Ecological Complexity 8 (2011) 92–97 Contents lists available at ScienceDirect Ecological Complexity journal homepage: www.elsevier.com/locate/ecocom Intransitive loops in ecosystem models: From stable foci to heteroclinic cycles John Vandermeer * Department of Ecology and Evolutionary Biology, University of Michigan, Ann Arbor, MI 48109, United States A R T I C L E I N F O A B S T R A C T Article history: Received 24 July 2010 Received in revised form 11 August 2010 Accepted 13 August 2010 Available online 18 September 2010 Several forms of intransitive loops are investigated with regard to their stabilizing properties in simple dynamic models. The intransitive loops are stabilizing when there is an odd number of species in the system. Furthermore, an intransitive loop can stabilize a chain of negative interactions that would otherwise be unstable. ß 2010 Published by Elsevier B.V. Keywords: Competition Intransitive loops Heteroclinic cycles 1. Introduction Standard ecological theory treats competitive communities as if they exist along a gradient from those in which competition coefficients are small (presumably because niches do not overlap too much) to those in which competition coefficients are large. The first, dominance-controlled communities (Yodzis, 1978), are perhaps the most common in nature, if indeed Gause’s principle is approximately true. The second, founder-controlled communities, are perhaps less likely in that standard ecological theory predicts their downfall (i.e., all but one species are doomed to competitive displacement with the standard model). In this communication I explore the consequences of various structures that defy the basic conclusion of Gause’s principle, allowing many species to coexist even though their competition coefficients may be extremely large. The fundamental idea stems from the notion of an intransitive loop. May and Leonard (1975) first noted the unusual state of a competitive arrangement that mimicked the classical children’s game of rock, scissors, paper (the rock dulls the scissors, the scissors cuts the paper, the paper covers the rock). Using the simple Lotka–Volterra form, dx1 ¼ x1 ð1 x1 a12 x2 a13 x3 Þ dt (1a) dx2 ¼ x2 ð1 x2 a21 x1 a23 x3 Þ dt (1b) dx3 ¼ x3 ð1 x3 a32 x2 a31 x1 Þ dt (1c) * Corresponding author. E-mail address: [email protected]. 1476-945X/$ – see front matter ß 2010 Published by Elsevier B.V. doi:10.1016/j.ecocom.2010.08.001 they effectively noted that if a12 a13, a23 a21, and a31 a32, this would represent a case of intransitive competition (2 beats 1, 3 beats 2, but 1 beats 3). If the inequalities are sufficiently large, a heteroclinic cycle would arise, in which the carrying capacity of each species would be alternatively approached as the zero values for the other species are approached. The underlying dynamics are well-known mathematically (see May and Leonard, 1975) and can be described by simple qualitative graphs. There exists a surface intersecting the three axis at (0,0,1), (0,1,0) and (1,0,0) and any trajectory rapidly moves to this surface (the surface itself is a stable manifold). Once arriving at that surface the system oscillates with each species alternatively taking the dominant role. If the above inequalities are not very large, the system exhibits relaxed oscillations on the manifold (see Fig. 1a). But as the inequalities increase, a bifurcation point is reached in which the stable focal point on the manifold reverses stability and the system oscillates in an unstable fashion, continually approaching more closely the points intersected by the stable manifold (Fig. 1b). Note that the manifold itself is still stable, but the non-zero equilibrium points have become unstable. I will refer to this type of bifurcation as a ‘‘heteroclinic bifurcation’’ (Vandermeer, 2006), in that the three equilibria on the axes (the carrying capacities, are saddles, and trajectories approach them along one axis, but are repelled along another axis). Metaphorically, the surface to which all trajectories are attracted (the manifold) is itself with an internal gravitation, concave when equalities are small, such that oscillations tend to converge on its center, but reversing to become convex at the bifurcation point, such that the ‘‘high point’’ on the surface tends to repel the oscillations (compare Fig. 1a and b). Several important modifications of this basic bifurcation framework enter as we increase the dimensionality of the system. First, it is not possible to construct a stable or heteroclinic system with an even number of species. Second, with more than three [(Fig._1)TD$IG] J. Vandermeer / Ecological Complexity 8 (2011) 92–97 93 Fig. 1. The manifold on which an intransitive change is manifest. species, the Hopf bifurcation (point at which the manifold surface changes from concave to convex) results in a limit cycle and then, with further increase in the size of the inequalities, a heteroclinic cycle emerges, focused on the carrying capacities. It appears to be the case that all cases of an odd number of species follow this basic pattern, whereas all cases of even number of species result in extinction of half of the components, leaving the other half living independently at their carrying capacities. 2. Multidimensional niches and intransitive loops The metaphor of normal-like distributions along an environmental gradient representing competition continues to motivate undergraduate ecology students. If overlap of these distributions is taken as proportional to competition intensity, it is evident that intransitive structures are not possible. However, using the elementary idea of niche overlap coupled with Hutchinson’s notion of the niche as hypervolume, intransitive loops can emerge quite simply. Formally, if E represents an environmental gradient and W(E) represents some estimate of fitness at E, then the competition coefficient can be thought of as, cðW i \ W j Þ 0 W i ðEÞ dE ai j ¼ R 1 (1) where c is a scaling factor that indicates the intensity of competition in areas of the environmental gradient with overlap. It is evident from (1) that it is not possible to obtain the following simultaneous relations, a21 > a12 (2a) a32 > a23 (2b) a13 > a31 (2c) which define an intransitive loop. Since Wi \ Wj = Wj \ Wi, it follows from (2a) that Z 1 W 1 ðEÞ dE > 0 Z 1 W 2 ðEÞ dE 0 and from (2b) that Z 1 W 2 ðEÞ dE > 0 Z 0 1 W 3 ðEÞ dE whence the requirement from (2c) that, Z 1 W 3 ðEÞ dE > 0 Z 1 W 1 ðEÞ dE 0 cannot be true. Thus, with the formulation 1 all competitive interactions are transitive. It follows from this that permanent communities with strong competition (founder-controlled communities) are not possible, that the interactions will culminate in the complete takeover by species i, if aji > aij for all values of j. However, if there are two niche axes, E1 and E2, with fitness functions 1 Wi ðE1 Þ and 2 Wi ðE2 Þ for species i, then a simple extension of 1 is ai j ¼ c 2 " 1 Wi ðE1 Þ \ 1 W j ðE1 Þ R1 0 1 Wi ðE1 Þ dE # þ 2 Wi ðE2 Þ \ 2 W j ðE2 Þ R1 0 2 Wi ðE2 Þ dE whence condition 2 is easily satisfied, a graphical example of which is shown in Fig. 2. 3. Stabilizing a large competitive chain with a small intransitive loop Extremely simplified conceptual models of resource competition are frequently employed to investigate generalized consequences of particular competitive arrangements. One of the most popular, as noted above, is envisioning a series of species along a single resource gradient in which competition results in some sort of species packing rule. Thus, for example, beginning with the arrangement in Fig. 3a, if niche overlap (the intersections of the various curves) results in sufficiently high competition coefficients, the species become packed as in Fig. 3b. This is sometimes referred to as the ‘‘limiting similarity’’ of species on a resource gradient. However, a rather large number of species can be held together even if undergoing extremely high levels of competition if an intransitive loop is involved somewhere within the overall structure (which, as shown above, is possible if there is more than one niche dimension). To see the basic principle involved, we begin with a simple specific form as follows: dx1 ¼ x1 ð1 x1 ax2 Þ dt (3a) [(Fig._2)TD$IG] J. Vandermeer / Ecological Complexity 8 (2011) 92–97 94 Fig. 2. Intransitive loop in a two dimensional niche space. Species 1 and 2 interact in niche axis 1 with species 1 dominant (since its overall niche occupancy in this dimension is larger than species 2). Species 2 and 3 interact in niche axis 2, with species 2 dominant. Species 1 and 3 interact in niche axis 1, with species 3 dominant. dx2 ¼ x2 ð1 x2 ax3 Þ dt (3b) dx3 ¼ x3 ð1 x3 ax1 Þ dt (3c) dxi ¼ xi ð1 xi axi1 Þ dt for i ¼ 4 . . . s (3d) Setting a = 1.1 (ensuring complete dominance by species i over species i 1 for all species indices >3), it is evident that an arbitrarily large value of s can be sustained. If any of the three species in the intransitive loop are eliminated (i.e., i = 1, 2, or 3) the system decomposes to a system of s/2 species, with every other species extinct (as in Fig. 3b). Thus, the competitive sorting process [(Fig._3)TD$IG] Fig. 3. (a) Representation of the niches of seven species on a single niche axis. (b) Resolution of the strong competition represented in (a). Conceptually the arrangement in (b) is maximal species packing, with overlaps between adjacent niche occupancies small. that leads from a situation like Fig. 3a to the species packing and limiting similarity of Fig. 3b, is effectively cancelled by the introduction of the intransitive loop. Consider the simple example of Eqs. (3) with s = 6, as diagrammed in Fig. 4a. If a is less than the Hopf bifurcation point for the subsystem {1,2,3}, the system as a whole, with all six species, is maintained at a focal point equilibrium (Fig. 4b). On the other side of the Hopf bifurcation the system oscillates with each of the three transitive species perfectly out of phase with one another (Fig. 4c). When in the heteroclinic orbit phase it is intuitively obvious what is happening. If species 3 is at a low point in its cycle, species 4 is able to increase, thus putting pressure on species 5, which consequently is less able to put competitive pressure on species 6. However, when species 3 increases in its oscillatory cycle it puts competitive pressure on species 4, which releases the pressure on species 5, which can then put more competitive pressure on species 6. Basically the same dynamics is operative to the left of the Hopf bifurcation (Fig. 4b), but it is less intuitively obvious since the system as a whole approaches, in the limit, a nonoscillatory situation. It is evident that the s in Eq. (3d) can be arbitrarily large. That is, the species in a transitive nonreciprocal chain (as in Fig. 3a, but with competition going in one direction, and connected to an intransitive loop) can be maintained in an arbitrarily large assemblage even with strong competition coefficients, as long as the system is not to the right of the Hopf bifurcation, at which point the heteroclinic oscillations eventually drive one of the species close enough to zero to conclude that a local extinction has occurred (formally, the actual arrival at zero population density is only in the limit). It is also evident that the intransitive loop can be much larger and give precisely the same results, but only if there is an odd [(Fig._4)TD$IG] J. Vandermeer / Ecological Complexity 8 (2011) 92–97 95 Fig. 4. Illustration of the stabilizing potential of an intransitive loop. (a) basic structure of system (equation system 2). (b) Time series of all six species prior to Hopf bifurcation. (c) Time series of all six species after the Hopf bifurcation. number of species in the loop. However, in three dimensions the Hopf bifurcation leads immediately to a heteroclinic cycle, whereas in all odd dimensions greater than three, the Hopf bifurcation leads first to a stable limit cycle, which, with increasing competition coefficients, eventually leads to a heteroclinic cycle. A complete bifurcation diagram for a 5D intransitive loop is shown in Fig. 5. coefficient. Thus the equations, dx1 ¼ x1 ð1 x1 ax2 Þ dt (4a) dx2 ¼ x2 ð1 x2 ax3 Þ dt (4b) dx3 ¼ x3 ð1 x3 ax1 bx4 Þ dt (4c) dx4 ¼ x4 ð1 x4 ax6 bx3 Þ dt (4d) 4. Coupling intransitive loops Since intransitive loops are inherently oscillatory, it makes sense to treat two or more of them as coupled oscillators. That is, a competitive community can be constructed in which two (or more) [(Fig._5)TD$IG]3D loops are coupled with a simple extension of the competition Fig. 5. Bifurcation diagram for the basic 5 species intransitive loop. Two bifurcation diagrams are superimposed here, the first on an arithmetic scale (the left y axis) showing local maxima and local minima, the second on a logarithmic scale (the right y axis), showing only the local minimum. The small curved arrows indicate which bifurcation diagram is associated with which axis. [(Fig._6)TD$IG] 96 J. Vandermeer / Ecological Complexity 8 (2011) 92–97 Fig. 6. Limit cycle resulting from weak coupling of two three dimensional intransitive loops. Note the ‘‘ghost’’ of the two original focal point attractors in the cycle. dx5 ¼ x5 ð1 x5 ax4 Þ dt (4e) dx6 ¼ x6 ð1 x6 ax5 Þ dt (4f) represent a coupling of two 3D intransitive loops through competition between species 3 and 4. As with other coupled oscillators, this coupling of species 3 and 4 brings the two oscillating systems into coordination with one another. In this particular form, for almost any strength of the coupling, b, the system moves into a phase locked state with the pairs {1,5}, {2,6} and [(Fig._7)TD$IG] {3,4} in complete reverse phase with one another. Given the almost perfect out-of-phase locking of the system, its overall behavior can be examined more easily with the following transformations, D1;5 ¼ x1 x5 D2;6 ¼ x2 x6 With a set such that the uncoupled system is in its stable oscillatory mode, increasing b causes the system to move rapidly to a limit cycle. The details of that cycle reflect the original two Fig. 7. The heteroclinic cycle resulting from an increase in the strength of the coupled nontransitive loop. (a) General structure of the coupled system. (b) Phase plane in the delta space (see text) for two of the three transformed variables. (c) 3D phase plane in the delta space. (d) Time series illustrating the increasing time spent near the zero and 1.0 points for each of the delta variables (note the increasing time spent near one of the equilibrium points, a characteristic of heteroclinic cycles). J. Vandermeer / Ecological Complexity 8 (2011) 92–97 focal point attractors (Fig. 6). Increasing the coupling (increasing the value of b in system 4), the system retains the overall structure of the original coupled cycle (Fig. 6) but rapidly moves toward a heteroclinic cycle (Fig. 7). In coupling larger intransitive loops, similar qualitative behavior is observed, namely, at low levels of coupling and either focal point attractors or simple limit cycles in the original loops, the coupled system synchronizes. As the coupling increases, the oscillations become more complicated and frequently become chaotic. Finally, with yet further increases in coupling intensity there is a heteroclinic bifurcation and some component(s) of the system are effectively lost (approach the value of zero in the limit). 5. Conclusions For any ecosystem composed of strongly competitive species, as that ecosystem becomes large, it is likely that intransitive loops will exist. Here the dynamic nature of such loops is discussed. Somewhat surprisingly, an intransitive loop can act as a stabilizing force, in that other species may be influenced by it in a way that prevents what would otherwise be a competitive exclusion. This may result in a competitive chain that is almost completely transitive, yet stabilized by a single intransitive loop. Thus, the classical notion that species arranged along a single environmental gradient will sort out such that overall competition is somehow minimized (the species packing effect) may be dramatically altered if an intransitive loop is involved. However, not all intransitive loops generate this stabilizing effect. Loops containing an even number of species are inherently unstable, with half of the species surviving by virtue of the fact that the other half disappear. Thus, only loops with an odd number of 97 species have this effect. Furthermore, a generalized pattern is observed as the competition coefficients connecting the species in the loop become large. As graphically illustrated in Fig. 1, the system moves from a stable oscillation to a heteroclinic cycle. More detailed examination (Fig. 5) shows a pattern of a stable focal point passing through a Hopf bifurcation leading to a limit cycle and eventually to a heteroclinic cycle. Elementary considerations of the relationship between competition coefficients and the ecological niche reveal that intransitive loops cannot exist on a single niche dimension, but can easily be envisioned on two or more niche dimensions. Since intransitive loops are likely to emerge in large ecosystems and since they are fundamentally oscillatory, the nature of the oscillations under coupling is of interest. Generally, coupling intransitive loops generates complicated cycles, frequently chaotic. However, a generalization seems to emerge in that coupling intransitive loops (with odd numbers of species) seems to follow a similar bifurcation pattern as is observed when increasing the competition coefficient in a simple odd species loop (Fig. 5). A stable focus yields a limit cycle which goes through cascading bifurcations to chaos and eventually leads to a heteroclinic cycle, sometimes a very complicated one (Fig. 7). References May, R.M., Leonard, W.J., 1975. Nonlinear aspects of competition between three species. SIAM J. Appl. Math. 29, 243–253. Vandermeer, J., 2006. Cascading thresholds to heteroclinicity in an ecosystem model. Int. J. Ecodyn. 1, 373–379. Yodzis, P., 1978. Competition for Space and the Structure of Ecological Communities. Springer-Verlag, Berlin.
© Copyright 2026 Paperzz