7.2.1 Introduction

Measures of complexity in ecological communities are often based on the composition of the community represented by the number and relative abundance of species. These measures, which include the myriad diversity indices, do not tell us very much about the functioning of communities and their properties, such as stability. To quantify functional aspects of the community and understand how they may respond to perturbations, we need to quantify the interactions between species. The potential complexity of such interactions in real ecological communities is illustrated by the food webs in Fig. 7.2.

The food web can be characterized by the number of component species and the average number of links between them. The latter is known as linkage density. This idea can be extrapolated to other interactions between species, such as mutualisms. A motivation of this work has been to understand how communities with different linkage densities might respond to environmental perturbations, for example pollution events. A naive view of this suggests two alternative outcomes. A highly connected community (high linkage density) might be predicted to be more resistant to perturbations, for example because any one species will have many possible food sources. Alternatively we might imagine that a highly connected community will be less resistant because harmful effects on just a few species will potentially

No. of observations

- ioi-iooo

Festuca rubra

Alopecurus pratensis

Tetramesa brevioornis

Tetramesa brevioo!!is

Tetramesa brevioornis

Tetramesa brevioo!!is

Braoon spJ

Eurytoma sp. nr festuoae2. Syoophila sp.1

Braoon spJ

Eurytoma sp. nr festuoae2. Syoophila sp.1

Ahtoia atra -Tetramesa angustipennis

Tetramesa !inearis

Eurytoma sp.

Chlorooytus pu!ohrip< Syoophila sp.3 Homoporus febriou!osus3 Eurytoma tapio Endromopoda sp.3

I Pediobius festusoae

I Pediobius festusoae

Eíymus repens i

Ammophiía arenaria ,

Alamagrostis epigejos J

Tetramesa !inearis

Tetramesa hya!ipennis

Tetramesa oornuta

Tetramesa eximia

Tetramesa oa!amagrostidis Eurytoma sp.

Syoophi!a sp.4

Eupe!mus

Eurytoma f!avimana4 atropurpureus

Eurytoma sp.4 nr apioa!is Homoporus sp.4

Braoon erythorstiotus |
nil | |||||||||||||||||||||||||

Eurytoma roseni |
Ch!orooytus agropyri5 Pediobius a!aspharus5 Eurytoma sp. nr steffani Homoporus sp6 Homoporus fu!viventris Syoophi!a sp.6 Endromopoda sp. \Maoroneura vesiou!aris Ch!orooytus harmo!itae - Eurytoma danuvioa6 '//'' Syntomaspis baudysi Braoon sp.7 Endromopoda sp. Pediobius oa!amagrostidis Homoporus !uniger7 Eurytoma po!!ux Eschampsia cespitosa Dactylis glomerata . Phalaris arundinacea . Phleum pratense - Hypodium sylvaticum Tetramesa airae -Tetramesa ionguia -_ Tetramesa iongicomis 1 „ Tetramesa aibomacuiata --- " Tetramesa phieicoia "" — — _ Tetramesa fuivicoiiis 1 Pediobius Eurytoma appendigaster Homoporus sp. desohampiae Endromopoda sp. Pediobius sp. nr phra!aridis8 Ch!orooytus sp.8 Eurytoma oastor8 Pediobius daoty!ioo!a Eurytoma erdoesi Braoon sp. Ch!orooytus desohampiae \Mesopolopus graminum Ch!orooytus pha!aridis Eurytoma pha!aridis Endromopoda sp. Pediobius sp. Pediobius pha!aridis Chorooytus u!tioonus Pediobius o!aridgei Braoon sp. Eurytoma oo!!aris Pediobius p!aniventris ## Chorooytus formosusFig. 7.2 Examples of food webs. (a) A grass-insect herbivore-parasitoid community used to demonstrate the effects of sampling effort (Martinez et al. 1999). (b) A diagrammatic view of a salt-marsh food web illustrating the complexity with inclusion of parasites (Lafferty et al. 2006, 2008). Parasites are light-coloured and free-living species are dark-coloured. Pediobius eubius Fig. 7.2 Continued Table 7.1 A matrix of interaction coefficients for a hypothetical four-species community. On species . . . Interaction coefficient Table 7.1 A matrix of interaction coefficients for a hypothetical four-species community. On species . . . Interaction coefficient
cause major problems throughout the community. It is these types of questions that we will address in this chapter. One way of describing the strength of interactions in a community is to construct a matrix of interaction coefficients. For example, a hypothetical four-species community can be represented as shown in Table 7.1. The representation can include both feeding and non-feeding interactions. For example, the effect of species A on species B is negative with a strength of 0.5. The intraspecific interactions (e.g. A on A) are all negative. Some of the interactions are symmetric, for example a positive effect of A on D and D on A, which is a fully mutualistic interaction. A value of 0 indicates no link between the species. The analysis of ecological community stability is part of a wider debate about the stability or not of complex systems. Gardner and Ashby (1970) asked whether large systems (biological or otherwise) which were assembled 126 CHAPTER 7 1 0.80 0.20 0 20 40 60 80 100 Connectance (%) Fig. 7.3 The relationship between stability, connectance and component richness. A breakpoint or threshold of stability is observed at higher numbers of components. From Gardner and Ashby (1970). at random would be stable. In their words, they were concerned with 'an airport with 100 planes, slum areas with 104 persons or the human brain with 1010 neurons . . . [where] stability is of central importance'. Forty years on and these questions are no less pertinent; indeed, we may add an order of magnitude to some of these problems and present entirely new problems for solution; for example, global access to the Internet. Gardner and Ashby showed that for small numbers of components (n; e.g. neurons or species) stability declines with connectance between components (Fig. 7.3). As the number of components increases, the system moves rapidly to a breakpoint situation when a small change in connectance will result in a switch from stability to complete instability. These results were developed by May (1972) in an ecological context. He concluded that increased numbers of species do not automatically imply community stability and in fact may produce just the opposite effect (see Jansen & Kokkoris 2003 and references therein for further debate on these results). Increased stability with complexity was promoted by Elton (1958), whose conclusions were partly based on detailed case studies of invasive species, such as the giant snail Achatina fulica into Hawaii and the red deer Cervus elaphus into New Zealand, which contributed to dramatic declines in the endemic species of those islands (see May 1984 and Pimm 1984 for a critique of Elton's views). May also demonstrated that a species that interacts widely with many other species (high connectance) does so weakly (small interaction coefficient) and, conversely, those that interact strongly with 0.80 0.20 0 20 40 60 80 100 Connectance (%) Fig. 7.3 The relationship between stability, connectance and component richness. A breakpoint or threshold of stability is observed at higher numbers of components. From Gardner and Ashby (1970). others do so with a smaller number of species. He predicted that communities which are compartmentalized into blocks (effectively communities within communities) may be stable while the whole may not be. These ideas were explored further by Tregonning and Roberts (1979) who examined the stability of a randomly constructed model community in which the interaction coefficients were non-zero and the values chosen randomly. They began by running the model with 50 species, and used two methods of species elimination: a species was either chosen at random or the species with the most negative equilibrium value selected. Therefore, in the latter case they removed the most ecologically unrealistic species, as all species needed to have a positive equilibrium value. This process was continued until all species had a positive equilibrium value. This was defined by Tregonning and Roberts as the homeostatic system: one that was ecologically feasible and at equilibrium. Under selective removal the mean number of species comprising a homeo-static system was 25 and the largest 29. However, if elimination was random then the largest homeostatic system was 4 and the mean 3.3. Further understanding of these results requires a deeper insight into the nature of the community matrix used by May, Tregonning and Roberts and others. This is the subject of the next section. 7.2.2 Construction and stability of the community matrix Levins (1968) first devised a matrix of Lotka-Volterra competition coefficients to describe community structure and predict community stability. This was a multi-species version of the two-species competition model (described earlier). This idea was developed by May (1972, 1973a) giving a general version of the matrix called the community matrix (sometimes referred to as the stability matrix) which expressed the effect of species j on species i near equilibrium. The community matrix makes some assumptions about the dynamics of its constituent species; in particular, it assumes that prey and competitors will be regulated so that in the absence of any interspecies interactions they will return to equilibrium. It is also assumed that predators decline exponentially in the absence of prey. The community matrix allows insights into the stability of the community, with the dynamics of each species described by a nonlinear first-order differential equation. The aim is to create a matrix M of all interactions between the species in the community. It is assumed that species will be regulated so that, in the absence of any interspecies interactions, they will return to equilibrium. The following notation will be used: aij the interaction coefficients between species i and j (expressed as the effect of species j on the growth rate of species i), including aii, the intraspecific interaction; the magnitude of a ranges from 1 to 0, which represents no interaction (the sign of a is considered below); Nj density or biomass of species i; Ti the intrinsic rate of change of species i; s the number of species. A generalized Lotka-Volterra model following Roberts (1974) and Tregon-ning and Roberts (1979) and referred to by them as the multi-species quadratic model, is summarized for any number of species by: where Ti is positive for a producer (prey, competitor) and negative for a consumer (predator, pathogen) following the convention in Chapter 6. Therefore consumers decline exponentially in the absence of producers. Producers show density-dependent regulation as illustrated by the one-species version of equation 7.1: This equation is equivalent to the logistic equation dN/dt = r1N1 - t1Ni2/K with an equal to -r1/K. Equation 7.2 shows why Tregonning and Roberts referred to the model as the (multi-species) quadratic model. In Chapter 5 it was shown that populations described by the logistic equation had a stable equilibrium of K. The equilibrium occurs at dN/dt = 0; therefore, for equation 7.2: where N1* is the equilibrium population size. Factorize the right-hand side to give N1*(r + a11N1*) and rearrange the non-trivial solution (r + a11N1*) = 0 to give: As r must be positive for a single producer species, a needs to be negative to give a positive value of N1*. The sign of a is important and we will return to it later. (Note that the trivial solution is N* = 0.) With two species, equation 7.1 gives: We can compare the parameters a11, a12, a21 and a22 with the parameters in the competition and predator-prey equations in Chapter 6. In comparison with the competition equations, a11 = -r1/K1, a12 = -P12T1/K1, a21 = -P21T2/K2 and a22 = -r2/K2. Generalizing for interactions between species i and j, a = -Pj fj/Kj and substituting aii for -ri/Ki, a,j = Pij«ii; that is, the competition coefficient multiplied by the intraspecific interaction coefficient. Compared with the predator-prey equations 6.3 and 6.4 (assuming N1 is prey and N2 is predator): a11 = -r1/K1, a12 = -a, a22 = 0 and a21 = b. Also r2 will be negative and r1 will be positive. We can therefore see how, with different values of r and a,j, equation 7.1 can provide a generalized description of Lotka-Volterra dynamics covering interactions such as competition and predation. The community matrix is then derived by considering the community at equilibrium. If we take the two-species example: The species densities can be evaluated at equilibrium: In matrix form this is The values of Ni* and N2* can be calculated using matrix algebra by finding the inverse of the matrix of coefficients and multiplying both sides to give: Vaiia22 - a2iai2 J vaiia22 a21a12 . Matrix equation 7.7 can be generalized for any number of species as: -r = AN* where A is the square matrix of interaction coefficients and r and N* are column matrices of intrinsic rates of change and equilibrium densities respectively. Equilibrium values can then be found by matrix algebra as for equation 7.7 (equivalent to the solution of s simultaneous equations): To determine the community matrix we need to linearize the population growth equation (7.1) at equilibrium. This is achieved with a Taylor expansion or series. A series is defined in mathematics as the sum of a sequence of numbers. We have seen how sequences may arise in ecological processes in Chapter 1 (Fibonacci sequence). Various functions such as e or sin(x) can be expressed as a series. Using a Taylor series a function f(x + h) can be expressed as: f (x + h) = f (x) + hf'(x) + h2/2!f"(x) + h3/3!f'"(x) + ... where x and h are both variables. f'(x) means the derivative of the function evaluated at x whereas f" is the second derivative. If x represents the equilibrium density of a population which is described by a nonlinear function then when h is small (a perturbation from the equilibrium) the function near the equilibrium can be expressed according to the Taylor series as: That is, ignoring terms with h2 and higher because h is relatively small. Equation 7.10 describes the linear tangent at equilibrium. In ecological communities we may be dealing with the abundances of many species, all of which have an equilibrium. In this case linearized dynamics are represented as partial derivatives in the Taylor series. Partial differentiation is a method of determining change in one variable while one or more other variables are held constant. Partial derivatives are indicated by 3. Returning to equation 7.1 and writing it in general terms as: the Taylor expansion around the equilibrium is: \ dt dNj second- and higher-order terms (7.11) where n is a small perturbation from equilibrium. Fj(N*) is 0 and second-and higher-order terms can be ignored. The partial derivatives 3(dN;/di)/3Nj at equilibrium are aN*; that is, the interaction coefficient multiplied by the equilibrium population density of the ith species. You could check this for the two-species example. For example, to find 3(dNi/di)/3Ni at N*: Substitute for N1* and N2* to give: aiia22 a2iai2 22 a2iai2 This reduces to: which is anN1*. The full community matrix for the two-species example is: The stability of the community is found by determining the eigenvalue(s) of the community matrix. This tells us about the growth of a perturbation (n) from equilibrium. If the sign of the largest eigenvalue of the community matrix is negative then the community is stable; that is, the perturbations reduce back in size towards the equilibrium. A positive value indicates growth of the perturbation away from the equilibrium. Therefore we can see why the community matrix is sometimes referred to as the stability matrix. The magnitude of the dominant eigenvalue determines the return time of the community (Pimm & Lawton 1977), which measures the time taken for a perturbation to decay to 1/e of its initial value. To conclude this section we link up the graphical interpretation of stability of the logistic equation with the analytical method of the community matrix, following May (1973 a) and Pimm (1982). Recall that there are two equilibria with the logistic equation (N* = 0 and N* = K). To examine the stability of those equilibria in Chapter 6 we used a graphical method to examine perturbations (displacements) from equilibrium and asked whether those displacements will become larger with time. If the perturbations do become larger then the equilibrium is locally unstable. From the community matrix analysis we expect that the stable equilibrium of the logistic model is given by a negative slope of dN/dt with respect to N at equilibrium; that is, the single 'eigenvalue' is negative. This is indeed the case (Fig. 7.4). Fig. 7.4 Logistic curve showing unstable and stable equilibria. Note the gradient of the curves at N* = 0 and N* = K. Fig. 7.4 Logistic curve showing unstable and stable equilibria. Note the gradient of the curves at N* = 0 and N* = K. Positive gradient at equilibrium at equilibrium Positive gradient at equilibrium at equilibrium |

Was this article helpful?

## Post a comment