrict our analysis of pattern formation to one distribution.
We only show the distribution of prey u for instance.
In Fig. 4(a), with s = 1.75, there is a pattern consisting of blue hexagons (minimum density of u) in a red (maximum
density of u) background, i.e., isolated zones with low population densities. We call this pattern as ‘‘holes’’. When increasing
s to s = 1.9, a few stripes emerge and the remainder of the holes pattern remains time-independent, i.e., it is a stripe–hole
mixtures pattern (cf., Fig. 4(b)). While increasing s to s = 2, the model dynamics exhibits a transition from stripes–holes
growth to stripes replication, i.e., holes delay and the stripes pattern emerges (cf., Fig. 4(c)). On increasing s = 2.45, a
few spots, associated with high population densities, fill in the stripes, i.e., the stripe–spot mixtures pattern emerges (cf.,
Fig. 4(d)). And the later random perturbations make these stripes decay, and ending with the time-independent regular
spots when s increasing to s = 3 (cf., Fig. 4(e)), which are isolated zones with high prey densities.
From Fig. 4, one can see that, on increasing the control parameter s, the pattern sequence ‘‘holes → stripe–hole mixtures
→ stripes → stripe–spot mixtures → spots’’ is observed.
From the view point of population dynamics, ‘‘spots’’ pattern (cf., Fig. 4(e)) shows that the prey population is driven by
the predator to a very low level in those regions. The final result is the formation of patches of high prey density surrounded
by areas of low prey densities [37]. That is to say, under the control of these parameters, the prey is predominant in the
domain. In contrast, ‘‘holes’’ pattern (cf., Fig. 4(a)) indicates that the predator is predominant in the domain.

4. Conclusions and remarks

In summary, in this paper, we have investigated the complex dynamics induced by Allee effect in a predator–prey model.
The value of this study lies in two aspects. First, it investigates the boundedness of the positive solutions, the existence
and stability of the equilibria of the non-spatial model, which indicates that the dynamics of the model with Allee effect
is complex. Second, it gives the analysis of Allee–diffusion-driven instability in the spatial domain, which shows that the
model dynamics exhibits complex pattern replication induced by Allee effect and diffusion together.
For the non-spatial model, the solutions of the model with/without Allee effect always exist and stay positive, and all
the non-negative solutions that start in R2+ are uniformly bounded. That is to say, Allee effect remains the boundedness of
the solutions. And in the case without Allee effect, there is a unique positive equilibrium, which is globally asymptotically
stable. In contrast, with Allee effect, none, one or two equilibria can exist at the interior of the first quadrant of the model.
And the model can have Hopf bifurcation and the existence of the limit cycle.
Especially, in the case with strong Allee effect, we find that there exists a separatrix curve determined by the stable
manifold of saddle point Es1 that separates the behavior of trajectories of the system. The orbits initiating the right of the
separatrix curve tend to the positive equilibrium that represents the coexistence of prey and predator, while the orbits
initiating the left of the separatrix curve tend to equilibrium (0, 0) that represents the extinction of prey and predator.
This means that, in this situation, the trajectories of the model can have different types of behavior strongly depending on
the initial conditions. In this viewpoint, to keep the prey and predator species coexist, we must regulate and control the
parameters in the right of the separatrix curve. Hence, it is important for ecologists to be aware of the kind of bistability
shown here.
For the spatial model, in the case without Allee effect, there is no effect on the stability of the positive equilibrium whether
with diffusion or not. That is to say, there is the nonexistence of diffusion-driven instability in the model without Allee effect.
And in the case with Allee effect, the positive equilibrium can be unstable. This instability is induced by Allee effect and
diffusion together, so we give a new definition called ‘‘Allee–diffusion-driven instability’’ and present the analysis of this
instability of the model in detail. To the best of our knowledge, this is the first reported case. Furthermore, via numerical
simulations, it is found that the model dynamics exhibits both Allee effect and diffusion controlled pattern formation
growth to holes, stripes–holes mixtures, stripes, stripes–spots mixtures, and spots replication. This indicates that the pattern
formation of the model with Allee effect is not simple, but rich and complex.
118 W. Wang et al. / Nonlinear Analysis: Real World Applications 16 (2014) 103–119

a b

c d

Fig. 4. Typical Turing patterns of u in model (9) with parameters α = 1, β = 0.3, γ = 0.3, b = 0.5, c = 0.6, m = 0.6, q = 0.35, d1 = 0.015, d2 = 1. (a)
s = 1.75; (b) s = 1.9; (c) s = 2; (d) s = 2.45; (e) s = 3.

The methods and results in the present paper related to the predator–prey model with Allee effect may enrich the
research of stability/instability analysis and pattern formation, and may be of some help to the investigation of other
reaction–diffusion systems.
 W. Wang et al. / Nonlinear Analysis: Real World Applications 16 (2014) 103–119 119

Acknowledgments

The authors would like to thank the anonymous referee for very helpful suggestions and comments which led to
improvements of their original manuscript. This research was supported by the National Science Foundation of China
(61373005) and Zhejiang Provincial Natural Science Foundation (LY12A01014).

References

[1] J.D. Murray, Mathematical Biology, Springer, 2003.

[2] R.M. Callaway, L.R. Walker, Competition and facilitation: a synthetic approach to interactions in plant communities, Ecology 78 (7) (1997) 1958–1965.
[3] S.D. Hacker, S.D. Gaines, Some implications of direct positive interactions for community species diversity, Ecology 78 (7) (1997) 1990–2003.
[4] G.W. Harrison, Multiple stable equilibria in a predator–prey system, Bull. Math. Biol. 48 (2) (1986) 137–148. Springer.
[5] G.W. Harrison, Comparing predator–prey models to Luckinbill’s experiment with didinium and paramecium, Ecology 76 (2) (1995) 357–374.
[6] M.E. Solomon, The natural control of animal populations, J. Anim. Ecol. 18 (1949) 1–35.
[7] A.J. Lotka, Elements of Physical Biology, Williams and Wilkins Co., 1924.
[8] C. Jost, Comparing predator–prey models qualitatively and quantitatively with ecological time-series data, Institut National Agronomique, Paris-
Grignon, 1998.
[9] H.I. Freedman, Stability analysis of a predator–prey system with mutual interference and density-dependent death rates, Bull. Math. Biol. 41 (1979)
67–78.
[10] W.C. Allee, Animal Aggregations: A Study in General Sociology, AMS Press, 1978.
[11] F. Courchamp, J. Berec, J. Gascoigne, Allee Effects in Ecology and Conservation, Oxford University Press, New York, 2008.
[12] B. Dennis, Allee effects: population growth, critical density, and the chance of extinction, Nat. Resour. Model. 3 (4) (1989) 481–538.
[13] M.A. McCarthy, The Allee effect, finding mates and theoretical models, Ecol. Model. 103 (1) (1997) 99–102.
[14] G. Wang, X.G. Liang, F.Z. Wang, The competitive dynamics of populations subject to an Allee effect, Ecol. Model. 124 (2) (1999) 183–192.
[15] M.A. Burgman, S. Ferson, H.R. Akçakaya, Risk Assessment in Conservation Biology, Vol. 12, Chapman and Hall, London, 1993.
[16] J. Wang, J. Shi, J. Wei, Dynamics and pattern formation in a diffusive predator–prey system with strong Allee effect in prey, J. Differential Equations
251 (2011) 1276–1304.
[17] J. Wang, J. Shi, J. Wei, Predator–prey system with strong Allee effect in prey, J. Math. Biol. 62 (3) (2011) 291–331.
[18] P.A. Stephens, W.J. Sutherland, Consequences of the Allee effect for behaviour, ecology and conservation, Trends Ecol. Evol. 14 (10) (1999) 401–405.
[19] P.A. Stephens, W.J. Sutherland, R.P. Freckleton, What is the Allee effect? Oikos 87 (1999) 185–190.
[20] A. Morozov, S. Petrovskii, B.L. Li, Spatiotemporal complexity of patchy invasion in a predator–prey system with the Allee effect, J. Theoret. Biol. 238
(1) (2006) 18–35.
[21] C. Çelik, O. Duman, Allee effect in a discrete-time predator–prey system, Chaos Solitons Fractals 40 (4) (2009) 1956–1962.
[22] S.R. Zhou, Y.F. Liu, G. Wang, The stability of predator–prey systems subject to the Allee effects, Theor. Popul. Biol. 67 (1) (2005) 23–31.
[23] J. Zu, M. Mimura, The impact of Allee effect on a predator–prey system with Holling type II functional response, Appl. Math. Comput. 217 (7) (2010)
3542–3556.
[24] E. González-Olivares, H. Meneses-Alcay, B. González-Yañez, et al., Multiple stability and uniqueness of limit cycle in a Gause-type predator–prey
model considering Allee effect on prey, Nonlinear Anal. RWA 12 (6) (2011) 2931–2942.
[25] Y. Cai, W. Wang, J. Wang, Dynamics of a diffusive predator–prey model with additive Allee effect, Int. J. Biomath. 5 (2) (2012) 1250023.
[26] C.E. Brassil, Mean time to extinction of a metapopulation with an Allee effect, Ecol. Model. 143 (1) (2001) 9–16.
[27] P. Aguirre, E. González-Olivares, E. Sáez, Two limit cycles in a Leslie–Gower predator–prey model with additive Allee effect, Nonlinear Anal. RWA 10
(3) (2009) 1401–1416.
[28] P. Aguirre, E. González-Olivares, E. Sáez, Three limit cycles in a Leslie–Gower predator–prey model with additive Allee effect, SIAM J. Appl. Math. 69
(2009) 1244.
[29] M.A. Aziz-Alaoui, M. Daher Okiye, Boundedness and global stability for a predator–prey model with modified Leslie–Gower and Holling-type II
schemes, Appl. Math. Lett. 16 (7) (2003) 1069–1075.
[30] S. Wiggins, Introduction to Applied Nonlinear Dynamical Systems and Chaos, Springer, 2003.
[31] A. Turing, The chemical basis of morphogenesis, Phil. Trans. R. Soc. Lond. Ser. B 237 (1952) 37–72.
[32] N.F. Britton, Essential Mathematical Biology, Springer, 2003.
[33] H. Malchow, S. Petrovskii, E. Venturino, Spatiotemporal Patterns in Ecology and Epidemiology–Theory, Models, and Simulation, in: Mathematical and
Computational Biology Series, Chapman & Hall/CRC, Boca Raton, 2008.
[34] Z.A. Wang, T. Hillen, Classical solutions and pattern formation for a volume filling chemotaxis model, Chaos 17 (2007) 037108.
[35] M.R. Garvie, Finite-difference schemes for reaction–diffusion equations modeling predator–prey interactions in MATLAB, Bull. Math. Biol. 69 (3) (2007)
931–956.
[36] A. Munteanu, R. Solé, Pattern formation in noisy self-replicating spots, Int. J. Bifurcation Chaos 16 (12) (2006) 3679–3683.
[37] D. Alonso, F. Bartumeus, J. Catalan, Mutual interference between predators can give rise to Turing spatial patterns, Ecology 83 (1) (2002) 28–34.