Research article

Circulation mechanisms responsible for wet or dry summers over Zimbabwe

  • Climate change has resulted in increased rainfall variability over many parts of the world including Southern Africa. As such, droughts and floods have become a frequent phenomenon in Zimbabwe and have potential to intensify socio-economic stressors. This study examined possible forcing factors behind the occurrence of extreme summer events using re-analysis datasets. Composite analysis and correlation methods were used to identify circulation mechanisms and their strength in determining rainfall patterns in Zimbabwe. Predominantly northerly airflow in the lower troposphere was found to favor wet while southerly airflow favors dry seasons. Negative geopotential anomalies (minimum of −20 hPa) to the west of Zimbabwe in the middle levels characterize wet summers which swing to positive anomalies (+24 hPa) during dry summers. Positive SST anomalies (maximum of 0.4) exist to the southwest of Madagascar extending to the western shore on the Angola-Namibian border characterize wet summers which swing to negative anomalies (−0.2 ºC) during dry summer seasons. SST anomalies in the South western Indian and South eastern Atlantic oceans are crucial in the determination of the strength of both the South Indian and Atlantic Ocean high pressure systems which in turn control moisture advection and convergence into Zimbabwe during the summer period. If these SST anomalies at lag times of about 3 months can be used to predict the incoming summer circulation patterns then the accuracy of summer seasonal outlook forecasts can be improved. Studying the mechanisms behind drought and flood occurrence is important to the country which is in the process of downscaling regional prediction products to improve the accuracy of seasonal forecasts. These findings are useful in crafting relevant measures to maximize the benefits and minimize the risks of extreme rainfall events.

    Citation: Moven Manjowe, Terence Darlington Mushore, Juliet Gwenzi, Collen Mutasa, Electdom Matandirotya, Emmanuel Mashonjowa. Circulation mechanisms responsible for wet or dry summers over Zimbabwe[J]. AIMS Environmental Science, 2018, 5(3): 154-172. doi: 10.3934/environsci.2018.3.154

    Related Papers:

    [1] Bin Zhen, Ya-Lan Li, Li-Jun Pei, Li-Jun Ouyang . The approximate lag and anticipating synchronization between two unidirectionally coupled Hindmarsh-Rose neurons with uncertain parameters. Electronic Research Archive, 2024, 32(10): 5557-5576. doi: 10.3934/era.2024257
    [2] Jun Guo, Yanchao Shi, Weihua Luo, Yanzhao Cheng, Shengye Wang . Exponential projective synchronization analysis for quaternion-valued memristor-based neural networks with time delays. Electronic Research Archive, 2023, 31(9): 5609-5631. doi: 10.3934/era.2023285
    [3] Xinling Li, Xueli Qin, Zhiwei Wan, Weipeng Tai . Chaos synchronization of stochastic time-delay Lur'e systems: An asynchronous and adaptive event-triggered control approach. Electronic Research Archive, 2023, 31(9): 5589-5608. doi: 10.3934/era.2023284
    [4] Caiwen Chen, Tianxiu Lu, Ping Gao . Chaotic performance and circuitry implement of piecewise logistic-like mapping. Electronic Research Archive, 2025, 33(1): 102-120. doi: 10.3934/era.2025006
    [5] Janarthanan Ramadoss, Asma Alharbi, Karthikeyan Rajagopal, Salah Boulaaras . A fractional-order discrete memristor neuron model: Nodal and network dynamics. Electronic Research Archive, 2022, 30(11): 3977-3992. doi: 10.3934/era.2022202
    [6] Huijun Xiong, Chao Yang, Wenhao Li . Fixed-time synchronization problem of coupled delayed discontinuous neural networks via indefinite derivative method. Electronic Research Archive, 2023, 31(3): 1625-1640. doi: 10.3934/era.2023084
    [7] Dandan Zuo, Wansheng Wang, Lulu Zhang, Jing Han, Ling Chen . Non-fragile sampled-data control for synchronizing Markov jump Lur'e systems with time-variant delay. Electronic Research Archive, 2024, 32(7): 4632-4658. doi: 10.3934/era.2024211
    [8] Jun Guo, Yanchao Shi, Shengye Wang . Synchronization analysis of delayed quaternion-valued memristor-based neural networks by a direct analytical approach. Electronic Research Archive, 2024, 32(5): 3377-3395. doi: 10.3934/era.2024156
    [9] Wanshun Zhao, Kelin Li, Yanchao Shi . Exponential synchronization of neural networks with mixed delays under impulsive control. Electronic Research Archive, 2024, 32(9): 5287-5305. doi: 10.3934/era.2024244
    [10] Yu-Jing Shi, Yan Ma . Finite/fixed-time synchronization for complex networks via quantized adaptive control. Electronic Research Archive, 2021, 29(2): 2047-2061. doi: 10.3934/era.2020104
  • Climate change has resulted in increased rainfall variability over many parts of the world including Southern Africa. As such, droughts and floods have become a frequent phenomenon in Zimbabwe and have potential to intensify socio-economic stressors. This study examined possible forcing factors behind the occurrence of extreme summer events using re-analysis datasets. Composite analysis and correlation methods were used to identify circulation mechanisms and their strength in determining rainfall patterns in Zimbabwe. Predominantly northerly airflow in the lower troposphere was found to favor wet while southerly airflow favors dry seasons. Negative geopotential anomalies (minimum of −20 hPa) to the west of Zimbabwe in the middle levels characterize wet summers which swing to positive anomalies (+24 hPa) during dry summers. Positive SST anomalies (maximum of 0.4) exist to the southwest of Madagascar extending to the western shore on the Angola-Namibian border characterize wet summers which swing to negative anomalies (−0.2 ºC) during dry summer seasons. SST anomalies in the South western Indian and South eastern Atlantic oceans are crucial in the determination of the strength of both the South Indian and Atlantic Ocean high pressure systems which in turn control moisture advection and convergence into Zimbabwe during the summer period. If these SST anomalies at lag times of about 3 months can be used to predict the incoming summer circulation patterns then the accuracy of summer seasonal outlook forecasts can be improved. Studying the mechanisms behind drought and flood occurrence is important to the country which is in the process of downscaling regional prediction products to improve the accuracy of seasonal forecasts. These findings are useful in crafting relevant measures to maximize the benefits and minimize the risks of extreme rainfall events.


    The Lyapunov exponent (LE), as introduced by Oseledets [1] in the context of his multiplicative ergodic theorem, serves as a quantifier of the divergence between two proximal trajectories over time within a dynamical system. In an n-dimensional dynamical system, there are n LEs available, with their quantity corresponding to the dimensionality of the system's phase space. Each LE characterizes the rate of convergence or divergence of the system's attractor in a specific direction. The spectrum of LEs offers a metric for assessing the local sensitivity of a system to initial conditions, as well as for providing crucial insights into the system's global dynamics. This spectrum facilitates the effective description and classification of system attractors based on their LEs. For stable equilibrium points, all LEs are real and negative; in the case of stable limit cycles, one LE is zero while the remainders are real and negative. An attractor is identified as a k-torus if the first k LEs are zero and the others are negative. The presence of positive values within the LEs spectrum signifies a chaotic attractor, which is further classified as hyperchaotic if two or more LEs are positive [2].

    The calculation or estimation of LEs constitutes a core aspect of research into nonlinear dynamical systems. Notably, the largest LE (LLE) holds particular significance as it directly influences the predictability of the system in question. Over the past several decades, a diverse array of studies focusing on the calculation schemes for LEs, especially the LLE, have been published. These studies have predominantly categorized them into two main methodologies: the determination of LEs from governing equations [3,4,5,6,7] and the estimation of LEs from time series data [8,9,10,11]. Benettin et al. [3] first introduced the method for calculating all LEs of dynamical systems, grounded in Oseledets's theory [1]. This methodology was later refined by Wolf et al. [8]. Additionally, Briggs [12] explored the calculation of LEs by using experimental data, proposing that the optimal estimation of Jacobian matrices in the presence of noisy data is attained through least-squares polynomial fitting. The QR decomposition (short for "QR factorization" is a process that decomposes a matrix into the product of an orthogonal matrix (Q) and an upper triangular matrix (R)) and the singular value decomposition (SVD) methods for determining all LEs of dynamical systems were developed by Von Bremen et al. [13] and Dieci and Elia [14], respectively. Dabrowski [15] numerically derived the LLE by calculating the LE in the direction of a disturbance, as based on the perturbation vector and its derivative's dot product. Liao [16] investigated the sensitivity gradients of the LLE in dynamical systems to address the deficiency of previous approaches, which were prohibitively time-consuming and resource-intensive for most optimization problems of reasonable size. Certain methods described above embody the direct approach, which quantifies the divergence growth rate between two trajectories that have an infinitesimal discrepancy in their initial conditions. Conversely, other methods aim to estimate the Jacobian matrices of systems, addressing the limitations inherent to direct approaches, especially in the context of noise [17]. However, approaches that rely on linear approximation often fall short of capturing nonlinear growth and typically necessitate laborious computations. In some instances, they may even result in erroneous LEs due to ill-conditioned Jacobian matrices [18]. Additionally, if the number of iterations is insufficient, the outcomes are likely to be imprecise [19]. Zhou et al. [20] introduced a groundbreaking method to derive the LLE by using two nearby pseudo-orbits, eliminating the need for phase space reconstruction and Jacobian matrix computation. Utilizing machine learning to forecast LEs from data was an approach adopted by Pathak et al. [21] and McAllister et al. [22]. The perturbation vectors method [23], the cloned dynamics approach [24], and the synchronization technique [25,26,27] have been developed as strategies to circumvent the direct calculation of Jacobian matrices or the need to solve variational equations. Recent advancements and applications of LLEs are documented in references [28].

    The analytical estimation of LEs within a dynamical system represents a compelling and significant topic. According to the definition of LEs, their analytical formulas are readily derivable only for steady-state solutions, including fixed points in nonlinear systems with a limited number of degrees of freedom and steady-state, spatially homogeneous solutions in spatially extended systems. Analytical expressions for LEs have been derived for neural oscillator models [29,30,31], as well as for certain simple models of nonlinear oscillators applied for synchronization problems triggered by common external noise [32,33,34]. Caponetto and Fazzino [7] introduced a semi-analytical approach, utilizing the differential transform method, to compute LEs in fractional order systems. Hramov et al. [35] first presented the analytical formula for the zero LE. A zero LE exists within the spectrum of LEs for flow systems, characterizing the perturbation evolution along the phase trajectory [36]. To derive the analytical expression for the zero LE, Hramov et al. analyzed a model system that describes the behavior of a driven periodic oscillator with noise near the synchronization onset. In chaotic systems, analytical approximations of LEs typically retain validity within a very narrow range of control parameter values, despite their derivability [37]. The analytical characterization of LEs for chaotic oscillators remains to be a formidable challenge.

    In this paper, we aim to derive an analytical expression for the LLE of a Rössler chaotic system [38] by utilizing the synchronization method [25,26,27,39]. It is established that synchronization between two diffusively coupled identical chaotic systems is invariably achievable with a sufficiently large coupling parameter [40]. A linear relationship exists between the synchronization threshold of the coupling parameter in two identical systems and the value of the LLE of the coupled systems. Consequently, the LLE can be estimated based on the critical coupling required for synchronization [39]. This paper focuses on the analytical criteria for synchronization between two identical Rössler chaotic systems from the perspective of the linear coupling of state variables. Unlike previous studies [41,42,43], we initially transform the synchronization error system between two identical Rössler chaotic systems into a set of Volterra integral equations, utilizing the Laplace transform and the convolution theorem. The critical coupling required for synchronization can be derived by applying the successive approximation method [44] within the framework of integral equation theory to resolve the error system's solution. Numerical simulations have been conducted to confirm the efficacy of our analytical estimation of the LLE for the Rössler chaotic system. Furthermore, this analytical estimation remains valid across a broad range of parameter variations.

    The remainder of the paper is structured as follows. Section 2 introduces the theoretical foundation of the estimation procedure for the LLE based on the synchronization method. Section 3 details the analytical estimation of the LLE for a Rössler system. Section 4 validates the analytical findings through numerical simulations. Finally, Section 5 provides the conclusions.

    Consider a set of ordinary differential equations

    ˙x=dxdt=f(x), (2.1)

    where xRn represents the state variables and f:RnRn is a smooth vector function. Assume that st(x0) is the solution of Eq (2.1) with the initial condition x=x0 which has the components (x10,x20,,xn0); one has

    dst(x0)dt=f[st(x0)],s0(x0)=x0. (2.2)

    Taking the variation with respect to x0 on both sides of Eq (2.2) yields

    dJt(x0)dt=f[st(x0)]xJt(x0), (2.3)

    where f[st(x0)]x=f(x)x|x=st(x0), Jt(x0)=st(x0)x0. Clearly, Jt(x0) can be obtained by solving Eq (2.3), which describes the influence of infinitesimal disturbance Δx0 to the initial condition x0 on the trajectory st(x0), that is,

    Δs(t)st(x0+Δx0)st(x0)=Jt(x0)Δx0. (2.4)

    Thus, the length of vector Δs(t) can be given as

    |Δs(t)|=Δs(t)TΔs(t)=ΔxT0Jt(x0)TJt(x0)Δx0, (2.5)

    where the notation T denotes the transpose of vectors. Since Jt(x0) is a real matrix, Jt(x0)TJt(x0) is real symmetric and positive semi-definite. Assume that ξi(t), i=1,2,,n, denotes the eigenvalues of the matrix Jt(x0)TJt(x0). Obviously, ξi(t)0. Assume that vi(t) is the corresponding eigenvector of ξi(t). If Δx0 has the same direction as vi(t), Eq (2.5) becomes

    |Δs(t)|=ξi(t)|Δx0|. (2.6)

    The definition of LEs denoted by λi, i=1,2,,n, in system (2.1) is given as

    λi=limtlnξi(t)t=limtln|ξi(t)|2t, i=1,2,,n. (2.7)

    From Eq (2.7), after long enough one has

    ξi(t)eλit, i=1,2,,n. (2.8)

    Substituting Eq (2.8) into Eq (2.6), leads to

    |Δsi(t)|=ξi(t)|Δxi0|=eλit|Δxi0|, i=1,2,,n. (2.9)

    The LEs are related to the expanding or contracting nature of different directions in phase space.

    Consider a chaotic system in the following form [25]

    ˙x=f(x), (2.10)

    where xRn and f:RnRn is a smooth functional vector. Two such identical oscillators couple to undergo unidirectional coupling, as follows:

    ˙x=f(x),˙y=f(y)+k(xy), (2.11)

    where x,yRn and kR is the coupling parameter. If k=0, two separate dynamical systems are obtained

    ˙x=f(x),˙y=f(y). (2.12)

    Assume that each system in Eq (2.12) evolves on an asymptotically stable chaotic attractor X. The solutions of Eq (2.12) starting from different initial conditions represent two independent trajectories on the atrractor X. If the two initial conditions are the same the two subsystems will exhibit identical behaviors (x=y). If the initial conditions for the two subsystems in Eq (2.12) have a small difference, then a state difference exists between the two subsystems during the time evolution, which is defined by the expression

    z=xy, (2.13)

    where zRn.

    Theorem 1. Assume that kmin>0 is the boundary value of the coupling parameter k that is required to cause synchronization in system (2.11), and that λmax is the LLE of system (2.10) such that λmaxkmin holds.

    Proof. To make further considerations easier, the following notations are first introduced:

    λj denotes the LEs in system (2.10) excluding λmax, j=1,2,,n1,

    Δλj=λmaxλj denotes the difference between λmax and other LEs λj, j=1,2,,n1,

    δ0 is initial distance in the λmax direction,

    δj0=mjδ0 denotes the initial distances in the λj direction, where mj denotes constant values, j=1,2,,n1.

    The norm of vector z is given by

    ||z||=(ni=1z2i)1/2. (2.14)

    Assume that z0=x0y0 is an initial distance between two trajectories of subsystems in system (2.12), where z0 has the components (z10,z20,,zn0). Obviously, z0 is the total λdistance vector, which is a sum of δ0 and δj0, j=1,2,,n1. From Eq (2.9), ||z|| can be written as

    ||z||=(δ20e2λmaxt+n1j=1δ2j0e2λjt)1/2=[δ20e2λmaxt(1+n1j=1m2je2Δλjt)]1/2. (2.15)

    Since Δλj<0 holds for j=1,2,,n1, the sum in Eq (2.15) finally decreases to zero during the time evolution; the norm of state difference z between two subsystems in Eq (2.12) approaches the following:

    ||z||=δ0eλmaxt. (2.16)

    This implies that the distance associated with the λmax direction becomes dominant after enough time.

    Next, we consider the case of k0 in Eq (2.11). For clarity, redefine the norm of the state difference between two subsystems in Eq (2.11) by Q. Clearly, Q0 for any values of x,y and k. From Eq (2.16), if k>0 one has

    ˙Q=||f(x)f(y)k(xy)||||f(x)f(y)||k||xy||=λmaxδ0eλmaxtkQ=(λmaxk)Q. (2.17)

    For k<0,

    ˙Q=||f(x)f(y)k(xy)||||f(x)f(y)||k||xy||=λmaxδ0eλmaxtkQ=(λmaxk)Q. (2.18)

    Solving Eqs (2.17) and (2.18) yields

    QQ0e(λmaxk)t,fork>0,QQ0e(λmaxk)t,fork<0, (2.19)

    where Q0 is a constant determined by the initial conditions given in Eq (2.11). If k>0 and the synchronization between two subsystems in Eq (2.11) is achieved, then Q0. From the first relation in Eq (2.19), it must follow that λmax<k. Assume that kmin>0 is the boundary value of the coupling parameter k that is required to cause synchronization in system (2.11); then, the following inequality should be held: λmax<kmin.

    On the contrary, if λmaxkmin, from the first inequality in Eq (2.19), the synchronization cannot be achieved. Therefore, one can have the following approximation

    λmaxkmin. (2.20)

    From Eq (2.18) and the second inequality in Eq (2.19), Q0 is impossible when λmax>0 and k<0. It implies that two identical chaotic systems in Eq (2.11) cannot synchronize with each other when k<0.

    The Rössler oscillator [38] is described as follows:

    ˙u=vw,˙v=u+av,˙w=b+w(uc), (3.1)

    where u, v, w are state variables, and a, b, c are parameters. For convenience, by moving the equilibrium (u0,v0,w0) of system (3.1) to the origin, system (3.1) can be rewritten as follows:

    ˙x=yzv0w0,˙y=x+ay+u0+av0,˙z=(xc)(z+w0)+u0(z+w0)+b, (3.2)

    where x=uu0, y=vv0, z=ww0. Consider two unidirectionally coupled Rössler systems as follows:

    ˙x1=y1z1v0w0,˙y1=x1+ay1+u0+av0,˙z1=(x1c)(z1+w0)+u0(z1+w0)+b,˙x2=y2z2v0w0+k(x1x2),˙y2=x2+ay2+u0+av0+k(y1y2),˙z2=(x2c)(z2+w0)+u0(z2+w0)+b+k(z1z2), (3.3)

    where k is a coupling parameter. Synchronization is said to occur in system (3.3) if

    ||x1x2||0, ||y1y2||0, ||z1z2||0for t. (3.4)

    By introducing

    e1=x1x2,e2=y1y2,e3=z1z2,
    e4=x1+x2,e5=y1+y2,e6=z1+z2,

    to system (3.3), the dynamical behavior of errors denoted by ei, i=1,2,3, can be described as follows:

    ˙e1=ke1e2e3,˙e2=e1(ka)e2,˙e3=w0e1(k+c)e3+e1e6+e3e4. (3.5)

    Then the synchronization condition given by Eq (3.4) becomes limt||ei||=0,i=1,2,3. Consider the Laplace transform defined as follows:

    Ei(s)=L[ei]=+0ei(t)estdt,ei(t)=L1[Ei]=12πjσ+jσjEi(s)estds,i=1,2,3. (3.6)

    By taking the Laplace transform on both sides of system (3.5), we obtain

    (sI3M)[E1E2E3]=[e10e20e30+W], (3.7)

    where e0i, i=1,2,3, denotes the given initial values of system (3.5), I3 is the 3×3 real identity matrix, and

    M=[k111ak0w00u0kc], (3.8)

    W is the Laplace transform of the nonlinear parts in the third equation in system (3.5)

    W=+0[e1e6+e3e4]estdt.

    Solving Eq (3.7) by using the Cramer's rule, one has

    E1=e10(s+ka)(s+k+c+u0)D(s)e20(s+k+c+u0)D(s)e30(s+ka)D(s)(s+ka)WD(s),E2=e10(s+k+c+u0)D(s)+e20(s+k)(s+k+c+u0)D(s)+e20w0e30D(s)WD(s),E3=e10w0(s+ka)D(s)e20w0e30D(s)+e30(s+k)(s+ka)D(s)+(s+k)(s+ka)WD(s)+WD(s), (3.9)

    where D(s)=s3+β1s2+β2s+β3 is the characteristic polynomial of the matrix of Eq (3.8), and

    β1=3k+cau0,β2=3k2+2(cau0)k+1+w0ac+au0,β3=k3+(cau0)k2+(1+w0ac+au0)k+caw0u0.

    To investigate whether ||ei||0, i=1,2,3, when t, we take the inverse Laplace transform on both sides of three equations in system (3.9) and consider the convolution theorem in the Laplace transform, which yields

    e1=e10γ5(t)e20γ2(t)e30γ1(t)t0γ1(tτ)[e1e6+e3e4]dτ,e2=e10γ2(t)+e20γ4(t)+(e20w0e30)γ6(t)t0γ6(tτ)[e1e6+e3e4]dτ,e3=e10w0γ1(t)(e20w0e30)γ6(t)+e30γ3(t)+t0(γ3(tτ)+γ6(tτ))[e1e6+e3e4]dτ, (3.10)

    where γ1=L1[s+kaD(s)], γ2=L1[s+k+c+u0D(s)], γ3=L1[(s+k)(s+ka)D(s)], γ4=L1[(s+k)(s+k+c+u0)D(s)], γ5=L1[(s+ka)(s+k+c+u0)D(s)], γ6=L1[1D(s)].

    Theorem 2. The necessary condition for e1,2,30 with t in Eq (3.10) is that all eigenvalues of the matrix of Eq (3.8) have negative real parts.

    Proof. Without loss of generality, consider the inverse Laplace transform of the following true fraction

    A1s2+A2s+A3D(s),

    where D(s) is the characteristic polynomial of the matrix of Eq (3.8) and Ai, i=1,2,3, denotes constants. There exist the following four cases:

    D(s) has 3 single real roots: s1,s2,s3

    A1s2+A2s+A3D(s)=B1ss1+B2ss2+B3ss3,

    where Bi=A1s2+A2s+A3D(s)(ssi)|s=si, i=1,2,3.

    L1[A1s2+A2s+A3D(s)]=B1es1t+B2es2t+B3es3t

    D(s) has a pair of conjugate complex roots s1,2=ω1±jω2 and a real root s3=ω3

    A1s2+A2s+A3D(s)=A1s2+A2s+A3(sω1jω2)(sω1+jω2)(sω3)=B1sω1jω2+B2sω1+jω2+B3sω3,

    where B1,2=A1s2+A2s+A3D(s)|s=ω1±jω2,B3=A1s2+A2s+A3D(s)(sω3)|s=ω3,

    L1[A1s2+A2s+A3D(s)]=B1e(ω1+jω2)t+B2e(ω1jω2)t+B3eω3t

    D(s) has 2 repeated real roots s=s0 and a single real root s=sk

    A1s2+A2s+A3D(s)=B1ss0+B2(ss0)2+B3ssk,

    where B1=12d2ds2[A1s2+A2s+A3D(s)(ss0)2]|s=s0, B2=A1s2+A2s+A3D(s)(ss0)2|s=s0,

    B3=A1s2+A2s+A3D(s)(ssk)|s=sk,

    L1[A1s2+A2s+A3D(s)]=(B1+B2t)es0t+B3eskt

    D(s) has 3 repeated real roots: s=s0

    A1s2+A2s+A3D(s)=B1ss0+B2(ss0)2+B3(ss0)3,

    where B(3i)=1i!didsi[A1s2+A2s+A3D(s)(ss0)3]|s=s0, i=1,2,

    B3=[A1s2+A2s+A3D(s)(ss0)3]|s=s0,

    L1[A1s2+A2s+A3D(s)]=(B1+B2t+B3t2)es0t

    It is evident that for ||ei|| to approach zero (i=1,2,3) in system (3.10), a necessary condition is that γj tends toward zero as time approaches infinity (j=1,2,3,4,5,6). This condition is satisfied if all roots of the equation D(s)=0 possess negative real parts, implying that all eigenvalues of the matrix of Eq (3.8) also have negative real parts.

    Under the condition given in Theorem 2, when t system (3.10) becomes as follows:

    e1=t0γ1(tτ)[e1e6+e3e4]dτ,e2=t0γ6(tτ)[e1e6+e3e4]dτ,e3=t0(γ3(tτ)+γ6(tτ))[e1e6+e3e4]dτ. (3.11)

    Theorem 3.2. e1,2,3=0 represents the unique continuous solutions to Eq (3.11).

    Proof. System (3.11) is a set of Volterra integral equations that can be solved by using the successive approximation method [44]. Consider the integral equation of the following form

    h(t)=Ψ(t)+t0g(tτ)H(τ,h(τ))dτ, (3.12)

    where g is an n×n matrix and Ψ(t) and H(t,h(t)) are vectors with n components. If the following conditions are satisfied

    |h|<;

    Ψ and h are continuous for 0<t<t0, where 0<t0<+;

    |g|L[0,ϵ] holds for any 0<ϵ<t0;

    ● For any η>0, if |h1h2|<η there must exist a constant κ>0 such that |H(t,h1)H(t,h2)|<κ,

    from the successive approximation method [44], Eq (3.12) has a unique continuous solution. Moreover the successive approximations given by

    ω0(t)=0,ωn+1(t)=Ψ(t)+t0g(tτ)H(τ,ωn(τ))dτ,n=0,1,2, (3.13)

    will uniformly converge to the unique continuous solution of Eq (3.12).

    Comparing Eq (3.11) with Eq (3.12), it is easy to verify that e1=e2=e3=0 constitutes the unique continuous solution of Eq (3.11).

    From Theorems 2 and 3, we have the following result:

    Theorem 4. The necessary condition for e1,2,30 in Eq (3.5) is that all eigenvalues of the matrix of Eq (3.8) have negative real parts.

    From the Routh-Hurwitz stability criterion, the necessary condition in {Theorem 4} is equivalent to the following condition:

    β1>0,β2>0,β3>0,β1β2β3>0, (3.14)

    where β1,2,3 has been defined in Eq (3.9). Since βi, i=1,2,3, denotes functions of k, from Eq (3.14) one can determine the boundary value of k.

    Theorem 5. The boundary value of k for synchronization in system (3.3) can be given as

    kc=max{m1,m2,m3}, (3.15)

    where max{} represents taking the maximum value of elements in the set,

    m1={ca3,a2+ac+c23(w0+1)0,ca3+a2+ac+c23(w0+1)3,a2+ac+c23(w0+1)>0, (3.16)

    and m2,3 are the maximum real roots of the equations

    β3=k3+(ca)k2+(1+w0ac)k+caw0=0,β1β2β3=k3+(ca)k2+(ca)2+1+w0ac4k+a2c(c2+1)a+cw08=0, (3.17)

    respectively.

    Proof. βi, i=1,2,3, denotes functions of k; it is easy to check that

    dβ1dk=3,dβ2dk=2β1,dβ3dk=β2.

    Therefore, one has

    d(β1β2β3)dk=d(β1β2)dkdβ3dk=dβ1dkβ2+β1dβ2dkβ2=2(β21+β2).

    If β1,2>0, then β3 and β1β2β3 are always monotonically increasing functions of k. β1>0 leads to

    k>cau03.

    Solving β2>0 yields the following:

    Ifa2+ac+c23(w0+1)>0,k<ca3a2+ac+c23(w0+1)3,ork>ca3+a2+ac+c23(w0+1)3Ifa2+ac+c23(w0+1)=0,kca3Ifa2+ac+c23(w0+1)<0,k(,+)

    Denote

    m1={ca3,a2+ac+c23(w0+1)0,ca3+a2+ac+c23(w0+1)3,a2+ac+c23(w0+1)>0.

    If k>m1, β3 and β1β2β3 are always monotonically increasing with an increase of k. Assume that k=m2,3 denotes the maximum real roots of equations β3 and β1β2β3, respectively, and that Eq (3.14) holds if and only if k>max{m1,m2,m3}.

    Remark. Suppose that system (3.1) has more than one equilibrium point (u01,v01,w01), (u02,v02,w02), , (u0n,v0n,w0n). For each equilibrium point (u0i,v0i,w0i), one from Eq (3.15) has one kic, i=1,2,,n. Then the boundary value of k for synchronization in system (3.3) is given by

    kmin=min{k1c,k2c,,knc}, (3.18)

    where min{} denotes the minimum value of elements in the set.

    In this section, numerical simulations are presented to illustrate the correctness of the result given by Eq (3.18). If a=0.15, b=0.2, and c=10.0, system (3.1) is chaotic [38]. Under such parameter conditions, the numerical result for the LLE of system (3.1) is λmax=0.092 [8,45], where the initial conditions are taken as u(0)=1, v(0)=1, and w(0)=1.

    Consider that the value of c is allowed to vary between 10.0 and 13.0. Using the numerical method proposed in [8], the LLEs of system (3.1) for different values of c can be obtained as shown in Figure 1, where the initial conditions are retained as u(0)=1, v(0)=1, and w(0)=1.

    Figure 1.  The LLE λmax versus t for system (3.1) with a=0.15, b=0.2, and (a) c=10.0 (b) c=11.0 (c) c=12.0, and (d) c=13.0. The initial conditions were taken as u(0)=1, v(0)=1, and w(0)=1.

    From Theorem 5, under the certain limitation of the parameters, the LLE of system (3.1) is just the maximum real root of the following equation (obtained from the second equation in Eq (3.17))

    H(k)=k3+h1k2+h2k+h3, (4.1)

    where

    h1=c(w0+1)a,h2=14{[(w0+1)2+w0]a2(2w0+3)ac+c2+w0+1},h3=18{[ac2(w0+(1+2w0)a2)c+a3w0(w0+1)+a(w20+1)]},w0=12a(c4ab+c2),

    a=0.15, b=0.2 and c varies in the range of 10 to 13.

    From Appendix, the analytic expression of the LLE of system (3.1) can be given as

    λmax=k=3Y1+3Y2h13,  (4.2)

    where

    Y1=3B2h1A+3Δ2,Y2=3B2h1A3Δ2A=h213h2,B=h1h29h3,C=h223h1h3,Δ=B24AC.

    Figure 2 illustrates the comparisons between the analytical results from Eq (4.2) and the numerical results obtained by using the method described in [8]. The numerical results exhibit minor differences relative to the analytical results as the value of c increases. Figure 3 displays the time series for x1,2, y1,2, and z1,2 in system (3.3) with varying c values to identify the critical synchronization conditions for k. The analytical estimation based on Eq (4.2) for the LLE of system (3.1) is confirmed to be valid and highly accurate, as evidenced by the data in Figures 2 and 3.

    Figure 2.  The values of the LLE λmax of system (3.1) with a=0.15, b=0.2, and c[10,13]. The red line denotes the analytical results obtained by using Eq (4.2). The blue line represents the numerical results obtained by using the method given in [8], where the initial conditions were taken as u(0)=1, v(0)=1, and w(0)=1.
    Figure 3.  The time series of x1,2, y1,2 and z1,2 in system (3.3) with a=0.15, b=0.2, and (a) c=10, k=0.083; (b) c=10, k=0.085; (c) c=11, k=0.083; (d) c=11, k=0.085; (e) c=12, k=0.08; (f) c=12, k=0.083; (g) c=13, k=0.08; and (h) c=13, k=0.086. x1, y1, z1 and x2, y2, z2 are denoted by blue solid lines () and red dashed lines (), respectively. The initial conditions were applied as x1(0)=1, y1(0)=1, z1(0)=1, x2(0)=1.1, y2(0)=1.1, and z2(0)=1.1.

    Consider that the value of a changes in the range of 0.15 to 0.2. Applying the numerical method in [8], one can obtain the LLEs of system (3.1) for different values of a, as depicted in Figure 4. The initial conditions were applied as u(0)=1, v(0)=1, and w(0)=1.

    Figure 4.  The LLE λmax versus t for system (3.1) with b=0.2, c=10.0 and (a) a=0.16 (b) a=0.17 (c) a=0.18, and (d) a=0.2. The initial conditions were applied as u(0)=1, v(0)=1, and w(0)=1.

    For the parameter values considered in this section, the analytical expression of the LLE of system (3.1) is still given by Eq (4.2). The comparisons between the analytical results obtained based on Eq (4.2) and the numerical results derived by using the method in [8] are depicted in Figure 5.

    Figure 5.  The values of the LLE λmax of system (3.1) with b=0.2, c=10.0 and a[0.15,0.2]. The red line denotes the analytical results obtained by using Eq (4.2). The blue line represents the numerical results obtained by using the method given in [8], where the initial conditions were taken as u(0)=1, v(0)=1 and w(0)=1.

    Figure 5 demonstrates that the analytical results linearly increase as the value of a rises. The numerical results, however, display slight fluctuations before reaching their peak. Figure 6 depicts the time series for x1,2, y1,2 and z1,2 in system (3.3) for various values of a and k. A comparison of Figures 5 and 6 reveals that the analytic approach outlined in Eq (4.2) is both effective and highly accurate as a tool to determine the LLE of system (3.1), especially when the value of a varies within the range of 0.15 to 0.2.

    Figure 6.  The time series of x1,2, y1,2 and z1,2 in system (3.3) with b=0.2, c=10.0 and (a) a=0.16, k=0.083; (b) a=0.16, k=0.089; (c) a=0.17, k=0.085; (d) a=0.17, k=0.09; (e) a=0.18, k=0.1; (f) a=0.18, k=0.108; (g) a=0.2, k=0.11; and (h) a=0.2, k=0.112. x1, y1, z1 and x2, y2, z2 are denoted by blue solid lines () and red dashed lines (), respectively. The initial conditions were applied as x1(0)=1, y1(0)=1, z1(0)=1, x2(0)=1.1, y2(0)=1.1, and z2(0)=1.1.

    Considering that the values of a and c vary within the ranges of 0.15 to 0.2 and 10 to 13, respectively, Figure 7 was constructed based on the calculated results of Eq (4.2) to illustrate the variation in the LLE of system (3.1) as a and c change. Equation (4.2) facilitates the examination of how wide-ranging parameter values influence the LLE in system (3.1).

    Figure 7.  The values of the LLE λmax of system (3.1) with b=0.2, a[0.15,0.2], and c[10.0,13.0] according to Eq (4.2).

    Although a linear relationship exists between the synchronization threshold of the coupling coefficient in two identical chaotic systems and their LLE, previous studies have primarily derived the boundary value of the coupling parameter by employing numerical methods to estimate the LLE, to the best of our knowledge. This paper has presented an approach to analytically estimate the LLE of the Rössler chaotic system by using the synchronization method. Unlike previous studies, this approach transforms the synchronization error system into a set of Volterra integral equations. The stability of these equations was then examined through the application of the successive approximation method in accordance with the theory of integral equations. Compared to the numerical results for the LLEs of Rössler chaotic systems, our analytical estimates demonstrate high accuracy. Moreover, these analytical estimates remain valid across a wide range of parameter variations.

    Our findings reveal that the value of the LLE for the Rössler chaotic system corresponds to the maximum real root of a cubic algebraic equation. This insight simplifies the challenge associated with analytically determining the LLE to solve such an equation. Our research introduces a novel approach for the analysis and management of the impact of parameter variations on the LLE value in the Rössler chaotic system.

    The authors declare that they have not used artificial intelligence tools in the creation of this article.

    The authors gratefully acknowledge the financial support from the National Natural Science Foundation of China grant numbers 11672185 and 11972327.

    The authors declare that there is no conflict of interest.

    In the Appendix, expressions of real roots of a cubic equation are provided according to the Cardano formula. Consider the following cubic equation:

    f(x)=ax3+bx2+cx+d=0, (A1)

    where a,b,c,d are real constants and a0. Denote

    A=b23ac,B=bc9ad,C=c23bd,Δ=B24AC,Y1,2=3aB2Ab±3aΔ2. (A2)

    ● If Δ>0, there is only one real root:

    x=3Y1+3Y2b3a.

    ● If Δ=0 and A=0, there are three equal real roots:

    x1=x2=x3=b3a.

    ● If Δ=0 and A>0, there are three real roots, where two roots are equal:

    x1=BAba, x2=x3=B2A.

    ● If Δ<0 and A>0, there are three different real roots:

    xi=2Acos(ϕ+2(i1)π3)b3a, ϕ=arccos(3aB2Ab2AA),i=1,2,3.

    [1] ADDIN Mendeley Bibliography CSL_BIBLIOGRAPHY Moyo EN, Nangombe SS (2015) Southern Africa's 2012–13 violent storms: role of climate change. Procedia IUTAM 17: 69–78. doi: 10.1016/j.piutam.2015.06.011
    [2] Wahid H, Ahmad S, Nor MSM, et al. (2017) Prestasi kecekapan pengurusan kewangan dan agihan zakat: perbandingan antara majlis agama islam negeri di Malaysia. J Ekon Malaysia 51: 39–54.
    [3] Mutasa M (2010) Zimbabwe' s Drought Conundrum : vulnerability and coping in Buhera and Chikomba districts. MS thesis, 163.
    [4] Ndlovu S (2010) Coping With Drought. Mainstreaming Livelihood Centred Approaches to Disaster Management.
    [5] Levey KM, Jury MR (1996) Composite intraseasonal oscillations of convection over Southern Africa. J Clim 9: 1910–1920. doi: 10.1175/1520-0442(1996)009<1910:CIOOCO>2.0.CO;2
    [6] Makarau A (1995) Intra-seasonal oscillatory modes of the southern Africa summer circulation, Univ. Cape T. (PhD thesis).
    [7] Zhakata W (1996) Impacts of climate variability and forecasting on agriculture (Zimbabwe experience), in Proc. Workshop on Reducing Climate-Related Vulnerability in Southern Africa, 131–139.
    [8] Jury MR (1993) The structure and possible forcing mechanisms of the 1991–1992 drought in Southern Africa, Suid-Afrikaanse Tydskr. vir Natuurwetenskap en Tegnol 12, 8–16.
    [9] Colberg F (2004) South Atlantic response to El Niño–Southern Oscillation induced climate variability in an ocean general circulation model. J Geophys Res 109: C12015. doi: 10.1029/2004JC002301
    [10] Chikozho C (2010) Applied social research and action priorities for adaptation to climate change and rainfall variability in the rainfed agricultural sector of Zimbabwe. Phys Chem Earth 35: 780–790. doi: 10.1016/j.pce.2010.07.006
    [11] Gwimbi P (2009) Cotton farmers' vulnerability to climate change in Gokwe District (Zimbabwe): impact and influencing factors. JÀMBÁ J Disaster Risk Stud 2: 81–92.
    [12] Alvarez TC, Delgado RC, González PA (2014) Aplicabilidad de los sistemas de triaje prehospitalarios en los incidentes con múltiples víctimas: De la teoría a la práctica. Emergencias 26: 147–154.
    [13] Tome S (2017) Weekly Bulletin on Outbreaks. Week 46, no. July, 1–7.
    [14] Bohle HG, Downing TE, Watts M J (1994) Climate Change and Social Vulnerability. Toward a Sociology and Geography of Food Insecurity. Glob Environ Chang 4: 37–48.
    [15] Mavhura E, Bernard S, Collins AE, et al. (2013) Author' s personal copy International Journal of Disaster Risk Reduction Indigenous knowledge, coping strategies and resilience to floods in Muzarabani, Zimbabwe. Int J Disaster Risk Reduct 5: 38–48. doi: 10.1016/j.ijdrr.2013.07.001
    [16] Mavhura E, Manatsa D, Mushore T (2015) Adaptation to drought in arid and semi-arid environments: Case of the Zambezi Valley, Zimbabwe. Jàmbá J Disaster Risk Stud 7: 1–7.
    [17] Mushore TD (2013) Uptake Of Seasonal Rainfall Forecasts In Zimbabwe. IOSR J Environ Sci Toxicol Food Technol 5: 31–37. doi: 10.9790/2402-0513137
    [18] Cook KH (2000) The South Indian convergence zone and interannual rainfall variability over Southern Africa. J Clim 13: 3789–3804. doi: 10.1175/1520-0442(2000)013<3789:TSICZA>2.0.CO;2
    [19] Mason S, Lindesay J, Tyson J (1994) Simulating drought in Southern Africa using sae surface temperature variations. Water SA 20: 15–22.
    [20] Beilfuss R (2012) A Risky Climate for Southern African Hydro. Int Rivers Berkely, 1–46.
    [21] Mamombe V, Kim WM, Choi YS (2017) Rainfall variability over Zimbabwe and its relation to large-scale atmosphere–ocean processes. Int J Climatol 37: 963–971. doi: 10.1002/joc.4752
    [22] Mwafulirwa ND (1999) Climate variability and predictability in tropical southern africa with a focus on dry spells over Malawi.
    [23] Manatsa D, Chingombe W, Matsikwa H, et al. (2008) The superior influence of Darwin Sea level pressure anomalies over ENSO as a simple drought predictor for Southern Africa. Theor Appl Climatol 92: 1–14. doi: 10.1007/s00704-007-0315-3
    [24] Manatsa D, Mukwada G (2012) Rainfall Mechanisms for the Dominant Rainfall Mode over Zimbabwe Relative to ENSO and/or IODZM. Sci World J 2012: 1–15.
    [25] Gaughan AE, Staub CG, Hoell A, et al. (2016) Inter- and Intra-annual precipitation variability and associated relationships to ENSO and the IOD in southern Africa. Int J Climatol 36: 1643–1656. doi: 10.1002/joc.4448
    [26] Ogwang BA, Guirong T, Haishan C (2012) Diagnosis of September - November Drought and the Associated Circulation Anomalies Over Uganda. Pakistan J Meteorol 9: 11–24.
    [27] Kabanda TA, Jury MR (1999) Inter-annual variability of short rains over northern Tanzania. Clim Res 13: 231–241. doi: 10.3354/cr013231
    [28] Nangombe S, Madyiwa S, Wang J (2018) Precursor conditions related to Zimbabwe's summer droughts. Theor Appl Climatol 131: 413–431. doi: 10.1007/s00704-016-1986-4
    [29] Manatsa D, Reason CJC, Mukwada G (2012) On the decoupling of the IODZM from southern Africa Summer rainfall variability. Int J Climatol 32: 727–746. doi: 10.1002/joc.2306
    [30] Simmonds I, Hope P (1997) Persistence Characteristics of Australian Rainfall Anomalies. Int J Climatol 17: 597–613. doi: 10.1002/(SICI)1097-0088(199705)17:6<597::AID-JOC173>3.0.CO;2-V
    [31] Manatsa D, Mushore T, Lenouo A (2017) Improved predictability of droughts over southern Africa using the standardized precipitation evapotranspiration index and ENSO. Theor Appl Climatol 127: 259–274. doi: 10.1007/s00704-015-1632-6
    [32] Department of Meterological Services (1981) Climate Handbook of Zimbabwe, Zimbabwe Meteorol. Serv. Harare, 48.
    [33] Mavhura E, Collins A, Bongo PP (2017) Flood vulnerability and relocation readiness in Zimbabwe. Disaster Prev Manag An Int J 26: 41–54. doi: 10.1108/DPM-05-2016-0101
    [34] Muzenda-Mudavanhu C, Manyena B, Collins AE (2015) Disaster risk reduction knowledge among children in Muzarabani District, Zimbabwe. Nat Hazards 84: 911–931.
    [35] Bohle HG, Downing TE, Watts MJ (1994) Köppen Climate Classification for Bucharest.
    [36] Janus Corporate Solutions (2016) Weather and Climate in Singapore. Longman.
    [37] Matarira CH (1990) Theoretical and Ap. plied Climatology Frequency and Tracks of Anticyclones and Their Effect on Rainfall Pat- terns over Zimbabwe. Theor Appl Climatol 66: 53–66.
    [38] Makarau A, Jury MR (1997) Seasonal Cycle of Convective Spells Over Southern. Int J Climatol 17: 1317–1332. doi: 10.1002/(SICI)1097-0088(199710)17:12<1317::AID-JOC197>3.0.CO;2-A
    [39] Manatsa D, Morioka Y, Behera SK, et al. (2013) Link between Antarctic ozone depletion and summer warming over southern Africa. Nat Geosci 6: 934–939. doi: 10.1038/ngeo1968
    [40] Levitus S, Boyer T (1998) NOAA/OAR/ESRL PSD, Boulder, Colorado, USA.
    [41] Dee DP, Uppala SM, Simmons AJ, et al. (2011) The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Q J R Meteorol Soc 137: 553–597. doi: 10.1002/qj.828
    [42] Kalnay E, Kanamitsu M, Kistler R, et al. (1996) The NCEP/NCAR 40-year reanalysis project. Bull Am Meteorol Soc 77: 437–471. doi: 10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2
    [43] Smith TM, Reynolds RW, Peterson TC, et al. (2008) Improvements to NOAA's historical merged land–ocean surface temperature analysis (1880–2006). J Clim 21: 2283–2296. doi: 10.1175/2007JCLI2100.1
    [44] Folland CK (1983) Regional-Scale International Variability Of Climate-A Northwest European-Perspective. Meteorol Mag 112: 163–183.
    [45] Mpeta EJ (1997) Intra-seasonal convection dynamics over southwest and northeast Tanzania: an observational study, University of Cape Town.
    [46] Naeraa M, Jury MR (1998) Meteorology and Atmospheric Physics Tropical Cyclone Composite Structure and Impacts over Eastern Madagascar During January-March 1994. Meteorol Atmos Phys 53: 43–53.
    [47] Jury MR, Levey KM (1997) Vertical structure of the atmosphere during wet spells over Southern Africa. Water SA 23: 51–55.
    [48] Hargraves R, Jury M (1997) Composite meteorological structure of flood events over the eastern mountains of South Africa. Water SA 23: 357–363.
    [49] Boschat G, Simmonds L, Purich A, et al. (2016) On the use of composite analyses to form physical hypotheses: An example from heat wave-SST associations. Sci Rep 6: 29599. doi: 10.1038/srep29599
    [50] Rocha A, Simmonds I (1997) Interannual Variability of South-Eastern African Summer Rainfall. Part Ii. Modelling the Impact of Sea-Surface Temperatures on Rainfall and Circulation. Int J Climatol 17: 267–290.
    [51] Simmonds I, Bi D, Hope P (1999) Atmospheric water vapor flux and its association with rainfall over China in summer. J Clim 12: 1353–1367. doi: 10.1175/1520-0442(1999)012<1353:AWVFAI>2.0.CO;2
    [52] Preston-Whyte RA, Tyson PD (1988) The atmosphere and weather of Southern Africa. Oxford University Press.
    [53] Rocha A, Simmonds I (1997) Interannual Variability of South-Eastern African Summer Rainfall. Part 1: Relationships With Air–Sea Interaction Processes. Int J Climatol 17: 235–265.
    [54] Nicholson SE, Entekhabi D (1987) Rainfall Variability in Equatorial and Southern Africa: Relationships with Sea Surface Temperatures along the Southwestern Coast of Africa. J Clim Appl Meteorol 26: 561–578. doi: 10.1175/1520-0450(1987)026<0561:RVIEAS>2.0.CO;2
  • Reader Comments
  • © 2018 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(5417) PDF downloads(998) Cited by(6)

Figures and Tables

Figures(7)

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog