It has been widely accepted that, in the temperate zones of the Northern Hemisphere, Pleistocene glaciations displaced forests to the south, where forest species remained in restricted refugia until post-glacial climatic amelioration allowed them to recolonize the continents (e.g. Huntley & Birks, 1983; Webb, 1987). However, estimations of the speed of this recolonizing process were incompatible with known dispersal mechanisms of the species involved (Reid’s Paradox), and demanded alternative mechanisms (Clark et al., 2003). Bennett (1988) suggested a different model for the glacial expansion of Fagus grandifolia in North America. From a pollen analysis, he concluded that post-glacial colonization did not take place as a ‘front’ from a southern refugium but ‘simultaneously at all sites’, starting from ‘population densities too small to be observed’ by pollen analysis. This author proposed that, during full-glacial conditions, F. grandiflora did not occur within localized refugia, but as a scarce widespread tree (Bennett, 1985). Thus, post-glacial forest re-establishment was not properly a recolonization, but rather a process of increasing the degree of occupancy of its geographic distribution. The refugia hypothesis was also adopted in the Southern Hemisphere, where the main forest refugia were in the north. In New Zealand, for example, during the Last Glacial Maximum (LGM), ‘continuous rainforest was restricted to the northernmost portion of the North Island, with forest decreasing rapidly south of latitude 38° S, only inhabiting microclimatically favoured locations’ (McGlone, 1985; see also Markgraf et al., 1995, p. 144). The term ‘microrefuge’ was introduced by Rull et al. (1988) to explain the persistence in time of high-mountain species and communities on the summits of table mountains (tepuis) from Venezuelan Guayana, in northern South America. Palaeoecological analyses of the older peats from some tepui summits revealed surprisingly rich and complex pollen assemblages, not to be expected from earlier successional stages, leading to the conclusion that they were derived from well-established plant communities. Two possible explanations were offered: the removal of older peats by groundwater erosion, or the existence of ‘small vegetation stands growing in certain favourable sites (microrefuges?) during Pleistocene dry phases’. Once the climate changed to a more humid one, ‘recolonization proceeded from these more or less complex stands’ (Rull et al., 1988). The biogeographical significance of these microrefugia for the Guayana region was discussed by Rull (2004).
Explicitly or not, the concept of microrefugia is widespread in the literature. For example, Leal (2001) used it to account for the continuity of tropical African rain forests under assumedly arid glacial conditions. According to this author, rain forest tree species are unlikely to have reached their present area of distribution from the currently postulated glacial refugia, and he proposed that ‘outside the postulated large areas of continuous forest, (macro)refugia, small patches of forest must have remained, acting as small scale refuge areas or microrefugia’ (p. 1077). This introduces the term ‘macrorefugia’ to designate the areas formerly called simply ‘refugia’. Other studies incorporated the concept but not the term. In Europe, Stewart & Lister (2001, p. 608) proposed that, during the late glacial, ‘the well-studied European southern and eastern refugia for thermophilous animal and plant taxa were supplemented by cryptic refugia in northern Europe during the Late Pleistocene. These northern refugia would have been in areas of sheltered topography that provided suitable stable microclimates’.
The diffuse approach of Bennett (1988) was adopted in part by Magri (2008), who proposed that the Pleistocene European Fagus refugia ‘were likely to have been a mosaic of sparse stands of small populations scattered in multiple regions’ (p. 450). A review of the potential location of microrefuge areas in Europe was attempted by Willis & van Andel (2004), based on the existing fossil and molecular evidence. In North America, McLachlan et al. (2005), using molecular phylogenetics, estimated that the northward late-glacial colonization by forest trees proceeded at rates significantly slower than previously thought and proposed that ‘populations of thermophilous trees may have been present in low abundance much farther north than the southern refuges depicted’ (p. 2089). McGlone & Clark (2005) used the terminology previously developed for the tropics and called these areas microrefugia, in contrast to macrorefugia, constituted as the main area of glacial forest refugia. This terminology was also adopted by Loehle (2007, p. 109), who proposed that ‘most temperate species could have survived across their current ranges at lower abundance by retreating to moist microsites. These would be microrefugia not easily detected by pollen records’. The potential existence of glacial microrefugia has also been proposed for animals, in order to explain their present speciation patterns (e.g. Joger et al., 2007). By contrast, some recent studies that model palaeoecological data from the Northern Hemisphere do not demand the existence of microrefugia and suggest that present-day forest composition can be explained by post-glacial recolonization from the main refugia (e.g. Montoya et al., 2007). These studies conclude that there is a manifest disequilibrium between climate change and forest expansion, which is notably delayed as a result of limitations in the dispersal of taxa, suggesting that glacial climates are still affecting diversity patterns (Svenning & Skov, 2007a, b).
The need for microrefugia is not necessarily linked to the glacial refugia hypothesis, which is widely accepted for temperate, but not for all tropical, regions. The refugia hypothesis has been severely questioned, especially for the Neotropics, where an alternative cold but not dry LGM scenario, with depleted CO2 atmospheric concentrations, has been proposed (Bush & De Oliveira, 2006). Based on palaeoecological evidence from the Amazon basin, the defenders of this ‘cold’ hypothesis propose that: (1) lowland rain forests were continuous both in space and in time, even during glacial periods, showing variations only in their taxonomic composition because of the downward migration of Andean taxa and (2) temperature-driven altitudinal displacements of sensitive taxa rather than large-scale contractions and expansions of entire biomes have been the rule during the glacial/interglacial alternation (Bush et al., 2001). In this context, microrefugia are needed to enable an understanding of the rapid post-glacial colonization of the Andean flanks by forests after glacier retreat. Here, the microrefugia approach is manifest in statements such as: ‘the combined effect of disjunct population nuclei surviving in isolated microhabitats and short migratory distances, could have resulted in rapid forest change with the onset of cooling or warming’ and ‘past range expansion (repeating with each cold or warm event) of species may have been accomplished by the expansion of relict isolates, allowing forest to respond rapidly to climatic fluctuation’ (Bush, 2002, p. 467). A consequence of the ‘cold’ hypothesis is the potential existence of both glacial and interglacial microrefugia. The Earth seems to have been glaciated 80% of the time during the Pleistocene. Indeed, a typical glaciation extends up to 90,000 years, whereas interglacials are comparatively short warmings of around 10,000 years. As a consequence, Pleistocene biota was adapted to ‘normal’ glacial conditions, and interglacial warm peaks can be viewed as ‘disturbances’, which sensitive taxa have to endure by upward migration to montane refugia (Bush et al., 2003; Bush, 2002). In this process, a number of species may become extinct, whereas others survive in microrefugia. The present interglacial, the Holocene, has provided numerous examples in the form of present-day relict populations. A classical example is the arctic/alpine shrub Dryas octopetala, which has populations that are highly fragmented as the result of a complex post-glacial history and specific microhabitat requirements (Skrede et al., 2006).
From this brief account, a general worldwide agreement about the potential occurrence of glacial and interglacial microrefugia, and their biogeographical significance for the establishment of present-day biotic patterns, is evident. Mountain regions seem especially well suited to hold microrefugia (Holderegger & Thiel-Egenter, in press). However, the number, location and extension of potential microrefugia remain elusive, as they are hardly detectable by palaeoecological analysis. So far, microrefugia are no more than a theoretical necessity for explaining Pleistocene recolonization patterns. Their precise characteristics, besides the also speculative but necessary occurrence of favourable microclimates, are unknown. The extensive agreement about the need for microrefugia is in contrast to the poor theoretical development of this concept, in particular with regard to which biogeographical and ecological features should be properly addressed. First, some terminological agreement is needed. So far, a number of names, for example ‘microrefuges’, ‘cryptic refugia’, ‘northern refugia’, ‘isolated microhabitats’, ‘relict isolates’, ‘sparse stands’, ‘humid microsites’, ‘intraglacial refugia’, have been used. It would be better to use a single term, and ‘microrefugia’ seems appropriate. It is especially important to differentiate the Latin terms ‘microrefugium’ (singular) and ‘microrefugia’ (plural), used in this paper, from ‘microrefuge(s)’, which is commonly employed to refer to, for example, microenvironments used as shelters by small animals during winter (e.g. Sinclair et al., 2003). One possible definition of microrefugium, which is proposed for discussion, is a small area with local favourable environmental features, in which small populations can survive outside their main distribution area (the macrorefugium), protected from the unfavourable regional environmental conditions. As defined in this way, the term is more general, and ‘glacial microrefugium’ and ‘interglacial microrefugium’ are two of the possible particular cases. According to the studies developed so far and mentioned above, three main types of microrefugia can be differentiated (Fig. 1): (1) distal or remote, occurring as isolates at great distances from the respective macrorefugia [as for example the nunataks (see Brochmann et al., 2003;Holderegger & Thiel-Egenter, in press)]; (2) diffuse or widespread, if isolates are more or less dispersed throughout inhospitable lands [as is the case for North American Fagus grandiflora (Bennett, 1985)]; and (3) proximal or ecotonal, when isolates are close to the macrorefugia, forming a gradual contact between it and the inhospitable lands [as proposed by Magri (2008) for the European Fagus sylvatica]. It is possible, however, that future studies will require the definition of more types or subtypes. For example, Holderegger & Thiel-Egenter (in press) describe three types of mountain refugia (nunataks, peripheral and lowland refugia), which they consider to fall within the microrefugia category.
Ecological characterization of microrefugia is not as yet possible, but some preliminary evidence suggests that large-seeded trees and thermophilous vertebrates survived preferably in the southern European Pleistocene macrorefugia, whereas species that occupied the northernmost microrefugia were mainly generalist, wind-dispersed plant species able to reproduce vegetatively, and generalist animals (Bhagwat & Willis, 2008). Modern ecological concepts and analyses of metapopulation dynamics may be of value in developing alternative models and in generating predictions, thus providing hypotheses to be tested with palaeoecological and molecular studies. The results of previous modelling surveys suggesting that modern diversity patterns can be achieved by migration from macrorefugia (Montoya et al., 2007; Svenning & Skov, 2007a,b) are not necessarily incompatible with the microrefugia hypothesis, and both models could be complementary. It would be interesting to run the same models considering the three types of microrefugia described above (Fig. 1), and then compare all the outputs. In addition, the detailed study of modern relict populations would help to improve our understanding of the dynamics of these isolates, in terms of Pleistocene patterns and processes.
The concept of microrefugia is also attractive in the area of biodiversity conservation in the context of ongoing climate change (Petit et al., 2008). For instance, Pearson (2006) considers that potential future microrefugia could help to mitigate the extent of the projected biotic extinction resulting from global warming, by providing suitable microhabitat conditions for threatened species. This would be especially beneficial for medium- and high-mountain endemic taxa, for which the risk of habitat loss by upward displacement is considerable (e.g. Rull & Vegas-Vilarrúbia, 2006). On the other hand, the protection of species endangered by habitat loss is commonly addressed through ex situ strategies, such as, for example, seed banks and botanical gardens, which in fact are man-made microrefugia.