On metapopulations and microrefugia: palaeoecological insights

Authors


Nicole A. Sublette Mosblech, Department of Biological Sciences, Florida Institute of Technology, Melbourne, FL 32901, USA.
E-mail: nsublett@my.fit.edu

Abstract

We highlight the importance of microrefugia in the light of population migration and genetic drift by synthesizing lessons learnt from metapopulation and palaeoecological studies. The concept of microrefugia is considered as a long-term variant of conventional metapopulations, in which microclimatic stability supersedes gene flow in determining species survival. Not all species can maintain populations in microrefugia. Life history traits such as small body size, the capacity for asexual reproduction, and species with light genetic loads favour survival. Microrefugia will facilitate faster rates of species responses to climate change than envisioned in diffusion models, and potentially provide a means to alleviate the negative effects posed by natural or anthropogenic barriers to migration. Predictive models based on relatively coarse-grained approaches that ignore microrefugia will lead to overestimates of extinction risk. Microrefugia should be identified and conserved, not for the species they contain, as these are likely to turn over with time, but as an important component of landscape diversity that will provide a safe haven for species not yet identified as at risk.

Introduction

Three of the most immediate challenges facing conservation biogeography, given the prospects of a rapidly changing climate, are to ascertain: (1) how species will respond to ongoing climate change, (2) whether rates of migration can match required range shifts (Hannah et al., 2002; Malcolm et al., 2002; Parmesan & Yohe, 2003), and thus (3) which species are most vulnerable to extinction. These newly recognized problems add uncertainty to the long-established attempts to quantify minimum viable populations and the space they require.

Since the inception of conservation biology as a science, conservation strategies have assumed that environments, in the absence of overt human activity, would remain more or less constant. A rich theoretical grounding for predictive reserve design has been established through island biogeography (MacArthur & Wilson, 1967; Diamond, 1975; Brown & Kodric-Brown, 1977), metapopulation studies (Levins, 1969; Hanski & Gyllenberg, 1993), conservation genetics (Avise & Hamrick, 1996) and neutral theory (Hubbell, 2001). However, these worldviews had at their core the expectation that populations were in equilibrium with climate.

Ecological niches are conserved as species track climatic change through time (e.g. Parmesan & Yohe, 2003; Visser, 2008). By considering the perceived niche of a species in terms of climatic needs, a species’ bioclimatic envelope, projected future range, or niche space can be modelled. Generally, such studies start by downscaling general circulation models to produce regional forecasts of bioclimatic envelopes. From those forecasts, projected range changes of key species are mapped under differing climate scenarios (e.g. Hannah et al., 2002). A further analytical step incorporates gap analysis to assess the existence of adequate habitat availability and the conservation status of those areas to support a particular population (e.g. Scott et al., 1993). All the above techniques are valuable, but all are essentially equilibrial views of ecosystems. Here, we draw from palaeoecological insights to focus attention on the importance of population outliers in cryptic habitats (Birks & Willis, 2008), or microrefugia [here defined as isolated populations surviving in unusual microclimates relative to the broader landscape (sensuRull et al., 1988; Rull, 2009)]. Within microrefugia, species experience climatic conditions that contrast with the average climatic envelope of the region yet are conducive to the presence and survival of the species.

Drawing on insights from Pleistocene biogeography and recognizing that macrorefugia could be large or small, the important characteristic within each Pleistocene macrorefugium was that the general climate of the region remained suitable for a particular species (Fig. 1). Within the southern peninsulas of eastern Europe, macrorefugial populations of many currently northern forest taxa persisted through glacial periods, although they may have escaped detection because of low population densities (Bennett et al., 1991). On the other hand, the more widespread presence of taxa, such as Picea in European and North American refugia (Davis, 1981; Bennett et al., 1991), throughout the last glacial indicates macrorefugial populations, in stark contrast to a small, remnant microrefugial population detected as having persisted in northern Ohio (Dachnowski, 1910).

Figure 1.

 Connectivity and environmental stability as complementary factors in metapopulation and microrefugial models. A hypothetical species is represented by metapopulations, with a main distribution and outlier local populations. When local populations are located near the macrorefugium (a and d) (mainland–island metapopulation), the local populations are recipients of immigrant genes, and genetic divergence among populations is extremely low. In a classic Levins model (b and e), where local populations have equal probabilities of colonization and extinction, local rates of dispersal between populations result in moderate genetic differentiation among populations. When local populations are far removed from the main population (c and f), and experience very rare dispersal between one another, the potential for genetic divergence is high. Survival in such microrefugia is greater within stable microhabitats, although still not guaranteed. Within the macrorefugium, some species may exist as small populations (a–c), whereas other species may exist throughout the range (d–f).

Over the last three decades, while biogeographers have wrestled with models of reserve design, palaeoecologists have developed an understanding that species migrated individualistically, and that communities were transient agglomerations of species (e.g. Davis, 1981; Webb, 1987). Key to this insight was the mapping of past pollen representations of species in North America and Europe using isopols (i.e. lines connecting locations with a similar percentage occurrence of a given taxon at a given time). In generating isopol maps, statistical cut-offs were adopted for what appeared to be significant representation of a species by its pollen; for example, the occurrence of > 25% for Pinus or 2% for Fagus had to be documented before the taxon was considered ‘present’ (e.g. Huntley & Birks, 1983; Webb, 1987; Shuman et al., 2002). Because pollen can be wafted long distances by wind, and rare grains contribute noise to data, such simplification using isopols was rational. As a consequence, occurrences of pollen falling below the cut-off were ignored. Subsequent genetic and fossil studies have revealed that in adopting such cut-offs, very important signals may have been discarded among the noise (McLachlan & Clark, 2004, 2005; Hu et al., 2009).

Predictions that thermophilous species in Europe and North America survived ice-age cooling in southern locations were substantiated through fossil pollen analysis (e.g. Delcourt et al., 1983; Huntley & Birks, 1983; Bennett et al., 1991). These localities have subsequently been referred to as macrorefugia. Where multiple refugia existed within a continent (e.g. in southern Europe: the Balkans, Iberian Peninsula and Mediterranean Italy), their populations have been found to be genetically distinct (e.g. Petit et al., 2003).

As species expanded out of refugia during interglacials, and contracted back during glacials, a legacy of that migration would be recorded in the population genotypes. The concept of northern purity and southern diversity developed as an expectation that founder effects would cause species to lose genetic diversity towards the expanding edge of their range, while maintaining diversity in the macrorefugium (Bennett et al., 1991; Hewitt, 2000; Hampe & Petit, 2005; Magri et al., 2006). Consequently, the initial colonists would have included only a few haplotypes, compared with many in the macrorefugial reservoir, and these haplotypes would have been disproportionately important to the spread of the population (Taberlet et al., 1998). Generally, the expectation of declining diversity with increased distance from a refugium was supported, as northerly populations that have expanded during the present interglacial are genetically less diverse and exhibit few unique haplotypes compared with refugial populations (Hewitt, 2000).

However, a significant problem with this simple model was that it called for unrealistically high plant migration rates (Clark et al., 1998). For populations to have moved across Europe or North America via diffusion, and to have occupied sites based on the chronology determined from fossil-pollen evidence, migration rates would have had to have been c. 100–1000 m per year. Such migrations would have been 2–5 times faster than those observed in modern studies (Clark et al., 2001). Furthermore, some taxa known to have very limited dispersal capabilities, for example woodland herbs estimated to migrate at c. 1 m per annum, appeared to keep pace with more easily dispersed trees (Cain et al., 1998). Early explanations for rapid rates of dispersal and recolonization of deglaciated landscapes emphasized the cumulative effect of chance, long-distance dispersal events (e.g. Birks, 1989), which seemed plausible for some species recolonizing previously glaciated landscapes (e.g. MacDonald & Cwynar, 1985; Cwynar & MacDonald, 1987). However, a more parsimonious explanation for the apparent mismatches in predicted and observed migration rates is the failure to include the effects of microrefugia.

Even as recently as 12,000 years ago, thousands of years after glacial ice began to recede, many European and eastern North American landscapes lacked most thermophilic trees. However, between 8000 and 4000 years ago, modern or near-modern communities were forming in many locations (e.g. Huntley & Birks, 1983; Davis & Shaw, 2001). At stake is whether the recolonization was an expansion out of southern refugia (with high rates of migration implied) or from northern microrefugia. This argument is critically important for predicting future responses to climate change.

The strongest genetic evidence supporting the existence of microrefugia comes from McLachlan & Clark (2005), who demonstrated that red maple (Acer rubrum) and beech (Fagus grandifolia) trees in northern North America had significant genetic differences between subpopulations, and that those subpopulations were not closely related to the classic macrorefugia in the states bordering the Gulf of Mexico. Rather, migration to those northern range margin sites appears to have begun within 200 km of the ice margin. Either palaeoclimatic reconstructions based on fossil pollen data exaggerated Last Glacial Maximum cooling close to the ice sheet (e.g. Loehle, 2007), or microrefugia offered locally modified microclimates that allowed persistence of populations in these locations. By accepting the presence of microrefugia, the inferred migrational rates for these species needed in the Holocene drop to a reasonable 80–90 m per year (McLachlan & Clark, 2005).

Data from European thermophilic trees are also consistent with the presence of microrefugia. Palaeoecological data of charred wood of thermophilous trees in Central Europe, dating to the Last Glacial Maximum (Willis & van Andel, 2004; Binney et al., 2009), suggest the ongoing presence of such taxa far from perceived macrorefugia. This growing body of evidence supports the long-held minority view that diffuse populations have been important in the recolonization of European forests (Beug, 1975; reviewed in Birks & Willis, 2008). Petit et al. (2003) document that the highest genetic diversity among populations of 22 European forest species occurs not within the classic refugial regions of the Balkans, Italy and Iberia, but north of the Alps. Clearly, this pattern is at odds with the expectation of a latitudinal decrease in genetic diversity, and Petit et al. (2003) attribute this to the area immediately to the north of the Alps being a ‘melting pot’ for recruits from the various macrorefugia. In the light of McLachlan & Clark's (2005) findings, the diversity peak north of the Alps could also represent diversification within microrefugia (Fig. 1).

Other evidence for the past importance of microrefugia comes from a wide array of sources. Phylogeographic studies of some species of small mammals and amphibians in Eastern Europe indicate survival within northerly microrefugia (Bhagwat & Willis, 2008), consistent with the idea that southern-facing slopes offered microhabitat protection for temperate-forest tree species and small animals as climates cooled in the last glacial period (Birks & Willis, 2008). Similarly, as climates warmed in the Holocene, cooler-than-matrix microrefugia have maintained unusual assemblages of species, such as saxifrages and other arctic–alpine plants on the northerly slopes of Cwm Idwal in North Wales, boreal elements in the bogs of central Ohio (Dachnowski, 1910), and the montane shrub Hedyosmum in lowland Amazon swamps of Madre de Dios, Peru (M.B. Bush, pers. obs.).

As data emerge pointing to the importance of these outlier populations, palaeoecologists are modifying their isopol-based view of migration to encompass the concept of micro- and macrorefugia (Birks & Willis, 2008; Rull, 2009). A parallel between micro- and macrorefugia exists in island biogeography theory, i.e. islands versus mainlands, and between isolated subpopulations and mainland populations in metapopulation theory. Here, we suggest that by explicitly incorporating time and stabilizing microclimatic factors with metapopulation dynamics we can enrich the connection between palaeoecological and conservation biology.

Metapopulations and microrefugia

Metapopulation models that include a mainland–island element assume that connectivity exists between the mainland and islands, and that gene flow between islands may also be important (Hanski, 2001). Large islands, with large populations, high fitness, and strong genetic connectivity to the mainland, are expected to be the most stable metapopulations, whereas those metapopulations with the opposite conditions are most likely to go extinct (Fig. 2). Because the size of a metapopulation is often set by physical factors, for example hilltop availability for checkerspot butterflies (e.g. Harrison, 1989), size is often viewed as being relatively fixed in the short term (i.e. on the decadal scale of most conservation planning). By contrast, a longer-term view (e.g. millennial scale) might well see substantial change in available niche space.

Figure 2.

 Contrasting population responses of a hypothetical montane species occupying macrorefugia (b–e) or microrefugia (f–i) to a glacial cycle. In a macrorefugial scenario, an interglacial population (a) of the species is in equilibrium with its environment. With cooler, glacial conditions, the species’ high-elevation populations go extinct. During full glacial conditions, only macrorefugial populations persist (c). With warming into an interglacial period, opportunities for recolonization are limited to expansion and long-distance dispersal from macrorefugia (d, e). In a microrefugial scenario, the high-elevation populations may contract into microrefugia (f), while populations in the low-elevation portion of the distribution expand into macrorefugia. The species may inhabit both micro- and macrorefugia during full glacial conditions (g). Subsequent recolonization in a warming climate occurs through the expansion of high-elevation microrefugial populations as well as from low-elevation populations (h, i). The net result is greater opportunities for genetic heterogeneity of subpopulations.

A parallel to the partial isolation of metapopulations exists in the genetic isolation of microrefugial populations. The areal extent of a microrefugium is, by definition, small, and (population) fitness is low. Taking these factors together, the prospects for population survival in microrefugia look bleak, but survive they do. We see many metapopulations with similar attributes failing on a decadal scale, but microrefugia persisting for > 10,000 years (Dachnowski, 1910).

The apparent key to the success of microrefugia is that their environment and microclimate are both very stable and also very different from those of the surrounding habitat, which we refer to as the matrix. Microclimatic stability may be provided by a reliable cold-water spring, or by a hot, dry habitat created by aspect and local rain shadow. In either case, the resulting microclimate is relatively invariable, and sufficiently distinct that matrix species are poor competitors for its resources.

Now consider how broad climate change influences microrefugial populations. During times of adversity, the population of a particular species of interest becomes isolated, having the attributes of a very threatened subpopulation but, as described, existing in a setting of unusual environmental stability (Fig. 2). When more benign conditions for that species occur, such as during shifts to interglacial periods, the geographic extent, population size and genetic connectivity increase, to the point where a previously isolated population now has the attributes of a large, ‘safe’ metapopulation, or even a mainland. As conditions deteriorate, for example through glacial periods, these trends are reversed until the microrefugium is re-established.

Thus a microrefugial population will, through time, pass through cycles of incorporation into, or exclusion from, the main population (Fig. 3). However, even though multiple populations expand to form an enlarged, potentially well-connected metapopulation during favourable times, the genetic identity of the microrefugial subsets may be retained.

Figure 3.

 Diagram introducing a millennial-scale dimension into metapopulation dynamics, and therefore influencing the probability of population extinction. Over time, small, isolated local populations (triangles) may go extinct owing to environmental, demographic and genetic stochasticity, while large, connected populations (circles) have lower chances of extinction. While experiencing changes at the regional scale that can be beneficial or adverse, microrefugial populations (crosses) oscillate between periods of higher and lower extinction probability. In microrefugia, populations face less environmental stochasticity, improving the probability of survival. Further stability is achieved if deleterious genes have been purged from the population.

The genetic importance of microrefugia

If macrorefugia were the means of survival during climatic adversity, relatively large populations would preserve heterozygosity. Genotypes adapted to the macrorefugial range of conditions (hot, cold, wet, dry, etc.) would persist, but adaptations gained during range expansion would be lost, especially among plants (Bennett et al., 1991). Although point mutations could add diversity as new conditions were encountered by an expanding population, such changes are thought to be rare (10−8 per generation) relative to the importance of founder effects that reduce diversity (Drake et al., 1998). Either through chance or adaptation, the spearhead of advance will be genetically impoverished compared with the main population forming the macrorefugium; that is, the expanding populations exhibit founder effects. Among plants, when the climate turns and these advance populations die back, the new mutations in the zone of population expansion are lost. For nearly all observed forest tree taxa, there is no migration back towards the macrorefugia, just a die-off where trees stand (Bennett et al., 1991). Consequently, the macrorefugial genotype is the conserved genotype, and forms the gene pool available for expansion with the next favourable climate cycle.

The microrefugial model is quite different in terms of genetic outcomes of climate change. Populations are scattered across the landscape and may be genetically connected via migration with one another, or with a macrorefugial population. In the most extreme cases, isolation is virtually complete. The populations are a remnant of the last population expansion and potentially represent different geographic, ecological, and genetically modified metapopulations. If the entire leading edge of migration is formed by a single haplotype, there is very little diversity. When the population contraction occurs as a result of climate change, rather than all the genetic differentiation being lost, any newly acquired genotypes could persist in the microrefugia.

When populations expand, the microrefugia form identifiable nuclei for expansion, to the extent that they are genetically differentiated. Importantly, given realistic migration rates, these microrefugia allow the species to colonize a much larger area, much faster, than if the expansion were limited to diffusional migration along a single expanding front.

Genotypic dispersal models point to the importance of rare but unusually long dispersal events. Clark et al.’s (2003) modification of fat-tailed distribution kernel models to include variable survival and reproductive rates are of direct relevance to understanding the dynamics of metapopulations and microrefugia. Clark et al. (2003) assume that mortality rates are higher and reproductive rates lower for individuals at the edges of a distribution than for those at the centre. Consequently, fitness is likely to be low in microrefugia when the matrix conditions are most hostile. However, when those constraints are released by climate change an exponential population increase is expected (Skellam, 1951; Clark et al., 2003). A further issue raised by Clark et al. (2003) is that accounting for the time to sexual maturity and peak reproductive output significantly slows the rate of spread of a population compared with assuming an instantaneous capacity to reproduce. In microrefugia, important pockets of sexually mature individuals can rapidly take advantage of a change in conditions.

A significant disadvantage of species without microrefugial populations is that, if a migrational barrier, such as a river, mountain chain or belt of developed land, lies between the macrorefugium and potential habitat, migration might falter or be slowed. If, on the other hand, the barrier was successfully passed previously (even just once), and a microrefugium is available in the colonized territory, then a population could persist on both sides of the barrier. In the classic macrorefugial model, subsequent expansions would be constrained until a migrant crossed the barrier, whereas in the microrefugial model the barrier ceases to be an expansion constraint.

Species traits favouring survival in microrefugia

Bhagwat & Willis (2008) identified traits apparently associated with species of inferred northerly versus southerly ice age distributions in Europe. Trees of northern distributions tended to be small-seeded and wind-dispersed, whereas southern species tended to be large-seeded and animal-dispersed. One of the largest-seeded species that Bhagwat and Willis considered, Fagus sylvatica, has a North American counterpart, Fagus grandifolia, which apparently occupied both macro- and microrefugia. Indeed, it appears that seed size is a poor predictor of colonization ability and, therefore, also an unreliable indicator of species survival in microrefugia. Fastie (1995), for example, documented Alnus, Picea sitchensis and Populus spreading after several centuries of recovery in the classic succession at Glacier Bay, Alaska, but almost no spread of Tsuga heterophylla. Environmental conditions such as soil type and texture were excluded by Fastie as the cause of stand composition (Fastie, 1995). Seed weights of these species or close congeners are: Alnus rubrac. 0.00068 g, Picea sitchensis c. 0.0022 g, Populus deltoidesc. 0.001 g and Tsuga heterophyllac. 0.0015 g (Hewitt, 1998). The point is that even a relatively small-seeded tree such as Tsuga can apparently be dispersal-limited in particular circumstances (Fastie, 1995).

If physical traits are ambiguous in determining which species will migrate quickly or survive in microrefugia, other aspects of reproduction may be important. In plants, clonal growth and the capacity for self-pollination are probably important traits (Bhagwat & Willis, 2008) that would allow long-term survival of very small populations. For animals, optimal strategies may be parthenogenesis and dichogamy (i.e. sequential hermaphrodites), both of which are common in invertebrates, lizards and fishes, or simply small body size (hence providing the potential for a large population in a small area). By inference, megafauna are far less likely to have microrefugia than, for example, beetles.

An alternative way to think of species that can survive in microrefugia is to consider their genetic requirements. A small population, perhaps completely isolated for thousands of years, either has to survive via asexual reproduction or to inbreed without suffering deleterious consequences. Inbreeding increases genetic load, lowering fitness or rendering the individual susceptible to strong environmental changes (Ralls et al., 1979). However, permanently rare species (i.e. endemic predators inhabiting small oceanic islands) are very likely to have purged deleterious recessives from their population (Barrett & Charlesworth, 1991), otherwise they could not have survived. Examples include elevated mortality during a winter storm among inbred song sparrows on Mandarte Island, Canada, relative to non-inbred lineages (Keller et al., 1994), and findings that inbreeding depression within outlier populations decreases with distance from ‘macrorefugial’ populations (Pujol et al., 2009). However, environmental stability reduces the problems associated with inbreeding (Hedrick & Kalinowski, 2000), and the combination of purging deleterious alleles and environmental stability may be enough to allow small populations of some species to persist in genetic isolation for long periods.

A naturally small population has a light genetic load. In contrast, a small population that was formerly part of a larger aggregation is probably at greater risk of extinction. These latter populations have a higher genetic load, which during bottleneck events reduces fitness. Emerging from the bottleneck, a species may have a reduced genetic load as recessive deleterious alleles have been purged (Byers & Waller, 1999). It is therefore theoretically possible to purge those alleles from the population, but purging recessive alleles takes time. A 10-generational study of Peromyscus mice revealed that three subspecies with different histories in the wild responded positively, neutrally and negatively, with respect to their genetic loads, to laboratory-controlled inbreeding (Lacy & Ballou, 1998). The important insight from this study is that the population history may determine the ability of a population to inbreed without negative consequences. The genotype forming the population when it collapses plays a very significant part in determining the subsequent potential for survival in isolation (Crnokrak et al., 2002). A direct prediction of this hypothesis is again that species that are usually rare, in which the genetic load is light, are more likely to survive in microrefugia than common species with heavy genetic loads.

An opportunity for diffusional spread of alleles exists after each deglaciation, resulting in a cline in haplotype diversity. If the haplotype at the leading edge of that diffusional gradient has purged some or all of its recessive alleles, it may be able to survive as a small population when climatic conditions cause population contraction, and if microrefugial settings are available. Indeed, it can be argued that no gene flow with the main population would be beneficial, lest purged deleterious genes be reintroduced. The subsequent expansion from the refugia will permit a more rapid occupation of the landscape than that via diffusion, but will still provide the same cline in haplotype diversity.

Successive retreats into microrefugia are possible, providing the conditions for survival do not change. Over time, the microrefugial populations would acquire neutral or nearly neutral mutations, which would serve to identify them as genetically distinct from the source haplotypes, as has been found in North American trees and insects (McLachlan & Clark, 2005; Walker et al., 2009) (Fig. 4). In this way, species that are abundant and have a high genetic load in their main population may produce lineages of low load that can survive in microrefugia.

Figure 4.

 Genetic consequences for populations that can exist (a) in microrefugia and macrorefugia versus (b) in macrorefugia alone in response to a warming event. (a) Northern haplotypes represent a ‘melting pot’ (as in Petit et al., 2003), with an unequal genetic contribution from each macrorefugium. Microrefugial haplotypes are primarily from one macrorefugium, as these individuals may more successfully reproduce asexually or may be more cold-adapted. Through the course of climate oscillations, if these haplotypes are purged of deleterious alleles, and small, inbreeding populations survive within microrefugia, then the microrefugial populations will be capable of rapidly expanding when favourable climatic conditions return. Through several of these climate oscillations, neutral genetic variations may accumulate to differentiate the genetics of each microrefugium. However, if inbreeding in small microrefugial populations reveals a high genetic load of deleterious alleles (b), only large macrorefugial populations survive and remain as haplotype sources. When climate conditions warm again, a different set of haplotypes may serve as the dominant genetic source.

Synthesizing these observations, it seems that not all species will be capable of surviving in microrefugia, either because of habitat needs, in the case of large predators, or because of their genetic load. Consequently, microrefugia are liable to offer safe haven to only a fraction of the original biota of a region. Overall diversity in the microrefugium may not be lower than that in the matrix, but the biota will probably represent a blend of microrefugial and matrix species. Indeed, it is likely that the higher trophic levels will be composed of species quite different from those of non-refugial times. The pressures that result from this imbalance could accentuate genetic selection away from the macrorefugial type. In doing so, the microrefugial species may experience less competitive and predatory pressures, increasing the probability of persistence within the microrefugia.

Microrefugia and conservation

Grappling with the synergistic consequences of habitat loss and climate change is the most pressing task in modern conservation. The species-based conservation policies that served us well for so long, because climates were assumed stable, will require radical revision. One change of perception is the recognition of the importance of microclimatic diversity within a landscape (e.g. Pearson & Dawson, 2003; Willis & Bhagwat, 2009), and the capacity of locations with locally atypical climates to be diversity reservoirs.

If microrefugia are not factored in, the modelling of population responses to future climate change will underestimate the capacity for range expansion and overestimate the likelihood of extinction (Luoto & Heikkinen, 2008; Randin et al., 2009). Modern distributions include microrefugia, some from the glacial period, which will probably not play much of a role in expansion as a response to projected warming. Other microrefugial populations may persist from the mid-Holocene climatic optimum, or even the mediaeval warm period, and these populations may already be expanding as a result of late 20th century warming, facilitating the general poleward and upslope movement of species already documented. In the majority of cases, no data will be available detailing whether these frontier populations have already begun to expand. Therefore, rather than the focus being entirely on species or populations, microrefugial areas should also be prioritized for conservation, because they will enable future expansion. The microrefugia at the other thermal extreme are no less important biologically, as they may see species turnover but continue to offer a microclimate at odds with the surrounding area. As ranges of species change, these cool microrefugia may warm, but their temperature differential relative to the surrounding area should be maintained. The microrefugial habitat will be unusual and may become a source of persistence for species that currently form the matrix around those sites, but are destined for regional extinction.

Most outputs predicting species responses to future climate change are derived from general circulation models (e.g. Cox et al., 2004; IPCC, 2007), which cover the land with a coarse grid (often 0.5–2° latitude). Downscaling these models is providing continual improvement in model quality, but critically important factors at the scale of microrefugia are missing (e.g. Pearson & Dawson, 2003; Luoto & Heikkinen, 2008; Thuiller et al., 2008; Randin et al., 2009). Complex topography, spring seeps, slope, aspect, even local cloud cover, are not realistically captured in these models, but it is this scale of landscape feature that usually provides the microclimate of the microrefugium. Consequently, microrefugia are not captured in general circulation models, and all the populations that could persist in those locations for the hundreds or thousands of years needed until our atmospheric carbon chemistry is restored are predicted to go extinct in the model output. Just as excluding microrefugia from climate models will exaggerate extinction rates, excluding them from conservation planning may exacerbate regional extinction rates.

Implicit in our conceptualization of the individualistic nature of species’ responses to ongoing climate change within diverse geographic microclimate landscapes is the re-mixing of species and the formation of non-analogue communities (e.g. Williams & Jackson, 2007). In the tropical Andes, such shifts in community composition are evident as climates cooled and warmed following the last glacial period (Valencia et al., 2010). Fossil pollen observations made at mid- and high elevations in the southern Peruvian Andes indicate that, despite disappearing from upslope settings as conditions cooled, Urticaceae/Moraceae pollen types did not become abundant further downslope. This important group of forest plants did not adhere to a diffusional pattern of migration. Nevertheless, Urticaceae/Moraceae rapidly recolonized with deglacial warming, suggesting that microrefugial survival facilitated population expansion (Valencia et al., 2010).

Conclusions

Past microrefugia have been important in shaping modern populations and are likely to be important for future responses to climate change. The detailed distributions of most species are relatively poorly known (the lower the latitude, the greater the uncertainty), but quantifying populations forming the discontinuous fringes of a species’ range is necessary to predict likely responses to climate change. Maintaining the conditions under which a population exists may be more important than maintaining the population itself because, while all communities are transient, the habitat that provides the anomalous conditions may be permanent. Some of the best examples of microrefugia are found near cooling spring waters, in which case maintaining flow from the aquifer must become the target of conservation efforts. With unprecedented rates of climate change, even established microrefugia are liable to see species turnover, but the sites will remain microrefugia for a different set of species. Thus, identifying and maintaining the integrity of microrefugial habitats, within a landscape, may lead to the provision of continual refugia for consistently high landscape-level biodiversity, despite changing conditions in the matrix.

During the peak of migration in response to the Pleistocene–Holocene transition, individualistic migration rates led to the formation of non-analogue communities (Jackson & Overpeck, 2000). Contemporary observations suggest that northern temperate tree species actually migrate 2–5 times slower than would be needed for species to have expanded from glacial macrorefugia and achieved their observed Holocene ranges. A reconciliation of this discrepancy is possible if microrefugia contributed to Holocene expansions. Multiple, scattered microrefugia both facilitate rapid, widespread population expansion and reduce the probability of extinction.

Because microrefugia may provide local relicts of a formerly prevalent regional climate, as those conditions reappear, expansion of populations from microrefugia allows some populations to respond rapidly to climate change. As we move into climates beyond the envelope of previous conditions, there will be no microrefugia to provide nuclei for expansion. Consequently, past migration rates may not be a good predictor for future migration rates. If migration rates lag rates of climate change more seriously than in the past, the imbalance between climate and the ecological optima of a given species will add uncertainty to conservation efforts.

Extinction rates will also be overestimated if microrefugia are not considered. Models that portray range changes in bioclimatic envelopes rarely include detailed topographic or fire-shelter effects, circumstances that can radically change the probability of a species existing in a microrefugium. While the model may be right in principle, showing a major range change, it may be wrong in assigning extinction to all the apparently ‘uninhabitable’ areas.

Not all species can survive in microrefugia, because of individual (body) size, resource requirements, and range constraints, particularly in the case of large predators. Another restriction relates to genetic constraints for species with high genetic loads that have in the recent past experienced large populations. Modern microrefugia are important targets for conservation – not for the species they currently hold, but for the attributes that make them so different from the matrix. These areas will always hold a different pool of species from the adjacent land, and protecting the physical features (e.g. the aquifer in the case of a spring) that provide this heterogeneity should be a conservation priority.

Acknowledgements

This work was funded by a grant from the Gordon and Betty Moore Foundation, NSF award BCS-0926973 to M.B.B. and a graduate research fellowship to N.A.S.M. This paper forms publication No. 6 of the Institute for Research on Global Climate Change, Florida Institute of Technology.

Biosketches

Nicole Sublette Mosblech is a doctoral student at the Florida Institute of Technology. She is working on the effect of climate change on Andean plant migrations via palaeoecological analyses.

Mark Bush is a professor of biology at the Florida Institute of Technology. His research focuses on fossil pollen analysis of Neotropical settings, environmental reconstructions of past climates and vegetation communities, and palaeoecological evidence of human responses to climate change.

Robert van Woesik is a professor of biology at Florida Institute of Technology. His research focuses on understanding coral-reef processes associated with climate change, including ecological questions relating to historical and contemporary thermal stresses, coral bleaching and subsequent population trajectories through time.

Editor: John Lambshead

Ancillary