DISCRETE AND CONTINUOUS doi:10.3934/dcdsb.

2021048
DYNAMICAL SYSTEMS SERIES B

PERSISTENCE OF MOSQUITO VECTOR AND DENGUE:

IMPACT OF SEASONAL AND DIURNAL
TEMPERATURE VARIATIONS

Naveen K. Vaidya
Department of Mathematics and Statistics, San Diego State University
San Diego, CA 92182, USA
Computational Science Research Center, San Diego State University
San Diego, CA 92182, USA
Viral Information Institute, San Diego State University
San Diego, CA 92182, USA

Feng-Bin Wang
Department of Natural Science in the Center for General Education
Chang Gung University, Guishan, Taoyuan 333, Taiwan
Community Medicine Research Center, Chang Gung Memorial Hospital
Keelung Branch, Keelung 204, Taiwan

(Communicated by Sze-Bi Hsu)

Abstract. Dengue, a mosquito-borne disease, poses a tremendous burden to

human health with about 390 million annual dengue infections worldwide. The
environmental temperature plays a major role in the mosquito life-cycle as well
as the mosquito-human-mosquito dengue transmission cycle. While previous
studies have provided useful insights into the understanding of dengue diseases,
there is little emphasis put on the role of environmental temperature variation,
especially diurnal variation, in the mosquito vector and dengue dynamics. In
this study, we develop a mathematical model to investigate the impact of sea-
sonal and diurnal temperature variations on the persistence of mosquito vector
and dengue. Importantly, using a threshold dynamical system approach to
our model, we formulate the mosquito reproduction number and the infection
invasion threshold, which completely determine the global threshold dynam-
ics of mosquito population and dengue transmission, respectively. Our model
predicts that both seasonal and diurnal variations of the environmental tem-
perature can be determinant factors for the persistence of mosquito vector and
dengue. In general, our numerical estimates of the mosquito reproduction num-
ber and the infection invasion threshold show that places with higher diurnal
or seasonal temperature variations have a tendency to suffer less from the bur-
den of mosquito population and dengue epidemics. Our results provide novel
insights into the theoretical understanding of the role of diurnal temperature,
which can be beneficial for the control of mosquito vector and dengue spread.

2020 Mathematics Subject Classification. 34D20, 37N25, 92B05, 92D30, 92D40.

Key words and phrases. Dengue epidemics, diurnal temperature, global stability, periodic mod-
els, seasonal temperature.

1
2 NAVEEN K. VAIDYA AND FENG-BIN WANG

1. Introduction. With the worldwide situation of 2.5 billion people living in areas
with risk of dengue, about 390 million annual new dengue infections, and 500,000
annual hospitalization [4, 5, 41], dengue fever poses serious global health concerns.
Moreover, dengue fever is rapidly spreading in the world [4, 35, 41] affecting more
than a hundred countries, and actual cause for its rapid spread is still in debate. Be-
cause of uncertainty in its transmission mechanism and absence of licensed vaccines
or specific therapeutics, studies on the transmission dynamics of dengue are be-
coming increasingly important. Mathematical modeling can reveal some important
insights into dengue transmission dynamics.
Increase in dengue cases, particularly in tropical and subtropical regions, might
be due to multiple factors, including inefficient control of its vector (the Aedes ae-
gypti mosquito), expansion of the vector, human population growth, urbanization,
and climate change [3, 11, 23, 27, 34]. Among various causes, the majority of the
cases are linked to climatic factors and human and vector mobility [4, 3, 23, 25].
Specifically, it is known that the environmental temperature can highly affect the
dengue transmission [4, 35, 24], and as a consequence there remains substantial
threat that continuous increase of global temperature can exacerbate the geograph-
ical expansion of the endemic range of dengue [3, 24, 10, 14, 15, 42]. Therefore, it
is critical to understand how the environmental temperature can impact the trans-
mission dynamics of dengue epidemics.
Dengue viruses are transmitted to humans through bites of infected Aedes
mosquitos and uninfected mosquitos can become infected when they bite infected
humans. These inter-species transmission rates as well as incubation periods fol-
lowing transmission have been found to depend on the environmental temperature
[9, 17]. More importantly, a change in the environmental temperature can also alter
many mosquito entomological parameters [45, 43], such as oviposition rate, matu-
ration rate, and mortality rate, eventually impacting the dengue transmission dy-
namics. Limiting mosquito population is a commonly practiced method of dengue
control as effective vaccines or therapeutics are yet to be developed [4, 45, 31].
Therefore, it is important to include such temperature-influenced entomological
and transmission parameters properly into the transmission dynamics models in
order to provide useful information for developing prevention and control measures.
As far as the vector-borne diseases are concerned, the temporal variation of the
environmental temperature constitutes an important feature impacting disease dy-
namics [37, 40, 46]. It is a common practice to consider constant temperatures
or intraannual seasonally varying temperatures to study the effects of environmen-
tal temperature on the insect populations and pathogen transmission [17, 22, 28].
However, in case of dengue virus transmission, only seasonal variation of the en-
vironmental temperature can not fully explain the disease epidemics; the diurnal
(daily) temperature variation also plays an important role in the dengue trans-
mission dynamics [17]. For example, a temporal change of dengue epidemics in
Thailand is not associated with seasonal variation of the temperature, rather it
is associated with variation of diurnal temperature fluctuation [17, 8, 20, 29, 30].
Moreover, in nature, mosquitoes and their pathogens undergo temperature condi-
tions that fluctuate throughout the day, on top of experiencing seasonal conditions.
Hence, in addition to seasonal variation, diurnal temperature variation needs to be
 PERSISTENCE OF MOSQUITO VECTOR AND DENGUE 3

considered in order to accurately evaluate the effects of environmental temperature

on dengue epidemics.
While the previous studies have explored temporal dynamics [17, 18, 21, 26, 44]
and spatial spread [35, 23, 24, 15, 7, 36] of dengue, those existing models lack the
temperature profiles with coupled diurnal and seasonal variations. How such two
different periodic natures embedded in the environmental temperature can impact
on disease outcomes and prevention effectiveness still remains unknown. In this
study, we develop a model to describe the transmission dynamics of dengue under
the environmental temperature with seasonal as well as diurnal variations. Non-
linear effects of the environmental temperature in our non-autonomous model are
incorporated through entomological and dengue dynamics parameters, which are
derived based on laboratory experimental data. We focus on the theoretical and
numerical aspects of the non-autonomous model analysis to study how seasonal
and diurnal variations of the environmental temperature impact the mosquito re-
production number and the infection invasion threshold, consequently affecting the
persistence of mosquito vector and dengue.

2. Mathematical model. We consider mutually exclusive compartments related

to aquatic (A), susceptible (Ms ), exposed (Me ), and infected (Mi ) groups of the
female mosquitos as well as susceptible (Hs ), exposed (He ), infected (Hi ) and recov-
ered (Hr ) groups of humans. We denote the total female mosquito population and
the total human population by NM (t) and NH (t), respectively. Interactions cap-
tured in our model are shown in a schematic diagram (Fig. 2.1) and the governing
equations are given by the following system:

dA A


dt = kδ(t) 1 − C NM − (θ(t) + µa (t))A,



 dMs bβm (t)Ms Hi
dt = θ(t)A − − µm (t)Ms ,



 NH
 dMe bβm (t)Ms Hi



 dt = NH − (γm (t) + µm (t))Me ,
 dMi



 dt = γ m (t)M e − µ m (t)Mi ,
 dHs = Λ − µ H − bβh (t)Hs Mi ,

dt h h s NH
dHe bβh (t)Hs Mi (2.1)


 dt = NH − (γ h + µh )He ,
dHi

dt = γh He − (αh + µh )Hi ,




dHr

dt = αh Hi − µh Hr ,








 A(0) ≥ 0, Ms (0) ≥ 0, Me (0) ≥ 0, Mi (0) ≥ 0,

Hs (0) ≥ 0, He (0) ≥ 0, Hi (0) ≥ 0, Hr (0) ≥ 0.


In this model, matured mosquitos (NM = Ms + Me + Mi ) lay eggs with the

per capita oviposition rate δ(t), among which a fraction k (a combination of the
fraction of eggs hatching to larvae and the fraction of female mosquitoes hatched
from all eggs) grow to the aquatic phase (larvae and pupae). Similar to the concept
used in the standard logistic equation, a parameter C is introduced to represent
the carrying capacity of the mosquito aquatic population, giving the net rate of
A
change of the aquatic mosquito population as kδ(t)(1 − C )NM per unit time. The
aquatic mosquitos develop into adult female mosquitoes at a maturation rate θ(t)
and die at a mortality rate µa (t). We assume that female mosquitos newly emerged
from the aquatic phase are all susceptible, which die with a natural mortality rate
4 NAVEEN K. VAIDYA AND FENG-BIN WANG

µm (t). Here, Λh and µh represent recruitment rate and natural mortality rate of
susceptible humans.

Figure 2.1. A schematic diagram of the dengue transmission

model. A, Ms , Me , Mi : aquatic, susceptible, exposed, and infected
mosquitos. Hs , He , Hi , Hr : susceptible, exposed, infected, and re-
covered humans. Solid arrows represent birth, maturation, infec-
tion, transfer, death, while dashed arrows indicate the effects of
time dependent periodic environmental temperature, T (t).

Dengue virus transmission occurs from infected humans to susceptible mosquitos

and from infected mosquitos to susceptible humans through mosquito bites. Trans-
mission related parameters b, βm (t), and βh (t) represent the per capita biting rate of
mosquitoes, the transmission probability from human to mosquito, and the trans-
mission probability from mosquito to human, respectively. An average extrinsic
period of mosquitos (i.e., an average duration of mosquitos in exposed class) and an
average intrinsic period of humans (i.e., an average duration of humans in exposed
class) are denoted by 1/γm (t) and 1/γh , respectively. Infected humans get recovered
from dengue at a rate αh , and once recovered, these recovered humans do not loose
immunity as mentioned in Bhatt et al. [4]. As described in [4, 31], dengue virus
is generally non-pathogenic. Therefore, our model does not include deaths due to
disease.
Based on experimental evidences [9, 17, 44], our model incorporates the effects of
environmental temperature via entomological and dengue dynamics parameters. In
particular, parameters δ(t), µa (t), θ(t), µm (t), γm (t), βm (t), and βh (t) of our model
are influenced by the diurnally and seasonally varying environmental temperature.
In the absence of detailed diurnal and seasonal temperature profile, i.e., homo-
geneous environment, our system (2.1) is similar to the dengue model proposed
previously in Pinho et al. [26].
 PERSISTENCE OF MOSQUITO VECTOR AND DENGUE 5

2.1. Entomological parameters. We denote the temperature profile by T (t) that

captures both diurnal and seasonal variation of the environmental temperature. We
used the experimental data [43, 44], which measure how a change in temperature
changes the oviposition rate, aquatic phase mortality rate, rate of emergence of
female mosquito from aquatic phase, and female mosquito mortality rate. As in our
previous study [36], we fitted the data to appropriate functional curves (Fig. 2.2),
which give the following relationships:
(
δ , T (t) < aδ ,
δ(t) = δm [T (t)−aδ ]Nδ (2.2)
δ + N δ Nδ
, T (t) ≥ aδ ,
δh +[T (t)−aδ ]

µa (t) = a0µa + a1µa T (t) + a2µa T (t)2 + a3µa T (t)3 + a4µa T (t)4 , (2.3)

 θ , p T (t) < a1θ ,
θ(t) = θ + a0θ T (t) (T (t) − a1θ ) a2θ − T (t), a1θ ≤ T (t) ≤ a2θ , (2.4)
θ , T (t) > a2θ ,


µm (t) = a0µm + a1µm T (t) + a2µm T (t)2 . (2.5)

Mortality rate (Aquatic Phase)

Experimental data Experimental data

Function δ(T) 0.8 Function µa(T)
12
Oviposition rate

0.6
9

0.4
6

3 0.2

10 15 20 25 30 35 10 20 30 40
Temperature (oC) Temperature (oC)

0.3 Experimantal data Experimental data

Mosquito Emergence rate

0.15
Mortality rate (Mosquito)

Function θ(T) Function µm(T)

0.25
0.12
0.2
0.09
0.15
0.06
0.1

0.05 0.03

0 10 20 30 40 50 10 15 20 25 30 35
Temperature (oC) Temperature (oC)

Figure 2.2. Best-fit curves provided by the experimental data [44]

for δ(T ) (oviposition rate), µa (T ) (aquatic phase mortality rate),
θ(T ) (mosquito emergence rate from acuatic phase), and µm (T )
(mosquito mortality rate).

2.2. Dengue dynamics parameters. We follow the similar techniques imple-

mented in our previous study [36] to reasonably estimate the functional form of
dengue dynamics parameters depending on the environmental temperature. Briefly,
based on an enzyme kinetic model [9, 17], the incubation time period of dengue, i.e.
6 NAVEEN K. VAIDYA AND FENG-BIN WANG

the average duration for which mosquitos stay in exposed class before they become
infectious, is given by
1 a1γm T (t)
γm (t) = e . (2.6)
a0γm
Lambrechts et al. [17] have provided reasonable estimates of dengue transmission
probability from humans to mosquitos as well as from mosquitos to humans. Fol-
lowing the trend identified by Lambrechts et al. [17], we use the E-max model (Fig.
2.3) to describe the temperature-dependent formula of the transmission probability
from humans to mosquitos as follows [36]:

 βm , T (t) < aβm ,
βm (t) = N
[T (t)−aβm ] βm (2.7)
 βm + Nβm Nβ
, T (t) ≥ aβm .
βmh +[T (t)−aβm ] m

1 Generated data
Transmission probability

Function βm(T)
(human to mosquito)

0.8

0.6

0.4

0.2

0
0 7 14 21 28 35
Temperature (oC)

Figure 2.3. Best-fit curve provided by the data generated from

the previous estimates [17] for the transmission probability from
human to mosquito, βm (T ).

Similarly, as done in Lambrechts et al. [17], we approximate the temperature-

dependent transmission probability from mosquitos to humans to the proportion of
midgut-infected mosquitoes transmitting virus, and obtain the following expression:

 βh , p T (t) < a1βh ,
βh (t) = βh + a0βh T (t) (T (t) − a1βh ) a2βh − T (t), a1βh ≤ T (t) ≤ a2βh ,
βh , T (t) > a2βh .

(2.8)

2.3. Temperature profile. As done in previous studies [37], seasonal variation of

the environmental temperature can be described well by the following sinusoidal
function of seasonal period τm :
 
2π
Tm (t) = T0 + m sin t + φm ,
τm
where T0 represents the annual mean temperature, m denotes the amplitude of
the seasonal temperature, and φm represents the phase shift in the seasonal tem-
perature. To the seasonal mean temperature, Tm (t), we now impose the diurnal
 PERSISTENCE OF MOSQUITO VECTOR AND DENGUE 7

temperature fluctuation Ψd (t) using another sinusoidal function of period τd :

 
2π
Ψd (t) = d sin t + φd ,
τd
where d and φd denote the amplitude and the phase shift in the diurnal temperature
variation. As a result, net time varying temperature profile introduced into the
model is:
T (t) = Tm (t) + Ψd (t).
Note that Tm (t) is τm -periodic and Ψd (t) is τd -periodic. Then choosing τ as the
least common multiple of τm and τd , the temperature function T (t) = Tm (t)+Ψd (t)
becomes τ -periodic. Thus, δ(t), µa (t), θ(t), µm (t), γm (t), βm (t), and βh (t) of our
model all become τ -periodic, implying that our system is also τ -periodic. A typical
example of the temperature profiles of Tm (t), Ψd (t), and T (t) are shown in Fig. 2.4.

40 8 40

6
35 35
4
30 30
2
Tm (t) [o C]

(t) [o C]

T(t) [o C]
25 0 25
d

-2
20 20
-4
15 15
-6

10 -8 10
0 200 400 600 0 0.5 1 1.5 2 0 200 400 600
Time in day Time in day Time in day

Figure 2.4. Temperature profile of Tm (t) [Left], Ψd (t) [Middle],

and T (t) [Right]. Parameters used are T0 = 25o C, m = 5o C, τm =
365 day, φm = 0, d = 5o C, τd = 1 day, and φd = 0.

The model parameters are given in Table 2.1.

3. Mathematical analysis. From [32, Theorem 5.2.1], we can show that for any
(A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) ∈ R8+ ,
the system (2.1) has a unique local nonnegative solution
(A(t), Ms (t), Me (t), Mi (t), Hs (t), He (t), Hi (t), Hr (t)) ∈ R8+ .
Since
NH (t) = Hs (t) + He (t) + Hi (t) + Hr (t), (3.1)
(2.1) implies
dNH (t)
= Λh − µh NH (t). (3.2)
dt
By the same arguments as in Zhao [47, Section 5.2], we have the following result:
Λh
Lemma 3.1. System (3.2) admits a unique positive constant H ∗ := µh such that
every solution NH (t) of (3.2) with NH (0) ≥ 0 satisfies
lim NH (t) = H ∗ . (3.3)
t→∞
8 NAVEEN K. VAIDYA AND FENG-BIN WANG

Table 2.1. Model parameters

Parameter Description Value Reference

k Fraction of female larvae from eggs 0.5 (0-1) [18, 26]
b Per capita biting rate 0.1 [6, 26]
µh Natural death rate of humans 4.22×10−5 d−1 Calculated, [16]
1/γh Intrinsic period 10 days [6, 16, 18, 26]
αh Human recovery rate 0.1 d−1 [18, 26]
δm In δ(t) 9.531 Data fitting
δh In δ(t) 22.55 Data fitting
Nδ In δ(t) 7.084 Data fitting
aδ In δ(t) 0 Data fitting
δ In δ(t) 10−6 Data fitting
a0µa In µa (t) 2.914 Data fitting
a1µa In µa (t) -0.4986 Data fitting
a2µa In µa (t) 0.03099 Data fitting
a3µa In µa (t) -0.0008236 Data fitting
a4µa In µa (t) 7.975×10−6 Data fitting
a0θ In θ(t) 8.044×10−5 Data fitting
a1θ In θ(t) 11.386 Data fitting
a2θ In θ(t) 40.1461 Data fitting
θ In θ(t) 10−6 Data fitting
a0µm In µm (t) 0.1901 Data fitting
a1µm In µm (t) -0.0134 Data fitting
a2µm In µm (t) 2.739×10−4 Data fitting
a0γm In γm (t) 5×104/3 Data fitting
a1γm In γm (t) 0.0768 Data fitting
βmh In βm (t) 18.9871 Data fitting
Nβm In βm (t) 7 Data fitting
βm In βm (t) 10−6 Data fitting
aβm In βm (t) 0 Data fitting
a0βh In βh (t) 1.044×10−3 Data fitting
a1βh In βh (t) 12.286 Data fitting
a2βh In βh (t) 32.461 Data fitting
βh In βh (t) 10−6 Data fitting

Since
NM (t) = Ms (t) + Me (t) + Mi (t), (3.4)
we are able to demonstrate the following mass conservation for aquatic phase A(t)
and the total population of female mosquitos NM (t) in (2.1). By computations, we
obtain that (A(t), NM (t)) satisfies the following coupled differential equations
 dA A


 dt = kδ(t) 1 − C NM − (θ(t) + µa (t))A,

dNM
dt = θ(t)A − µm (t)NM ,
(3.5)

0 0
A(0) = A , NM (0) = NM .


It is easy to verify that (0, 0) is a trivial solution of the system (3.5).

3.1. Mosquito reproduction number. Following the approaches provided by

Bacaër and Guernaoui [2] and Wang and Zhao [40] for analysis of models with
 PERSISTENCE OF MOSQUITO VECTOR AND DENGUE 9

seasonality, we derive the mosquito reproduction number, RM , for our periodic

compartmental system (3.5). Linearizing the system (3.5) at (0, 0), we get the
following linear system
 dA
 dt = kδ(t)NM − (θ(t) + µa (t))A,

dNM
dt = θ(t)A − µm (t)NM ,
(3.6)

0 0
A(0) = A , NM (0) = NM .


In view of system (3.6), we assume

 
M 0 kδ(t)
F (t) = , (3.7)
0 0
and
 
M θ(t) + µa (t) 0
V (t) = . (3.8)
−θ(t) µm (t)
Suppose ΨVM (·) (t) is the monodromy matrix of the linear τ -periodic differential
system dv(t)
dt = VM (t)v, and r(ΨVM (·) (τ )) is the spectral radius of ΨVM (·) (τ ).
Assume Z(t, s), t ≥ s, is the evolution operator of the linear τ -periodic system
dz(t)
= −VM (t)z, (3.9)
dt
that is, for each s ∈ R, the 2 × 2 matrix Z(t, s) satisfies
d
Z(t, s) = −VM (t)Z(t, s), ∀t ≥ s, Z(s, s) = I,
dt
where I is the 2 × 2 matrix. Thus, the fundamental solution matrix Ψ−VM (·) (t) of
(3.9) is equal to Z(t, 0), t ≥ 0.
We assume that ϕ(s), τ -periodic in s, is the initial distribution of adult mosquito
population. Then FM (s)ϕ(s) is the rate of new aquatic mosquitos produced by the
adult mosquitos who were introduced at time s. Given t ≥ s, then Z(t, s)FM (s)ϕ(s)
gives the distribution of those adult mosquitos who were newly produced in aquatic
phase at time s and remain in the adult compartments at time t. It follows that
Z t Z ∞
ψ M (t) := Z(t, s)FM (s)ϕ(s)ds = Z(t, t − a)FM (t − a)ϕ(t − a)da
−∞ 0
is the distribution of accumulative new adult mosquitos at time t produced by all
those adult mosquitos ϕ(s) introduced at time previous to t.
Let CτM be the ordered Banach space of all τ -periodic functions from R to R2 ,
M
which is equipped with the maximum norm k · k and the positive cone Cτ,+ := {ϕ ∈
CτM : ϕ(t) ≥ 0, ∀ t ∈ R}. Then we define a linear operator LM : CτM → CτM by
Z ∞
(LM ϕ)(t) = Z(t, t − a)FM (t − a)ϕ(t − a)da, ∀ t ∈ R, ϕ ∈ CτM . (3.10)
0
M
Then L represents the next mosquito-generation operator [40], and we define the
mosquito reproduction number as RM := r(LM ), the spectral radius of LM .
Here, if W M (t, s, λM ), t ≥ s, s ∈ R is the monodromy matrix of the linear
τ -periodic system on R2
FM (t)
 
dw M
= −V (t) + w, t ∈ R, (3.11)
dt λM
10 NAVEEN K. VAIDYA AND FENG-BIN WANG

with parameter λM ∈ (0, ∞), then by Theorem 2.1 of [40], we have the following
results.

Lemma 3.2. The following statements hold

(i) If r W M (τ, 0, λM ) = 1 has a positive solution λM M
0 , then λ0 is an eigenvalue
of operator LM , and hence, RM > 0.

(ii) If RM > 0, then λM = RM is the unique
solution of r W M (τ, 0, λM ) = 1.
(iii) RM = 0 if and only if r W M (τ, 0, λM ) < 1 for all λM > 0.

RM for homogeneous (constant temperature) case. We now briefly men-

tion the homogeneous case, in which the environmental temperature remains con-
stant over time. In this case, δ(t) ≡ δ, θ(t) ≡ θ, µa (t) ≡ µa , and µm (t) ≡ µm
are all positive constants. Then FM (t) ≡ FM and VM (t) ≡ VM become constant
matrices. Substituting constant matrices FM and VM , we obtain (see also [40, 38])

−1
 kδθ kδ 
FM V M = µm (θ+µa ) µm
.
0 0

Then RM can be expressed as the following explicit form


−1  kδθ
RM = r FM VM = .
µm (θ + µa )
We have the results related to mosquito population extinction as stated in the
following Lemma.

Lemma 3.3. [40, Theorem 2.2] The following statements hold.

(i) RM = 1 if and only if r(ΨFM (·)−VM (·) (τ )) = 1;
(ii) RM > 1 if and only if r(ΨFM (·)−VM (·) (τ )) > 1;
(iii) RM < 1 if and only if r(ΨFM (·)−VM (·) (τ )) < 1.
Thus, the mosquito-free equilibrium (0, 0) is locally asymptotically stable for system
(3.5) if RM < 1, and unstable if RM > 1.

Moreover, we have the following results related to the global dynamics of (3.5):

Lemma 3.4. Let ∆ := {(A, M ) ∈ R2+ : 0 ≤ A ≤ C}. Then the following statements
hold.
(i) If RM < 1, then the trivial solution (0, 0) is globally attractive in ∆ for (3.5);
(ii) If RM > 1, then the system (3.5) admits a unique positive τ -periodic solu-
tion (A∗ (t), M ∗ (t)) which is globally attractive in ∆\{(0, 0)}, that is, for any
(A(0), NM (0)) ∈ ∆\{(0, 0)}, we have
lim [(A(t), NM (t)) − (A∗ (t), M ∗ (t))] = 0, (3.12)
t→∞
n o
µm (t)[θ(t)+µa (t)] θ(t)
where A∗ (t) = C 1 − kδ(t)θ(t) , and M ∗ (t) = ∗
µm (t) A (t).

Proof. From [32, Theorem 5.2.1], we observe that ∆ is positively invariant for system
(3.5). We first prove the following claim.
Claim.
A(t) < C, ∀ t > 0. (3.13)
 PERSISTENCE OF MOSQUITO VECTOR AND DENGUE 11

Proof of the claim. Assume, by contradiction, that (3.13) is not true. Then there
exists a t0 > 0 such that A(t0 ) = C. It follows from the first equation in system
(3.5) that
 
dA A(t0 )
0= = kδ(t0 ) 1 − NM (t0 ) − [θ(t0 ) + µa (t0 )]A(t0 ),
dt t=t0 C
which implies that θ(t0 ) + µa (t0 ) = 0. This contradiction shows that (3.13) holds.
From (3.13), it is easy to see that system (3.5) is strongly monotone in ∆ (see,
e.g., [32]). On the other hand, we assume that
(
A


g1 (A, NM ) = kδ(t) 1 − C NM − (θ(t) + µa (t))A,
g2 (A, NM ) = θ(t)A − µm (t)NM .
Then for all A > 0, NM > 0, 0 < ϑ < 1, we have
g1 (ϑA, ϑNM ) > ϑg1 (A, NM ), g2 (ϑA, ϑNM ) = ϑg2 (A, NM ),
that is, system (3.5) is strictly subhomogeneous in ∆ (see, e.g., [47]). Then we can
use Lemma 3.3, [47, Theorem 2.3.4], and the similar arguments as in [39, Lemma
2.5] to complete the rest of the proof.

Let
X := {(A, Ms , Me , Mi , Hs , He , Hi , Hr ) ∈ R8+ : 0 ≤ A ≤ C}.
Then we can prove the following Lemma.
Lemma 3.5. X is positively invariant for system (2.1) and the system (2.1) has a
unique and bounded solution with the initial value in X. Further, the system (2.1)
admits a connected global attractor G on X in the sense that G attracts all positive
orbits in X.
Proof. From [32, Theorem 5.2.1], we can observe that X is positively invariant for
system (2.1). By (3.4), (3.5), Lemma 3.1, and Lemma 3.4, it follows that solutions
of the system (2.1) are uniformly and ultimately bounded. Also, by [12, Theorem
3.4.8], it follows that the system (2.1) admits a connected global attractor G on
X.

3.2. Infection invasion threshold. In order to find the disease-free periodic state
of (2.1), we set Me = Mi = He = Hi = Hr = 0. Then Hs (t) and (A, Ms ) satisfy
(3.2) and (3.6), respectively. By Lemma 3.4 and Lemma 3.1, we see that
E0 = (A, Ms , Me , Mi , Hs , He , Hi , Hr ) = (0, 0, 0, 0, H ∗ , 0, 0, 0)
always exists, and
E1 (t) = (A, Ms , Me , Mi , Hs , He , Hi , Hr ) = (A∗ (t), M ∗ (t), 0, 0, H ∗ , 0, 0, 0)
exists if RM > 1.
Theorem 3.1. Assume that (A(t), Ms (t), Me (t), Mi (t), Hs (t), He (t), Hi (t), Hr (t))
is a solution of the system (2.1) with initial value (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 )
∈ X. If RM < 1, then
lim (A(t), Ms (t), Me (t), Mi (t), Hs (t), He (t), Hi (t), Hr (t)) = E0 .
t→∞
12 NAVEEN K. VAIDYA AND FENG-BIN WANG

Proof. If RM < 1, then by (3.4), (3.5) and Lemma 3.4 (i), we see that
lim A(t) = lim Ms (t) = lim Me (t) = lim Mi (t) = 0.
t→∞ t→∞ t→∞ t→∞

This implies that He is asymptotic to

dHe
= −(γh + µh )He ,
dt
which gives limt→∞ He (t) = 0. Similarly, Hi is asymptotic to dH
dt = −(αh + µh )Hi ,
i

dHr
and hence, limt→∞ Hi (t) = 0. Finally, Hr is asymptotic to dt = −µh Hr , and
hence, limt→∞ Hr (t) = 0. Given these asymptotic behaviors for He , Hi , and Hr ,
we see that Hs is asymptotic to system (3.2). Then by the theory of asymptot-
ical semiflows (see, e.g., [48] or [47, section 3.2]) and Lemma 3.1, it follows that
limt→∞ Hs (t) = H ∗ . This completes the proof.
Linearizing system (2.1) at the disease-free periodic state E1 (t), we get the fol-
lowing system for the (Me , Mi , He , Hi ) components:
∗
= −[γm (t) + µm (t)]Me + bβm (t)M (t)
 dM

 dt
e
H ∗ Hi ,
 dMi
 dt = γm (t)Me − µm (t)Mi ,



dHe
dt = bβh (t)Mi − (γh + µh )He ,
(3.14)

 dHi = γh He − (αh + µh )Hi ,

dt



Me (0) ≥ 0, Mi (0) ≥ 0, He (0) ≥ 0, Hi (0) ≥ 0.


We next introduce infection invasion threshold R0 for our periodic compartmen-

tal epidemic model (2.1) by using the general theory in [40]. From (3.14), we define
∗
0 bβm (t)M (t) 

0 0 H∗
 0 0 0 0 
F(t) =  0 bβh (t) 0
,
0 
0 0 0 0
and
 
γm (t) + µm (t) 0 0 0
 −γ m (t) µm (t) 0 0 
V(t) =  .
 0 0 γh + µh 0 
0 0 −γh αh + µh
Suppose ΦV(·) (t) is the monodromy matrix of the linear τ -periodic differential sys-
tem dz(t)
dt = V(t)z, and r(ΦV(·) (τ )) is the spectral radius of ΦV(·) (τ ). Assume
Y (t, s), t ≥ s, is the evolution operator of the linear τ -periodic system
dy(t)
= −V(t)y, (3.15)
dt
that is, for each s ∈ R, the 4 × 4 matrix Y (t, s) satisfies
d
Y (t, s) = −V(t)Y (t, s), ∀ t ≥ s, Y (s, s) = I,
dt
where I is the 4 × 4 matrix. Thus, the fundamental solution matrix Φ−V(·) (t) of
(3.15) is equal to Y (t, 0), t ≥ 0.
We assume that φ(s), τ -periodic in s, is the initial distribution of infectious
individuals. Then F(s)φ(s) is the rate of new infections produced by the infected
 PERSISTENCE OF MOSQUITO VECTOR AND DENGUE 13

individuals who were introduced at time s. Given t ≥ s, then Y (t, s)F(s)φ(s) gives
the distribution of those infected individuals who were newly infected at time s and
remain in the infected compartments at time t. It follows that
Z t Z ∞
ψ(t) := Y (t, s)F(s)φ(s)ds = Y (t, t − a)F(t − a)φ(t − a)da
−∞ 0

is the distribution of accumulative new infections at time t produced by all those

infected individuals φ(s) introduced at time previous to t.
Let Cτ be the ordered Banach space of all τ -periodic functions from R to R4 ,
which is equipped with the maximum norm k · k and the positive cone Cτ+ := {φ ∈
Cτ : φ(t) ≥ 0, ∀ t ∈ R}. Then we define a linear operator L : Cτ → Cτ by
Z ∞
(Lφ)(t) = Y (t, t − a)F(t − a)φ(t − a)da, ∀ t ∈ R, φ ∈ Cτ . (3.16)
0

Then L represents the next infection operator [40], and we define infection invasion
threshold as R0 := r(L), the spectral radius of L.
Here, if W(t, s, λ0 ), t ≥ s, s ∈ R is the monodromy matrix of the linear τ -periodic
system on R4
 
dw F(t)
= −V(t) + 0 w, t ∈ R, (3.17)
dt λ

with parameter λ0 ∈ (0, ∞), then by Theorem 2.1 of [40], we have the following
results.

Lemma 3.6. The following statements hold

(i) If r W(τ, 0, λ0 ) = 1 has a positive solution λ00 , then λ00 is an eigenvalue of
operator L, and hence, R0 > 0.

(ii) If R0 > 0, then λ0 = R0 is the unique

solution of r W(τ, 0, λ0 ) = 1.
(iii) R0 = 0 if and only if r W(τ, 0, λ0 ) < 1 for all λ0 > 0.

R0 for homogeneous (constant temperature) case. We now briefly mention

that R0 formulated above can also recover the basic reproduction number, R̄0 ,
that we derived previously for the dynamics when the environmental temperature
remains constant over time [36]. In the homogeneous (constant temperature) case,
i.e. T (t) = T , all of δ(t) = δ, µa (t) = µa , θ(t) = θ, µm (t) = µm , γm (t) = γm , βm (t) =
βm , βh (t) = βh are
 constant. In this case, E1 = (Ā∗ , M̄s∗ , 0, 0, H̄ ∗ , 0, 0, 0), where
Ā = C 1 − RM , M̄s = θĀ /µm , and H̄ ∗ = Λh /µh . Clearly, E1 exists if RM > 1.
∗ 1 ∗ ∗

Here, the system linearized about E1 provides both F(t) ≡ F and V(t) ≡ V to be
constant matrices. Substituting constant matrices F and V, we obtain (see also
[40, 38])
R0 = r(L) = r(FV−1 ).
By computations, we can obtain that
 
0 0 Dγh /(γh + µh )(αh + µh ) D/(αh + µh )
0 0 0 0
FV−1
 
=
 Eγm /µm (γm + µm )
,
E/µm 0 0 
0 0 0 0
14 NAVEEN K. VAIDYA AND FENG-BIN WANG

bβm Ms∗
where D = H̄ ∗
and E = bβh . Then R0 can be expressed as the following explicit
form s
0 DEγm γh
R = .
µm (γm + µm )(γh + µh )(αh + µh )
Here, R0 = R̄0 , which is the basic reproduction number [36] that we derived using
the second generation matrix method [38] for constant temperature case.
We have the result about the local stability of the disease-free state E1 (t) as
stated in the following theorem.

Lemma 3.7. [40, Theorem 2.2] The following statements hold.

(i) R0 = 1 if and only if r(ΦF(·)−V(·) (τ )) = 1;
(ii) R0 > 1 if and only if r(ΦF(·)−V(·) (τ )) > 1;
(iii) R0 < 1 if and only if r(ΦF(·)−V(·) (τ )) < 1.
Thus, the disease-free state E1 (t) is locally asymptotically stable if R0 < 1, and
unstable if R0 > 1.

3.3. Threshold dynamics. Let A(t) be a continuous, cooperative, irreducible,

and τ -periodic k × k matrix function. Suppose ΦA(·) (t) is the monodromy matrix
of the linear ordinary differential system
dx(t)
= A(t)x, (3.18)
dt
and r(ΦA(·) (τ )) is the spectral radius of ΦA(·) (τ ). From [1, Lemma 2] (see also [13,
Theorem 1.1]), it follows that ΦA(·) (t) is a matrix with all entries positive for each
t > 0. By the Perron-Frobenius theorem, r(ΦA(·) (τ )) is the principal eigenvalue of
ΦA(·) (τ ) in the sense that it is simple and admits a positive eigenvector. Then we
have the following result.

Lemma 3.8. ([46, Lemma 2.1]) Let µ = τ1 lnr(ΦA(·) (τ )). Then there exists a
positive, τ -periodic function v(t) such that eµt v(t) is a solution of (3.18).

Suppose P : X → X is the Poincaré map associated with system (2.1), that is,
P (x0 ) = u(τ, x0 ), ∀ x0 := (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) ∈ X,
where u(t, x0 ) is the unique solution of system (2.1) with u(0, x0 ) = x0 . It is easy
to see that
P n (x0 ) = u(nτ, x0 ), ∀ n ≥ 0.
For convenience, we define E0 = E0 = (0, 0, 0, 0, H ∗ , 0, 0, 0) and E1 = E1 (0) =
(A∗ (0), M ∗ (0), 0, 0, H ∗ , 0, 0, 0). Let
X0 := {(A, Ms , Me , Mi , Hs , He , Hi , Hr ) ∈ X : Mi > 0}.
and
∂X0 := X\X0 = {(A, Ms , Me , Mi , Hs , He , Hi , Hr ) ∈ X : Mi = 0}.

Lemma 3.9. Assume that (A(t), Ms (t), Me (t), Mi (t), Hs (t), He (t), Hi (t), Hr (t)) is
a solution of the system (2.1) with initial value (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) ∈
X0 . Then (A(t), Ms (t), Me (t), Mi (t), Hs (t), He (t), Hi (t), Hr (t))  0, ∀ t > 0.
 PERSISTENCE OF MOSQUITO VECTOR AND DENGUE 15

Proof. Given an initial value (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) ∈ X0 . In view of
the fifth equation of system (2.1), it follows that
 Z t R 
s2
− 0t ς(s1 )ds1
R
ς(s1 )ds1 0
Hs (t) = e Λh e 0 ds2 + Hs , (3.19)
0

where
bβh (t)Mi (t)
ς(t) := µh + . (3.20)
NH (t)
Thus, Hs (t) > 0, ∀ t > 0. From the first equation of system (2.1), we see that
Z t R 
Rt s2
A(t) = e− 0 b(s1 )ds1 e 0 b(s1 )ds1 ζ(s2 )ds2 + A0 , (3.21)
0

where (
ζ(t) := kδ(t)NM ≥ 0,
kδ(t) (3.22)
b(t) := C Nm + θ(t) + µa (t).
On the contrary, we assume there exists t0 ≥ 0 such that A(t0 ) = 0. This implies that
A0 = 0 and ζ(t) := kδ(t)NM ≡ 0 on [0, t0 ]. This contradicts that Mi (0) = Mi0 > 0.
Thus, A(t) > 0, ∀ t > 0. From the second equation of system (2.1), we have
Z t R 
s2
− 0t c(s1 )ds1
R
c(s1 )ds1 0
Ms (t) = e e 0 θ(s2 )A(s2 )ds2 + Ms , (3.23)
0

where
bβm (t)Hi (t)
c(t) := + µm (t). (3.24)
NH (t)
This implies that Ms (t) > 0, ∀ t > 0.
By [32, Theorem 4.1.1] as generalized to nonautonomous systems, the irreducibil-
ity of the cooperative matrix
bβm (t)Ms (t)
 
−(γm (t) + µm (t)) 0 0 NH (t)
γm (t) −µm (t) 0 0
 
(3.25)
 
 bβh (t)Hs (t)
−(γ

 0 NH (t) h + µh ) 0 
0 0 γh −(αh + µh )

implies that (Me (t), Mi (t), He (t), Hi (t))T  0, ∀ t > 0.

Finally, it follows from the eighth equation in system (2.1) that
 Z t 
Hr (t) = e−µh t αh eµh s Hi (s)ds + Hr0 , (3.26)
0

which implies that Hr (t) > 0, ∀ t > 0. This completes the proof of the lemma.

Lemma 3.10. Let RM > 1 and R0 > 1. Then for j = 0, 1, there exists σj > 0
such that for any (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) ∈ X0 with
k(A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) − Ej k ≤ σj ,
we have
lim sup dist(P n (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ), Ej ) ≥ σj .
n→∞
16 NAVEEN K. VAIDYA AND FENG-BIN WANG

Proof. Since R0 > 1, Lemma 3.7 implies that r(ΦF(·)−V(·) (τ )) > 1. Thus, we may
choose ρ1 > 0 small enough such that r(ΦFρ1 (·)−V(·) (τ )) > 1, where
∗
0 bβm (t)(M (t)−ρ1 )
 
0 0 H ∗ +4ρ1
 0 0 0 0
 
Fρ1 (t) =  .

bβh (t)(H ∗ −ρ1 )
 0 H ∗ +4ρ1 0 0 
0 0 0 0
By the continuity of the solutions with respect to the initial values, there exists a
σ1 > 0 such that for all (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) ∈ X0 with
k(A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) − E1 k ≤ σ1 ,
there holds ku(t, (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 )) − u(t, E1 )k < ρ1 , ∀ t ∈ [0, τ ].
We first prove the case j = 1, i.e.,
lim sup dist(P n (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ), E1 ) ≥ σ1 .
n→∞
Assume, by contradiction, that the above conclusion does not hold. Then we have
lim sup dist(P n (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ), E1 ) < σ1 ,
n→∞

for some (A , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) ∈ X0 . Without loss of generality, we
0

assume that
dist(P n (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ), E1 ) < σ1 , ∀ n ≥ 0.
It follows that
ku(t, P n (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 )) − u(t, E1 )k < ρ1 , ∀ t ∈ [0, τ ], n ≥ 0.
For any t ≥ 0, let t = mτ + t0 , where t0 ∈ [0, τ ), and m is the largest integer less
than or equal to τt . Therefore, we have ku(t, (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 )) −
u(t, E1 )k = ku(t0 , P m (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 )) − u(t0 , E1 )k < ρ1 . Note
that (A(t), Ms (t), Me (t), Mi (t), Hs (t), He (t), Hi (t), Hr (t)) = u(t, (A0 , Ms0 , Me0 , Mi0 ,
Hs0 , He0 , Hi0 , Hr0 )) and u(t, E1 ) = E1 (t), ∀ t ≥ 0. It then follows that for all t ≥ 0,
we have
Ms (t) > M ∗ (t)−ρ1 , H ∗ +ρ1 > Hs (t) > H ∗ −ρ1 , ρ1 > He (t), ρ1 > Hi (t), ρ1 > Hr (t).
From the equations of Me , Mi , He and Hi in (2.1), it follows that
bβm (t)(M ∗ (t)−ρ1 )

dMe


 dt ≥ H ∗ +4ρ1 Hi − (γm (t) + µm (t))Me , ∀ t ≥ 0,
 dMi = γ (t)M − µ (t)M , ∀ t ≥ 0,

dt m e m i
∗
−ρ1 ) (3.27)
 dHe ≥ bβh (t)(H
∗ M i − (γ h + µh )He , ∀ t ≥ 0,

 dt H +4ρ1
 dHi = γ H − (α + µ )H , ∀ t ≥ 0.

dt h e h h i

Since (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) ∈ X0 , it follows from Lemma 3.9 that
(Me (t), Mi (t), He (t), Hi (t))  0, ∀ t > 0.
Thus, we may fix a t1 > 0 such that (Me (t1 ), Mi (t1 ), He (t1 ), Hi (t1 ))  0. By
Lemma 3.8, it follows that there exists a positive, τ -periodic function J(t) and
˜ := b̃eµ̃(t−t1 ) J(t) is a solution of
µ̃ = τ1 ln [r(ΦFρ1 (·)−V(·) (τ ))] such that J(t)
dx(t)
= (Fρ1 (t) − V(t)) x(t),
dt
 PERSISTENCE OF MOSQUITO VECTOR AND DENGUE 17

˜ 1 ) := b̃J(t1 ) ≤ (Me (t1 ), Mi (t1 ), He (t1 ), Hi (t1 )). The standard

where b̃ satisfies J(t
comparison theorem (see, e.g., [33, Theorem B.1]) implies that
˜
(Me (t), Mi (t), He (t), Hi (t)) ≥ J(t), ∀ t ≥ t1 .
In particular, there exists n1 such that
˜ ), ∀ n ≥ n1 .
(Me (nτ ), Mi (nτ ), He (nτ ), Hi (nτ )) ≥ J(nτ
˜ ) → ∞ as n → ∞. Thus, (Me (nτ ), Mi (nτ ), He (nτ ),
Since µ̃ > 0, it follows that J(nτ
Hi (nτ )) → ∞ as n → ∞. This contradiction completes the proof of the case j = 1.
Next, we will use the fact RM > 1 to show that the conclusion is also true for the
case j = 0. Since RM > 1, Lemma 3.3 implies that r(ΨFM (·)−VM (·) (τ )) > 1. Thus,
we may choose ρ0 > 0 small enough such that r(ΨFM M
ρ0 (·)−Vρ0 (·)
(τ )) > 1, where

0 k(1 − ρC0 )δ(t)

 
FM
ρ0 (t) = , (3.28)
0 0
and
!
θ(t) + µa (t) 0
VρM0 (t) = bβm (t)ρ0 . (3.29)
−θ(t) H ∗ −ρ0 + µm (t)

By the continuity of the solutions with respect to the initial values, there exists a
σ0 > 0 such that for all (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) ∈ X0 with
k(A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) − E0 k ≤ σ0 ,
there holds ku(t, (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 )) − u(t, E0 )k < ρ0 , ∀ t ∈ [0, τ ].
For the case j = 0, we need to prove that
lim sup dist(P n (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ), E0 ) ≥ σ0 ,
n→∞

where, dist(A, B) is the distance between A and B.

Assume, by contradiction, that the above conclusion does not hold. Then we
have
lim sup dist(P n (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ), E0 ) < σ0 ,
n→∞
for some (A , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 )
0
∈ X0 . Without loss of generality, we
assume that
dist(P n (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ), E0 ) < σ0 , ∀ n ≥ 0.
It follows that
ku(t, P n (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 )) − u(t, E0 )k < ρ0 , ∀ t ∈ [0, τ ], n ≥ 0.
For any t ≥ 0, let t = `τ + t00 , where t00 ∈ [0, τ ), and ` is the largest integer less
than or equal to τt . Therefore, we have ku(t, (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 )) −
u(t, E0 )k = ku(t00 , P ` (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 )) − u(t00 , E0 )k < ρ0 . Note
that (A(t), Ms (t), Me (t), Mi (t), Hs (t), He (t), Hi (t), Hr (t)) = u(t, (A0 , Ms0 , Me0 , Mi0 ,
Hs0 , He0 , Hi0 , Hr0 )) and u(t, E0 ) = E0 = E0 , ∀ t ≥ 0. It then follows that for all t ≥ 0,
we have
A(t) < ρ0 , Ms (t) < ρ0 , ∀ t ≥ 0.
By Lemma 3.1, it follows that there exists a t0 > 0 such that
NH (t) > H ∗ − ρ0 , ∀ t ≥ t0 .
18 NAVEEN K. VAIDYA AND FENG-BIN WANG

Then it follows from the equations of A and Ms in (2.1) that

(
dA ρ0


dt ≥ kδ(t) 1 − C Ms − (θ(t) + µa (t))A, t ≥ t0 ,
(3.30)
dMs
dt ≥ θ(t)A − [ bβm (t)Ms ρ0
H ∗ −ρ0 + µm (t)]Ms , t ≥ t0 .
Since (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) ∈ X0 , it follows from Lemma 3.9 that
(A(t0 ), Ms (t0 ))  0.
By Lemma 3.8, it follows that there exists a positive, τ -periodic function Q(t) and
˜ κ(t−t0 ) Q(t) is a solution of
κ = τ1 ln [r(ΨFM (·)−VM (·) (τ ))] such that Q̃(t) := de
dy(t)
= FM (t) − VM (t) y(t),


dt
˜ ˜ 0 ) ≤ (A(t0 ), Ms (t0 )). The standard comparison
where d satisfies Q̃(t0 ) := dQ(t
theorem (see, e.g., [33, Theorem B.1]) implies that
(A(t), Ms (t)) ≥ Q̃(t), ∀ t ≥ t0 .
In particular, there exists n0 such that
(A(nτ ), Ms (nτ )) ≥ Q̃(nτ ), ∀ n ≥ n0 .
Since κ > 0, it follows that Q̃(nτ ) → ∞ as n → ∞. Thus, (A(nτ ), Ms (nτ )) → ∞
as n → ∞. This contradiction completes the proof of the case j = 0.
Now we prove that R0 is a threshold index for disease persistence if RM > 1 as
stated in the following theorem.
Theorem 3.2. Assume that RM > 1. Then the following statements hold.
(i) If R0 < 1, then the disease-free periodic state E1 (t) is globally attractive for
system (2.1) in the sense that if (A0 , Ms0 ) 6= (0, 0), we have
lim [(A(t), Ms (t), Me (t), Mi (t), Hs (t), He (t), Hi (t), Hr (t)) − E1 (t)] = (0, 0, 0, 0, 0, 0, 0, 0);
t→∞

(ii) If R0 > 1, there exists an η > 0 such that for any solution
(A(t), Ms (t), Me (t), Mi (t), Hs (t), He (t), Hi (t), Hr (t))
with initial value (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) ∈ X0 satisfies
lim inf Mi (t) ≥ η.
t→∞

Further, system (2.1) admits at least one positive τ -periodic solution

(Ã(t), M̃s (t), M̃e (t), M̃i (t), H̃s (t), H̃e (t), H̃i (t), H̃r (t)).
Proof. Part (i). We first consider the case where R0 < 1. From Lemma 3.7, it
follows that r(ΦF(·)−V(·) (τ )) < 1. Now we choose ξ0 > 0 sufficiently small such that
r(ΦFξ0 (·)−V(·) (τ )) < 1, where
∗
0 bβm (t)(M (t)+ξ0 )
 
0 0 H ∗ −ξ0
 0 0 0 0
 
Fξ0 (t) =  .

bβh (t)(H ∗ +ξ0 )
 0 H ∗ −ξ0 0 0 
0 0 0 0
Assume that
(A(t), Ms (t), Me (t), Mi (t), Hs (t), He (t), Hi (t), Hr (t))
 PERSISTENCE OF MOSQUITO VECTOR AND DENGUE 19

is a nonnegative solution of system (2.1) in X. Note that NH (t) and (A(t), NM (t))
satisfy (3.3) (Lemma 3.1) and (3.12) (Lemma 3.4), respectively. Then there is a
T > 0 such that for any t ≥ T , we have
Ms (t) ≤ NM (t) ≤ M ∗ (t) + ξ0 , Hs (t) ≤ NH (t) ≤ H ∗ + ξ0 , NH (t) ≥ H ∗ − ξ0 .
By Lemma 3.8, it follows that there exists a positive, τ -periodic function v(t)
and µ = τ1 ln [r(ΦFξ0 (·)−V(·) (τ ))] such that v̄(t) := āeµt v(t) is a solution of
dx(t)
= (Fξ0 (t) − V(t)) x(t),
dt
where ā satisfies v̄(T ) := āv(T ) ≥ (Me (T ), Mi (T ), He (T ), Hi (T )).
From the equations of Me , Mi , He and Hi in (2.1), it follows that
bβm (t)(M ∗ (t)+ξ0 )

dMe
 dt ≤

 H ∗ −ξ0 Hi − (γm (t) + µm (t))Me , ∀ t ≥ T,
 dMi = γ (t)M − µ (t)M , ∀ t ≥ T,

dt m e m i
dHe bβh (t)(H ∗ +ξ0 ) (3.31)
dt ≤ H ∗ −ξ0 Mi − (γh + µh )He , ∀ t ≥ T,



 dHi = γ H − (α + µ )H , ∀ t ≥ T.

dt h e h h i

The standard comparison theorem (see, e.g., [33, Theorem B.1]) implies that
(Me (t), Mi (t), He (t), Hi (t)) ≤ v̄(t), ∀ t ≥ T.
Since µ < 0, it follows that v̄(t) → 0 as t → ∞. Thus, (Me (t), Mi (t), He (t), Hi (t))
→ 0 as t → ∞. This implies that (A, Ms ) is asymptotic to (3.6). By the theory of
asymptotically periodic semiflows (see, e.g., [48] or [47, section 3.2]) and Lemma 3.4,
it follows that limt→∞ [(A(t), Ms (t)) − (A∗ (t), M ∗ (t))] = (0, 0). This completes the
proof of Part (i).
Part (ii). We next consider the case where R0 > 1. From Lemma 3.5, it follows
that the discrete-time system {P n }n≥0 admits a global attractor in X. Now we prove
that {P n }n≥0 is uniformly persistent with respect to (X0 , ∂X0 ). By Lemma 3.9, it
follows that X and X0 are positively invariant. Clearly, ∂X0 is relatively closed in
X. Let
M∂ = {(A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) ∈ ∂X0 ,
: P n (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) ∈ ∂X0 , ∀ n ≥ 0}.
We are going to prove that
M∂ := {(A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) ∈ X : Mi0 = 0}, (3.32)
0
for which, it suffices to prove that for any (A , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 )
∈ M∂ ,
we have Mi (nτ ) = 0, ∀ n ≥ 0. If it is not true, there exists an n1 ≥ 0 such that
Mi (n1 τ ) > 0. (3.33)
It is easy to see from A equation of (2.1) that
Z t R 
− nt τ b(s1 )ds1 s2
R
b(s1 )ds1
A(t) = e 1 e 1
n τ
ζ(s2 )ds2 + A(n1 τ ) , (3.34)
n1 τ

where ζ(t) and b(t) are defined in (3.22). Thus, A(t) > 0, ∀ t > n1 τ . Similarly,
from Ms equation of (2.1) we have
Z t R 
− nt τ c(s1 )ds1 s2
R
c(s1 )ds1
Ms (t) = e 1 e 1
n τ
θ(s2 )A(s2 )ds2 + Ms (n1 τ ) , (3.35)
n1 τ
20 NAVEEN K. VAIDYA AND FENG-BIN WANG

where c(t) is defined in (3.24). This implies that Ms (t) > 0, ∀ t > n1 τ . By [32,
Theorem 4.1.1] as generalized to nonautonomous systems, the irreducibility of the
cooperative matrix (3.25) implies that

(Me (t), Mi (t), He (t), Hi (t)  0, ∀ t > n1 τ,

where the initial value (Me (n1 τ ), Mi (n1 τ ), He (n1 τ ), Hi (n1 τ )) > 0. In particu-
lar, we have (Me (nτ ), Mi (nτ ), He (nτ ), Hi (nτ ))  0, ∀ n > n1 , which contra-
dicts the fact that (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) ∈ M∂ . This implies that
(3.32) holds. It is clear that there are two fixed points of P in M∂ , which are
E0 = E0 = (0, 0, 0, 0, H ∗ , 0, 0, 0) and E1 = E1 (0) = (A∗ (0), M ∗ (0), 0, 0, H ∗ , 0, 0, 0).
If (A(t), Ms (t), Me (t), Mi (t), Hs (t), He (t), Hi (t), Hr (t)) is a solution of system (2.1)
initiating from M∂ , it follows from system (2.1) and the fact Mi (t) ≡ 0 that
(A(t), Ms (t), Me (t), Mi (t), Hs (t), He (t), Hi (t), Hr (t)) approaches E0 or E1 (t) as t
approaches ∞.
It follows from Lemma 3.10 that {E0 } ∪ {E1 } is an isolated invariant set in X
and W s (Ej ) ∩ X0 = ∅, j = 0, 1, where W s (Ej ) is the stable set of Ej . Note that
every orbit in M∂ approaches to {E0 } ∪ {E1 }, and {E0 } ∪ {E1 } is acyclic in M∂ . By
[47, Theorem 1.3.1], it follows that {P n }n≥0 is uniformly persistent with respect
to (X0 , ∂X0 ). By [47, Theorem 3.1.1], the solutions of system (2.1) are uniformly
persistent with respect to (X0 , ∂X0 ), that is, there exists an η > 0 such that for any
solution
(A(t), Ms (t), Me (t), Mi (t), Hs (t), He (t), Hi (t), Hr (t))

with initial value (A0 , Ms0 , Me0 , Mi0 , Hs0 , He0 , Hi0 , Hr0 ) ∈ X0 satisfies

lim inf Mi (t) ≥ η.

t→∞

Furthermore, [47, Theorem 1.3.6] implies that P has a fixed point

(Ã(0), M̃s (0), M̃e (0), M̃i (0), H̃s (0), H̃e (0), H̃i (0), H̃r (0)) ∈ X0 ,

and hence, M̃i (0) > 0. By the same arguments as those in Lemma 3.9, one can
show that

(Ã(t), M̃s (t), M̃e (t), M̃i (t), H̃s (t), H̃e (t), H̃i (t), H̃r (t))  0.

This completes the proof of Part (ii).

4. Numerical computation. In this section we present some numerical results

demonstrating how mean temperature, seasonal temperature variations, and diurnal
temperature variations can impact the mosquito reproduction number (RM ) and
the infection invasion threshold (R0 ). We used Lemma 3.11 to compute RM and
Lemma 3.17 to compute R0 numerically as done in previous studies ([37, 40, 19]).
Our results show that the value of RM is greater than R0 for every temperature
profile. This indicates that RM < 1 (extinction of mosquito population) implies
R0 < 1 (eradication of dengue), which is practically true as mosquito bites are the
only route of dengue transmission.
 PERSISTENCE OF MOSQUITO VECTOR AND DENGUE 21

4.1. Effects of mean temperature, T0 . We performed computations to study

how varying annual mean temperature, T0 , from 10 o C to 45 o C affects RM and R0
(Fig. 4.1). On increasing T0 , both RM and R0 increase, reach the corresponding
maximum values and then decrease, with RM < 1 and R0 < 1 for extremely low
mean temperatures as well as extremely high mean temperatures. Therefore, there
exist optimal ranges of the mean temperature for both persistence of mosquito
population and persistence of dengue fever. However, the interval of the mean
temperature for mosquito persistence is larger than that for the dengue persistence.
For example, in our computation (Fig. 4.1), mosquito persists (i.e., RM > 1) for
13.5o C ≤ T0 ≤ 44.5o C, while dengue persists (i.e., R0 > 1) for 16.5o C ≤ T0 ≤
36.5o C. Note that there are small ranges of the mean temperature during which
mosquito persists, but dengue does not persist.

Time-varying temperature Time-varying temperature

Constant temperature Constant temperature
90
RM = 1 7 R0 = 1
Mosquito reproduction number

80

Infection invasion threshold

6
70
5
60

50 4

40
3
30
2
20
1
10

15 20 25 30 35 40 45 15 20 25 30 35 40
Mean temperature Mean temperature

Figure 4.1. Mosquito reproduction number (RM ) [Left] and in-

fection invasion threshold (R0 ) [right] for different values of the
mean temperature (T0 ) with amplitudes of seasonal temperature
and diurnal temperature fixed at m = 5 o C and d = 5 o C, re-
spectively. For comparison purposes, RM and R0 for the constant
temperature (i.e., m = d = 0 o C) are also plotted.

For comparison purposes, we also computed both the mosquito reproduction

number and the infection invasion threshold for the constant temperature (Figure
4.1). For the parameter range considered in this study, the mosquito population
hardly persists for temperatures maintained constant at values greater than 40 o C.
However, for the time-varying temperature, the mosquito population may persist
also for the mean temperatures slightly higher than 40 o C. It should be noted that
for the time-varying periodic temperature, even though the mean temperature is 40
o
C, the temperature in our base case computation can reach as low as 30 o C due
to the seasonal and diurnal variations. Therefore, the time-varying temperature
provides the values of RM and R0 different from their values with constant tem-
perature. Our estimates show that the constant temperature underestimates RM
and R0 for low and high mean temperatures while it overestimates for the middle
temperature range.

4.2. Effects of seasonal temperature variation, m . In this subsection, we

use the parameter m , the amplitude of the seasonal temperature, to evaluate the
22 NAVEEN K. VAIDYA AND FENG-BIN WANG

effects of seasonal temperature variation on the mosquito reproduction number and

the infection invasion threshold. We compute RM (Fig. 4.2, left column) and R0
(Fig. 4.2, right column) for various values of m with all other parameters fixed.
As we discussed above, since the values of RM and R0 may highly depend on
the mean temperature with maximum values occurring at some temperature, we
consider three cases: a lower mean temperature (T0 = 16 o C) (Fig. 4.2, top row),
an optimal mean temperature (T0 = 28 o C) (Fig. 4.2, middle row), and a higher
mean temperature (T0 = 38 o C) (Fig. 4.2, bottom row).
For the amplitude of seasonal temperature considered, the mosquito reproduction
number mostly remains larger than 1 (Fig. 4.2, right column). This shows that the
mosquito population persists for this range of temperature as observed in reality.
We observed that in each of the three mean temperatures considered, there is a
decreasing trend of RM for an increasing amplitude of the seasonal temperature
(Fig. 4.2, right column). This shows that in places, where the temperature remains
relatively stable over the seasons, the growth of mosquito population is favored,
compared to the places with large seasonal variation of temperature. This effect
is mainly pronounced in the case of the optimal mean temperature (T0 = 28 o C)
(Fig. 4.2, middle row). At T0 = 28 o C, RM decreases from 65 to 13 when m
increases from 5 o C to 15 o C, while at T0 = 16 o C and T0 = 38 o C, the change of
RM is from 4.9 to 2 and from 13.6 to 6, respectively.
Computed values of R0 for varying m (Fig. 4.2, left column) indicate that R0
can be less than 1 for some values of m . Therefore, the seasonal variation of the
temperature can be a determinant factor for the dengue epidemic to die out (R0 < 1)
or to persist (R0 > 1). In general, a larger variation of seasonal temperature
provides a smaller value of the infection invasion threshold. For example, R0 < 1
for m > 11 o C, 16 o C, and 18.5 o C at the mean temperature T0 = 16 o C, T0 = 28
o
C, and T0 = 38 o C, respectively. However, note that for T0 = 16 o C and T0 = 38
o
C, smaller amplitudes also make R0 < 1. These results indicate that the places
with a larger variation in seasonal temperature has a less likelihood for a dengue
epidemic to occur.

4.3. Effects of diurnal temperature variation, d . We now vary d , the ampli-

tude of diurnal temperature, from 5o C to 25o C, and compute the mosquito repro-
duction number (RM ) (Fig. 4.3, left column) and the infection invasion threshold
R0 (Fig. 4.2, right column) for those values of m . Again, we consider three different
seasonal mean temperature, T0 = 16 o C (Fig. 4.3, top row), T0 = 28 o C (Fig. 4.3,
middle row), and T0 = 38 o C (Fig. 4.3, bottom row). Our results show that the
patterns of RM and R0 (Fig. 4.3) for an increasing d are similar to those observed
for an increasing m .
An increase in diurnal temperature variation decreases the mosquito reproduction
number with a bigger effect seen in the mean temperature corresponding to the peak
RM (i.e. T0 = 28 o C). In our calculation, the increase in d by 10 o C (from 5 o C
to 15 o C) caused a decrease of the mosquito reproduction number by 3 (from 5 to
2), 55 (from 65 to 10), and 11 (from 13 to 2) when the mean temperature was set
to T0 = 16 o C, 28 o C, and 38 o C, respectively. Thus, places with higher diurnal
temperature variation are less favorable for the growth of mosquito population. It
is worth noting that at T0 = 38 o C, a higher amplitude of diurnal temperature,
 PERSISTENCE OF MOSQUITO VECTOR AND DENGUE 23

70 6

Mosquito reproduction number

Infection invasion threshold

60
5
50
4
40
3
30
2
20

10 1

3 5 7 9 11 13 5 7 9 11 13
Amplitude of seasonal temperature Amplitude of seasonal temperature

70 6
Mosquito reproduction number

Infection invasion threshold

60
5
50
4
40
3
30
2
20

10 1

6 9 12 15 18 21 24 6 9 12 15 18 21 24
Amplitude of seasonal temperature Amplitude of seasonal temperature

70 6
Mosquito reproduction number

Infection invasion threshold

60
5
50
4
40
3
30
2
20

10 1

6 9 12 15 18 21 24 6 9 12 15 18 21 24
Amplitude of seasonal temperature Amplitude of seasonal temperature

Figure 4.2. Mosquito reproduction number (RM ) [left column]

and infection invasion threshold (R0 ) [right column] for different
values of the amplitudes of seasonal temperature (m ) with the
amplitude of diurnal temperature fixed at d = 5 o C and the mean
temperature fixed at T0 = 16 o C [top row], T0 = 28 o C [middle
row], and T0 = 38 o C [bottom row].

for example, d > 20 o C, can bring RM to a value less than 1, resulting in the
extinction of mosquito population.
Similarly, our results show that higher diurnal temperature can bring the in-
fection invasion threshold, R0 , to a value less than 1, avoiding dengue epidemics.
According to our computations, the dengue can not persist for amplitude of di-
urnal temperature, d , greater than 13.5 o C, 19.5 o C, and 13.5 o C, for the mean
temperature T0 = 16 o C, 28 o C, and 38 o C, respectively. Therefore, places with
24 NAVEEN K. VAIDYA AND FENG-BIN WANG

environmental temperature with higher diurnal fluctuation are less vulnerable to

dengue epidemics.

70 6
Mosquito reproduction number

Infection invasion threshold

60
5
50
4
40
3
30
2
20

10 1

3 5 7 9 11 13 5 7 9 11 13
Amplitude of diurnal temperature Amplitude of diurnal temperature

70 6
Mosquito reproduction number

Infection invasion threshold

60
5
50
4
40
3
30
2
20

10 1

6 9 12 15 18 21 24 6 9 12 15 18 21 24
Amplitude of diurnal temperature Amplitude of diurnal temperature

70 6
Mosquito reproduction number

Infection invasion threshold

60
5
50
4
40
3
30
2
20

10 1

6 9 12 15 18 21 24 6 9 12 15 18 21 24
Amplitude of diurnal temperature Amplitude of diurnal temperature

Figure 4.3. Mosquito reproduction number (RM ) [left column]

and infection invasion threshold (R0 ) [right column] for different
values of the amplitudes of diurnal temperature (d ) with the am-
plitude of seasonal temperature fixed at m = 5 o C and the mean
temperature fixed at T0 = 16 o C [top row], T0 = 28 o C [middle
row], and T0 = 38 o C [bottom row].

5. Discussion and conclusion. It is known that the environmental temperature

can have a substantial impact on the life cycle of mosquito and the mosquito-human-
mosquito transmission cycle of dengue virus [4, 35, 24, 9, 17, 45, 43]. Because of
complexity in the variation of environmental temperature, especially existence of
 PERSISTENCE OF MOSQUITO VECTOR AND DENGUE 25

diurnal temperature fluctuations on top of the seasonal variations, there remains

much uncertainty on the understanding of the role of temperature on mosquito
population and dengue transmission [17, 28]. The major objective of this study was
to develop a mathematical model to evaluate the impact of seasonal and diurnal
temperature variation on the persistence of mosquito vectors and dengue.
Using techniques from the dynamical system theory, we used our nonautonomous
model to establish two thresholds, the mosquito reproduction number (RM ) and
the infection invasion threshold (R0 ), which fully determine whether the mosquito
population and dengue, respectively, persist or die out in a community. Specifically,
RM > 1 implies the persistence of mosquito population and R0 > 1 implies the
persistence of dengue. Importantly, the formulated expressions show that both of
RM and R0 depend on seasonal and diurnal temperature variations, highlighting
the importance of considering these factors in the study of mosquito abundance and
dengue spread.
Analysis of threshold dynamics and numerical computations of the threshold
values provide some interesting results related to mosquito vectors and dengue epi-
demics. First, RM is always greater than R0 , indicating that a certain temperature
can result in the persistence of mosquitos while avoiding dengue epidemics, as ob-
served in real life situations. This also indicates eradicating the mosquito population
can be much difficult compared to eradicating the dengue. Therefore, relying on
only mosquito population control may not be enough for the successful control of
dengue. Second, we identified that there exists a certain range of annual mean
temperature for the persistence of mosquito population, i.e., the mosquito popu-
lation can not persist when the average environmental temperature is too low or
too high. Similarly, the dengue persistence is favored for a certain range of annual
mean temperature. The temperature range (interval) for the persistence of dengue
is subset of the temperature range (interval) for the persistence of mosquito pop-
ulation. These temperature intervals provide the ranges of temperature for which
mosquitos persist but the dengue epidemic is avoided.
Next, our results show that both seasonal and diurnal temperature variations
are critical for the persistence of mosquito population (RM > 1) as well as for the
persistence of dengue (R0 > 1). This explains why only the mean temperature was
unable to describe the observed epidemics in some places, such as Thailand, where
the epidemics are correlated mainly with diurnal temperature variations [17, 8, 20,
29, 30]. In general, our numerical estimates show that a sufficiently large seasonal
or diurnal temperature variation can result in RM < 1 and R0 < 1, implying that
places with larger temperature variations suffer less from the mosquito population
burden and dengue epidemics.
We acknowledge the several limitations of our study. Our computations are based
on the parameters estimated using limited data sets. Therefore, we note that our
computations should be considered primarily for the purpose of qualitative results,
and may need to be improved for quantitative applications in real life. More data
sets, particularly those related to diurnal temperature variations, can help improve
the calculation of RM and R0 . We primarily focused on the threshold dynamics
predicted by the model. However, it might be important to study detailed temporal
dynamics of the dengue, especially for the case when R0 > 1. For this, the future
work with numerical comparison of the model prediction with real data on dengue
26 NAVEEN K. VAIDYA AND FENG-BIN WANG

cases will strengthen the validity of our model predictions. We are also unable to
explicitly formulate the endemic equilibrium, which requires more theoretical and
higher computational exercises. Furthermore, models that combine spatial variation
along with the temporal variation of the environmental temperature may be needed
to more accurately describe the rapid worldwide spread of dengue.
In summary, the model developed here is capable of capturing the effects of di-
urnal and seasonal temperature variations on the dynamics of mosquito population
and dengue transmission. As revealed in the results from our model, in addition to
the seasonal temperature variation, the diurnal temperature variation also plays a
significant role in the persistence of mosquito vectors and the persistence of dengue.
Thus, both seasonal as well as diurnal temperature variations should be considered
in the study of mosquito population as well as dengue transmission, control, and
prevention.

Acknowledgments. This work was funded by NSF grants DMS-1951793 (NKV),

DMS-1616299 (NKV), DMS-1836647 (NKV), and DEB-2030479 (NKV) from Na-
tional ScienceFoundation of USA, and UGP award (NKV) and the start-up fund
(NKV) from San Diego State University. Research of FBW is supported in part
by Ministry of Science and Technology, Taiwan; and National Center for Theo-
retical Sciences, National Taiwan University; and Chang Gung Memorial Hospital
(CRRPD3H0011, BMRPD18, NMRPD5J0201 and CLRPG2H0041). NKV would
like to thank National Center for Theoretical Sciences, National Taiwan University
for providing support for NCTS visit, during which some of the work was carried
out. Finally, the authors would like to thank Mr. Pavan Chalumuri for his help in
computations and Dr. Eyal Oren for useful discussion.

REFERENCES

[1] G. Aronsson and R. B. Kellogg, On a differential equation arising from compartmental anal-
ysis, Math. Biosci., 38 (1978), 113–122.
[2] N. Bacaër and S. Guernaoui, The epidemic threshold of vector-borne diseases with seasonality:
the case of cutaneous leishmaniasis in Chichaoua, Morocco, J. Math. Biol., 53 (2006), 421–
436.
[3] S. Banu, W. Hu, C. Hurst and S. Tong, Dengue transmission in the asia-pacific region: Im-
pact of climate change and socio-environmental factors, Tropical Medicine and International
Health, 11 (2011), 598–607.
[4] S. Bhatt, P. W. Gething, O. J. Brady, J. P. Messina, A. W. Farlow, et al., The global
distribution and burden of dengue, Nature, 496 (2013), 504–507.
[5] O. J. Brady, P. W. Gething, S. Bhatt, J. P. Messina, J. S. Brownstein, et al., Refining the
global spatial limits of dengue virus transmission by evidence-based consensus, PLoS Negl.
Trop. Dis., 6 (2012), e1760.
[6] G. Chowell, P. Diaz-Dueñas, J. C. Miller, A. Alcazar-Velazco, J. M. Hyman, P. W. Fenimore
and C. Castillo-Chavez, Estimation of the reproduction number of dengue fever from spatial
epidemic data, Math. Biosci., 208 (2007), 571–589.
[7] N. C. Dom, Z. A. Latif, A. H. Ahmad, R. Ismail and B. Pradhan, Manifestation of gis tools
for spatial pattern distribution analysis of dengue fever epidemic in the city of Subang Jaya,
Malaysia, Environment Asia, 5 (2012), 82–92.
[8] T. P. Endy, A. Nisalak, S. Chunsuttiwat, D. H. Libraty, S. Green, et al., Spatial and temporal
circulation of dengue virus serotypes: A prospective study of primary school children in
Kamphaeng Phet, Thailand, Am. J. Epidemiol., 156 (2002), 52–59.
 PERSISTENCE OF MOSQUITO VECTOR AND DENGUE 27

[9] D. A. Focks, E. Daniels, D. G. Haile and J. E. Keesling, A simulation model of the epidemi-
ology of urban dengue fever: Literature analysis, model development, preliminary validation,
and samples of simulation results, Am. J. Trop. Med. Hyg., 53 (1995), 489–506.
[10] A. K. Githeko, S. W. Lindsay, U. E. Confalonieri and J. A. Patz, Climate change and vector-
borne diseases: A regional analysis, Bulletin of the World Health Organization, 78 (2000),
1136–1147.
[11] D. J. Gubler, Epidemic dengue/dengue hemorrhagic fever as a public health, social and
economic problem in the 21st century, Trends in Microbiology, 10 (2002), 100–103.
[12] J. K. Hale, Asymptotic Behavior of Dissipative Systems, American Mathematical Society
Providence, RI, 1988.
[13] M. W. Hirsch, Systems of differential equations that are competitive or cooperative II: Con-
vergence almost everywhere, SIAM J. Math. Anal., 16 (1985), 423–439.
[14] M. J. Hopp and J. A. Foley, Global-scale relationships between climate and the dengue fever
vector, aedes aegypti, Climatic Change, 48 (2001), 441–463.
[15] S. Karl, N. Halder, J. K. Kelso, S. A. Ritchie and G. J. Milne, A spatial simulation model for
dengue virus infection in urban areas, BMC Infec. Dis., 14 (2014), p447.
[16] A. Khan, M. Hassan and M. Imran, Estimating the basic reproduction number for single-
strain dengue fever epidemics, Infectious Diseases of Poverty, 3 (2014), p12.
[17] L. Lambrechts, K. P. Paaijmans, T. Fansiri, L. B. Carrington, L. D. Kramer, M. B. Thomas
and T. W. Scott, Impact of daily temperature fluctuations on dengue virus transmission by
aedes aegypti, PNAS , 108 (2011), 7460–7465.
[18] M.-T. Li, G.-Q. Sun, L. Yakob, H.-P. Zhu, Z. Jin and W.-Y. Zhang, The driving force for
2014 dengue outbreak in Guangdong, China, PLoS ONE , 11 (2016), e0166211.
[19] L. Liu, X.-Q. Zhao and Y. Zhou, A tuberculousis model with seasonality, Bull. Math. Biol.,
72 (2010), 931–952.
[20] A. Nisalak, T. P. Endy, S. Nimmannitya, S. Kalayanarooj, U. Thisayakorn, et al., Serotype-
specific dengue virus circulation and dengue disease in Bangkok, Thailand from 1973 to 1999,
Am. J. Trop. Med. Hyg., 68 (2003), 191–202.
[21] M. Oki and T. Yamamoto, Climate change, population immunity, and hyperendemicity in
the transmission threshold of dengue, PLoS ONE , 7 (2010), e48258.
[22] K. P. Paaijmans, A. F. Read and M. B. Thomas, Understanding the link between malaria
risk and climate, PNAS , 106 (2009), 13844–13849.
[23] A. Pakhare, Y. Sabde, A. Joshi, R. Jain, A. Kokane and R. Joshi, A study of spatial and
meteorological determinants of dengue outbreak in bhopal city in 2014, PLoS Negl. Trop.
Dis., 53 (2014), 225–233.
[24] W. G. Panhuisa, M. Choisyb, X. Xionga, N. S. Choka, P. Akarasewid, et al., Region-wide
synchrony and traveling waves of dengue across eight countries in southeast asia, Proc. Nat.
Acad. Sci., 112 (2015), 13069–13074.
[25] J. A. Patz, D. Campbell-Lendrum, T. Holloway and J. A. Foley, Impact of regional climate
change on human health, Nature, 438 (2005), 310–317.
[26] S. T. R. Pinho, C. P. Ferreira, L. Esteva, F. R. Barreto, V. C. Morato e Silva and M. G. L.
Teixeira, Modelling the dynamics of dengue real epidemics, Philos. Trans. R. Soc. Lond. Ser.
A Math. Phys. Eng. Sci., 368 (2010), 5679–5693.
[27] V. Racloz, R. Ramsey, S. Tong and W. Hu, Surveillance of dengue fever virus: A review of
epidemiological models and early warning systems, PLoS Negl. Trop. Dis., 6 (2012), e1648.
[28] D. J. Rogers and S. E. Randolph, Climate change and vector-borne diseases, Adv. Parasitol.,
62 (2006), 345–381.
[29] T. W. Scott, A. C. Morrison, L. H. Lorenz, G. G. Clark, D. Strickman, et al., Longitudi-
nal studies of aedes aegypti (diptera: Culicidae) in Thailand and puerto rico: Population
dynamics, J. Med. Entomol., 37 (2000), 77–88.
[30] P. M. Sheppard, W. W. Macdonald, R. J. Tonnand and B. Grab, The dynamics of an adult
population of aedes aegypti in relation to dengue haemorrhagic fever in bangkok, J. Anim.
Ecol., 38 (1969), 661–702.
[31] C. P. Simmons, J. J. Farrar, N. van Vinh Chau and B. Wills, Dengue, J. Vector Borne Dis,
6 (2012), e1648.
28 NAVEEN K. VAIDYA AND FENG-BIN WANG

[32] H. L. Smith, Monotone Dynamical Systems: An Introduction to the Theory of Competitive

and Cooperative Systems, American Mathematical Society Providence, RI, 1995.
[33] H. L. Smith and P. Waltman, The Theory of the Chemostat, CCambridge University Press,
Cambridge, 1995.
[34] R. W. Sutherst, Global change and human vulnerability to vector-borne diseases, N. Engl. J.
Med., 366 (2012), 1423–1432.
[35] M. Teurlai, C. E. Menkés, V. Cavarero, N. Degallier, E. Descloux, J.-P. Grangeon, et al.,
Socio-economic and climate factors associated with dengue fever spatial heterogeneity: A
worked example in new caledonia, PLoS Negl. Trop. Dis., 9 (2015), e0004211.
[36] N. K. Vaidya, X. Li and F.-B. Wang, Impact of spatially heterogeneous temperature on the
dynamics of dengue epidemics, Discrete Contin. Dyn. Syst. Ser. B , 24 (2019), 321–349.
[37] N. K. Vaidya and L. M. Wahl, Avian influenza dynamics under periodic environmental con-
ditions, SIAM J. Appl. Math., 75 (2015), 443–467.
[38] P. van den Driessche and J. Watmough, Reproduction numbers and sub-threshold endemic
equilibria for compartmental models of disease transmission, Math. Biosci., 180 (2002), 29–
48.
[39] F.-B. Wang, S.-B. Hsu and W. Wang, Dynamics of harmful algae with seasonal temperature
variations in the cove-main lake, Discrete Contin. Dyn. Syst. Ser. B , 21 (2016), 313–335.
[40] W. Wang and X.-Q. Zhao, Threshold dynamics for compartmental epidemic models in periodic
environments, J. Dynam. Differential Equations, 20 (2008), 699–717.
[41] WHO, Dengue Guidelines for Diagnosis, Treatment, Prevention, and Control, TDR: World
Health Organization, (2009).
[42] R. E. Woodruff and T. McMichael, Climate change and human health: All affected bit
somevvmore than others, Social Alternatives, 23 (2004), 17–22.
[43] H. M. Yang, M. de L. da G. Macoris, K. C. Galvani and M. T. M. Andrighetti, Follow up esti-
mation of aedes aegypti entomological parameters and mathematical modellings, Biosystems,
103 (2011), 360–371.
[44] H. M. Yang, M. L. G. Macoris, K. C. Galvani, M. T. M. Andrighetti and D. M. V. Wanderley,
Assessing the effects of temperature on dengue transmission, Epidemiol. Infect., 137 (2009),
1179–1187.
[45] H. M. Yang, M. L. G. Macoris, K. C. Galvani, M. T. M. Andrighetti and D. M. V. Wanderley,
Assessing the effects of temperature on the population of aedes aegypti, the vector of dengue,
Epidemiol. Infect., 137 (2009), 1188–1202.
[46] F. Zhang and X.-Q. Zhao, A periodic epidemic model in a patchy environment, J. Math.
Anal. Appl., 325 (2007), 496–516.
[47] X.-Q. Zhao, Dynamical Systems in Population Biology, Springer, New York, 2003.
[48] X.-Q. Zhao, Asymptotic behavior for asymptotically periodic semiflows with applications,
Comm. Appl. Nonlinear Anal., 3 (1996), 43–66.

Received August 2020; revised January 2021.

E-mail address: [email protected]
E-mail address: [email protected], [email protected]