Professional Documents
Culture Documents
ISSN No:-2456-2165
Abstract:- The purpose of the current manuscript is to reversed females bearing two Y chromosomes, i.e.
propose a generic method that causes the local feminized supermales (r), at a constant rate µ to a target
extinction of a harmful invasive species. Eradication is population containing f and m. Mating between the
achieved via introduction of phenotypically modified introduced r and the wild-type m generates a
organisms into a target population. Here we propose a disproportionate number of males over time. The higher
model without the logistic type term, of which the incidence of males decrease the female to male ratio.
reaction terms may change sign, and so the solutions are Ultimately, the number of f decline to zero, causing local
not bounded a priori. We prove global existence of extinction. This theoretical method of eradication is known
solutions via a Lyapunov function method, and show as Trojan Y Chromosome strategy (TYC), [13]. Note, if an
existence of a finite dimensional (L2(Ω);H2(Ω)) global invasive species is used as a biological weapon, one would
attractor that supports states of extinction, improving aim at maximum damage, by choosing a species that might
current results in the literature. We also conduct populate very rapidly, and not grow according to the
numerical simulations to investigate the decay rate of logistic control terms (at least in certain time windows),
the female species. Lastly we apply optimal control assumed traditionally [51, 52]. There is a large literature of
techniques to compare the effectiveness of various such rapid population actuations in the so called case of an
reaction terms on species extinction. insect \outbreak" [4]. Furthermore, past models have not
considered the effects of directed movements, such as
Keywords:- Reaction Diffusion System, Global Existence, movement of the males and supermales towards high
Global Attractor, Optimal Control, Invasive Species, concentrations of females, or avoidance of high
Biological Control. concentrations, of each other. Thus such situations also
need to be considered in our setting.
I. INTRODUCTION
The TYC model has been intensely investigated
An exotic species commonly referred to as invasive recently [13, 42, 41, 43, 14, 66, 63, 59, 64, 44, 45, 61, 15],
species, is any species capable of propagating into a non- and in the case of the classical TYC model, we now know
native environment. As a result of globalization, exotic the attractor is actually in Hs, 8s _ 0, [66]. However, a
species are being introduced to ecosystems around the number of fundamental questions remain unanswered
world at an unprecedented pace, in many cases causing concerning existence of solutions as well as the existence
harm to the environment, human health, and/or the and regularity of a global attractor, in the case that the
economy [35, 34]. Once an exotic species is established in reaction terms are \bad", that is say without logistic control
a new environment, its detrimental potential might be terms, so that no a priori bounds on the solutions are
realized in the form of economic losses or threats to public possible. In [46] we began a program where we study TYC
health. Eradication initiatives in these cases frequently type models for biological control, where we remove the
require continuous efforts for long periods of time. A small logistic type term. We also assume nonlinear and
fraction of the estimated 50,000 exotic species in the US is functionally dependent birth and death rates, instead of the
harmful, but they inflict considerable damage [50, 16]. constant coefficient birth and death rates, assumed earlier.
Studies indicate losses of about $120 billion/year by 2004 In this case the system poses serious mathematical
[50]. A strategy for eradication of exotic species in which a difficulties, as the nonlinearities change sign, and the
\Trojan individual" is strategy is relevant to species with an components of the solution are not priori bounded in some
XY sex-determination system, in which males are the Lp space. There is extensive literature on such problems [1,
heterogametic sex (carrying one X chromosome and one Y 18, 20, 21, 25, 39, and 65]. In [46] we were able to use an
chromosome, XY) and females are the homogametic sex elegant Lyapunov functional to prove global existence of
(carrying two X chromosomes, XX). solutions as well prove the existence of a finite dimensional
(L2(Ω);L2(Ω)) global attractor to a TYC type model. An
Variations in the sex chromosome number can be immediate mathematical question is: Is it is possible to
produced through genetic manipulation; for example, a improve the regularity of the attractor for such a class of
phenotypically normal and fertile male fish bearing two Y models? Also, from a more practical perspective one might
chromosomes termed supermales (s) [2, 6, 7, 8]. Additional ask, what is the decay rate of the female species?
manipulations through hormone treatments can reverse the
sex, resulting in a feminized YY supermale [36, 33, 22].
The eradication strategy involves the addition of sex-
attractor.
We perform numerical simulations to investigate the see []. The solution of (4) with the given boundary
decay rate of the female species, showing numerical conditions exists globally in time and is bounded on
evidence of exponential attraction. We also explore
optimal control scenarios for extinction of the invasive r (t,.) r (t ), in R , (9)
species, for different reaction type terms.
The control method described above is modelled via R+. The primary difficulty to prove the global existence of
the following system of reaction diffusion equations: a solution to (1)-(4), is that the reaction terms given by (7)
can change sign, and thus the solutions to (1)-(3) are not
t f-d1f=r1 (f,m,s,r), (1) bounded a priori.
t m-d 2 m=r2 (f,m,s,r), (2)
III. NOTATIONS AND PRELIMINARY
t s-d3 s=r3 (f,m,s,r), (3) OBSERVATIONS
t r-d 4 r =r4 (f,m,s,r), (4)
in R+Ω with the boundary conditions For the definition of a strong solution we give the
f=m=s=r=0 on R+Ω (5) following (see for example [29])
Since the nonlinear right hand side of (1)-(4) is IV. GLOBAL EXISTENCE
4
continuously differentiable on R+, then for any initial data
For the global existence of the system (1)-(3), we
in C () or Lp(), p 1,+ ( )
, it is easy to check introduce the following functional used in S. Kouachi []
directly its Lipschitz continuity on bounded subsets of the Lp (t ) H p ( f (t , x), m(t , x), s(t , x))dx, (14)
domain of a fractional power of the operator
where
d1 0
0 0 p q q i
i qi pq
H (f,m,s)= C C f m s
0 d 0
2
0
. (12)
p
q=0 i=0
p q i q
. (15)
0 0 d3 0
The sequences and
{ } { }
are real and
0 0 0 d4
i
iN
q
qN
It is well known that to prove global existence of solutions positive satisfying
to (1)-(3) (see, for example [19]), there are several methods
such as the method of comparison with corresponding ii 2 2
ordinary differential equations, method of invariant regions d3, i 1,..., q, (16)
and functional methods based on a priori estimates. This
i 1
2
time by application (to the reaction terms) of the well q q 2 2 2 2 2 2
known regularizing effect (see for example [9]) which is 2 d 1 i 2i 2 d 3 d 2 d 1 d 3 ,
also called LpL smoothing effect of the heat operator q 1 i 1 (17)
(i.e. the diffusion equation has an instantaneous i 1,..., q, q 1,..., p,
regularizing effect in the sense that the above solution u
belongs to L 0,Tmax ,L () regardless of the
[ [ where
di d j
regularity of the initial data and that of the reaction to dk , i j k, i, j , k 1, 2,3.
belong to L 0,Tmax ,Lp () for some p>N/2). The
[ [ 2 di d j
proof is based on the Riesz-Thorin interpolation Theorem (18)
(see e.g. [11]). Rigorously it suffices to derive a uniform Remark 3 Conditions (16)-(17) imply that the sequences
estimate of each ri ( f , m, s, r , , 1i4 on [0,Tmax i+1 q+1
p [ and are increasing and the
for some p>N/2 and deduce that the solution to (1)-(3) is in i iN q qN
sequences i
[
L () for all t 0,Tmax , where Tmax denotes the [ { } and {q} and can be chosen as
iN qN
follows
eventual blow-up time in L (). Under these assumptions,
the following local existence result is well known (see i2 q2
[19,12,51,and 58]). i K d k 3 and i K d 1 ,
(19)
Proposition 1 The system (1)-(4) admits a unique, classical i, q 0,1,...,
solution (f,m,s,r) on [0,T [ . Furthemore if where K and K are any positive constants.
max
Tmax< then We suppose that the polynomials g , g , g and g
2 6 9 10
lim
f (t ,.) m(t ,.) s (t ,.) r (t ,.) (13) (not all constant) are sufficiently large, that is in term of
t Tmax limits
1
g3 fm g 4 fs 12 g5mr
where T
max
denotes the eventual blow-up time in L (). lim 2
,
f msr g2 f
Remark 1 In our setting a classical solution to (1)-(4) can and (20)
be proved to be a strong solution. However, we refrain
2 g6
from this at present time.
lim ,
f msr g1 f
Remark 2 The uniqueness of the solution which is a fixed and
point of a nonlinear operator, is obtained by using
standard arguments (Fixed Point Theorem) and the fact
blow-up approach to controlling invasive populations. i 1,..., q, q 1,..., p,
Thus we can state the following result where
Theorem 4.1 Let ( f (t ,.), m(t ,.), s (t ,.), r (t ,.)) be any ii 2 2 q q 2 2
i d3 , q d1
positive solution of the problem (1)-(4) and suppose that
2 2
the polynomials g2, g6, g9 and g10 are sufficiently large i 1 q 1
3
(conditions (20)-(23)), then under conditions (16)-(18) the and this gives from (17) >0. Consequently we have I0.
functional Lp(t) given by (14) is decreasing on the interval iq
For the second integral we have many ways to prove that
[0,Tmax[ . J0, but we choose only two ways:
Proof. The first, since
Following the same reasoning as in S. Kouachi [28], that is p 2 q
by differentiating Lp with respect to t we got I p( p 1) Cqi C pq2 iq f i mq i s p 1q dx,
q 0 i 0
' f m s
Lp(t)=
fHp t+mHp t +sHp t dx where
q 1 i 1
iq r1 r2 r3 qi
=
(afHpf+bmHpm+csHps)dx (24) q i
with i 1,..., q, q 1,..., p .
+
(r1fHp+r2mHp+r3sHp)dx Replacing the reactions r1, r2 and r3 by their respective
values given by (7), we get
=I+J. iq q 1
Using Green’s formula and the boundary conditions via(5), Gi 12 g 7 mr g8 rs g9 s,
we obtain qi q
p 2 q
I p ( p 1) [C pq 2Cqi BiqT .T ] f i m q i s p 2q dx,
where
i 1 1
q 0 i 0 Gi g fm g2 f 12 g3 fm g 4 fs 12 g5mr g6m
(25) i 2 1
(27)
i+1 i+1
gi ( f , m, s, r) C2 ( f , m, s, r )[1 f m s r ]l on R3 , (29),
g1fmg6m<0< g f g312fmg4fsg512mr,
2 2
i i i 0,1,..., p,
and also this is satisfied if where C2 positive and bounded function on bounded
4
12g fm+g fs+12g mr 2g m subsets of R+ we have the following
3 4 5 i+1 6
,
g f
2 i
g fm
1 Proposition 2 If the reaction terms are of polynomial
growth with g2, g6, g9 and g10 are sufficiently large, then
then under condition (20), we can choose the sequence
{ } i
satisfying (16). Secondly by choosing the all positive solutions of (1)-(4) with initial data in L ()
iN are global.
sequence q { } satisfying Proof: From corollary 1, there exists a positive constant C
3
qN such that
q 1 1
2 g6m 12 g3 fs 12 g5mr g4 f 12 g7 mr g8rs g9 s 0, (1 f (t, x) m(t, x) s(t, x)) dx C ,
p
3 on [0, Tmax [,
q
(30)
for all p1 and from (26) we have
which can be chosen under condition (21).
ri f , m, s, r l2 C2 f , m, s, r (1 f m s r ) p ,
p
The second way is that we choose on [0,Tmax [,
(31)
q+1 i+1 * p [ [
Since f,m,s and r are in L ( 0,T ;L ()), for all
q i
g112fm+g312fm+g4fs+g512mrg9s p N
p1, then we can choose p1 such > and from the
l+2 2
< preliminary observations the solution is global.
q+1 i+1 Remark 6 The global existence can be proved under more
q i 2
0< g f +g6mg712mrg8rs, general boundary conditions including homogeneous and
nonhomogeneous Dirichlet, Neumann and mixed boundary
conditions (see [28]). Also note, because the non linear
that is semi-group S(t) in this case is regularizing [59], for initial
data say u L2(), for some r>0, S(r)u Lp(). We can
0 0
1
g7 mr g8rs q 1 g9 s
2
. now use the constructed functional (14) with initial data
i 1 S(r)u0, which is in Lp(), so the local solution is in Lp(),
g f g6 m q 1 i 1
g fm 2 g3 fm g 4 fs 2 g5mr
1 1
i 2 2
i 1 thus can’t blow up and becomes global. Thus we have a
p
priori L (0,;L ()) bounds for data in L ().
2
As the gi 's are polynomials with positive coefficients, then
condition (22) together with (23) permit us to choose the V. BOUNDED ABSORBING SETS AND FURTHER A
sequence q PRIORI ESTIMATES
{ } satisfying (17). This ends the proof of
qN
5.1 Bounded absorbing sets
the Theorem. By application of the preliminary
In this section we aim to investigate the asymptotic
observations, we have the following,
behavior of (1)-(4). We use the functional Lp to show the
Corollary 1 Suppose that the reaction terms are
4 existence of bounded absorbing sets. Using the fact that the
continuously differentiable on R+, then all positive
matrices Biq, i= 0,q ,q= 0,p are positive definite,
solutions of (1)-(4) with initial data in Lp() are in we can find a constant C4 such that
L(0,T ;Lp()) for all p1.
max p2 2
Proof: If p is an integer, the proof is an immediate Ip(p1)C
4 (
df+m) | (f+m+s)| dx, (32)
consequence of Theorem 3.1 and the trivial inequality
( f (t, x) m(t, x) s(t, x)) dx C L(t),
p
1 on [0, Tmax [, (28) and this gives
absorbing set in Lp(), for all p1, and so in particular for Note for any 2>>0, there exists a t=T*(), s.t
Thus given 2>>0, we can always find >0 via (45), so 4( 2 1) 4 2
* 4 f 4( 2 1)
4 m 4 C.
that t1<T*(). 2
(53)
* 2 1
Finally we choose t1=max(t0,t1), and we have Now using the Sobolev embedding of H ()↪H0(), we
C obtain
2 1
||f||21+ , t>t1. (48) d 2 2
2 ||f||2+C1||f||2C.
1 dt
We next demonstrate next the H0() estimates with f. (54)
We integrate (37) in the time interval from [t,t+1] for any Grönwall Lemma via integration in the time interval [t,t]
tt1, to obtain yields the following uniform bound
CC1 1 e (t t *) e (t t *) f t * 2
2
f
t+1 2
2 2 2
||f(t+1)||2+
2
||f|| 2dsC5+||f(t)||2C6, tt1. (49)
t CC1 C6 , t t* t1.
C1 (55)
Remark 9 Note the C6 absorbs C5 and 1+ from (48),
2
This follows via (51).
C1
so 1+ +C5<C6. .
2
Remark 10 Note, (51) holds for any t>t*. However, the
Thus we have the following uniform integral in time bound *
t+1 reason we use t , is to first derive a uniform in time bound
2
||f||2dsC6, tt1, on the ||f(t)|| 2 via (51), so that the e(tt ))||f(t)|| 2
(50) 2 2
t
using the Mean Value Theorem for integrals, there term can be absorbed, uniformly in time for times tt t1,
*
exists t [t,t+1] such that for all t>t1, we obtain using (51).
* 2
||f(t )||2dsC6, (51) 1
Remark 11 The strategy for the uniform H0() estimates
We next multiply (1) by f and integrate by parts over .
For such higher order Sobolev estimates, we will assume f for the m,s,r components is similar. That is we can derive a
m 2 m
and f satisfy the same boundary conditions, and similarly finite time t3 s.t. ||m||2C, for t>t3 . Here the finiteness
the same is true for the other components. Thus we obtain
m
of the time t , comes via the methods similar to (36)-(55),
3
where we use the equation for m via (2). Similarly we can
respectively. We denote t by
3 Note the regularizing properties of the semigroup yield Lp
m s r 2
t =max(t1,t3 ,t3,t3). (p>2) bounds on the solution, for initial data in L . We now
3
integrate (65) above from t to t to obtain
There exists a constant C independent of time and initial 3
data, and depending only on the parameters in (1)-(4), such t t
2
||f||2ds
Cds.
that for any t>t the following uniform a priori estimates
3
hold: t t
3 3
In particular choosing
2 (56)
||f||2 C, (56) t=t +1,
3
2
||f|| C, (57) yields (57)
2
2 t +1
||m||2 C, (58)
1
3
2
(58)
2
|f|2dsC.
||m||2 C, (59) (t +1)t (59)
3 3 t
2 3
||s||2 C, (60) (60)
2 Hence using a mean value Theorem for integrals we obtain
||s||2 C, (61) (61)
2 that there exists a time t t ,t +1] such that the following
3 3 3
||r||
2
C, (62) (62)
estimate holds uniformly
2 2
||r|| C. (63) ||f(t )|| C. (63)
2 3 2
5.2 Local In Time a Priori Estimate for f. Remark 12 The strategy for the uniform integral in time
Our goal now is to show that we can derive a priori
H2() estimates for the m,s,r components is similar. That
2
H () bounds on the solutions to (1)-(4). To this end we m
next multiply (1) by f and integrate by parts over to is we can derive a finite time t3* s.t.
obtain m
t3*+1
f 2 d1 f a1 f 1 1m 1 1 (f )dx
1 d 2 2
2 m
2 dt 2
||m||2dtC, for t>t3*. Similarly we can derive a
(64) m
a2 f 2 1 2
m (f )dx, t
3*
s
finite time t s.t.
3*
Then employing Young’s inequality yields, s
1d 2 2 t3*+1
||f|| +d ||f|| 2 s
2 dt 2 1 2
||s||2dtC, for t>t3*,
d1 2 4( 1+1) 4(1+1)
||f|| +2||f|| +2||m|| s
4 2 4(1+1) 4(1+1) t
3*
s
t3*+1
+1 1
(a2f 2 (2)m 2 m+a2m 2(2+1)f 2f)(f)dx
2 s
||s||2dtC t>t3*
s
d1 t3*
2 2 2
C2+ ||(f)||2+C3||f||2+C4||m||2. (66) (66)
2
We denote t by
4
This follows via the a priori Lp() bounds on the
* m s
t =max(t3,t3*,t3*).
4
solution, the embedding of H2()↪W1,4() , Cauchy-
Schwartz and Young’s inequalities. Now using the Then there exists a constant C independent of time and
embedding of H3()↪H2() we obtain, initial data, and depending only on the parameters in (1)-
(4), such that for any t>t the following uniform a priori
2 4
d||f||2 d1 estimates hold:
2 2 2
+ ||f||2C2+C3||f||2+C4||m||2. (67)
dt 2
||f(t)|| 2
H ()
C, tt4+1,
Now we recall the Uniform Grönwall Lemma
||m(t)||H2() C, tt4+1,
Lemma 5.2. (Uniform Gronwall Lemma) Let , and h
1
be nonnegative functions in L [0, [. Assume that is ||s(t)|| 2
H ()
C, tt4+1.
loc
absolutely continuous on ]0, [ and the following Thus the existence of a bounded absorbing set in H2() has
differential inequality is satisfied. also been established.
d
+h, for t>0. (68) f
dt 5.4. Uniform a Priori Estimates for
If there exists a finite time t1>0 and some r>0 such that From (1) via brute force we obtain t
2
t r t r t r
f
d A,
t
d B
t
and h d C ,
t
(69) t2
2
d1f+a1f1+1m1+1a2f2+1m2
for any t>t , where A, B and C are some positive constants,
1 =
( ) dx
then
(t) r +CeA, for any t>t1+r.
B
(70) 2 4(1+1) 4(1+1)
C||f||2+C1||f|| +C2||m||
Thus using 4(1+1) 4(1+1)
H ( ) s
f C f (t ) 2 ds
2 * *
m(t ) 2
C, t6* , H () 6 6
t 2
t6* 2
m t6* 1
m 2 f 2
2
s
2
C, t6* , t6* 1
f 2
t * 2 * 2
2
C f (t ) m(t ) ds. (73)
s 2
6 H 2 () 6 H 2 ()
r
2
t6*
C, t *
t
6
2 C.
This easily follows via the estimates in Lemma 5.5. We
f This follows via the regularizing properties of the
next make a local in time estimate on We take the
t semigroup, Lemma 5.5, Lemma 5.4 and the embedding of
H2()↪L()and hence using a mean value Theorem for
partial derivative w.r.t t of (1) and multiply the resulting
f
equation by and integrate by parts over to obtain integrals we obtain that there exists a time
t
1 d f
2
f
2 t6
**
t6 , t6 1
* *
such that the following estimate holds
d1 uniformly
2 dt t 2 t 2 2
f (t6** )
C.
+1 m +1 f f t
a1(1+1)f 1 m 1 t +a1(1+1)f 1m 1 t tdx 2
f
2
f f m
2
m
a2 2 f 2 1m 2 1 ( 2 1) f 2 m 2 dx C, tt6 ,
t t t t L2 ( )
r
2
C, tt6 ,
t
m 2 f
2 L2 ( )
2(1 1) 2( 1 1) 2( 2 1) 2 2
C max f m , f m s
2
t 2 t
2 C, tt6 .
t L2 ( )
d1 f
2
.
2 t 2 VI. EXISTENCE OF GLOBAL ATTRACTOR
d1 ( t t6** ) i) A is compact in X.
e C C. (74)
t 2 t d1 ii) A is invariant, i.e., S(t)A=A, t0.
2 iii) If B is bounded in H, then
t t ** dist (S(t)B,A)0, t.
6 . X
We can make similar estimates for the other components,
m s r Definition 6.2. (Asymptotic compactness) The semi-
and derive similarly absorbing times t ,t ,t , where group {S(t)} associated with a dynamical system is said
6 6 6
m s r t0
(t ,t ,t ) are the absorption times for
6 6 6 2
to be asymptotically compact in H () if for any
m(t ) r (t ) s(t )
, , . 2
t L2 ( ) t L2 ( ) t L2 ( ) { }n=1 bounded in L () and a sequence of times
f0,n
f=m=s=0,t (0, ,)
1
H=(f+m) 2+ f'+ m'+ s'+ r'. (96)
2 1 2 3 4
Figure 2: Here we consider a superlinear source term
2
=(r)=r . In this case r, will blow up in finite time. We We use the Hamiltonian to find a differential equation of
investigate if this source term can cause a faster decay in the adjoint ,i=1,2,3,4.
i
the female species, in comparison to a constant perse.
Surprisingly this is not so. We fix a spatial location and
look at the decay of the trajectory of the female f. Here we 7.3 g=1,(r)=
are comparing (88) -(91) with g=1 to (88) -(91) with g=1
and =r2, and same parameter set as in Fig.1 We '1 t =1-1 ma1 b1 -2 ma2 +sb 2 ,
observe that there is a sharper decay in f with a constant ,
'2 t =1-1a1 f -2 fa2 +rc 2 e2 -3 ra 3 ,
than with =r2.
'3 t =-2 b 2 f -3 rb3 +e3 , (97)
1(T)=2(T)=3(T)=4(T)=0 (102)
Now considering the optimality conditions, the
Hamiltonian function is differentiated with respect to
control variable resulting in
H
= (103)
4
Then a compact way of writing the optimal control is
*(t)=max(0,4) (104)
7.5 g=1,(r)=r2
1(T)=2(T)=3(T)=4(T)=0 (106)
Now considering the optimality conditions, the
Hamiltonian function is differentiated with respect to
control variable resulting in f+m+s+r
Fig 4:- Here, we simulate case 2 where g=1 and
H 2 K
=r (107)
4 (r)=.
Then a compact way of writing the optimal control is
f+m+s+r
Fig 8: Here, we simulate case 2 where g=1 and
K
(r)=.
Fig 5:- Here, we simulate case 3 where g=1 and (r)=r2.
A B
Figure 6: We take the initial conditions as
f =20,m =20,s =3,r =1 and compare the decay
0 0 0 0
Fig 9:- Here, we simulate case 3 where g=1 and (r)=r2.
rate of both females and males in each case.
The existence of a bounded absorbing set indicates The analysis of global attractors can be helpful to
that for either eradication or invasion the final state of the estimate times to extinction in complex spatial domains.
population is stable. Via Theorem 6.8 we showed that given We have determined that for Dirichlet boundary conditions
the initial population and values for the parameters, it is on a connected domain there exists an extinction state as a
possible to find an explicit time such that for times greater result of the introduction of s. However, more complicated
than this, an attractor is reached, in which the population is geometries or boundary conditions could have an influence
confined to finite sized sets in H. That is there is a compact in coexistence or extinction. Increasing the level of
subset of the phase space, that attracts all trajectories, in the sophistication of the eradication strategy, the distribution of
dynamical system. Furthermore, knowing the analytical s individuals could be variable in space as opposed to the
form of the bounded absorbing sets, helps guide the constant level that has been studied, i.e. s could be a
exploration of the parameter space. This and related population density dependent function intended to
problems are the subject of current investigation [33, 32]. minimize the introduction of s individuals and therefore
We numerically see that the attractor is a one point minimize costs of implementation. Also under our
attractor, depending on parameter values. Furthermore, proposed strategy, the system can always be driven to the
what is also tried in the numerical experiments, is that the extinction state, at an exponential rate. This is seen
form of is changed from a constant, to being state numerically as well in Fig1, where we compare the
dependent. trajectories of normal males and females to the function
Also we tried numerical experiments, where we use a e(0.012)t. Clearly, exponential attraction to the extinction
state is seen. It would be interesting to try and rigorously
bad source term such as r2. In this case we compare (88) - prove the existence of an exponential attractor, in this
(91), with g as a logistic term, with the same system when setting.
is replaced with r2. Surprisingly, this causes a slower
decay in the female species, than if one were to use , see The viability of YY individuals remains an open
question. The supermale model assumes that phenotypes
Fig2. We also tried simulations with r3 and r4. What we are stable after maturation, but this could be problematic
observe is that as the power on the source term increases, for species whose sex determination involves many genes,
the decay rate of the female species gets slower. The point or when there is environmental pressure to feminization or
here is one can always stop the influx of the feminized masculinization. To incorporate this we choose death
supermale, before the actual blow up, if the bad source coefficients so that the supermales die at a faster rate than
actually “sped up" the extinction, but this is not seen to the normal males, as these are not considered as fit as their
happen. It is however worth investigating other source normal counterparts, [34], and so fitness penalty should be
terms, in this context. The optimal control experiments tell exercised. Another potential problem is hybridization with
us that in general, whether the initial population of the compatible species, which would extend the eradication
invasive species is large or small, the classic TYC strategy pressure beyond the initial target; however, this effect
(case 2) is always effective from the perspective of should disappear by the interruption of the influx of s. Also
eradication of the invasive species. The larger the initial it would make for very interesting future work if we could
population of the invasive species, the larger the perhaps place sufficient restrictions on the reaction terms in
introduction of superfemales (s) has to be to achieve question to show via Simon-Lojasiewicz gradient
eradication. Optimal control of the TYC system under inequality techniques [49], to show that convergence to the
various parameter regimes is also an area of current spatially homogenous equilibrium state is guaranteed.
investigations [63] that should be pursued further.
To summarize, we have rigorously shown that
We would like to point out that since g ,1i10 are all introduction of phenotypically manipulated supermales into
i
constants in some of our simulations, the conditions for an established population can lead to local extinction.
global existence via (20)-(23) are not met. That is there can Moreover, this can be done even if the population dynamics
be various range of large initial data, and parameters, for