This paper shows that the epidemic model, previously proposed under ordinary differential equation theory, can be generalized to fractional order on a consistent framework of biological behavior. The domain set for the model in which all variables are restricted is established. Moreover, the existence and stability of equilibrium points are studied. We present the proof that endemic equilibrium point when reproduction number R0>1 is locally asymptotically stable. This result is achieved using the linearization theorem for fractional differential equations. The global asymptotic stability of disease-free point, when R0<1, is also proven by comparison theory for fractional differential equations. The numeric simulations for different scenarios are carried out and data obtained are in good agreement with theoretical results, showing important insight about the use of the fractional coupled differential equations set to model babesiosis disease and tick populations.
1. Introduction
Bovine babesiosis is transmitted by the sting of ticks and is the most important disease to attack bovine populations in tropical regions. In warm and hot regions there is great economic loss due to bovine death by bovine babesiosis, with decrease of bovine products and by-products. Moreover, the climate conditions in those regions favor the survival and reproduction of ticks and, consequently, bovines have a permanent contact with these vectors [1]. Furthermore, the vertical transmission in bovines and ticks is possible provided that the ovaries of the female ticks are infected by parasites [1]. The behavior dynamics of diseases has been studied for a long time and is an important issue in the real world. The most important model that can be used to interpret the disease characteristic of epidemics is a susceptible-infected-recuperated model (SIR) that was developed by Kermack and McKendrick [2], and various types of diseases are studied by this type of ordinary differential equation system. Aranda et al. [3] introduced the epidemiological model for bovine babesiosis and tick populations disease. In this work the qualitative dynamics behavior is determined by the basic reproduction number, R0. If the threshold parameter, R0<1, is proved by LaSalle-Lyapunov theorem then the solution converges to the disease free equilibrium point. However, if R0>1, the convergence is to the endemic equilibrium point by numerical simulations. In recent years, the theory of networks in epidemiological model has been introduced in the literature. The purpose of this modification is to have better understanding and prediction of epidemic patterns and intervention measures. For more details see [4–6].
The notion of fractional calculus was introduced by Leibniz, one of the founders of standard calculus, in a letter written in 1695. In recent decades, fractional differential equations are one of the most important topics in mathematics and have received attention due to the possibility of describing nonlinear systems, thus attracting much attention and increasing interest due to its potential applications in physics, control theory, and engineering (see [7–15]). The advantage of fractional-order differential equation systems is that they allow greater degrees of freedom and incorporate the memory effect in the model. Due to this fact, they have been introduced in epidemiological modeling systems. In [16], a fractional order for the dynamics of A(H1N1) influenza disease is studied by numerical simulations. Pooseh et al. [17] and Diethelm [18] have introduced fractional dengue models. In this paper the parameters of the equations obtained in the field research do not reproduce well the evolution of the disease in the case of entire order model. However, when we consider the fractional system, with the same parameters obtained in the field, the data are better adjusted which shows an advantage of the fractional system. In [11] the parameter θ is associated with a memory effect. In [19], the authors attribute to θ the memory information of the dengue disease’s. In this paper, we consider the fractional-order system associated with the evolution of bovine babesiosis disease and tick populations. We introduce a generalization of the classical model presented by Aranda et al. [3]. The generalization is obtained by changing the ordinary derivative by fractional Caputo derivative. It is easy to see that when θ=1 we return to the classical model. For the construction of this model by Aranda et al. [3], the compartments of populations and the biological hypothesis are used. This argument is well established in the disease transmission theory. In Aranda et al., theorems well established in the literature for ordinary differential systems are used. To prove our results, it is necessary to use different tools to those used for the integer order. This is due to the fact that the versions of La-Salle invariance theorem used by Aranda et al. are not found in the literature for fractional-order systems. Therefore, we emphasize that the work presents a collaboration in this direction as when using the comparison theory for fractional-order systems to prove the global stability of the equilibrium free point of the disease by introducing a new type of results in the literature. On the other hand, we also have a test on the local asymptotic stability of endemic equilibrium point, a result that is just enunciated in Aranda et al. [3]. We obtain a generalization of all results in [3]. Our simulation shows that the fractional model has great potential to describe the real problem without the need for adjustment of parameters obtained in field research. This is due to a greater flexibility of adjustment obtained with the introduction of the new parameter. This paper is organized in four sections. Introduction is the first section. In Section 2, we mention a few results and notations related to the theory of fractional differential equations; in Section 3, we consider the fractional-order model associated with the dynamics of bovine babesiosis and ticks populations. Qualitative dynamics of the model is determined by the basic reproduction number. We give a detailed analysis for the global asymptotical stability of disease-free equilibrium point and the local asymptotical stability of the endemic equilibrium point. Finally, in Section 4, numerical simulations are presented to verify the main results.
2. Preliminaries
For many years, there have been several definitions that fit the concept of fractional derivatives [10, 20]. In this paper the Riemann-Liouville fractional derivative and Caputo fractional derivative definitions are presented. Firstly, we introduce the definition of Riemann-Liouville fractional integral(1)Jθft=1Γθ∫0tt-sθ-1fsds,where θ>0, f∈L1(R+), and Γ(·) is the Gamma function.
The Riemann-Liouville derivative is given by (2)DRθft=dndtnJn-θft=1Γn-θdndtn∫0tt-sn-θ-1fsds,n-1≤θ<n.The Caputo fractional derivative is given as follows:(3)DCθft=Jn-θfnt=1Γn-θ∫0tt-sn-θ-1fnsds,where n is the first integer which is not less than θ.
The Laplace transform of the Caputo fractional derivative is given by(4)LDCθft=sθFs-∑k=0n-1fk0sθ-k-1.
The Mittag-Leffler function is defined by the following infinite power series:(5)Eα,βz=∑k=0∞zkΓαk+β.The Laplace transform of the functions is(6)Ltβ-1Eα,β±atα=sα-βsα∓a.
Let α,β>0 and z∈C, and the Mittag-Leffler functions satisfy the equality given by Theorem 4.2 in [10](7)Eα,βz=zEα,α+βz+1Γβ.
Definition 1.
A function f is Hölder-continuous if there are nonnegative constants C, ν such that (8)fx-fy≤Cx-yν,for all x, y in the domain of f and ν is the Hölder exponent. We represent the space of Hölder-continuous functions by C0,ν.
We develop a generalized inequality, wherein the underlying comparison system is a vector fractional-order system.
A nonnegative (resp., positive) vector v means that every component of v is nonnegative (resp., positive). We denote a nonnegative (resp., positive) vector by 0≤≤v (resp., 0<<v).
Consider the fractional-order system:(9)DCθut=ft,u,u0=u0,where DCθu(t)=(DCθu1(t),DCθu2(t),…,DCθum(t))T, 0<θ<1, u(t)∈M⊂Rm, t∈[0,T)(T≤+∞), M is an open set, 0∈M, and f:[0,T)×M→Rm is continuous in t and satisfies the Lipschitz condition:(10)ft,u′-ft,u′′≤Lu′-u′′,t∈0,T,for all u′,u′′∈Ω⊂M, where L>0 is a Lipschitz constant.
Theorem 2 (see [15]).
Let u(t), t∈[0,T), be the solution of system (9). If there exists a vector function v=(v1,v2,…,vm)T:[0,T)→M such that vi∈C0,ν,θ<ν<1,i=1,…,m and(11)DCθvt≤≤ft,vt,t∈0,T.If v(0)≤≤u0,u0∈M, then v(t)≤≤u(t),t∈[0,T).
Now, we will introduce a Theorem of stability for linear systems of fractional order. Let A∈Mm×m(R), and we define the linear system homogeneous equation:(12)DCθxt=Axt,x0=x0.
Definition 3.
We say that linear system (12) is stable if for all ϵ>0, δ>0 exists such that x0<δ; then x(t)<ϵ, for all t≥0; linear system (12) is asymptotically stable if limt→∞x(t)=0.
The next result establishes the stability of the fractional linear system similarly to the theory of ordinary differential equation.
Theorem 4 (see [21]).
System (12) origin is asymptotically stable if and only if arg(λi)>θπ/2 is satisfied for all eigenvalues of the matrix A. Moreover, this system is stable if and only if arg(λi)≥θπ/2 is satisfied for all eigenvalues of the matrix A, and the eigenvalues satisfying arg(λi)=θπ/2 have geometric multiplicity equal to one.
Let f:M→Rm, M∈Rm; we consider the following system of fractional order:(13)DCθxt=fx,x0=x0.
Definition 5.
We say that E is an equilibrium point for (13), if and only if f(E)=0.
Remark 6.
When θ∈(0,1), the fractional system DCθx(t)=f(x) has the same equilibrium points as the system x′(t)=f(x).
Definition 7.
The equilibrium point E of autonomous system (13) is said to be stable if for all ϵ>0, δ>0 exists such that if x0-E<δ, then x(t)-E<ϵ, t≥0; the equilibrium point E of autonomous system (13) is said to be asymptotically stable if limt→∞x(t)=E.
Theorem 8 (see [12]).
The equilibrium points of system (13) are locally asymptotically stable if all eigenvalues λi of Jacobian matrix J, calculated in the equilibrium points, satisfy arg(λi)>θπ/2.
3. Mathematical Model
In this section, we introduce the fractional model for the babesiosis disease in bovine and tick populations. We use the assumptions in Aranda et al. [3] and introduce the following hypotheses.
The total of bovine population N¯B(t) is divided into three subpopulations:
bovines that may become infected (susceptible S¯B(t));
bovines infected by Babesia parasite (infected I¯B(t));
bovines that have been treated for the babesiosis (controlled C¯B(t)).
The parameter μB is the birth rate of bovine. The birth rate μB is assumed to be equal to the natural death.
The total population of ticks N¯T(t) is divided into two subpopulations:
ticks which may become infected by the disease S¯T(t);
ticks infected by the Babesia parasite I¯T(t).
The parameter μT is the birth rate of the ticks and it is assumed to be equal to the death rate.
A susceptible bovine can transit to the infected subpopulation I¯B(t) because of an effective transmission due to a sting of an infected tick at a rate βB.
A susceptible tick can be infected if there exists an effective transmission when it stings an infected bovine, at rate βT.
We assumed a hundred percent vertical transmission in the bovine populations μB. In the tick populations it occurs with probability 1-p, where p is the probability that a susceptible tick was born from an infected one.
A fraction λB of the infected bovine is controlled, that is, treated against Babesia parasite.
A fraction αB of the controlled bovine may return to the susceptible state.
Homogeneous mixing is assumed; that is, all susceptible bovines have the same probability to be infected and all susceptible ticks have the same probability to be infected.
Under the above assumptions, the transmission dynamics of babesiosis disease to bovine and tick population can be modeled by the following system nonlinear ordinary differential equations [3]:(14)S¯B′t=μBS¯Bt+C¯Bt+αBC¯Bt-μBS¯Bt-βBS¯BtI¯TtN¯Tt,I¯B′t=μBI¯Bt+βBS¯BtI¯TtN¯Tt-μBI¯Bt-λBI¯Bt,C¯B′t=λBI¯Bt-μB+αBC¯Bt,S¯T′t=μTS¯Tt+pI¯Tt-βTS¯TtI¯BtN¯Bt-μTS¯Tt,I¯T′t=βTS¯TI¯BtN¯Bt+1-pμTI¯Tt-μTI¯Tt.
In recent years, a considerable interest in the fractional calculus has been shown, which allows us to consider integration and differentiation of any order. To a large extent this is due to the applications of the fractional calculus to problems in different areas of research. The advantage of fractional-order differential equation systems is that they allow greater degrees of freedom and incorporate memory effect in the model. Now we describe the new system of fractional differential equations to model the babesiosis disease in bovine and tick populations, and in this system, θ∈(0,1):(15)DCθS¯Bt=μBS¯Bt+C¯Bt+αBC¯Bt-μBS¯Bt-βBS¯BtI¯TtN¯Tt,DCθI¯Bt=μBI¯Bt+βBS¯BtI¯TtN¯Tt-μBI¯Bt-λBI¯Bt,DCθC¯Bt=λBI¯Bt-μB+αBC¯Bt,DCθS¯Tt=μTS¯Tt+pI¯Tt-βTS¯TtI¯BtN¯Bt-μTS¯Tt,DCθI¯Tt=βTS¯TI¯BtN¯Bt+1-pμTI¯Tt-μTI¯Tt.Simplifying the system (15) and using the bovine populations constant equal N¯B and tick populations is N¯T and introducing the proportions(16)SBt=S¯BtN¯Bt,IBt=I¯BtN¯Bt,CBt=C¯BtN¯Bt,STt=S¯TtN¯Tt,ITt=I¯TtN¯Tt,we obtain the following fractional system that describes the dynamics of the proportion of bovines in each class:(17)DCθSBt=μB+αB1-SBt-IBt-βBSBtITt,DCθIBt=βBSBtITt-λBIBt,DCθITt=βT1-ITtIBt-μTpITt,defined in the region Ω={(SB,IB,IT):0≤SB+IB≤1,0≤IT≤1}. Next, we show all variables of the babesiosis model living in Ω for all time t≥0. To establish our first result we introduce the following lemma.
Lemma 9 (see [22]).
Let the function f∈C[t0,t1] and its fractional derivative DCθf(t)∈C(t0,t1] for 0≤θ<1, and t0,t1∈R; then one has (18)ft=ft0+1ΓαDCθfτt-t0α,for all t∈(t0,t1], where t0≤τ<t.
Thus, considering the interval [0,t1] for any t1>0, this theorem implies that the function f:[0,t1]→R+ is nonincreasing on (0,t1) if DCθf(t)≤0 for all t∈(0,t0) and nondecreasing on [0,t0] if DCθf(t)≥0 for all t∈(0,t0).
Proposition 10.
The region Ω={(SB,IB,IT):0≤SB+IB≤1,0≤IT≤1} is a positive invariant set for system (17).
Proof.
By Theorem 3.1 and Remark 3.2 in [23] we obtain the global existence and uniqueness of the solutions of (17).
We denote by Ω+={(SB,IB,IT):SB≥0,IB≥0 and IT≥0}. If (SB(0),IB(0),IT(0))∈SB-axis ={(SB,0,0):SB≥0} (with the same form we define IB-axis and IT-axis). The vector field from (17) confined in SB-axis assumes the form F(SB,IB,IT)=((μB+αB)-(μB+αB)SB(t),0,0), by the Laplace transform properties (6), and we obtain the solution(19)SBt,IBt,ITt=tθEθ,θ+1-μB+αBtθμB+αB+Eθ,1-μB+αBtθSB0,0,0∈SB-axis.
By the same argument, if (SB(0),IB(0),IT(0))∈IB-axis we obtain (20)SBt,IBt,ITt=0,Eθ,1-λBtθIB0,0∈IB-axisand if (SB(0),IB(0),IT(0))∈IT-axis, we have (21)SBt,IBt,ITt=0,0,Eθ,1-μTptθIT0∈IT-axis.This proves that axes SB, IB, and IT are solutions and positive invariants sets.
Now, we will prove that Ω+ is a positive invariant set. By way of contradiction, suppose there exists a solution (SB,IB,IT) such that (SB(0),IB(0),IT(0))∈Ω+ and the solution (SB(t),IB(t),IT(t)) to escape of Ω+. From the previous argument and by the unicity of solutions (SB(t),IB(t),IT(t)) do not cross the axis. From the previous conclusion we have three possibilities.
If the solution (SB(t),IB(t),IT(t)) escapes by the plane SB=0, then there exists t0 such that SB(t0)=0, IB(t0)>0 and IT(t0)>0 and for all t>t0 sufficiently near t0 we have SB(t)<0. On the other hand, DCθSB(t)|t=t0=(μB+αB)(1-IB(t0))>(μB+αB)>0. From Lemma 9, we obtain SB(t)≥SB(t0)≥0 for all t sufficiently near t0, and this is absurd.
If the solution (SB(t),IB(t),IT(t)) escape by IB=0, then there exists t0 such that SB(t0)>0, IB(t0)=0, and IT(t0)>0 and for all t>t0 sufficiently near t0 we have IB(t)<0. Again, DCθIB(t)|t=t0=βBSB(t0)IT(t0)>0. From Lemma 9, we obtain IB(t)≥IB(t0)≥0 for all t sufficiently near t0, and this is a contradiction.
If the solution (SB(t),IB(t),IT(t)) escape by IT=0, then there exists t0 such that SB(t0)>0, IB(t0)>0 and IT(t0)=0 and for all t>t0 sufficiently near t0 we have IT(t)<0. We obtain DCθIT(t)|t=t0=βTIB(t0)>0 and by Lemma 9, we have IT(t)≥IB(t0)≥0 for all t sufficiently near t0, and this is false.
Therefore, we obtain SB(t)≥0, IB(t)≥0 and IT(t)≥0, for all t≥0.
If 0≤SB(0)+IB(0)≤1, from the two first equations of system (17), we get(22)DCθSBt+IBt=μB+αB-μB+αBSBt+IBt-λBIBt≤μB+αB-μB+αBSBt+IBt.Applying the Laplace transform in the previous inequality, we have (23)λθLSBt+IBt-λθ-1SB0+IB0≤μB+αB1λ-μB+αBLSBt+IBt,that can be written as(24)LSBt+IBt≤μB+αBλθ-1+θλθ+μB+αB+λθ-1λθ+μB+αBSB0+IB0.From the Laplace transform properties (6) and equality (7) we infer (25)SBt+IBt≤tθEθ,θ+1-μB+αBtθμB+αB+Eθ,1-μB+αBtθSB0+IB0≤tθEθ,θ+1-μB+αBtθμB+αB+Eθ,1-μB+αBtθ=1.Therefore, we have that 0≤SB(t)+IB(t)≤1.
On the other hand, if 0≤IT(0)≤1, from system (17), we obtain (26)DCθITt=βT1-ITtIBt-μTpITt≤βT+μTp-βT+μTpITt.The proof of 0≤IT(t)≤1 is similar to the previous case. Finally, we conclude that Ω is a positive invariant set.
In the following result we study the existence and stability of the equilibrium points of system (17). Motivated by Aranda et al. [3], we will use the following threshold parameter. For more details on the threshold parameter, see [24, 25]:(27)R0=βBβTλBμTp.
The next result is similar to Proposition 1 in [3], and so we omit its proofs.
Theorem 11.
System (17) has the disease-free equilibrium point: (28)E1=SB1,IB1,IT1=1,0,0,for all the values of the parameters in this system, whereas only if R0>1, there is (unique) endemic equilibrium point: (29)E2=SB2,IB2,IT2,where(30)SB2=λBαB+μBβT+αB+μB+λBλBβTαBβB+λB+λBμB+βBλB+μB,IB2=μB+αBβTβB-λBμTpβTαBβB+λB+μBλB+βBμB+λB,IT2=αB+μBβBβT-λBμTpαB+μBβBβT+αB+μB+λBβBμTp,in the interior of Ω.
Computing the Jacobian matrix of system (17) evaluated at the disease-free point, one gets (31)JE1=-μB+αB-μB+αB-βB0-λBβB0βT-μTp,and consequently, the eigenvalues of J(E1) are (32)λ1=-μB+αB,λ2=-λB+μTp+Δ2,λ3=-λB+μTp-Δ2,where Δ=(λB-μTp)2+4βBβT. It is easy to see that λ1 and λ3 are negative numbers. If R0<1 we observe (33)Δ=λB-μTp2+4βBβT=λB2+μT2p2-2λBμTp+4βBβT<λB2+μT2p2+2λBμTp=λB+μTp2.We infer that (34)λ2=-λB+μTp+Δ2<-λB+μTp+λB+μTp2=0.Therefore, λ2<0; then we have that all eigenvalues of the Jacobian matrix at E1 are negative: that is, arg(λi)=π, i=1,2,3, and from Theorem 8, we have that disease-free equilibrium point E1 is locally asymptotically stable. Consequently, we have the following Theorem.
Theorem 12.
If R0<1, then the disease-free point E1 is locally asymptotically stable.
In the next result we prove the global asymptotical stability of the disease-free equilibrium point.
Theorem 13.
If R0<1, then the disease-free point E1 is globally asymptotically stable.
Proof.
Suppose that (SB(t),IB(t),IT(t)) is the solution of system (17). Making the change of variables LB=1-SB we obtain the new system:(35)DCθLBt=-μB+αBLBt+μB+αBIBt+βBITt-βBLBtITt,DCθIBt=βB1-LBtITt-λBIBt,DCθITt=βT1-ITtIBt-μTpITt.It is easy to see that(36)-μB+αBLBt-IBt+βBITt-LBtITt≤-μB+αBLBt-IBt+βBITt,βB1-LBtITt-λBIBt≤βBITt-λBIBt,βT1-ITtIBt-μTpITt≤βTIBt-μTpITt.From the above, it follows that the solutions (LB(t),IB(t),IT(t)) of system (35) satisfy the differential inequality:(37)DCθLBt≤-μB+αBLBt+μB+αBIBt+βBITt,DCθIBt≤βBITt-λBIBt,DCθITt≤βTIBt-μTpITt.Moreover, motivated by (37), let (X(t),Y(t),Z(t)) be the solution of fractional linear system:(38)DCθXt=-μB+αBXt+μB+αBYt+βBZt,DCθYt=βBZt-λBYt,DCθZt=βTYt-μTpZt,with initial conditions (X(0),Y(0),Z(0))=(X0,Y0,Z0)∈Ω.
The eigenvalues of system (38) are given by (39)-μB+αBμB+αBβB0-λBβB0βT-μTp.Similar to the proof of Theorem 12, we infer that all the eigenvalues are negatives; thus, arg(xi)=π, i=1,2,3, and from Theorem 4, we can conclude that limt→∞X(t)=0, limt→∞Y(t)=0, and limt→∞Z(t)=0.
From the previous discussion and the comparison principle, Theorem 2, we have (40)LBt,IBt,ITt≤≤Xt,Yt,Zt.This implies limt→∞(LB(t),IB(t),IT(t))=(0,0,0), and it follows that (SB(t),IB(t),IT(t)) converge to the disease-free equilibrium point E1=(1,0,0), when R0<1. This ends the proof.
Now we show the local stability of the endemic equilibrium point E2, and we give the definition of an additive compound matrix. For more details see [26, 27].
Definition 14.
Let A be any n×m matrix of real and complex numbers, and let ai1,…,jk be the minor of A determined by the rows (i1,…,ik) and the columns (j1,…,jk),1≤i1<i2,…,<ik≤n,1≤j1<j2,…,<jk≤m. The kth multiplicative compound matrix of Ak of A is the nk×nk matrix whose entries, written in a lexicographic order, are ai1,…,jk. When A is a n×m matrix with columns a1,a2,…,ak,Ak is the exterior product a1∧a2∧⋯∧ak.
Definition 15.
If A=aij is a n×n matrix, its kth additive compound A[k] of the A is the nk×nk matrix given by A[k]=D(I+hA)(k)=0, where D is a differentiation with respect to h. For any integers i=1,…,nk, let (i)=(i1,…,ik) be the ith member in the lexicographic ordering of all k-tuples of integers such that 1≤i1<i2<⋯<ik≤in. Then(41)bij=ai1i1+⋯+aikikifi=j,-1r+saisir,if one entry of is of i does not occur in j and js does not occur in i,0,if i differs from j in two or more entries.
Remark 16.
For n=3, the matrices A[k] are as follows: (42)A1=A,A2=a11+a22a23-a13a32a11+a33a12-a31a21a22+a33,A3=a11+a22+a33.
The next lemma is stated and proved in [28].
Lemma 17.
Let M be a 3×3 real matrix. If tr(M)<0, det(M)<0, and det(M[2])<0 are all negative, then all eigenvalues of M have negative real part.
Theorem 18.
If R0>1, μB+αB>βT, and μB+αB>βB, then endemic equilibrium point E2 is locally asymptotically stable.
Proof.
The Jacobian matrix of systems (17) in the endemic equilibrium point is given by(43)JE2=-μB+αB+βBIT-μB+αB-βBSBβBIT-λBβBSB0βT1-IT-IBβT-μTp.
From J(E2), we have tr(J(E2))=-(μB+αB+βBIT)-λB-IBβT-μTp<0.
To show the detJ(E2)<0, we will make a simplification into system (17), where it comes from(44)-μB+αB=-βBSBIT1-SB-IB,λB=βBSBITIB,μTp=βT1-ITIBIT.Substituting (44) in the matrix (43), we obtain (45)detJE2=-βBIT1-IB1-SB-IB-βBSBIT1-SB-IB-βBSBβBIT-βBSBITIBβBSB0βT1-IT-βTIBIT.Then (46)detJE2=-1-SB-IBIBIT1-IBβBSBβTIT1-SB-IBIBIT1-SB-IB-βBSBβBITβT1-IT-βBSBITβBITβTIB1-SB-IBIT=-βBSBβTIT1-SB-IBIBIT1-IB1-SB-IBIBIT1-SB-IB+βB1-IT-βBSBβTITβBITIB1-SB-IBIT.Therefore, as all are constant positive parameters, it follows that det(J(E2))<0.
Let J[2](E2) be the additive compound matrix: (47)J2E2=M-λBβBSBβBSBβT-βTITM+K-μB+αB0βBIT-λB+K,where M=-(μB+αB+βBIT) and K=-(IBβT+μTp). From the hypothesis 0≤1-IT≤1, we get (48)detJ2E2=-μB+αB+βBIT+λB·μB+αB+βBIT+IBβT+μTp·λB+IBβT+μTp+βBSBβT1-ITITβB+λB+IBβT+μTp-μB+αB+βBIT+λBμB+αBβBIT≤-μB+αB+βBIT+λB·μB+αB+βBIT+IBβT+μTp·λB+IBβT+μTp+βBβTITβB+λB+IBβT+μTp-μB+αB+βBIT+λBμB+αBβBIT=-λB+IBβT+μTpμB+αB+βBIT+λB·μB+αB+βBIT+IBβT+μTp-βBβT-βBITμB+αB+βBIT+λBμB+αB-βBβT.Analyzing the terms of equality above, we have (49)μB+αB+βBIT+λB·μB+αB+βBIT+IBβT+μTp>βBβT,μB+αB+βBIT+λBμB+αB>βBβT.Then det(J[2](E2))<0 and from Lemma 17, the endemic equilibrium point (E2) is locally asymptotically stable. This concludes the proof.
4. Numerical Simulations
In this section, we simulate different possible scenarios to check the effect that some values of fractional exponent θ have on the dynamics of bovine babesiosis disease and tick populations. For comparison purposes, we will use the same parameters as Aranda et al. [3]. To solve a nonlinear differential equation set with fractional order, a method based on the classical Adams-Bashforth-Moulton approach was used, as presented in [29]: (50)fj,m=fj,1+J,in which j=1,2, and 3 represents population number: SB, IB, and IT, respectively. The time is defined as tm=(m-1)H in which 2<m<N+1 and N=T/H, with T equal to the final time. The fractional integral is determined by modified trapezoidal rule as (51)J=∑k=0n-1fi,kwμ+fi,k-fi,k+1gμh,in which μ=n-k, h=tm/n, and tk=kh, (52)wμ=hθγθ+1μθ-μ-1θ,gμ=hθ+1γθ+2μθ+1-μ+θμ-1θ.
In this work H=40 and n=400. More details about the numerical integration algorithm can be found in [29, 30].
Figure 1 shows the dynamics of the bovine babesiosis disease and tick populations, with initial condition of SB=0.3756, IB=0.5184, and IT=0.6000, and reproduction number R0=67.54. As can be seen, following the course of the disease, the system evolves to the endemic equilibrium point with population number of SB2=0.04967, IB2=0.7894, and IT2=0.7019, as determined by (30). The convergence to the equilibrium point, when R0>1, is predicted by Theorem 18. The variables SB, IB, and IT drop to less than 1% of the equilibrium values above 6280 years, when a veterinary intervention was simulated making R0 less than 1 (R0=0.6754). This new R0 value was obtained with βB equal to 1/10 of the initial value. Now the system gets out of endemic equilibrium point and evolves to the disease-free equilibrium point (1,0,0), as predicted by Theorems 12 and 13. The control parameters of differential equation set are presented in Table 1.
The control parameters.
Parameter
Value
μB
0.0002999
αB
0.001
βB
0.006
λB
0.000265
βT
0.00048
μT
0.001609
p
0.1
Dynamic of the bovine babesiosis disease. IB (continuous line) together with SB (dashed line) and IT (dotted line) were shown as function of the time.
A comparison between two different values of the fractional order is shown in Figure 2, with the same control parameter shown in Table 1. Figure 2 shows a different behavior for θ=0.9 and θ=1, with a maximum value of SB and a minimum value of IB, that does not appear when θ=1. For both cases, the disease evolves to the endemic equilibrium point; however, it is slower when θ=0.9.
Dynamic of the bovine babesiosis disease, with θ=1 (continuous line) and θ=0.9 (dotted line).
Table 2 shows the time τ1% in which variables drop to less than 1% of the equilibrium values. These times were obtained with different values of θ. As we can see, the time τ1% increases when θ decreases. The time τ1% as function θ was adjusted by two linear equations, τ1%×1/θ and ln(τ1%)×1/θ. The first case is consistent with exponential behavior and the second case with power law t-θ. After the statistical analysis based on the correlation coefficient, 0.90013 against 0.98018, one concludes that the system decays to equilibrium condition like power law t-θ. This result was previously proven under theoretical assumptions [21].
Relaxation time.
θ
τ1%/years
1
5120
0.975
8640
0.95
15200
0.925
26720
0.9
42840
0.875
49560
0.85
65840
0.825
70880
0.8
96680
5. Conclusions
We did not find global stability results for fractional differential order equations in the literature. This way, we obtain a new result for global asymptotical stability of disease-free equilibrium using comparison theory of fractional differential equations since R0<1, and therefore the proof that endemic equilibrium point, when R0>1, μB+αB>βC, and μB+αB>βC, is locally asymptotically stable was achieved using the linearization theorem for fractional differential equations. Therefore, if R0<1 so the system evolves to endemic equilibrium point. To return to disease-free status, the R0 value should be greater than 1. The R0<1 is achieved when parameters βB and βC are very small or when parameters λB, μC, and p are very large. Therefore, biological strategy to combat babesiosis disease would have to focus on one of these parameters. These results were confirmed by numerical simulations using the extension of Adams-Bashforth-Moulton algorithm.
Numeric simulations of improved epidemic model with arbitrary order have shown that fractional order is related to relaxation time, in other words, the time taken to reach equilibrium. Numerical simulations with different order show that the system decays to equilibrium condition like power law t-θ, as previously established in [21]. This result provides an important insight about the use of fractional order to model the dynamics of babesiosis disease and tick population. The proof shown here should be used as a guide in the study of equilibrium conditions in similar problems, such as tuberculosis [28], malaria [31], or toxoplasmosis disease [32].
Conflict of Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
Acknowledgments
The authors would like to thank the editor and the anonymous reviewers for their valuable comments and constructive suggestions. José Paulo Carvalho dos Santos is partially supported by FAPEMIG/Brazil under Grant CEX-APQ-00748-12. Lislaine Cristina Cardoso and Nelson H. T. Lemes are supported by FAPEMIG/Brazil.
BenavidesE.Considerations with respect to the epizootilogia of anaplasmosis and babesiosis in the bovines198531411KermackW. O.McKendrickA. G.A contribution to the mathematical theory of epidemics192711577270072110.1098/rspa.1927.0118ArandaD. F.TrejosD. Y.ValverdeJ. C.VillanuevaR. J.A mathematical model for babesiosis disease in bovine and tick populations201235324925610.1002/mma.1544MR28768072-s2.0-84856260051KeelingM. J.EamesK. T. D.Networks and epidemic models20052429530710.1098/rsif.2005.00512-s2.0-33646893242WangY.JinZ.YangZ.ZhangZ.-K.ZhouT.SunG.-Q.Global analysis of an SIS model with an infective vector on complex networks201213254355710.1016/j.nonrwa.2011.07.0332-s2.0-80054951532WangY.CaoJ.Global dynamics of a network epidemic model for waterborne diseases spread201423747448810.1016/j.amc.2014.03.148MR32011472-s2.0-84899119114CaputoM.1992Roma, ItalyDipartimento di Fisica, Universitá La SapienzaLecture NotesCiesielskiM.LeszczynskiJ.Numerical simulations of anomalous diffusionProceedings of the 15th Conference on Computer Methods in Mechanics (Wisla Polonia), http://arxiv.org/abs/math-ph/0309007DemirciE.OzalpN.A method for solving differential equations of fractional order2012236112754276210.1016/j.cam.2012.01.005MR2891361ZBL1243.340032-s2.0-84857790460DiethelmK.2004SpringerDuM.WangZ.HuH.Measuring memory with the order of fractional derivative20133, article 343110.1038/srep034312-s2.0-84890107623LiC.MaY.Fractional dynamical system and its linearization theorem201371462163310.1007/s11071-012-0601-1MR30301252-s2.0-84878697727LorenzoC. F.HartleyT. T.Initialization, conceptualization, and application in the generalized (fractional) calculus200735644755310.1615/critrevbiomedeng.v35.i6.102-s2.0-68749115189McCallP. J.KellyD. W.Learning and memory in disease vectors2002181042943310.1016/s1471-4922(02)02370-x2-s2.0-0036790831WangZ.YangD.MaT.SunN.Stability analysis for nonlinear fractional-order systems based on comparison principle2014751-238740210.1007/s11071-013-1073-7MR31448602-s2.0-84891556704González-ParraG.ArenasA. J.Chen-CharpentierB. M.A fractional order epidemic model for the simulation of outbreaks of influenza A(H1N1)201437152218222610.1002/mma.2968MR32647212-s2.0-84908117489PoosehS.RodriguesH.TorresD.Fractional derivatives in Dengue epidemics2011138973910.1063/1.3636838DiethelmK.A fractional calculus based model for the simulation of an outbreak of dengue fever201371461361910.1007/s11071-012-0475-2MR30301242-s2.0-84878363605SardarT.RanaS.ChattopadhyayJ.A mathematical model of dengue transmission with memory2015221–351152510.1016/j.cnsns.2014.08.009MR32824412-s2.0-84916633344PodlubnyI.2005Logan, Utah, USAUtah State UniversityMatignonD.Stability results for fractional differential equations with applications to control processing19962963968OdibatZ. M.ShawagfehN. T.Generalized Taylor's formula2007186128629310.1016/j.amc.2006.07.1022-s2.0-33947277891LinW.Global existence theory and chaos control of fractional differential equations2007332170972610.1016/j.jmaa.2006.10.040MR23196932-s2.0-34247154280Cintrón-AriasA.Castillo-ChávezC.BettencourtL. M.LloydA. L.BanksH. T.The estimation of the effective reproductive number from disease outbreak data20096226128210.3934/mbe.2009.6.261MR25320162-s2.0-67650458331DietzK.The estimation of the basic reproduction number for infectious diseases199321234110.1177/0962280293002001032-s2.0-0027736832MarshK.OtooL.HayesR. J.CarsonD. C.GreenwoodB. M.Antibodies to blood stage antigens of Plasmodium falciparum in rural Gambians and their relation to protection against infection198983329330310.1016/0035-9203(89)90478-12-s2.0-0024349561TumwiineJ.MugishaJ. Y. T.LuboobiL. S.A mathematical model for the dynamics of malaria in a human host and mosquito vector with temporary immunity200718921953196510.1016/j.amc.2006.12.0842-s2.0-34249096551McCluskeyC.van den DriesscheP.Global analysis of tuberclosis models200416139166HoffmanJ.FrankelS.2001CRC PressBaleanuD.DiethelmK.ScalasE.TrujilloJ.20123World Scientific10.1142/9789814355216MR2894576NgwaG. A.ShuW. S.A mathematical model for endemic malaria with variable human and mosquito populations2000327-874776310.1016/S0895-7177(00)00169-2MR1792096ZBL0998.920352-s2.0-0033752986González-ParraG. C.ArenasA. J.ArandaD. F.VillanuevaR. J.JódarL.Dynamics of a model of Toxoplasmosis disease in human and cat populations200957101692170010.1016/j.camwa.2008.09.012ZBL1186.340632-s2.0-65049091290