Are all plant populations metapopulations?

The nature of the regional dynamics of a species is critical to its survival. Imagine, for instance, that we are attempting to design a programme to conserve the regional population of a species. If the regional population is a metapopulation, then conservation needs to be aimed at conserving habitat. This is because, as a consequence of the balance between colonization and extinction, metapopulations do not occupy all suitable habitat. Metapopulation persistence relies on the existence of a certain amount of suitable but currently unoccupied habitat. In contrast, in regional ensembles and non-equilibrium metapopulations, conservation efforts have to be aimed at individual populations, because colonization and the creation of new populations is rare, and the survival of the ensemble would be sensitive to increased extinction. Alternatively, in spatially extended populations the availability of suitable habitat will be the key and populations will be sensitive to factors that change average suitability across large areas, such as fragmentation or climate change. We use this hypothetical example to make the point that different types of populations differ fundamentally and qualitatively in the nature of their regional dynamics. This rationale, we believe, is crucial to the future development and application of plant population ecology at a large scale, and is fundamental to our response to Ehrlén & Eriksson (2003). In our review of plant regional dynamics (Freckleton & Watkinson 2002) we highlighted these differences as a consequence of different types of regional population structure by reference to a series of case studies. Ehrlén & Eriksson argue that our distinction between such forms is unnecessary – we are not wedded to these typologies but believe it is important to recognize that not all plant populations have a metapopulation structure, even when individuals are distributed across the landscape in a patchy manner. Their critique of our paper is divided into three sections which address the following issues: (i) that the temporal or spatial scale of study may affect whether populations are deemed to be metapopulations or not; (ii) whether regional processes are important and whether positive relationships between occupancy and isolation imply a balance between colonization and extinction; and (iii) that the metapopulation notion may be useful for plants if the theory is modified appropriately. In the following sections we respond to these points in order. ‘Anything you can do, I can do meta’, McIntosh (1995) quoted in Reich & Grimm (1996) Ehrlén & Eriksson argue that our classification for plant regional dynamics is unnecessary and that the examples we cite as not being metapopulations could be regarded as metapopulations simply by expanding the spatial or temporal scales at which we observe the systems. For instance, they suggest that by observing a regional ensemble, such as the populations of Vulpia ciliata studied by Watkinson et al. (2000), over long time periods (many decades) then it might be possible to use a metapopulation-like model, predicting dynamics from decade to decade or century to century. This suggestion seems to us to be unhelpful for several reasons. Firstly, predictions of dynamics are generally required for ecological systems over much shorter time scales. Secondly, over larger time-scales it seems exceedingly likely that much higher order processes (e.g. climate change, land-use change, and biotic changes such as succession) will occur. Thus, average colonization or extinction rates will not remain constant, and the amount or nature of suitable habitat will change, with the consequence that extinction and colonization will not balance. Thirdly, this re-scaling of dynamics ignores important details of dynamics over the short term. Stochasticity and unpredictability in changes in the number of populations may be a characteristic of the dynamics of regional ensembles and non-equilibrium metapopulations over short time-scales (i.e. decades), which is likely to be important in predicting persistence. Finally, given that observations are typically not available for populations for more than a few years, it seems impractical to attempt to study population dynamics within a conceptual framework which requires measurements of colonization and extinction over periods of decades or more. Ehrlén & Eriksson also suggest that changing spatial scale may be used in order to re-classify some populations as metapopulations. In our original paper (section entitled ‘A patchy population is not a metapopulation’) we highlighted that this is commonly done. Specifically, metapopulation models are often used to model spatial dynamics within local populations. The important contribution of metapopulation theory, or of studying the regional dynamics of a population, is that it makes possible prediction of the total population size of a species within a landscape or large region. Applied at too small a scale, such models only predict local population sizes and contribute nothing to understanding population dynamics at a larger scale. There are important problems in deciding what constitute local and regional populations. However these problems have arguably been exacerbated by shoehorning populations into the restrictive framework of metapopulation terminology, as pointed out by Berryman (2002) and Camus & Lima (2002). Thus, we do not believe the notion of a metapopulation used in our paper is too restrictive. We do not repeat here the definition of a metapopulation we used, or our reasons for choosing it. These may be found in our paper or, in more detail, in Hanski & Simberloff (1997). We stress that this definition of a metapopulation is not arbitrary and arises from the assumptions and predictions of metapopulation models over an ecologically realistic time frame. In this regard we believe that it is equally as important to state that a population is not a metapopulation, as to state that a population is a metapopulation. In contrast, by arguing that all populations are metapopulations simply avoids important problems of describing and predicting large-scale population dynamics over short to medium time-scales. Describing all populations as metapopulations defines away many of the problems of predicting large-scale plant population dynamics. We did not argue that regional processes are unimportant in determining population dynamics in general, but that the key characteristics of systems, such as regional ensembles at the regional scale, are determined by local processes. Ehrlén & Eriksson have thus misread our paper in this regard, although this point was made clear with our example of Vulpia ciliata. To summarize this example, the main characteristics of this population at the regional scale are (i) very low rates of extinction, (ii) this being despite the small size of individual populations, with most being only a few metres in extent, and (iii) very low rates of colonization and dependence of new patch colonization on unpredictable factors (Watkinson et al. 2000). The first of these characteristics result from the strong local-scale density dependence that buffers populations against changes, as well as low rates of extrinsically determined extinction (arguably a regional process). The second two result from highly restricted local dispersal and no mechanisms of long-distance dispersal between areas of suitable habitat. However, that does not mean that regional dynamics do not play a role, for instance in the creation of new populations. This was highlighted in our earlier paper (Watkinson et al. 2000). However in the landscape within which Vulpia ciliata currently exits, regional processes are weak because colonization rates are very low, and we are not aware that recolonization ever occurs, indicating that persistence and extinction rates are determined by local processes. Ehrlén & Eriksson make great play of the inference that the fact that isolation, occupancy and colonization are linked in several case studies implies that a colonization–extinction balance drives regional dynamics. It seems unsurprising that most populations show such relationships because (i) suitable habitat tends to be clumped and hence we would expect occupancy and colonization to be higher in areas where suitable habitat is more frequent, and (ii) very long-distance migration is rare in plants, even in those with specialized dispersal mechanisms. Even in regional ensembles in which colonization is very low, we would expect new populations to arise closer to existing populations than further away. Relationships between occupancy and isolation in plant regional dynamics should not be taken to infer the existence of a balance between colonization and extinction. Indeed it has been argued that even when there are relationships between occupancy and isolation, habitat configuration may not be the prime factor driving species’ distributions. For example, Dupré & Ehrlén 2002) state that: ‘… our results suggest that the quality of the habitat may often be more important for explaining species incidence than its spatial configuration’, Dupré & Ehrlén (2002, p. 804). Species’ occurrence in this example is largely a function of the availability of habitat of suitable quality. In fact the system studied by these authors would appear to be a previously continuous population/community (akin to our spatially extended population in an area of suitable habitat) that has become fragmented. On the basis that isolation and colonization are linked, Ehrlén & Eriksson cited this study in support of the notion that the regional distribution of species is determined mainly by the balance between colonization and extinction. It appears to explicitly contradict this interpretation. In terms of the role of regional vs. local processes, we suggest that abundance–occupancy relationships may provide more useful insights (e.g. reviewed by Gaston et al. 2000). Mobile species (e.g. birds) show strong positive abundance–occupancy relationships both intra- and interspecifically. Positive relationships of this sort are a prediction of metapopulation models and models in which dispersal between habitat types is very common (e.g. Hanski 2000; Watkinson et al. 2003) and result when the following three conditions hold: (i) new suitable habitat is colonized rapidly, (ii) increasing local habitat quality (e.g. through decreasing mortality or increasing resources) leads to an increase in local density, and (iii) increasing local density leads to an increase in the number of dispersal units (e.g. see Watkinson et al. 2003). The latter two of these conditions tend to hold for plant populations, because increasing habitat quality tends to increase densities of plants (at least until densities are reached at which self-thinning occurs), and increasing plant density tends to increase seed production and hence the production of dispersal units. However, the speed with which new habitat is colonized depends on the distribution of suitable habitat as well as on colonization ability. Thus if suitable habitat is rare, or if colonization of suitable habitat is low, condition (i) will not hold, and positive abundance–occupancy relationships will not occur. In Freckleton & Watkinson (2002) we argued that, at a regional scale, suitable habitat is likely to be rare for many species, and colonization rates are likely to be low. Thus it would be predicted that, for many plants, positive abundance–occupancy relationships are unlikely, and indeed this is the case at the national scale in the United Kingdom (Thompson et al. 1998). Thus, the evidence would seem to point to the conclusion that limited migration, together with patchy distributions of suitable habitat, often limits the ability of plants to respond to changes in the regional abundance of suitable habitat. These are exactly the ingredients that we argued were likely to lead to regional ensembles. In species where habitat is more continuously distributed and dispersal occurs over longer distances, we would expect positive relationships to result. In support of this, Riis & Sand-Jensen (2002) showed that in stream vegetation, in which habitat is continuously distributed and dispersal may occur over large distances, positive abundance–occupancy relationships occur, and the strength of the relationship depends on dispersal ability. Thus we believe that studying abundance–occupancy relationships in relation to plant life-histories are likely to provide more information about the nature of regional dynamics than isolation–occupancy relationships. The problem of characterizing the nature of regional dynamics is not one of semantics. Its importance is in determining the way in which we attempt to study and predict regional dynamics. By adopting an incorrect framework, false or misleading interpretations may be drawn. To illustrate this, consider the following statement in a recently published paper on the large-scale dynamics of forest plants: ‘In order to avoid problems with the metapopulation approach arising from variations in the species pool, study sites should be carefully selected to be as homogeneous as possible with respect to habitat quality, i.e. similar climatic and edaphic parameters. However, this will reduce the potential application of such models to larger geographical areas or to vegetation types covering larger environmental gradients’, Dupré & Ehrlén (2002; p. 804). In order to study the regional dynamics of these species using a metapopulation approach, it appears that study populations have to be selected carefully from the regional pool of available populations. This is clearly nonsense, and underlines the problem of deciding a priori whether the metapopulation approach is the most appropriate. Rather, given that so little is known about the regional dynamics of plants, the importance and applicability of metapopulation theory should be tested; the scaling of local to regional dynamics is a key element in this. Ehrlén & Eriksson argue that it is wrong to develop terminology describing the nature of regional dynamics, citing Thomas & Kunin (1999) who point out that in many cases different ecologists would frequently describe the same population in different ways. In contrast, we believe that greater rigour in the application of theory would avoid disagreements. We have already argued (Freckleton & Watkinson 2002) that lack of rigour in applying theory is a common problem when studying local phenomena. Loosely applying the concept of metapopulation structure to any spatially structured population at any scale is another example of a lack of such rigour. We do, however, agree with Ehrlén & Eriksson that if there is to be an advance in our understanding of the determinants of the regional abundance of plant species then further studies are required to quantify dispersal, colonization, recolonization and extinction processes in a range of plant populations. Where we disagree with them is in the necessity to shoehorn all plant populations into the metapopulation framework. At heart, the fundamental prediction of metapopulation theory is very simple: at a large scale, populations persist as a consequence of a balance between colonization and extinction. It seems to us reasonable to state that if the regional population is either (i) not in such a balance, or (ii) is simply filling large tracts of suitable habitat, then the metapopulation theory is not applicable. This need to hang on to the term metapopulation in such cases is puzzling to us. In the introductory paragraph of this paper we highlighted the fact that such differences in regional population structure impact meaningfully on population persistence. Thus, this argument is not simply an issue of semantics.