Applied Mathematics Letters 112 (2021) 106756

Contents lists available at ScienceDirect

Applied Mathematics Letters

www.elsevier.com/locate/aml

Influence of the fear factor on the dynamics of a stochastic

predator–prey model
Qun Liu a , Daqing Jiang b,c ,∗
a
School of Mathematics and Statistics, Key Laboratory of Applied Statistics of MOE, Northeast Normal
University, Changchun 130024, Jilin Province, PR China
b
College of Science, China University of Petroleum, Qingdao 266580, Shandong Province, PR China
c
Nonlinear Analysis and Applied Mathematics (NAAM)-Research Group, King Abdulaziz
University, Jeddah, Saudi Arabia

article info abstract

Article history: In this paper, we construct and analyze a stochastic predator–prey system where
Received 24 July 2020 apart from direct predation the prey population is affected by the fear induced
Received in revised form 3 September from predators. We obtain sufficient criteria for the existence and uniqueness of an
2020 ergodic stationary distribution of positive solutions to the system by establishing a
Accepted 4 September 2020 series of suitable Lyapunov functions. In a biological interpretation, the existence
Available online 10 September 2020
of a stationary distribution indicates that both the prey and predator species will
Keywords: be persistent and coexistent in the long term.
Stochastic predator–prey model © 2020 Elsevier Ltd. All rights reserved.
Fear effect
Nonlinear perturbation
Stationary distribution
Ergodicity

1. Introduction

A large number of ecological studies have concentrated on the fundamental predator–prey interaction
that is governed by direct killing of prey in the presence of predators. However, both theoretical studies and
experimental data have pointed out that the presence of predator can change the behavior of the prey even
more powerfully than the influence of direct predation [1,2]. Studies have shown that the mental state of
the adolescent prey can be mediated by fear induced from predators and the alternation causes deleterious
outcomes on their adult’s survival [3]. Therefore, many evolutionary ecologists and theoretical biologists have
realized that cost of fear or indirect effects of predators along with direct predation should be incorporated
in prey–predator interaction [4]. Recently, numerous mathematical models which include the cost of fear in
interacting population species have been set up [3,5–7]. In particular, Zhang et al. [5] proposed a predator–
prey model with Holling type II functional response incorporating a prey refuge and fear effect. If we do not

∗ Corresponding author at: College of Science, China University of Petroleum, Qingdao 266580, Shandong Province, PR China.
E-mail address: [email protected] (D. Jiang).

https://doi.org/10.1016/j.aml.2020.106756
0893-9659/© 2020 Elsevier Ltd. All rights reserved.
2 Q. Liu and D. Jiang / Applied Mathematics Letters 112 (2021) 106756

consider the prey refuge, i.e., m = 0, then the system can be written as follows:
dx αx βxy
⎧
⎪
⎨ = − bx2 − ,
dt 1 + Ky 1 + ax (1.1)
⎩ dy = −γy + cβxy ,
⎪
dt 1 + ax
where the variables x(t) and y(t) denote the densities of prey and predator populations at time t, respectively.
The parameters α, K, b, β, a, γ and c are assumed to be positive constants, α denotes the intrinsic growth
rate, K models the fear level induced by predator, b represents the strength of competition among individuals
of species x, γ denotes the death rate of the predator, β denotes the search rate and a is the half saturation
constant, c represents the energy conversion rate. The term βx/(1+ax) denotes the Holling-type-II functional
response.
Evidently, system (1.1) is derived based on the assumption that all input variables follow deterministic
laws and they are deterministic functions of time. However, the deterministic system has its limitation in
mathematical modeling of ecosystems since the parameters involved in the system are not able to capture
the influence of random environmental fluctuations [3]. Therefore, it is reasonable to introduce the white
noise term into the corresponding deterministic model to study the effects of demographic stochasticity on
the entire ecosystem.
Following the idea of Liu et al. [8], in this paper, we consider the following stochastic predator–prey model
with fear effect and nonlinear perturbation:
( )
αx βxy
⎧
2
⎪
⎪ dx = − bx − dt + x(σ11 + σ12 x)dB1 (t),
⎨ 1 + Ky 1 + ax
( ) (1.2)
cβxy
⎩dy = −γy +
⎪
⎪ dt + y(σ21 + σ22 y)dB2 (t),
1 + ax
where B1 (t) and B2 (t) are mutually independent standard Brownian motions defined on a complete
probability space (Ω , F, {Ft }t≥0 , P) with a filtration {Ft }t≥0 satisfying the usual conditions (i.e., it is
increasing and right continuous and F0 contains all P-null sets) [9], σ12 and σ22 are the intensities of the
white noise. If f is a bounded function on [0, ∞), define f u = supt≥0 f (t).
To study the stochastic dynamics of system (1.2), we should first present some conditions under which
system (1.2) has a unique global positive solution. Since the proof is similar to the statement of Lemma 2.1
in Liu et al. [8], we only state the result without proof.

Lemma 1.1. For any initial value (x(0), y(0))T ∈ R2+ , there is a unique solution (x(t), y(t))T to system (1.2)
on t ≥ 0 and the solution will remain in R2+ with probability one, namely, the solution (x(t), y(t))T ∈ R2+ for
all t ≥ 0 almost surely (a.s), where R2+ = {x = (x1 , x2 )T ∈ R2 : xi > 0, i = 1, 2}.

2. Main results

In this section, we will give our main results of system (1.2).

Let X(t) be a regular time-homogeneous Markov process in Rd described by the stochastic differential
equation
∑k
dX(t) = f (X(t))dt + gr (X)dBr (t).
r=1
The diffusion matrix of the process X(t) is defined as follows
k
∑
A(x) = (aij (x)), aij (x) = gri (x)grj (x).
r=1
 Q. Liu and D. Jiang / Applied Mathematics Letters 112 (2021) 106756 3

Lemma 2.1 ([10]). The Markov process X(t) has a unique ergodic stationary distribution µ(·) if there exists
a bounded open domain D ⊂ Rd with regular boundary Γ , having the following properties:
(B.1) In the domain D and some neighborhood thereof, the smallest eigenvalue of the diffusion matrix
A(x) is bounded away from zero.
(B.2) If x ∈ Rd \ D, the mean time τ at which a path issuing from x reaches the set D is finite, and
supx∈U Ex τ < ∞ for every compact subset U ⊂ Rd .

Lemma 2.2 ([11]). For any x > 0, the following inequality holds
√
x(1 − x) + 2x ≤ 2 x.

We now proceed to give our main theorem which is stated as follows.

2
σ11 2
2
σ11 b2 cβ(α− 2 ) 2
σ21
Theorem 2.1. If α > 2 and λ := ασ 2
− (γ + 2 ) > 0, then system (1.2) admits a
α(b+aα)(b+σ11 σ12 + 2b12 )2
unique stationary distribution µ(·) and it has the ergodic property.

Proof . To prove Theorem 2.1, we only need to validate conditions (B.1) and (B.2) in Lemma 2.1. We first
prove the condition (B.1). The diffusion matrix of system (1.2) is given by
( 2
x (σ11 + σ12 x)2
)
0
A= .
0 y 2 (σ21 + σ22 y)2

Evidently, the matrix A is positive definite for any compact subset of R2+ , then the condition (B.1) in
Lemma 2.1 holds.
Now we prove the condition (B.2). In view of system (1.2), we obtain

αx βxy
L(x) = − bx2 −
1 + Ky 1 + ax
≤αx − bx2 (2.1)
α α(b + aα)
≤ − (1 + ax) + ,
a ab
where the second inequality holds due to the fact that b(x − αb )2 ≥ 0. In addition, we have

αx βxy
L(x) = − bx2 −
1 + Ky 1 + ax (2.2)
≤αx − bx2 ,

α βy σ2 σ2
L(− ln x) = − + bx + + 11 + σ11 σ12 x + 12 x2
1 + Ky 1 + ax 2 2
2
αKy βy σ11 σ2
=−α+ + bx + + + σ11 σ12 x + 12 x2 (2.3)
1 + Ky 1 + ax 2 2
2 2
σ σ
≤ − α + (b + σ11 σ12 )x + (αK + β)y + 11 + 12 x2
2 2
and
cβx σ2 σ2
L(− ln y) = − + γ + 21 + σ21 σ22 y + 22 y 2 . (2.4)
1 + ax 2 2
Define
V1 (x, y) = − ln y + c1 x,
4 Q. Liu and D. Jiang / Applied Mathematics Letters 112 (2021) 106756

where c1 is a positive constant which will be determined later. According to (2.1) and (2.4), we derive
cβx c1 α σ2 c1 α(b + aα) σ2
LV1 ≤ − − (1 + ax) + γ + 21 + + σ21 σ22 y + 22 y 2
1√+ ax a 2 ab 2
2 2
(2.5)
cc1 αβ √ σ21 c1 α(b + aα) σ22
≤−2 x+γ+ + + σ21 σ22 y + y2 .
a 2 ab 2
Then define ( 2
)
b b σ12
V2 (x) = x+ 2 − ln x + x ,
2α2 α(b + σ11 σ12 +
ασ12 2b
2b )
then in view of (2.2) and (2.3), we have
2
( ) ( )
b b b b σ11 b(αK + β)
LV2 ≤ x 1 − x + x − 2 α − + y
2α α α ασ 2 ασ 2
α(b + σ11 σ12 + 2b12 ) α(b + σ11 σ12 + 2b12 )
√
2
(2.6)
b√
( )
b σ11 b(αK + β)
≤ x− 2 α− + y,
α ασ12 2 ασ 2
α(b + σ σ +
11 12 )
2b α(b + σ σ + 12 ) 11 12 2b

where in the second inequality we have used Lemma 2.2. Next, define
√ ( √ )
c1 cα2 β 1 2b(αK + β) c1 cα2 β
V3 (x, y) = V1 (x, y) + 2 V2 (x) + σ21 σ22 + y,
ab γ ασ 2 ab
α(b + σ σ + 12 )
11 12 2b

then from (2.5) and (2.6) it follows that

2
σ11 √
2
2b(α − 2 ) c1 cα2 βσ21 c1 α(b + aα)
LV3 ≤ − 2 + +γ+
α(b +
ασ12
σ11 σ12 + 2b ) ab 2 ab
( √ )
cβ 2b(αK + β) c1 cα2 β
+ σ21 σ22 + 2
γ ασ
α(b + σ σ + 12 ) ab
11 12 2b
σ2
× xy + 22 y 2 .
2
Choose 2
σ11
ab3 cβ(α − 2 )
2
c1 = 2 ,
ασ12
α2 (b + aα)2 (b + σ11 σ12 + 2b )
2

then 2
σ11
b2 cβ(α − 2 )
2 2
σ21
LV3 ≤ − 2 +γ+
α(b + aα)(b + σ11 σ12 +
ασ12 2 2
2b )
√
2
( )
cβ 2b(αK + β) c1 cα2 β σ22
+ σ21 σ22 + 2 xy + y2 (2.7)
γ ασ12
α(b + σ11 σ12 + 2b ) ab 2
√
σ2
( )
cβ 2b(αK + β) c1 cα2 β
=−λ+ σ21 σ22 + 2 xy + 22 y 2 ,
γ ασ
α(b + σ11 σ12 + 2b12 ) ab 2

where 2
σ11
b2 cβ(α − 2 2
( )
2 ) σ21
λ := − γ + > 0.
ασ 2 2
α(b + aα)(b + σ11 σ12 + 2b12 )2
Define
Q p
V4 (y) = y ,
p
 Q. Liu and D. Jiang / Applied Mathematics Letters 112 (2021) 106756 5

where p ∈ (0, 1) is a sufficiently small number and Q is a sufficiently large positive constant which will be
determined later. Then we have
( )
cβx Q(1 − p) p
LV4 =Qy p −γ + − y (σ21 + σ22 y)2
1 + ax 2
(2.8)
p Q(1 − p) 2 p+2
≤cβQxy − σ22 y .
2
From (2.7) and (2.8) it follows that
( √ )
cβ 2b(αK + β) c1 cα2 β
L(V3 + V4 ) ≤ − λ + σ21 σ22 + 2 xy + cβQxy p
γ ασ12
α(b + σ11 σ12 + 2b ) ab
2
σ Q(1 − p) 2 p+2
+ 22 y 2 − σ22 y .
2 2

Remark 2.1. Let h(y) = A1 y 2 − A2 y p+2 for y > 0 and any positive constants A1 and A2 . Hence with
1 A 1 2 A 2 p
2
ȳ = ( p+2 )p (A2
) we obtain h′ (ȳ) = 0 and supy>0 h(y) = ( p+2
1 p 2
)p (A1 p
2
) ( p+2 )A1 .

Let
{ 2 } 2
( ) p2
σ22 2 Q(1 − p) 2 p+2 pσ22 2
Ap = sup y − σ22 y = .
y>0 2 2 2(p + 2) Q(p + 2)(1 − p)
According to Remark 2.1, we get

2 2
( ) p2
σ22 2 Q(1 − p) 2 p+2 pσ22 2
y − σ22 y ≤ .
2 2 2(p + 2) Q(p + 2)(1 − p)

Thus, in view of (2.9), we derive

2
( ) p2 ( √ )
pσ22 2 cβ 2b(αK + β) c1 cα2 β
L(V3 + V4 ) ≤ − λ + + σ21 σ22 + xy
2(p + 2) Q(p + 2)(1 − p) γ ασ 2 ab
α(b + σ11 σ12 + 2b12 )
+ cβQxy p .
p
p 2 2
2p 2 σ22 pσ22
Choose Q = p p+2 such that −λ + 2
2(p+2) ( Q(p+2)(1−p) )
p = − λ2 := −λ̄ < 0, then we obtain
λ 2 (2+p) 2 (1−p)

( √ )
cβ 2b(αK + β) c1 cα2 β
L(V3 + V4 ) ≤ −λ̄ + σ21 σ22 + 2 xy + cβQxy p . (2.9)
γ ασ12
α(b + σ11 σ12 + 2b ) ab

Next, define
xp yp
V5 (x, y) = + ,
p p
where p ∈ (0, 1) is a sufficiently small number. Then we have
( ) ( )
p α βy 1−p p 2 p cβx 1−p p
LV5 =x − bx − − x (σ11 + σ12 x) + y −γ + − y (σ21 + σ22 y)2
1 + Ky 1 + ax 2 1 + ax 2
σ 2 (1 − p) p+2 σ 2 (1 − p) p+2
≤αxp − 12 x + cβxy p − 22 y .
2 2
(2.10)
Define
V̄ (x, y) = M (V3 (x, y) + V4 (y)) + V5 (x, y),
6 Q. Liu and D. Jiang / Applied Mathematics Letters 112 (2021) 106756

where M is a sufficiently large positive constant satisfying −M λ̄ + f1u + f2u ≤ −2 and functions fi will be
determined later, i = 1, 2. In addition, note that V̄ (x, y) is not only continuous, but also tends to ∞ as (x, y)
approaches the boundary of R2+ . Hence it should be lower bounded and achieves this lower bound at a point
(x0 , y0 ) in the interior of R2+ . Then we define a nonnegative C 2 -function V : R2+ → R+ {0} by
⋃

V (x, y) =V̄ (x, y) − V̄ (x0 , y0 )

=M (V3 (x, y) + V4 (y)) + V5 (x, y) − V̄ (x0 , y0 ).

Therefore, according to (2.9) and (2.10), we have

( √ )
M cβ 2b(αK + β) c1 cα2 β
LV ≤ − M λ̄ + σ21 σ22 + 2 xy + cβ(M Q + 1)xy p
γ ασ12
α(b + σ11 σ12 + 2b ) ab
2
σ (1 − p)
+ αxp − 12 xp+2
2
σ 2 (1 − p) p+2
− 22 y
2 √
( )
M cβ 2b(αK + β) c1 cα2 β
=f1 (x) + f2 (y) − M λ̄ + σ21 σ22 + 2 xy + cβ(M Q + 1)xy p ,
γ α(b + σ σ +
ασ12
) ab
11 12 2b

where
2
σ12 (1 − p) p+2 σ 2 (1 − p) p+2
f1 (x) := − x + αxp , f2 (y) := − 22 y .
2 2
Denote
( √ )
M cβ 2b(αK + β) c1 cα2 β
G(x, y) = f1 (x) + f2 (y) − M λ̄ + σ21 σ22 + 2 xy + cβ(M Q + 1)xy p .
γ ασ12
α(b + σ11 σ12 + 2b ) ab

Then ⎧
⎨G(∞, y) → −∞, as x → ∞,
G(x, y) ≤ G(x, ∞) → −∞, as y → ∞,
⎩ u
f1 + f2u − M λ̄ ≤ −2, as x → 0+ or y → 0+ .
Thus, we can take 0 < ϵ < 1 sufficiently small such that

LV (x, y) ≤ −1 for any (x, y) ∈ R2+ \ Dϵ ,

where Dϵ = [ϵ, 1ϵ ] × [ϵ, 1ϵ ]. This shows that the condition (B.2) in Lemma 2.1 also holds. By Lemma 2.1, we
derive that system (1.2) admits a unique stationary distribution µ(·) and it has the ergodic property. This
completes the proof.

Acknowledgments

This work is supported by the National Natural Science Foundation of China (No. 11871473), Shandong
Provincial Natural Science Foundation, China (No. ZR2019MA010) and the Fundamental Research Funds
for the Central Universities, China (No. 2412020QD024).

References

[1] X. Wang, L. Zanette, X. Zou, Modelling the fear effect in predator–prey interactions, J. Math. Biol. 73 (2016) 1–26.
[2] X. Wang, X. Zou, Modeling the fear effect in predator–prey interactions with adaptive avoidance of predators, Bull.
Math. Biol. 79 (2017) 1–35.
 Q. Liu and D. Jiang / Applied Mathematics Letters 112 (2021) 106756 7

[3] J. Roy, D. Barman, S. Alam, Role of fear in a predator–prey system with ratio-dependent functional response in
deterministic and stochastic environment, Bio Syst. (2020) http://dx.doi.org/10.1016/j.biosystems.2020.104176.
[4] D. Barman, J. Roy, S. Alam, Trade-off between fear level induced by predator and infection rate among prey species,
J. Appl. Math. Comput. (2020) http://dx.doi.org/10.1007/s12190-020-01372-1.
[5] H. Zhang, Y. Cai, S. Fu, W. Wang, Impact of the fear effect in a prey-predator model incorporating a prey refuge,
Appl. Math. Comput. 356 (2019) 328–337.
[6] J. Wang, Y. Cai, S. Fu, W. Wang, The effect of the fear factor on the dynamics of a predator–prey model incorporating
the prey refuge, Chaos 29 (2019) 083109.
[7] K. Sarkar, S. Khajanchi, Impact of fear effect on the growth of prey in a predator–prey interaction model, Ecol. Complex.
42 (2020) 100826.
[8] Q. Liu, D. Jiang, T. Hayat, B. Ahmad, Stationary distribution and extinction of a stochastic predator–prey model with
additional food and nonlinear perturbation, Appl. Math. Comput. 320 (2018) 226–239.
[9] X. Mao, Stochastic Differential Equations and Applications, Horwood Publishing, Chichester, 1997.
[10] R. Khasminskii, Stochastic Stability of Differential Equations, Springer Heidelberg Dordrecht London, New York, 2012.
[11] Q. Liu, D. Jiang, N. Shi, T. Hayat, A. Alsaedi, Dynamical behavior of a stochastic HBV infection model with logistic
hepatocyte growth, Acta Math. Sci. 37 (B) (2017) 927–940.