Abstract
In this paper, we examine a stochastic avian influenza model with a nonlinear incidence rate within avian populations and the psychological effect within the human population, where susceptible humans reduce their contact with infected avians as the number of infected humans increases. For the deterministic model, the basic reproduction number \(\mathscr {R}_0\), possible equilibria, and related asymptotic stability are first studied. Then, for the stochastic model, we obtain a critical value \(\mathscr {R}_0^S\), which can determine the persistence and extinction of avian influenza. It is theoretically proved that the stochastic model has a unique stationary distribution \(\varpi (\cdot )\) if \(\mathscr {R}_0^S>1\), but the disease will go to extinction when \(\mathscr {R}_0^S<1\). Taking stochasticity into account, a quasi-endemic equilibrium \(\overline{T}^*\) related to the endemic equilibrium of the deterministic model is defined. We develop an important lemma for solving the special Fokker–Planck equation and derive the explicit expression of the density function of the distribution \(\varpi (\cdot )\) around the equilibrium \(\overline{T}^*\). Numerical simulations verify our theoretical results, and we study the impact of noise and the psychological effect on the transmission dynamics of avian influenza.
Similar content being viewed by others
Data Availability
Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.
References
Agusto, F.B.: Optimal isolation control strategies and cost-effectiveness analysis of a two-strain avian influenza model. BioSystems 113(3), 155–164 (2013)
Allen, L.J.S.: An introduction to stochastic epidemic models. Math. Epidemiol. 1945(4), 81–130 (2008)
Bai, N., Song, C., Xu, R.: Mathematical analysis and application of a cholera transmission model with waning vaccine-induced immunity. Nonlinear Anal. RWA 58, 103232 (2021)
Cai, Y., Kang, Y., Banerjee, M., Wang, W.M.: A stochastic epidemic model incorporating media coverage. Commun. Math. Sci. 14(4), 893–910 (2016)
Cai, Y., Jiao, J., Gui, Z., Liu, Y., Wang, W.: Environmental variability in a stochastic epidemic model. Appl. Math. Comput. 329, 210–226 (2018)
Capasso, V., Serio, G.: A generalization of the Kermack-McKendrick deterministic epidemic model. Math. Biosci. 42, 43–61 (1978)
Caraballo, T., Fatini, M.E., Khalifi, M.E.: Analysis of a stochastic distributed delay epidemic model with relapse and Gamma distribution kernel. Chaos. Soliton. Fract. 133, 109643 (2020)
Chen, Q.: A new idea on density function and covariance matrix analysis of a stochastic SEIS epidemic model with degenerate diffusion. Appl. Math. Lett. 103, 106200 (2019)
Chong, N.S., Tchuenche, J.M., Smith, R.J.: A mathematical model of avian influenza with half-saturated incidence. Theory Biosci. 133(1), 2338 (2014)
Chowell, G., Ammon, C.E., Hengartner, N.W., Hyman, J.M.: Transmission dynamics of the great influenza pandemic of 1918 in geneva, switzerland: assessing the effects of hypothetical interventions. J. Theor. Biol. 241, 193–204 (2006)
Du, N.H., Nhu, N.N.: Permanence and extinction of certain stochastic SIR models perturbed by a complex type of noises. Appl. Math. Lett. 64, 223–230 (2017)
Du, N.H., Dieu, N.T., Nhu, N.N.: Conditions for permanence and ergodicity of certain SIR epidemic models. Acta. Appl. Math. 160, 81–99 (2019)
Du, Y., Kang, T., Zhang, Q.: Asymptotic behavior of a stochastic delayed avian influenza model with saturated incidence rate. Math. Biosci. Engine. 17(5), 5341–5368 (2020)
Gardiner, C.W.: Handbook of stochastic methods for physics, chemistry and the natural sciences. Springer, Berlin (1983)
Han, B., Jiang, D., Zhou, B., et al.: Stationary distribution and probability density function of a stochastic SIRSI epidemic model with saturation incidence rate and logistic growth. Chaos. Soliton. Fract. 142(5), 110519 (2020)
Higham, D.J.: An algorithmic introduction to numerical simulation of stochastic differential equations. SIAM Rev. 43, 525–546 (2001)
Ikeda, N., Watanade, S.: A comparison theorem for solutions of stochastic differential equations and its applications. Osaka J. Math. 14, 619–633 (1977)
Iwami, S., Takeuchi, Y., Liu, X.: Avian-human influenza epidemic model. Math. Biosci. 207, 1–25 (2007)
Jiang, X., Yang, Y., Meng, F., Xu, Y.: Modelling the dynamics of avian influenza with nonlinear recovery rate and psychological effect. J. Appl. Anal. Comput. 10(3), 1170–1192 (2020)
Kang, T., Zhang, Q.: Dynamics of a stochastic delayed avian influenza model with mutation and temporary immunity. Int. J. Biomath. 14(5), 2150029 (2021)
Keeling, M.J., Rohani, P.: Modeling Infectious Diseases in Humans and Animals. Princeton University Press, Princeton (2008)
Khan, M.A., Farhan, M., Islam, S., et al.: Modeling the transmission dynamics of avian influenza with saturation and psychological effect. Discrete Contin. Dyn. Sys. Ser. S 12(3), 455–474 (2019)
Khasminskii, R.: Stochastic Stability of Differential Equations. Springer, Berlin (2011)
Lee, H., Lao, A.: Transmission dynamics and control strategies assessment of avian influenza A (H5N6) in the Philippines. Infect Dis Model 3, 35–59 (2018)
Lin, Y., Jiang, D., Xia, P.: Long-time behavior of a stochastic SIR model. Appl. Math. Comput. 236, 1–9 (2014)
Lipster, R.: A strong law of large numbers for local martingales. Stochastics 3, 217–228 (1980)
Liu, S., Pang, L., Ruan, S., et al.: Global dynamics of avian influenza epidemic models with psychological effect. Comput. Math. Meth. Med. 2015, 1–12 (2015)
Liu, S., Ruan, S., Zhang, X.: Nonlinear dynamics of avian influenza epidemic models. Math. Biosci. 283, 118–135 (2017)
Liu, Q., Jiang, D., Hayat, T., et al.: Dynamics of a stochastic predator-prey model with stage structure for Predator and Holling Type II functional response. J. Nonlinear. Sci. 28, 1151–1187 (2018)
Liu, Q., Jiang, D., Hayat, T., et al.: Stationary distribution and extinction of a stochastic HIV-1 infection model with distributed delay and logistic growth. J. Nonlinear. Sci. 30(1), 369–395 (2020)
Ma, Z., Zhou, Y., Wu, J.: Modeling and Dynamics of Infectious Diseases. Higher Education Press, Beijing (2009)
Ma, Z., Zhou, Y., Li, C.: Qualitative and Stability Methods for Ordinary Differential Equations. Science Press, Beijing (2015). ((In Chinese))
Mao, X.: Stochastic Differential Equations and Applications. Horwood Publishing, Chichester (1997)
Nguyen, D.H., Yin, G., Zhu, C.: Long-term analysis of a stochastic SIRS model with general incidence rates. SIAM J. Appl. Math. 80, 814–838 (2020)
OIE-World Organisation for Animal Health. (n.d.): Retrieved November 07 (2017), from http://www.oie.int/animal-health-in-the-world/update-onavian-influenza/
Oksendal, B.: Stochastic Differential Equations: An Introduction with Applications. Springer, Heidelberg, New York (2000)
Qi, K., Jiang, D.: The impact of virus carrier screening and actively seeking treatment on dynamical behavior of a stochastic HIV/AIDS infection model. Appl. Math. Model. 85, 378–404 (2020)
Roozen, H.: An asymptotic solution to a two-dimensional exit problem arising in population dynamics. SIAM J. Appl. Math. 49, 1793 (1989)
Shi, Z., Zhang, X., Jiang, D.: Dynamics of an avian influenza model with half-saturated incidence. Appl. Math. Comput. 355, 399–416 (2019)
Tian, X., Xu, R., Lin, J.: Mathematical analysis of a cholera infection model with vaccination strategy. Appl. Math. Comput. 361, 517–535 (2019)
Tuncer, N., Martcheva, M.: Modeling seasonality in avian influenza H5N1. J. Biol. Syst. 21(4), 1–30 (2013)
World Organization for Animal Health, Avian Influenza Portal (2013), from http://www.oie.int/en/animal-health-in-the-world/web-portal-on-avian-influenza
Zhang, X.: Global dynamics of a stochastic avian-human influenza epidemic model with logistic growth for avian population. Nonlinear Dyn. 90, 2331–2343 (2017)
Zhang, X., Yuan, R.: A stochastic chemostat model with mean-reverting Ornstein-Uhlenbeck process and Monod-Haldane response function. Appl. Math. Comput. 394, 125833 (2021)
Zhang, X., Shi, Z., Wang, Y.: Dynamics of a stochastic avian-human influenza epidemic model with mutation. Phys. A 534, 121940 (2019)
Zhao, Y., Jiang, D.: The threshold of a stochastic SIS epidemic model with vaccination. Appl. Math. Comput. 243, 718–727 (2014)
Zhou, B., Zhang, X., Jiang, D.: Dynamics and density function analysis of a stochastic SVI epidemic model with half saturated incidence rate. Chaos. Soliton. Fract. 137, 109865 (2020)
Zhou, B., Han, B., Jiang, D., et al.: Ergodic stationary distribution and extinction of a hybrid stochastic SEQIHR epidemic model with media coverage, quarantine strategies and pre-existing immunity under discrete Markov switching. Appl. Math. Comput. 410, 126388 (2021)
Zhou, B., Jiang, D., Han, B., Hayat, T.: Threshold dynamics and density function of a stochastic epidemic model with media coverage and mean-reverting Ornstein-Uhlenbeck process. Math. Comput. Simulat. 196, 15–44 (2022)
Zhu, C., Yin, G.: Asymptotic properties of hybrid diffusion systems. SIAM J. Control. Optim. 46, 1155–1179 (2007)
Acknowledgements
This work is supported by the National Natural Science Foundation of China (No. 11871473), Shandong Provincial Natural Science Foundation (No. ZR2019MA010) and the Fundamental Research Funds for the Central Universities (No. 22CX03030A).
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Additional information
Communicated by Amy Radunskaya.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix A: Proof of Theorem 3.1
Unless specifically stated, throughout Appendix A, a nonnegative \(C^2\)-function \(\mathscr {G}(x):\ \mathbb {R}_+\rightarrow \mathbb {R}_+\) is defined by
Below we divide the proof of Theorem 3.2 into two parts.
(I) (Proof for case (a) in Theorem 3.1) We define a \(C^2\)-Lyapunov function as
Calculating the derivative of \(\mathscr {F}_1(t)\) along positive solutions of system (3.1) yields
Combining the equality \(\Pi =\mu (S_0+V_0)\) and assumption \((\textbf{H}_1)\), we obtain
Using the inequality of arithmetic and geometric means, one easily derives from (A.1) that \(\mathscr {F}_1'(t)\le 0\) when \(\mathscr {R}_0<1\), and the sign holds if and only if \((S(t),V(t),I(t))=P_0\). According to the Lyapunov stability theorem (Bai et al. 2021; Tian et al. 2019), \(P_0\) is GAS if \(\mathscr {R}_0<1\).
(II) (Proof for case (b) in Theorem 3.1) We divide the proof into two steps. The first step is to verify the existence and uniqueness of \(P^*\), and the second is to obtain the global asymptotic stability of \(P^*\).
Step 1. (Existence and Uniqueness)
Combining Eq. (3.2), the existence and uniqueness of \(P^*\in \mathbb {R}_+^3\) is equivalent to the equation \(g_1(I)=0\) having a unique root on \((0,\infty )\). According to assumption \((\textbf{H}_1)\), it can be derived that \(g_1(I)\) is a monotonically decreasing function, which satisfies
implying that the equation \(g_1(I)=0\) has a unique solution \(I^*>0\) if \(\mathscr {R}_0>1\).
Step 2. (Global stability)
We define a nonnegative \(C^2\)-Lyapunov function \(\mathscr {F}_2(t)\) by
Taking the derivative of \(\mathscr {F}_2(t)\) along the solution of system (3.1), one has
Using the equality \(\mu +\alpha =\frac{\beta _1S^*}{f_1(I^*)}+\frac{\beta _2V^*}{f_2(I^*)}\). (A.2) can then be rewritten as
where
This together with assumption \((\textbf{H}_2)\) and the Lagrange mean value theorem yields:
where \(\xi _j\) and \(\eta _j\) are both between I and \(I^*\), \(j=1,2\).
According to the inequality of arithmetic and geometric means, we combine (A.3–A.4) to determine that \(\mathscr {F}_2'(t)\le 0\) if \(\mathscr {R}_0>1\), and the sign holds if and only if \((S(t),V(t),I(t))=P^*\). By the Lyapunov stability theorem, \(P^*\) is GAS when \(\mathscr {R}_0>1\). This completes the proof.
Appendix B: Proof of Theorem 3.2
We divide the proof of Theorem 3.2 into two steps.
Step 1 (Proof for case (a) in Theorem 3.2) The Jacobi matrix of system (1.2) at the disease-free equilibrium \(T_0\) is:
Letting \(|\lambda \textbf{1}_5-J_a(T_0)|=0\), we have the following characteristic equation
It is clear to see that Eq. (B.1) has five real roots, which include \(\lambda _1=\lambda _2=-\mu \), \(\lambda _3=-\mu _h\), \(\lambda _4=-(\mu +\alpha _h+\delta )\) and \( \lambda _5=(\mu +\alpha )(\mathscr {R}_0-1)\). When \(\mathscr {R}_0<1\), all of the eigenvalues of \(J_a(T_0)\) have negative real parts. Using Definition 2.1 and the Routh–Hurwitz criterion (Ma et al. 2015), we determine that \(J_a(T_0)\in \overline{RH}(5)\) and \(T_0\) is LAS. Note that \(\lambda _5>0\) if \(\mathscr {R}_0>1\); thus, \(T_0\) is unstable if \(\mathscr {R}_0>1\).
Step 2 (Proof for case (b) in Theorem 3.2) The Jacobi matrix of system (1.2) at the equilibrium \(T^*\) is:
where
Below we define
On the one hand, by assumptions \((\textbf{H}_1)\)-\((\textbf{H}_2)\), we obtain that \(f_j'(x)\ge 0\) and \((\frac{x}{f_k(x)})'=\frac{f_k(x)-xf_k'(x)}{(f_k(x))^2}\ge 0\) for any \(x\ge 0;j=1,2,3;k=1,2\). Thus, \((b_{11},b_{13},b_{22},b_{23},b_{31},b_{32},\overline{b}_{11},\overline{b}_{12},\overline{b}_{21},\overline{b}_{22})\in \mathbb {R}_+^{10}\). Combining the equality \(\frac{\beta _1S^*}{f_1(I^*)}+\frac{\beta _2V^*}{f_2(I^*)}=\mu +\alpha \) and assumption \((\textbf{H}_1)\), we have
By direct calculation, the characteristic polynomial of \(Q_1\) is
where \(b_1=b_{11}+b_{22}+b_{33}\), \( b_2=b_{11}(b_{22}+b_{33})+(b_{22}b_{33}+b_{23}b_{32})+b_{13}b_{31}\) and \( b_3=b_{11}(b_{22}b_{33}+b_{23}b_{32})+b_{13}b_{22}b_{31}\). Clearly, \(b_i>0\ (\forall \ i=1,2,3)\). Moreover,
Using Definition 2.1, we have \(Q_1\in \overline{RH}(3)\).
On the other hand, we calculate that \(\psi _{Q_2}(\lambda )=\lambda ^2+\overline{b}_1\lambda +\overline{b}_2\), where \(\overline{b}_1=\overline{b}_{11}+\overline{b}_{22}>0\), and
Thus, \(Q_2\in \overline{RH}(2)\).
According to (B.2) and the form of \(J_a(T^*)\), we obtain
This implies that all of the eigenvalues of \(J_a(T^*)\) have negative real parts, i.e., \(J_a(T^*)\in \overline{RH}(5)\). Hence, \(T^*\) is LAS when \(\mathscr {R}_0>1\). This completes the proof.
Appendix C
We present some preliminaries of SDE in part (I). In part (II), we prove that \(\mathscr {L}U_0(S,V,I,S_h,I_h)\le -1\) for any \((S,V,I,S_h,I_h)\in \bigcup _{j=1}^{10}\mathbb {D}_{\epsilon ,j}^c\).
(I) (Preliminaries of SDE) It is assumed that \(B_c(t)\) is an n-dimensional standard Brownian motion defined on the complete probability space \( \{\Omega ,\varGamma ,\{\varGamma _t\}_{t\ge 0},\mathbb {P}\} \). Let Z(t) be the solution of the following SDE,
with initial value \( Z(0)\in \mathbb {R}^n \).
The Itô’s differential operator \( \mathscr {L}\) (Mao 1997) is given by:
Let \(C^{2,1}(\mathbb {R}^n\times \mathbb {R}_+;\mathbb {R})\) be the space of all real-valued functions V(z, t) on \(\mathbb {R}^n\times \mathbb {R}_+\), which are twice continuously differentiable with respect to z and continuously differentiable with respect to t. We can then derive that
where \( V_t=\frac{\partial V}{\partial t}\), \( V_Z=(\frac{\partial V}{\partial z_1},...,\frac{\partial V}{\partial z_n})\) and \( V_{ZZ}=(\frac{\partial ^2V}{\partial z_i \partial z_j})_{n\times n} \). Then the Itô’s formula is
(II) We divide the proof of (4.15) into the following seven cases.
Case 1 If \((S,V,I,S_h,I_h)\in \bigcup _{j=1}^{4}\mathbb {D}_{\epsilon ,j}^c\), by (4.9–4.10), we obtain
Case 2 If \((S,V,I,S_h,I_h)\in \mathbb {D}_{\epsilon ,5}^c\), combining (4.9) and (4.11), we have
Case 3 If \((S,V,I,S_h,I_h)\in \mathbb {D}_{\epsilon ,6}^c\), by (4.9) and (4.12), we have
Case 4 If \((S,V,I,S_h,I_h)\in \mathbb {D}_{\epsilon ,7}^c\), in view of (4.9) and (4.12), we obtain
Case 5 If \((S,V,I,S_h,I_h)\in \mathbb {D}_{\epsilon ,8}^c\), by (4.9) and (4.12), we obtain
Case 6 For any \((S,V,I,S_h,I_h)\in \mathbb {D}_{\epsilon ,9}^c\), combining (4.9) and (4.13), we have
Case 7 If \((S,V,I,S_h,I_h)\in \mathbb {D}_{\epsilon ,10}^c\), in view of (4.9) and (4.14), we obtain
In summary, for a sufficiently small \(\epsilon \) satisfying (4.10–4.14),
This completes the proof of (4.15).
Appendix D: Proof of Lemma 2.4
For simplicity, for the same dimensional symmetric matrices \(Q_1\) and \(Q_2\), we define
In this sense, \(Q_1\succ \textbf{0}\) if \(Q_2\succ \textbf{0}\) and \(Q_1\succeq Q_2\). According to the theory of matrix algebra, we list two basic results, which include that (i) the positive definiteness of a real symmetric matrix is not affected by congruence transformation, and (ii) the similarity transformation does not change the eigenvalues of the matrix. Thus, for any invertible matrix P, it is easily derived that \(P\Sigma P^{\tau }\succ \textbf{0}\) if \(\Sigma \succ \textbf{0}\), and \(PA_0P^{-1}\in \overline{RH}(n)\) if \(A_0\in \overline{RH}(n)\).
We define \(\Sigma _k\) as the solutions of the following algebraic equations
where \(M_1=\textrm{diag}\{1,0,0,0,0\}\), \( M_2=\textrm{diag}\{0,1,0,0,0\}\), \(M_3=\textrm{diag}\{0,0,1,0,0\}\), \(M_4=\textrm{diag}\{0,0,0,1,0\}\) and \( M_5=\textrm{diag}\{0,0,0,0,1\}\).
Using the finite independent superposition principle, we have \(\Sigma _0=\sum _{j=1}^{5}\rho _j\Sigma _j\). The proof of \(\Sigma _0\succ \textbf{0}\) is divided into two steps. The first step is to prove that there is a positive constant \(\xi _1\) such that \(\Sigma _1\succeq \textrm{diag} \{\xi _1,0,0,0,0\}\), and the second is to find a matrix \(Q_0\succ \textbf{0}\) satisfying \(\Sigma _0\succeq Q_0\).
Step 1 For the algebraic equation
after which the related proof can be divided into the following two conditions:
Case 1 If \((\mathscr {A}_1)\) is satisfied, an intuitive calculation shows that
By letting the matrix \(F_1{:=}(a_{ij})_{\{2\le i,j\le 4\}}\), the characteristic polynomial of A is that \(\psi _A(\lambda )=(\lambda -a_{11})|\lambda \mathcal {I}_4-F_1|\). Clearly, A has an eigenvalue \(\lambda =a_{11}\). Combining \(A\in \overline{RH}(5)\) and \(a_{11}\in \mathbb {R}\), we have \(a_{11}<0\), implying that \(\varTheta _{11}\succeq \textbf{0}\) and
Case 2 If \((\mathscr {B}_1)\) is satisfied, then \( a_{21}\ne 0\) or \(a_{31}\ne 0\) or \(a_{41}\ne 0\) or \(a_{51}\ne 0\). Below we need to illustrate that the elements \(a_{j1}\ (j=2,3,4,5)\) have the equivalent status in A. Let \({\widetilde{A}}=J_1AJ_1^{-1}{:=}(\widetilde{a}_{ij})_{5\times 5}\), \({\widehat{A}}=J_2AJ_2^{-1}{:=}(\widehat{a}_{ij})_{5\times 5}\), \(\overline{A}=J_3AJ_3^{-1}{:=}(\overline{a}_{ij})_{5\times 5}\), \( \widetilde{\Sigma }_1=J_1\Sigma _1J_1^{\tau }\), \(\widehat{\Sigma }_1=J_2\Sigma _1J_2^{\tau }\) and \(\overline{\Sigma }_1=J_3\Sigma _1J_3^{\tau }\), where the invertible matrices \(J_1,J_2\) and \(J_3\) take the form
Equation (D.1) can then be equivalently transformed into the following three forms:
It is clear to see that (i) \(J_iM_1J_i^{\tau }=M_1\ (\forall \ i=1,2,3)\), (ii) \(\Sigma _1\), \(\widetilde{\Sigma }_1\), \(\widehat{\Sigma }_1\) and \(\overline{\Sigma }_1\) have the same positive definiteness, and (iii) \({\widetilde{A}}\), \({\widehat{A}}\), \(\overline{A}\in \overline{RH}(5)\). Moreover, \(\widetilde{a}_{21}=a_{31}\), \(\widehat{a}_{21}=a_{41}\) and \(\overline{a}_{21}=a_{51}\). Thus, we can use the similarity transformation to transform \(a_{31}\) (or \(a_{41},\ a_{51}\)) into the position of the first element of the second line of A. That is, we only need to consider the case \(a_{21}\ne 0\), which is equivalent to \((\mathscr {B}_1)\).
Let \(B=J_4AJ_4^{-1}{:=}(b_{ij})_{5\times 5}\), where \(J_4\) is called the first elimination matrix. Direct calculation shows that
where \(b_{i1}=0\ (\forall \ i=3,4,5)\). In view of \(J_4M_1J_4^{\tau }=M_1\) and \(b_{21}=a_{21}(\ne 0)\), Eq. (D.1) can then be equivalently rewritten as:
Below we similarly consider two cases of the parameters \((b_{32},b_{42},b_{52})\):
Case 2-1. If \((\mathscr {A}_2)\) is satisfied, we calculate that
where the symmetric matrix \(W_0=(\phi _{ij})_{2\times 2}\) with
Let the matrix \(F_2=(b_{ij})_{\{3\le i,j\le 5\}}\); then, the characteristic polynomial of A is:
Since \(A\in \overline{RH}(5)\), all of the roots of the equation \(\lambda ^2-(b_{11}+b_{22})\lambda +(b_{11}b_{22}-b_{12}b_{21})=0\) have negative real part. Using Definition 2.1, we obtain that \(b_{11}+b_{22}<0\) and \(b_{11}b_{22}-b_{12}b_{21}>0\). Combined with \( b_{21}\ne 0\), we have
This implies that \(W_0\succ \textbf{0}\) and \(J_4\Sigma _1J_4^{\tau }\succeq \textbf{0}\).
Define two positive semi-definite matrices \(\varTheta _{12}\) and \(\widetilde{\varTheta }_{12}\) by
Then,
Using \(\widetilde{\varTheta }_{12}\succeq \textbf{0}\), we obtain that \(J_4^{-1}\widetilde{\varTheta }_{12}(J_4^{-1})^{\tau }\succeq \textbf{0}\) and
Case 2-2. If \((\mathscr {B}_2)\) is satisfied, then \( b_{32}\ne 0\) or \(b_{42}\ne 0\) or \(b_{52}\ne 0\). We define \( \widetilde{B}=J_5BJ_5^{-1}\), \(\overline{B}=J_6BJ_6^{-1}\), \(\widetilde{\Sigma }_1=J_5\Sigma _1J_5^{-1}\) and \( \overline{\Sigma }_1=J_6\Sigma _1J_6^{-1}\), where \(J_5\) and \(J_6\) are both invertible matrices, and they are
We can then equivalently transform (D.3) into the following algebraic equations
It can be noticed that \(J_5M_1J_5^{\tau }=J_6M_1J_6^{\tau }=M_1\). By a method similar to that of Case 2, we can determine that the elements \(b_{j2}\ (j=3,4,5)\) have the equivalent status in B. Below we only discuss the case \(b_{32}\ne 0\), which is equivalent to \((\mathscr {B}_2)\).
Let \(C=J_7BJ_7^{-1}{:=}(c_{ij})_{5\times 5}\), where \(J_7\) is called the second elimination matrix. By direct calculation, C and the invertible matrix \(J_7\) are as follows:
where \(c_{i1}=c_{j2}=0\ (\forall \ i=3,4,5;j=4,5)\). Note that \(c_{21}=b_{21}(=a_{21}\ne 0)\), \( c_{32}=b_{32}(\ne 0)\) and \((J_7J_4)M_1(J_7J_4)^{\tau }=M_1\), Eq. (D.3) (or (D.1)) is then equivalently transformed into
Similar to cases \((\mathscr {A}_i)\) and \((\mathscr {B}_i)\), \(i=1,2\), the analysis of Eq. (D.5) can be divided into the following two cases:
Case 2-2-1. If \((\mathscr {A}_3)\) is satisfied, by defining
then the characteristic polynomial of A is
where \(c_1=-(c_{11}+c_{22}+c_{33})\), \(c_2=c_{11}(c_{22}+c_{33})+c_{22}c_{33}-c_{23}c_{32}-c_{12}c_{21}\) and \(c_3=c_{11}(c_{23}c_{32}-c_{22}c_{33})+c_{21}(c_{12}c_{33}-c_{13}c_{32})\). Using \(A\in \overline{RH}(5)\), all of the root of equation \(\lambda ^3+c_1\lambda ^2+c_2\lambda +c_3=0\) has three negative real part. By Definition 2.1, one has
Define two nonsingular transformed matrices \(P_1\) and \(J_8\) by
Let the matrices \( F_7=P_1F_6{:=}(\overline{c}_{ij})_{\{1\le i\le 3,4\le j\le 5\}}\) and \(\overline{C}=J_8CJ_8^{-1}\). Note that \(J_8^{-1}=\left( \begin{array}{cc} P_1^{-1}&{} \mathbb {O}_{3,2} \\ \mathbb {O}_{2,3}&{} \textbf{1}_2 \end{array} \right) \), we obtain that \((J_8J_7J_4)M_1(J_8J_7J_4)^{\tau }=\textrm{diag}\{c_{21}^2c_{32}^2,0,0,0,0\}=(c_{21}c_{32})^2M_1\) and
Thus, Eq. (D.5) (or (D.1)) can be equivalently rewritten as
The solution of Eq. (D.8) is unique, and it satisfies
Combining (D.7) and Han et al. (2020), it is easily derived that \(F_8\succeq \textbf{0}\). Consider the following two positive semi-definite matrices
we obtain
Clearly, \((c_{21}c_{32})^2(J_8J_7J_4)^{-1}\widetilde{\varTheta }_{13}[(J_8J_7J_4)^{-1}]^{\tau }\succeq \textbf{0}\). This implies that
Case 2-2-2. When \((\mathscr {B}_3)\) is satisfied, by the method similar to those of Case 2 and Case 2-2, we obtain that \(c_{43}\ne 0\) is equivalent to the condition \(c_{53}\ne 0\). Thus, we only need to analyze the case \(c_{43}\ne 0\).
Let \(D=J_9CJ_9^{-1}{:=}(d_{ij})_{5\times 5}\), where \(J_9\) is called the third elimination matrix. By direct calculation, D and the invertible matrix \(J_9\) are obtained by
where \(d_{i1}=d_{k2}=d_{43}=0\ (\forall \ i=3,4,5;k=4,5)\). In view of \((J_9J_7J_4)M_1(J_9J_7J_4)^{\tau }=M_1\), \(d_{21}=c_{21}(=a_{21}\ne 0)\), \( d_{32}=c_{32}(=b_{32}\ne 0)\) and \(d_{43}=c_{43}(\ne 0)\). Hence, Eq. (D.5) (or (D.1)) can be equivalently transformed into
Based on the value of \(d_{54}\), we analyze the following two cases:
Case 2-2-2-1. If \(d_{54}=0\), for simplicity, we define \(D=\left( \begin{array}{cc} F_9&{} F_{10} \\ \mathbb {O}_{1,4}&{} d_{55} \end{array} \right) \), where
Then, the characteristic polynomial of A is
where \(d_1=-(d_{11}+d_{22}+d_{33}+d_{44})\), \(d_2=d_{11}(d_{22}+d_{33}+d_{44})+d_{22}(d_{33}+d_{44})+d_{33}d_{44}-d_{34}d_{43}-d_{12}d_{21}-d_{23}d_{32}\), \(d_3=d_{22}(d_{34}d_{43}-d_{33}d_{44})+d_{32}(d_{23}d_{44}-d_{24}d_{43})-d_{11}[d_{33}d_{44}-d_{34}d_{43}+d_{22}(d_{33}+d_{44})-d_{23}d_{32}]+d_{12}d_{21}(d_{33}+d_{44})-d_{13}d_{21}d_{32}\) and \(d_4=d_{21}d_{32}(d_{13}d_{44}-d_{14}d_{43})-d_{12}d_{21}(d_{33}d_{44}-d_{34}d_{43})-d_{11}[d_{22}(d_{34}d_{43}-d_{33}d_{44})+d_{32}(d_{23}d_{44}-d_{24}d_{43})]\).
Combining Definition 2.1 and \(A\in \overline{RH}(5)\), we obtain that the equation \(\lambda ^4+d_1\lambda ^3+d_2\lambda ^2+d_3\lambda +d_4=0\) has roots with all negative real components, i.e.,
Define two invertible transformed matrices \(P_2\) and \(J_{10}\) by
where \(m_1=d_{21}d_{32}d_{43}\ne 0\), \(m_2=(d_{22}+d_{33}+d_{44})d_{32}d_{43}\), \(m_3=d_{43}(d_{23}d_{32}+d_{34}d_{43}+d_{33}d_{44}+d_{33}^2+d_{44}^2)\) and \(m_4=d_{24}d_{32}d_{43}+(d_{33}+d_{44})d_{34}d_{43}+(d_{34}d_{43}+d_{44}^2)d_{44}\). By choosing \(F_{11}=P_2F_{10}{:=}(\overline{d}_{i5})_{\{1\le i\le 4\}}\) and \(\overline{D}=J_{10}DJ_{10}^{-1}\), we get that \((J_{10}J_9J_7J_4)M_1(J_{10}J_9J_7J_4)^{\tau }=\textrm{diag}\{m_1^2,0,0,0,0\}=m_1^2M_1\) and
Thus, Eq. (D.10) (or (D.1)) is equivalently rewritten as:
By letting \(F_{12}=\frac{1}{m_1^2}(J_{10}J_9J_7J_4)\Sigma _1(J_{10}J_9J_7J_4)^{\tau }\), direct calculation shows that the solution of Eq. (D.12) is unique, and it satisfies
We define \(\varTheta _{14}=\textrm{diag}\{\frac{1}{2d_1},0,0,0,0\}\) and \(\widetilde{\varTheta }_{14}=F_{12}-\varTheta _{14}\). Note that the determinants of all principal submatrices of \(\widetilde{\varTheta }_{14}\) are nonnegative; then, \(\widetilde{\varTheta }_{14}\succeq \textbf{0}\). Moreover,
This implies that
Case 2-2-2-2. If \((\mathscr {B}_4)\) is satisfied, we construct a nonsingular transformed matrix
where \(\nu _2=d_{32}d_{43}d_{54}\), \(\nu _3=d_{43}d_{54}(d_{33}+d_{44}+d_{55})\), \(\nu _4=d_{54}(d_{34}d_{43}+d_{45}d_{54}+d_{44}d_{55}+d_{44}^2+d_{55}^2)\), \(\nu _5=d_{35}d_{43}d_{54}+d_{45}d_{54}(d_{44}+2d_{55})+d_{55}^3\), \(\gamma _1=d_{21}d_{32}d_{43}d_{54}\ne 0\), \(\gamma _2=d_{22}\nu _2+d_{32}\nu _3\), \(\gamma _3=d_{23}\nu _2+d_{33}\nu _3+d_{43}\nu _4\), \(\gamma _4=d_{24}\nu _2+d_{34}\nu _3+d_{44}\nu _4+d_{54}\nu _5\) and \(\gamma _5=d_{25}\nu _2+d_{35}\nu _3+d_{45}\nu _4+d_{55}\nu _5\).
Direct calculation shows that \((P_3J_9J_7J_4)M_1(P_3J_9J_7J_4)^{\tau }=\textrm{diag}\{\gamma _1^2,0,0,0,0\}=\gamma _1^2M_1\). Thus, Eq. (D.10) (or (D.1)) can be equivalently transformed into
where \({\widetilde{D}}=P_3DP_3^{-1}\). For convenience, let \(F_{12}{:=}\frac{1}{\gamma _1^2}(P_3J_9J_7J_4)\Sigma _1(P_3J_9J_7J_4)^{\tau }\), and the characteristic polynomial of A is assumed as
According to (2.1–2.2) in Definition 2.1, the sufficient and necessary conditions of \(A\in \overline{RH}(5)\) are
That is,
Using (D.15), it can be derived that
This together with \(\psi _{{\widetilde{D}}}(\lambda )=\psi _A(\lambda )\) and Eq. (D.14) yields that
where \(\theta _{11}=\frac{a_2(a_3a_4-a_2a_5)-a_4(a_1a_4-a_5)}{2\varDelta _4}\), \(\theta _{22}=\frac{a_3a_4-a_2a_5}{2\varDelta _4}\), \(\theta _{33}=\frac{a_1a_4-a_5}{2\varDelta _4}\), \(\theta _{44}=\frac{\varDelta _2}{2\varDelta _4}\) and \(\theta _{55}=\frac{\varDelta _3}{2a_5\varDelta _4}\).
Similar to the relevant description of Han et al. (2020), a matrix \(A_0\) is called standard \(\textbf{R}_5\) matrix if there exists an invertible matrix \(H_0\) satisfying \(H_0AH_0^{-1}={\widetilde{D}}\). Clearly, A under Case 2-2-2-2 is a standard \(\textbf{R}_5\) matrix.
By (D.14–D.16), one obtains that \(\theta _{ii}>0\) for any \(i=2,3,4,5\). Below we define a constant
and a real symmetric matrix
By direct calculation, we have
which implies that \(\overline{\theta }_{11}=\frac{1}{\theta _{55}}[(\overline{\theta }_{11}\theta _{55}-\theta _{33}^2)+\theta _{33}^2]>\frac{1}{\vartheta _{55}}[\frac{a_3^2}{4a_1a_5\varDelta _4}+\theta _{33}^2]>0\). In view of (D.18), we obtain
Based on the theory of matrix algebra, \(L_0\succeq \textbf{0}\) if and only if the determinants of its principal submatrices are all nonnegative. To proceed, let \(L_0(k_1,k_2,...,k_i)\) be the principal submatrix constructed by the same rows and columns \(k_j\ (j=1,2,..,i)\) of \(L_0\). Using (D.18–D.19), we obtain that the determinants of the first-order and second-order principal submatrices of \(L_0\) are all positive. Furthermore,
(i). The determinants of the third-order principal submatrices of \(L_0\):
(ii). The determinants of the fourth order principal submatrices of \(L_0\):
(iii). The determinant of \(L_0\):
Thus, the proof of \(L_0\succeq \textbf{0}\) is completed.
We consider an anti-diagonal matrix \( J_0 \) and a symmetric matrix \( \widetilde{\varTheta }_{15}\), which take the form
Clearly, \(\widetilde{\varTheta }_{15}\succeq \textbf{0}\). Let \(\varTheta _{15}=\textrm{diag}\{\frac{1}{2a_1},0,0,0,0\}\), then \(F_{12}=\varTheta _{15}+\widetilde{\varTheta }_{15}\) and
Thus,
According to (D.2), (D.4), (D.9), (D.13) and (D.20), a constant \(\xi _1>0\) always exists such that
Step 2. For the following four algebraic equations,
We give four invertible transformation matrices \(L_j\ (j=2,3,4,5)\), where \(L_3=L_2^2\), \(L_4=L_2^3\), \(L_5=L_2^4\) with
Equations (D.22) can then be equivalently rewritten as
Direct calculation shows that \(L_iM_iL_i^{\tau }=M_1\) for any \(i=2,3,4,5\). By defining \(A_i=L_iAL_i^{-1}\) and \(\overline{\Sigma }_i=L_i\Sigma _iL_i^{\tau }\), Eqs. (D.23) can then be rewrritten as
Since \(A\in \overline{RH}(5)\), we determine that \(A_k\in \overline{RH}(5)\) for any \(k=2,3,4,5\). By a method similar to that of Eq. (D.1) in Step 1, one can obtain that there are four positive numbers \(\xi _i>0\ (i=2,3,4,5)\) satisfying
which means that
Let \(Q_0{:=}\textrm{diag}\{\rho _1\xi _1,\rho _2\xi _2,\rho _3\xi _3,\rho _4\xi _4,\rho _5\xi _5\}\). Combining (D.21) and (D.24), we have
Thus, \(\Sigma _0\succ \textbf{0}\). Moreover, the uniqueness and expression of \(\Sigma _0\) can be obtained by the above analysis. This completes the proof of Lemma 2.4. For completeness, an important corollary is supplemented here. Let \(|F_{12}^{(k)}|\) be the k-th leading principal minor of \(F_{12}\), \(k=1,2,...,5\). Combining \(\theta _{11}-\overline{\theta }_{11}=\frac{1}{2a_1}>0\), we calculate that
Hence, \(F_{12}\succ \textbf{0}\). In fact, as shown in (D.21) and (D.24), it is evident that \(\Sigma _i\succeq \textbf{0}\) for any \(i=1,2,...,5\). If at least one of the five matrices \(A{:=}A_1\) and \(A_j\ (j=2,3,4,5)\) is a standard \(\textbf{R}_5\) matrix, then there is an invertible matrix \(H_j\) such that \(\Sigma _j=H_jF_{12}H_j^{\tau }\), implying that \(\Sigma _0\succ \textbf{0}\). Thus, the condition \(M_c=\textrm{diag}\{\rho _1,\rho _2,\rho _3,\rho _4,\rho _5\}\) is only a sufficient condition determining \(\Sigma _0\succ \textbf{0}\), but is not close to a necessary condition.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Zhou, B., Jiang, D., Dai, Y. et al. Threshold Dynamics and Probability Density Function of a Stochastic Avian Influenza Epidemic Model with Nonlinear Incidence Rate and Psychological Effect. J Nonlinear Sci 33, 29 (2023). https://doi.org/10.1007/s00332-022-09885-8
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00332-022-09885-8
Keywords
- Stochastic avian influenza model
- Extinction
- Stationary distribution
- Fokker–Planck equation
- Density function