You are on page 1of 23

Discrete and Continuous Dynamical Systems - Series B

doi:10.3934/dcdsb.2022131

DISEASE TRANSMISSION DYNAMICS OF AN


EPIDEMIOLOGICAL PREDATOR-PREY SYSTEM IN OPEN
ADVECTIVE ENVIRONMENTS

Shuai Li and Sanling Yuan∗


University of Shanghai for Science and Technology
Shanghai 200093, China

Hao Wang
Department of Mathematical and Statistical Sciences, University of Alberta
Edmonton, Alberta T6G 2G1, Canada

(Communicated by Yuan Lou)

Abstract. This paper delves into the dynamics of a spatial eco-epidemiological


system with disease spread within the predator population in open advective
environments. The disease-free subsystem is first discussed, and the net repro-
ductive rate RP is established to determine whether the predator can invade
successfully. The impacts of advection rate on RP are also discussed. Then
for the scenario of successful invasion of the predator, sufficient conditions for
the prevalence of disease and the local stability of disease-free attractor are
obtained by dint of persistence theory and comparison theorem. Finally, we
present a special numerical example, in which the basic reproduction ratio R0
of the disease is established in the absence or presence of periodic perturbation.
Our theoretical and numerical results both indicate that the advection rate in
an intermediate interval can favor the coexistence of prey and healthy predator
as well as the eradication of disease.

1. Introduction. The two fields of population dynamics and infectious disease


dynamics were studied separately for decades. However, they have some com-
mon features. The study based on the mathematical models with regard to eco-
epidemiological frameworks has attracted the wide attention of many authors after
the seminal work of Hadeler and Freedman [14] and Chattopadhyay and Arino [8].
In the realm of eco-epidemiology, researchers study the effects of transmissible dis-
ease in an ecological system. The eco-epidemiological models have been applied to
study disease either in predators [5, 11, 17] or in prey [15, 22, 47] or in both popula-
tions [18, 29]. For example, Hilker and Schmitz observed that disease transmission
among predator populations can stabilize fluctuations of predator and prey [17].

2020 Mathematics Subject Classification. Primary: 92D25, 92D30, 35A01, 35B35.


Key words and phrases. Predator-prey system, advective environments, eco-epidemiology, basic
reproductive ratio, persistence.
The second author is partially supported by the National Natural Science Foundation of China
(Grant 12071293). The third author is partially supported by the Natural Sciences and Engineering
Research Council of Canada (Discovery Grant RGPIN-2020-03911 and Accelerator Grant RGPAS-
2020-00090).
∗ Corresponding author: Sanling Yuan.

1
2 SHUAI LI, SANLING YUAN AND HAO WANG

Hethcote et al. pointed out that predation on infected prey can lead to the eradica-
tion of disease [15]. Hsieh and Hsiao found that the endemic states can also exhibit
periodic perturbation when the disease can spread in both populations [18]. Notice
that the classical eco-epidemiological systems usually assume that the population
is well mixed and the spatial impact on the dynamics of disease transmission has
not been studied in depth.
It is acknowledged that many organisms live in advective environments such as
rivers and streams. Recently, because of anthropogenic influence and rising water
temperatures induced by climate change, the severity and incidence of fish diseases
induced by viruses have increased, posing a grave menace to the supply of fish
and leading to local extinctions in many river systems [7, 34]. For example, the
regulations of stream and river (such as water transfer and dam construction) have
a great influence on multifarious properties of flow regime including flow speed
which can greatly influence the fish behaviors and hence the disease transmission
[16, 35]. Moreover, there is ample experimental evidence that water velocity can
affect the spread of disease in the fish population. For example, the authors in [4]
showed that increasing water velocity leading to a turnover rate of 4.5/h can cause
the elimination of Ich. The experiment in [3] demonstrated that the lower water
velocity could induce higher Ceratomyxa shasta infection prevalence in Manayunkia
Speciosa and increase infection severity in fish. Timmerhaus et al. thought that the
optimum water velocity for Atlantic salmon is from 1BL/s to 1.8BL/s. Salmons’
growth potential has not been fully used and resistance to disease has not been
formed when the water velocity is under 1BL/s [43]. Hence, it is essential to study
the impacts of advection rate on the disease spread in the predator-prey system.
In fact, many researchers have constructed reaction-diffusion-advection systems
to study the dynamics of species in advective environments [16, 20, 24, 32, 36, 37,
39, 41, 44, 45, 48, 54, 26]. How individuals prevent flow-induced washout and persist
in advective environments has puzzled ecologists for decades. Speirs and Gurney
conjectured that random diffusion plays an important role in the phenomenon that
extinction is avoidable when there is only a transport process. By developing a
reaction-diffusion-advection model, they displayed that the single species cannot
survive when the random dispersal rate is too large or too small [39]. The authors
in [44] proved that there admits a critical advection rate below which the species
will persist. By picking water velocity as a trait, Lou et al. demonstrated that
the species with a larger advection rate has disadvantages against one with a lower
advection speed if the other parameters are the same for a competition system
[26]. Hilker and Lewis studied a predator-prey system with Danckwert’s boundary
conditions and discern the flow speed scenarios about coexistence and extinction of
predator and prey [16]. Prevention or manipulations of the water environment have
been suggested as the best ways to control disease [4]. By developing an advection-
reaction eco-epidemic model, in [36] the authors found that additional dam release
can reduce the disease prevalence up to 40% under current conditions. However,
little work has been focused on modelling dynamics for the spread of infectious
disease among biotic populations in rivers.
Motivated by the aforementioned discussions, in this paper, we will construct an
advection-diffusion-reaction model to try to elucidate the impacts of water velocity
on disease transmission within predators. We would like to address the following
ecological issues:
• Whether the unidirectional water flow can cause disease extinction?
ADVECTIVE EPIDEMIOLOGICAL PREDATOR-PREY DYNAMICS 3

• How to define and characterize the basic reproduction ratio R0 if the global at-
tractor of the disease-free dynamics is a non-constant steady state in advective
environments?
We will estimate the optimum interval of advection rate to ensure the eradication
of disease and the coexistence of prey and susceptible predator.
The rest of this paper is structured as follows. In section 2, we formulate a
novel system and present some useful results on the eigenvalue problem. The well-
posedness of the developed system is addressed in section 3. In section 4, we perform
the analysis of the subsystem without infected predator and obtain the conditions
for the invasion of predators. In section 5, we study the persistence and extinction of
the developed system. A numerical example is provided to protrude the correctness
of our developed results in section 6, and finally in section 7, some conclusions and
discussions are given to summarize the paper.

2. Model formulation and some preliminaries. Assume that the interactive


populations inhabit open advective environments with length `. The boundary
conditions are Danckwert’s boundary conditions, which mean that the stream source
satisfies zero-flux conditions but the stream outflow meets free-flow conditions at the
same rate as the advection rate [1]. Let U (x, t), S(x, t) and I(x, t) be respectively
the population densities of prey, susceptible predator and infected predator species
at location x and moment t. We further assume that (H1) the prey population
is regulated by logistic growth when the predator is absent; (H2) the predation
ability of infected predator is reduced to be negligible due to disease; (H3) only
the horizontal transmission is considered; and (H4) the lifespan of predators is not
long so that immunity or loss of immunity is ignored. By considering the above
assumptions, we formulate the epidemiological predator-prey model as follows:
  U

 Ut = δ1 Uxx − ϑUx + γU 1 − − g(U )S, x ∈ (0, `), t > 0,
k





 St = δ2 Sxx − ϑSx + g(U )S − h(I)S − σS, x ∈ (0, `), t > 0,





 I = δ I − ϑI + h(I)S − (σ + ξ)I, x ∈ (0, `), t > 0,

t 3 xx x
(1)
 ϑU (0, t) − δ1 Ux (0, t) = 0, Ux (`, t) = 0, t > 0,






ϑS(0, t) − δ2 Sx (0, t) = 0, Sx (`, t) = 0, t > 0,







ϑI(0, t) − δ3 Ix (0, t) = 0, Ix (`, t) = 0, t > 0,

where γ and k denote respectively the per-capita intrinsic growth rate and carrying
capacity of the prey; (< 1) is the conversion rate; σ and ξ stand respectively for
the natural death rate and disease-induced death rate of predators; δi , i = 1, 2, 3
are respectively the diffusion coefficients of prey, susceptible and infected predators;
ϑ is the advection rate. We further assume that the functional response function
g ∈ C 2 (R+ , R+ ) and the incidence function h ∈ C 2 (R+ , R+ ) satisfy
(A1) g(0) = 0, g 0 (U ) > 0 and g 00 (U ) ≤ 0 for all U ≥ 0;
(A2) h(0) = 0, h0 (I) > 0 and h00 (I) ≤ 0 for all I ≥ 0.
In the remainder of this section, we present some useful results on the linear
eigenvalue problem and the predator-absent subsystem as preliminaries. We first
4 SHUAI LI, SANLING YUAN AND HAO WANG

consider an auxiliary linear eigenvalue problem


(
δwxx − ϑwx + l(x)w = λw, 0 < x < `,
(2)
ϑw(0) − δwx (0) = 0, wx (`) = 0,

where l(x) ∈ C([0, `], R+ ). The problem (2) has a principal eigenvalue λ1 (δ, ϑ, l(x))
and the associated eigenfunction w1 (δ, ϑ, l(x)) is positive by Krein-Rutman theorem
[19, 23]. It follows from the variational characterization of the principal eigenvalue
that
R ` −ϑx
0
e δ (−δwx2 + l(x)w2 )dx − ϑw2 (0)
λ1 (δ, ϑ, l(x)) = sup R ` −ϑx . (3)
w6=0,w∈H 1 (0,`)
0
e δ w2 dx
From [6] and results in [24, 25], we can reach that λ1 (δ, ϑ, l(x)) shares the following
properties.
Lemma 2.1. The principal eigenvalue λ1 (δ, ϑ, l(x)) meets following statements:
(1) λ1 (δ, ϑ, l1 (x)) ≥ λ1 (δ, ϑ, l2 (x)) if l1 (x) ≥ l2 (x) and the equality holds if l1 (x) ≡
l2 (x);
(2) λ1 (δ, ϑ, l(x)) decreases strictly as ϑ increases;
(3) λ1 (δ, ϑ, l0 ) increases strictly as δ increases and limδ→∞ λ1 (δ, ϑ, l0 ) = l0 − ϑ`
where l0 is a constant.
Next, we are concerned with the following system when the predator is absent:
U
 

 Ut = δ1 Uxx − ϑUx + γU 1 − , 0 < x < `, t > 0,

 k
ϑU (0, t) − δ1 Ux (0, t) = 0, Ux (`, t) = 0, t > 0, (4)



U (x, 0) = U0 (x) ≥6≡ 0, 0 ≤ x ≤ `.

With the aim of understanding the dynamics of system (4), we linearize (4) at zero
and let U = eλt φ(x), then we have
(
δ1 φxx − ϑφx + γφ = λφ, x ∈ (0, `),
(5)
ϑφ(0) − δ1 φx (0) = 0, φx (`) = 0,
where δ1 > 0, γ > 0 and ϑ ≥ 0. We denote by λ1 (δ1 , ϑ, γ) the principal eigenvalue
of system (5) to stress its dependence on the coefficients √ δ1 , ϑ, γ. From [27], there
admits a unique threshold value ϑ∗ = ϑ∗ (δ1 , γ) ∈ (0, 2 δ1 γ) such that
λ (δ , ϑ, γ) > 0, if 0 ≤ ϑ < ϑ∗ ,

 1 1


λ1 (δ1 , ϑ, γ) = 0, if ϑ = ϑ∗ , (6)



λ1 (δ1 , ϑ, γ) < 0, if ϑ > ϑ .

Hence, we have the following result from Lemma 5.4 in [25].


Lemma 2.2. Suppose δ1 , ϑ, γ > 0 are fixed. Then system (4) admits a unique
globally asymptotically stable positive steady state (denoted by U ∗ (x, δ1 , ϑ)) if 0 ≤
ϑ < ϑ∗ , and U = 0 is globally asymptotically stable for ϑ ≥ ϑ∗ . Furthermore, 0 <
U ∗ (x, δ1 , ϑ) < k on [0, `] and 0 < Ux∗ (x, δ1 , ϑ) < δϑ1 U ∗ (x, δ1 , ϑ) in (0, `). Besides,
U ∗ (x, δ1 , ϑ) is decreasing as ϑ increases.
ADVECTIVE EPIDEMIOLOGICAL PREDATOR-PREY DYNAMICS 5

Finally, we consider the following eigenvalue problem:


χψ = δ2 ψxx − ϑψx + (g(U ∗ (x, δ1 , ϑ)) − σ)ψ, x ∈ (0, `),
(
(7)
ϑψ(0) − δ2 ψx (0) = 0, ψx (`) = 0.
The following results follow the arguments of Lemma 2.4 in [33].
Lemma 2.3. Let χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ)) be the principal eigenvalue of the eigenvalue
problem (7). Then,
(1) χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ)) strictly decreases as ϑ increases in [0, ϑ∗ ) with
χ1 (δ2 , 0, U ∗ (x, δ1 , 0)) = g(k) − σ, lim χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ)) = λ1 (δ2 , ϑ∗ , −σ);
ϑ→ϑ∗−
(2) the associated principal eigenfunction ψ1 meets (ψ1 )x > 0 in (0, `).

3. Well-posedness. In this section, we narrow our focus to explore the well-


posedness of system (1). In the sequel of this paper, we denote Xν = C([0, `], Rν+ )
as the positive cone in the functional space C([0, `], Rν ) endowed with the usual
supremum norm; here ν can be chosen as 1, 2 or 3. Let u1 = U, u2 = S, u3 = I and
u = (u1 , u2 , u3 ), and the initial functions in system (1) meeting u0 = (u01 , u02 , u03 ) =
(U 0 , S 0 , I 0 ) ∈ X3 . By appealing to the method developed in [30], we can obtain the
following conclusion about the local existence of solutions to system (1).
Lemma 3.1. Suppose that the initial data u0 ∈ X3 . Then system (1) admits a
unique mild solution u(x, t, u0 ) ∈ X3 on [0, Tu0 ), where Tu0 ≤ ∞. Moreover, this
solution is a classical solution.
To obtain the global existence and ultimate boundedness of the solutions u(x, t,
u0 ), we first recall a result of [46], which is a direct outcome of Theorem 1 in [12]
Lemma 3.2. Consider the following parabolic system with no-influx boundaries:

 ∂vi
= ∇ · Di (x)∇vi + qi (x) · ∇vi + gi (x, t, v, ∇vi ), x ∈ Ω ⊂ Rn , t > 0,
∂t
 v (x, 0) = v 0 (x), x ∈ Ω, 1 ≤ i ≤ k,
i i

where v = (v1 , v2 , · · · , vk ), Ω is an open set in Rn associated with smooth boundary,


Di (x) > di > 0 and qi (x) is bounded for i = 1, 2, · · · , k. Suppose that for each
i = 1, 2, · · · , k, the functions gi satisfy
k
X
|gi (x, t, v, ζ)| ≤ C1 |vi |p + C2 |ζ|q + C3 , (8)
i=1
for some constants Cj ≥ 0, 1 ≤ j ≤ 3, p > 0 and q ≥ 0. Let α be the positive
constant such that
n n 2(q − 1) o
α > max 0, p − 1, .
2 (q − 2)
If there exists a positive function Cα (v0 ) such that ||v(x, t)||Lα (Ω) ≤ Cα (v0 ), ∀t ∈
(0, Tv0 ), then the solution exists globally and there exists a function Cα (v0 ) such
that ||v(x, t)||L∞ (Ω) ≤ C∞ (v0 ). Moreover, if there exist T and CT independent of
initial function such that ||v(x, t)||Lα (Ω) ≤ CT , ∀t ≥ T , then there exists a positive
number C∞ , independent of initial datum, such that ||v(x, t)||L∞ (Ω) ≤ C∞ , t ≥ T .
Thereupon, we obtain the following main result of this section.
Theorem 3.3. For any initial data u0 ∈ X3 , the solution of system (1) is ultimately
bounded, positive and exists globally for t ∈ [0, ∞).
6 SHUAI LI, SANLING YUAN AND HAO WANG

Proof. From the first equation of system (1), we know U (x, t) satisfies
 U
Ut ≤ δ1 Uxx − ϑUx + γU 1 − , x ∈ (0, `), t > 0.
k
By the standard comparison arguments and results in [32, 33], we conclude that
there exists M1 > 0 which hinges on the initial datum U 0 , such that
0 < U (x, t) ≤ M1 , x ∈ [0, `], t > 0.
Thereupon, letting Z(x, t) = U (x, t) + S(x, t) + I(x, t) and then following from
system (1), we obtain that
d `
Z
Z(x, t)dx ≤ −ϑU (`, t) − ϑS(`, t) − ϑI(`, t)
dt 0
Z `  Z `
U 
+ γU 1 − + σU dx − σ Z(x, t)dx
0 k 0
R`
≤ M1 (γ + σ)` − σ 0 Z(x, t)dx,
from which we can easily have
Z `
M1 (γ + σ)`
lim sup Z(x, t)dx ≤ . (9)
t→∞ 0 σ
Denote the reaction terms of system (1) by g = (g1 (x, t, u), g2 (x, t, u), g3 (x, t, u)).
We then calculate from Young’s inequality that
U 2 + S2
|g1 (x, t, u)| ≤ γk + |g 0 (0)| ,
2
U 2 + S2 S2 + I 2 S2 + 1
|g2 (x, t, u)| ≤ |g 0 (0)| + |h0 (0)| +σ
2 2 2
0
|g (0)| 2  |g (0)| + |h (0)| + σ 2 |h0 (0)| 2 σ
0 0 
= U + S + I + ,
2 2 2 2
2 2 2
S + I 1 + I
|g3 (x, t, u)| ≤ |h0 (0)| + (σ + ξ)
2 2
|h0 (0)| 2 |h0 (0)| + (σ + ξ) 2 (σ + ξ)
= S + I + .
2 2 2
By taking n = 1, k = 3, p = 2, q = 0 and α = 1, we can then check that the
conditions in Lemma 3.2 satisfy. Therefore, we know that for any initial data u0 ,
system (1) admits a unique positive solution u(x, t, u0 ) on [0, +∞).
Before going into the investigation of system (1), we will first study the dynamics
of its subsystem in the absence of infected predator.

4. The subsystem without infected predator. We study system (1) when


infected predator is absent in this section. That is, we consider the following sub-
system:
U
 

 U t = δ 1 U xx − ϑU x + γU 1 − − g(U )S, x ∈ (0, `), t > 0,
k




 S = δ S − ϑS + g(U )S − σS, x ∈ (0, `), t > 0,

t 2 xx x
(10)
ϑU (0, t) − δ U (0, t) = 0, U (`, t) = 0, t > 0,



 1 x x



ϑS(0, t) − δ2 Sx (0, t) = 0, Sx (`, t) = 0, t > 0,

ADVECTIVE EPIDEMIOLOGICAL PREDATOR-PREY DYNAMICS 7

and its corresponding steady state system:


U
 

 δ 1 Uxx − ϑU x + γU 1 − − g(U )S = 0, x ∈ (0, `),
k




 δ S − ϑS + g(U )S − σS = 0, x ∈ (0, `),

2 xx x
(11)
ϑU (0) − δ1 Ux (0) = 0, Ux (`) = 0,







ϑS(0) − δ2 Sx (0) = 0, Sx (`) = 0.

It then follows from Lemma 2.2 that predator absent steady state solution of system
(10) is given by (U ∗ (x, δ1 , ϑ), 0) which is unique if λ1 (δ1 , ϑ, γ) > 0. Linearizing
system (10) at (U ∗ (x, δ1 , ϑ), 0), we obtain the following equations with Danckwert’s
boundary conditions:

 Ut = δ1 Uxx − ϑUx + γ 1 − 2U (x, δ1 , ϑ) U − g(U ∗ (x, δ1 , ϑ))S, x ∈ (0, `), t > 0,
  
k
St = δ2 Sxx − ϑSx + g(U ∗ (x, δ1 , ϑ))S − σS, x ∈ (0, `), t > 0.

(12)
Substituting U = exp(χt)ϕ(x), S = exp(χt)ψ(x) into Eqs. (12), we have the corre-
sponding eigenvalue problem:
2U ∗ (x, δ1 , ϑ) 
 

 χϕ = δ 1 ϕxx − ϑϕ x + γ 1 − ϕ − g(U ∗ (x, δ1 , ϑ))ψ, x ∈ (0, `),
k




 χψ = δ ψ − ϑψ + (g(U ∗ (x, δ , ϑ)) − σ)ψ, x ∈ (0, `),

2 xx x 1

ϑϕ(0) − δ1 ϕx (0) = 0, ϕx (`) = 0,









ϑψ(0) − δ2 ψx (0) = 0, ψx (`) = 0.

(13)
It is obvious that the eigenvalues of the following two operators comprise the eigen-
values of system (13) with Danckwerts boundary conditions:
d2 d  2U ∗ (x, δ1 , ϑ) 
L1 (ϑ) = δ1 2 − ϑ +γ 1− ,
dx dx k
d2 d
L2 (ϑ) = δ2 2 − ϑ + (g(U ∗ (x, δ1 , ϑ)) − σ).
dx dx

Observing that λ1 (δ1 , ϑ, γ(1 − U (x,δ k
1 ,ϑ)
)) = 0, we deduce that all eigenvalues of
the operator L1 (ϑ) are negative [54]. Therefore, the positive or negative sign of
χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ)) of L2 (ϑ) can respectively determine the instability or stability
of (U ∗ (x, δ1 , ϑ), 0).
Now, we define the next generation operator with regard to the linearized system
(12). We can determine whether susceptible predators can persist in a stream or not
by exploring the properties of the next generation operator. Assume that healthy
predators are introduced into the stream with density S 0 (x) when the density of
prey is U ∗ (x, δ1 , ϑ). Then the healthy predators undergo dispersal in the advective
environments until they are extirpated. Let s(x, t) be semiflow generated by the
following system with Danckwerts boundary condition:
st = δ2 sxx − ϑsx − σs, x ∈ (0, `), t > 0.

Then function g(U (x, δ1 , ϑ))s(x, t) refers to the rate of reproduction by the ini-
tially introduced individuals at time t and at location x through predation. Hence,
the total reproduction by initial individuals during their lifetime is governed by
8 SHUAI LI, SANLING YUAN AND HAO WANG

R +∞
0
g(U ∗ (x, δ1 , ϑ))s(x, t)dt := L(S 0 (x)), which is the next generation distribu-
tion of the initial distribution S 0 (x). Define the net reproductive rate RP by
RP := r(L),
where r(L) is the spectral radius of the linear operator L [10]. The following lemma
states the relationship between RP and χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ)) according to Theorem
2.11 in [31].
Lemma 4.1. RP − 1 and χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ)) have the same sign.
Then, we present the following arguments of extinction and permanence for sys-
tem (10).
Theorem 4.2. Assume λ1 (δ1 , ϑ, γ) > 0 and v(x, t, v0 ) := (U (x, t), S(x, t)) is a
solution to system (10) with v(·, 0, v0 ) = v0 ∈ X2 . Then the following results hold:
(i) If RP < 1, then the predator-absent steady state (U ∗ (x, δ1 , ϑ), 0) for system
(10) is globally attractive.
(ii) If RP > 1, there is a ζ > 0 such that the solution of system (10) meets
limt→∞ inf(U (x, t, v0 ), S(x, t, v0 )) ≥ (ζ, ζ) uniformly on x ∈ [0, `]. Moreover,
system (10) admits at least one positive steady state.
Proof. (i) It follows from Lemma 4.1 that χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ)) < 0 if RP < 1. By
continuity, we obtain that there admits a small enough ε > 0 such that χ1 (δ2 , ϑ,
U ∗ (x, δ1 , ϑ) + ε) < 0. From the first equation of system (10), we can see that
 U
Ut ≤ δ1 Uxx − ϑUx + γU 1 − , x ∈ (0, `), t > 0.
k
By the standard parabolic comparison principle, we conclude that there exists t0 =
t0 (v0 ) > 0 such that when t > t0 ,
U (x, t, v0 ) ≤ U ∗ (x, δ1 , ϑ) + ε, x ∈ [0, `].
Thus, we have
St ≤ δ2 Sxx − ϑSx + g(U ∗ (x, δ1 , ϑ) + ε)S − σS, x ∈ (0, `), t > t0 .
Let ψ ε be the strictly positive eigenfunction of problem (7) associated with eigenva-
lue χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ)+ε). It is easy to deduce that v ε = exp{χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ)
+ ε)t}ψ ε is a solution of the following equation:
− ϑvxε + g(U ∗ (x, δ1 , ϑ) + ε)v ε − σv ε , x ∈ (0, `), t > t0 ,
( ε ε
vt = δ2 vxx

ϑv ε (0, t) − δ2 vxε (0, t) = 0, vxε (`, t) = 0, t > t0 .


Notice that there must have a C > 0 such that for given v0 ∈ X2 , S(x, t0 , v0 ) ≤
C exp{χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ) + ε)t0 }ψ ε . The parabolic comparison principle implies
that
S(x, t, v0 ) ≤ C exp{χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ) + ε)t}ψ ε , t ≥ t0 ,
and hence, limt→∞ S(x, t, v0 ) = 0 uniformly for x ∈ [0, `]. Therefore, the U equa-
tion in system (10) is asymptotic to the scale equation (4). By appealing to the
theory of asymptotically autonomous semiflows, we know that limt→∞ U (x, t, v0 ) =
U ∗ (x, δ1 , ϑ) uniformly for x ∈ [0, `].
(ii) Denote X0 = {v(·) = (v1 (·), v2 (·)) ∈ X2 : vi (·) 6≡ 0, i = 1, 2} and ∂X0 =
{v(·) = (v1 (·), v2 (·)) ∈ X2 : v1 (·) ≡ 0 or v2 (·) ≡ 0}. Let Φ(t) : X2 → X2 be
ADVECTIVE EPIDEMIOLOGICAL PREDATOR-PREY DYNAMICS 9

the semiflow generated by system (10). Maximum principle implies that for any
v0 ∈ X0 , we have v(x, t, v0 ) > 0, that is Φ(t)X0 ⊂ X0 . Set
M∂ = {v0 ∈ ∂X0 : Φ(t)v0 ∈ ∂X0 , t ≥ 0},
and ω(v0 ) the omega limit set of the forward orbit γ + (v0 ) := {Φ(t)v0 : t ≥ 0}. Let
E0 := {(0, 0)} and E1 := {(U ∗ (x, δ1 , ϑ), 0)}. We divide the proof into four steps.
Step 1. v0 ∈M∂ ω(v0 ) = E0 E1 .
S S

Assume v0 ∈ ∂X0 . According to the definition of ∂X0 , we have v(x, t, v0 ) =


Φ(t)v0 ∈ ∂X0 for all t ≥ 0. By the maximum principle, we have either v1 (x, t, v0 ) ≡
0 or v2 (x, t, v0 ) ≡ 0 for all t ≥ 0 and x ∈ [0, `]. If v2 (x, t, v0 ) ≡ 0, v1 satisfies system
0 ∗
(4), and hence from Lemma 2.2, we obtain that S limt→∞ v01 (x, t, vS) = U (x, δ1 , ϑ)
0
or limt→∞ v1 (x, t, v ) = 0. That implies that v0 ∈M∂ ω(v ) = E0 E1 .
Step 2. E0 is a uniform weak repeller for X0 . That is, we can find a constant ρ0 > 0
such that
lim sup k Φ(t)v0 − E0 k≥ ρ0 , v0 ∈ X0 .
t→∞
If this is not true, then for any small enough ρ > 0 there admits a t1 > 0, vρ ∈ X0
such that
U (x, t, vρ ) < ρ, S(x, t, vρ ) < ρ, t > t1 .
It then follows from the equation of U in system (10) that
 2γρ 
Ut > δ1 Uxx − ϑUx + γU − + g 0 (0)ρ U, x ∈ (0, `), t > t1 .
k
Since U (x, t, vρ ) > 0 for all t > 0 and x ∈ (0, `), there admits a C1 > 0 such that
n  2γρ  o
U (x, t1 , vρ ) > C1 exp λ1 (δ1 , ϑ, γ − + g 0 (0)ρ) t1 ψ ρ , x ∈ [0, `],
k
where ψ ρ is the positive eigenfunction corresponding to λ1 (δ1 , ϑ, γ − ( 2γρ 0
k + g (0)ρ)).
2γρ
It is easy to check that exp{λ1 (δ1 , ϑ, γ − ( k + g 0 (0)ρ))t}ψ ρ is a solution of the
following equation:
e − 2γρ + g 0 (0)ρ U
  
 U exx − ϑU
et = δ1 U ex + γ U e , x ∈ (0, `), t > t1 ,
k
ϑU (0, t) − δ1 U
 e ex (0, t) = 0, U
ex (`, t) = 0, t > t1 .
By appealing to the comparison principle and the continuity of eigenvalue, we obtain
that
n   2γρ  o
U (x, t, vρ ) ≥ C1 exp λ1 δ1 , ϑ, γ − + g 0 (0)ρ t ψ ρ , for all t ≥ t1 , x ∈ [0, `],
k
which indicates U (x, t, vρ ) is unbounded, a contradiction.
Step 3. E1 is a uniform weak repeller for X0 . That is, to prove that we can find a
ς > 0 meeting
lim sup k Φ(t)v0 − E1 k≥ ς, v0 ∈ X0 .
t→∞
Suppose it is false, there admits a t2 > 0, vς ∈ X0 and for any small enough ς such
that
U (x, t, vς ) > U ∗ (x, δ1 , ϑ) − ς, 0 < S(x, t, vς ) < ς, t > t2 .
Thus, by the monotonicity of g(U ), we obtain
St > δ2 Sxx − ϑSx + g(U ∗ (x, δ1 , ϑ) − ς)S − σS, x ∈ (0, `), t > t2 .
10 SHUAI LI, SANLING YUAN AND HAO WANG

Since S(x, t, vς ) > 0 for all t > 0 and x ∈ (0, L), there exists a C2 > 0 such that
S(x, t2 , vς ) > C2 exp{χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ) − ς)t2 }ψ ς , x ∈ [0, `],
where ψ ς is the positive eigenfunction corresponding to χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ) − ς).
We can check that exp{χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ) − ς)t}ψ ς is a solution of the following
equation:

 Set = δ2 Sexx − ϑSex + g(U ∗ (x, δ1 , ϑ) − ς)S(x,
e t) − σ S(x,
e t), x ∈ (0, `), t > t2 ,
 e t) − δ2 Sex (0, t) = 0, Sex (`, t) = 0, t > t2 .
ϑS(0,
By appealing to the comparison principle again, we obtain that
S(x, t, vς ) ≥ C2 exp{χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ) − ς)t}ψ ς , for all t ≥ t2 , x ∈ [0, `],
which implies S(x, t, vς ) is unbounded leading to a contradiction.
Step 4. Define a function % : X2 → [0, +∞) by
n o
%(v) := min min v1 (x), min v2 (x) .
x∈[0,`] x∈[0,`]

It is easy to see that %−1 (0, +∞) ⊂ X0 and % satisfies that if %(v0 ) > 0 or v0 ∈ X0
with %(v0 ) = 0, then %(Φ(t)v0 ) > 0, t > 0. That is, % is a generalized distance
function for the semiflow Φ(t) : X2 → X2 [53]. Observe that any forward orbit of
Φ(t) in M∂ converges to E0 or E1 . Taking into account the above steps, we obtain
that E0 and E1 are isolated invariant sets for Φ(t) in X0 and W s (E0 ) ∩ X0 = ∅,
W s (E1 ) ∩ X0 = ∅, where W s (E0 ) and W s (E1 ) are the stable set of E0 and E1 .
Furthermore, there is no subset of E0 and E1 forms a cycle in ∂X0 . With the aid of
Theorem 1.3.2 in [53], we get that there exists an δ̃ > 0 such that min{%(v) : v ∈
ω(v0 )} > δ̃ for any v0 ∈ X0 . The uniform persistence is proved. The existence of
positive solution to system (11) can be obtained by Theorem 4.7 in [28].
In the remainder of this section, we choose advection rate ϑ as the bifurcation
parameter to perform some bifurcation analysis on the steady state solutions ob-
tained in Theorem 4.2. We first obtain a priori estimate of positive solutions for
system (11).
Lemma 4.3. Let (U (x), S(x)) be a nonnegative solution of system (11) with U (x)
6≡ 0 and S(x) 6≡ 0. Then 0 ≤ ϑ < ϑ∗ , and σ < g(k).
Proof. We can obtain that U (x) > 0, S(x) > 0 on [0, `] because of the strong
maximum principle. From lemma 2.1, we can know that
 γU (x) g(U (x))S(x) 
0 = λ1 δ1 , ϑ, γ − − < λ1 (δ1 , ϑ, γ).
k U (x)
It then follows from (6) that 0 ≤ ϑ < ϑ∗ . Using Eq. (3), it is easy to deduce from
Lemma 2.2 and comparison arguments that
0 = λ1 (δ2 , ϑ, g(U (x)) − σ) < λ1 (δ2 , ϑ, g(k) − σ) = λ1 (δ2 , ϑ, 0) + g(k) − σ.
Proposition 2.1 in [1] implies that
λ1 (δ2 , ϑ, 0) < 0 for ϑ > 0, and λ1 (δ2 , ϑ, 0) = 0 for ϑ = 0.
Therefore, we have σ < g(k).
The following lemma reveals the relationship between advection rate ϑ and RP .
ADVECTIVE EPIDEMIOLOGICAL PREDATOR-PREY DYNAMICS 11

Lemma 4.4. For any δi > 0, (i = 1, 2) and 0 < ϑ < ϑ∗ , RP < 1 if σ ≥ g(k).
If σ < g(k), then there exists ϑ0 ∈ (0, ϑ∗ ) such that RP < 1 if ϑ ∈ (ϑ0 , ϑ∗ ) and
RP > 1 if ϑ ∈ [0, ϑ0 ). Moreover, RP increases as ϑ decreases.
Proof. If σ ≥ g(k), it follows from Lemma 2.3 that for any ϑ ∈ (0, ϑ∗ ),
χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ)) < χ1 (δ2 , 0, U ∗ (x, δ1 , 0)) = g(k) − σ ≤ 0,
which implies RP < 1. If σ < g(k), we have χ1 (δ2 , 0, U ∗ (x, δ1 , 0)) = g(k) − σ >

0. Notice that limϑ→ϑ∗− χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ)) = λ1 (δ2 , ϑ∗− , −σ) → −σ − ϑ` <
0 as δ2 → ∞. The monotonicity of λ1 (δ2 , ϑ∗ , −σ) about δ2 (see Lemma 2.1) yields
that limϑ→ϑ∗− χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ)) < 0 for any δ2 > 0. We can then obtain that
there admits a unique ϑ0 ∈ (0, ϑ∗ ) such that


 RP > 1, if 0 ≤ ϑ < ϑ0 ,


RP = 1, if ϑ = ϑ0 , (14)


RP < 1, if ϑ0 < ϑ < ϑ∗ .

Next, we will show RP increases as ϑ decreases. In fact, we know from [53] that
RP = µ11 , where µ1 is the unique positive eigenvalue with a positive eigenfunction
for the following linear problem:
−δ2 Ψxx + ϑΨx + σΨ = µ1 g(U ∗ (x, δ1 , ϑ))Ψ, x ∈ (0, `),
(
(15)
ϑΨ(0) − δ2 Ψx (0) = 0, Ψx (`) = 0.
We now adopt the method used in [24] to obtain that µ1 is a strictly increasing
function of ϑ. We first turn (15) into the following problem by the transformation
ϑ
Ψ = e δ2 x Φ:
−δ2 Φxx − ϑΦx + σΦ = µ1 g(U ∗ (x, δ1 , ϑ))Φ, x ∈ (0, `),
(
(16)
Φx (0) = Φ(`) = 0.
We then differentiate both sides of (16) with respect to ϑ and denote ∂
∂ϑ :=0 yielding
−δ2 (Φxx )0 − ϑ(Φx )0 − Φx + σΦ0 = µ01 g(U ∗ (x, δ1 , ϑ))Φ





+ µ  ∂g(U ) (U ∗ (x, δ , ϑ))0 Φ


 1 ∂U 1
(17)



 + µ1 g(U (x, δ1 , ϑ))Φ0 , x ∈ (0, `),



Φ0x (0) = Φ0 (`) = 0.

ϑ
Multiplying the first equation of (16) by e δ2 x Φ0 and the first equation of (17) by
ϑ
e δ2 x Φ, integrating over (0, `) by parts, and then subtracting the two equations we
obtain
Z ` Z `
 ϑ x ∂g(U ) ∗  ϑ x
− e δ2 ΦΦx − µ1  (U (x, δ1 , ϑ))0 Φ2 dx = µ01 g(U ∗ (x, δ1 , ϑ))e δ2 Φ2 dx.
0 ∂U 0
(18)
Z `
Φ2 (0) ϑ ` δϑ x Φ2
 Z
ϑ
x
Since − e ΦΦx dx = δ2
+ e 2 dx > 0 and (U ∗ (x, δ1 , ϑ))0 < 0
0 2 δ 2 0 2
(see Lemma 2.2), we can deduce from (18) that µ01 > 0 and thus RP is a strictly
decreasing function with respect to ϑ.
Based on Lemma 4.4 we can further have the following result.
12 SHUAI LI, SANLING YUAN AND HAO WANG

Lemma 4.5. Suppose δi > 0, (i = 1, 2) and σ < g(k). Then system (11) exists
no positive solution if ϑ ≥ ϑ0 .
ϑ
Proof. Multiplying the second equation of system (13) by e− δ2 x S(x) and the equa-
ϑ
tion for S of system (11) by e− δ2 x ψ1 , integrating over (0, `) and then subtracting
the two equations, we can deduce
R` − δϑ x R` − δϑ x
χ1 (δ2 , ϑ, U ∗ (x, δ1 , ϑ)) 0 e 2 ψ1 S(x)dx = 0 (g(U
∗ (x, δ
1 , ϑ)) − g(U ))e 2 ψ1 Sdx > 0
∗ ∗
since U < U . This is a contradiction to χ1 (δ2 , ϑ, U (x, δ1 , ϑ)) ≤ 0, when ϑ ∈
[ϑ0 , ϑ∗ ).

Based on previous discussion, we will explore the existence of non-constant pos-


itive steady state solution of system (10).
Theorem 4.6. Suppose δi > 0 (i = 1, 2) and σ < g(k) fixed. Then system (10)
undergoes a backward supercritical transcritical bifurcation at ϑ = ϑ0 , and has at
least one positive steady state solution provided 0 ≤ ϑ < ϑ0 .
Proof. Set X = W 2,p (0, `)×W 2,p (0, `) and Y = Lp (0, `)×Lp (0, `) with p > 1. Then
X ,→ C 1 [0, `] × C 1 [0, `]. Define F : R+ × X → Y × R4 by
  U 
δ1 Uxx − ϑUx + γU 1 − − g(U )S
 k 
δ2 Sxx − ϑSx + (g(U ) − σ)S
 
 
 
ϑU (0) − δ1 Ux (0)
 
F(ϑ, U, S) =  .
 

 ϑS(0) − δ2 Sx (0) 

 

 Ux (`) 

Sx (`)

Set D(U,S) F(ϑ, U (x, δ1 , ϑ), 0) to be the Fréchet derivative of F(ϑ, U, S) with regard
to (U, S) at (U ∗ (x, δ1 , ϑ), 0). We can verify D(U,S) F(ϑ, U ∗ (x, δ1 , ϑ), 0) is a Fredholm
operator with index zero. In fact, D(U,S) F(ϑ, U ∗ (x, δ1 , ϑ), 0)(ϕ, ψ) = 0 has the
following form
2γU ∗ (x, δ1 , ϑ) 
 

 δ1 ϕxx − ϑϕ x + γ − ϕ − g(U ∗ (x, δ1 , ϑ))ψ = 0, x ∈ (0, `),
k




 δ ψ − ϑψ + (g(U ∗ (x, δ , ϑ)) − σ)ψ = 0, x ∈ (0, `),

2 xx x 1

ϑϕ(0) − δ1 ϕx (0) = 0, ϕx (`) = 0,









ϑψ(0) − δ2 ψx (0) = 0, ψx (`) = 0.

Picking ϑ = ϑ0 , by (14), we get χ1 (δ2 , ϑ0 , U ∗ (x, δ1 , ϑ0 )) = 0. The kernel of


D(N,S) F(ϑ0 , U ∗ (x, δ1 , ϑ0 ), 0) is as follows
N (D(U,S) F(ϑ0 , U ∗ (x, δ1 , ϑ0 ), 0)) = span{(φ0 , ψ0 )},
where ψ0 is a positive principal eigenfunction with regard to the eigenvalue problem
(13) with ϑ = ϑ0 , and φ0 = [L1 (ϑ0 )]−1 (g(U ∗ (x, δ1 , ϑ0 ))ψ0 ). Obviously, in view of

λ1 (δ1 , ϑ, γ(1 − 2U (x,δ
k
1 ,ϑ)
)) < 0, we get that φ0 < 0.
Next we prove that the range of D(U,S) F(ϑ0 , U ∗ (x, δ1 , ϑ0 ), 0) is
{(U, S, ς1 , ς2 , ς3 , ς4 ) ∈ Y × R4 : l(U, S, ς1 , ς2 , ς3 , ς4 ) = 0},
ADVECTIVE EPIDEMIOLOGICAL PREDATOR-PREY DYNAMICS 13

where l : Y × R4 → R in (Y × R4 )∗ which is defined by


Z ` −ϑ x −ϑ0 `
0
l(U, S, ς1 , ς2 , ς3 , ς4 ) = e δ2 ψ0 Sdx + ς3 ψ0 (0) − δ2 ς4 ψ0 (`)e δ2 .
0

Suppose (U, S, ς1 , ς2 , ς3 , ς4 ) ∈ R(D(U,S) F(ϑ0 , U ∗ (x, δ1 , ϑ0 ), 0)). (φ, ψ) ∈ X meeting


2U ∗ (x, δ1 , ϑ0 ) 
 

 δ 1 ϕxx − ϑ 0 ϕx + γ 1 − ϕ − g(U ∗ (x, δ1 , ϑ0 ))ψ = U, x ∈ (0, `),
k




 δ ψ − ϑ ψ + (g(U ∗ (x, δ , ϑ )) − σ)ψ = S, x ∈ (0, `),

2 xx 0 x 1 0

ϑ0 ϕ(0) − δ1 ϕx (0) = ς1 , ϕx (`) = ς2 ,









ϑ0 ψ(0) − δ2 ψx (0) = ς3 , ψx (`) = ς4 .

(19)
Notice that
δ2 (ψ0 )xx − ϑ0 (ψ0 )x + (g(U ∗ (x, δ1 , ϑ0 )) − σ)ψ0 = 0, x ∈ (0, `),
(
(20)
ϑ0 (ψ0 )(0) − δ2 (ψ0 )x (0) = 0, (ψ0 )x (`) = 0.
ϑ0
Multiplying the second equation of (19) by e− δ2 x ψ0 and the first equation of (20)
ϑ0
by e− δ2 x ψ, integrating over (0, `) by parts, and then subtracting the two equations
we obtain
Z ` −ϑ x −ϑ0 `
0
e δ2 ψ0 Sdx + ς3 ψ0 (0) − δ2 ς4 ψ0 (`)e δ2 = 0.
0
Now, we check the transversality condition in the sense that
Dϑ(U,S) F(ϑ0 , U ∗ (x, δ1 , ϑ0 ), 0)(φ1 , ψ0 )T ∈
/ R(D(U,S) F(ϑ0 , U ∗ (x, δ1 , ϑ0 ), 0)).
Direct computation indicates
Dϑ(U,S) F(ϑ0 , U ∗ (x, δ1 , ϑ0 ), 0)(φ1 , ψ0 )T
2γ∂U ∗ (x, δ1 , ϑ) ∂U ∗ (x, δ1 , ϑ)
 
 −(φ0 )x − φ0 − g 0 (U ∗ (x, δ1 , ϑ0 )) ψ0
k∂ϑ ∂ϑ

ϑ=ϑ0 ϑ=ϑ0 
 ∗ 
∂U (x, δ1 , ϑ)

−(ψ0 )x + εg 0 (U ∗ (x, δ1 , ϑ0 ))
 
 ψ0 
∂ϑ

 ϑ=ϑ0 
 
=
 −φ0 (0) .

 

 −ψ0 (0) 

 

 0 

0
and
l(Dϑ(U,S) F(ϑ0 , U ∗ (x, δ1 , ϑ0 ), 0)(φ0 , ψ0 )T )

`
∂U ∗ (x, δ1 , ϑ0 )
Z −ϑ0  
x
= e δ2
ψ0 − (ψ0 )x + εg 0 (U ∗ (x, δ1 , ϑ0 )) dx − ψ02 (0) < 0.
∂ϑ

0 ϑ=ϑ0

Hence, there admits a ε  1 such that positive solutions (ϑ(τ ), U (τ, x), S(τ, x))(τ ∈
(0, ε), x ∈ [0, L]) of system (11) bifurcates from (U ∗ (x, δ1 , ϑ), 0) at ϑ = ϑ0 . These
solutions are on a smooth curve
Γ = {(ϑ(τ ), U (τ, x), S(τ, x)) : τ ∈ (0, ε), x ∈ [0, L]},
14 SHUAI LI, SANLING YUAN AND HAO WANG

where U (τ, x) = U ∗ (x, δ1 , ϑ) + τ φ0 (x) + τ ω1 (τ, x), S(τ, x) = τ ψ0 (x) + τ ω2 (τ, x),
ϑ(0) = ϑ0 , and ωi (x, 0) = 0 for i = 1, 2. Since
2
D(U,S) F(ϑ0 , U ∗ (x, δ1 , ϑ0 ), 0)[(φ0 , ψ0 )T , (φ0 , ψ0 )T ]
−2φ20 − 2g 0 (U ∗ (x, δ1 , ϑ0 ))ψ0 φ0
 

2g 0 (U ∗ (x, δ1 , ϑ0 ))ψ0 φ0


 
 
 
0
 
 
=
 ,
0

 
 
0
 
 
0
we can deduce that
2
< l, D(U,S) F(ϑ0 , U ∗ (x, δ1 , ϑ0 ), 0)[(φ0 , ψ0 )T , (φ0 , ψ0 )T ] >
ϑ0 (0) = −
2 < l, Dϑ(U,S) F(ϑ0 , U ∗ (x, δ1 , ϑ0 ), 0)(φ0 , ψ0 )T >
R ` −ϑ0
 0 e δ2 x g 0 (U ∗ (x, δ1 , ϑ0 ))ψ02 φ0 dx
= R −ϑ0 ∗
< 0.
` δ2 x 0 (U ∗ (x, δ , ϑ )) ∂U (x,δ1 ,ϑ)
 2 (0)
0
e ψ0 (ψ 0 ) x − g 1 0 ∂ϑ ψ 0 dx + ψ0
ϑ=ϑ0
By appealing to the stability theorem in [9], the stability of the bifurcation solution
is determined by the sign of −τ ϑ0 (s)χ1ϑ (δ2 , ϑ0 , U ∗ (x, δ1 , ϑ0 )). We can verify that
χ1ϑ (δ2 , ϑ0 , U ∗ (x, δ1 , ϑ0 )) < 0, hence the bifurcation solution (ϑ(τ ), U (τ, x), S(τ, x))
exists for ϑ ∈ (ϑ0 − ε, ϑ0 ) and the solution is locally asymptotically stable for
τ ∈ (0, ε). Moreover, we can check that the branch Γ of positive solutions to system
(11) bifurcated from (ϑ0 , U ∗ (x, δ1 , ϑ0 ), 0) can extend to ϑ = 0 according to Theorem
4.4 in [38].
Remark 1. Crandall and Rabinowitz put forward the bifurcation theorem of simple
eigenvalues for infinite dimensional dynamical systems [9]. Then, Shi and Wang
further refined such theory [38]. Afterward, the bifurcation theorem is widely used
to study the existence of positive steady state solution to parabolic systems since
the theory can provide more information around the bifurcation point, such as the
direction as well as the local stability of the bifurcation solution. We also notice that
when the functional response function g(U ) is Holling type I, we can employ the
direct method [33] or the topological degree theory [21, 50] to obtain the uniqueness
of the bifurcating non-constant steady state.

5. The long-time behavior of the full system (1). We are mainly concerned
with the conditions for the prevalence of disease and those for the local stability
of the disease-free attractor as well as those for the extinction of both prey and
predator species for system (1) in this section.
5.1. Uniformly persistence. Because of the advection and nonlinearity, the
uniqueness and global stability of positive steady state for subsystem (10) with-
out infected predator are unknown. We will provide some conditions to ensure that
there exists at least one positive coexistence state of system (1) by appealing to per-
sistence theory. Let A0 be the global attractor of disease-free system (10), whose
existence can be guaranteed by Theorem 3.3. Let P : X2 → X1 be the projection
on X2 which is defined by P(U, S) = S for (U, S) ∈ X2 . Therefore, we prove the
following results on the coexisting steady state for system (1).
ADVECTIVE EPIDEMIOLOGICAL PREDATOR-PREY DYNAMICS 15

Theorem 5.1. Let u(x, t, u0 ) = (U (x, t, u0 ), S(x, t, u0 ), I(x, t, u0 )) be the solution


of system (1) with u0 ∈ X3 and Ψ(t) be the semiflow generated by system (1). If
λ1 (δ1 , ϑ, γ) > 0, RP > 1 and λ1 (δ3 , ϑ, h0 (0) inf S∈P(A0 ) S(x)−σ−ξ) > 0, then system
(1) admits at least one positive steady state, and we can find an η > 0 such that for
any u0 ∈ X3 with u0 6≡ 0, one can obtain lim inf t→∞ (U (x, t, u0 ), S(x.t, u0 ), I(x, t,
u0 )) ≥ (η, η, η) uniformly on x ∈ [0, `].
Proof. Due to λ1 (δ1 , ϑ, γ) > 0 and RP > 1, we can obtain from Theorem 4.2 (ii)
e 0 < ζe < ζ, there has a T > 0 such that (U (x, t, v0 ), S(x, t, v0 )) ≥
that for given ζ,
(ζ, e for all t > T, x ∈ [0, `]. Obviously, (U (x, t, v0 ), S(x, t, v0 ), 0) is a solution of
e ζ),
system (1) for all t ∈ R. Let
W0 := {u ∈ X3 : ui (·) 6≡ 0, i = 1, 2, 3},
and
∂W0 = X3 \W0 = {u ∈ X3 : u1 (·) ≡ 0 or u2 (·) ≡ 0 or u3 (·) ≡ 0}.
Define
N∂ = {u0 ∈ ∂W0 : Ψ(t)u0 ∈ ∂W0 , t ≥ 0}.
Denote E1 = {(0, 0, 0)}, E2 = {(U ∗ (x, δ1 , ϑ), 0, 0)}, E3 = A0 × {0} and let ω(u0 )
be the omega limit set of the forward orbit γ + (u0 ) := {Ψ(t)u0 : t ≥ 0} for u0 ∈ X3 .
We can check that Ψ(t)W0 ⊂ W0 . By similar proof as Theorem 4.2, the following
two arguments hold.
S S S
Claim 1. u0 ∈N∂ ω(φ) = E1 E2 E3 .
Claim 2. Ei , i = 1, 2 are uniform weak repellers for W0 . That is, there exists a
constant κ > 0 such that lim supt→∞ k Ψ(t)u0 − Ei k≥ κ, i = 1, 2 for all u0 ∈ W0 .
Moreover, we can further have the following claim on E3 .
Claim 3. E3 is a uniform weak repellers for W0 . That is, there exists a constant
ι > 0 such that lim supt→∞ k Ψ(t)u0 − E3 k≥ ι for all u0 ∈ W0 . In fact, suppose
this is not true, then there exist a t3 > 0 and a φ0 ∈ W0 such that k Ψ(t)φ0 −E3 k< ι
for all t > t3 . By Assumption (A2), it is easy to check I(x, t, φ0 ) meets
It > δ3 Ixx − ϑIx + h0 (ι) inf (S(x) − ι)I − (σ + ξ)I, x ∈ (0, `), t > t3 ,
(
S∈P(A0 )
ϑI(0, t) − δ3 Ix (0, t) = 0, Ix (`, t) = 0, t > t3 .
Since λ1 (δ3 , ϑ, h0 (0) inf S∈P(A0 ) S(x) − σ − ξ) > 0, we can take ι sufficiently small
such that λ1 (δ3 , ϑ, h0 (ι) inf S∈P(A0 ) (S(x) − ι) − σ − ξ) > 0. Let wι be the positive
eigenfunction with respect to λ1 (δ3 , ϑ, h0 (ι) inf S∈P(A0 ) (S(x) − ι) − σ − ξ). Then
W(x, t) = exp{λ1 (δ3 , ϑ, h0 (ι) inf S∈P(A0 ) (S(x) − ι) − σ − ξ)t}wι (x) is a solution of
the following equation:
Wt = δ3 Wxx − ϑWx + h0 (ι) inf (S(x) − ι)W − (σ + ξ)W, x ∈ (0, `), t > t3 ,
(
S∈P(A0 )
ϑW(0, t) − δ3 Wx (0, t) = 0, Wx (`, t) = 0, t > t3 .
It follows from the comparison principle that there exists a constant C3 > 0 such
that
I(x, t, φ0 ) ≥ C3 exp{λ1 (δ3 , ϑ, h0 (ι) inf (S(x) − ι) − σ − ξ)t}wι (x), for all t ≥ t3 .
S∈P(A0 )

λ1 (δ3 , ϑ, h0 (ι) inf S∈P(A0 ) (S(x)−ι)−σ −ξ) > 0 indicates that I(x, t, φ0 ) is unbounded
for large t, which leads to a contradiction.
16 SHUAI LI, SANLING YUAN AND HAO WANG

Following the similar arguments as in the last step in the proof of Theorem 4.2,
the proof is thus completed.
5.2. Extinction results. In this subsection, we delve into the conditions for the
local stability of the global attractor without disease and the global extinction of
both predator and prey species by dint of the method adopted in [52].
Theorem 5.2. (i) If λ1 (δ1 , ϑ, γ) > 0 and RP < 1, then for any u0 = (U 0 , S 0 , I 0 )
with U 0 6≡ 0, we have limt→∞ (U (x, t, u0 ), S(x, t, u0 ), I(x, t, u0 )) = (U ∗ (x, δ1 , ϑ), 0,
0). (ii) If λ1 (δ1 , ϑ, γ) > 0, RP > 1 and λ1 (δ3 , ϑ, h0 (0) supS∈P(A0 ) S(x) − σ − ξ) < 0,
then there exists an ζ > 0 such that for u0 ∈ X3 with U 0 6≡ 0, S 0 6≡ 0, I 0 6≡ 0  1,
we have limt→∞ (U (x, t, u0 ), S(x, t, u0 )) ≥ (ζ, ζ) and limt→∞ I(x, t, u0 ) = 0.
Proof. From Theorem 4.2 we can easily know that (i) naturally holds. Now we
proceed to prove (ii). Obviously, there exists a small % > 0 such that
λ1 (δ3 , ϑ, h0 (0) sup (S(x) + %) − σ − ξ) < 0. (21)
S∈P(A0 )

We can obtain from system (1) that (U (x, t), I(x, t)) meets the following equations
with Danckwert’s boundary conditions:
 U = δ U − ϑU + γU 1 − U − g(U )S, x ∈ (0, `), t > 0,
  
t 1 xx x
k
St ≤ δ2 Sxx − ϑSx + g(U )S − σS, x ∈ (0, `), t > 0.

This means that the first two equations in system (1) can be controlled by disease-
free subsystem (10). Hence, there exists a t4 > 0 such that when t > t4 , we have
S(x, t) ≤ sup (S(x) + %).
S∈P(A0 )

It then follows from the third equation of I(x, t) in system (1) that when t > t4 ,
It ≤ δ3 Ixx − ϑIx + h0 (0) sup (S(x) + %)I − (σ + ξ)I, 0 < x < `,
(
S∈P(A0 ) (22)
ϑI(0, t) − δ3 Ix (0, t) = 0, Ix (`, t) = 0.
Noticing (21), with the aid of comparison principle we can deduce from (22) that
limt→∞ I(x, t) = 0.
Now we can view I(x, t) as a function of t. As a result, (U (x, t), S(x, t)) satisfies
the following nonautonomous reaction-diffusion-advection equation with Danckw-
ert’s boundary conditions:
 U = δ U − ϑU + γU 1 − U − g(U )S, x ∈ (0, `), t > 0,
  
t 1 xx x
k (23)
St = δ2 Sxx − ϑSx + g(U )S − h(I)S − σS, x ∈ (0, `), t > 0.

Obviously, system (23) is asymptotic to system (10) due to limt→∞ I(x, t) = 0. By


appealing to the theory for asymptotically autonomous semiflows [42], A0 is also
the global compact attractor of the solution semiflow associated with system (23).
Theorem 4.2 implies that there admits an ζ such that
lim inf (S(x, t, u0 ), I(x, t, u0 )) ≥ (ζ, ζ).
t→∞

Similarly to the arguments above, we obtain the following theorem about the
global extinction of both the prey and predators for full system (1).
ADVECTIVE EPIDEMIOLOGICAL PREDATOR-PREY DYNAMICS 17

Theorem 5.3. Assume λ1 (δ1 , ϑ, γ) ≤ 0. Then for any initial datum u0 ∈ X3 , the
solution (U (x, t, u0 ), S(x, t, u0 ), I(x, t, u0 )) of system (1) satisfies
lim (U (x, t, u0 ), S(x, t, u0 ), I(x, t, u0 )) = (0, 0, 0).
t→∞

6. Numerical examples. In this section, we provide a numerical illustration to


U
support theoretical results. Here we choose g(U ) = ~+U and h(I) = βI and use the
following parameter values from references [2, 16, 52]:
` = 1, γ = 2, k = 1, ~ = 0.5,  = 0.9, ξ = 0.2, δ1 = δ2 = 0.4, δ3 = 0.2.
By selecting suitable σ and ϑ, we can observe that system (10) without infected
predator has a unique global positive state (U ∗ (x), S ∗ (x)). Hence, we obtain the
following equations for the infection component by linearizing system (1) at (U ∗ (x),
S ∗ (x)):
It = δ3 Ixx − ϑIx + βS ∗ (x)I − (σ + ξ)I, x ∈ (0, `), t > 0,
(
(24)
ϑI(0, t) − δ3 Ix (0, t) = 0, Ix (`, t) = 0, t > 0.
Suppose that state variables are near equilibrium (N ∗ (x), S ∗ (x)) and the distri-
bution of initial infection individuals is governed by I 0 (x). Denote by Q(t) the
C0 −semigroup generated by the following system:
zt = δ3 zxx − ϑzx − (σ + ξ)z, x ∈ (0, `), t > 0,
(

ϑz(0, t) − δ3 zx (0, t) = 0, zx (`, t) = 0, t > 0.


0
Then Q(t)I (x) is the distribution of infectious cases as time evolves because of the
mortality and mobility of the infected members. Hence, the distribution of new
infection at time t is βS ∗ (x)Q(t)I 0 (x). Therefore, the distribution of accumulated
infectious cases during the infection period is governed by
Z ∞
0
L(I (x))(x) = βS ∗ (x)Q(t)I 0 (x)dt,
0
where L is the next generation operator. The basic reproduction ratio is defined by
the spectral radius of L, that is
R0 = r(L).
In virtue of Theorem 2.11 in [31], we have the following result.
Lemma 6.1. R0 − 1 and λ1 (δ3 , ϑ, βS ∗ (x) − σ − ξ) have the same sign.
Due to the lack of explicit expression, we numerically compute R0 , RP by em-
ploying the direct and efficient method developed in [49]. For (σ, ϑ, β) = (0.35, 0.13,
1), we can evaluate that RP = 1.2021 > 1 and R0 = 1.0789 > 1. According to
Theorem 5.1, the disease persists, which is confirmed in Fig. 1. For (σ, ϑ, β) =
(0.35, 0.14, 1), we can calculate RP = 1.1727 > 1 and R0 = 0.9758 < 1. According
to Theorem 5.1, the disease should be eradicated from the system when the initial
datum of infected predator is small, which is confirmed in Fig. 2. Now, picking
advection rate ϑ as a bifurcation parameter, we observe that the moderate interval
can favor the eradication of disease and coexistence of prey and healthy predator
for σ = 0.35 and β = 1, as shown in Fig. 3.
Next, we fix σ = 0.1, ϑ = 0.14, and we observe that the dynamics of disease-free
system are ruled by periodic perturbation. In order to obtain the basic reproductive
RT
number R0 , we replace S ∗ (x) by S(x) = T1 0 S(x, t)dt in (24) where T is the period
18 SHUAI LI, SANLING YUAN AND HAO WANG

Figure 1. The temporal-spatial evolution of system (1) with


(σ, ϑ, β) = (0.35, 0.13, 1).

Figure 2. The temporal-spatial evolution of system (1) with


(σ, ϑ, β) = (0.35, 0.14, 1).

1
Umin
0.9
S min

0.8 I min

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4

Figure 3. Bifurcation diagram about ϑ with σ = 0.35, β = 1.


ADVECTIVE EPIDEMIOLOGICAL PREDATOR-PREY DYNAMICS 19

of S [2]. When β = 0.460, we can calculate RP = 2.3545 > 1, R0 = 1.0070 > 1.


Theorem 5.1 means the density of infected individuals I(x, t) is persistent (see Fig.
4). When β = 0.445, we obtain RP = 2.3545 > 1 and R0 = 0.9742 < 1. Theorem
5.2 indicates that the infected individuals I(x, t) converges to zero as time evolves
when the initial scale of infected predator is small (see Fig. 5). Now choosing
advection rate ϑ as a bifurcation parameter, we discover the moderate advection
rate can favor the eradication of disease and coexistence of prey and healthy predator
for σ = 0.1 and β = 0.445 (see Fig. 6).

Figure 4. The temporal-spatial evolution of system (1) with


(σ, ϑ, β) = (0.10, 0.14, 0.460).

Figure 5. The temporal-spatial evolution of system (1) with


(σ, ϑ, β) = (0.10, 0.14, 0.445).

7. Conclusions and discussions. It is acknowledged that prevention or manip-


ulations of the water environment have been suggested as the best ways to control
the disease. In an attempt to gain deeper insight into disease spread in rivers or
streams, we formulated and analyzed an epidemiological predator-prey system with
disease in predators in open advective environments to explore the effect of ad-
vection rate on the transmission of disease. In contrast to the system in [40, 51],
our PDE framework allows incorporating biological and environmental factors to
20 SHUAI LI, SANLING YUAN AND HAO WANG

Umin
1.2 S min
I min

0.8

0.6

0.4

0.2

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

Figure 6. Bifurcation diagram about ϑ with σ = 0.1, β = 0.445.

explore the coupled processes that reveal the general view of disease transmission
dynamics in rivers or streams.
The well-posedness of the system is studied to guarantee the system admits a
global attractor. In the absence of predator, there admits a critical advection rate
ϑ∗ . The prey species are extirpated provided ϑ > ϑ∗ . For the disease-free subsys-
tem, we introduce the net reproductive rate RP which is the spectral radius of the
next generation operator of L. We rigorously show that RP = 1 is the threshold
for the invasion of predators. The susceptible predator can survive if RP > 1. The
results indicate that RP > 1 if the death rate of healthy predators and water flow
speed is small. We also performed the bifurcation analysis of steady state solutions
for the disease-free subsystem by picking advection rate ϑ as the bifurcation param-
eter. By appealing to the theory of persistence, we obtain sufficient conditions for
disease transmission. The conditions on the local stability of disease-free attractor
are also obtained using the principle of comparison. To illustrate the obtained the-
oretical results, we present a numerical example, where a basic reproduction ratio
R0 is defined under the circumstances that the disease-free system admits a global
positive state or disease-free dynamics are ruled by periodic perturbation. We adopt
a direct algorithm developed in [49] to numerically compute R0 by solving the asso-
ciated principal eigenvalue problem. It turns out that the moderate advection rate
can favor the eradication of disease and coexistence of prey and healthy predator.
That is to say, there exists such a situation that water current can also convey
benefits which is pointed out by Frank Hilker and Mark Lewis [16].
We also notice that a discontinuous jump occurs from a periodic disease-free
solution to a positive steady state of the full system in Fig. 6. We conjugate
that system (1) may undergo a backward bifurcation at R0 = 1 (corresponding
to a forward subcritical transcritical bifurcation when the advection rate is equal
to the critical value). We take system (1) without diffusions and advection as an
illustration. We take ξ as a bifurcation parameter by fixing other parameters as in
Fig. 6. We then plot the bifurcation diagram I with respect to R0 and ξ respectively
in Fig. 7. Indeed, system can exhibit a backward bifurcation at R0 = 1 (ξ = 0.3693).
We further observe that system (1) without diffusions and advection can undergo a
saddle-node bifurcation at ξ = 0.4006. That is, the equilibrium I can exhibit a jump
from zero to a positive constant as ξ crosses through the saddle-node point in the
negative direction. From a biological standpoint, we need to adjust the advection
ADVECTIVE EPIDEMIOLOGICAL PREDATOR-PREY DYNAMICS 21

rate according to the scale of infected predator. For rivers without diseases, we just
require the advection rate ϑ satisfying R0 < 1 to prevent the outbreak of disease.
However, the advection rate should be larger to eradicate disease when the density
of infected predator is higher.

1.2 1.2

1 1

0.8 0.8

0.6 0.6
LP LP
0.4 0.4

0.2 0.2

0 BP 0 BP
0.8 0.9 1 1.1 1.2 0 0.1 0.2 0.3 0.4 0.5

(a) Bifurcation diagram about R0 (b) Bifurcation diagram about ξ

Figure 7. Bifurcation diagram of I with respect to R0 and ξ.

There are a lot of interesting problems deserving further investigation. The


advection speed and other biological parameters may differ at different locations
[37]. How does the advection rate alter the dynamics of the system in spatially
heterogeneous environments? We can easily employ the Lyapunov-Schmidit method
to investigate the transcritical and saddle-node bifurcation of the system (1) without
diffusions and advection since the equilibrium has an explicit expression [13]. How
to adopt the Lyapunov-Schmidit method to study these bifurcations of system (1)
seems difficult. Moreover, the uniqueness and stability of interior steady state for
the disease-free subsystem and the full system are mathematically challenging. We
leave these questions as open problems.

Acknowledgments. We would like to thank the editor and two anonymous refer-
ees whose comments and suggestions have led to improvements in the manuscript.

REFERENCES

[1] M. Ballyk, L. Dung, D. A. Jones and H. L. Smith, Effects of random motility on microbial
growth and competition in a flow reactor, SIAM J. Appl. Math., 59 (1999), 573–596.
[2] A.-M. Bate and F. M. Hilker, Predator–prey oscillations can shift when diseases become
endemic, J. Theor. Biol., 316 (2013), 1–8.
[3] S. J. Bjork and J. L. Bartholomew, The effects of water velocity on the Ceratomyxa shasta
infectious cycle, J. Fish Dis., 32 (2009), 131–142.
[4] L. R. Bodensteiner, R. J. Sheehan, P. S. Wills, A. M. Brandenburg and W. M. Lewis, Flowing
water: An effective treatment for Ichthyophthiriasis, J. Aquat. Anim. Health, 12 (2000), 209–
219.
[5] I. M. Bulai and F. M. Hilker, Eco-epidemiological interactions with predator interference and
infection, Theor. Popul. Biol., 130 (2019), 191–202.
[6] R. S. Cantrell and C. Cosner, Spatial Ecology Via Reaction-Diffusion Equations, John Wiley
& Sons, 2003.
[7] L. Carraro, L. Mari, M. Gatto, A. Rinaldo and E. Bertuzzo, Spread of proliferative Kidney
disease in fish along stream networks: A spatial metacommunity framework, Freshw. Biol.,
63 (2018), 114–127.
22 SHUAI LI, SANLING YUAN AND HAO WANG

[8] J. Chattopadhyay and O. Arino, A predator-prey model with disease in the prey, Nonlinear
Anal., 36 (1999), 747–766.
[9] M.-G. Crandall and P.-H. Rabinowitz, Bifurcation, perturbation of simple eigenvalues and
linearized stability, Arch. Ration. Mech. Anal., 52 (1973), 161–180.
[10] O. Diekmann, J. A. P. Heesterbeek and J. A. J. Metz, On the definition and the computation of
the basic reproduction ratio R0 in models for infectious diseases in heterogeneous populations,
J. Math. Biol., 28 (1990), 365–382.
[11] T. Dondè, Uniform persistence in a prey–predator model with a diseased predator, J. Math.
Biol., 80 (2020), 1077–1093.
[12] L. Dung, Dissipativity and global attractors for a class of quasilinear parabolic systems,
Commun. Partial. Differ. Equ., 22 (1997), 413–433.
[13] S. J. Guo and J. H. Wu, Bifurcation Theory of Functional Differential Equations, vol. 10,
Springer, 2013.
[14] K. P. Hadeler and H. I. Freedman, Predator-prey populations with parasitic infection, J.
Math. Biol., 27 (1989), 609–631.
[15] H. W. Hethcote, W. D. Wang, L. T. Han and Z. E. Ma, A predator–prey model with infected
prey, Theor. Popul. Biol., 66 (2004), 259–268.
[16] F. M. Hilker and M. A. Lewis, Predator–prey systems in streams and rivers, Theor. Ecol., 3
(2010), 175–193.
[17] F. M. Hilker and K. Schmitz, Disease-induced stabilization of predator–prey oscillations, J.
Theor. Biol., 255 (2008), 299–306.
[18] Y.-H. Hsieh and C.-K. Hsiao, Predator–prey model with disease infection in both populations,
Math. Med. Biol., 25 (2008), 247–266.
[19] S.-B. Hsu and Y. Lou, Single phytoplankton species growth with light and advection in a
water column, SIAM J. Appl. Math., 70 (2010), 2942–2974.
[20] Q. H. Huang, H. Wang and M. A. Lewis, A hybrid continuous/discrete-time model for invasion
dynamics of zebra mussels in rivers, SIAM J. Appl. Math., 77 (2017), 854–880.
[21] D. H. Jiang, H. Nie and J. H. Wu, Crowding effects on coexistence solutions in the unstirred
chemostat, Appl. Anal.,, 96 (2017), 1016–1046.
[22] B. W. Kooi and E. Venturino, Ecoepidemic predator–prey model with feeding satiation, prey
herd behavior and abandoned infected prey, Math. Biosci., 274 (2016), 58–72.
[23] M. G. Krein and M. A. Rutman, Linear operators leaving invariant a cone in a Banach space,
Uspekhi Mat. Nauk (N.S.), 3 (1948), 3–95.
[24] Y. Lou and F. Lutscher, Evolution of dispersal in open advective environments, J. Math.
Biol., 69 (2014), 1319–1342.
[25] Y. Lou, H. Nie and Y. Wang, Coexistence and bistability of a competition model in open
advective environments, Math. Biosci., 306 (2018), 10–19.
[26] Y. Lou, D. M. Xiao and P. Zhou, Qualitative analysis for a Lotka-Volterra competition system
in advective homogeneous environment, Discrete Continuous Dyn. Syst. Ser. B , 36 (2016),
953–969.
[27] Y. Lou and P. Zhou, Evolution of dispersal in advective homogeneous environment: The effect
of boundary conditions, J. Differential Equations, 259 (2015), 141–171.
[28] P. Magal and X.-Q. Zhao, Global attractors and steady states for uniformly persistent dy-
namical systems, SIAM J. Math. Anal., 37 (2005), 251–275.
[29] A. P. Maiti, C. Jana and D. K. Maiti, A delayed eco-epidemiological model with nonlin-
ear incidence rate and Crowley–Martin functional response for infected prey and predator,
Nonlinear Dyn., 98 (2019), 1137–1167.
[30] R. H. Martin and H. L. Smith, Abstract functional-differential equations and reaction-diffusion
systems, Trans. Amer. Math. Soc., 321 (1990), 1–44.
[31] H. W. Mckenzie, Y. Jin, J. Jacobsen and M. A. Lewis, R0 analysis of a spatiotemporal model
for a stream population, SIAM J. Appl. Dyn. Syst., 11 (2012), 567–596.
[32] H. Nie, C. R. Liu and Z. G. Wang, Global dynamics of an ecosystem in open advective
environments, Internat. J. Bifur. Chaos Appl. Sci. Engrg. , 31 (2021), Paper No. 2150087,
24 pp.
[33] H. Nie, B. Wang and J. H. Wu, Invasion analysis on a predator–prey system in open advective
environments, J. Math. Biol., 81 (2020), 1429–1463.
ADVECTIVE EPIDEMIOLOGICAL PREDATOR-PREY DYNAMICS 23

[34] J. A. Patz, P. Daszak, G. M. Tabor, A. A. Aguirre, M. Pearl, J. Epstein, N. D. Wolfe, A. M.


Kilpatrick, J. Foufopoulos, D. Molyneux et al., Unhealthy landscapes: Policy recommenda-
tions on land use change and infectious disease emergence, Environ. Health Perspect., 112
(2004), 1092–1098.
[35] L.-E. Polvi, L. Lind, H. Persson, A. Miranda-Melo, F. Pilotto, X. L. Su and C. Nilsson, Facets
and scales in river restoration: Nestedness and interdependence of hydrological, geomorphic,
ecological, and biogeochemical processes, J. Environ. Manage., 265 (2020), 110288.
[36] V. Schakau, F. M. Hilker and M. A. Lewis, Fish disease dynamics in changing rivers:
Salmonid Ceratomyxosis in the Klamath River, Ecol. Complex., 40 (2019), 100776.
[37] Y. Shao, J. B. Wang and P. Zhou, On a second order eigenvalue problem and its application,
J. Differential Equations, 327 (2022), 189–211.
[38] J. P. Shi and X. F. Wang, On global bifurcation for quasilinear elliptic systems on bounded
domains, J. Differential Equations, 246 (2009), 2788–2812.
[39] D.-C. Speirs and W.-S. Gurney, Population persistence in rivers and estuaries, Ecology, 82
(2001), 1219–1237.
[40] M. Su and C. Hui, An eco-epidemiological system with infected predator, in 2010 3rd Inter-
national Conference on Biomedical Engineering and Informatics, vol. 6, 2010, 2390–2393.
[41] D. Tang and Y. M. Chen, Global dynamics of a Lotka–Volterra competition-diffusion system
in advective heterogeneous environments, SIAM J. Appl. Dyn. Syst., 20 (2021), 1232–1252.
[42] H. R. Thieme, Convergence results and a Poincaré-Bendixson trichotomy for asymptotically
autonomous differential equations, J. Math. Biol., 30 (1992), 755–763.
[43] G. Timmerhaus, C. C. Lazado, N. A. R. Cabillon, B. K. M. Reiten and L. H. Johansen, The
optimum velocity for Atlantic salmon post-smolts in RAS is a compromise between muscle
growth and fish welfare, Aquaculture, 532 (2021), 736076.
[44] O. Vasilyeva and F. Lutscher, How flow speed alters competitive outcome in advective envi-
ronments, Bull. Math. Biol., 74 (2012), 2935–2958.
[45] J. F. Wang and H. Nie, Invasion dynamics of a predator-prey system in closed advective
environments, J. Differential Equations, 318 (2022), 298–322.
[46] X. Y. Wang, R. W. Wu and X.-Q. Zhao, A reaction-advection-diffusion model of cholera
epidemics with seasonality and human behavior change, J. Math. Biol., 84 (2022), Paper No.
34, 30 pp.
[47] Y. N. Xiao and L. S. Chen, Modeling and analysis of a predator–prey model with disease in
the prey, Math. Biosci., 171 (2001), 59–82.
[48] X. Yan, H. Nie and P. Zhou, On a competition-diffusion-advection system from river ecology:
Mathematical analysis and numerical study, SIAM J. Appl. Dyn. Syst., 21 (2022), 438–469.
[49] T. H. Yang and L. Zhang, Remarks on basic reproduction ratios for periodic abstract func-
tional differential equations, Discrete Continuous Dyn. Syst. Ser. B , 24 (2019), 6771–6782.
[50] J. M, Zhang, J. P. Shi and X. Y. Chang, A mathematical model of algae growth in a pelagic-
benthic coupled shallow aquatic ecosystem, J. Math. Biol., 76 (2018), 1159–1193.
[51] J.-F. Zhang, W.-T. Li and X.-P. Yan, Hopf bifurcation and stability of periodic solutions in
a delayed eco-epidemiological system, Appl. Math. Comput., 198 (2008), 865–876.
[52] L. Zhang, W.-X. Shi and S.-M. Wang, A nonlocal and time-delayed reaction–diffusion eco-
epidemiological predator–prey model, Comput. Math. Appl., 77 (2019), 2534–2552.
[53] X.-Q. Zhao, Dynamical Systems in Population Biology, 2ed edition, Springer, Cham, 2017.
[54] P. Zhou and Q. H. Huang, A spatiotemporal model for the effects of toxicants on populations
in a polluted river, SIAM J. Appl. Math., 82 (2022), 95–118.

Received February 2022; revised May 2022; early access July 2022.
E-mail address: xynu sx lishuai@163.com
E-mail address: sanling@usst.edu.cn
E-mail address: hao8@ualberta.ca

You might also like