Skip to main content

Mathematical modeling of forest ecosystems by a reaction–diffusion–advection system: impacts of climate change and deforestation


We present an innovative mathematical model for studying the dynamics of forest ecosystems. Our model is determined by an age-structured reaction–diffusion–advection system in which the roles of the water resource and of the atmospheric activity are considered. The model is abstract but constructed in such a manner that it can be applied to real-world forest areas; thus it allows to establish an infinite number of scenarios for testing the robustness and resilience of forest ecosystems to anthropic actions or to climate change. We establish the well-posedness of the reaction–diffusion–advection model by using the method of characteristics and by reducing the initial system to a reaction–diffusion problem. The existence and stability of stationary homogeneous and stationary heterogeneous solutions are investigated, so as to prove that the model is able to reproduce relevant equilibrium states of the forest ecosystem. We show that the model fits with the principle of almost uniform precipitation over forested areas and of exponential decrease of precipitation over deforested areas. Furthermore, we present a selection of numerical simulations for an abstract forest ecosystem, in order to analyze the stability of the steady states, to investigate the impact of anthropic perturbations such as deforestation and to explore the effects of climate change on the dynamics of the forest ecosystem.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9
Fig. 10
Fig. 11


  • Anderegg WR, Trugman AT, Badgley G, Konings AG, Shaw J (2020) Divergent forest sensitivity to repeated extreme droughts. Nat Clim Change 10:1–5

    Article  Google Scholar 

  • Antonovsky MY, Fleming R, Kuznetsov YA, Clark W (1990) Forest-pest interaction dynamics: the simplest mathematical models. Theor Popul Biol 37(2):343–367

    MathSciNet  Article  Google Scholar 

  • Botkin DB et al (1993) Forest dynamics: an ecological model. Oxford University Press on Demand

  • Cantin G, Verdière N (2020) Networks of forest ecosystems: mathematical modeling of their biotic pump mechanism and resilience to certain patch deforestation. Ecol Complex 43:100850

    Article  Google Scholar 

  • Cantrell RS, Cosner C (2004) Spatial ecology via reaction–diffusion equations. Wiley, Hoboken

    Book  Google Scholar 

  • Cook-Patton SC, Leavitt SM, Gibbs D, Harris NL, Lister K, Anderson-Teixeira KJ, Briggs RD, Chazdon RL, Crowther TW, Ellis PW et al (2020) Mapping carbon accumulation potential from global natural forest regrowth. Nature 585(7826):545–550

    Article  Google Scholar 

  • Cosandey C (1992) Influence de la forêt sur le cycle de l’eau: conséquences d’une coupe forestière sur le bilan d’écoulement annuel. Hydrologie continentale 7(1):13–22

  • Creed IF, Weber M, Accatino F, Kreutzweiser DP (2016) Managing forests for water in the anthropocene-the best kept secret services of forest ecosystems. Forests 7(3):60

    Article  Google Scholar 

  • Dubreuil V, Arvor D, Funatsu BM, Nédélec V, Mello-Théry NAd (2020) Les changements climatiques en Amazonie, une approche multiscalaire. In: Mercier D (ed) Les impacts spatiaux du changement climatique. ISTE, pp 247–270

  • Fernandes K, Giannini A, Verchot L, Baethgen W, Pinedo-Vasquez M (2015) Decadal covariability of Atlantic SSTs and western Amazon dry-season hydroclimate in observations and CMIP5 simulations. Geophys Res Lett 42(16):6793–6801

    Article  Google Scholar 

  • Filippov AF (2013) Differential equations with discontinuous righthand sides: control systems, vol 18. Springer, Berlin

    Google Scholar 

  • Flannigan MD, Stocks BJ, Wotton BM (2000) Climate change and forest fires. Sci Total Environ 262(3):221–229

    Article  Google Scholar 

  • Gillett N, Weaver A, Zwiers F, Flannigan M (2004) Detecting the effect of climate change on Canadian forest fires. Geophys Res Lett 31(18):L18211

    Article  Google Scholar 

  • Hansen AJ, Burns P, Ervin J, Goetz SJ, Hansen M, Venter O, Watson JE, Jantz PA, Virnig AL, Barnett K et al (2020) A policy-driven framework for conserving the best of earth’s remaining moist tropical forests. Nat Ecol Evol 41:1–8

  • Hecht F, Pironneau O, Le Hyaric A, Ohtsuka K (2005) Freefem++ manual

  • Henry D (2006) Geometric theory of semilinear parabolic equations, vol 840. Springer, Berlin

    Google Scholar 

  • Keenan RJ (2015) Climate change impacts and adaptation in forest management: a review. Ann For Sci 72(2):145–167

    MathSciNet  Article  Google Scholar 

  • Kohyama T (1992) Size-structured multi-species model of rain forest trees. Funct Ecol 6:206–212

    Article  Google Scholar 

  • Kolobov A, Frisman EY (2016) Individual-based model of spatio-temporal dynamics of mixed forest stands. Ecol Complex 27:29–39

    Article  Google Scholar 

  • Kuznetsov YA, Antonovsky MY, Biktashev V, Aponina E (1994) A cross-diffusion model of forest boundary dynamics. J Math Biol 32(3):219–232

    MathSciNet  Article  Google Scholar 

  • Le Huy C, Tsujikawa T, Yagi A (2006) Asymptotic behavior of solutions for forest kinematic model. Funkcialaj Ekvacioj 49(3):427–449

    MathSciNet  Article  Google Scholar 

  • Le Huy C, Tsujikawa T, Yagi A (2009) Stationary solutions to forest kinematic model. Glasgow Math J 51(1):1–17

    MathSciNet  Article  Google Scholar 

  • Magal P, Zhang Z (2017) Competition for light in forest population dynamics: from computer simulator to mathematical model. J Theor Biol 419:290–304

    MathSciNet  Article  Google Scholar 

  • Nobre CA, Obregón GO, Marengo JA, Fu R, Poveda G (2009) Characteristics of Amazonian climate: main features. In: Amazonia and global change. American Geophysical Union, pp 149–162 (AGU)

  • North M, Stephens S, Collins B, Agee J, Aplet G, Franklin J, Fule PZ (2015) Reform forest fire management. Science 349(6254):1280–1281

    Article  Google Scholar 

  • Noss RF (2001) Beyond Kyoto: forest management in a time of rapid climate change. Conserv Biol 15(3):578–590

    Article  Google Scholar 

  • Oyama MD, Nobre CA (2003) A new climate-vegetation equilibrium state for Tropical South America. Geophys Res Lett 30(23):2199

    Article  Google Scholar 

  • Running SW, Nemani RR, Peterson DL, Band LE, Potts DF, Pierce LL, Spanner MA (1989) Mapping regional forest evapotranspiration and photosynthesis by coupling satellite data with ecosystem simulation. Ecology 70(4):1090–1101

    Article  Google Scholar 

  • Schertzer E, Staver A, Levin SA (2015) Implications of the spatial dynamics of fire spread for the bistability of savanna and forest. J Math Biol 70(1–2):329–341

    MathSciNet  Article  Google Scholar 

  • Schwetlick HR (2002) Limit sets for multidimensional nonlinear transport equations. J Differ Equ 179(1):356–368

    MathSciNet  Article  Google Scholar 

  • Shirai T, Chuan L, Yagi A (2007) Dynamical system for forest kinematic model under Dirichlet conditions. Scientiae Mathematicae japonicae 66(2):275–288

    MathSciNet  MATH  Google Scholar 

  • Shukla J, Dubey B (1997) Modelling the depletion and conservation of forestry resources: effects of population and pollution. J Math Biol 36(1):71–94

    MathSciNet  Article  Google Scholar 

  • Smith HL, Thieme HR (2011) Dynamical systems and population persistence, vol 118. American Mathematical Society, Providence

    MATH  Google Scholar 

  • Spittlehouse DL, Stewart RB (2004) Adaptation to climate change in forest management. J Ecosyst Manag 4(1)

  • Strang G (1968) On the construction and comparison of difference schemes. SIAM J Numer Anal 5(3):506–517

    MathSciNet  Article  Google Scholar 

  • Veldkamp E, Schmidt M, Powers JS, Corre MD (2020) Deforestation and reforestation impacts on soils in the tropics. Nat Rev Earth Environ 1:1–16

    Article  Google Scholar 

  • Verzelen N, Picard N, Gourlet-Fleury S (2006) Approximating spatial interactions in a model of forest dynamics as a means of understanding spatial patterns. Ecol Complex 3(3):209–218

    Article  Google Scholar 

  • Webb T (1986) Is vegetation in equilibrium with climate? How to interpret late-Quaternary pollen data. Vegetatio 67(2):75–91

    Article  Google Scholar 

  • Wilson KB, Hanson PJ, Mulholland PJ, Baldocchi DD, Wullschleger SD (2001) A comparison of methods for determining forest evapotranspiration and its components: sap-flow, soil water budget, eddy covariance and catchment water balance. Agric For Meteorol 106(2):153–168

    Article  Google Scholar 

  • Yagi A (2009) Abstract parabolic evolution equations and their applications. Springer, Berlin

    MATH  Google Scholar 

  • Yatat V, Couteron P, Tewa JJ, Bowong S, Dumont Y (2017) An impulsive modelling framework of fire occurrence in a size-structured model of tree–grass interactions for savanna ecosystems. J Math Biol 74(6):1425–1482

    MathSciNet  Article  Google Scholar 

  • Yousefpour R, You B, Hanewinkel M (2019) Simulation of extreme storm effects on regional forest soil carbon stock. Ecol Model 399:39–53

    Article  Google Scholar 

  • Zemp DC, Schleussner C-F, Barbosa HM, Hirota M, Montade V, Sampaio G, Staal A, Wang-Erlandsson L, Rammig A (2017) Self-amplified Amazon forest loss due to vegetation-atmosphere feedbacks. Nat Commun 8(1):1–10

    Article  Google Scholar 

  • Zheng Z, Huang W, Li S, Zeng Y (2017) Forest fire spread simulating model using cellular automaton with extreme learning machine. Ecol Model 348:33–43

    Article  Google Scholar 

  • Zhou G, Meng C, Jiang P, Xu Q (2011) Review of carbon fixation in bamboo forests in China. Bot Rev 77(3):262

    Article  Google Scholar 

Download references

Author information

Authors and Affiliations


Corresponding author

Correspondence to Guillaume Cantin.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.



In this appendix, we present the proofs of Theorems 12, 3, and we briefly show numerical results which underpin the approximation of the non-stationary advection equation (1) by the stationary advection equation (2).

Proof of Theorem 1

The proof of Theorem 1 is divided into three steps. First, we use the method of characteristics to solve the stationary equation (19), so as to prove that the operator \(\psi \) is well defined. Since \(v \in L^\infty (\varOmega )\) may be discontinuous, we use the theory of ordinary differential equations of Carathéodory. In the second step, we prove that the operator \(\psi \) is continuous, using an integral Gronwall lemma. Finally, in the third step, we prove that the operator \(\psi \) is differentiable. We remark that its derivative is uniquely determined as the solution of an ODE.

First step. Let \(\big \lbrace \xi (x_0,\,s)\big \rbrace _{0 \le s \le S(x_0)}\) denote one characteristic line of the advection field \(\varvec{a}\), starting at \(x_0 \in \varGamma \) and ending at \(\xi \big (x_0,\,S(x_0)\big ) \in \partial \varOmega {\setminus } \varGamma \). We introduce the function \({\tilde{\rho }}\) defined along the latter characteristic line by \({\tilde{\rho }}(s) = \rho \circ \xi (x_0,\,s)\), for \(0 \le s \le S(x_0)\), with the lightened notation (18). Then it is seen that \({\tilde{\rho }}\) satisfies the following ODE:

$$\begin{aligned} {\left\{ \begin{array}{ll} \dfrac{d {\tilde{\rho }}}{ds} + \sigma {\tilde{\rho }} = \varphi ({\tilde{\rho }}){\tilde{v}}, \quad 0 < s \le S(x_0), \\ {\tilde{\rho }}(0) = m(x_0). \end{array}\right. } \end{aligned}$$

Now we introduce the function \(\theta \) defined on \([0,\,S(x_0)]\) by \(\theta (s) = {\tilde{\rho }}(s)e^{\sigma s}\). We easily prove that \(\theta \) satisfies the following ODE:

$$\begin{aligned} {\left\{ \begin{array}{ll} \dfrac{d \theta }{ds} = g(s,\,\theta ), \quad 0 < s \le S(x_0), \\ \theta (0) = m(x_0), \end{array}\right. } \end{aligned}$$

where g is given by

$$\begin{aligned} g(s,\,\theta ) = \varphi \big (\theta e^{-\sigma s} \big ) e^{\sigma s} {\tilde{v}}(s). \end{aligned}$$

Since \(v \in L^\infty (\varOmega )\), the function g is continuous in its second variable \(\theta \) but discontinuous in its first variable s. For that reason, we solve Eq. (44) as an ODE of Carathéodory type (see e.g. Filippov 2013, Chapter 1). We directly verify the following properties:

  • the function \(g(s,\,\theta )\) is defined in \([0,\,S(x_0)] \times {\mathbb {R}}\), continuous in \(\theta \) for almost every s;

  • for all \(\theta \), the function \(g(\cdot ,\,\theta )\) is measurable;

  • \(\left| g(s,\,\theta )\right| \le M(s)\) with \(M(s) = \varphi _0 e^{\sigma s} \left| {\tilde{v}}(s)\right| \in L^\infty \big ([0,\,S(x_0)]\big )\);

  • \(\left| g(s,\,\theta _1) - g(s,\,\theta _2)\right| \le L(s) \left| \theta _1 - \theta _2\right| \) with \(L(s) = \varphi _0 \left| {\tilde{v}}(s)\right| \in L^\infty \big ([0,\,S(x_0)]\big )\).

By virtue of Theorems 1 and 2 in Filippov (2013), Chapter 1, Eq. (44) admits a unique local solution, defined on \([0,\,{\bar{s}}]\) with \({\bar{s}} > 0\). That solution is given by

$$\begin{aligned} \theta (s) = m(x_0) + \int _0^s g\big (\tau ,\,\theta (\tau )\big )d\tau , \quad 0 \le s \le {\bar{s}}, \end{aligned}$$

and it is absolutely continuous on each compact interval included in \([0,\,{\bar{s}}]\). Next, the non-negativity of \(\theta (s)\) follows from Proposition A.17 in Smith and Thieme (2011) (note that the continuity is not necessary in the proof of that proposition). We continue by proving that \(\theta \) is global in \([0,\,S(x_0)]\). This simply follows from the differential inequality

$$\begin{aligned} \dfrac{d\theta }{ds} \le \varphi _0 \left\| v\right\| _\infty \theta , \quad s > 0, \end{aligned}$$

which leads to

$$\begin{aligned} \theta (s) \le m(x_0)e^{\varphi _0 \left\| v\right\| _\infty S(x_0)}, \quad \forall s \in [0,\,{\bar{s}}]. \end{aligned}$$

Thus we have \({\bar{s}} = S(x_0)\).

Now we are brought back to equation (43), which is solved by setting

$$\begin{aligned} {\tilde{\rho }}(s) = \theta (s) e^{-\sigma s}, \quad s \in [0,\,S(x_0)]. \end{aligned}$$

As constructed, \({\tilde{\rho }}\) is absolutely continuous on each compact interval included in \([0,\,S(x_0)]\), and uniquely determined by the representation formula

$$\begin{aligned} {\tilde{\rho }}(s) = m(x_0)e^{-\sigma s} + \int _0^s \varphi \big ({\tilde{\rho }}(\tau )\big ){\tilde{v}}(\tau )e^{-\sigma (s-\tau )}d\tau , \quad s \in [0,\,S(x_0)], \end{aligned}$$

which is the lightened expression for Eq. (21). Since Eq. (43) has been solved along each characteristic line of the advection field \(\varvec{a}\), we finally solve Eq. (19) in \(\varOmega \) by setting

$$\begin{aligned} \rho (x) = {\tilde{\rho }}\big (\zeta _1(x),\,\zeta _2(x)\big ), \quad x \in \varOmega , \end{aligned}$$

with the notation (18), where \(\zeta _1\) and \(\zeta _2\) are defined by (17). As constructed, \(\rho \) is continuous and non-negative along the characteristic lines of the advection field \(\varvec{a}\). Furthermore, the principle of continuity of the solution of Eq. (43) with respect to a variation of the initial condition (see Theorem 6 in Filippov 2013, Chapter 1) and the continuity of the advection field \(\varvec{a}\) [see Eq. (15)] guaranty that \(\rho \) is continuous in \({\bar{\varOmega }}\), thus belongs to \(L_+^\infty (\varOmega )\). We have proved that the operator \(\psi \) is well defined.

Second step. Let \(v, h \in L_+^\infty (\varOmega )\). We set \(\rho = \psi (v)\) and \(\rho ^h = \psi (v+h)\). By virtue of the representation formula (21), we have:

$$\begin{aligned} \begin{aligned}&\left| \psi (v+h)(x) - \psi (v)(x)\right| \\&\quad =\left| \int _0^s \varphi \big ({\tilde{\rho }}^h(\tau )\big )({\tilde{v}}+{\tilde{h}})(\tau )e^{-\sigma (s-\tau )}d\tau - \int _0^s \varphi \big ({\tilde{\rho }}(\tau )\big ){\tilde{v}}(\tau )e^{-\sigma (s-\tau )}d\tau \right| \\&\quad =\left| \int _0^s \Big [\varphi \big ({\tilde{\rho }}^h(\tau )\big ) - \varphi \big ({\tilde{\rho }}(\tau )\big ) \Big ]{\tilde{v}}(\tau )e^{-\sigma (s-\tau )}d\tau + \int _0^s \varphi \big ({\tilde{\rho }}^h(\tau )\big ){\tilde{h}}(\tau )e^{-\sigma (s-\tau )}d\tau \right| \\&\quad \le \left\| v\right\| _\infty \int _0^s \left| \varphi \big ({\tilde{\rho }}^h(\tau )\big ) - \varphi \big ({\tilde{\rho }}(\tau )\big )\right| d\tau + \left\| h\right\| _\infty \varphi _0 \int _0^s e^{-\sigma (s-\tau )}d\tau \\&\quad \le \left\| v\right\| _\infty \varphi _0\int _0^s \left| {\tilde{\rho }}^h(\tau ) - {\tilde{\rho }}(\tau )\right| d\tau + \left\| h\right\| _\infty \dfrac{\varphi _0}{\sigma }, \end{aligned} \end{aligned}$$

where we have used properties of the function \(\varphi \) given in (13). Now we introduce the function p defined by

$$\begin{aligned} p(s) = \left| {\tilde{\rho }}^h(s) - {\tilde{\rho }}(s)\right| , \quad s \ge 0. \end{aligned}$$

We have proved above that

$$\begin{aligned} p(s) \le \left\| v\right\| _\infty \varphi _0\int _0^s p(\tau )d\tau + \left\| h\right\| _\infty \dfrac{\varphi _0}{\sigma }. \end{aligned}$$

By virtue on the integral Gronwall lemma, we obtain

$$\begin{aligned} p(s) \le \left\| h\right\| _\infty \dfrac{\varphi _0}{\sigma }e^{\left\| v\right\| _\infty \varphi _0 s} \le \left\| h\right\| _\infty \dfrac{\varphi _0}{\sigma }e^{\left\| v\right\| _\infty \varphi _0 {\bar{S}}}, \end{aligned}$$

where we have used the uniform bound (16). We then obtain

$$\begin{aligned} \left\| \psi (v+h)-\psi (v)\right\| _\infty \le \left\| h\right\| _\infty \times \frac{\varphi _0}{\sigma }e^{\varphi _0\left\| v\right\| _\infty {\bar{S}}}, \end{aligned}$$

which proves the continuity of the operator \(\psi \).

Third step. Finally, let us introduce the operator \(L_v\) uniquely defined for all \(v \in L_+^\infty (\varOmega )\) by the integral formula

$$\begin{aligned} L_v\,{\tilde{h}}(s) = \int _0^s \varphi \big ({\tilde{\rho }}(\tau )\big ){\tilde{h}}(\tau )e^{-\sigma (s-\tau )}d\tau + \int _0^s\varphi '\big ({\tilde{\rho }}(\tau )\big ){\tilde{v}}(\tau )L_v\,{\tilde{h}}(\tau ) e^{\sigma (s-\tau )} d\tau , \end{aligned}$$

for all \(h \in L_+^\infty (\varOmega )\), where \(\rho = \psi (v)\). As constructed, the operator \(L_v\) is obviously linear. Now, let \(v, h \in L_+^\infty (\varOmega )\). We set again \(\rho = \psi (v)\) and \(\rho ^h = \psi (v+h)\), and we introduce the function p defined by

$$\begin{aligned} p(s) = \left| {\tilde{\rho }}^h(s) - {\tilde{\rho }}(s) - L_v\,h(s)\right| , \quad s \ge 0. \end{aligned}$$

We use the continuity and differentiability of the function \(\varphi \) in order to write:

$$\begin{aligned} \begin{aligned}&\varphi ({\tilde{\rho }}^h) = \varphi ({\tilde{\rho }}) + o(1) ~\text {as}~\left\| h\right\| _\infty \rightarrow 0,\\&\varphi ({\tilde{\rho }}^h) = \varphi ({\tilde{\rho }}) + \varphi '({\tilde{\rho }}) ({\tilde{\rho }}^h - {\tilde{\rho }}) + o\big (\left\| h\right\| _\infty \big ) ~\text {as}~\left\| h\right\| _\infty \rightarrow 0,\\ \end{aligned} \end{aligned}$$

which leads to, after basic computations:

$$\begin{aligned} p(s) \le \left\| v\right\| _\infty \int _0^s p(\tau )d\tau + o\big (\left\| h\right\| _\infty \big ) \times \dfrac{\left\| v\right\| _\infty +1}{\sigma }, \quad s \ge 0. \end{aligned}$$

Using the integral Gronwall lemma again and Eq. (16) leads to

$$\begin{aligned} \left| {\tilde{\rho }}^h(s) - {\tilde{\rho }}(s) - L_v\,h(s)\right| \le o\big (\left\| h\right\| _\infty \big ) \times \dfrac{\left\| v\right\| _\infty +1}{\sigma } e^{\left\| v\right\| _\infty {\bar{S}}}. \end{aligned}$$

Furthermore, we easily observe that

$$\begin{aligned} \left| L_v\,{\tilde{h}}(s)\right| \le \left\| v\right\| _\infty \int _0^s \left| L_v\,{\tilde{h}}(\tau )\right| + \dfrac{\varphi _0 \left\| h\right\| _\infty }{\sigma }. \end{aligned}$$

Using a third time the integral Gronwall lemma leads to

$$\begin{aligned} \left| L_v\,{\tilde{h}}(s)\right| \le \dfrac{\varphi _0 \left\| h\right\| _\infty }{\sigma }e^{\left\| v\right\| _\infty {\bar{S}}}. \end{aligned}$$

In this way, we have proved that \(\psi \) is differentiable in \(L^\infty (\varOmega )\). Its derivative is given by

$$\begin{aligned} D\psi (v)\,h = L_v\,h, \end{aligned}$$

for all \(v, h \in L_+^\infty (\varOmega )\) and satisfies (24).

Proof of Theorem 2

By virtue of Theorem 4.4 in Yagi (2009), it suffices to prove an estimation of the type

$$\begin{aligned} \begin{aligned}&\left\| F(U) - F({\tilde{U}})\right\| _X \\&\quad \le k\big (\left\| U\right\| _X + \left\| {\tilde{U}}\right\| _X\big ) \Big [\left\| A^\eta (U-{\tilde{U}})\right\| _X + (\left\| U\right\| _X+\left\| {\tilde{U}}\right\| _X)\left\| U-{\tilde{U}}\right\| _X \Big ], \end{aligned} \end{aligned}$$

for each U, \({\tilde{U}}\) in \({\mathcal {D}}(A^\eta )\), where k is an increasing function defined in \({\mathbb {R}}^+\) with values in \({\mathbb {R}}^+\). Thus we consider \(U = (u,\,v,\,w)^T\), \({\tilde{U}} = ({\tilde{u}},\,{\tilde{v}},\,{\tilde{w}})^T\) in \({\mathcal {D}}(A^\eta )\). We begin by estimating the norm of the difference \(\left\| F_1(U)-F_1({\tilde{U}})\right\| _\infty \). To that aim, we write:

$$\begin{aligned} \begin{aligned} \left\| F_1(U)-F_1({\tilde{U}})\right\| _\infty&\le \beta \delta \left\| w-{\tilde{w}}\right\| _\infty \\&\quad + \left\| \gamma _0(v)u - \gamma _0({\tilde{v}}){\tilde{u}}\right\| _\infty \\&\quad + \left\| \mu \big (\psi (v)\big )u - \mu \big (\psi ({\tilde{v}})\big ){\tilde{u}}\right\| _\infty . \end{aligned} \end{aligned}$$

Next, we use the expression of \(\gamma _0(v)\) given in (9):

$$\begin{aligned} \begin{aligned} \left\| \gamma _0(v)u - \gamma _0({\tilde{v}}){\tilde{u}}\right\| _\infty \le a\left\| v^2 u - {\tilde{v}}^2 {\tilde{u}}\right\| _\infty + 2ab \left\| v u - {\tilde{v}} {\tilde{u}}\right\| _\infty + (ab^2+c) \left\| u - {\tilde{u}}\right\| _\infty . \end{aligned} \end{aligned}$$

But we have

$$\begin{aligned} \begin{aligned} \left\| v u - {\tilde{v}} {\tilde{u}}\right\| _\infty&\le \left\| v u - v{\tilde{u}}\right\| _\infty + \left\| v{\tilde{u}} - {\tilde{v}} {\tilde{u}}\right\| _\infty \\&\le \left\| v\right\| _\infty \left\| u - {\tilde{u}}\right\| _\infty + \left\| {\tilde{u}}\right\| _\infty \left\| v - {\tilde{v}}\right\| _\infty \\&\le (\left\| U\right\| _X+\left\| {\tilde{U}}\right\| _X)\left\| U-{\tilde{U}}\right\| _X. \end{aligned} \end{aligned}$$

Similarly, we have:

$$\begin{aligned} \begin{aligned} \left\| v^2 u - {\tilde{v}}^2 {\tilde{u}}\right\| _\infty&\le \left\| v\right\| ^2_\infty \left\| u - {\tilde{u}}\right\| _\infty + \left\| {\tilde{u}}\right\| _\infty (\left\| v\right\| _\infty + \left\| {\tilde{v}}\right\| _\infty )\left\| v-{\tilde{v}}\right\| _\infty \\&\le (\left\| U\right\| ^2_X+\left\| {\tilde{U}}\right\| ^2_X)\left\| U-{\tilde{U}}\right\| _X. \end{aligned} \end{aligned}$$

Afterwards, we write:

$$\begin{aligned} \begin{aligned} \left\| \mu \big (\psi (v)\big )u - \mu \big (\psi ({\tilde{v}})\big ){\tilde{u}}\right\| _\infty&\le \left\| \mu \big (\psi (v)\big )u - \mu \big (\psi (v)\big ){\tilde{u}}\right\| _\infty \\&\quad + \left\| \mu \big (\psi (v)\big ){\tilde{u}} - \mu \big (\psi ({\tilde{v}})\big ){\tilde{u}}\right\| _\infty \\&\le \mu _0 \left\| u - {\tilde{u}}\right\| _\infty + \left\| {\tilde{u}}\right\| _\infty \times \mu _0 \left\| \psi (v)-\psi ({\tilde{v}})\right\| _\infty \\&\le \mu _0 \left\| u - {\tilde{u}}\right\| _\infty + \left\| {\tilde{u}}\right\| _\infty \times \left\| v-{\tilde{v}}\right\| _\infty \frac{ \mu _0 \varphi _0}{\sigma }e^{\varphi _0\left\| v\right\| _\infty {\bar{S}}}, \end{aligned} \end{aligned}$$

where we have used estimation (22) at the last step. Combining the above inequalities leads to

$$\begin{aligned} \left\| F_1(U) - F_1({\tilde{U}})\right\| _\infty \le k_1\big (\left\| U\right\| _X + \left\| {\tilde{U}}\right\| _X\big ) (\left\| U\right\| _X+\left\| {\tilde{U}}\right\| _X)\left\| U-{\tilde{U}}\right\| _X, \end{aligned}$$

where \(k_1\) is a continuous increasing function. We estimate \(\left\| F_2(U)-F_2({\tilde{U}})\right\| _\infty \) analogously.

It remains to estimate the norm of the difference \(\left\| F_3(U)-F_3({\tilde{U}})\right\| _2\). We write:

$$\begin{aligned} \begin{aligned}&\left\| \alpha \big (\psi (v)\big )v - \alpha \big (\psi ({\tilde{v}})\big ){\tilde{v}}\right\| _2 \\&\quad \le \left\| \alpha \big (\psi (v)\big )v - \alpha \big (\psi ({\tilde{v}})\big )v\right\| _2 + \left\| \alpha \big (\psi ({\tilde{v}})\big )v - \alpha \big (\psi ({\tilde{v}})\big ){\tilde{v}}\right\| _2 \\&\quad \le \left\| v\right\| _\infty \left\| \alpha \big (\psi (v)\big ) - \alpha \big (\psi ({\tilde{v}})\big )\right\| _2 + \left\| \alpha \big (\psi ({\tilde{v}})\big )\right\| _\infty \left\| v - {\tilde{v}}\right\| _2 \\&\quad \le \alpha _0 \left\| v\right\| _\infty \left\| \psi (v) - \psi ({\tilde{v}})\right\| _2 + \alpha _0 \left\| v - {\tilde{v}}\right\| _2, \end{aligned} \end{aligned}$$

where we have used the properties of the function \(\alpha \) given in (13). Now we use the embedding \(L^\infty (\varOmega ) \subset L^2(\varOmega )\) and estimation (22), which leads to

$$\begin{aligned} \begin{aligned} \left\| \alpha \big (\psi (v)\big )v - \alpha \big (\psi ({\tilde{v}})\big ){\tilde{v}}\right\| _2&\le C \left[ \left\| v\right\| _\infty \left\| v-{\tilde{v}}\right\| _\infty \frac{\varphi _0}{\sigma }e^{\varphi _0\left\| v\right\| _\infty {\bar{S}}} + \left\| v - {\tilde{v}}\right\| _2\right] , \end{aligned} \end{aligned}$$

where C denotes a positive constant. Since \(\left\| v - {\tilde{v}}\right\| _2\le C \left\| A^\eta (U - {\tilde{U}})\right\| _X\), we obtain:

$$\begin{aligned} \begin{aligned}&\left\| \alpha \big (\psi (v)\big )v - \alpha \big (\psi ({\tilde{v}})\big ){\tilde{v}}\right\| _2\\&\quad \le k_3\big (\left\| U\right\| _X + \left\| {\tilde{U}}\right\| _X\big ) \Big [\left\| A^\eta (U-{\tilde{U}})\right\| _X + (\left\| U\right\| _X+\left\| {\tilde{U}}\right\| _X)\left\| U-{\tilde{U}}\right\| _X \Big ], \end{aligned} \end{aligned}$$

where \(k_3\) denotes a continuous increasing function. Combining the above estimates leads to the desired conclusion.

Proof of Theorem 3

If \(s(0) > 0\), the spectral condition \(\text {Spec}(-{{\overline{A}}}) \subset \{z\in {\mathbb {C}}:\;\mathfrak {R}(z)\le -\varepsilon \}\) for some \(\varepsilon > 0\) follows directly from Lemma 1 and the above properties of the function s. Next, the local stability of the stationary state \(\big ({\bar{\rho }}(x),\,0,\,0,\,0\big )\) is guaranteed by the results of Yagi (2009) (Section 6.2).

If \(s(0) < 0\), if follows again from Lemma 1 that \(\text {Spec}(-{\bar{A}})\) contains a real positive value. To prove that the stationary state \(\big ({\bar{\rho }}(x),\,0,\,0,\,0\big )\) is unstable, by virtue of Corollary 5.1.6 in Henry (2006), it suffices to check that

$$\begin{aligned} \left\| F({\bar{U}} + z) - F({\bar{U}}) - F'({\bar{U}})z\right\| _X = O\big (\left\| z\right\| _X^p\big ), \end{aligned}$$

for z in a neighborhood of 0 and \(p>1\). To this end, we easily compute

$$\begin{aligned} F({\bar{U}} + z) - F({\bar{U}}) - F'({\bar{U}})z = \begin{pmatrix} -\gamma \big (z_2, \psi (z_2)\big )z_1 + \gamma \big (0, \psi (0)\big ) z_1 \\ -h\big (\psi (z_2)\big )z_2 + h\big (\psi (0)\big )z_2 \\ \alpha \big (\psi (z_2)\big )z_2 - \alpha \big (\psi (0)\big )z_2 \end{pmatrix}, \end{aligned}$$

for all \(z = (z_1,\,z_2,\,z_3)^T \in X\). Recall that \(X = L^\infty (\varOmega ) \times L^\infty (\varOmega ) \times L^2(\varOmega )\). We have

$$\begin{aligned} \begin{aligned} \left\| \gamma \big (z_2, \psi (z_2)\big )z_1 - \gamma \big (0, \psi (0)\big ) z_1\right\| _\infty&\le \left\| \Big [a(z_2-b)^2+c-(ab^2+c) \Big ]z_1\right\| _\infty \\&\quad + \left\| \Big [\mu \big (\psi (z_2)\big ) - \mu \big (\psi (0)\big )\Big ]z_1\right\| _\infty . \end{aligned} \end{aligned}$$

First, it is easily seen that

$$\begin{aligned} \left\| \Big [a(z_2-b)^2+c-(ab^2+c) \Big ]z_1\right\| _\infty = O\big (\left\| z\right\| _X^2\big ), \end{aligned}$$

as \(\left\| z\right\| _X\) tends to 0. Furthermore, since the function \(\mu \) and the operator \(\psi \) are differentiable, we have

$$\begin{aligned} \left\| \mu \big (\psi (z_2)\big ) - \mu \big (\psi (0)\big )\right\| _\infty = O\big (\left\| z\right\| _X\big ), \end{aligned}$$

from which it follows that

$$\begin{aligned} \left\| \Big [\mu \big (\psi (z_2)\big ) - \mu \big (\psi (0)\big )\Big ]z_1\right\| _\infty = O\big (\left\| z\right\| _X^2 \big ), \end{aligned}$$

as \(\left\| z\right\| _X\) tends to 0. We obtain

$$\begin{aligned} \begin{aligned} \left\| \gamma \big (z_2, \psi (z_2)\big )z_1 - \gamma \big (0, \psi (0)\big ) z_1\right\| _\infty = O \big (\left\| z\right\| _X^2) \end{aligned} \end{aligned}$$

as \(\left\| z\right\| _X\) tends to 0. Since the functions h and \(\alpha \) are similarly differentiable, it is shown analogously that

$$\begin{aligned} \left\| -h\big (\psi (z_2)\big )z_2 + h\big (\psi (0)\big )z_2\right\| _\infty = O \big (\left\| z\right\| _X^2) \end{aligned}$$


$$\begin{aligned} \left\| \alpha \big (\psi (z_2)\big )z_2 - \alpha \big (\psi (0)\big )z_2\right\| _2 = O \big (\left\| z\right\| _X^2), \end{aligned}$$

which proves (45) with \(p=2\). The proof is complete.

Approximation of the non-stationary advection equation (1) by the stationary advection equation (2)

In order to support the approximation of the non-stationary advection equation (1) by the stationary advection equation (2), we compute numerically the solution of the following system with stationary advection

$$\begin{aligned} {\left\{ \begin{array}{ll} \begin{array}{rcl} \varvec{a} \cdot \nabla \rho &{}=&{} -\sigma \rho + \varphi (\rho )v, \\ \dfrac{\partial u}{\partial t} &{}=&{} \beta \delta w - \gamma (v,\,\rho )u - f u, \\ \dfrac{\partial v}{\partial t} &{}=&{} f u - h(\rho ) v, \\ \dfrac{\partial w}{\partial t} &{}=&{} d \varDelta w - \beta w + \alpha (\rho ) v, \end{array} \end{array}\right. } \end{aligned}$$

and in parallel, we compute the solution of the following system with non-stationary advection

$$\begin{aligned} {\left\{ \begin{array}{ll} \begin{array}{rcl} \dfrac{\partial {{\tilde{\rho }}}}{\partial t} + \varvec{a} \cdot \nabla {{\tilde{\rho }}} &{}=&{} -\sigma {{\tilde{\rho }}} + \varphi ({{\tilde{\rho }}}){{\tilde{v}}}, \\ \dfrac{\partial {{\tilde{u}}}}{\partial t} &{}=&{} \beta \delta {{\tilde{w}}} - \gamma ({{\tilde{v}}},\,{{\tilde{\rho }}})u - f {{\tilde{u}}}, \\ \dfrac{\partial {{\tilde{v}}}}{\partial t} &{}=&{} f {{\tilde{u}}} - h({{\tilde{\rho }}}) {{\tilde{v}}}, \\ \dfrac{\partial {{\tilde{w}}}}{\partial t} &{}=&{} d \varDelta {{\tilde{w}}} - \beta {{\tilde{w}}} + \alpha ({{\tilde{\rho }}}) {{\tilde{v}}}, \end{array} \end{array}\right. } \end{aligned}$$

with the same initial conditions and the same parameters. In both cases, the solutions are attracted to the stationary heterogeneous solution \(\big ({{\bar{\rho }}}(x),\,0,\,0,\,0\big )\) given in Proposition 3, as depicted in the following Figs. 12, 13, 14 and 15.

Fig. 12
figure 12

Numerical results for \(\rho (x)\) and \({\tilde{\rho }}(x)\). Left: stationary advection. Right: non-stationary advection

Fig. 13
figure 13

Numerical results for u(x) and \({\tilde{u}}(x)\). Left: stationary advection. Right: non-stationary advection

Fig. 14
figure 14

Numerical results for v(x) and \({\tilde{v}}(x)\). Left: stationary advection. Right: non-stationary advection

Fig. 15
figure 15

Numerical results for w(x) and \({\tilde{w}}(x)\). Left: stationary advection. Right: non-stationary advection

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Cantin, G., Ducrot, A. & Funatsu, B.M. Mathematical modeling of forest ecosystems by a reaction–diffusion–advection system: impacts of climate change and deforestation. J. Math. Biol. 83, 66 (2021).

Download citation

  • Received:

  • Revised:

  • Accepted:

  • Published:

  • DOI:


  • Forest model
  • Water resource
  • Climate change
  • Reaction–diffusion–advection
  • Stability analysis

Mathematics Subject Classification

  • 92B05
  • 35K58
  • 35K65
  • 35P15