ARTICLE IN PRESS

Nonlinear Analysis ( ) –

Contents lists available at ScienceDirect

Nonlinear Analysis
journal homepage: www.elsevier.com/locate/na

Models of plant quality and larch budmoth interaction

Sophia R.-J. Jang a , Jui-Ling Yu b,∗
a
Department of Mathematics and Statistics, Texas Tech University, Lubbock, TX 79409-1042, United States
b
Department of Applied Mathematics, Providence University, Taichung, Taiwan

article info a b s t r a c t
Keywords: We present and investigate two discrete-time models of leaf quality and larch budmoth
Larch budmoth population
interaction. In particular, existence and stability of equilibria are studied. For one model,
Hopf bifurcation
the moth population may grow unboundedly large and the interior steady state is always
Period-doubling bifurcation
locally asymptotically stable when it exists. A Hopf bifurcation may occur for the other
model when the interior steady state becomes unstable.
© 2009 Elsevier Ltd. All rights reserved.

1. Introduction

The population of larch budmoth (LBM), Zeiraphera diniana, in the Swiss Alps is well known for its periodic outbreaks
and regular oscillations. Plant quality is considered as a necessary effect to induce cycles [1–4]. It usually takes several years
for the leaf quality to recover after a serious outbreak. Field and laboratory also showed that poor plant quality has a strong
effect on budmoth survival and reproduction. Hence one strong explanation of LBM cycles is based on their interaction with
foliage quality.
The goal of this study is to present and investigate two models of interaction between larch moth and plant quality.
Since LBM has an annual life cycle with non-overlapping generations, the models are systems of difference equations. We
investigate the existence of simple stationary solutions of the systems and their stability. Local bifurcation analysis for each
of the models will also be performed. We will compare and contrast these analytical results. See [5] for analysis of other
larch budmoth population models. In the following section we study a leaf quality-larch budmoth model with exponential
growth of the moth. Section 3 investigates a plant quality-moth model with Ricker equation. The final section provides some
biological conclusions.

2. A simple plant quality-moth model

Let Nt denote the larch budmoth density at time t, t = 0, 1, . . ., where the moth population density is measured in
terms of number of third instar larvae per kg of larch branches. The leaf quality at time t denoted by Qt is an index of plant
−15
quality measured in terms of average needle length at time t, Lt , from field studies using a linear conversion Qt = Lt 15 ,
where the average minimum and maximum needle lengths of larch trees observed in the data collected are 15 and 30 mm,
respectively [3]. As a result, Qt is dimensionless and lies between 0 and 1. The dynamics of leaf quality in the absence of the
moth population is modeled by a simple linear equation Qt +1 = (1 − α) + α Qt , where 1 − α is the recovery rate of plant
quality. The plant quality decays to 0 at a rate α if the plant cannot renew itself.
In the absence of the plant population, the dynamics of moth population is governed by a simple linear equation
Nt +1 = λNt , where λ > 0 is the per capita growth rate. The moth population either grows to infinity or decays to zero

∗ Corresponding author.
E-mail address: [email protected] (J.-L. Yu).

0362-546X/$ – see front matter © 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.na.2009.02.091

Please cite this article in press as: S.R.-J. Jang, J.-L. Yu, Models of plant quality and larch budmoth interaction, Nonlinear Analysis (2009),
doi:10.1016/j.na.2009.02.091
 ARTICLE IN PRESS
2 S.R.-J. Jang, J.-L. Yu / Nonlinear Analysis ( ) –

exponentially, depending on whether λ is greater or less than one. When the plant population is present, it is assumed that
the effect of leaf quality is on the budmoth’s intrinsic growth rate, and the effect of plant quality on budmoth is modeled by
a simple expression b+ Q
Q
, where b > 0 is the half saturation constant for the plant quality. The uptake rate of the moth upon
the plant is modeled using a simple Michaelis-Menton form d+ N
N
with maximal uptake rate c > 0, where d > 0 is the half
saturation constant for the moth population.
Under the above biological consideration, the interaction between plant quality and larch budmoth population can be
described by the following system of two-dimensional first order difference equations:

Qt

 N = λNt
 t +1

b+Q

t
cNt (2.1)
 Qt +1 = (1 − α) + α Qt −

 d + Nt
N0 , Q0 ≥ 0


where parameters λ, b, c , d > 0 and 0 < α < 1. In order for solutions of (2.1) to remain nonnegative we assume

c + α < 1. (2.2)

Notice Nt = 0 for t ≥ 1 if either N0 = 0 or Q0 = 0. In this case we have limt →∞ Qt = 1. Moreover, Nt > 0 and Qt > 0 for
t > 0 if N0 > 0 and Q0 > 0.
Since (1 − α − c ) + α Qt < Qt +1 ≤ (1 − α) + α Qt for t ≥ 0, it follows that the Q -component of solutions of (2.1) satisfy

1−α−c
≤ lim inf Qt ≤ lim sup Qt ≤ 1. (2.3)
1−α t →∞ t →∞

Clearly (2.1) always has a trivial steady state E0 = (0, 1) where

 the consumer population goes extinct. The Jacobian matrix of
λ
0
system (2.1) evaluated at E0 has the form J (E0 ) = b + 1 . It follows that E0 is locally asymptotically stable if λ < b + 1
−c /d α
and unstable if λ > b + 1 with stable manifold lying on the Q -axis. Define

b(1 − α)
λ0 = 1 + . (2.4)
1−α−c
Notice λ0 > b + 1. The following theorem provides sufficient conditions for which the moth population either goes extinct
or becomes unboundedly large.

Theorem 2.1. If λ < b + 1, then E0 = (0, 1) is globally asymptotically stable for (2.1). If λ > λ0 , then solutions of (2.1) satisfy
limt →∞ Nt = ∞ and limt →∞ Qt = 1−α−
1−α
c
if N0 > 0 and Q0 > 0.

Proof. The proof of global asymptotic stability of E0 is straightforward if λ ≤ 1. Suppose 1 < λ < b + 1. Let (Nt , Qt ) be an
arbitrary solution of (2.1). Then given any  > 0 there exists t0 > 0 such that Qt < 1 +  for all t ≥ t0 . We choose  > 0 so
λ(1+)
that b+1+ < 1. As a result, Nt +1 < λNt b+1+
1+
for t ≥ t0 implies limt →∞ Nt = 0 and hence limt →∞ Qt = 1. Therefore E0 is
globally attracting in R2+ and thus globally asymptotically stable in R2+ .
Suppose λ > λ0 . Let (Nt , Qt ) be a solution of (2.1) with N0 > 0 and Q0 > 0. Since lim inft →∞ Qt ≥ 1−α−
1−α
c
, for any  > 0
there exists t1 > 0 such that Qt > 1−α−
1−α
c
−  for all t ≥ t1 . Using λ > λ0 , we can choose  > 0 such that

1 − α − c − (1 − α)
λ > 1. (2.5)
b(1 − α) + 1 − α − c − (1 − α)
Then
1−α−c
−
Nt +1 > λNt
1−α
for t ≥ t1
b+ 1−α−c
1−α
−
implies limt →∞ Nt = ∞ by (2.5). Therefore for any M > 0 there exists t2 > 0 such that Nt > M for t ≥ t2 . Hence
1−α− cM
Qt +1 < (1 − α) + α Qt − dcM
+M
for t ≥ t2 and we obtain lim supt →∞ Qt ≤ 1−α
d+M
. Letting M → ∞, we have
1−α−c 1−α−c
lim supt →∞ Qt ≤ 1−α ≤ lim inft →∞ Qt . Therefore, limt →∞ Qt = 1−α and the proof is complete. 

A simple calculation yields that the Q -component of an interior steady state (N̄ , Q̄ ) is Q̄ = λ−
b
1
and the N-component
satisfies
d(1 − α)(1 − Q̄ )
N̄ = .
c − (1 − α)(1 − Q̄ )
Please cite this article in press as: S.R.-J. Jang, J.-L. Yu, Models of plant quality and larch budmoth interaction, Nonlinear Analysis (2009),
doi:10.1016/j.na.2009.02.091
 ARTICLE IN PRESS
S.R.-J. Jang, J.-L. Yu / Nonlinear Analysis ( ) – 3

If λ ≤ b + 1, then Q̄ ≥ 1 and N̄ ≤ 0, and there is no interior steady state for (2.1). If λ > b + 1 then Q̄ < 1 and thus
d(1 − α)(1 − Q̄ ) > 0. Notice
c − (1 − α)(1 − Q̄ ) > 0 if and only if λ < λ0 .
We conclude that system (2.1) has a unique interior steady state Ē = (N̄ , Q̄ ) if λ ∈ (b + 1, λ0 ) and E0 = (0, 1) is the only
steady state if λ 6∈ (b + 1, λ0 ).

Theorem 2.2. If λ ∈ (b + 1, λ0 ), then (2.1) has a unique interior steady state Ē = (N̄ , Q̄ ) which is moreover locally
asymptotically stable.
Proof. The existence and uniqueness of the interior steady state follow from the above discussion. To show Ē is locally
asymptotically stable, consider the Jacobian matrix of system (2.1) evaluated at Ē:

λbN̄
 
1
J (Ē ) =  (b + Q̄ )2  .
 
−cd
α
 
(d + N̄ )2
It follows from the Jury conditions [6] that Ē is locally asymptotically stable if |tr J | < 1 + det J < 2, which results in

cdλbN̄
1+α <1+α+ < 2. (2.6)
(d + N̄ )2 (b + Q̄ )2
The first inequality in (2.6) is always true and the second inequality is equivalent to
(1 − c − α)λ2 + 2(cb + 1 − α − b + α b + c )λ + (1 − α)(1 + b)2 − c (1 + b) > 0.
Let the left hand side of the above inequality be denoted by g (λ). Then Ē is locally asymptotically stable if g (λ) > 0 and
unstable if g (λ) < 0.
Notice g is concave up with g (1 + b) = 3(1 − α) + cb2 + cb > 0, and the vertex of the parabola y = g (λ) is at
α+b−cb−1−α b−c
1−c −α
which is less than 1 + b by a straightforward calculation. We conclude that g (λ) > 0 for λ ∈ (b + 1, λ0 ) and
Ē is locally asymptotically stable whenever it exists. 

3. A plant quality-moth model with Ricker equation

We next consider a model proposed by Turchin [4]. Unlike model (2.1), in the absence of the plant population, it is
rNt
assumed that the dynamics of the moth population is governed by a Ricker equation Nt +1 = Nt e− K , where K > 0 is the
carrying capacity. Moreover, the effect of food source upon the moth population is through intrinsic growth rate. Specifically,
the interaction between larch budmoth and plant quality is now given by the following system:
  
Nt

 N
 t +1
 = Nt exp r Qt −
 K
cNt (3.1)
 Qt +1 = (1 − α) + α Qt −

 d + Nt
N0 , Q0 ≥ 0,


where parameters r , K , c , d > 0 and 0 < α < 1. Similar to system (2.1), we impose condition (2.2) on the parameters c and
α so that solutions of (3.1) remain nonnegative for all forward time. It is clear that limt →∞ (Nt , Qt ) = (0, 1) if N0 = 0, and
Nt > 0, Qt > 0 for t > 0 if N0 > 0. Moreover, solutions of (3.1) satisfy (2.3), i.e., 1−α−
1−α
c
≤ lim inft →∞ Qt ≤ lim supt →∞ Qt ≤
K
1. Since lim supt →∞ Qt ≤ 1, we have lim supt →∞ Nt ≤ r
er −1 and solutions of (3.1) are bounded.

Lemma 3.1. Solutions of (3.1) remain nonnegative for t > 0 and are bounded.

 steadystate E0 = (0, 1) always exists where the moth population is extinct. The Jacobian
Similar to model (2.1), the trivial
er 0
matrix of (3.1) evaluated at E0 is −c /d α .Since r > 0, E0 is always unstable with stable manifold lying on the Q -axis. We
proceed to discuss the existence of an interior steady state. The Q -component of a steady state satisfies
cN
Q =1− . (3.2)
(1 − α)(d + N )
Let h(x) = 1 − (1−α)(
cx
d+x)
. Then h(0) = 1, h0 (x) < 0 for x ≥ 0 and limx→∞ h(x) = 1−α−
1−α
c
> 0. Using Q = h(N ), we see that
the N-component of an interior steady state must satisfy
h(N ) − N /K = 0. (3.3)

Please cite this article in press as: S.R.-J. Jang, J.-L. Yu, Models of plant quality and larch budmoth interaction, Nonlinear Analysis (2009),
doi:10.1016/j.na.2009.02.091
 ARTICLE IN PRESS
4 S.R.-J. Jang, J.-L. Yu / Nonlinear Analysis ( ) –

Since h is monotonically decreasing with h(0) = 1, (3.3) always has a positive solution N ∗ which is moreover unique. Since
(2.2) holds, we have Q > 0 for all Q in (3.2). As a consequence, (3.2) always has a unique interior steady state E ∗ = (N ∗ , Q ∗ ),
where N ∗ solves (3.3) and Q ∗ is given in (3.2). Observe that N ∗ < K , Q ∗ < 1, and both N ∗ and Q ∗ are independent of r.
The Jacobian matrix of system (3.1) evaluated at E ∗ has the following form:
1 − rN ∗ /K
 
rN ∗
∗
J = −cd .
α
(d + N ∗ )2
It follows from the Jury condition [6] that E ∗ is locally asymptotically stable if |tr J ∗ | < 1 + det J ∗ < 2, where tr J ∗ < 1 + det J ∗
if and only if − Kr N ∗ < − αKr N ∗ + (dcdrN
∗
+N ∗ )2
which is clearly true since 0 < α < 1. Therefore E ∗ is locally asymptotically stable
if det J ∗ < 1, i.e., if
 r  cdrN ∗
α 1 − N∗ + <1 (3.4)
K (d + N ∗ )2
and if −1 − det J < tr J ∗ , i.e., if
∗

 r  r cdrN ∗
− 1 − α 1 − N∗ + N∗ − < 1 + α. (3.5)
K K (d + N ∗ )2
Notice if r ≤ 1, then the left hand side of (3.4) is less than α(1 − Kr N ∗ ) + cr ≤ α + c − α Kr N ∗ < 1, i.e., (3.4) holds, and (3.5)
is also clearly true since N ∗ < K . Hence E ∗ is locally asymptotically stable if r ≤ 1.

Proposition 1. System (3.1) always has two steady states E0 = (0, 1) and E ∗ = (N ∗ , Q ∗ ) with N ∗ < K and Q ∗ < 1. Steady
state E0 is always unstable and E ∗ is locally asymptotically stable if r ≤ 1.
Since N ∗ and Q ∗ are independent of r, we use r as our bifurcation parameter. Define
1−α
rh =   (3.6)
cd α
N∗ (d+N ∗ )2
− K

and
2(1 + α)
rp =  . (3.7)
α+1 cd
N∗ K
− (d+N ∗ )2
Notice E ∗ is locally asymptotically stable for all r > 0 if rh ≤ 0 and rp ≤ 0. Also, E ∗ is locally asymptotically stable if r < rh
and r < rp whenever rh > 0 and rp > 0. We proceed to discuss possible bifurcations of system (3.1) when r = rh or r = rp .
We separate our discussion into the following cases: (i) (d+cd N ∗ )2
< Kα , (ii) Kα < (d+cdN ∗ )2 < α+
K
1
and (iii) α+
K
1
< (d+cdN ∗ )2 .
For case (i) we have rh < 0 < rp . Hence det J ∗ < 1 holds and at r = rp we have −1 − det J ∗ = tr J ∗ . Therefore, one of the
eigenvalues is −1 and the other eigenvalue lies between −1 and 1. Hence a period-doubling bifurcation [6] may occur at
r = rp . For case (ii) we see that both rh and rp are positive. If rh < rp , then at r = rh we have det J ∗ = 1 and |tr J ∗ | < 1 + det J ∗ .
Therefore J ∗ has two complex eigenvalues with modulus one and a Hopf bifurcation (Navier–Sacker bifurcation) [6,7] may
occur in this situation. On the other hand if rp < rh , then similar to case (i) we see that a period-doubling bifurcation may
occur at r = rp . Similarly for case (iii) we have rh > 0 and rp < 0 and thus J ∗ has two complex eigenvalues with modulus
one at r = rh . Consequently, a Hopf bifurcation (Navier–Sacker bifurcation) may occur at r = rh .
We next use numerical simulations to study system (3.1). Let K = 0.5, c = 0.2, d = 0.2 and α = 0.7. Then
rh = −0.8064516129 and rp = 3.497942388. Let r = 3.498. Fig. 1(a) provides a solution with initial condition (0.3, 0.4).
The first 1000 iterations are truncated and the figure plots the next 1000 iterations. It can be seen from Fig. 1(a) that the
trajectory is a closed curve in the N–Q plane. On the other hand, if we let K = 0.9, c = 0.8, d = 0.2 and α = 0.04,
then rh = 6.052975059 and rp = 7.076651528. Therefore, according to our earlier analysis, a periodic-doubling bifurcation
may occur when the interior steady state loses its stability. Using r = 6.053 and initial condition (0.3, 0.4), Fig. 1(b) plots
the solution when the first 1000 iterations were truncated. It can be seen from Fig. 1(b) that the system has a period two
solution.

4. Discussion

In this manuscript we have studied two discrete-time leaf quality-LBM population models. In the first model, the moth
population in the absence of food source either grows or decays exponentially and thus the food resource serves as a
population regulator. In this situation, the interior steady state is always locally asymptotically stable whenever it exists.
Therefore the model cannot generate periodic cycles as observed in nature. For the second model, the system may either
undergo a Hopf bifurcation or a period-doubling bifurcation when the interior steady state becomes unstable. It is easy to
see that system (2.1) is over simplified. However, system (3.1) can capture some of the observed cycling phenomenon in the
leaf quality-LBM interaction that has occurred in the Alps environment. We plan to continue the study of LBM population
by incorporating Allee effects [8–11] into the model.

Please cite this article in press as: S.R.-J. Jang, J.-L. Yu, Models of plant quality and larch budmoth interaction, Nonlinear Analysis (2009),
doi:10.1016/j.na.2009.02.091
 ARTICLE IN PRESS
S.R.-J. Jang, J.-L. Yu / Nonlinear Analysis ( ) – 5

Fig. 1. (a) plots one solution of system (3.1) when a period-doubling bifurcation occurs, while (b) plots one solution when a Hopf bifurcation occurs. The
initial condition is (0.3, 0.4) for both plots. Parameter values are presented in the text.

Acknowledgments

Jui-Ling Yu was supported by National Science Council under Grant Nos. P97-2914-I-126-008-A1 and Providence
University under Grant Nos. 95-11100-C10 in Taiwan.

References

[1] W. Baltensweiler, A. Fischlin, The larch budmoth in the Alps, in: A. Berryman (Ed.), Dynamics of Forest Insect Populations: Patterns, Causes,
Implications, Plenum, New York, 1988, pp. 331–351.
[2] A.A. Berryman, What causes population cycles of forest Lepidoptera? Trends Ecol. Evol. 11 (1996) 28–32.
[3] P. Turchin, S. Wood, S. Ellner, B. Kendall, W. Murdoch, A. Fischlin, E. MccAuley, C. Briggs, Dynamical effects of plant quality and parasitism on population
cycling of larch budmoth, Ecology 84 (2003) 1207–1214.
[4] P. Turchin, Complex Population Dynamics, Princeton University Press, New Jersey, 2003.
[5] S.R.-J. Jang, D.M. Johnson, Dynamics of discrete-time larch budmoth population models, J. Biol. Dyn. (in press).
[6] S. Elaydi, Discrete Chaos, Prentice Hall, 2001.
[7] J.E. Marsden, M. McCracken, Hopf Bifurcation and its Applications, Springer, New York, 1976.
[8] W.C. Allee, The Social Life of Animals, William Heinemann, 1938.
[9] S.R.-J. Jang, Allee effects in a discrete-time host–parasitoid model, J. Difference Equ. Appl. 12 (2006) 165–181.
[10] S.R.-J. Jang, Allee effects in a discrete-time host–parasitoid model with stage structure in the host, Discrete Contin. Dyn. Syst., Ser. B 8 (2007) 145–159.
[11] S.R.-J. Jang, S.L. Diamond, A host–parasitoid interaction with Allee effects on the host, Comput. Math. Appl. 53 (2007) 89–103.

Please cite this article in press as: S.R.-J. Jang, J.-L. Yu, Models of plant quality and larch budmoth interaction, Nonlinear Analysis (2009),
doi:10.1016/j.na.2009.02.091