You are on page 1of 17

Bifurcations, chaos, and multistability in a

nonautonomous predator–prey model with


fear
Cite as: Chaos 31, 123134 (2021); https://doi.org/10.1063/5.0067046
Submitted: 13 August 2021 • Accepted: 06 December 2021 • Published Online: 30 December 2021

Mainul Hossain, Saheb Pal, Pankaj Kumar Tiwari, et al.

ARTICLES YOU MAY BE INTERESTED IN

A framework for causal discovery in non-intervenable systems


Chaos: An Interdisciplinary Journal of Nonlinear Science 31, 123128 (2021); https://
doi.org/10.1063/5.0054228

Complex network analysis of spatiotemporal dynamics of premixed flame in a Hele–Shaw cell:


A transition from chaos to stochastic state
Chaos: An Interdisciplinary Journal of Nonlinear Science 31, 123133 (2021); https://
doi.org/10.1063/5.0070526

Catalytic feed-forward explosive synchronization in multilayer networks


Chaos: An Interdisciplinary Journal of Nonlinear Science 31, 123130 (2021); https://
doi.org/10.1063/5.0060803

Chaos 31, 123134 (2021); https://doi.org/10.1063/5.0067046 31, 123134

© 2021 Author(s).
Chaos ARTICLE scitation.org/journal/cha

Bifurcations, chaos, and multistability in a


nonautonomous predator–prey model with fear
Cite as: Chaos 31, 123134 (2021); doi: 10.1063/5.0067046
Submitted: 13 August 2021 · Accepted: 6 December 2021 ·
Published Online: 30 December 2021 View Online Export Citation CrossMark

Mainul Hossain,1 Saheb Pal,1 Pankaj Kumar Tiwari,2 and Nikhil Pal1,a)

AFFILIATIONS
1
Department of Mathematics, Visva-Bharati, Santiniketan 731235, India
2
Department of Basic Science and Humanities, Indian Institute of Information Technology, Bhagalpur 813210, India

a)
Author to whom correspondence should be addressed: nikhil.pal@visva-bharati.ac.in

ABSTRACT
Classical predator–prey models usually emphasize direct predation as the primary means of interaction between predators and prey.
However, several field studies and experiments suggest that the mere presence of predators nearby can reduce prey density by forcing them to
adopt costly defensive strategies. Adoption of such kind would cause a substantial change in prey demography. The present paper investigates
a predator–prey model in which the predator’s consumption rate (described by a functional response) is affected by both prey and predator
densities. Perceived fear of predators leads to a drop in prey’s birth rate. We also consider both constant and time-varying (seasonal) forms
of prey’s birth rate and investigate the model system’s respective autonomous and nonautonomous implementations. Our analytical studies
include finding conditions for the local stability of equilibrium points, the existence, direction of Hopf bifurcation, etc. Numerical illustra-
tions include bifurcation diagrams assisted by phase portraits, construction of isospike and Lyapunov exponent diagrams in bi-parametric
space that reveal the rich and complex dynamics embedded in the system. We observe different organized periodic structures within the
chaotic regime, multistability between multiple pairs of coexisting attractors with intriguing basins of attractions. Our results show that even
relatively slight changes in system parameters, perturbations, or environmental fluctuations may have drastic consequences on population
oscillations. Our observations indicate that the fear effect alters the system dynamics significantly and drives an otherwise irregular system
toward regularity.
Published under an exclusive license by AIP Publishing. https://doi.org/10.1063/5.0067046

Classical predator–prey models have revolved primarily around analytically. With the help of advanced numerical tools and tech-
direct predation. However, in recent years, the indirect effects niques, we explore the role of different system parameters by
of predators on their prey have gained considerable attention using bifurcation diagrams assisted by phase portraits. We also
among ecological researchers. Several field studies have been construct several isospike and Lyapunov exponent diagrams in
carried out to assess the impact of prey’s perceived fear of the bi-parametric space of fear and birth rate parameters. In this
being killed by its predators as opposed to just direct preda- way, we illustrate the intricate and rich dynamics woven into our
tion. Furthermore, theoretical studies have revealed many aspects system. Our observations include the presence of different orga-
of the role of fear effect on the stability and coexistence across nized periodic structures (shrimp-shaped periodic islands and
a wide range of ecological contexts. Inspired by recent works Arnold tongues) within the chaotic regime. We also detect multi-
on ecological systems incorporating fear effect, we investigate stability between several pairs of coexisting attractors. It appears
a nonautonomous predator–prey model with a time-varying from their basins of attractions that they are of fractal nature. Our
birth rate and direct predation being governed by Cosner func- results suggest that the parameter values and the initial popula-
tional response. Here, fear is introduced as a decreasing factor tion sizes of prey and predator play a vital role in the dynamics
to prey’s birth rate. We explore both autonomous and nonau- of the system; a slight alteration in either of these can lead to a
tonomous implementations of our model system. This includes drastic change in the outcome. We observe that the fear effect
finding the conditions for the local stability of equilibrium points, has a stabilizing role on the autonomous system, whereas in the
examining the existence and direction of Hopf bifurcation, etc., nonautonomous system, the fear effect drives the system toward

Chaos 31, 123134 (2021); doi: 10.1063/5.0067046 31, 123134-1


Published under an exclusive license by AIP Publishing
Chaos ARTICLE scitation.org/journal/cha

regularity. In summary, we conclude that the nonautonomous prey, and h is the handling time per prey. Unlike the other predator-
ecological systems are inherently complex and exhibit intricate dependent functional responses (such as ratio-dependent or Bed-
dynamical behaviors. dington–DeAngelis), the functional response (1.2) is increasing in
predator density y since it presumed that predators were foraging in
a group. For the convenience of mathematical analysis, we consider
I. BIOLOGICAL BACKGROUND AND MODEL α as the product of C and e0 in (1.2). Then, the model system (1.1)
FORMULATION can be written in the form

Evolution in organisms is shaped by interspecies interactions. 


αxy2
Among them, predator–prey interactions are a widely studied phe-  dx

dt
= bx − dx − ax2 − ,
 1 + hαxy
nomena in the scientific community. At its core, the interaction (1.3)
is between an organism and its natural enemies. Plant–herbivore, µαxy2
 dy =


dt
− my.
host–parasitoid, herbivore–carnivore, and host–pathogen interac- 1 + hαxy
tions are a few examples of prey–predator interactions. Recent
trends in population biology are to develop several mathemati- Recent empirical and theoretical studies suggested that preda-
cal models and study the long-term behavior of the species, their tor has both direct and indirect impacts (consumptive and non-
coexistence, and how they impact each other’s densities.1,2 Two- consumptive) upon their prey. In a direct way, the predator kills and
species mathematical models look simple; however, they exhibit rich consumes prey, but in an indirect way, the predator induces fear in
and complex dynamics. A prototypical two-species predator–prey the prey population and changes the prey’s behavior. Due to fear
model introducing x(t) and y(t) as the prey and predator population of predation, the prey population enhances vigilance, limits forag-
densities at time t, respectively, can be written as ing activity, sacrifices higher intake zone and feeds in a safer place,
( dx adjust reproductive cycle, etc.12–16 These adjustments create trade-
dt
= bx − dx − ax2 − 8(x, y)y, offs such that prey will enhance their anti-predator behavior and
dy
(1.1) limit their foraging intensity or reproductive behavior. Therefore,
dt
= µ8(x, y)y − my,
prey compensates for mortality risk, which ultimately affects the
where b and d are the birth and death rates of prey, respectively, growth rate of prey species. Several researchers empirically observed
a denotes the intraspecific competition rate of prey, µ represents such types of phenomena in different predator–prey systems.17,18
the conversion factor of prey biomass into predator biomass, 8(x, y) Here, we assume that due to fear of predation risk, the birth rate
describes the prey consumption rate of individual predator depend- of the prey population is reduced. To incorporate the fear phe-
ing on both the predator and prey densities, and m is the predator nomenon, we multiply the birth rate (b) of prey population with a
mortality rate. Functional response is a basic modeling unit in decreasing function of the predator density (y) and the intensity of
a predator–prey system. It describes the intake rate at which a fear (k). k measures the level of fear exerted by a particular predator
predator consumes prey. The behavioral characteristic of predator on a particular prey. It may be understood this way: if the predators
species reflects through the term 8(x, y), and predation is typically are very aggressive and fearsome, k is higher, and if the predators are
described by the functional response. less fearsome, k assumes a lesser value. Although it may vary from
The most widely used functional responses in describing preda- species to species depending on many biological factors (age, terrain,
tor–prey interactions are of Holling types,3 which depend on prey etc.), for simplicity, we assume it to remain constant over time for all
density only. However, such prey-dependent functional responses age groups of the prey and predator. Now, from a biological point of
do not consider the interference among predators.4,5 Some biolo- view, the fear function 2(k, y) must have the following properties:
gists have claimed that in many situations, especially when predators
have to search for food (and therefore, have to share or compete 1. 2(0, y) = 1 : if there is no fear of predator, then the birth rate of
for food), the functional response in a predator–prey model should prey remains unchanged.
be predator-dependent.6–9 However, the appropriate form of func- 2. 2(k, 0) = 1 : if there is no predator in the system, then there is
tional response has been debated. For a better understanding of no reduction in prey reproduction due to fear.
the proper representation of predation processes, detailed observa- 3. limk→∞ 2(k, y) = 0 : when strength of fear is very large and
tions of environmental factors that influence functional response predator density y > 0, then prey reproduction decreases and
are required. Cosner et al.10 proposed a functional response that is ultimately becomes zero.
predator-dependent and describes the schooling behavior of both 4. limy→∞ 2(k, y) = 0 : when predator density is very large and
the predator and prey populations. A biological example regarding strength of fear k > 0, then prey reproduction decreases and
this type of behavior is that a school of tuna searches, contacts, and ultimately becomes zero due to large strength of fear.
then hunts a herd of prey.11 The functional form of such response 5. ∂2(k,y)
∂k
< 0 : the reproduction of prey decreases with an increase
function is given by in the strength of fear.
Ce0 xy 6. ∂2(k,y)
∂y
< 0 : the reproduction of prey decreases with an increase
8(x, y) = , (1.2) in predator density.
1 + hCe0 xy
where C is the amount of prey killed by a predator per encounter, The simplest form of 2(k, y) for which the above properties
1 19–23
e0 is the total encounter coefficient between the predator and the hold well is given by 1+ky .

Chaos 31, 123134 (2021); doi: 10.1063/5.0067046 31, 123134-2


Published under an exclusive license by AIP Publishing
Chaos ARTICLE scitation.org/journal/cha

Thus, the model (1.3) with a fear phenomenon becomes


bx αxy2

dx 2
 dt = 1 + ky − dx − ax − 1 + hαxy ,


(1.4)
µαxy2
 dy =


dt
− my.
1 + hαxy
All the parameters involved in the system (1.4) are non-
negative. We investigate the behaviors of the above model both
analytically and numerically.

II. MODEL ANALYSIS


Mathematical analysis of the model gives important informa-
tion about the steady state (where the system does not change with
time), coexistence, and extinction condition of each species. The
predator–prey model (1.4) always exhibits the trivial equilibrium
E0 = (0, 0), where both the species extinct
 from the system. Also, the
predator-free equilibrium Ek = b−d a
, 0 exists for the model when
b > d.
For the coexistence equilibrium, where both the prey and
predator populations survive, we have to obtain the positive solu- FIG. 1. Figure shows non-trivial nullclines of the system (1.4) for k = 0.2,
tion(s) of the following system of algebraic equations: k = 0.7132, and k = 1.2. Other parameter values are b = 1.4, d = 0.3,
a = 0.4, α = 0.2, h = 2, µ = 0.7, and m = 0.1. Here, prey and predator
(
b αy2 nullclines are represented by red and green curves, respectively.
1+ky
− d − ax − 1+hαxy = 0,
µαxy
(2.1)
1+hαxy
− m = 0.
value of k. It is to be noted that the predator nullcline is independent
From the second equation of (2.1), we have
of the fear level.
(µ − mh)αxy − m = 0, Now, to determine the stability properties of coexistence equi-
librium Ē = (x̄, ȳ), we have to check the signs of trace and deter-
which demonstrates that model (1.4) has no coexistence equilibrium minant values of the corresponding variational matrix LĒ . The
for µ ≤ mh. From the ecological point of view, both the populations algebraic expression of LĒ is given by
can persist in the system if the conversion factor from prey biomass  
to predator biomass is sufficiently large compared to the product L L12
LĒ = 11 ,
of handling time and mortality rate of predators. Thus, we consider L21 L22
µ > mh, and the prey density of coexistence equilibrium is given by
m
x̄ = (µ−mh)αȳ , where the predator density ȳ is a positive root of the where
following biquadratic equation: hα 2 x̄ȳ3 kbx̄ α x̄ȳ(2 + hα x̄ȳ)
L11 = −ax̄ + , L12 = − − ,
ρ0 y4 + ρ1 y3 + ρ2 y2 + ρ3 y + ρ4 = 0, (1 + hα x̄ȳ)2 (1 + kȳ)2 (1 + hα x̄ȳ)2
µαȳ2 µα x̄ȳ
with L21 = 2
, L22 = .
(1 + hα x̄ȳ) (1 + hα x̄ȳ)2
2 2 2 2
ρ0 = kα (µ − mh) , ρ1 = α (µ − mh) ,
The trace (T) and determinant (D) values of the above variational
ρ2 = µdkα(µ − mh), matrix are given by
ρ3 = µamk + µα(µ − mh)(d − b), ρ4 = µam. T(x̄, ȳ) = L11 (x̄, ȳ) + L22 (x̄, ȳ), (2.2)
Note that all the coefficients are always positive for µ > mh
except ρ3 . A necessary (but not sufficient) condition for ρ3 to be neg- D(x̄, ȳ) = L11 (x̄, ȳ)L22 (x̄, ȳ) − L12 (x̄, ȳ)L21 (x̄, ȳ). (2.3)
ative is that b > d. The above biquadratic polynomial equation in y
has at most two positive roots according to Descartes’ rule of signs. Thus, the coexistence equilibrium Ē = (x̄, ȳ) is locally asymptotically
Thus, the model (1.4) exhibits at most two coexistence equilibria. In stable (LAS) if T(x̄, ȳ) < 0 and D(x̄, ȳ) > 0.
Fig. 1, we draw non-trivial prey and predator nullclines for different The local stability of extinction and predator-free equilibria is
values of k to visualize the existence of interior equilibrium points, straightforward, as the eigenvalues of the corresponding variational
where all other parameter values are fixed. We observe that below matrix are located along the diagonal. In Table I, we summarize all
a threshold value of k, both prey and predator nullclines intersect the steady states, their feasibility, and local stability of the model
at exactly two points; the nullclines do not cross above a threshold system (1.4).

Chaos 31, 123134 (2021); doi: 10.1063/5.0067046 31, 123134-3


Published under an exclusive license by AIP Publishing
Chaos ARTICLE scitation.org/journal/cha

TABLE I. Existence conditions and stability criteria of different steady states of the For the existence of a pair of purely imaginary roots of the
model (1.4). above equation, we must have D(x̄(k),
p ȳ(k)) > 0 at k = kh . Thus, we
consider λ1,2 = ±iν0 , where ν0 = D(x̄(k), ȳ(k))|k=kh .
Steady state Feasibility Stability Now, we check the transversality condition that confirms that
the eigenvalues cross the imaginary axis of the complex plane with
E0 = (0, 0)  Always exists LAS if b < d
non-zero speed. For this, let at any neighboring point k of kh , the
Ek = b−d ,0 b>d Always LAS if it
a eigenvalues of the variational matrix are λ1,2 = ξ(k) ± iν(k), where
exists
Coexistence state µ > mh and ρ 3 < 0 LAS if T(x̄, ȳ) < 0 T(x̄(k), ȳ(k))
Ē = (x̄, ȳ) and D(x̄, ȳ) > 0 ξ(k) =
2
and
r
In Fig. 2, we plot the densities of the prey population at equi- T2 (x̄(k), ȳ(k))
librium by changing the parameter k with two different values of ν(k) = D(x̄(k), ȳ(k)) − .
4
b(= 0.9/1.4), where other parameter values are fixed at d = 0.3,
a = 0.4, α = 0.2, h = 2, µ = 0.7, and m = 0.1. Note that the Now,
system has at most two coexisting (positive) equilibrium points.  
d 1 d
The equilibrium point that moves along the magenta curve always ξ(k)
= T(x̄(k), ȳ(k)) .
exhibits saddle behavior. In the present work, the dynamics of the dk k=kh 2 dk k=kh
system (1.4) are explored with the help of numerical simulations
d
around the equilibrium point, where the equilibrium point moves Thus, the transversality condition is satisfied if dk T(x̄(k), ȳ(k))
along the green curve. A nice animation is also made available on the 6= 0 at k = kh . Therefore, the system (1.4) undergoes Hopf bifurca-
electronic version of this article (see the additional high-resolution tion at k = kh . 
movies in the supplementary material), which explains the existence
and stability of the equilibrium points. B. Direction of the Hopf bifurcation
At this point, we analyze the stability properties of the bifurcat-
A. Hopf bifurcation ing periodic solutions originating from the coexisting equilibrium
Nonlinear mathematical models of interacting populations point Ē = (x̄, ȳ) via the Hopf bifurcation. For this, we calculate the
show rich and complex dynamical behaviors even when system 1st Lyapunov coefficient by obtaining the normal form of the Hopf
complexity is low (two or three species). Oscillating behavior is the bifurcation. As mentioned earlier, at k = kh , the variational matrix
most frequent dynamical property in population dynamics. Oscil- LĒ has eigenvalues λ1,2 = ±iν0 .
lating behavior or the existence of a limit cycle leads to the Hopf Now, to obtain the normal form, first, we perform a linear
bifurcation of the system. The Hopf bifurcation is defined as the coordinate shift by introducing two new variables w1 and w2 ,
appearance or disappearance of a periodic orbit through a local
change in the stability properties of an equilibrium point. Here, we w1 = x − x̄, w2 = y − ȳ.
explore the possibility of occurrence of the Hopf bifurcation and
Then, the equilibrium point Ē = (x̄, ȳ) of the system (1.4)
its direction around an arbitrary coexistence equilibrium Ē = (x̄, ȳ)
transforms into the origin, and the system (1.4) becomes
with respect to bifurcating parameter k.
The following theorem gives the necessary and sufficient con- 
b(w1 + x̄)
ditions for Hopf bifurcation of the system (1.4).  dw

dt
1
= − d(w1 + x̄) − a(w1 + x̄)2



 1 + k(w2 + ȳ)
Theorem II.1. The necessary and sufficient conditions for which α(w1 + x̄)(w2 + ȳ)2


the system (1.4) undergoes the Hopf bifurcation at k = kh around − ,
 1 + hα(w1 + x̄)(w2 + ȳ)
the coexistence equilibrium point Ē are that D(x̄(k), ȳ(k)) > 0 and

 2
 dw2 = µα(w1 + x̄)(w2 + ȳ) − m(w + ȳ).

d

T(x̄(k), ȳ(k)) 6= 0 at k = kh .

dk  dt 2
1 + hα(w1 + x̄)(w2 + ȳ)
Proof. The characteristic equation of the variational matrix LĒ
is given by Expanding Taylor’s series of the above system at (w1 , w2 )
= (0, 0) up to terms of order 3 produces the following system:
λ2 − T(x̄(k), ȳ(k))λ + D(x̄(k), ȳ(k)) = 0,
ẇ1 = e10 w1 + e01 w2 + e20 w21 + e11 w1 w2

where T and D are trace and determinant values as given in (2.2)


+e02 w22 + e30 w31 + e21 w21 w2 + e12 w1 w22


and (2.3), respectively. For the Hopf bifurcation, the matrix LĒ must



+e03 w32 + O(|z|4 ),

have a pair of purely imaginary eigenvalues. At a critical value

(2.4)
k = kh , T(x̄(k), ȳ(k)) = 0. Therefore, the above characteristic 2
ẇ2 = f10 w1 + f01 w2 + f20 w1 + f11 w1 w2

equation becomes 


 +f02 w22 + f30 w31 + f21 w21 w2 + f12 w1 w22


λ2 + D(x̄(k), ȳ(k))|k=kh = 0. +f03 w32 + O(|z|4 ),

Chaos 31, 123134 (2021); doi: 10.1063/5.0067046 31, 123134-4


Published under an exclusive license by AIP Publishing
Chaos ARTICLE scitation.org/journal/cha

FIG. 2. Figure shows the variations of


coexistence equilibrium densities of the
prey population of the system (1.4) with
respect to the fear level k. In the left figure,
b = 0.9, and in the right figure, b = 1.4,
and the remaining parameter values
are d = 0.3, a = 0.4, α = 0.2, h = 2,
µ = 0.7, and m = 0.1.

where the upper dot denotes the derivative with respect to time t, where W = (w1 , w2 )T and
and

e20 w21 + e11 w1 w2 + e02 w22 + e30 w31


 
hα 2 ȳ3
 
e10 e01  
 +e21 w21 w2 + e12 w1 w22 + e03 w32 
= LĒ , e20 = −a + , M1
f10 f01 (1 + hα x̄ȳ)3 M=  f20 w2 + f11 w1 w2 + f02 w2 + f30 w3  .
= 
M2 1 2 1
bk 2αȳ +f21 w21 w2 + f12 w1 w22 + f03 w32
e11 = − 2
− ,
(1 + kȳ) (1 + hα x̄ȳ)3
bk2 x̄ α x̄ The eigenvector v̄ of variational matrix LĒ corresponding to the
e02 = − ,
(1 + kȳ)3 (1 + hα x̄ȳ)3 eigenvalue iν0 at k = kh is v̄ = (e01 , iν0 − e10 )T . Now, define
h2 α 3 ȳ4 3hα 2 ȳ2
e30 = − 4
, e21 = ,  
(1 + hα x̄ȳ) (1 + hα x̄ȳ)4 R = (Re(v̄), −Im(v̄)) =
e01 0
.
−e10 −ν0
bk2 α(1 − 2hα x̄ȳ)
e12 = 3
− ,
(1 + kȳ) (1 + hα x̄ȳ)4
Let W = RV or V = R−1 W, where V = (v1 , v2 )T . Under this
bk3 x̄ hα 2 x̄2
e03 = − + , transformation, system (2.5) becomes
(1 + kȳ)4 (1 + hα x̄ȳ)4
µhα 2 ȳ3 2µαȳ V̇ = (R−1 LĒ R)V + R−1 M(RV).
f20 = − , f11 = ,
(1 + hα x̄ȳ)3 (1 + hα x̄ȳ)3
µα x̄ µh2 α 3 ȳ4 This can be written as
f02 = 3
, f30 = ,
(1 + hα x̄ȳ) (1 + hα x̄ȳ)4
3µhα 2 ȳ2
      1 
µα(1 − 2hα x̄ȳ) v˙1 0 −ν0 v1 F (v1 , v2 ; k = kh )
f21 = − , f12 = , = + 2 , (2.6)
(1 + hα x̄ȳ)4 (1 + hα x̄ȳ)4 v˙2 ν0 0 v2 F (v1 , v2 ; k = kh )
µhα 2 x̄2
f03 = − .
(1 + hα x̄ȳ)4 where F1 and F2 are nonlinear in v1 and v2 and are given by

Neglecting the higher-order terms of degree 4 and above, the 1


F1 (v1 , v2 ; k = kh ) = M1 ,
system (2.4) can be written in the vector form as e01
1
F2 (v1 , v2 ; k = kh ) = − (e10 M1 + e01 M2 ),
Ẇ = LĒ W + M(W), (2.5) ν0 e01

Chaos 31, 123134 (2021); doi: 10.1063/5.0067046 31, 123134-5


Published under an exclusive license by AIP Publishing
Chaos ARTICLE scitation.org/journal/cha

with
M1 = (e20 e201 − e11 e01 e10 + e02 e210 )v21 + ν0 (2e02 e10 − e11 e01 )v1 v2
+ ν02 e02 v22 + (e30 e301 − e21 e201 e10 + e12 e01 e210 − e03 e310 )v31
+ ν0 (2e12 e10 e01 − e21 e201 − 3e03 e210 )v21 v2
+ ν02 (e12 e01 − 3e03 e10 )v1 v22 − ν03 e03 v32 ,
M2 = (f20 e201 − f11 e01 e10 + f02 e210 )v21 + ν0 (2f02 e10 − f11 e01 )v1 v2
+ ν02 f02 v22 + (f30 e301 − f21 e201 e10 + f12 e01 e210 − f03 e310 )v31
+ ν0 (2f12 e01 e10 − f21 e201 − 3f03 e210 )v21 v2
+ ν02 (f12 e01 − 3f03 e10 )v1 v22 − ν03 f03 v32 .
Now, we calculate the 1st Lyapunov coefficient, based on the
normal form (2.6), which is given by [Wiggins,24 Eq. (20.2.14), p.
385]
1  1
Fv1 v1 v1 + F1v1 v2 v2 + F2v1 v1 v2 + F2v2 v2 v2

l1 =
16
1  1
Fv1 v2 F1v1 v1 + F1v2 v2 − F2v1 v2 F2v1 v1 + F2v2 v2
 
+
16ν0
−F1v1 v1 F2v1 v1 + F1v2 v2 F2v2 v2 ,


where all the partial derivatives are calculated at the bifurcation


point; i.e., (v1 , v2 ; k) = (0, 0; kh ).
By applying the result given in Wiggins’s book,24 the Hopf
bifurcation is supercritical if l1 < 0 and subcritical if l1 > 0. Remark
that, when l1 = 0, the system (1.4) exhibits a generalized Hopf
(GH) bifurcation. The GH bifurcation point separates branches

FIG. 4. Figure shows bifurcation diagrams for both populations [prey in (a) and
predator in (b)] of the system (1.4) with respect to prey’s birth rate (b). Depending
on the nature of stability, we break it into two groups: stable equilibrium (gray area)
and limit cycle (blue area). In the first portion of the gray region, predators cannot
survive moving the system into a predator-free equilibrium. At a sufficiently higher
birth rate, both species coexist (last part of the gray region) and in the blue region
populations show periodic oscillations. Here, other parameter values are chosen
from (3.1), and the initial condition is fixed at (0.5, 1.0).

of subcritical and supercritical Hopf bifurcations in the parameter


plane.
FIG. 3. Figure shows the phase portrait of the system (1.4) in the absence of fear
effect, i.e., at k = 0. Other parameter values are chosen from (3.1), and the initial
condition is fixed at (0.5, 1.0). Here, both prey and predator species coexist in III. NUMERICAL SIMULATIONS
an oscillatory manner forming a stable limit cycle around the unstable equilibrium
point (0.3889, 2.5712). In this section, we analyze the system (1.4) with the help of
various numerical tools and techniques. We use the renowned RK4

Chaos 31, 123134 (2021); doi: 10.1063/5.0067046 31, 123134-6


Published under an exclusive license by AIP Publishing
Chaos ARTICLE scitation.org/journal/cha

method for solving the differential equations. We solve the system


for at least T = 20 000 (enough for discarding initial fluctuations) for
plotting our results. The programs are written and run on MAT-
LAB. Our primary goal here is to extract more insight into the
system dynamics, which is almost impossible to get from an ana-
lytical approach. This also serves as a way to verify some of our
analytical results.
To start with, we first choose a biologically reasonable set of
parameter values given by the following:
b = 1.4, d = 0.3, a = 0.4, α = 0.2,
(3.1)
h = 2, µ = 0.7, m = 0.1.
We also fix the initial condition at (0.5, 1.0), unless otherwise
stated.
With parameter values chosen from (3.1) and in the absence of
fear effect (i.e., with k = 0), a phase-portrait diagram (see Fig. 3) of
the autonomous system (1.4) shows that the two interacting species
oscillate periodically forming a stable limit cycle around the unsta-
ble equilibrium point (0.3889, 2.5712). The birth rate (b) of the prey
(x) plays an important role in its survivability as well as its coexis-
tence with predator species. A bifurcation study of the system (1.4)
with respect to b, keeping the initial condition fixed at (0.5, 1.0),
reveals that the system shows three types of behaviors: prey-only
state, stable coexistence, and oscillatory coexistence. From Fig. 4,
we can draw a conclusion that the predator’s survivability is largely
dependent on prey’s birth rate (b), and only a sufficiently higher
value of b (namely, b > 0.87) makes the coexistence possible. For
0.87 < b < 1.011, the populations show stable coexistence and then
switch to limit cycle behavior via a Hopf bifurcation at b = 1.011.
We draw another phase portrait of the system (1.4) for b = 0.9
to reaffirm the stable coexistence of the two interacting species as

FIG. 6. Figure shows bifurcation diagrams of system (1.4) with respect to k, at


parameter values given in (3.1). (a) shows the change in the prey population, and
(b) shows the change in the predator population. Here, we observe that with a
gradual increase in the value of k, the amplitudes of periodic oscillations shrink
down and ultimately form a stable equilibrium via a Hopf bifurcation occurring at
k = 0.206 and thereafter remains stable. Further increase in k leads to predator’s
extinction.

indicated by the bifurcation diagrams (Fig. 4). Figure 5 shows that


among the two coexisting equilibrium states, (1.2831, 0.7794) and
(0.6471, 1.5454), the former one is saddle and the latter is spirally
stable (the corresponding variational matrix has a pair of complex
conjugate eigenvalues with negative real part).
FIG. 5. Figure shows the phase portrait of the system (1.4) reaffirming the stable
coexistence of the two interacting species at b = 0.9, keeping other parameters
Now, we move our focus to the effect of fear (k) in the sys-
fixed at those given in (3.1). tem (1.4). Fear could have both stabilizing and destabilizing effects.25
At first, we explore the situation when initially the system shows a

Chaos 31, 123134 (2021); doi: 10.1063/5.0067046 31, 123134-7


Published under an exclusive license by AIP Publishing
Chaos ARTICLE scitation.org/journal/cha

FIG. 8. Figure shows the bifurcation diagram of system (3.2) with respect to θ.
We have plotted the maximum peaks in blue and minimums in red. The system
shows chaotic dynamics along with some periodic and higher periodic windows.
Here, b = 1.4, k = 0, w = 1, and other parameters are chosen from (3.1).

A. The nonautonomous case


A large volume of literature exists in the study of a
mathematical model that describes the predator–prey relationship,
and in the last few decades, much progress has been seen in this
direction. Essentially, all these models consider within the frame-
work of an unvarying, deterministic environment. However, such

FIG. 7. Figure shows bifurcation diagrams of system (1.4) with respect to k, at


b = 0.9 and other parameter values chosen from (3.1). (a) shows the change
in prey population and (b) shows the change in predator population. Here, we
observe that with a gradual increase in the value of k, the stable equilibrium
remains stable and a further increase in k leads to predator’s extinction.

periodic oscillation. In Fig. 6, we have plotted two bifurcation dia-


grams of system (1.4) with respect to k, at parameter values given
in (3.1). It is observed that as we gradually increase the intensity of
fear, the amplitudes of periodic oscillations shrink down and ulti-
mately form a stable equilibrium via a Hopf bifurcation occurring at
k = 0.206 and thereafter remains stable. Further increase in k leads
FIG. 9. Figure shows the bifurcation diagram of system (3.2) with respect to θ.
to predator’s extinction. A similar study with b = 0.9 shows that the We have plotted the maximum peaks in blue and minimums in red. Here, b = 0.9,
stable equilibrium remains stable until further increase in k leads to k = 0, w = 1, and other parameter values are chosen from (3.1).
predator’s extinction (see Fig. 7).

Chaos 31, 123134 (2021); doi: 10.1063/5.0067046 31, 123134-8


Published under an exclusive license by AIP Publishing
Chaos ARTICLE scitation.org/journal/cha

FIG. 11. Figure shows isoperiodic regions in θ-k biparametric space of system
FIG. 10. Figure shows the bifurcation diagram of system (3.2) with respect to (3.2). Here, period means the periodicity of the peaks of respective time-series
k. We have plotted the maximum peaks in blue and minimums in red. Here, evaluations discarding initial fluctuations. EP denotes the equilibrium state. Apart
the system shows chaotic dynamics along with some periodic and higher peri- from equilibrium, we have presented regions of period 1 to period 18 in distinct
odic windows. Here, θ = 0.5, w = 1, and other parameter values are chosen colors and the rest in black color, which includes chaotic attractors. A sidebar rep-
from (3.1). resents the periods corresponding to each color. Here, we can also see some
shrimp-shaped structures and another structure that seems to be like Arnold-
tongue. During this simulation, we fixed our initial condition at (0.5, 1.0), and
other parameter values are chosen from (3.1).
an unvarying or constant environment cannot reflect the reality of
nature. Most natural environments are physically highly variable, reason, time-dependent predator–prey systems have been widely
and in response, the population’s birth and death rates, carrying investigated.
capacities, competition coefficient, and many other parameters that It is well accepted that temporal fluctuations in the phys-
characterize the deterministic model vary greatly in time.26 For this ical environment are major drivers of population fluctuations.

FIG. 12. Figure shows the largest (a) and the second-largest (b) Lyapunov exponents (LE) of system (3.2) in θ-k biparametric space. Other parameter values are chosen
from (3.1).

Chaos 31, 123134 (2021); doi: 10.1063/5.0067046 31, 123134-9


Published under an exclusive license by AIP Publishing
Chaos ARTICLE scitation.org/journal/cha

FIG. 13. Figure shows different versions of attractor-following isospike regions in θ-k biparametric space of system (3.2). All four subfigures are drawn in an attractor-following
manner. (a) is drawn by first solving system (3.2) for θ = 0 and for all k values (vertical line θ = 0). The last point of the solution is used as the initial condition for the next
value of θ. This process is repeated until we reach θ = 0.8 (horizontal forward). (b) is drawn changing θ backward from 0.8 to 0 (horizontal backward) following the same
process. Similarly, in (c), we first take the horizontal line k = 0 for all θ and then increase k until we reach k = 0.25 (vertical forward). In (d), we just change k backward from
0.25 to 0 (vertical backward). For the first iteration, we used (0.5, 1.0) as the initial state of system (3.2). Here, the last points of the solutions of each iteration are used as
the initial states of the next iteration. Other parameter values are chosen from (3.1).

Theoretical evidence to date indicates that many populations and Seasonal variation in birth rates is common among many
community patterns represent intricate interactions between biol- species. There could be many factors contributing to such fluc-
ogy and variation in the physical environment (for a brief review, tuations, for example, temperature, humidity, rainfall, change in
see Chesson27 and related papers in the same issue). When envi- daylight period, an abundance of food, etc.28–30 It may happen due to
ronmental fluctuation is taken under consideration, a model must some seasons being evolutionarily beneficial and favorable for giving
be nonautonomous and thus more difficult to analyze in general. birth and raising the young. According to most research, seasonal
However, one can take advantage of the properties of those varying breeding is driven by dynamics related to resource availability. Many
parameters by assuming that the parameters are periodic or almost wild mammals31–33 and birds29,34,35 show seasonal breeding. In North
periodic for seasonal reasons. America, during the spring and early summer months, when food is

Chaos 31, 123134 (2021); doi: 10.1063/5.0067046 31, 123134-10


Published under an exclusive license by AIP Publishing
Chaos ARTICLE scitation.org/journal/cha

FIG. 14. Figure shows the zoomed-in view of two different modes of the attractor following the isospike diagram in θ − k biparametric space. (a) is drawn in by taking the
vertical line θ = 0.65 and moving it backward (horizontal backward). (b) is drawn in by taking the horizontal line k = 0.06 and moving it backward (vertical backward). Other
parameter values are chosen from (3.1).

plentiful and it is not too cold, mule deer give birth.36,37 In savanna For b = 0.9 and k = 0, we draw another bifurcation diagram
ecosystems, the variation of resource availability is greatly depen- of system (3.2) with respect to θ (see Fig. 9). Here, we see that
dent on rainfall. Ryan et al.38 observed that African buffalos are more the system moves from a stable to periodic state at the very first
likely to give birth during the wet season of every year. In Canada, instant where θ becomes nonzero and remains the same except for
the song sparrow lays its eggs and hatches them between April and an increase in the amplitude of oscillation (due to an increase in θ ).
June every year.17 There is ample evidence of seasonal breeding in In Fig. 10, we draw the bifurcation diagram of the system (3.2)
marine and coastal ecosystems from several studies.39–41 with respect to the fear parameter k with θ = 0.5, w = 1, and other
We, therefore, take another step by introducing the periodic parameter values are chosen from (3.1). We see that with an ade-
birth rate into the system (1.4). Thus, we replace the birth rate b with quate increase in the strength of fear, the system (3.2) becomes
b + θ cos(wt) making system (1.4) into a nonautonomous system periodic from a chaotic state and remains there. One can also check
and get that beyond a certain level of fear, the predator population (y)
becomes extinct.
(b + θ cos wt)x αxy2

dx
− dx − ax2 − , Although bifurcation diagrams are very useful in revealing
 dt =


1 + ky 1 + hαxy the rich dynamics of a system such as (3.2), it is quite limited to
(3.2)
µαxy 2 only showing changes for a single parameter value. Variation of
 dy =


dt
− my. one more parameter gives us extra advantages and often unfolds
1 + hαxy
many organized periodic structures in the chaotic regime that can-
Here, θ is the amplitude of variation and w represents the corre- not be seen otherwise.42 More parameters can be varied at once,
sponding frequency. We also set w = 1 for all subsequent simula- but it is very hard to visualize on the plane of a paper. For this,
tions of system (3.2). we draw an isospike diagram of the system (3.2) in θ − k bipara-
Now, we simulate the nonautonomous system (3.2) by first metric space keeping other parameter values fixed at those given in
changing θ and keeping k = 0 and choosing other parameter values (3.1). (Fig. 11). We also fix our initial condition at (0.5, 1.0) through-
from (3.1). Figure 8 shows the corresponding bifurcation diagram out the simulation. Here, period means the periodicity of the peaks
of system (3.2). Here, the blue dots represent the local maximums (spikes) of the time-series data for each pair of parameter values.
of the time-series data, and the red ones represent the local mini- We have discarded the first T = 20 000 time-series data to remove
mums. We observe that the system moves from the initial periodic initial fluctuations. In this rectangular region ([0, 0.8] × [0, 0.25]),
state to chaotic oscillations along with some periodic windows. We filled with 600 × 600 colored dots, each color is associated with
have added a colorbar below along with a sidebar on the right to unique periodicity mentioned in the sidebar on the right. Here, EP
represent the periodicity of the peaks corresponding to each color. represents the equilibrium state. Chaotic and periodic regions with
Here, black dots are associated with the periodicity higher than 18, periodicity higher than 18 are represented in black. One can now
which also includes the chaotic regimes. easily correlate Figs. 8 and 10 with Fig. 11 and verify the results.

Chaos 31, 123134 (2021); doi: 10.1063/5.0067046 31, 123134-11


Published under an exclusive license by AIP Publishing
Chaos ARTICLE scitation.org/journal/cha

FIG. 15. Figure shows sections of time-series evaluations and corresponding basins of attraction for three different pairs of (θ, k) values for which system (3.2)
shows multistability between multiple coexisting attractors. (a) and (b) corresponding to (θ, k) = (0.4, 0.053) show the multistability between attractors of peri-
odicities 1, 2, and 3. Similarly, (c) and (d) corresponding to (θ, k) = (0.4, 0.048) show the multistability between attractors of periodicities 1, 2, and 6, and (e)
and (f) corresponding to (θ, k) = (0.452, 0.048) show the multistability between attractors of periodicities 1, 2, and 12. Here, other parameter values are chosen
from (3.1).

Chaos 31, 123134 (2021); doi: 10.1063/5.0067046 31, 123134-12


Published under an exclusive license by AIP Publishing
Chaos ARTICLE scitation.org/journal/cha

FIG. 16. Figure shows the largest Lyapunov diagram of two different segments from θ − k biparametric space (Fig. 12) containing different organized periodic structures of
system (3.2). (a) shows shrimp-shaped structures, and (b) shows Arnold tongues. Here, other parameter values are chosen from (3.1).

Here, we see many organized periodic structures including shrimp- of the line segment. We then use the last points of the solutions
shaped periodic islands inside the chaotic regime. We also see the as the initial states of the neighboring line segment and continue
overlapping of different periodic zones indicating the possible pres- this process until we reach the other boundary. Depending on the
ence of multistability between multiple coexisting attractors.43 We choice of starting line, we have four different variations presented in
discuss more about these phenomena in Subsection III A 1. Fig. 13. In Fig. 13(a), we choose the vertical line θ = 0 and proceed
Lyapunov exponents are very well-known measures of expo- in a forward direction (θ increasing), and in Fig. 13(b), we use the
nential convergence or divergence of infinitesimally close orbits. vertical line θ = 0.8 and move backward to 0 (θ decreasing). Simi-
In general, with very few exceptions, a positive value of the largest larly, in Figs. 13(c) and 13(d), we, respectively, choose the horizontal
Lyapunov exponent (LLE) indicates the chaotic nature of a system. lines k = 0 and k = 0.25 and proceed in forward (k increasing) and
Similarly, a negative value of LLE represents a stable equilibrium backward directions (k decreasing), correspondingly.
state. Regular oscillations give zero LLE. Although very useful, the From the four subfigures in Fig. 13, one can easily point
computational cost of calculating the Lyapunov spectrum of a sys- out their color differences and distinguish some of the prominent
tem such as (3.2) for a large set of parameter values is huge. Here, multistability zones. To take a closer look, we redraw the region
we draw the largest [Fig. 12(a)] and second-largest [Fig. 12(b)] enclosed by the white dotted rectangle ([0.35, 0.65] × [0.016, 0.06]
Lyapunov exponents (LE) of system (3.2) in θ − k biparametric in θ − k biparametric space) in Fig. 13(b) (horizontal backward
space ([0, 0.8] × [0, 0.25]). We used Wolf’s algorithm involving version) and Fig. 13(d) (vertical backward version) and display in
variational equations. This involves discarding of first T = 20 000 Fig. 14. Here, we can clearly observe the overlapping of multiple
time-series data to make sure the absence of any initial fluctuations pairs of coexisting attractors of different periodicities. These differ-
and averaging relative exponential divergence rates of nearby tra- ences strengthen our claim for the presence of multistability in our
jectories for another T = 10 000. A side-by-side comparison with system (3.2).
Fig. 13 and some easy time-series evaluations reveal that the regions To be accurate, we further pick three (θ , k) pairs from three
where the system has positive LLE shows chaotic nature. different multistable regions displayed in Fig. 14 and explore their
corresponding time-series evaluations for three different sets of
initial conditions. We also draw their basins of attraction. Those
1. Multistability
three (θ , k) pairs are (0.4, 0.053) from a red-green overlapping
To confirm the multistable nature of our system (3.2) in some zone, (0.4, 0.048) from a deep green-green overlapping zone, and
certain zones in the abovesaid biparametric space, we draw four dif- (0.452, 0.048) from a violet-green overlapping zone. In Fig. 15, we
ferent versions of Fig. 11 using an attractor-following technique. In plot the time-series evolutions on the first column, and correspond-
this technique, we first choose a vertical or horizontal line segment ing basins of attraction are placed on their right. Figures 15(a)
from the four boundary lines of the rectangular area and solve the and 15(b) corresponding to (θ , k) = (0.4, 0.053) show the multi-
system for each pair of parameter values on the line segment. In stability between attractors of periodicities 1, 2, and 3. Similarly,
this step, we use a fixed initial state (here [0.5, 1.0]) for each point Figs. 15(c) and 15(d) corresponding to (θ , k) = (0.4, 0.048) show

Chaos 31, 123134 (2021); doi: 10.1063/5.0067046 31, 123134-13


Published under an exclusive license by AIP Publishing
Chaos ARTICLE scitation.org/journal/cha

FIG. 17. Figure shows self-similar patterns in different zoomed-in versions of the fractal basin presented in Fig. 15(b) corresponding to (θ, k) = (0.4, 0.0453). (a) represents
the area inside the white dotted rectangle in Fig. 15(b). (b) is the zoomed-in view of (a). (c) represents the area inside a smaller rectangle (at the left), and (b) and (d) represent
the larger one (at the right). Here, other parameter values are chosen from (3.1).

the multistability between attractors of periodicities 1, 2, and 6, and different rectangular sections (violet-bordered boxes) of the (θ , k)
Figs. 15(e) and 15(f) corresponding to (θ , k) = (0.452, 0.048) show plane presented in Fig. 12. In Fig. 16(a), we scatter plot the LLE for
the multistability between attractors of periodicities 1, 2, and 12. a grid of 500 × 500 equispaced pairs of parameter values (θ , k) ∈
Since the birth rate of the prey varies periodically with time, the sys- [0.37, 0.4] × [0.03, 0.04]. Here, we observe multiple shrimp-shaped
tem has no stable equilibrium point. All the subfigures presented in organized periodic structures inside the chaotic regime. It is inter-
Fig. 15 are drawn in terms of the prey population (x). In the blue esting to note that one can explore the period-bubbling transition
region, predators cannot survive and prey oscillates due to the peri- to chaos near the inner boundaries of the shrimp.42,44 Similarly,
odic nature of its birth rate. The readers may notice the self-similar for (θ , k) ∈ [0.095, 0.225] × [0.065, 0.095], we see the presence of
pattern of the basin of attraction in all three cases. We explore it Arnold tongues in Fig. 16(b). The colorbars represent the LLE values
further in Sec. III A 3. for colors attached to them.
2. Organized periodic structures
In Fig. 12, we observe different organized periodic 3. Fractal basin
structures. Among them, we found shrimp-shaped structures and In all three basins of attraction presented in Fig. 15, we see the
Arnold tongues particularly noteworthy to report. We selected two presence of self-similar (fractal) patterns. To demonstrate this, we

Chaos 31, 123134 (2021); doi: 10.1063/5.0067046 31, 123134-14


Published under an exclusive license by AIP Publishing
Chaos ARTICLE scitation.org/journal/cha

pick a rectangular region from Fig. 15(b) (indicated by the white conclude that the nonautonomous ecological systems are inherently
dotted area) and draw it again on Fig. 17(a). Figure 17(b) shows complex and exhibit intricate dynamical behaviors.
a zoomed-in view of the white rectangular area at the center of
Fig. 17(a). It displays a similar pattern. We again plot two dif- SUPPLEMENTARY MATERIAL
ferent rectangular sections of the basin of attraction of Fig. 17(b)
in Figs. 17(c) and 17(d) (in the left to right order). Both exhibit In the supplementary material, we include two movies show-
self-similar patterns. This kind of phenomenon is quite rare in ing the qualitative change in the dynamics of the model (1.4) when
continuous-time models. Here, each area is divided into 600 × 600 the strength of fear (k) is varied. We plot the nullclines of both
equispaced grid points, and then the system (3.2) is solved for all species in the first quadrant to show the coexisting equilibrium
360 000 pairs of initial conditions and for time T = 30 000. We ana- points. Throughout the animations, the red curve represents the
lyzed the system behavior for the last T = 5000 (i.e., discarding first prey nullcline, and the green one represents the predator nullcline.
T = 25 000 time-series data) to detect the periodicity of the attractors We added the vector fields corresponding to each value of k by using
and plot them in their respective colors. the “quiver” command in MATLAB. We also plotted solution curves
for three arbitrarily chosen initial conditions to show where the solu-
tions converge. This gives a clear understanding of the stability of the
coexisting equilibrium points.
IV. CONCLUSION The movie named nullcline1 shows nullclines of the system
In the present study, we have investigated the impact of the fear (1.4) for different values of k, ranging from 0 to 1 where b = 1.4
effect in a predator–prey model. We considered that due to the fear and other parameter values are chosen from (3.1). The other movie,
of predation risk, the birth rate of the prey population reduces in named nullcline2, does the same except for b = 0.9 and varying k in
both autonomous and nonautonomous versions of the model. The the range of 0–0.1. In both cases, we see that there are at most two
whole investigation of the system is mainly centered on two impor- coexisting equilibrium states in the system (1.4), which are repre-
tant factors, namely, the birth rate of prey and the fear effect. Using sented by the intersection of the nullclines. Among these two, one
phase-portrait, bifurcation, isospike, and Lyapunov exponent dia- is always saddle. In the present work, the dynamics of the system
grams, we explored the rich and complex dynamical behaviors of (1.4) are explored with the help of numerical simulations around
the system. First, we explored how changes in seasonal fluctuations the non-saddle equilibrium point.
in the birth rate influenced the population cycles. In contrast to the
autonomous system, which shows periodic behaviors, the nonau- ACKNOWLEDGMENTS
tonomous system exhibits chaotic oscillations. Seasonal variations The authors are thankful to the anonymous reviewers and the
in the level of birth rate lead to higher periodic and chaotic oscilla- handling editor for their valuable comments and suggestions, which
tions and produce dynamical complexities in the ecosystem. It was helped us improve the quality of the manuscript. M. Hossain is
also noticed that increased fear of predation risk potentially sup- grateful to the Department of Science and Technology (DST), India,
presses the irregular oscillations in the system. We also explored for providing financial support under the INSPIRE Fellowship
the global behavior of the systems in the biparameter space (birth (No. IF-170522) program.
rate and fear effect) by varying the control parameters. We found
various types of several periodic islands in the chaotic regimes.
AUTHOR DECLARATIONS
Therefore, unexpected shifts in the dynamics of the system can hap-
pen in a small variation of the control parameters. The well-known Conflict of Interest
organized structures such as shrimp-shaped structures and Arnold’s The authors have no conflicts to disclose.
tongues are detected in the biparameter space. It is to be noted
that period-bubbling transition to chaos can be found due to the
Author Contributions
existence of shrimp-shaped structures. Furthermore, in the isospike
diagram, overlapping of multiple periodic windows is found, and All authors contributed equally to this work.
multistability between topologically different coexisting attractors is
detected. The complexities of the multistable attractors are exam- DATA AVAILABILITY
ined by drawing the basin of attractions. We drew basins of several
The data that support the findings of this study are available
coexisting attractors and found fractal basin boundaries. Here, the
within the article.
parameter values and the initial population sizes of prey and preda-
tor play a vital role in the dynamics of the system, and the solution
trajectories can converge to different non-equivalent attractors. It is REFERENCES
observed that the persistence of the predator population also highly
1
depends on the initial population densities. Therefore, small changes J. Duarte, C. Januário, N. Martins, and J. Sardanyés, “Chaos and crises in a model
in the parameters, perturbations, or environmental fluctuations can for cooperative hunting: A symbolic dynamics approach,” Chaos 19, 043102
(2009).
have drastic outcomes in the oscillations of the populations. In the 2
S. Pal, N. Pal, S. Samanta, and J. Chattopadhyay, “Effect of hunting cooperation
present model system, the fear effect has a stabilizing role on the and fear in a predator-prey model,” Ecol. Complex. 39, 100770 (2019).
autonomous system, whereas in the nonautonomous system, the 3
C. S. Holling, “Some characteristics of simple types of predation and parasitism,”
fear effect drives the system toward regularity. In summary, we Can. Entomol. 91, 385–398 (1959).

Chaos 31, 123134 (2021); doi: 10.1063/5.0067046 31, 123134-15


Published under an exclusive license by AIP Publishing
Chaos ARTICLE scitation.org/journal/cha

4 26
R. Arditi and L. R. Ginzburg, “Coupling in predator-prey dynamics: Ratio- R. M. May, “Stability in randomly fluctuating versus deterministic environ-
dependence,” J. Theor. Biol. 139, 311–326 (1989). ments,” Am. Nat. 107, 621–650 (1973).
5 27
P. A. Abrams and L. R. Ginzburg, “The nature of predation: Prey dependent, P. Chesson, “Understanding the role of environmental variation in population
ratio dependent or neither?,” Trends Ecol. Evol. 15, 337–341 (2000). and community dynamics,” Theor. Popul. Biol. 3, 253–254 (2003).
6 28
J. R. Beddington, “Mutual interference between parasites or predators and its F. E. Wilson and R. S. Donham, “Daylength and control of seasonal reproduc-
effect on searching efficiency,” J. Anim. Ecol. 44, 331–340 (1975). tion in male birds,” in Processing of Environmental Information in Vertebrates,
7
D. L. DeAngelis, R. Goldstein, and R. O’Neill, “A model for tropic interaction,” Proceedings in Life Sciences, edited by M. H. Stetson (Springer, New York, 1988),
Ecology 56, 881–892 (1975). pp. 101–119.
8 29
R. Arditi and H. R. Akçakaya, “Underestimation of mutual interference of J. F. Cockrem, “Timing of seasonal breeding in birds, with particular reference
predators,” Oecologia 83, 358–361 (1990). to New Zealand birds,” Reprod. Fertil. Dev. 7, 1–19 (1995).
9 30
R. Arditi, N. Perrin, and H. Saïah, “Functional responses and heterogeneities: An F. H. Bronson, “Climate change and seasonal reproduction in
experimental test with cladocerans,” Oikos 60, 69–75 (1991). mammals,” Philos. Trans. R. Soc. Lond. B: Biol. Sci. 364, 3331–3340
10
C. Cosner, D. L. DeAngelis, J. S. Ault, and D. B. Olson, “Effects of spatial (2009).
31
grouping on the functional response of predators,” Theor. Popul. Biol. 56, 65–75 G. A. Lincoln, “Biology of seasonal breeding in deer,” in The Biology of Deer,
(1999). edited by R. D. Brown (Springer, New York, 1992), pp. 565–574.
11 32
B. L. Partridge, J. Johansson, and J. Kalish, “The structure of schools of giant B. Malpaux, C. Viguié, D. C. Skinner, J. C. Thiéry, J. Pelletier, and P. Chemineau,
bluefin tuna in Cape Cod Bay,” Environ. Biol. Fishes 9, 253–262 (1983). “Seasonal breeding in sheep: Mechanism of action of melatonin,” Anim. Reprod.
12
A. Sih, “Optimal behavior: Can foragers balance two conflicting demands?,” Sci. 42, 109–117 (1996).
33
Science 210, 1041–1043 (1980). H. Dardente, D. Lomet, V. Robert, C. Decourt, M. Beltramo, and M.-T. Pellicer-
13
J. F. Gilliam and D. F. Fraser, “Habitat selection under predation hazard: Test Rubio, “Seasonal breeding in mammals: From basic science to applications and
of a model with foraging minnows,” Ecology 68, 1856–1862 (1987). back,” Theriogenology 86, 324–332 (2016).
14 34
J. S. Brown, J. W. Laundré, and M. Gurung, “The ecology of fear: Optimal P. J. Sharp, “Photoperiodic regulation of seasonal breeding in birds,” Ann. N. Y.
foraging, game theory, and trophic interactions,” J. Mammal. 80, 385–399 (1999). Acad. Sci. 1040, 189–199 (2005).
15 35
S. L. Lima and L. M. Dill, “Behavioral decisions made under the risk of G. Passuni, C. Barbraud, A. Chaigneau, H. Demarcq, J. Ledesma, A. Bertrand, R.
predation: A review and prospectus,” Can. J. Zool. 68, 619–640 (1990). Castillo, A. Perea, J. Mori, V. A. Viblanc, J. Torres-Maita, and S. Bertrand, “Sea-
16
J. M. Jeschke, “Density-dependent effects of prey defenses and predator sonality in marine ecosystems: Peruvian seabirds, anchovy, and oceanographic
offenses,” J. Theor. Biol. 242, 900–907 (2006). conditions,” Ecology 97, 182–193 (2016).
17 36
L. Y. Zanette, A. F. White, M. C. Allen, and M. Clinchy, “Perceived predation R. T. Bowyer, “Timing of parturition and lactation in southern mule deer,”
risk reduces the number of offspring songbirds produce per year,” Science 334, J. Mammal. 72, 138–145 (1991).
37
1398–1401 (2011). K. L. Monteith, V. C. Bleich, T. R. Stephenson, B. M. Pierce, M. M. Con-
18
J. P. Suraci, M. Clinchy, L. M. Dill, D. Roberts, and L. Y. Zanette, “Fear of large ner, J. G. Kie, and R. T. Bowyer, “Life-history characteristics of mule deer:
carnivores causes a trophic cascade,” Nat. Commun. 7, 10698 (2016). Effects of nutrition in a variable environment,” Wildl. Monogr. 186, 1–62
19
X. Wang, L. Zanette, and X. Zou, “Modelling the fear effect in predator–prey (2014).
38
interactions,” J. Math. Biol. 73, 1179–1204 (2016). S. Ryan, C. Knechtel, and W. Getz, “Ecological cues, gestation length, and
20
P. Panday, N. Pal, S. Samanta, and J. Chattopadhyay, “Stability and bifurcation birth timing in African buffalo (Syncerus caffer),” Behav. Ecol. 18, 635–644
analysis of a three-species food chain model with fear,” Int. J. Bifurcation Chaos (2007).
39
28, 1850009 (2018). T. McClanahan, “Seasonality in East Africa’s coastal waters,” Mar. Ecol. Prog.
21
J. Wang, Y. Cai, S. Fu, and W. Wang, “The effect of the fear factor on the dynam- Ser. 44, 191–199 (1988).
40
ics of a predator-prey model incorporating the prey refuge,” Chaos 29, 083109 A. Clarke, “Seasonality in the Antarctic marine environment,” Comp. Biochem.
(2019). Physiol. Part B: Comp. Biochem. 90, 461–473 (1988).
22 41
S. Samaddar, M. Dhar, and P. Bhattacharya, “Effect of fear on prey–predator R. Coma, M. Ribes, J.-M. Gili, and M. Zabala, “Seasonality in coastal benthic
dynamics: Exploring the role of prey refuge and additional food,” Chaos 30, ecosystems,” Trends Ecol. Evol. (Amst.) 15, 448–453 (2000).
42
063129 (2020). M. Hossain, N. Pati, S. Pal, S. Rana, N. Pal, and G. Layek, “Bifurcations and
23
M. Hossain, N. Pal, S. Samanta, and J. Chattopadhyay, “Fear induced stabi- multistability in a food chain model with nanoparticles,” Math. Comput. Simul.
lization in an intraguild predation model,” Int. J. Bifurcation Chaos 30, 2050053 190, 808–825 (2021).
43
(2020). N. Pati, S. Garai, M. Hossain, G. Layek, and N. Pal, “Fear induced mul-
24
S. Wiggins, Introduction to Applied Nonlinear Dynamical Systems and Chaos, tistability in a predator-prey model,” Int. J. Bifurcation Chaos 31, 2150150
2nd ed. (Springer Science & Business Media, New York, 2003). (2021).
25 44
S. Pal, S. Majhi, S. Mandal, and N. Pal, “Role of fear in a predator–prey model N. Pati, G. Layek, and N. Pal, “Bifurcations and organized structures in a
with Beddington–DeAngelis functional response,” Z. Naturforsch., A 74, 581–595 predator-prey model with hunting cooperation,” Chaos, Solitons Fractals 140,
(2019). 110184 (2020).

Chaos 31, 123134 (2021); doi: 10.1063/5.0067046 31, 123134-16


Published under an exclusive license by AIP Publishing

You might also like