Copyright ©1995 Vera C.S. Vidigal-Jones
Do not use or copy without permission of the author

Metapopulation dynamics may facilitate cladogenesis

It has been argued that speciation may occur more rapidly and more commonly when a metapopulation structure is present than in a singly panmictic population (Levin 1995). I will argue complementarily that 1) A population within a metapopulation may also have higher probability of undergoing cladogenesis more rapidly than a single panmictic population of same size; 2) Tropical forests provide a suitable metapopulation environment; and therefore 3) The great animal diversity present in tropical forests may be correlated both with the tropical forest's environment and with metapopulation dynamics. I will review metapopulation dynamics and cladogenesis to address these hypotheses.

Metapopulation Dynamics

A metapopulation is a collection of local populations loosely connected by migration and isolated from the remainder of the species (Levin, 1995). The dynamics of a metapopulation occurs over time and is characterized by successions of random local population extinction events, dispersal, and colonization of empty patches. Hanski et al. (1995) further narrows the definition by specifying that most individuals of local populations stay in their natal patch, and that migration must be strongly distance dependent. The interconnected local populations may differ in size, and may be subject to different degrees of emigration, immigration, gene flow, genetic drift, local adaptation, and temporal persistence (Hanski, 1991; Patton, da Silva, and Malcolm in press).

I will follow Hanski's (1991) analysis of single-species metapopulation dynamics because of its comprehensiveness. According to this analysis, persistence of a metapopulation in a variable environment requires asynchronous dynamics of local populations, dispersal among them, and an optimal density at the local population level. Asynchronous dynamics increases a metapopulation persistence because it decreases the metapopulation vulnerability to regional stochasticity.

Hanski defines four kinds of stochasticities a metapopulation is subject to: demographic, environmental, immigration-extinction, and regional. Demographic and environmental stochasticities are the chances of individuals' death and birth in a local population. The chances are uncorrelated between individuals in demographic stochasticity and correlated in environmental stochasticity. Immigration-extinction and regional stochasticities are the chances of populations' extinction and persistence in a metapopulation. The chances are uncorrelated between populations in immigration-extinction stochasticity and correlated in regional stochasticity. The effect of demographic stochasticity in persistence is significant for local populations of small size. Immigration-extinction stochasticity is significant when the fraction of occupied patches is small (i.e., when the number of local populations is small). It is further increased if local population's extinction rate is high because all local populations might coincidentally go extinct at once. Environmental and regional stochasticities can occur unconditionally, affecting respectively several members of up to a whole local population at once and the whole metapopulation at once. Regional stochasticity is significant when there is little isolation between local populations.

Contingent to the species life history, "persistence requires that for a given extinction rate, the colonization rate exceeds a threshold value, and that for a given colonization rate the extinction rate is smaller than a threshold value." ... "Extinction rate decreases with increasing area of habitat patches" (Hanski, 1991). Extinction also decreases with increasing fraction of occupied patches because this increase is accompanied by increase in average population size (i.e., rescue effect).

Persistence also "requires that the average patch area is greater than a threshold value, and for a given average area of patches, their average degree of isolation must be smaller than a threshold value." Thus "a species may be missing from systems of small habitat patches, and from systems in which the average degree of isolation is great, even if the patches may offer temporary support to local populations." A small area of patches "may be compensated by little isolation, and great average isolation may be compensated by large areas" (Hanski, 1991).

Colonization rate of empty patches and migration between patches depend on dispersal rate, and it decreases with increasing isolation between patches, according to 'stepping-stone' dispersal. The more isolated patches are, mortality of dispersers increases, and gene flow between patches decreases. Dispersal rate has been found to be caused by natural selection, whereby with frequent extinctions of local populations, vacant patches become suitable to colonization, which selects for increased dispersal rate to compensate for the extinctions. The evolution of dispersal rate is not necessarily in the direction of an optimal fraction of patches occupied, because individual or genetic fitness when different may antagonize species fitness. Most dispersers are not successful in giving rise to a new local population.

The fraction of patches that are occupied depends on the ratio of extinction to successful colonization rates. The likelihood of colonization and occupancy of a patch is contingent on spatial variables (e.g., distance between patches and patch size). Extinction may increase with increasing fraction of occupied patches if the habitat is heterogeneous, however. This is because with a highly heterogeneous habitat, only a few favorable habitats may be available, thus the metapopulation structure will be of only a few large invulnerable local populations in the few favorable habitats, surrounded by relatively unstable populations in smaller habitat patches.

Average population size of a metapopulation depends on emigration and population growth rates and on dispersers' mortality rate, which is directly related to isolation between patches: Average population size increases with decreasing isolation due to decreasing mortality during dispersal. Average population size also increases with increasing fraction of occupied patches. Local population size increases with immigration and local population growth and decreases with emigration. Immigration significantly increases the growth rate of small local populations. Hanski's model assumes emigration to be density-independent. I question this assumption on the basis that as a patch population becomes very dense, high density should induce emigration.

Hanski (1991) also sets apart four kinds of metapopulation states with respect to average local population size and to fraction of occupied patches (Table 1). Table 1 summarizes the dynamics between the persistence parameters discussed above for four basic possible states of a metapopulation. All the states have been found in nature, although there is controversy as to the persistence allowed by the conditions of each state. A metapopulation can vary from one state to another through time depending on stochasticities, thus I will consider all states as valid possibilities, taking into consideration that a whole metapopulation may go extinct if it cannot evolve to a more stable and favorable state. A satellite species is likely to become rapidly extinct, but it may be a species in the beginning of its evolutionary history as a clade, right at or past the completion of its speciation process. It may therefore be ascending to a more favorable state, instead of in declining, even though its current state make the species most vulnerable. I also question the statement that rural or urban species are rare and likely to become rapidly extinct as discussed by Hanski. Rural or urban species may also be in a transition to a more favorable state. In addition, not sufficient fundamental life history and population structure information is known about a wide number of species likely to exist in a metapopulation dynamics (e.g., in the Amazon. Patton, Da Silva, and Malcolm, in press, address this issue) to regard these alternative states unviable in long term persistence. Persistence of the states summarized in Table 1, as well as any variation between them, is also contingent to how isolated patches are. I add that the magnitude of each of these variables varies according to aspects of life history of the organism in question (for example, body size and basic metabolic needs), although the core of the above relationships may hold true.

Table 1. Temporal/spatial characteristics of four metapopulation states.
Species StateFraction of occupied patchesAvg. Local population sizePopulation growthRatio of extinction to colonization
Core Specieslargelargehighlow
Satellite Speciessmallsmalllowhigh
Rural Specieslargesmalllowlow
Urban Speciessmalllargehighhigh

Population growth rate is the ratio of growth rate to emigration rate.

Hanski's analysis demonstrates that metapopulation dynamics are highly complex. Any study of or discussion about metapopulations should take the several interrelated parameters discussed above into consideration.


Bush (1994) defines speciation as the splitting of lineages. "A process whereby gene flow is reduced sufficiently between sister populations to allow each to become irrevocably committed to different evolutionary paths." In this discussion, I use the term cladogenesis and speciation interchangeably in agreement with this definition. I focus on cladogenesis processes in sexual organisms, following the review by Rice and Hostert (1993) and Bush (1994).

There are five models of speciation: basic allopatry model of vicariant speciation, reinforcement model of allo-parapatric speciation, bottleneck model of peripatric speciation, and divergence-with-gene flow model of either parapatric or sympatric speciation. Allopatry, reinforcement, and bottleneck models all involve the assumption of zero gene flow for the onset of the speciation process.

In the basic allopatry model a species is physically divided by a temporal and/or spatial barrier, so that no gene flow exists between individuals from the two divided subpopulations. To fit this model, the subpopulations must be "sufficiently large to exclude inbreeding as a factor in the speciation process" (Bush, 1994). Adaptation to different environments and/or genetic drift generate genetic differentiation that incidentally produces prezygotic and/or postzygotic reproductive isolation as a byproduct. Prezygotic reproduction isolation occurs by positive assortative mating, which reduces the production of hybrids. It is "due to pleiotropy of genes built up directly via selection or indirectly via tight linkage and genetic hitchhiking," and genetic drift may either "contribute to or detract from isolation among populations" (Rice and Hostert, 1993). Postzygotic reproduction isolation is the production of inferior hybrids with reduction of viability and/or fertility of hybrids.

In the reinforcement model of allo-parapatric speciation, the onset of speciation also occurs in allopatry, but the temporal/spatial barrier breaks down before complete reproductive isolation has evolved. Heterotypic matings produce low-fitness hybrids, and this selects for positive assortative mating. Once positive assortative mating is complete, the speciation process is complete.

I couple bottleneck model and peripatric speciation because both processes have as initial condition a sudden circumstance in which the population size is extremely low, be it caused by drastic reduction of population size in the bottleneck model or by colonization by one or a very small number of founding dispersers. Genetic differentiation between this small population and the extinct large population (in the case of bottleneck) or the original stock population (in the case of peripatry) occurs by inbreeding, selection, genetic drift, genome reorganization, and morphological and ecological shifts. These then lead to reproductive isolation from the original population as a byproduct.

Divergence-with-gene flow model of parapatric speciation occurs when two populations adapting to adjacent habitats are subject to disruptive runaway selection due to different environmental conditions. This model has two contingencies: 1) This is the single-variation model of isolation by pleiotropy/hitchhiking, whereby genetic variation caused by disruptive selection must produce prezygotic isolation as a byproduct of pleiotropy and/or genetic hitchhiking, causing the antagonism between selection and the homogenizing effect of recombination to be bypassed; 2) Disruptive selection must be sufficiently high relative to gene flow to overcome the antagonism between selection and the homogenizing effect of gene flow caused by migration between contiguous populations.

In the divergence-with-gene flow model of sympatric speciation cladogenesis occurs within the dispersal range of a population's offspring. Although there is the least amount of temporal/spatial segregation between populations in this model, some segregation still occurs, and it must be enough to cause runaway selection, which generates genetic based differences in habitat preference and fitness between each population. Positive assortative mating is again caused by incidental pleiotropy/hitchhiking. This model is subject to the same contingencies listed for parapatric speciation. Environmentally dependent postzygotic isolation may also occur in both divergent-with-gene-flow speciation models as a byproduct of pleiotropy/hitchhiking associated with genes that adapt populations to different environmental conditions.

Cladogenesis in a metapopulation context

There is much controversy about which are the main natural modes of cladogenesis. In discussing cladogenesis in a metapopulation context, I will take the models above into consideration, addressing the controversies as it appears appropriate. In a metapopulation context all models of speciation are possible, but the likelihood of each of them changes positively or negatively depending on the given metapopulation dynamics.

Cladogenesis requires a long period of time, genomic reorganization, and concurrent maintenance of genetic variation. Given that "major demographic change or dynamic is a prerequisite for significant evolutionary progress" (Slatkin, 1989 cited in Levin 1995), a metapopulation may be the best domain for cladogenesis to occur. Metapopulation dynamics allow for greater speciation occurrence and in less time than in a single population because a metapopulation is more likely to have longer persistence time, because it retains genetic variation more readily, and because it offers more opportunity for genomic reorganization through stochastic and selection processes (Levin, 1995). Metapopulations of local population size N will persist longer than a single population of size N, and, if colonization rate is sufficient and environmental variance experienced is not strongly correlated, a metapopulation of size N composed by several x populations of size N/x will also have persistence time longer than a single population of size N (Levin, 1995). In addition, time scale of local dynamics is much faster than the time scale of metapopulation dynamics (Hanski, 1991).

Due to the frequent local extinction and colonization events in a metapopulation and because selection is acting at each local population for a novel structural homozygote, the fixation rate of genetic novelty throughout a metapopulation is determined by the effective size of its local populations. Fixation of novelty is therefore greatly facilitated by an average small effective size of local populations, independent of the fraction of populations comprising the metapopulation. In addition, in a metapopulation with many local populations, several different chromosomal rearrangements may be at various stages in the fixation process, so a metapopulation may accumulate multiple chromosomal changes faster than single large panmictic populations (Levin, 1995).

In metapopulations with a large fraction of occupied patches, local populations may contain as much genetic variance as in a panmictic population of same size as the metapopulation. Thus subdivision of individuals into semi-isolated populations may increase the total genetic variance in the system as a whole at the expense of variation within each single population.

Genetic drift may affect the process of cladogenesis in two main ways. Genetic variance may be lost from a population through genetic drift, however in a metapopulation of several local populations with minimal gene flow between them, erosion of genetic variance by genetic drift occurs at a slower rate in the metapopulation as a whole than in a panmictic population of same size. This differential increases if periodic bottleneck events occur because the effective population size of a panmictic population will be affected more (Levin, 1995). Sudden erosion of genetic variance may highly affect the genetic health of a single population through inbreeding depression. In a metapopulation, the negative effects of inbreeding depression are constrained by the higher possibility that more than one local population survive a single bottleneck event, thus genetic variation erosion may occur at a pace conducive to speciation, but not fast enough to sacrifice the health of the metapopulation as a whole. The positive effect of genetic drift is that it can lead a population from one adaptive domain to another, then selection will direct the population to a new adaptive peak, according to Wright's shifting balance theory of evolution. A metapopulation structure of small local populations connected by low rate of migration is more conducive to evolutionary radiation into new adaptive domains than single populations (Lande 1980, Wright 1940, 1982 cited in Levin, 1995) because the smaller the effective population size the faster the transition to a new adaptive peak and the greater the likelihood of its occurrence. The successive colonization events of a metapopulation structure are very effective in exporting novel genotypes because there is no dilution by a recipient population (Levin 1995).

Metapopulation dynamics can cause a species to evolve to be generalists, if there is high degree of gene flow and a large fraction of occupied patches (e.g., a core species), or it can cause a species to evolve to be specialists, if the metapopulation undergoes bottleneck periods often, or if there is low degree of gene flow due to isolation between patches, so that isolated local populations can each evolve to be specialists to their habitat. In a patchy environment, high genetic diversity and metapopulation structure may enhance population growth rate and ability to adapt to new environmental conditions.

Taylor (1991) affirms that "isolated populations would be less correlated than populations linked in a metapopulation, and these would be less correlated than patches within a single population." The context of this statement is Taylor's a priori definition that a 'true metapopulation' is one in which "all sub-populations are essentially similar and equally dependent on dispersal." This statement does not seem to take into consideration the population size. It is theoretically true that two populations with no gene flow would be less correlated than populations linked in a metapopulation, but effective population sizes have to be taken into account. For a fair comparison, we need to consider the same population size. Stochasticities cause persistence of single small isolated populations highly unlikely, while two local populations of the same small size may persist in a metapopulation structure. In this case, the two isolated populations will not have long term evolutionary significance because they are too short lived. Additionally, there is a wide continuum in the extent of gene flow in a natural metapopulation, as well as in the extent of other parameters, that it may be unrealistic or too reductionist to arbitrarily select a single circumstance to be a 'true metapopulation.'

Cladogenesis at the metapopulation level

Hanski (1991) remarks that the dominant metapopulation structure in nature is of one or more large and practically invulnerable local populations in large and/or favorable habitat patches, surrounded by relatively unstable populations in smaller habitat patches (Harrison, 1991 as cited in Hanski, 1991).

Given this scenario, only the few large local populations are sources of colonists, each under genetic drift and directional selection according to Wright's shifting balance theory. Consequently metapopulation with short lived local populations will lose genetic variance faster than permanent populations because the fewer colonists and their source populations, the less genetic variance will be present at a faster rate. This loss of genetic variance could ultimately lead to genetic identification of all individuals in all local populations. This may result in the emergence of prezygotic and postzygotic isolation as a byproduct of pleiotropy/genetic hitchhiking between the metapopulation and the remainder of the species (Levin, 1995).

Levin (1995) thus proposes that the unit of speciation is the entire metapopulation. This process is homologous to the basic allopatry model described by Bush (1994) and Rice and Hostert (1993), whereby the metapopulation is isolated from the remainder of the species.

Cladogenesis within a metapopulation

A population that undergoes cladogenesis within a metapopulation buds off the network of the metapopulation. Given the small effective size of a local population, it can be a domain for much and fast evolutionary change, yet it may be protected from inbreeding depression by a minimum amount of gene flow. Cladogenesis of a local population is also contingent on long persistence time of the local population. This circumstance is more likely to occur in a metapopulation structure different from the one considered in Levin speciation process. Habitats may be heterogeneous in very distinct ways , and Levin's heterogeneity circumstance of few good patches may only be one possible situation. Another possible situation that may be conducive to speciation within a metapopulation is a fair number of favorable patches with heterogeneous distributions, whereby some patches are little isolated from each other and others are relatively isolated.

A population within a metapopulation may speciate by founder effect if a successful colonization of a new patch occurs. This is more likely to happen if the regional environment is patchy and variable but if at the same time resource richness and minimum environmental conditions for disperser survival does not decrease with distance. Thus colonization of distant favorable patches may allow peripatric speciation to occur as described by Bush (1994) and Rice and Hostert (1993).

Cladogenesis of a local population may also occur in sympatry or in parapatry. Parapatric speciation process may occur given that a metapopulation may be genetically diverse because local populations are undergoing chromosomal rearrangements and fixation uncorrelated between populations, yet variation within a single local population may be very low. It may be possible that a local population experiences enough amount of chromosomal rearrangement and speed of fixation, in a metapopulation with low gene flow (e.g., due to low migration rate or isolation between patches) and in a favorable patch that it becomes irreversibly committed to a different evolutionary path from the remainder of the metapopulation. Sympatric speciation may also occur according to Rice and Hostert's (1993) divergence-with-gene-flow model. A metapopulation in satellite or rural states may be conducive of sympatric speciation, as long as the patches are within dispersal range of each other.

Tropical forest diversity and metapopulation dynamics

Widespread populations of generalists connected by high degrees of gene flow allow for speciation of the whole metapopulation from the remainder of the species. Populations that are less phenotypicly plastic in adapting to patchy varying habitat may be conducive to cladogenesis within the metapopulation.

Areas like the Amazon basin are highly variable patchy environments. The patchy nature make it a prime environment for metapopulation dynamics to occur (Alvarez-Buylla and Garcia-Barrios, 1993). The resource richness and more constant climate of the low altitude Amazonian rain forest may provide a more favorable environment for species in rural, satellite, and such less commonly held favorable state variations to persist long enough to be a domain for cladogenesis within a metapopulation structure.

Interspecific interactions (for example, predator-prey systems) should be taken into consideration for a more comprehensive understanding of speciation within or at the metapopulation level. These important considerations are beyond the scope of this discussion.

Literature cited

Alvarez-Buylla, E. R., and R. Garcia-Barrios. 1993. Models of patch dynamics in tropical forests. Trends Ecol. Evol. 8:201-204.

Bush, G. L. 1994. Sympatric speciation in animals: new wine in old bottles. Trends Ecol. Evol. 9:285-288.

Hanski, I., J. Poyry, T. Pakkala, M. Kuussaari. 1995. Multiple equilibria in metapopulation dynamics. Nature 377:618-621.

Hanski, I. 1991. Single-species metapopulation dynamics: concepts, models and observations. Biological Journal of the Linnean Society 42:17-38.

Hanski, I., and M. Gilpin. 1991. Metapopulation dynamics: brief history and conceptual domain. Biological Journal of the Linnean Society 42:3-16.

Levin, D. A. 1995. Metapopulations: an arena for local speciation. Journal of Evolutionary Biology 8:635-644.

Patton, J. L., M. N. F. Da Silva, and J. R. Malcolm. 1995. Hierarchical genetic structure and gene flow in three sympatric species of Amazonian rodents. Molecular Ecology, In press.

Rice, W. R., and E. E. Hostert. 1993. Laboratory experiments on speciation: what have we learned in 40 years? International Journal of Organic Evolution 47:1637-1653.

Taylor, A. D. 1991. Studying metapopulation effects in predator-prey systems. Biological Journal of the Linnean Society 42:305-323.