Elsevier

Ecological Modelling

Volume 440, 15 January 2021, 109381
Ecological Modelling

A minimalistic model of vegetation physiognomies in the savanna biome

https://doi.org/10.1016/j.ecolmodel.2020.109381Get rights and content

Highlights

  • A Tree–Grass model with parameters depending on the Mean Annual Rainfall is studied.

  • Several ecological thresholds shape the dynamics of this fire-mediated model.

  • Our model verifies that increasing the fire frequency decreases the woody biomass.

  • Bifurcation diagrams related to fire frequency and Mean Annual Rainfall are given.

  • A user-friendly tool is available at https://gitlab.com/cirad-apps/tree-grass.

Abstract

We present and analyze a model aiming at recovering, as dynamical outcomes of fire-mediated tree–grass interactions, the wide range of vegetation physiognomies observable in the savanna biome along rainfall gradients at regional/continental scales. The model is based on two ordinary differential equations (ODE), for woody and grass biomass. It is parameterized from literature with respect to the African context and retains mathematical tractability, since we restricted it to the main processes, notably tree–grass asymmetric interactions (either facilitative or competitive) and the grass-fire feedback. We used a fully qualitative analysis to derive all possible long term dynamics and express them in a bifurcation diagram in relation to mean annual rainfall and fire frequency. We delineated domains of monostability (forest, grassland, savanna), of bistability (e.g. forest–grassland or forest–savanna) and even tristability. Notably, we highlighted regions in which two savanna equilibria may be jointly stable (possibly in addition to forest or grassland). We verified that common knowledge about decreasing woody biomass with increasing fire frequency is verified for all levels of rainfall, contrary to previous attempts using analogous ODE frameworks. Thus, our framework appears able to render more realistic and diversified outcomes than often thought of regarding ODE. Our model can help figure out the ongoing dynamics of savanna vegetation in large territories for which local data are sparse or absent. To explore the bifurcation diagram with different combinations of the model parameters, we have developed a user-friendly R-Shiny application freely available at : https://gitlab.com/cirad-apps/tree-grass.

Introduction

Savannas, as broadly defined as systems where tree and grass coexist (Scholes and Archer, 1997), occupy about 20% of the Earth land surface and are observed in a large range of Mean Annual Precipitation (MAP). In Africa, they particularly occur between 100 mm and 1500 mm (and sometimes more) of total mean annual precipitation (Lehmann et al., 2011, Baudena and Rietkerk, 2013), that is along a precipitation gradient leading from dense tropical forest to desert. There is widespread evidence that fire and water availabilities are variables which can exert determinant roles in mixed tree–grass systems (Scholes and Archer, 1997, Yatat Djeumen et al., 2018b and references therein). Empirical studies showed that vegetation properties such as biomass, leaf area, net primary production, maximal tree height and annual maximum standing crop of grasses vary along gradients of precipitation (Penning de Vries and Djitèye, 1982, Abbadie et al., 2006). It is widely accepted that water availability directly limits woody vegetation in the driest part of the rainfall gradient, see e.g. Sankaran et al. (2005). However. in the mesic and humid parts of this gradient, rainfall is known to influence indirectly the fire regime through what can be referred to as the grass-fire feedback (Yatat Djeumen et al., 2018b, Scholes, 2003 and references therein): grass biomass that grows during rainfall periods is fuel for fires occurring in the dry months. Sufficiently frequent and intense fires are known to prevent or at least delay the development of woody vegetation (Yatat Djeumen et al., 2018b, Govender et al., 2006), thereby preventing trees and shrubs to depress grass production through competition for light and nutrients. The grass-fire feedback is widely acknowledged in literature as a force able to counteract the asymmetric competition of trees onto grasses, at least for climatic conditions that enables sufficient grass production during wet months.

Dynamical processes underlying savanna vegetation have been the subject of many models. Some of them explicitly considered the influence of soil water resource on the respective productions of grass and woody vegetation components (see the review of Yatat Djeumen et al. (2018b)). Most of the models also incorporated the grass-fire positive feedback, several of them distinguishing fire-sensitive small trees and shrubs from non-sensitive large trees (Higgins et al., 2000, Beckage et al., 2009, Baudena et al., 2010, Staver et al., 2011, Yatat Djeumen et al., 2014, Yatat Djeumen et al., 2018b), while the rest stuck to the simplest formalism featuring just grass and tree state variables (Van Langevelde et al., 2003, D’Odorico et al., 2006, Higgins et al., 2010, Accatino et al., 2010, Beckage et al., 2011, Yu and D’Odorico, 2014, Tchuinté Tamen et al., 2014, see also the review of Yatat Djeumen et al. (2018b) and Fig. 1).

Models featuring the grass-fire feedback have shown that complex physiognomies displaying tree–grass coexistence (i.e. savannas) may be stable (Van Langevelde et al., 2003, D’Odorico et al., 2006, Baudena et al., 2010, Accatino et al., 2010, Yatat Djeumen et al., 2014, Tchuinté Tamen et al., 2014) along with more “trivial” equilibria such as desert, dense forest or open grassland. There are also field observations that report contrasted savanna–forest mosaics at landscape scale (see e.g. Fig. 2) that suggest bistability which is indeed among the outcomes of some models (see for instance Accatino et al. (2010), Staver et al. (2011), Tchuinté Tamen et al. (2014), Yatat Djeumen et al., 2014, Yatat Djeumen et al., 2018b and references therein). However, the ability to predict, along the whole rainfall gradient, all the physiognomies that are suggested by observations as possible stable or multi-stable outcomes was not fully mastered and established. Indeed, most models focused on specific contexts or questions and often feature parameters difficult to assess over large territories, especially in Africa (Accatino et al., 2010, Higgins et al., 2010, Baudena et al., 2010, De Michele et al., 2011, Beckage et al., 2011, Yu and D’Odorico, 2014). Nonetheless, the (Accatino et al., 2010)’s attempt was a seminal step in that direction but with some notable imperfections.

The Accatino et al. (2010) model was pioneering in the sense that it allowed these authors to provide a “broad picture”, by delimiting stability domains for a variety of possible vegetation equilibria or steady states as functions of gradients in rainfall, the most limiting resource, and fire frequency, the most widespread disturbance. This result was especially interesting and the considered model was sufficiently simple (two vegetation variables, i.e. grass and tree covers) to provide analytical projections. However, results from Accatino et al. (2010) were questionable regarding the role of fire return time. In fact, all over the rainfall gradient their model predicted that increasing fire frequency would lead to an increase in woody cover which contradicts empirical knowledge on the subject. The features of the model that led to this problem were barely debated in the ensuing publications. And more recent papers instead either devised more complex models or shift to stochastic modeling (see the review of Yatat Djeumen et al. (2018b)) that did not allow much analytical exploration of their fundamental properties.

In this paper, we aim to account for a wide range of physiognomies and dynamical outcomes of the tree–grass interactions system as observable at both regional and continental scales by relying on a simple model that explicitly address some essential processes that are: (i) limits put by rainfall on woody and grassy biomasses development, (ii) asymmetric interactions between woody and herbaceous plant life forms, (iii) positive feedback between grass biomass and fire intensity, and decreased fire impact with tree height.

Starting from Yatat Djeumen et al. (2018b), we explicitly express the growth of both woody and herbaceous vegetation as functions of the mean annual rainfall, with the aim to study model predictions in direct relation to rainfall and fire frequency gradients. Through the present contribution we aim at extending and improving a framework for modeling vegetation in the savanna biome through an ODE-based model, that is minimal (in terms of state variables and parameters), mathematically tractable and generic in the sense that its structure does not pertain to particular locations in the savanna biome.

An idiosyncrasy of our minimalistic tree–grass model is that we considered the fire-induced loss of woody biomass by mean of two independent non-linear functions, namely ω (see (3)) and ϑ (see (4)). Introducing these two functions, Tchuinté Tamen et al. (2017) showed that the previous model substantially improve previously published results on tree–grass dynamical systems (see also Yatat Djeumen et al. (2018b). For example, they showed that increasing fire return period systematically leads to woody biomass build-up with possible switch from grassland/savanna to forest. This result is entirely consistent with field observations (Bond et al., 2005, Yatat Djeumen et al., 2018b) and references therein). From this sound basis, we introduced improvements in the model which are exposed in the present paper. Notably, we now let influences of trees on grasses range from facilitation to competition according to climate.

The goal of the present paper is to present the improved version of the minimalistic tree–grass ODE model (Section 2) and show through a complete theoretical analysis (Section 3 and appendices) that it is able to provide, at broad scales, an array of sensible predictions about possible vegetation physiognomies that was not to date attained by tree–grass models of similar levels of complexity (in terms of the number of equations and/or the types of non-linearities). Predictions sensitivity to parameters ranges was assessed in Section 4. Relying only on qualitative results, we will construct a bifurcation diagram (Section 6) depicting the possible vegetation types along the rainfall vs. fire frequency gradients. Last but not least, in order to render our approach easy-to-use, we have developed a R-Shiny application (Section 5) to build the bifurcation diagram taking into account all the model parameters as to let them been changed easily according to the reader’s wish.

Section snippets

The minimalistic ODE model formulation

Our model features two coupled ordinary differential equations (Eq. (6) below) expressing the dynamics of tree and grass biomasses. Each equation entails a term of logistic growth (with parameters depending on MAP, Section 2.1) and terms of biomass suppression by external agents (e.g. herbivory grazers or browsers) and fire. Coupling of the equations occurs because fire intensity impacts woody biomass as a non-linear increasing function of grass biomass (see Section 2.3), while the grass

Qualitative analysis results

Our approach has kept the model amenable to a complete qualitative analysis of equilibria and stability thereof, as developed in the appendices. Equilibria embodying the long-term behavior of system (6) are summarized in Table 2, Table 3 in the case of competitive and facilitative influences of trees on grasses, respectively. Tables 2–3 result from the theoretical analysis of system (6) provided in Appendix A. For reader convenience, we recall in the following some key findings from the

Sensitivity analyses of model (6)

Interpretation of results from mathematical models of biological systems is often complicated by the presence of uncertainties in experimental data that are used to estimate parameter values (Marino et al., 2008). Moreover, some parameters are liable to vary in space, even in a given reference area. Sensitivity analysis (SA) is a method for measuring uncertainty in any type of complex model by identifying critical inputs and quantifying how input uncertainty impacts model outcomes. Different SA

Bifurcation diagrams and numerical simulations

We first provide bifurcation diagrams, based on the thresholds computation for the following set of parameters (see Table 4). We secondly present numerical simulations (also based on Table 4 values) to illustrate bifurcations in relation to mean annual rainfall (W) and fire frequency (f).

Thanks to the qualitative analysis of system (6)) (see Appendix A), any version of the bifurcation diagrams (see for instance Figs. 7–8, in terms of the fire frequency and the MAP, summarize the outcomes of the

Meaningful and diversified outcomes from a simple ODE framework

The present line of modeling aimed at demonstrating that meaningful and diversified outcomes can be expected from a full qualitative analysis of a parsimonious 2-dimensional models of grassy and woody biomasses interactions in the savanna biome. On the basis of a simple ODE framework, realistic results were indeed reached regarding how vegetation physiognomies change in relation to MAP and fire frequency. The model is liable to predict “trivial” equilibria, i.e. desert, grassland and forest as

Conclusion

In this paper, we presented and analyzed an improved version of a ‘minimalistic’ tree–grass model that addresses the influence of fire and rainfall (MAP) in tree–grass ecosystems. The model is minimalistic in terms of state variables and parameters, by only explicitly addressing essential processes that are: logistic growth of woody and grassy biomasses, asymmetric direct interactions thereof (both MAP-modulated), positive grass-fire feedback and decreased fire impact on large woody biomass. To

Supplementary materials

The R package containing the source code of the ‘Tree–Grass’ application is available at https://gitlab.com/cirad-apps/tree-grass.

CRediT authorship contribution statement

I.V. Yatat Djeumen: Conceptualization, Methodology, Formal analysis, Software, Visualization, Writing - original draft, Writing - review & editing. Y. Dumont: Conceptualization, Methodology, Software, Visualization, Writing - original draft, Writing - review & editing. A. Doizy: Software, Visualization, Writing - original draft, Writing - review & editing. P. Couteron: Conceptualization, Methodology, Validation, Writing - original draft, Writing - review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

VY and YD were supported by the DST/NRF SARChI Chair in Mathematical Models and Methods in Biosciences and Bioengineering at the University of Pretoria, South Africa (grant 82770). This work benefited from ongoing field investigation in Cameroon supported by Nachtigal Hydropower Company, Cameroon (Contract no C006C007-DES-2017). YD is funded by the European Union Agricultural Fund for Rural Development, by the Conseil Régional de La Réunion, the Conseil Départemental de La Réunion, and by the

References (90)

  • LefeverR. et al.

    Deeply gapped vegetation patterns: Oncrown/root allometry, criticality and desertification

    J. Theo. Ecol.

    (2009)
  • MarinoS. et al.

    A methodology for performing global uncertainty and sensitivity analysis in systems biology

    J. Theoret. Biol.

    (2008)
  • Martínez-GarcíaR. et al.

    Spatial patterns in mesic savannas: the local facilitation limit and the role of demographic stochasticity

    J. Theoret. Biol.

    (2013)
  • SchefferM. et al.

    Catastrophic regime shifts in ecosystems: linking theory to observation

    Trends Ecol. Evol.

    (2003)
  • StapeJ.L. et al.

    The Brazil eucalyptus potential productivity project: Influence of water, nutrients and stand uniformity on wood production

    Forest Ecol. Manag.

    (2010)
  • Tchuinté TamenA. et al.

    Tree–grass interaction dynamics and pulsed fires: Mathematical and numerical studies

    Appl. Math. Model.

    (2016)
  • Tchuinté TamenA. et al.

    A minimalistic model of tree–grass interactions using impulsive differential equations and non-linear feedback functions of grass biomass onto fire-induced tree mortality

    Math. Comput. Simulation

    (2017)
  • Van de KoppelJ. et al.

    Catastrophic vegetation shifts and soil degradation in terrestrial grazing systems

    Trends Ecol. Evol.

    (1997)
  • AbbadieL. et al.

    Lamto: Structure, Functioning, and Dynamics of a Savanna Ecosystem, vol. 179

    (2006)
  • AndronovA. et al.

    Theory of Bifurcations of Dynamic Systems on a Plane

    (1971)
  • ArchibaldS. et al.

    What limits fire? An examination of drivers of burnt area in Southern Africa

    Global Change Biol.

    (2009)
  • ArchibaldS. et al.

    Southern african fire regimes as revealed by remote sensing

    Int. J. Wildland Fire

    (2010)
  • AugierP. et al.

    Modelisation mathematique en ecologie. Cours et exercices corrigés

    (2010)
  • BanasiakJ. et al.

    Spreading speeds and traveling waves for monotone systems of impulsive reaction-diffusion equations: application to tree-grass interactions in fire-prone savannas

    (2019)
  • BarbierN. et al.

    Spatial decoupling of facilitation and competition at the origin of gapped vegetation patterns

    Ecology

    (2008)
  • BaudenaM. et al.

    An idealized model for tree-grass coexistence in savannas: the role of life stage structure and fire disturbances

    J. Ecol.

    (2010)
  • BaudenaM. et al.

    Complexity and coexistence in a simple spatial model for arid savanna ecosystems

    Theor. Ecol.

    (2013)
  • BeckageB. et al.

    Vegetation, fire and feedbacks: a disturbance-mediated model of savannas

    Am. Nat.

    (2009)
  • BondW.J.

    What limits trees in C4 grasslands and savannas?

    Annu. Rev. Ecol. Evol. Syst.

    (2008)
  • BondW.J. et al.

    The global distribution of ecosystems in a world without fire

    New Phytol.

    (2005)
  • BorgognoF. et al.

    Mathematical models of vegetation pattern formation in ecohydrology

    Rev. Geophys.

    (2009)
  • Braun, H.M.H., 1972a. Primary production in the Serengeti: purpose, methods and some results of research. In: IBP...
  • BraunH.M.H.

    Botanische Samenstelling Van De Vegetaties in De Serengeti PlainsTypescript Report to Wageningen University and the Serengeti Research Institute

    (1972)
  • BuciniG. et al.

    A continental-scale analysis of tree cover in African savannas

    Global Ecol. Biogeogr.

    (2007)
  • ChangW. et al.

    shiny: Web Application Framework for R

    (2020)
  • ColganM.S. et al.

    Topo-edaphic controls over woody plant biomass in South African savannas

    Biogeosciences

    (2012)
  • DioufA. et al.

    Relationships between fire history, edaphic factor and woody vegetation structure and composition in a semi-arid savanna landscape (Niger, West Africa)

    Appl. Veg. Sci.

    (2012)
  • D’OdoricoP. et al.

    A probabilistic analysis of fire-induced tree-grass coexistence in savannas

    Am. Nat.

    (2006)
  • FavierC. et al.

    Abrupt shifts in African savanna tree cover along a climatic gradient

    Global Ecol. Biogeogr.

    (2012)
  • FrostP.

    The ecology of miombo woodlands

  • GautierL.

    Contact forêt-savane en Côte d’Ivoire centrale: évolution de la surface forestière de la réserve de Lamto (sud du V-Baoulé)

    Bull. Soc. Bot. Fr. Act. Bot.

    (1989)
  • GovenderN. et al.

    The effect of fire season, fire frequency, rainfall and management on fire intensity in savanna vegetation in South Africa

    J. Appl. Ecol.

    (2006)
  • HigginsS.I. et al.

    Effects of four decades of fire manipulation on woody vegetation structure in savanna

    Ecology

    (2007)
  • HigginsS. et al.

    Fire, resprouting and variability: a recipe for grass–tree coexistence in savanna

    J. Ecol.

    (2000)
  • HigginsS.I. et al.

    The stability of African savannas: insights from the indirect estimation of the parameters of a dynamic model

    Ecology

    (2010)
  • Cited by (7)

    • Spatio-temporal modelling of tree-grass dynamics in humid savannas: Interplay between nonlocal competition and nonlocal facilitation

      2022, Applied Mathematical Modelling
      Citation Excerpt :

      However, as pointed out in Yatat-Djeumen et al. [2], modelling biomasses, instead of covers like in [37,38], helps to take into account the fact that plant types are not mutually exclusive at a given point in space since field studies suggested that grass often develops under scattered tree crowns (see Yatat-Djeumen et al. [2] and references therein). Moreover, [37,38] emphasized the effect of precipitation on possible vegetation transitions while Yatat-Djeumen et al. [39] suggested that, interplay between fire and water availability may give more realistic scenarios of vegetation distribution or transitions. Recently, Patterson et al. [40] proposed to bridge the gap between ecological models with macroscopic viewpoints (deterministic models) and microscopic descriptions of stochastic transitions (stochastic models).

    • A general ODE-based model to describe the physiological age structure of ectotherms: Description and application to Drosophila suzukii

      2021, Ecological Modelling
      Citation Excerpt :

      In addition, the versatility of the ODEs allows a more reliable description of the phenomenon, since it is easier to improve the equations by adding novel functions related to other biological aspects. The use of ODEs to describe biological phenomena has been recently and successfully applied also in systems involving multiple actors, such as in the case of Xylella fastidiosa (Brunetti et al., 2020) or vegetation physiognomies in the interaction between grass and trees in the savanna (Yatat Djeumen et al., 2021). These two models may represent a great potential for future extensions of the general theory introduced with this work.

    • A MODEL OF SEASONAL SAVANNA DYNAMICS

      2023, SIAM Journal on Applied Mathematics
    View all citing articles on Scopus
    View full text