Review

Meta-analysis identifies metabolic sensitivities to ocean acidification

Running title: Ocean acidification impacts metabolic function
  • Received: 11 July 2017 Accepted: 22 September 2017 Published: 21 November 2017
  • Ocean acidification is expected to have wide-ranging and complicated impacts on organismal physiology, notably metabolism. Effects on metabolism may have numerous consequences at the whole-organism level, in particular costs to growth, locomotion, reproductive output, and homeostasis. Negative effects on these metrics may further cascade up to impact ecosystem structure and function, and thus human society. As such, the study of metabolism in response to ocean acidification is a pivotal research avenue within the study of global ocean change. Here, the metabolic responses of marine species to ocean acidification are reviewed and examined through meta-analysis. We reviewed a total of 44 published studies and used a traditional meta-analysis to identify broad-scale trends in the metabolic responses of species to ocean acidification. Results from this study indicate varied metabolic strategies in response to OA, further complicating our predictive power to forecast ecosystem-level consequences of ongoing CO2 increases. However, strong effects were observed with respect to ontogeny, marine ecosystem, motility, and taxonomic origin, thereby reinforcing the need for a multi-faceted approach to both management of sensitive species and mitigation of future impacts.

    Citation: Amanda L. Kelley, Jay J. Lunden. Meta-analysis identifies metabolic sensitivities to ocean acidification[J]. AIMS Environmental Science, 2017, 4(5): 709-729. doi: 10.3934/environsci.2017.5.709

    Related Papers:

    [1] Tao-Li Kang, Hai-Feng Huo, Hong Xiang . Dynamics and optimal control of tuberculosis model with the combined effects of vaccination, treatment and contaminated environments. Mathematical Biosciences and Engineering, 2024, 21(4): 5308-5334. doi: 10.3934/mbe.2024234
    [2] Laurenz Göllmann, Helmut Maurer . Optimal control problems with time delays: Two case studies in biomedicine. Mathematical Biosciences and Engineering, 2018, 15(5): 1137-1154. doi: 10.3934/mbe.2018051
    [3] Zhenzhen Liao, Shujing Gao, Shuixian Yan, Genjiao Zhou . Transmission dynamics and optimal control of a Huanglongbing model with time delay. Mathematical Biosciences and Engineering, 2021, 18(4): 4162-4192. doi: 10.3934/mbe.2021209
    [4] Na Pang . Nonlinear neural networks adaptive control for a class of fractional-order tuberculosis model. Mathematical Biosciences and Engineering, 2023, 20(6): 10464-10478. doi: 10.3934/mbe.2023461
    [5] Xin Gao, Yue Zhang . Bifurcation analysis and optimal control of a delayed single-species fishery economic model. Mathematical Biosciences and Engineering, 2022, 19(8): 8081-8106. doi: 10.3934/mbe.2022378
    [6] Jerzy Klamka, Helmut Maurer, Andrzej Swierniak . Local controllability and optimal control for\newline a model of combined anticancer therapy with control delays. Mathematical Biosciences and Engineering, 2017, 14(1): 195-216. doi: 10.3934/mbe.2017013
    [7] Heping Ma, Hui Jian, Yu Shi . A sufficient maximum principle for backward stochastic systems with mixed delays. Mathematical Biosciences and Engineering, 2023, 20(12): 21211-21228. doi: 10.3934/mbe.2023938
    [8] Ya-Dong Zhang, Hai-Feng Huo, Hong Xiang . Dynamics of tuberculosis with fast and slow progression and media coverage. Mathematical Biosciences and Engineering, 2019, 16(3): 1150-1170. doi: 10.3934/mbe.2019055
    [9] H. J. Alsakaji, F. A. Rihan, K. Udhayakumar, F. El Ktaibi . Stochastic tumor-immune interaction model with external treatments and time delays: An optimal control problem. Mathematical Biosciences and Engineering, 2023, 20(11): 19270-19299. doi: 10.3934/mbe.2023852
    [10] Eduardo Ibarguen-Mondragon, Lourdes Esteva, Leslie Chávez-Galán . A mathematical model for cellular immunology of tuberculosis. Mathematical Biosciences and Engineering, 2011, 8(4): 973-986. doi: 10.3934/mbe.2011.8.973
  • Ocean acidification is expected to have wide-ranging and complicated impacts on organismal physiology, notably metabolism. Effects on metabolism may have numerous consequences at the whole-organism level, in particular costs to growth, locomotion, reproductive output, and homeostasis. Negative effects on these metrics may further cascade up to impact ecosystem structure and function, and thus human society. As such, the study of metabolism in response to ocean acidification is a pivotal research avenue within the study of global ocean change. Here, the metabolic responses of marine species to ocean acidification are reviewed and examined through meta-analysis. We reviewed a total of 44 published studies and used a traditional meta-analysis to identify broad-scale trends in the metabolic responses of species to ocean acidification. Results from this study indicate varied metabolic strategies in response to OA, further complicating our predictive power to forecast ecosystem-level consequences of ongoing CO2 increases. However, strong effects were observed with respect to ontogeny, marine ecosystem, motility, and taxonomic origin, thereby reinforcing the need for a multi-faceted approach to both management of sensitive species and mitigation of future impacts.


    1. Introduction

    Tuberculosis (TB) is the second leading cause of death from an infectious disease worldwide [32]. Active TB refers to disease that occurs in someone infected with Mycobacterium tuberculosis. It is characterized by signs or symptoms of active disease, or both, and is distinct from latent tuberculosis infection, which occurs without signs or symptoms of active disease. Only individuals with active TB can transmit the infection. Many people with active TB do not experience typical TB symptoms in the early stages of the disease. These individuals are unlikely to seek care early, and may not be properly diagnosed when seeking care [31].

    Delays to diagnosis of active TB present a major obstacle to the control of a TB epidemic [27], it may worsen the disease, increase the risk of death and enhance tuberculosis transmission to the community [24,26]. Both patient and the health system may be responsible for the diagnosis delay [24]. Efforts should be done in patient knowledge/awareness about TB, and health care systems should improve case finding strategies to reduce the delay in diagnosis of active TB [15,24,25].

    Mathematical models are an important tool in analyzing the spread and control of infectious diseases [13]. There are several mathematical dynamic models for TB, see, e.g., [5,6,11,18]. In this paper we consider the mathematical model for TB proposed in [11]. We introduce a discrete time delay which represents the delay on the diagnosis of individuals with active TB and commencement of treatment. The stability of the disease free and endemic equilibriums is analyzed for any time delay.

    Optimal control theory has been successfully applied to TB mathematical models (see, e.g., [19,22,23] and references cited therein). We propose and analyze an optimal control problem where the control system is the mathematical model from [11], but with a time delay in the state variable that represents individuals with active TB, and introduce two control functions. The control functions represent the fraction of early and persistent latent individuals that are treated for TB. Treatment of latent TB infection greatly reduces the risk that TB infection will progress to active TB disease. Certain groups are at very high risk of developing active TB disease once infected. Every effort should be made to begin appropriate treatment and to ensure completion of the entire course of treatment for latent TB infection [33]. Treatment of latent TB infection should be initiated after the possibility of TB disease has been excluded. It can take 2 to 8 weeks after TB infection for the body's immune system to react to tuberculin and for the infected to be detected, which justifies the introduction of a time delay on the control associated to treatment of early latent individuals. On the other hand, delays in the treatment of latent TB may also occur due to clinical and demographic patient and health care services characteristics. For these reasons, we consider discrete time delays in both control functions. To our knowledge, this work is the first to apply optimal control theory to a TB model with time delay in state and control variables.

    The paper is organized as follows. In Section 2 we formulate the TB model with state delay. The stability of the disease free equilibrium is analyzed in Section 3 while stability of the endemic equilibrium is investigated in Section 4. Optimal control of TB with state and control delays is carried out in Section 5 and some numerical results given in Section 6. We end with Section 7 of conclusions.


    2. TB model with state delay

    In this section we consider a TB mathematical model proposed in [11], where reinfection and post-exposure interventions for tuberculosis are considered. The model divides the total population into five categories: susceptible (S); early latent (L1), i.e., individuals recently infected (less than two years) but not infectious; infected (I), i.e., individuals who have active TB and are infectious; persistent latent (L2), i.e., individuals who were infected and remain latent; and recovered (R), i.e., individuals who were previously infected and have been treated. As in [11], we assume that the total population, N, with N=S(t)+L1(t)+I(t)+L2(t)+R(t), is constant in time. In other words, we assume that the birth and death rates are equal and there are no TB-related deaths. We introduce a discrete time-delay in the state variable I, denoted by dI, that represents the delay to diagnosis and commencement of treatment of active TB infection,

    {˙S(t)=μNβNI(t)S(t)μS(t),˙L1(t)=βNI(t)(S(t)+σL2(t)+σRR(t))(δ+τ1+μ)L1(t),˙I(t)=ϕδL1(t)+ωL2(t)+ωRR(t)τ0I(tdI)μI(t),˙L2(t)=(1ϕ)δL1(t)σβNI(t)L2(t)(ω+τ2+μ)L2(t),˙R(t)=τ0I(tdI)+τ1L1(t)+τ2L2(t)σRβNI(t)R(t)(ωR+μ)R(t). (1)

    The initial conditions for system (1) are

    S(θ)=φ1(θ),L1(θ)=φ2(θ),I(θ)=φ3(θ),L2(θ)=φ4(θ),R(θ)=φ5(θ), (2)

    dIθ0, where φ=(φ1,φ2,φ3,φ4,φ5)TC with C the Banach space C([dI,0],R5) of continuous functions mapping the interval [dI,0] into R5. From biological meaning, we further assume that φi(0)>0 for i=1,,5.

    Throughout this paper, we focus on the dynamics of the solutions of (1) in the restricted region

    Ω={(S,L1,I,L2,R)R5+0|0S+L1+I+L2+R=N}.

    In this region, the usual local existence, uniqueness and continuation results apply [12,14]. Hence, a unique solution (S(t),L1(t),I(t),L2(t),R(t)) of (1) with initial condition (2) exists for all time t0. Consider the solutions of (1) with (φ1(θ),,φ5(θ))IntΩ, i=1,,5, for all θ[dI,0]. Then the solutions stay in the interior of the region Ω for all time t0, i.e., the region Ω is positively invariant with respect to system (1) (see, e.g., [14]).

    A mathematical model has a disease free equilibrium if it has an equilibrium point at which the population remains in the absence of the disease [28]. The model (1) has a disease free equilibrium given by E0=(N,0,0,0,0).

    The basic reproduction number R0 represents the expected average number of new TB infections produced by a single TB active infected individual when in contact with a completely susceptible population [28]. For model (1) it is given by

    R0=βμωR(ω+τ2+μ)τ1+δ[(ωR+μ)(ϕμ+ω)+(ωR+ϕμ)τ2](τ0+μ+ωR)(ω+τ2+μ)(δ+τ1+μ)=ND. (3)

    Note that in [11] the basic reproduction number is deduced under the assumption that τ1=τ2=0. The expression (3) generalizes the one given in [11].


    3. Stability of the disease free equilibrium

    It is important to analyze the stability of the disease free equilibrium, as it indicates whether the population will remain in the absence of the disease, or the disease will persist for all time [28,29]. System (1) is equivalent to

    {˙S(t)=μNβNI(t)S(t)μS(t),˙L1(t)=βNI(t)(S(t)+σL2(t)+σR(NS(t)L1(t)I(t)L2(t)))(δ+τ1+μ)L1(t),˙I(t)=ϕδL1(t)+ωL2(t)+ωR(NS(t)L1(t)I(t)L2(t))τ0I(tdI)μI(t),˙L2(t)=(1ϕ)δL1(t)σβNI(t)L2(t)(ω+τ2+μ)L2(t), (4)

    where the equation for R(t) is derived from R(t)=NS(t)L1(t)I(t)L2(t). The disease free equilibrium of system (4) is given by (S0,L10,I0,L20)=(N,0,0,0). To discuss its local asymptotic stability, let us consider the coordinate transformation s(t)=S(t)¯S, l1(t)=L1(t)¯L1, i(t)=I(t)¯I, l2(t)=L2(t)¯L2, where (¯S,¯L1,¯I,¯L2) denotes any equilibrium of (4). Hence, we have that the corresponding linearized system of (4) is of the form

    {˙s(t)=β¯I+μNNs(t)βN¯Si(t)˙l1(t)=βN¯I(σR1)s(t)βN(σR¯I+(δ+τ1+μ)N)l1(t)βN(¯Sσ¯L2σR(N+¯S+¯L1+2¯I+¯L2))i(t)+β¯I(σσR)Nl2(t)˙i(t)=ωRs(t)+(ϕδωR)l1(t)(ωRμ)i(t)+(ωωR)l2(t)τ0i(tdI)˙l2(t)=(1ϕ)δl1(t)βNσ¯L2i(t)βN(σ¯I+(ω+τ2+μ)N)l2(t). (5)

    We then express system (5) in matrix form as follows:

    ddt(s(t)l1(t)i(t)l2(t))=A1(s(t)l1(t)i(t)l2(t))+A2(s(tdI)l1(tdI)i(tdI))l2(tdI))

    with A1 the 4×4 matrix

    A1=[β¯I+μNN0β¯SN0β¯I(1σR)Nβ¯IσR+c1NNβ(¯S+σ¯L2+σR(N+¯S+¯L1+2¯I+¯L2))Nβ¯I(σσR)NωRϕδωRωRμωωR0(1+ϕ)δσβ¯L2Nβ¯Iσ+c2NN],

    where c1=δ+τ1+μ and c2=ω+τ2+μ, and A2 the 4×4 diagonal matrix A2=diag(0,0,τ0,0). The transcendental characteristic equation of system (5) is defined by Δ(λ)=det(λIA1eλdIA2)=0 and is given by

    Δ(λ)=P(λ)+Q(λ)=0, (6)

    where

    P(λ)=λ4+a3λ3+a2λ2+a1λ+a0,Q(λ)=τ0(λ+μ)(λ+c1)(λ+c2)(eλdI1),

    with a0=DN,

    a1=2μD+μ2(c1+c2+c4)c4c5c6β(τ1ωR+ωδ+δϕ(ωR+τ2+2μ)),a2=c4c5+3μ(c1+c2+c4)+c6(c4+c5)βϕδ,a3=c1+c2+c3+μ,

    and c3=ωR+τ0+μ, c4=τ0+ωR, c5=τ2+ω, c6=δ+τ1.

    Remark 1. For any dI0 and a3>0, if R0<1, then a0>0.

    Recall that an equilibrium point is asymptotically stable if all roots of the corresponding characteristic equation have negative real parts [1].

    Lemma 3.1 If R0>1, then the disease free equilibrium E0 is unstable for any dI0.

    Proof. The characteristic equation (6) satisfies Δ(0)=DN. Assuming R0>1, then Δ(0)<0. Since limλ+Δ(λ)=+, there exists at least one positive root of (6).

    Lemma 3.2. If (ⅰ) R0<1, (ⅱ) a1>0, (ⅲ) a2>0, (ⅳ) a3a2>a1, and (ⅴ) a3a2a1>a21+a23a0, then the disease free equilibrium E0 is locally asymptotically stable for dI=0.

    Proof. When dI=0, the associated transcendental characteristic equation (6) of system (5) at (N,0,0,0)=(¯S,¯L1,¯I,¯L2) becomes Δ(λ)=P(λ)=0. Using the Routh--Hurwitz criterion for fourth-order polynomials, all the roots of P(λ) have negative real part if all coefficients satisfy an>0 for n=0,,3, a3a2>a1, and a3a2a1>a21+a23a0.

    In the case dI>0, by Rouché's theorem [8], if instability occurs for a particular value of the delay dI, then a characteristic root of (6) must intersect the imaginary axis. Our aim is to prove that the polynomial (6) does not have purely imaginary roots for all positive delays (see, e.g., [2,7]). The complex ib, b>0, is a root of (6) if and only if b4ia3b3a2b2+ia1b+a0+τ0(ib+μ)(ib+c1)(ib+c2)(eibdI1)=0. By using Euler's formula expibdI=cos(bdI)isin(bdI), and by separating real and imaginary parts, we have

    a3b3+τ0bc1c2a1bτ0b3+τ0μc1b+τ0μbc2=Acos(bdI)Bsin(bdI)b4+a2b2τ0b2c1a0τ0b2c2μτ0b2+τ0μc1c2=Asin(bdI)+B(cos(bdI))

    with A=τ0b(c1c2b2+μ(c1+c2)) and B=τ0(μc1c2b2(μ+c1+c2)). Adding up the squares of both equations, we obtain that

    b8+α3b6+α2b4+α1b2+α0=0, (7)

    where α0=a0(a02μτ0c1c2),

    α1=2τ0(μ(a0+a2c1c2a1(c1+c2))+a0(c1+c2)a1c1c2)2a2a0+a21,α2=2τ0(μ(a3(c1+c2)a2c1c2)a2(c1+c2)+a3c1c2+a1)+2a0+a222a3a1,α3=2τ0(μ+c1+c2)+a322(a3τ0+a2).

    Let z=b2. Then (7) becomes

    z4+α3z3+α2z2+α1z+α0=0. (8)

    By the Routh--Hurwitz criterion, (8) has no positive real roots if αi>0, i=0,,3, α3α2>α1, and α3α2α1>α21+α23α0.

    For the parameter values of Table 1 and β=40 (R0=0.880827), the Routh-Hurwitz criterion does not hold. In fact, equation (8) takes the form

    z4+241.429794z3+31.065028z2221.270089z0.037233=0 (9)
    Table 1. Parameter values.
    SymbolDescriptionValue
    βTransmission coefficient[50,150]
    μDeath and birth rate1/70yr1
    δRate at which individuals leave L112yr1
    ϕProportion of individuals going to I0.05
    ωEndogenous reactivation rate for persistent latent infections0.0002yr1
    ωREndogenous reactivation rate for treated individuals0.00002yr1
    σFactor reducing the risk of infection as a result of acquired
    immunity to a previous infection for L2
    0.25
    σRRate of exogenous reinfection of treated patients0.25
    τ0Rate of recovery under treatment of active TB2yr1
    τ1Rate of recovery under treatment of early latent individuals L12yr1
    τ2Rate of recovery under treatment of persistent latent individuals L21yr1
    NTotal population30,000
    TTotal simulation duration5 yr
    ϵ1Efficacy of treatment of early latent L10.5
    ϵ2Efficacy of treatment of persistent latent TB L20.5
     | Show Table
    DownLoad: CSV

    and we immediately see that the coefficient α1=221.270089 is not positive. Moreover, the roots of equation (9) are approximately 0.89, 0.00017, 1.03, 241.30, therefore there exists a positive imaginary root given by 0.89i. Thus, there exists at least one time delay such that the disease free equilibrium is unstable. We have just proved the following result.

    Lemma 3.3 Let R0<1. Then there exists at least one positive time delay dI>0 such that the disease free equilibrium (N,0,0,0) is unstable.

    Remark 2.Observe that there may exist specific time delays for which the disease free equilibrium (N,0,0,0) is locally asymptotically stable when R0<1. As we show next, this is the case for the time delay dI=0.1. Indeed, consider the parameter values of Table 1 and β such that R0<1. For example, let β=40, for which R0=0.880827. The results are given in Figure 1, where to compute the trajectories we have used the MATLAB routine dde23, which solves delay differential equations with constant delays. For the theoretical results that underlie this solver we refer to [21]. The characteristic equation (6) takes the form χ(λ)=0 with

    χ(λ)=λ4+17.057363λ3+20.733305λ2+4.489748λ+0.048755+2(λ+170)(λ+98170)(λ+1.014486)(eλ0.11).
    Figure 1. Disease free equilibrium with basic reproduction number R0=0.88 (β=40, dI=0.1 and the other values from Table 1).

    The derivative χ(λ) has exactly four zeros: ˉλ1=21.408183, ˉλ2=12.680307, ˉλ3=0.748206, and ˉλ4=0.151719. For all λ],ˉλ1[ one has χ(λ)>0 and since limλχ(λ)= and χ(ˉλ1)>0, we conclude that there exists a unique λ1 in the interval ],ˉλ1[ such that χ(λ1)=0. Analogously, we prove that there exists exactly five roots of χ(λ)=0: λ1=23.481727, λ2=18.106597, λ3=1.024343, λ4=0.320880, λ5=0.011482. We conclude that there are no positive real roots and that the disease free equilibrium (N,0,0,0) is locally asymptotically stable for dI=0.1 and the considered parameter values.

    In this paper we assume that the time delay dI associated to the diagnosis of active TB is equal to 0.1, that is, 36.5 days. This value makes sense from the epidemiological point of view, since it fits in the intervals available in the literature for the delay in the diagnosis of active TB. For instance, in [24] the reported overall patient delay is similar to the health system delay for diagnosis of active TB, 31.03 and 27.2 days, respectively. The average (median or mean) patient delay and health system delay range from 4.9 to 162 days and 2 to 87 days, respectively, in both low and middle income countries and high income countries.


    4. Stability of the endemic equilibrium

    System (1) has an unique endemic equilibrium such that I(t)>0 for any t>0. The analytic expression is cumbersome and not useful for our purposes. Consider the parameter values from Table 1 with β=100. Then the basic reproduction number is R0=2.202067. The endemic equilibrium E is given by I=11.006448, L1=36.111397, L2=402.155827, S=8407.668384. The matrices A1 and A2 associated to the linearized system (5) at the endemic equilibrium E are computed as

    A1=[0.050974028.02556100.02751614.02345845.97073400.0000200.5999800.0143060.000180011.4000000.3351301.023658]

    and A2=diag(0,0,2,0). The transcendental characteristic equation is given by

    λ4+15.112395λ312.243801λ228.331139λ0.966336+(30.196179λ2+30.244462λ+2λ3+1.463482)eλdI=0. (10)

    When dI=0, we have the following characteristic equation:

    λ4+17.112395λ3+17.952378λ2+1.913323λ+0.497146=0. (11)

    The roots of (11) are 1.029896, 15.997555, 0.0424720.168435i, 0.042472+0.168435i. All the roots of (11) have negative real part, thus the endemic equilibrium E is asymptotical stable. Consider now the case dI>0 and suppose that (10) has a purely imaginary root bi, with b>0. Separating the real and imaginary parts in (10), we have

    b8+281.828573b651.906667b41.236501b20.000246=0. (12)

    It is easy to verify that b=0.453220 is a root of equation (12). Thus, by Rouché's theorem, there exists at least a time delay dI>0 such that the endemic equilibrium E is unstable. In the specific case dI=0.1, the characteristic equation is given by

    λ4+15.112395λ312.243801λ228.331139λ0.966336+(2λ3+30.196179λ2+30.244462λ+1.463482)e0.1λ=0. (13)

    Similarly to Remark 2, it follows from Bolzano's theorem and the monotonicity of the characteristic function associated to (13) that all roots of equation (13) have a negative real part. Therefore, the endemic equilibrium E is locally asymptotically stable for dI=0.1 and R0>1.


    5. Optimal control of a tuberculosis model with state and control delays

    We now consider the TB model (1) with a time delay in the state variable I(t) and introduce two control functions u1() and u2() and two real positive model constants ϵ1 and ϵ2. The control u1 represents the effort on early detection and treatment of recently infected individuals L1 and the control u2 represents the application of chemotherapy or post-exposure vaccine to persistent latent individuals L2. The parameters ϵi(0,1), i=1,2, measure the effectiveness of the controls ui, i=1,2, respectively, i.e., these parameters measure the efficacy of post-exposure interventions for early and persistent latent TB individuals, respectively. Since after TB infection the human immune system can take from 2 to 8 weeks to react, it takes at least the same time to detect the infection by the medical test [33]. Hence, we introduce a time delay du1 in the control u1 which represents the delay in the diagnosis of latent TB and commencement of latent TB treatment. The prophylactic treatment of persistent latent individuals may also suffer from a delay due to personal reasons of the patient, who may be resistant to treatment having spent more than two years with latent infection without passing to active disease, or the delay may be caused by priorities given to early latent and active infectious individuals from the health care system. Based on these facts, for numerical simulations we shall consider the following delays with bounds on the control delays:

    dI=0.1,du1,du2[0.05,0.2]. (14)

    The resulting model is given by the following system of nonlinear ordinary delay differential equations:

    {˙S(t)=μNβNI(t)S(t)μS(t),˙L1(t)=βNI(t)(S(t)+σL2(t)+σRR(t))(δ+τ1+ϵ1u1(tdu1)+μ)L1(t),˙I(t)=ϕδL1(t)+ωL2(t)+ωRR(t)τ0I(tdI)μI(t),˙L2(t)=(1ϕ)δL1(t)σβNI(t)L2(t)(ω+ϵ2u2(tdu2)+τ2+μ)L2(t). (15)

    Recall that the recovered population is determined by R(t)=N(S(t)+L1(t)+I(t)+L2(t)), which formally gives the equation

    ˙R(t)=τ0I(tdI)+(τ1+ϵ1u1(tdu1))L1(t)+(τ2+ϵ2u2(tdu2))L2(t)σRβNI(t)R(t)(ωR+μ)R(t).

    Note, however, that this equation is not needed in the subsequent optimal control computations. We prescribe the following initial conditions for the state variables (S,L1,L2) and, due to the delays, initial functions for the state variable I and controls u1 and u2:

    S(0)=(76/120)N,L1(0)=(36/120)N,L2(0)=(2/120)N,R(0)=(1/120)N,I(t)=(5/120)NfordIt0,uk(t)=0fordukt<0(k=1,2). (16)

    In the case du1=du2=0 of no control delays, the last condition is void. The following box constraints are imposed on the control variables:

    0uk(t)1t[0,T](k=1,2). (17)

    Let us denote the state and control variable of the control system (15), respectively, by x=(S,L1,I,L2)R4 and u=(u1,u2)R2. We shall consider two types of objectives: either the L1-type objective

    J1(x,u)=T0(I(t)+L2(t)+W1u1(t)+W2u2(t))dt, (18)

    which is linear in the control variable u, or the L2-type objective

    J2(x,u)=T0(I(t)+L2(t)+W1u21(t)+W2u22(t))dt, (19)

    which is quadratic in the control variable. In both objectives, W1>0, W2>0 are appropriate weights to be chosen later. L2-type functionals like (19) are often used in economics to describe, e.g., productions costs, but are not appropriate in a biological framework; cf. the remarks in [20]. The L1 functional J1(x,u) incorporates the total amount of drug used as a penalty and thus appears to be more realistic. For that reason, we shall mainly focus on the functional J1(x,u).

    The optimal control problem then is defined as follows: determine a control function u=(u1,u2)L1([0,T],R2) that minimizes either the cost functional J1(x,u) in (18) or J2(x,u) in (19) subject to the dynamic constraints (15), initial conditions (16) and control constraints (17). Necessary optimality conditions for optimal control problems with multiple time delays in control and state variables may be found, e.g., in [10]. Here, we discuss the Maximum Principle in order to display the controls and the switching functions in a convenient way. To define the Hamiltonian for the delayed control problem, we introduce the delayed state variable y3(t)=x3(tdI)=I(tdI) and the delayed control variables vk(t)=uk(tdu1), k=1,2. Using the adjoint variable λ=(λS,λL1,λI,λL2)R4, the Hamiltonian for the objective J1 and the control system (15) is given by

    H(x,y3,λ,u1,u2,v1,v2)=(I+L2+W1u1+W2u2)+λS(μNβNISμS)+λL1(βN(S+σL2+σRR)(δ+τ1+ϵ1v1+μ)L1)+λI(ϕδL1+ωL2+ωRRτ0y3μI)+λL2((1ϕ)δL1σβNIL2(ω+ϵ2v2+τ2+μ)L2).

    We obtain the adjoint equations ˙λS(t)=HS[t], ˙λL1(t)=HL1[t], ˙λL2(t)=HL2[t], and ˙λI(t)=HI[t]+χ[0,TdI]Hy3[t+dI], where subscripts denote partial derivatives and χ[0,TdI] is the characteristic function in the interval [0,TdI] [10]. Note that only the equation for ˙λI(t) contains the advanced time t+dI. Since the terminal state x(T) is free, the transversality conditions are

    λS(T)=λL1(T)=λI(T)=λL2(T)=0. (20)

    To characterize the optimal controls u1 and u2, we introduce the following switching functions for k=1,2:

    ϕk(t)=Huk[t]+χ[0,Tduk](t+duk)Hvk[t+duk]={WkϵkλLk(t+duk)Lk(t+duk)for0tTduk,WkforTduktT. (21)

    Then the maximum condition for the optimal controls u1(t), u2(t) is equivalent to the maximization ϕk(t)uk(t)=max0uk1ϕk(t)uk, k=1,2, which gives the control law

    uk(t)={1ifϕk(t)>0,0ifϕk(t)<0,singularifϕk(t)=0onIs[0,T],k=1,2. (22)

    We do not discuss singular controls further, since both in the non-delayed and the delayed control problem we did not find singular arcs. In view of the transversality conditions (20), the terminal value of the switching function is ϕk(T)=Wk for k=1,2. Hence, we may conclude from the control law (22) that u1(t)=u2(t)=0 holds on a terminal interval.


    6. Numerical results for the non-delayed and the delayed optimal control problem

    We choose the numerical approach "First Discretize then Optimize" to solve both the non-delayed and delayed optimal control problem. The discretization of the control problem on a fine grid leads to a large-scale nonlinear programming problem (NLP) that can be conveniently formulated with the help of the Applied Modeling Programming Language AMPL [9]. AMPL can be linked to several powerful optimization solvers. We use the Interior-Point optimization solver IPOPT developed by Wächter and Biegler [30]. Details of discretization methods for delayed control problems may be found in [10]. The subsequent computations for the terminal time tf=5 have been performed with N=2500 to N=5000 grid points using the trapezoidal rule as integration method. Choosing the error tolerance tol=1010 in IPOPT, we can expect that the state variables are correct up to 7 or 8 decimal digits. Since Lagrange multipliers are computed a posteriori in IPOPT, we cannot expect more than 6 correct decimal digits in the adjoint variables.

    Also, the control package NUDOCCCS developed by Büskens [3] (cf. also [4]) provides a highly efficient method for solving the discretized control problem, because it allows to implement higher order integration methods. However, so far NUDOCCCS can only be implemented for non-delayed control problems. For the non-delayed TB control problem, we obtained only bang-bang controls. An important feature of NUDOCCCS is the fact that it provides an efficient method for optimizing the switching times of bang-bang controls using the arc-parametrization method [16]. This approach is called the Induced Optimization Problem (IOP) for bang-bang controls. NUDOCCCS then allows for a check of second-order sufficient conditions of the IOP, whereby the second-order sufficient conditions for bang-bang controls can be verified with high accuracy; cf. [16,17].


    6.1. Optimal control solution of the non-delayed TB model

    First, we consider the optimal control of non-delayed TB model, where formally we put dI=du1=du2=0. The numerical solutions serve as reference solutions, which later will be compared with the solutions to the delayed control problem. We choose the weights W1=W2=50 in the objective (18) and the parameter β=100 (R0=2.2) in Table 1. The discretization approach shows that controls uk(t) are bang-bang and with only one switch at tk, k=1,2:

    uk(t)={1for0ttk,0fortk<tT,k=1,2. (23)

    To obtain a refined solution, we solve the IOP with respect to the switching times t1 and t2 using the arc-parametrization method [16] and the code NUDOCCCS [3]. We get the following numerical results:

    J1(x,u)=28390.73,t1=3.677250,t2=4.866993,S(T)=1034.634,L1(T)=53.59586,I(T)=25.89556,L2(T)=780.7667,R(T)=28105.11. (24)

    The initial value of the adjoint variable λ=(λs,λL1,λI,λL2) is computed as

    λ(0)=(0.376159,0.452761,4.03059,0.394839).

    The control and state trajectories are displayed in Figure 2. The Hessian HessL of the Lagrangian for the IOP is positive definite: HessL=(453.98387.42387.42385.69)>0. It can be seen from Figure 2, top row (a) and bottom row (a), that the switching functions satisfy the strict bang-bang property (cf. [16,17]) corresponding to the Maximum Principle: ϕk(t)>0 for 0t<tk, ˙ϕk(tk)<0, ϕk(t)<0 for tk<tT(k=1,2). Hence, the bang-bang controls (23) characterized by the data (24) satisfies the second-order sufficient conditions in [17,Chap. 7] and thus provides a strict strong minimum. Figure 3 displays the comparison of the optimal controls for the functionals J1(x,u) (18) and J2(x,u) (19). The state variables are nearly identical, since the control variables differ only a terminal interval. Also the objective values are very close: J1(x,u)=28390.73, J2(x,u)=28382.37. Note that the controls for J2(x,u) are continuous, since the strict Legendre-Clebsch condition holds and the Hamiltonian has a unique minimum. The proof, that second-order sufficient conditions (SSC) are satisfied for the controls corresponding to J2(x,u), is quite elaborated since it requires to test whether an associated matrix Riccati equation has a bounded solution; cf. [17]. When we increase the weights W1 and W2, the control u1 stays to be bang-bang with only one switching time which gets smaller, whereas the control u2 may have an additional zero arc at the beginning. E.g., for W1=W2=150 we obtain the controls

    (u1(t),u2(t))={(1,0)for0tt1,(1,1)fort1<tt2,(0,1)fort2<tt3,(0,0)fort3<tT,k=1,2.
    Figure 2. Optimal control and state variables of the non-delayed TB model with L1 objective (18) and weights W1=W2=50. Top row: (a) control u1 (23) and (scaled) switching function ϕ1 (21) satisfying the control law (22) for k=1, (b) susceptible individuals S and recovered individuals R, (c) infectious individuals I. Bottom row: (a) control u2 (23) and (scaled) switching function ϕ2 (21) satisfying the control law (22) for k=2, (b) early latent L1, (c) persistent latent L2.
    Figure 3. Comparison of controls u1 and u2 for the L1-type objective (18) and L2-type objective (19) with weights W1=W2=50.

    The objective value and the switching times are computed as J1(x,u)=29175.97, t1=0.00260, t2=2.662, and t3=4.633. The optimal controls are shown in Figure 4.

    Figure 4. Optimal controls u1 and u2 for the L1-type objective (18) with weights W1=W2=150.

    6.2. Optimal control solution of the TB model with control and state delays

    To see more distinctively the difference between delayed and non-delayed solutions, we consider state and control delays with values at their upper bounds in (14), that is, dI=0.1, du1=0.2, and du2=0.2. Again, we choose the weights W1=W2=50 in the objective (18) and the parameter β=100 in Table 1, for which R0=2.2. The discretization approach with N=5000 grid points and the trapezoidal rule as integration method yields the following bang-bang controls with only one switch as in the non-delayed case:

    uk(t)={1for0ttk,0fortk<tT,k=1,2. (25)

    We obtain the numerical results J1(x,u)=26784.60, t1=3.108, t2=4.581, S(T)=1234.598, L1(T)=24.93928, I(T)=11.71451, L2(T)=469.8865, and R(T)=28258.86. However, in contrast to the non-delayed case, we are not able to optimize the switching times directly because the time-transformation in the arc-parametrization method [16] cannot be applied to the delayed problem. The initial value of the adjoint variable is computed as λ(0)=(0.3789,0.4682,3.6412,0.4263). When comparing the results for the delayed problem with those in (24) for the non-delayed problem, we notice the surprising fact that the terminal values L1(T), I(T), L2(T) and the switching times t1, t2 are significantly smaller in the delayed problem on the expense that the terminal value S(T) is significantly higher. As in the non-delayed problem, the switching functions satisfy the strict bang-bang property related to the Maximum Principle:

    ϕk(t)>0for0t<tk,˙ϕk(tk)<0,ϕk(t)<0fortk<tT(k=1,2).

    However, we are not aware in the literature of any type of sufficient conditions which could be applied to the extremal solution shown in Figure 5.

    Figure 5. Optimal control and state variables of the delayed TB model with L1-objective (18), W1=W2=50 and delays dI=0.1,du1=du2=0.2. Top row: (a) control u1 (25) and (scaled) switching function ϕ1 (21) satisfying the control law (22) for k=1, (b) susceptible individuals S and recovered individuals R, (c) infectious individuals I. Bottom row: (a) control u2 (25) and (scaled) switching function ϕ2 (21) satisfying the control law (22) for k=2, (b) early latent L1, (c) persistent latent L2.

    We also compared the extremal solutions for the L1-type objective (18) and the L2-type objective (19). Since the controls are very similar to those in Figure 3, they are not displayed here. Figure 6 shows the extremal controls for the L1-objective (18) for the increased weights W1=W2=150.

    Figure 6. Extremal controls for the delayed TB model with L1 objective (18), W1=W2=150 and delays dI=0.1, du1=du2=0.2. (a) control u1 (25) and (scaled) switching function ϕ1 (21) satisfying the control law (22) for k=1, (b) control u2 (25) and (scaled) switching function ϕ2 (21) satisfying the control law (22) for k=2.

    The most significant influence on the optimal controls is exerted by the transmission coefficient β. It can be clearly seen that the increase of the transmission coefficient β triggers a substantial increase in the switching times tk of the bang-bang controls uk for k=1,2 (cf. Figure 7) as could be expected from the equation for ˙S. Let us perform in this case a more detailed sensitivity analysis of the trajectories with respect to the parameter β[50,150]. We compute the extremal solutions for a sufficiently fine grid of parameters β[50,150] by a homotopy method and plot the objective J1(x,u) and the terminal state (S(T),L1(T),I(T),L2(T),R(T)) as a function of β. The numerical results are shown in Figure 8.

    Figure 7. Comparison of extremal controls for parameters β=50 and β=150 in the delayed TB model with L1 objective (18), weights W1=W2=150 and delays dI=0.1,du1=du2=0.2.
    Figure 8. Homotopic solutions of the delayed TB model with L1 objective (18) and weights W1=W2=50 for parameters β[50,150]. Displayed are the objective value J1(x,u) and the terminal states S(T), R(T), I(T), L1(T), L2(T).

    7. Conclusions

    We introduced a discrete time delay dI in a mathematical model for tuberculosis (TB), which represents the delay on the diagnosis of active TB infection and commencement of treatment. The delay on the diagnosis of active TB has important negative consequences on TB control and eradication. The later the treatment of active TB starts, the more people can be infected and may die from TB. The introduction of a time delay on the state variable of active TB infected individuals I is therefore justified from the epidemiological point of view.

    When a time delay is introduced into a mathematical model, the stability of its disease free and endemic equilibriums may change. We proved that the disease free equilibrium (DFE) of the TB model with delay in the state variable I is unstable for any time delay dI0, whenever the basic reproduction number is greater than one. We derived conditions under which the model is locally asymptotically stable for dI=0 and proved that there exists at least one positive time delay dI>0 such that the DFE is unstable for R0<1. Despite of this, we also proved that for the concrete time delay dI=0.1, the set of parameters of Table 1, and a transmission coefficient for which R0<1, the DFE is locally asymptotically stable. The value dI=0.1 (36.5 days) fits the data reported in the literature for the delay in the diagnosis of active TB. For the endemic equilibrium (EE), we considered that the transmission coefficient β is such that R0>1 and proved the local stability of the specific EE associated to our set of parameters and discrete time delay dI=0.1.

    We proposed an optimal control problem where the control system is the mathematical model for TB with time delay in the state variable I and where the control functions u1 and u2 represent, respectively, the effort on early detection and treatment of recently infected individuals L1, and the application of prophylactic treatment to persistent latent individuals L2. The objective is to minimize the number of individuals with active and persistent latent TB as well as the cost associated to the implementation of the control measures. Human immune system can take from two to eight weeks to react to TB infection and detection. Moreover, prophylactic treatment of early and persistent latent individuals may face resistance from patients and health care services. Based on these facts, we introduced discrete time delays in the controls u1 and u2. To our knowledge, this is the first time an optimal control problem for TB with delays in the state and control variables is investigated.

    Firstly, we considered the non-delayed case (dI=du1=du2=0) and compared the solutions for L1 and L2 objectives. Our results show that the optimal state variables are nearly identical for both objectives, the control variables differing only on a terminal interval. The optimal values of the objective functionals are also very close. We observed that for the L1 objective the optimal control u2 may have an additional zero arc at the beginning if the weight constants associated to the implementation of the control policies u1 and u2 are big enough. For the delayed optimal control problem, our focus was on the L1 objective since it seems to be the more realistic and, comparing the extremal solutions of L1 and L2 objectives, the differences are similar to the non-delayed case. In the delayed case, the switching times t1 and t2 are significantly smaller and the costs are also smaller, when compared to the non-delayed case. From an epidemiological point of view, when we introduce delays in the TB model, the optimal solutions for the reduction of the number of individuals with active and persistent latent TB infection are associated to control policies that are less costly and can be implemented in a shorter period of time. Associated to these positive facts, we observed that if a delay is introduced in the state variable I and in the controls, the number of individuals with active TB infection at the end of five years is reduced in approximately 45.2 per cent, when compared to the non-delayed case. Similarly, the number of persistent latent individuals at the end of five years is also reduced in approximately 60.1 per cent. Moreover, the terminal number of susceptible and recovered individuals is bigger in the delayed case. Through a sensitivity analysis, we observed that the transmission coefficient β has a significant influence on the optimal controls and the cost functional, and that the number of active infected individuals and the number of early and persistent latent individuals also increase with β. The number of recovered individuals increases for β[50,100] and decreases for β[100,150], which means that for R0>3.1 the control measures are no longer effective and should be rethought.


    Acknowledgments

    This work was partially supported by FCT within project TOCCATA, reference PTDC/EEI-AUT/2933/2014. Silva and Torres were also supported by CIDMA and project UID/MAT/04106/2013; Silva by the post-doc grant SFRH/BPD/72061/2010. The authors are very grateful to anonymous reviewers for their careful reading and helpful comments.


    [1] Kroeker KJ, Micheli F, Gambi MC, et al. (2011) Divergent ecosystem responses within a benthic marine community to ocean acidification. PNAS 108: 14515-14520. doi: 10.1073/pnas.1107789108
    [2] Fabricius KE, Langdon C, Uthicke S, et al. (2011) Losers and winners in coral reefs acclimatized to elevated carbon dioxide concentrations. Nature Climate Change 1: 165-169. doi: 10.1038/nclimate1122
    [3] Hale R, Calosi P, McNeill L, et al. (2011) Predicted levels of future ocean acidification and temperature rise could alter community structure and biodiversity in marine benthic communities. Oikos 120: 661-674. doi: 10.1111/j.1600-0706.2010.19469.x
    [4] Fabry VJ, Seibel BA, Feely RA, et al. (2008) Impacts of ocean acidification on marine fauna and ecosystem processes. ICES Journal of Marine Science: J du Conseil 65: 414-432. doi: 10.1093/icesjms/fsn048
    [5] Caldeira K, Wickett ME (2003) Oceanography: anthropogenic carbon and ocean pH. Nature 425: 365-365. doi: 10.1038/425365a
    [6] Raven J, Caldeira K, Elderfield H, et al. (2005) Ocean acidification due to increasing atmospheric carbon dioxide: The Royal Society.
    [7] Hönisch B, Ridgwell A, Schmidt DN, et al. (2012) The geological record of ocean acidification. Science 335: 1058-1063. doi: 10.1126/science.1208277
    [8] Orr JC, Fabry VJ, Aumont O, et al. (2005) Anthropogenic ocean acidification over the twenty-first century and its impact on calcifying organisms. Nature 437: 681-686. doi: 10.1038/nature04095
    [9] Kleypas JA, Buddemeier RW, Archer D, et al., (1999) Geochemical consequences of increased atmospheric carbon dioxide on coral reefs. Science 284: 118-120. doi: 10.1126/science.284.5411.118
    [10] IPCC, Climate Change 2013: The Physical Science Basis: Working Group I Contribution to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. 2014: Cambridge University Press.
    [11] Walther GR, Post E, Convey P, et al. (2002) Ecological responses to recent climate change. Nature 416: 389-395. doi: 10.1038/416389a
    [12] Hofmann GE, Todgham AE (2010) Living in the now: Physiological mechanisms to tolerate a rapidly changing environment. Annu Rev Physiol 72: 127-145. doi: 10.1146/annurev-physiol-021909-135900
    [13] Pespeni MH, Sanford E, Gaylord B, et al. (2013) Evolutionary change during experimental ocean acidification. PNAS 110: 6937-6942. doi: 10.1073/pnas.1220673110
    [14] Kelly MW, Hofmann GE (2013) Adaptation and the physiology of ocean acidification. Funct Ecol 27: 980-990. doi: 10.1111/j.1365-2435.2012.02061.x
    [15] Bell G, Collins S (2008) Adaptation, extinction and global change. Evol Appl 1: 3-16. doi: 10.1111/j.1752-4571.2007.00011.x
    [16] Gienapp P, Teplitsky C, Alho J, et al. (2008) Climate change and evolution: disentangling environmental and genetic responses. Mol ecol 17: 167-178. doi: 10.1111/j.1365-294X.2007.03413.x
    [17] Hoegh-Guldberg O, Mumby P, Hooten A, et al. (2007) Coral reefs under rapid climate change and ocean acidification. science 318: 1737-1742. doi: 10.1126/science.1152509
    [18] Parker L, Ross PM, O'Connor WA (2011) Populations of the Sydney rock oyster, Saccostrea glomerata, vary in response to ocean acidification. Mar Biol 158: 689-697. doi: 10.1007/s00227-010-1592-4
    [19] Hofmann GE, Evans TG, Kelly MW, et al. (2014) Exploring local adaptation and the ocean acidification seascape–studies in the California Current Large Marine Ecosystem. Biogeosci 11: 1053-1064.
    [20] Prosser CL (1991) Comparative Animal Physiology, Environmental and Metabolic Animal Physiology: Wiley-Liss.
    [21] Somero G (2005) Linking biogeography to physiology: Evolutionary and acclimatory adjustments of thermal limits. Front Zool 2: 1. doi: 10.1186/1742-9994-2-1
    [22] Somero GN (2002) Thermal physiology and vertical zonation of intertidal animals: optima, limits, and costs of living. Integr comp biol 42: 780-789. doi: 10.1093/icb/42.4.780
    [23] Stillman JH, Somero GN (2000) A comparative analysis of the upper thermal tolerance limits of eastern Pacific porcelain crabs, genus Petrolisthes: influences of latitude, vertical zonation, acclimation, and phylogeny. Physiol Biochem Zool 73: 200-208. doi: 10.1086/316738
    [24] Prosser C (1955) Physiological variation in animals. Biol Rev 30: 229-261.
    [25] Lannig G, Eilers S, Portner HO, et al. (2010) Impact of ocean acidification on energy metabolism of oyster, Crassostrea gigas--changes in metabolic pathways and thermal response. Mar Drugs 8: 2318-2339. doi: 10.3390/md8082318
    [26] Randall D, Burggren WW, French K (2002) Eckert animal physiology: Macmillan.
    [27] Alberts B, Bray D, Lewis J, et al. (2008) Molecular biology of the cell. Garland, New York: 139-194.
    [28] Comeau S, Jeffree R, Teyssia JL, et al. (2010) Response of the Arctic pteropod Limacina helicina to projected future environmental conditions. PLoS ONE 5: e11362-e11362. doi: 10.1371/journal.pone.0011362
    [29] Thomsen Jr, Melzner F (2010) Moderate seawater acidification does not elicit long-term metabolic depression in the blue mussel Mytilus edulis. Mar Biol 157: 2667-2676. doi: 10.1007/s00227-010-1527-0
    [30] Stumpp M, Wren J, Melzner F, et al. (2011) CO 2 induced seawater acidification impacts sea urchin larval development I: elevated metabolic rates decrease scope for growth and induce developmental delay. Comparative Biochemistry and Physiology Part A: Molecular & Integrative Physiology 160: 331-340.
    [31] Hand SC (1991) Metabolic dormancy in aquatic invertebrates. Advances in comparative and environmental physiology: Springer. 1-50.
    [32] Langenbuch M, Pörtner HO (2002) Changes in metabolic rate and N excretion in the marine invertebrate Sipunculus nudus under conditions of environmental hypercapnia identifying effective acid-base variables. J Exp Biol 205: 1153-1160.
    [33] Pörtner H, Langenbuch M, Reipschläger A (2004) Biological Impact of Elevated Ocean CO2 concentrations: lessons from animal physiology and earth history. J Oceanogr 60: 705-718. doi: 10.1007/s10872-004-5763-0
    [34] Pan TCF, Applebaum SL, Manahan DT (2015) Experimental ocean acidification alters the allocation of metabolic energy. PNAS 112: 4696-4701. doi: 10.1073/pnas.1416967112
    [35] Bennett A (1991) The evolution of activity capacity. J Exp Biol 160: 1-23.
    [36] Seibel BA, Drazen JC (2007) The rate of metabolism in marine animals: environmental constraints, ecological demands and energetic opportunities. Philosophical Transactions of the Royal Society B: Biological Sciences 362: 2061-2078. doi: 10.1098/rstb.2007.2101
    [37] Melzner F, Gutowska M, Langenbuch M, et al. (2009) Physiological basis for high CO2 tolerance in marine ectothermic animals: pre-adaptation through lifestyle and ontogeny? Biogeosci 6: 2313-2331.
    [38] Seibel BA, Walsh PJ (2001) Potential Impacts of CO2 Injection on Deep-Sea Biota. Science 294: 319-320. doi: 10.1126/science.1065301
    [39] Melzner F, Gutowska MA, Langenbuch M, et al. (2009) Physiological basis for high CO2 tolerance in marine ectothermic animals: pre-adaptation through lifestyle and ontogeny? Biogeosciences Discuss 6.
    [40] Seibel BA, Walsh PJ (2003) Biological impacts of deep-sea carbon dioxide injection inferred from indices of physiological performance. J Exp Biol 206: 641-650. doi: 10.1242/jeb.00141
    [41] Widdicombe S, Spicer JI (2008) Predicting the impact of ocean acidification on benthic biodiversity: What can animal physiology tell us? J Exp Mar Biol Ecol 366: 187-197. doi: 10.1016/j.jembe.2008.07.024
    [42] Kurihara H (2008) Effects of CO2-driven ocean acidification on the early developmental stages of invertebrates. Mar Ecol Prog Ser 373: 275-284.
    [43] Peñuelas J, Estiarte M (1998) Can elevated CO 2 affect secondary metabolism and ecosystem function? Trends ecol evol 13: 20-24. doi: 10.1016/S0169-5347(97)01235-4
    [44] Sanford E, Gaylord B, Hettinger A, et al. (2014) Ocean acidification increases the vulnerability of native oysters to predation by invasive snails. P Roy Soc B: Biol Sci 281: 20132681.
    [45] Kroeker KJ, Micheli F, Gambi MC (2013) Ocean acidification causes ecosystem shifts via altered competitive interactions. Nature Climate Change 3: 156-159.
    [46] Osenberg CW, Sarnelle O, Cooper SD (1997) Effect size in ecological experiments: the application of biological models in meta-analysis. Am Nat 150: 798-812. doi: 10.1086/286095
    [47] Sorte CJ, Ibáñez I, Blumenthal DM, et al. (2013) Poised to prosper? A cross-system comparison of climate change effects on native and non-native species performance. Ecol Lett 16: 261-270.
    [48] Gibson R, Atkinson R, Gordon J, et al. (2011) Impact of ocean warming and ocean acidification on marine invertebrate life history stages: vulnerabilities and potential for persistence in a changing ocean. Oceanogr Mar Biol Annu Rev 49: 1-42.
    [49] Nakamura M, Ohki S, Suzuki A, et al. (2011) Coral larvae under ocean acidification: survival, metabolism, and metamorphosis. PLoS ONE 6: e14521-e14521. doi: 10.1371/journal.pone.0014521
    [50] Pimentel M, Pegado M, Repolho T, et al. (2014) Impact of ocean acidification in the metabolism and swimming behavior of the dolphinfish (Coryphaena hippurus) early larvae. Mar Biol 161: 725-729. doi: 10.1007/s00227-013-2365-7
    [51] Rivest EB, Hofmann GE (2014) Responses of the Metabolism of the Larvae of Pocillopora damicornis to Ocean Acidification and Warming. PloS one 9: e96172. doi: 10.1371/journal.pone.0096172
    [52] Albright R (2011) Reviewing the effects of ocean acidification on sexual reproduction and early life history stages of reef-building corals. J Mar Biol 2011.
    [53] Storey KB, Storey JM (2004) Oxygen limitation and metabolic rate depression. Functional metabolism: regulation and adaptation: 415-442.
    [54] Langenbuch M, Pörtner H-O (2004) High sensitivity to chronically elevated CO 2 levels in a eurybathic marine sipunculid. Aquat Toxicol 70: 55-61. doi: 10.1016/j.aquatox.2004.07.006
    [55] Hand SC, Hardewig I (1996) Downregulation of cellular metabolism during environmental stress: mechanisms and implications. Annu rev physiol 58: 539-563. doi: 10.1146/annurev.ph.58.030196.002543
    [56] Beckerman A, Benton TG, Ranta E, et al. (2002) Population dynamic consequences of delayed life-history effects. Trends ecol evol 17: 263-269. doi: 10.1016/S0169-5347(02)02469-2
    [57] Green MA, Jones ME, Boudreau CL, et al. (2004) Dissolution mortality of juvenile bivalves in coastal marine deposits. Limnol Oceanogr 49: 727-734. doi: 10.4319/lo.2004.49.3.0727
    [58] Hofmann GE, Smith JE, Johnson KS, et al. (2011) High-frequency dynamics of ocean pH: a multi-ecosystem comparison. PloS one 6: e28983. doi: 10.1371/journal.pone.0028983
    [59] Kapsenberg L, Kelley AL, Shaw EC, et al. (2015) Near-shore Antarctic pH variability has implications for the design of ocean acidification experiments. Sci Rep 5.
    [60] Evans TG, Chan F, Menge BA, et al. (2013) Transcriptomic responses to ocean acidification in larval sea urchins from a naturally variable pH environment. Mol ecol 22: 1609-1625. doi: 10.1111/mec.12188
    [61] Clarke A (2003) Costs and consequences of evolutionary temperature adaptation. Trends ecol evol 18: 573-581. doi: 10.1016/j.tree.2003.08.007
    [62] Albright R, Langdon C (2011) Ocean acidification impacts multiple early life history processes of the Caribbean coral Porites astreoides. Global Change Biol 17: 2478-2487.
    [63] Byrne M, Przeslawski R (2013) Multistressor impacts of warming and acidification of the ocean on marine invertebrates' life histories. Integr comp biol 53: 582-596. doi: 10.1093/icb/ict049
    [64] Maas A, Wishner K, Seibel B (2012) The metabolic response of pteropods to acidification reflects natural CO2/sub-exposure in oxygen minimum zones. Biogeosci 9: 747-757. doi: 10.5194/bg-9-747-2012
    [65] Henry RP, Wheatly MG (1992) Interaction of respiration, ion regulation, and acid-base balance in the everyday life of aquatic crustaceans. Am Zool 32: 407-416. doi: 10.1093/icb/32.3.407
    [66] Waldbusser G, Voigt E, Bergschneider H, et al. (2011) Biocalcification in the Eastern Oyster (Crassostrea virginica) in Relation to Long-term Trends in Chesapeake Bay pH. Estuar Coast 34: 221-231.
    [67] Waldbusser GG, Hales B, Langdon CJ, et al. (2015) Ocean acidification has multiple modes of action on bivalve larvae. PloS one 10: e0128376. doi: 10.1371/journal.pone.0128376
    [68] Flynn EE, Bjelde BE, Miller NA, et al. (2015) Ocean acidification exerts negative effects during warming conditions in a developing Antarctic fish. Conservation Physiology 3: cov033. doi: 10.1093/conphys/cov033
    [69] Todgham AE, Hofmann GE (2009) Transcriptomic response of sea urchin larvae Strongylocentrotus purpuratus to CO2-driven seawater acidification. J Exp Biol 212: 2579-2594. doi: 10.1242/jeb.032540
    [70] Evans TG, Watson-Wynn P (2014) Effects of seawater acidification on gene expression: resolving broader-scale trends in sea urchins. Biol Bull 226: 237-254.
    [71] Kaniewska P, Campbell PR, Kline DI, et al. (2012) Major cellular and physiological impacts of ocean acidification on a reef building coral. PloS one 7: e34659. doi: 10.1371/journal.pone.0034659
    [72] Ogawa D, Bobeszko T, Ainsworth T, et al. (2013) The combined effects of temperature and CO2 lead to altered gene expression in Acropora aspera. Coral Reefs 32: 895-907. doi: 10.1007/s00338-013-1046-9
    [73] Brown JH, Gillooly JF, Allen AP, et al. (2004) Toward a metabolic theory of ecology. Ecology 85: 1771-1789. doi: 10.1890/03-9000
    [74] Helmuth B (2009) From cells to coastlines: how can we use physiology to forecast the impacts of climate change? J Exp Biol 212: 753-760. doi: 10.1242/jeb.023861
    [75] Kooijman SALM (2010) Dynamic energy budget theory for metabolic organisation: Cambridge university press.
    [76] Rosa R, Trübenbach K, Pimentel MS, et al. (2014) Differential impacts of ocean acidification and warming on winter and summer progeny of a coastal squid (Loligo vulgaris). J exp biol 217: 518-525.
    [77] Fernández-Reiriz MJ, Range P, Álvarez-Salgado XA, et al. (2011) Physiological energetics of juvenile clams Ruditapes decussatus in a high CO2 coastal ocean. Mar Ecol Prog Ser 433: 97-105. doi: 10.3354/meps09062
    [78] Ferrari MC, Dixson DL, Munday PL, et al. (2011) Intrageneric variation in antipredator responses of coral reef fishes affected by ocean acidification: implications for climate change projections on marine communities. Global Change Biol 17: 2980-2986. doi: 10.1111/j.1365-2486.2011.02439.x
    [79] Manríquez PH, Jara ME, Mardones ML, et al. (2013) Ocean acidification disrupts prey responses to predator cues but not net prey shell growth in Concholepas concholepas (loco). PLoS One 8: e68643. doi: 10.1371/journal.pone.0068643
    [80] Li W, Gao K (2012) A marine secondary producer respires and feeds more in a high CO2 ocean. Mar pollut bull 64: 699-703. doi: 10.1016/j.marpolbul.2012.01.033
    [81] Pörtner HO, Farrell AP (2008) Physiology and climate change. Science 322: 690-692. doi: 10.1126/science.1163156
    [82] Russell BD, Harley CD, Wernberg T, et al. (2011) Predicting ecosystem shifts requires new approaches that integrate the effects of climate change across entire systems. Biol Lett: rsbl20110779.
    [83] Mumby PJ, Steneck RS (2008) Coral reef management and conservation in light of rapidly evolving ecological paradigms. Trends ecol evol 23: 555-563. doi: 10.1016/j.tree.2008.06.011
    [84] McElhany P, Shallin Busch D Appropriate pCO2 treatments in ocean acidification experiments. Mar Biol 160: 1807-1812.
    [85] Folt C, Chen C, Moore M, et al. (1999) Synergism and antagonism among multiple stressors. Limnol Oceanogr 44: 864-877. doi: 10.4319/lo.1999.44.3_part_2.0864
  • This article has been cited by:

    1. Eenezer Bonyah, Fahad Al Basir, Santanu Ray, 2020, Chapter 14, 978-981-15-0927-8, 301, 10.1007/978-981-15-0928-5_14
    2. J. Frédéric Bonnans, Justina Gianatti, E. Augeraud, M. Banerjee, J.-S. Dhersin, A. d'Onofrio, T. Lipniacki, S. Petrovskii, Chi Tran, A. Veber-Delattre, E. Vergu, V. Volpert, Optimal control techniques based on infection age for the study of the COVID-19 epidemic, 2020, 15, 0973-5348, 48, 10.1051/mmnp/2020035
    3. Abdelfatah Kouidere, Driss Kada, Omar Balatif, Mostafa Rachik, Mouhcine Naim, Optimal control approach of a mathematical modeling with multiple delays of the negative impact of delays in applying preventive precautions against the spread of the COVID-19 pandemic with a case study of Brazil and cost-effectiveness, 2021, 142, 09600779, 110438, 10.1016/j.chaos.2020.110438
    4. C. J. Silva, D. F. M. Torres, E. Venturino, M. Ruzhansky, P. Agarwal, I. Area, E. T. Karimov, Optimal Spraying in Biological Control of Pests, 2017, 12, 1760-6101, 51, 10.1051/mmnp/201712305
    5. U. Roman, Z. Gul, I. Saeed, U. Hakeem, S. Shafie, Control strategies of avian influenza pandemic model with time delay, 2017, 10, 1793-5245, 1750104, 10.1142/S1793524517501042
    6. Diana Rocha, Cristiana J. Silva, Delfim F. M. Torres, Stability and optimal control of a delayed HIV model, 2018, 41, 01704214, 2251, 10.1002/mma.4207
    7. S. Rosa, P. Rebelo, C.M. Silva, H. Alves, P.G. Carvalho, Optimal control of the customer dynamics based on marketing policy, 2018, 330, 00963003, 42, 10.1016/j.amc.2018.02.027
    8. Rossella Della Marca, Alberto d’Onofrio, Volatile opinions and optimal control of vaccine awareness campaigns: chaotic behaviour of the forward-backward sweep algorithm vs. heuristic direct optimization, 2021, 98, 10075704, 105768, 10.1016/j.cnsns.2021.105768
    9. Ana P. Lemos-Paião, Cristiana J. Silva, Delfim F. M. Torres, 2020, Chapter 12, 978-3-030-49895-5, 323, 10.1007/978-3-030-49896-2_12
    10. Adnane Boukhouima, Khalid Hattaf, El Mehdi Lotfi, Marouane Mahrouf, Delfim F.M. Torres, Noura Yousfi, Lyapunov functions for fractional-order systems in biology: Methods and applications, 2020, 140, 09600779, 110224, 10.1016/j.chaos.2020.110224
    11. Filipe Rodrigues, Cristiana J. Silva, Delfim F. M. Torres, Helmut Maurer, Optimal control of a delayed HIV model, 2018, 23, 1553-524X, 443, 10.3934/dcdsb.2018030
    12. Weronika Wojtak, Cristiana J. Silva, Delfim F.M. Torres, Abdon Atangana, Gisèle Mophou, Jordan Hristov, Zakia Hammouch, Uniform asymptotic stability of a fractional tuberculosis model, 2018, 13, 0973-5348, 9, 10.1051/mmnp/2018015
    13. Duo-Qing Sun, Xiao-Ying Ma, Characteristic Model-Based Adaptive Control for a Class of MIMO Uncertain Nonaffine Nonlinear Systems Governed by Differential Equations, 2020, 2020, 1076-2787, 1, 10.1155/2020/3649472
    14. Legesse Lemecha Obsu, Shiferaw Feyissa Balcha, Optimal control strategies for the transmission risk of COVID-19, 2020, 14, 1751-3758, 590, 10.1080/17513758.2020.1788182
    15. Cristiana J. Silva, Delfim F. M. Torres, Modeling and optimal control of HIV/AIDS prevention through PrEP, 2018, 11, 1937-1179, 119, 10.3934/dcdss.2018008
    16. Riccardo Bonalli, Bruno Hérissé, Emmanuel Trélat, Continuity of Pontryagin Extremals with Respect to Delays in Nonlinear Optimal Control, 2019, 57, 0363-0129, 1440, 10.1137/18M119121X
    17. Anthony A. E. Losio, Steady Mushayabasa, Modeling the Effects of Spatial Heterogeneity and Seasonality on Guinea Worm Disease Transmission, 2018, 2018, 1110-757X, 1, 10.1155/2018/5084687
    18. Ana R.M. Carvalho, Carla M.A. Pinto, Non-integer order analysis of the impact of diabetes and resistant strains in a model for TB infection, 2018, 61, 10075704, 104, 10.1016/j.cnsns.2018.01.012
    19. Arianna Cecchetti, Pietro Ferragamo, Alice Zampini, Ezio Venturino, Optimal Control of an Ecogenetic Model, 2017, 3, 2349-5103, 473, 10.1007/s40819-017-0365-x
    20. Ismail Gad Ameen, Hegagi Mohamed Ali, 2020, chapter 3, 9781799831228, 41, 10.4018/978-1-7998-3122-8.ch003
    21. Nasser Sweilam, Seham AL-Mekhlafi, Salma Shatta, Dumitru Baleanu, Numerical study for a novel variable-order multiple time delay awareness programs mathematical model, 2020, 158, 01689274, 212, 10.1016/j.apnum.2020.07.016
    22. Malik Zaka Ullah, Abdullah K. Alzahrani, Dumitru Baleanu, An efficient numerical technique for a new fractional tuberculosis model with nonsingular derivative operator, 2019, 13, 1658-3655, 1147, 10.1080/16583655.2019.1688543
    23. Cristiana J. Silva, Carla Cruz, Delfim F. M. Torres, Alberto P. Muñuzuri, Alejandro Carballosa, Iván Area, Juan J. Nieto, Rui Fonseca-Pinto, Rui Passadouro, Estevão Soares dos Santos, Wilson Abreu, Jorge Mira, Optimal control of the COVID-19 pandemic: controlled sanitary deconfinement in Portugal, 2021, 11, 2045-2322, 10.1038/s41598-021-83075-6
    24. N H Sweilam, S M AL–Mekhlafi, Optimal control for a time delay multi-strain tuberculosis fractional model: a numerical approach, 2019, 36, 0265-0754, 317, 10.1093/imamci/dnx046
    25. Mlyashimbi Helikumi, Moatlhodi Kgosimore, Dmitry Kuznetsov, Steady Mushayabasa, Backward Bifurcation and Optimal Control Analysis of a Trypanosoma brucei rhodesiense Model, 2019, 7, 2227-7390, 971, 10.3390/math7100971
    26. Nasser Hassan Sweilam, Seham Mahyoub Al-Mekhlafi, Taghreed Abdul Rahman Assiri, Numerical Study for Time Delay Multistrain Tuberculosis Model of Fractional Order, 2017, 2017, 1076-2787, 1, 10.1155/2017/1047384
    27. Simão P. S. Santos, Natália Martins, Delfim F. M. Torres, Noether currents for higher-order variational problems of Herglotz type with time delay, 2018, 11, 1937-1179, 91, 10.3934/dcdss.2018006
    28. Teklebirhan Abraha, Fahad Al Basir, Legesse Lemecha Obsu, Delfim F.M. Torres, Pest control using farming awareness: Impact of time delays and optimal use of biopesticides, 2021, 146, 09600779, 110869, 10.1016/j.chaos.2021.110869
    29. Cristiana J. Silva, Stability and optimal control of a delayed HIV/AIDS-PrEP model, 2022, 15, 1937-1632, 639, 10.3934/dcdss.2021156
    30. Zain Ul Abadin Zafar, Samina Younas, Sumera Zaib, Cemil Tunç, An efficient numerical simulation and mathematical modeling for the prevention of tuberculosis, 2022, 15, 1793-5245, 10.1142/S1793524522500152
    31. Zain Ul Abadin Zafar, Sumera Zaib, Muhammad Tanveer Hussain, Cemil Tunç, Shumaila Javeed, Analysis and numerical simulation of tuberculosis model using different fractional derivatives, 2022, 160, 09600779, 112202, 10.1016/j.chaos.2022.112202
    32. LEGESSE LEMECHA OBSU, OPTIMAL CONTROL ANALYSIS OF A TUBERCULOSIS MODEL, 2022, 30, 0218-3390, 837, 10.1142/S0218339022500309
    33. Danylo Malyuta, Taylor P. Reynolds, Michael Szmuk, Thomas Lew, Riccardo Bonalli, Marco Pavone, Behcet Acikmese, Convex Optimization for Trajectory Generation: A Tutorial on Generating Dynamically Feasible Trajectories Reliably and Efficiently, 2022, 42, 1066-033X, 40, 10.1109/MCS.2022.3187542
    34. S. L. Chasanah, Y. T. Utami, D. E. Nurvazly, B. F. H. E. Syachrannie, 2022, 2563, 0094-243X, 050004, 10.1063/5.0103442
    35. Willi Semmler, Jérôme Henry, Helmut Maurer, Pandemic meltdown and economic recovery – A multi-phase dynamic model, empirics, and policy, 2023, 6, 2590051X, 100106, 10.1016/j.resglo.2022.100106
    36. Hanis Nasir, A time-delay model of diabetic population: Dynamics analysis, sensitivity, and optimal control, 2021, 96, 0031-8949, 115002, 10.1088/1402-4896/ac1473
    37. Yujiang Liu, Shujing Gao, Zhenzhen Liao, Di Chen, Dynamical behavior of a stage-structured Huanglongbing model with time delays and optimal control, 2022, 156, 09600779, 111830, 10.1016/j.chaos.2022.111830
    38. Mohamed A. Zaitri, Cristiana J. Silva, Delfim F. M. Torres, Stability Analysis of Delayed COVID-19 Models, 2022, 11, 2075-1680, 400, 10.3390/axioms11080400
    39. Lela Alkhazishvili, Medea Iordanishvili, The local formula of representation of a solution for a functional differential equation with the mixed initial condition considering perturbations of delays containing in the phase coordinates and in controls, 2022, 29, 1072-947X, 1, 10.1515/gmj-2021-2123
    40. Juliana Couras, Iván Area, Juan J. Nieto, Cristiana J. Silva, Delfim F. M. Torres, 2021, Chapter 23, 978-981-16-2449-0, 509, 10.1007/978-981-16-2450-6_23
    41. Abou Bakari Diabaté, Boureima Sangaré, Ousmane Koutou, Optimal control analysis of a COVID-19 and Tuberculosis (TB) co-infection model with an imperfect vaccine for COVID-19, 2023, 2254-3902, 10.1007/s40324-023-00330-8
    42. A. Sakkoum, M. Lhous, M. Rachik, A. Tridane, Discrete mathematical modeling and optimal control of the marital status: Islamic polygamous marriage model case, 2023, 10, 23129794, 748, 10.23939/mmc2023.03.748
    43. Oluwatosin Babasola, Evans Otieno Omondi, Kayode Oshinubi, Nancy Matendechere Imbusi, Stochastic Delay Differential Equations: A Comprehensive Approach for Understanding Biosystems with Application to Disease Modelling, 2023, 3, 2673-9909, 702, 10.3390/appliedmath3040037
    44. C. Yalçın Kaya, Helmut Maurer, Optimization over the Pareto front of nonconvex multi-objective optimal control problems, 2023, 0926-6003, 10.1007/s10589-023-00535-7
    45. Sibaliwe Maku Vyambwera, Peter Witbooi, Victor Kovtunenko, Optimal Intervention Strategies on TB Epidemiology in a Prison Population, 2024, 2024, 1085-3375, 10.1155/aaa/6482099
    46. Krzysztof Michalak, Knowledge-based optimization in epidemics prevention, 2025, 1432-7643, 10.1007/s00500-025-10494-2
    47. Sayooj Aby Jose, Raja Ramachandran, Jirawattanapanit Anuwat, Jinde Cao, Ravi P. Agarwal, A fractional derivative approach to infectious disease dynamics: modeling and optimal control strategies, 2025, 11, 2363-6203, 10.1007/s40808-025-02394-z
  • Reader Comments
  • © 2017 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(5640) PDF downloads(1207) Cited by(15)

Article outline

Figures and Tables

Figures(5)  /  Tables(3)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog