Journal of Mathematical Analysis and Applications 258, 733᎐754 Ž2001.

doi:10.1006rjmaa.2001.7514, available online at http:rrwww.idealibrary.com on

Analysis of a Three Species Eco-Epidemiological Model1

Yanni Xiao and Lansun Chen

Academy of Mathematics & System Sciences, Academia Sinica,

Beijing 100080, P.R. China
E-mail: xyn@math03.math.ac.cn, lschen@math08.math.ac.cn

Submitted by Horst R. Thieme

Received September 13, 2000

This paper formulates and analyzes a predator᎐prey model with disease in the
prey. Mathematical analyses of the model equations with regard to invariance of
nonnegativity, boundedness of solutions, nature of equilibria, permanence, and
global stability are analyzed. It is also shown that for some parameter values, the
positive equilibrium is asymptotically stable, but for other parameter values, it
is unstable and a surrounding periodic solution appears by Hopf bifurcation.
A concluding discussion with numerical simulation is then presented. 䊚 2001
Academic Press
Key Words: predator᎐prey model; global stability; permanence; Hopf bifurcation.

1. INTRODUCTION

After the pioneering work of Kermack᎐McKendrick on susceptible᎐

infective᎐removal᎐susceptible ŽSIRS., epidemiological models have re-
ceived much attention from scientists. Relevant references in this context
are also vast and we shall mention here a survey paper w1x and some books
Žsee w2᎐4x, to mention a few.. Assuming that the species has no relation
with other species Žthat is to say, these epidemic models were formulated
to describe only the spread of the disease., they obtained some threshold
results.
In the natural world, however, species do not exist alone, it is of more
biological significance to study the persistence-extinction threshold of each
population in systems of two or more interacting species subjected to
parasitism. But little attention has been paid so far to merge these two

1
This work is supported by the National Natural Science Foundation of China.

733
0022-247Xr01 $35.00
Copyright 䊚 2001 by Academic Press
All rights of reproduction in any form reserved.
734 XIAO AND CHEN

areas of research w5᎐7x. In order to study the influence of disease on an

environment where two or more interacting species are present.
We consider a three species, eco-epidemological system, namely, sound
prey Žsusceptible ., infected prey Žinfective., and predator. We consider the
case where the predator mainly eats the infected prey. This is in accor-
dance with the fact that the infected individuals are less active and can be
caught more easily, or the behavior of the prey is modified such that they
live in parts of the habitat which are accessible to the predator Žfish and
aquatic snails staying close to water surface, snails staying on the top of
the vegetation rather than under the plant cover. w8x. Peterson and Page
w9x have indicated wolf attacks on moose are more often successful if the
moose is heavily infected by ‘‘Echinococcus granulosus.’’
We have two populations:
1. The prey, whose total population density is denoted by N.
2. The predator, whose population density is denoted by Y.
We make the following assumptions:
Ž H1 . In the absence of disease the prey population density grows
according to a logistic curve with carrying capacity K Ž K ) 0., with an
intrinsic birth rate constant r Ž r ) 0..
dN N
s rN 1 y
ž / . Ž 1.1.
dt K
Ž H2 . In the presence of disease we assume that the total prey popula-
tion N is composed of two population classes: one is the class of suscepti-
ble prey, denoted by S, and the other is the class of infected prey, denoted
by I. Therefore, at any time t, the total density of prey population is

NŽ t . s SŽ t . q IŽ t . . Ž 1.2.
ŽNote: In the following we are always referring to population densities; we
may omit the word ‘‘density’’ for the sake of simplicity..
Ž H3 . We assume that only susceptible prey S are capable of reproduc-
ing with logistic law ŽEq. Ž1.1..; i.e., the infected prey I are removed by
death Žsay its death rate is a positive constant c . or by predation before
having the possibility of reproducing. However, the infective population I
still contributes with S to population growth toward the carrying capacity.
Ž H4 . We assume that the disease is spread among the prey population
only and the disease is not genetically inherited. The infected populations
do not recover or become immune. The incidence is assumed to be the
simple mass action incidence ␤ SI, where ␤ ) 0 is called the transmission
coefficient.
 ECO-EPIDEMIOLOGICAL MODEL ANALYSIS 735

Ž H5 . We assume that the predator eats only the infected prey with
predator response function ␩ Ž I .. The predator has a death rate constant
d Ž d ) 0.. The coefficient in conversing prey into predator is constant
k Ž0 - k F 1.. It is assumed that ␩ Ž I . is a positive, increasing, and bounded
function of I, and ␩ Ž0. s 0 Žsee w6, 10x..
From the above assumptions, the model equations are
dS SqI
s rS 1 y
ž / y ␤ SI
dt K
dI
s ␤ SI y cI y ␩ Ž I . Y Ž 1.3.
dt
dY
s Y Ž yd q k␩ Ž I . . .
dt
Chattopadhyay and Arino w6x have considered a predator᎐prey model with
disease in the prey. They assumed that the sound prey population grows
according to a logistic law involving the whole prey population; i.e.,
dS SqI
s rŽS q I. 1 y
ž / y ␤ SI y ␥ Ž S . Y ,
dt K
where ␥ Ž S . is the predator response function. They further assumed that
the intrinsic growth rate r large enough such that S q I approaches the
environment carrying capacity K, and using S q I s K, they reduced the
three-dimensional system to a two-dimensional system. Hence they studied
the local stability, extinction, and Hopf-bifurcation in a two-dimensional
system. By applying a Poincare ´ map they observed the connection between
the reduced and the original system. System Ž1.3. is also similar to the one
studied by Venturino w7x. However, there are two differences: Ži. Venturino
does not allow for density dependence in the host regulation except for the
disease. Žii. In our model we consider the effect of the predator response
function. So we believe that this is the first time that eco-epidemiology
model has been formulated and analyzed.
In analogy to the threshold result of epidemic theory Žsee, for example,
w11x., we obtain the threshold parameters which govern the development of
the disease or the growth of the predator population. Furthermore, we
discuss the permanence of the system and determine conditions for which
the system enters a Hopf-type bifurcation.
For the sake of simplicity, we put in dimensionless form the model
equations Ž1.3. by rescaling the variables on the carrying capacity value K ;
i.e.,
S I Y
ss , is , ys Ž 1.4.
K K K
736 XIAO AND CHEN

and then using as dimensionless time, ␻ s ␤ Kt. This leads to the dimen-
sionless equations
ds
s as Ž 1 y Ž s q i . . y si
d␻
di
s si y b 2 i y ␩ Ž i . y Ž 1.5.
d␻
dy
s y Ž yb1 q k␩ Ž i . . ,
d␻
where

r d c ␩ Ž iK .
as , b1 s , b2 s , s ␩Ž i. Ž 1.6.
␤K ␤K ␤K ␤K

are the dimensionless parameters. The initial condition for Eq. Ž1.5. may
3
be any point in the nonnegative orthant of Rq of R 3 , where by Rq3
we
mean
3
Rq s  Ž s, i , y . g R 3 : s G 0, i G 0, y G 0 4 .

For convenience, in the following we replace ␻ by t for the dimensionless

time.

2. PRELIMINARIES

Positi¨ e In¨ ariance

Let us put Eq. Ž1.5. in a vector form by setting

X s col Ž s, i , y . g R 3 Ž 2.1.
F1 Ž X . as Ž 1 y Ž s q i . . y si

 0
F Ž X . s F2 Ž X .
F3 Ž X .
s si y b 2 i y ␩ Ž i . y
yb1 y q k␩ Ž i . y
,
0 Ž 2.2.

where F: R 3 ª R 3 and F g C⬁Ž R 3 .. Then Eq. Ž1.5. becomes

Ẋ s F Ž X . , Ž 2.3.
where ⭈s dtd and with X Ž0. s X 0 g Rq3
. It is easy to check in Eq. Ž2.2. that
whenever choosing X g Rq such that x i s 0, then Fi Ž x .< x i Ž t .s0 G 0, i s
3

1, 2, 3. Due to the well-known theorem by Nagumo w12x any solution of Eq.

 ECO-EPIDEMIOLOGICAL MODEL ANALYSIS 737

Ž2.3. with X 0 g Rq
3
, say X Ž t . s X Ž t, X 0 ., is such that X Ž t . g Rq
3
for all
t ) 0.

Boundedness of Solutions
LEMMA 2.1. Assume that the initial condition of Eq. Ž1.5. satisfies s0 q i 0
G 1. Then either Ži. sŽ t . q iŽ t . G 1 for all t G 0 and therefore as t ª q⬁,
Ž sŽ t ., iŽ t ., y Ž t .. ª E10 s Ž1, 0, 0., or Žii. there exists a t 1 ) 0 such that
sŽ t . q iŽ t . - 1 for all t ) t 1. Finally Žiii. if s0 q i 0 - 1, then sŽ t . q iŽ t . - 1
for all t G 0.
Proof. We consider first sŽ t . q iŽ t . G 1 for all t G 0. From the first
two equations of Ž1.5. we get

d
Ž s Ž t . q i Ž t . . s as 1 y Ž s q i . y b 2 i y ␩ Ž i . y. Ž 2.4.
dt

Hence, for all t G 0, we have that s⬘Ž t . q i⬘Ž t . F 0. Let

lim Ž s Ž t . q i Ž t . . s ␰ . Ž 2.5.
tª⬁

If ␰ ) 1, then by the Barbalat

˘ lemma, we have
d
0 s lim Ž sŽ t . q iŽ t . .
tªq⬁ dt
s lim as Ž t . Ž 1 y s Ž t . y i Ž t . . y b 2 i Ž t . y ␩ Ž i Ž t . . y Ž t .
tªq⬁

s lim as Ž t . Ž 1 y ␰ . y b 2 i Ž t . y ␩ Ž i Ž t . . y Ž t .
tªq⬁

F ymin  a Ž ␰ y 1 . , b 2 4 lim Ž sŽ t . q iŽ t . .
tªq⬁

s y␰ min  b 2 , a Ž ␰ y 1 . 4 - 0.

This contradiction shows that ␰ s 1; i.e.,

lim Ž s Ž t . q i Ž t . . s 1. Ž 2.6.
tªq⬁

Let us denote by g Ž t . s sŽ t . q iŽ t . for t g w0, q⬁.. Of course, g Ž t . is

differentiable and g ⬘Ž t . uniformly continuous for t g Ž0, q⬁.. Thus, with
Eq. Ž2.6. all the assumptions of Barbalat ˘ lemma hold true and, therefore,
d
lim Ž s Ž t . q i Ž t . . s 0. Ž 2.7.
tªq⬁ dt
738 XIAO AND CHEN

Since from the first two equations of Ž1.5.

d
Ž s Ž t . q i Ž t . . s as Ž t . 1 y Ž s Ž t . q i Ž t . .
dt
y b2 i Ž t . y ␩ Ž i Ž t . . y Ž t . Ž 2.8.
then Eq. Ž2.6. implies that

d
lim Ž s Ž t . q i Ž t . . s y lim Ž b 2 i Ž t . q ␩ Ž i Ž t . . y Ž t . . . Ž 2.9.
tªq⬁ dt tªq⬁

Hence, Eqs. Ž2.7. and Ž2.8. are in agreement if and only if lim t ªq⬁ iŽ t . s 0,
which jointly with Eq. Ž2.6. imply lim t ªq⬁ sŽ t . s 1. From the third equa-
tion of Ž1.5., we have y Ž t . tends to zero as t ª q⬁. This completes the
case Ži..
Assume that assumption Ži. is violated. Then there exists t 0 ) 0 at which
for the first time sŽ t 0 . q iŽ t 0 . s 1. According to Eq. Ž2.8. we have

d
Ž s Ž t . q i Ž t . . < tst s yb2 i Ž t 0 . y ␩ Ž i Ž t 0 . . y Ž t 0 . - 0.
0
dt

This implies that once a solution with s q i has entered into the interval
Ž0, 1. then it remains bounded there for all t ) t 0 ; i.e.,

s Ž t . q i Ž t . - 1 for all t ) t 0 . Ž 2.10.

Finally, if s0 q i 0 - 1, applying the previous argument, it follows that
sŽ t . q iŽ t . - 1 for all t ) 0; i.e., Žiii. holds true. This completes the proof.
LEMMA 2.2. There is an M ) 0 such that for any positi¨ e solution
Ž sŽ t ., iŽ t ., y Ž t .. of Eq. Ž1.5.

y Ž t . - M for all large t , Ž 2.11.

k
where M s b0 , k 0 s k Ž1 q ⑀ . q k Ž b1 y b 2 ., ⑀ ) 0 is sufficiently small.
1

Proof. Lemma 2.1 implies that for any Ž s0 , i 0 , y 0 . such that s0 q i 0 G 1,

then either a time t 0 ) 0 exists for which sŽ t . q iŽ t . - 1 for all t ) t 0 or
lim t ªq⬁ sŽ t . s 1, lim t ªq⬁ iŽ t . s 0. Furthermore, if s0 q i 0 - 1 then sŽ t .
q iŽ t . - 1 for all t ) 0. Hence, in any case a nonnegative time, say t*,
exists such that iŽ t . - 1, sŽ t . - 1 q ⑀ , for all t ) t*, then we have ␩ Ž i . -
␩ Ž1. for t ) t*. Set

V s ki Ž t . q y Ž t . . Ž 2.12.
 ECO-EPIDEMIOLOGICAL MODEL ANALYSIS 739

Calculating the derivative of V along the solutions to Eq. Ž1.5., we find for
t ) t*

V̇ s ks Ž t . i Ž t . y kb2 i Ž t . y b1 y Ž t .
F k Ž 1 q ⑀ . y b1 Ž ki Ž t . q y Ž t . . q k < b1 y b 2 <
F yb1V q k 0 .

Recall that iŽ t . - 1 for all t ) t*. There exists positive constant M,

depending only on the parameters of system Ž1.5., such that V Ž t . - M for
all large t. The assertion of Lemma 2.2 now follows and the proof is
completed.
Let ⍀ be the following subset of Rq 3

⍀ s  Ž s, i , y . g Rq
3
: s q i F 1, y F M4. Ž 2.13.

THEOREM 2.1. The set ⍀ is a global attractor in Rq

3
and, of course, it is
positi¨ ely in¨ ariant.
Proof. Due to Lemma 2.1 and Lemma 2.2 for all initial conditions in
Rq 3
such that Ž s0 , i 0 , y 0 . f ⍀, either there exists a positive time, say
T, T s max t 1 , t*4 , such that the corresponding solution Ž sŽ t ., iŽ t ., y Ž t .. g
int ⍀ for all t ) T, or the corresponding solution is such that
Ž sŽ t ., iŽ t ., y Ž t .. ª E10 s Ž1, 0, 0. as t ª q⬁. But E10 g ⭸ ⍀. Hence the
global attractivity of ⍀ in Rq 3
has been proved.
Assume now that Ž s0 , i 0 , y 0 . g int ⍀. Then Lemma 2.1 implies that
sŽ t . q iŽ t . - 1 for all t ) 0 and also by Lemma 2.2 we know that y Ž t . - M
for all large t. Let us remark that if Ž s0 , i 0 , y 0 . g ⭸ ⍀, because s0 q i 0 s 1
or y 0 s M or both, then still the corresponding solution Ž sŽ t ., iŽ t ., y Ž t ..
must immediately enter int ⍀ or coincide with E10 .

3. EQUILIBRIA, LOCAL STABILITY,

AND HOPF BIFURCATION

The model equations Ž1.5. has the following nonnegative equilibria,

namely

aŽ 1 y b2 .
E0 s Ž 0, 0, 0 . , E10 s Ž 1, 0, 0 . , E20 s b 2 , , 0 J Ž s, i , y .
ž 1qa /
740 XIAO AND CHEN

and E2 s Ž s*, i*, y*., where

a y Ž a q 1 . i* b1 Ž s* y b 2 . i*
s* s , ␩ Ž i* . s , y* s . Ž 3.1.
a k ␩ Ž i* .

The boundary equilibrium E20 exists if b 2 - 1 and the existence condi-

aŽ1 y b . b
tion for the positive equilibrium E* is ␩ Ž a q 1 2 . ) k1 ; i.e.,

1qa b1
b2 - 1 y ␩y1 ž / J bU2 . Ž 3.2.
a k
It is clear that b 2 - bU2 is necessary for the existence of component of
y* of the positive equilibrium. It is to be also noted that this condition
implies the existence of E20 . Hence, we conclude that the existence of E*
implies the existence of E20 , but the reverse is not true. It is also
interesting to observe that the equilibrium E20 arises from E10 for the
value of the parameter b 2 equal to 1 and persists for all b 2 - 1, while E*
arises from E20 when b 2 reaches the value bU2 and persists below this
value. We summarize these results.
THEOREM 3.1. Whene¨ er b 2 g w1, q⬁. the equilibria of Eq. Ž1.5. are E0
and E10 . Whene¨ er b 2 g w bU2 , 1., besides E0 and E10 , equilibrium E20 is
feasible. As b 2 ª 1y, then E20 ª E10 and E20 s E10 when b 2 s 1. When-
e¨ er b 2 g Ž0, bU2 . besides E0 , E10 , and E20 , the positi¨ e equilibrium E* is
feasible. As b 2 ª bUy 2 then E* ª E20 and E* s E20 when b 2 s bU2 .
Let
1 k aŽ 1 y b2 .
R0 s , ␴s ␩ ž / . Ž 3.3.
b2 b1 aq1

From Theorem 3.1, we know that if R 0 ) 1, then the equilibrium point

E20 exists, and the positive equilibrium E* is feasible if ␴ ) 1. In particu-
lar, parameter R 0 , which controls whether or not the disease persists, is
closely related to the ‘‘basic reproduction number’’ of epidemic theory w11x,
and parameter ␴ governs whether or not the positive equilibrium exists.
Let Eˆ s Ž ˆ
s, ˆ
i, ˆ
y . be arbitrary equilibrium. Then the characteristic equa-
ˆ
tion about E is given by

a y 2 asˆ y Ž a q 1 . ˆ
iy␭ y Ž a q 1. ˆ
s 0
det ˆi ˆs y b 2 y ␩ ⬘ Ž ˆi . ˆy y ␭ y␩ Ž ˆ
i.
0 k␩ ⬘ Ž ˆ
i. ˆ
y yb1 q k␩ Ž ˆ
i. y ␭

s 0. Ž 3.4.
 ECO-EPIDEMIOLOGICAL MODEL ANALYSIS 741

For equilibrium E0 s Ž0, 0, 0., Ž3.4. reduces to

Ž ␭ y a. Ž ␭ q b 2 . Ž ␭ q b1 . s 0.
Obviously, the above formula has a root with a positive real part. Hence
E0 is a saddle point.
In a similar manner, the disease-free equilibrium E10 is locally asymp-
totically stable if b 2 ) 1, and it is unstable if b 2 - 1. Equilibrium E20 is
locally asymptotically stable if b 2 ) bU2 , and it is unstable if b 2 - bU2 .
For positive equilibrium E* s Ž s*, i*, y*., characteristic equation Ž3.4.
gives
␭3 q Q1 ␭2 q Q 2 ␭ q Q 3 s 0,
where the coefficients Q i , i s 1, 2, 3, are
Q1 s ys* q b 2 q ␩ ⬘ Ž i* . y* q as*
Q 2 s as* Ž ys* q b 2 q ␩ ⬘ Ž i* . y* . q ky*␩ ⬘ Ž i* . ␩ Ž i* . q Ž a q 1 . s*i*
Q 3 s as*ky*␩ ⬘ Ž i* . ␩ Ž i* . .

From assumption Ž H5 . and Ž1.6., we know that ␩ Ž i*. ) 0, ␩ ⬘Ž i*. ) 0, then

Q 3 ) 0. Define

⌬Ž2. s Q1 Q 2 y Q 3
s Ž A q as* . as*A q Ž a q 1 . s*i* q ky*␩ Ž i* . ␩ ⬘ Ž i* .
y as*ky*␩ ⬘ Ž i* . ␩ Ž i* .
s Ž A q as* . as*A q as* Ž 1 y s* . q Ab1 y*␩ ⬘ Ž i* .
2
␩ ⬘ Ž i* . i* y ␩ Ž i* . 2
␩ ⬘ Ž i* . i* y ␩ Ž i* .
s as* Ž s* y b 2 . q
ž ␩ Ž i* . / ␩ Ž i* .

2
␩ ⬘ Ž i* . b1 i*
= Ž s* y b 2 . Ž as* . y as* Ž 1 y s* . q Ž s* y b 2 .
␩ Ž i* .
2
q Ž as* . Ž 1 y s* . ,
where
␩ ⬘ Ž i* . i*
A s ys* q b 2 q ␩ ⬘ Ž i* . y* s y Ž s* y b 2 . q Ž s* y b 2 .
␩ Ž i* .
␩ ⬘ Ž i* . i* y ␩ Ž i* .
s Ž s* y b 2 . .
␩ Ž i* .
742 XIAO AND CHEN

We classify the following three cases to investigate the local properties

of E*
␩ ⬘ Ž i* . i* y ␩ Ž i* .
Case Ži.. ␩ Ž i* .
G 0, in which case A ) 0, then Q1 ) 0, ⌬Ž2. ) 0.
Hence E* is locally asymptotically stable according to Routh᎐Hurwitz
criterion.
as* ␩ ⬘ Ž i* . i* y ␩ Ž i* .
Case Žii.. y s* y b - ␩ Ž i* .
- 0, in which case Q1 ) 0. Further,
2

2
␩ ⬘ Ž i* . i* y ␩ Ž i* . 2
as*
⌬ ) as* Ž s* y b 2 . y Ž s* y b 2 .
Ž2.
ž ␩ Ž i* . / s* y b 2

2
␩ ⬘ Ž i* . b1 i* 2
= Ž as* . q as* Ž 1 y s* . q Ž s* y b 2 . q Ž as* . Ž 1 y s* .
␩ Ž i* .
2
␩ ⬘ Ž i* . i* y ␩ Ž i* . 2 2
s as* Ž s* y b 2 . y Ž as* .
ž ␩ Ž i* . /
␩ ⬘ Ž i* . b1 i*
y Ž s* y b 2 . .
␩ Ž i* .

Then, E* is locally asymptotically stable provided

2
Ž ␩ ⬘ Ž i* . i* y ␩ Ž i*. . Ž s* y b 2 . 2
2
) Ž as*␩ Ž i* . . q ␩ ⬘ Ž i* . ␩ Ž i* . b1 i* Ž s* y b 2 . Ž 3.5.
holds true.
␩ ⬘ Ž i* . i* y ␩ Ž i* . as*
Case Žiii.. ␩ Ž i* .
- y s* y b , in which case Q1 - 0, hence E* is
2

unstable. Therefore, we have the following proposition.

PROPOSITION 3.1. E* is locally asymptotically stable pro¨ ided one of the
following two cases holds
␩ ⬘ Ž i* . i* y ␩ Ž i* .
Ž1.
␩ Ž i* .
G0

Ž2. y as* - ␩ ⬘ Ž i* . i* y ␩ Ž i* . - 0 and Ž3.5. holds true.

s* y b ␩ Ž i* .
2

To study the asymptotic properties of positive equilibrium and Hopf

bifurcation at E* in detail, we consider the response function as Holling
type II, given by
mi
␩Ž i. s , mk ) b1 , Ž 3.6.
nqi
 ECO-EPIDEMIOLOGICAL MODEL ANALYSIS 743

where m, n are positive constants. Thus Condition Ž3.2. which ensures the
existence of E* gives
Ž a q 1 . b1 n
b2 - 1 y J bU2 . Ž 3.7.
a Ž mk y b1 .
To consider the local stability analysis of the positive equilibrium E* as
b 2 varies in Ž0, bU2 ., we recall that the stability properties of E* depend on
the susceptible dimensionless concentration s*, which we shall rename as
␰ s s*, ␰ g Ž b2 , 1. . Ž 3.8.
The characteristic equation Ž3.4. about E* gives
␭3 q Q1 Ž ␰ . ␭2 q Q 2 Ž ␰ . ␭ q Q 3 Ž ␰ . s 0, Ž 3.9.
where the coefficients Q i Ž ␰ ., i s 1, 2, 3 are
mny*
Q1 Ž ␰ . s y␰ q b 2 q 2
q a␰
Ž n q i* .
mny* km2 ni*y*
Q 2 Ž ␰ . s a ␰ y␰ q b 2 q q
ž Ž n q i* .
2 / Ž n q i* .
3
Ž 3.10.
q Ž a q 1 . s*i*
m2 n
Q 3 Ž ␰ . s aks*y*i* 3
, ␰ g Ž b2 , 1. .
Ž n q i* .
mny*
Denote AŽ ␰ . s y␰ q b 2 q . Applying equations as*Ž1 y s*. s
Ž n q i* . 2
my* kmi*
Ž a q 1. s*i*, s* y b 2 s and b1 s n q i* , we have
n q i*

n b1
AŽ ␰ . s y Ž ␰ y b2 . q Ž ␰ y b2 . s y Ž ␰ y b 2 . Ž 3.11.
n q i* mk
and then Ž3.10. can be written as
b1 b1 b 2
Q1 Ž ␰ . s A Ž ␰ . q a ␰ s a y ž / ␰q
mk mk
i*
Q 2 Ž ␰ . s A Ž ␰ . a ␰ q b1 1 y ž /Ž ␰ y b2 . q a␰ Ž 1 y ␰ .
n q i*
b1 Ž mk y b1 .
s AŽ ␰ . a␰ q a␰ Ž 1 y ␰ . q Ž ␰ y b2 .
mk
mn b1 a ␰ Ž mk y b1 .
Q 3 Ž ␰ . s a ␰ b1 y* 2
q Ž ␰ y b2 . .
Ž n q i* . mk
744 XIAO AND CHEN

Obviously, Q 3 Ž ␰ . ) 0 for all ␰ g Ž b 2 , 1.. If Q1Ž ␰ . ) 0, then we have

AŽ ␰ . ) ya ␰ . As for Q1Ž ␰ . we have two cases:

Ži. b1 F amk Ži.e., ya F ␩ ⬘ Ž i* . i* y ␩ Ž i* . - 0., in which case Q1Ž ␰ . ) 0

␩ Ž i* .
for all ␰ g Ž b 2 , 1.;
␩ ⬘ Ž i* . i* y ␩ Ž i* .
Žii. b1 ) amk Ži.e., ␩ Ž i* .
- ya., in which case Q1Ž ␰ . - 0 for
as* ␩ ⬘ Ž i* . i* y ␩ Ž i* .
all ␰ g Ž ␰ 1 , 1., Q1Ž ␰ . ) 0 for all ␰ g Ž b 2 , ␰ 1 . Ži.e., y s* y b - ␩ Ž i* .
.
2
and Q1Ž ␰ 1 . s 0 at ␰ s ␰ 1 , where

b1 b 2
␰1 s . Ž 3.12.
b1 y amk

By applying the result w13x we obtain the following theorem.

y1 q '1 q 8 b and ab - 1 y b
Assume b 2 - bU2 - ␰ 1 -
2
THEOREM 3.2. 2 2 2
b n
- ab1 . Then a single Hopf bifurcation occurs at the unique ¨ alue ␰ 0 g Ž b 2 , ␰ 1 .
2

for increasing ␰ ; i.e., the positi¨ e equilibrium E* is locally asymptotically

stable in Ž b 2 , ␰ 0 . and unstable in Ž ␰ 0 , 1..
Proof. Since b 2 - bU2 - ␰ 1 , then we have ␰ g Ž b 2 , ␰ 1 .; it follows from
Žii. that Q1Ž ␰ . ) 0. Moreover, Q 3 Ž ␰ . ) 0 for all ␰ g Ž b 2 , 1.. Now let us
look at

⌬Ž2. Ž ␰ . s Q1 Ž ␰ . Q 2 Ž ␰ . y Q 3 Ž ␰ .
s Ž AŽ ␰ . q a␰ . Ž a␰ AŽ ␰ . q a␰ Ž 1 y ␰ . .
A Ž ␰ . b1 Ž mk y b1 .
q Ž ␰ y b2 .
mk
2
b1 2
b1
s ž / a␰ Ž ␰ y b2 . y Ž ␰ y b2 .
mk mk

2
b1 Ž mk y b1 .
= Ž a␰ . q a␰ Ž 1 y ␰ . q Ž ␰ y b2 .
mk
2
q Ž a␰ . Ž 1 y ␰ . .

b
Obviously, ⌬Ž2. Ž b 2 . s Ž ab2 . 2 Ž1 y b 2 . ) 0, ⌬Ž2. Ž ␰ 1 . s yŽ mk1 . 2 Ž mk y b1 .
Ž ␰ y b 2 . 2 - 0. Since ⌬Ž2. Ž ␰ . is continuous on Ž b 2 , ␰ 1 ., a value ␰ 0 g Ž b 2 , ␰ 1 .
must exist at which ⌬Ž2. Ž ␰ 0 . s 0.
 ECO-EPIDEMIOLOGICAL MODEL ANALYSIS 745
d 2⌬Ž2. Ž ␰ .
Now, we check the sign of for ␰ g Ž b 2 , ␰ 1 ..
d␰ 2

d 2⌬Ž2. Ž ␰ . b1 b1 b1 b 2 b1
s 6a ž q1 /ž y a ␰ y 2a / ž q1 /ž ya /
d␰ 2
mk mk mk mk
2 b1 b1 b1 Ž mk y b1 .
y ab2 ž q1 q /
mk mk mk
b1 b1 b1 b 2 b1
- 6a ž q1 /ž y a ␰1 y 2 a/ ž q1 /ž ya /
mk mk mk mk
2 b1 b1 b1 Ž mk y b1 .
y ab2 ž q1 q /
mk mk mk
b1 b 2 b1 b1 b 2 b1
s y2 a ž q1 /ž ya y / ž q1 /
mk mk mk mk
2 b1 b1 b1
y Ž mk y b1 . y ab2 ž q1 /
mk mk mk
2
2 ab1 b1 1 b1 n
-y Ž 1 y ab2 y b 2 . y 2 1q /ž y ab2
mk ž /žmk a 1 y b2 /
- 0,

which implies the value ␰ 0 is unique.

Furthermore,

d⌬Ž2. Ž ␰ .
d␰ ␰s ␰ 0

b1
s yA Ž ␰ 0 . a ␰ 0 ž q 2 q a Ž A Ž ␰ 0 . q 1 .Ž A Ž ␰ 0 . q 2 a ␰ 0 .
/
mk

2
1 y ␰0 A Ž ␰ 0 . b1
y Ž a␰ 0 . q3 q Ž mk y b1 .
ž ␰ 0 y b2 / mk

2
b1 2
1 y ␰0
- Ž a␰ 0 . q 2 q 2 a2␰ 0 y Ž a ␰ 0 . q3
ž mk / ž ␰ 0 y b2 /
A Ž ␰ 0 . b1
q Ž mk y b1 .
mk
746 XIAO AND CHEN

2
b1 a2␰ 0
s Ž a␰ 0 . ž y1 q/ Ž ␰ 02 q ␰ 0 y 2 b2 .
mk ␰ 0 y b2
A Ž ␰ 0 . b1
q Ž mk y b1 .
mk
- 0.

Hence ⌬Ž2. Ž ␰ . ) 0 in Ž b 2 , ␰ 0 . and, according to Routh᎐Hurwitz criterion,

E* is locally asymptotically stable in Ž b 2 , ␰ 0 .. Furthermore, according to
the results w13x we have a single Hopf bifurcations toward periodic solu-
tions for increasing ␰ , being ⌬Ž2. Ž ␰ . - 0 in Ž ␰ 0 , ␰ 1 .. Of course, E* is
unstable in Ž ␰ 0 , ␰ 1 .. Further, Q1Ž ␰ . - 0 for all ␰ g Ž ␰ 1 , 1., by
Routh᎐Hurwitz criterion we know that E* is also unstable in Ž ␰ 1 , 1.. This
completes the proof.

4. GLOBAL STABILITY RESULTS: PERMANENCE

In Section 3, we have shown that whenever the parameter b 2 - bU2 then

the positive equilibrium E* is feasible and the boundary one E20 s Ž s, i, y .
is unstable. However, as b 2 increases to bU2 and when b 2 s bU2 the positive
equilibrium E* collapses into E20 , whereas for bU2 - b 2 - 1 the positive
equilibrium is not feasible and boundary one E20 becomes locally asymp-
totically stable. As b 2 increases to 1 and when b 2 s 1 the boundary
equilibrium E20 collapses into E10 s Ž1, 0, 0., whereas for b 2 ) 1 the
equilibrium E20 is not feasible and the one E10 is locally asymptotically
stable. In the following, we show the global stability of E10 and E20 .

THEOREM 4.1. If b 2 G 1, then the boundary equilibrium E10 s Ž1, 0, 0. is

globally asymptotically stable.

Proof. Let ⍀ be set Ž2.13. of Rq 3

. We proved that any solution to
Eq. Ž1.5. starting outside ⍀ Žin Rq . either enters into ⍀ at some finite
3

time, say t 0 ) 0, and then it remains in its interior ⍀ for all t ) t 0 or

tends to the boundary equilibrium E10 g ⭸ ⍀. It is therefore sufficient to
prove that E10 is asymptotically stable with respect to int ⍀ to prove
3
global asymptotic stability in Rq .
Let Rqs s Ž s, i, y . g Rq, s ) 0, i G 0, y G 04 and consider scalar
3 3

function V: Rqs3
ªR

V Ž t . s s y 1 y ln s q i. Ž 4.1.
 ECO-EPIDEMIOLOGICAL MODEL ANALYSIS 747

From system Ž1.5. we get

V̇ Ž t . s yaŽ 1 y s . Ž 1 y s y i . q Ž 1 y b 2 . i y y␩ Ž i .
F yaŽ 1 y s . Ž 1 y s y i . q Ž 1 y b 2 . i.
Clearly, the first term on the right of the above formula is always negative
in int ⍀. If b 2 ) 1, then V˙Ž t . is negative in int ⍀ and vanishes if and only
if Ž s, i . s Ž1, 0.. If b 2 s 1, then we have

V̇ Ž t . F yaŽ 1 y s . Ž 1 y s y i . F 0
in ⍀. However, in this case, we have the largest positively invariant subset
of the set, where V˙Ž t . s 0 is Ž s, i . s Ž1, 0.. Hence, for all solutions to
system Ž1.5. starting in int ⍀, we know lim t ªq⬁ sŽ t . s 1, lim t ªq⬁ iŽ t . s 0.
Thus we have y Ž t . tends to zero as t ª q⬁ from the third equation of
system Ž1.5.. This implies global asymptotic stability in int ⍀, and hence in
3
Rq too, of E10 . This completes the proof.
THEOREM 4.2. If bU2 - b 2 - 1, then all solutions to system Ž1.5. starting
in ⍀ approach the boundary equilibrium E20 s Ž s, i, y . as t ª q⬁.
Proof. Because of the positivity of solutions, we have
di Ž t .
F s Ž t . i Ž t . y b2 i Ž t . . Ž 4.2.
dt
Consider the comparison equations

˙1 Ž t . s au1 Ž t . 1 y Ž u1 Ž t . q u 2 Ž t . . y u1 Ž t . u 2 Ž t .
u
Ž 4.3.
˙2 Ž t . s u1 Ž t . u 2 Ž t . y b 2 u 2 Ž t . .
u
aŽ1 y b .
It is easy to see that for b 2 - 1, Ž b 2 , 1 q a 2 . is a unique positive equilib-
rium of system Ž4.3. which is globally asymptotically stable. Let u1Ž0. G s0 ,
u 2 Ž0. G i 0 . If Ž u1Ž t ., u 2 Ž t .. is a solution to Ž4.3. with initial value
Ž u1Ž0., u 2 Ž0.., then by comparison theorem we have sŽ t . F u1Ž t ., iŽ t . F
u 2 Ž t ., for t ) 0, and hence
aŽ 1 y b2 .
lim sup i Ž t . F .
tªq⬁ 1qa
1q a
␩y1 Ž k1 ., we can choose ⑀ ) 0 small enough such
b
Since 1 ) b 2 ) 1 y a
that
aŽ 1 y b2 .
b1 ) k␩ q⑀ .
ž 1qa / Ž 4.4.
748 XIAO AND CHEN

Let T ) 0 be sufficiently large such that

aŽ 1 y b2 .
iŽ t . - q ⑀ for t ) T .
1qa
Then we have, for t ) T

aŽ 1 y b2 .
˙y Ž t . - yb1 y Ž t . q k␩ ž q ⑀ yŽ t. .
/
1qa

By Ž4.4. and comparison theorem we know

lim sup y Ž t . F 0.
tªq⬁

Therefore lim t ªq⬁ y Ž t . s 0. So all solutions to system Ž1.5. approach the

s y i plane first. Since Ž s, i . is globally asymptotically stable in s y i plane,
then for any solution Ž sŽ t ., iŽ t ., y Ž t .. to system Ž1.5. starting in ⍀ we have
lim t ª⬁ sŽ t . s s, lim t ªq⬁ iŽ t . s i, lim t ªq⬁ y Ž t . s 0. This completes the
proof.
In the following we shall prove that the instability of E10 and E20
implies the system Ž1.5. is permanent Žsee definition in w16x..
THEOREM 4.3. System Ž1.5. is permanent pro¨ ided b 2 - bU2 .
In order to prove Theorem 4.3, we present the persistence theory for
infinite dimensional systems from w17x. Let X be a complete metric space.
Suppose that X 0 ; X, X 0 ; X, X 0 l X 0 s ⭋. Assume that T Ž t . is a C0
semigroup on X satisfying

T Ž t. : X0ªX0
Ž 4.5.
T Ž t. : X0 ª X0 .

Let Tb Ž t . s T Ž t .< X 0 and let A b be the global attractor for Tb Ž t .. The

following is a variant of Theorem 4.1 in paper w17x.
LEMMA 4.1. Suppose that T Ž t . satisfies Ž4.5. and we ha¨ e the following:
Ži. there is a t 0 G 0 such that T Ž t . is compact for t ) t 0
Žii. T Ž t . is point dissipati¨ e in X
Žiii. Aˆb s Dx g A ␻ Ž x . is isolated and has an acyclic co¨ ering M,
ˆ
b
where
ˆ s  M1 , M 2 , . . . , M n 4 .
M

Živ. W s Ž Mi . l X 0 s ⭋ for i s 1, 2, . . . , n.
 ECO-EPIDEMIOLOGICAL MODEL ANALYSIS 749

Then X 0 is a uniform repellor with respect to X 0 ; i.e., there is an ⑀ ) 0 such

that for any x g X 0 , lim t ªq⬁ inf d ŽT Ž t . x, X 0 . G ⑀ , where d is the distance of
T Ž t . x from X 0 .

We are now able to state the proof of Theorem 4.3.

Proof of Theorem 4.3. We begin by showing that the boundary planes

3
of Rq repel the positive solutions to system Ž1.5. uniformly. Let us define

C1 s  Ž s, i , y . g Rq
3
: s Ž 0. s 04

C2 s  Ž s, i , y . g Rq
3
: i Ž 0 . s 0, s Ž 0 . / 0 4

C3 s  Ž s, i , y . g Rq
3
: y Ž 0 . s 0, s Ž 0 . i Ž 0 . / 0 4 .

If C0 s C1 j C2 j C3 and C 0 s int Rq 3
, it suffices to show that there
exists an ⑀ 0 ) 0 such that for any solution u t to system Ž1.5. initiating from
C 0 , lim t ªq⬁ inf d Ž u t , C0 . G ⑀ 0 . To this end, we verify below that the
conditions of Lemma 4.1 are satisfied. It is easy to see that C 0 and C0 are
positively invariant. Moreover, Conditions Ži. and Žii. of Lemma 4.1 are
clearly satisfied. Thus we only need to verify Conditions Žiii. and Živ..
There are three constant solutions E0 , E10 , and E20 in C0 , corresponding,
respectively, to sŽ t . s iŽ t . s y Ž t . s 0, sŽ t . s 1, iŽ t . s y Ž t . s 0, and sŽ t . s
s, iŽ t . s i, y Ž t . s 0. If Ž sŽ t ., iŽ t ., y Ž t .. is a solution to system Ž1.5. initiating
di Ž t .
from C1 then dt s yb 2 i y ␩ Ž i . y F yb 2 i. Hence iŽ t . ª 0 and y Ž t . ª 0
as t ª q⬁. If Ž sŽ t ., iŽ t ., y Ž t .. is a solution initiating from C2 with sŽ0. ) 0,
it follows that sŽ t . ª 1 as t ª q⬁ and y Ž t . ª 0 as t ª q⬁. If
Ž sŽ t ., iŽ t ., y Ž t .. is a solution to system Ž1.5. initiating from C3 with sŽ0. ) 0,
iŽ0. ) 0, it is easy to prove that sŽ t . ª s and iŽ t . ª i as t ª q⬁. This
shows that if invariant sets E0 , E10 , and E20 are isolated,  E0 , E10 , E20 4 is
isolated and is an acyclic covering. It is obvious that E0 is an isolated
invariant. The isolated invariance of E10 and E20 will follow from the
following proof.
We now show that W s Ž E0 . l C 0 s ⭋, W s Ž E10 . l C 0 s ⭋, and
W s Ž E20 . l C 0 s ⭋, we restrict our attention to the second and third
equations, since the proof for the first is simple. Assume that contrary, i.e.,
W s Ž E10 . l C 0 / ⭋, then there exists a positive solution Ž sŽ t ., iŽ t ., y Ž t .. to
system Ž1.5. such that

Ž s Ž t . , i Ž t . , y Ž t . . ª Ž 1, 0, 0. as t ª q⬁.
750 XIAO AND CHEN

If we let Ž3.6. hold, we can choose ⑀ ) 0 small enough such that

n Ž 1 y b2 . an2
⑀ - min ½ m
,
m 5 . Ž 4.6.

Let t 0 ) 0 be sufficiently large such that

y⑀ - y Ž t . - ⑀ for t ) t 0 .

Then we have, for t ) t 0 ,

ds Ž t .
s s Ž t . aŽ 1 y s Ž t . . y Ž a q 1. i Ž t .
dt
Ž 4.7.
di Ž t . m⑀
G i Ž t . s Ž t . y b2 y
ž / .
dt n q iŽ t .

Let us consider

˙x 1 Ž t . s x 1 Ž t . a Ž 1 y x 1 Ž t . . y Ž a q 1 . x 2 Ž t .
m⑀ Ž 4.8.
˙x 2 Ž t . s x 2 Ž t . x 1 Ž t . y b 2 y
ž / .
n q x2 Ž t .

Let ¨ s Ž ¨ 1 , ¨ 2 . and let ␮ ) 0 be small enough such that

␮¨ 1 - s Ž t 0 . , ␮¨ 2 - i Ž t 0 . .

If Ž x 1Ž t ., x 2 Ž t .. is a solution to system Ž4.8. satisfying x i Ž t 0 . s ␮¨ i , i s 1, 2,

we know by comparison theorem sŽ t . G x 1Ž t ., iŽ t . G x 2 Ž t . for all t ) t 0 . It
is easy to see that system Ž4.8. has a unique positive equilibrium Ž xU1 , xU2 .,
where
m⑀
xU1 s b 2 q ,
n q xU2

xU2 s
a Ž 1yb 2 . y n Ž aq1 . q 'Ž aŽ1yb . ynŽ aq1. . q4 aŽ aq1. Ž nŽ1yb . ym⑀ . ,
2
2
2

2 Ž aq1 .

which is globally asymptotically stable for ⑀ given by Ž4.6.. Note that

sŽ t . G x 1Ž t ., iŽ t . G x 2 Ž t . for t ) t 0 and lim t ª⬁ x 2 Ž t . s xU2 . This is a contra-
diction. Hence W s Ž E10 . l C 0 s ⭋.
 ECO-EPIDEMIOLOGICAL MODEL ANALYSIS 751

Suppose that there exists a positive solution Ž sŽ t ., iŽ t ., y Ž t .. to system

Ž1.5. such that

Ž s Ž t . , i Ž t . , y Ž t . . ª Ž s, i , 0 . as t ª q⬁.
Since b 2 - bU2 , we can choose ␰ ) 0 small enough such that
a
b 2 - bU2 y ␰. Ž 4.9.
aq1
Let t 1 ) 0 be sufficiently large such that

i y ␰ - i Ž t . - i q ␰ , for t ) t 1 .

Then we have, for t ) t 1

dy Ž t .
G yb1 y Ž t . q k␩ Ž i y ␰ . y Ž t . .
dt
Let us consider

˙Ž t . s yb1 u Ž t . q k␩ Ž i y ␰ . u Ž t . .
u Ž 4.10.
Let u 0 ) 0 be small enough such that

u0 - y Ž 0. s y 0 .

If uŽ t . is a solution to Ž4.10. satisfying uŽ t 1 . s u 0 , since Ž4.9. Ži.e., b1 -

k␩ Ž i y ␰ .., it follows that uŽ t . ª q⬁ as t ª q⬁. Note that y Ž t . G uŽ t .
for t ) t 1 we have y Ž t . ª q⬁ as t ª q⬁. This contradicts Lemma 2.2.
At this time, we are able to conclude from Lemma 4.1 that C0 repels the
positive solutions to Ž1.5. uniformly. Incorporating into Lemmas 2.1 and
2.2, we know that system Ž1.5. is permanent.

5. DISCUSSION

In this paper we have proposed and analyzed a model of a three species,

eco-epidemiological system, namely, sound prey, infected prey and preda-
tor. We have pointed out the well-known phenomenon of ‘‘exchange of
stability’’ through simple bifurcation at the crossing point of E10 to E20 as
well as at the crossing point of E20 to E*. We have shown that E10 is
globally asymptotically stable for b 2 G 1 and E20 is globally asymptotically
stable in ⍀ for b* - b 2 - 1, and we have pointed out the instability of
boundary equilibria implies the permanence of the system and existence of
752 XIAO AND CHEN

the positive equilibrium. Comparing the results w6, 7x in which the nature
of equilibria is proved locally, we obtain the global stability of equilibria
E10 and E20 .
Furthermore, we observed that the strictly positive equilibrium enters a
Hopf-type bifurcation under the conditions of Theorem 3.2. See Fig. 1.
Two threshold parameters R 0 and ␴ are found in this paper which
control the development of the disease and the growth of the predator
population, respectively. The parameter R 0 governs whether or not the
disease persists. In the absence of disease, the prey population approaches
the carrying capacity K. The threshold parameter

1 ␤K
R0 s s
b2 c

is the average number of adequate contacts, when the prey population is

K, of an infective during the mean infectious period 1c . In a totally
susceptible population of K, the parameter R 0 is the average number of
new infections Žsecondary cases. produced per infective, so it is often
called the basic reproduction number. We know that if R 0 F 1 Ži.e.,

FIG. 1. a s 16 , b1 s 1, k s 12 , m s 4, n s 18 , and b 2 s 0.10005.

 ECO-EPIDEMIOLOGICAL MODEL ANALYSIS 753

b 2 G 1., the disease dies out, and of course the predator population goes
into extinction; that is to say the disease-free equilibrium E10 is globally
asymptotically stable. Whereas if R 0 ) 1 Ži.e., b 2 - 1. the infective does
not tend to zero.
The parameter
k aŽ 1 y b2 . k
␴s ␩ ž / s ␩ Ž iK .
b1 aq1 d

governs whether or not the positive equilibrium exists. In the absence of

predator, according to the proof of Theorem 4.2, we know that the infected
aŽ1 y b .
prey population approaches the constant i s 1 q a 2 . Moreover, dk repre-
sents the survival rate of predator. Hence the threshold parameter ␴
simply says that the predator population can only survive if the number
k␩ Ž iK . of new predator is enough to maintain the predator with the given
mortality d, which is just another way to write the condition b 2 - bU2 ; that
is, if ␴ ) 1, the positive equilibrium E* exists and the system is perma-
nent.

ACKNOWLEDGMENTS

The authors thank Professor H. Thieme and a referee for their several useful suggestions.

REFERENCES

1. H. W. Hethcote, A thousand and one epidemic models, in ‘‘Frontiers in Mathematical

Biology’’ ŽS. A. Levin, Ed.., Lecture Notes in Biomathematics 100, pp. 504᎐515,
Springer-Verlag, BerlinrHeidelberg, New York, 1994.
2. R. M. Anderso and R. M. May, ‘‘Infectious Disease of Humans, Dynamics and Control,’’
Oxford Univ. Press, London, 1991.
3. N. J. T. Bailey, ‘‘The Mathematical Theory of Infectious Disease and Its Applications,’’
Griffin, London, 1975.
4. O. Diekmann, J. A. P. Hecsterbeck, and J. A. J. Metz, The legacy of Kermack and
Mckendrick, in ‘‘Epidemic Models: Their Structure and Relation to Data’’ ŽD. Mollision,
Ed.., Cambridge Univ. Press, Cambridge, UK, 1994.
5. K. P. Hadeler and H. I. Freedman, Predator᎐prey population with parasitic infection, J.
Math. Biol. 27 Ž1989., 609᎐631.
6. J. Chattopadhyay and O. Arino, A predator᎐prey model with disease in the prey,
Nonlinear Anal. 36 Ž1999., 749᎐766.
7. E. Venturino, The influence of disease on Lotka-Volterra systems, Rockymount. J. Math.
24 Ž1994., 389᎐402.
8. J. C. Holmes and W. M. Bethel, Modification of intermediate host behavior by parasites,
in ‘‘Behavioural Aspects of Parasite Transmission’’ ŽE. V. Canning and C. A. Wright,
Eds.., Zool. f. Linnean Soc. 51 Suppl. 1 Ž1972., 123᎐149.
754 XIAO AND CHEN

9. R. O. Peterson and R. E. Page, Wolf density as a predictor of predation rate, Swedish

Wildlife Research Suppl. 1 Ž1987., 771᎐773.
10. L. S. Chen and J. Chen, ‘‘Nonlinear Biological Dynamic Systems,’’ Science Press, Beijing,
1993.
11. R. M. Anderson and R. M. May, The population dynamics of microparasites and their
intervebrates hosts, Proc. Roy. Soc. Lond. B 291 Ž1981., 451᎐463.
12. N. Nagumo, Uber die Lage der Integralkurven gewonlicher Differantialgleichungen,
Proc. Phys. Math. Soc. Japan 24 Ž1942., 551᎐567.
13. W. M. Liu, Criterion of Hopf bifurcation without using eigenvalues, J. Math. Anal. Appl.
182 Ž1994., 250᎐255.
14. H. I. Freedman and V. Sree Hari Rao, The trade-off between mutual interference and
time lags in predator᎐prey-systems, Bull. Math. Biol. 45 Ž1983., 991᎐1003.
15. J. K. Hale, ‘‘Theory of Functional Differential Equations,’’ Springer-Verlag, New York,
1977.
16. Y. Kuang, ‘‘Delay Differential Equations,’’ Academic Press, San Diego, 1993.
17. J. K. Hale and P. Waltman, Persistence in infinite-dimensional systems, SIAM J. Math.
Anal. 20 Ž1989., 388᎐396.