Fossil evidence and phylogeography of temperate species: ‘glacial refugia’ and post-glacial recolonization

Authors

  • Robert S. Sommer,

    1. 1 Ecology Centre, Christian-Albrechts-University of Kiel, Olshausenstrasse 40, 24118 Kiel, Germany and 2Zoological Institute, Christian-Albrechts-University of Kiel, Olshausenstrasse 40, 24118 Kiel, Germany
    Search for more papers by this author
  • and 1 Frank E. Zachos 2

    1. 1 Ecology Centre, Christian-Albrechts-University of Kiel, Olshausenstrasse 40, 24118 Kiel, Germany and 2Zoological Institute, Christian-Albrechts-University of Kiel, Olshausenstrasse 40, 24118 Kiel, Germany
    Search for more papers by this author

*Robert S. Sommer, Ecology Centre, Christian-Albrechts-University of Kiel, Olshausenstrasse 40, 24118 Kiel, Germany. E-mail: rsommer@ecology.uni-kiel.de

Abstract

We present a short synthesis of the Pleistocene distribution dynamics and phylogeographic recolonization hypotheses for two temperate European mammal species, the red deer (Cervus elaphus) and the roe deer (Capreolus capreolus), for which high-resolution patterns of fossil evidence and genetic data sets are available. Such data are critical to an understanding of the role of hypothesized glacial refugia. Both species show a similar pattern: a relatively wide distribution in the southern part of Central Europe 60,000–25,000 years ago, and a strong restriction to areas in southern Europe for nearly 10,000 years during the Last Glacial Maximum (LGM) and the early Late Glacial (25,000–14,700 years ago). With the beginning of Greenland Interstadial 1 (Bølling/Allerød warming, c. 14,700–11,600 years ago) a sudden range expansion into Central Europe is visible, but the colonization of most of Central Europe, including the northern European Lowlands, only began in the early Holocene. In a European context, regions where the species were distributed during the LGM and early Late Glacial are most relevant as potential origins of recolonization processes, because during these c. 10,000 years distribution ranges were smaller than at any other time in the Late Quaternary. As far as the present distribution of temperate species and their genetic lineages is concerned, so-called ‘cryptic refugia’ are important only if the species are actually confirmed there during the LGM, as otherwise they could not possibly have contributed to the recolonization that eventually resulted in the present distribution ranges.

Introduction

The model of glacial refugia as core areas for the survival of thermophilous and/or temperate animal and plant species during unfavourable Pleistocene environmental conditions and as the sources of post-glacial recolonization processes is widely accepted in biogeography (Hewitt, 2000; Willis & Whittaker, 2000). Over the last decade there has been an increasing interest in the importance of glacial refugia in maintaining diversity under changing climates (e.g. Taberlet et al., 1998; Hewitt, 2000; Willis & Whittaker, 2000; Stewart & Lister, 2001; Widmer & Lexer, 2001; Taberlet & Cheddadi, 2002; Tzedakis et al., 2002, 2003; Willis & van Andel, 2004; Provan & Bennett, 2008; Stewart & Cooper, 2008; Stewart & Darlén, 2008). Genetic data may help in the identification of glacial refugia of animal and plant species (e.g. Hewitt, 1996; Taberlet et al., 1998; Schmitt & Krauss, 2004; Brito, 2005; Deffontaine et al., 2005; Schönswetter et al., 2005; Kotlík et al., 2006; Loehr et al., 2006) but cannot resolve their precise location. Fossil remains of plants (e.g. Bennett et al., 1991; Willis & van Andel, 2004) and animals (e.g. Sommer & Nadachowski, 2006; Sommer et al., 2008) are indispensable in identifying the spatio-temporal pattern of these refugia.

The term ‘glacial refugia’ has often been used with very different meanings, leading to recent attempts to clarify the terminology (McGlone & Clark, 2005; Bennett & Provan, 2008; Holderegger & Thiel-Egenter, 2009; Rull, 2009). To optimize understanding of the function and significance of glacial refugia it is important to obtain a deeper knowledge of climatic influences on the range dynamics of species during the Late Quaternary. We present a short synthesis of the Pleistocene distribution dynamics and phylogeographic recolonization hypotheses for two temperate European mammal species – the red deer, Cervus elaphus Linnaeus, 1758, and the roe deer, Capreolus capreolus (Linnaeus, 1758) – for which comprehensive fossil and genetic data sets are available. These two species are suitable models to illustrate how Late and post-glacial aged refugia may have determined the pattern of post-glacial recolonization. Our conclusions refer primarily to temperate zones and to Europe in particular.

Important climatic changes in the Late Quaternary

When reviewing the cold epochs of the Late Quaternary, two major cold events are evident: the first maximum glacial cooling during marine isotope stage 4 (MIS-4), c. 74,000–60,000 years ago, and the second maximum glacial cooling, the Last Glacial Maximum (LGM), c. 25,000–18,000 years ago, which occurred within marine isotope stage 2 (MIS-2) of the Weichselian Glacial (for a review see Van Andel, 2003; see Appendix S1 in Supporting Information).

The first glacial maximum was much milder than the second (LGM), as can be seen in the extension of the northern ice shield: during the first glacial maximum, Great Britain was not glaciated, and the Scandinavian ice shield did not reach Central Europe (Van Andel, 2003).

Marine isotope stage 3 (MIS-3, c. 60,000–25,000 years ago) includes various important climate warmings (Greenland Interstadials, GI), such as the well-known interstadials ‘Hengelo’ (GI 12, c. 48,000–44,000 years ago) and ‘Denekamp’ (GI 8, c. 39,000–36,000 years ago), which led to fundamental environmental changes for 4000–5000 years in each case (other interstadials lasted only for about 1000–1500 years, see Shackleton et al., 2004; Svensson et al., 2006; Andersen et al., 2006). These changes resulted in temperate grasslands and mixed forest mosaics in Central Europe.

Temperate mammal species reached northern Central Europe during the warmer Greenland Interstadials between the first and the second glacial maxima (Sommer & Benecke, 2005a; Sommer et al., 2008, 2009). The peak of the Devensian/Weichselian glaciation was the LGM ‘sensu stricto’, c. 24,000 years ago (Fig. 1), when, for a space of nearly 1000 years, the annual mean temperatures in Greenland were about 21°C lower than they are today (Cuffey et al., 1995) and the sea-level was 121 ± 5 m lower than it is at present (Lambeck et al., 2002). During the LGM ‘sensu lato’ (Fig. 1) the mean annual temperature in the northern regions was around −8°C, and the southern limit of sporadic permafrost extended across central France into southern Germany (Huijzer & Vandenberghe, 1998). The climate was extremely cold and dry, and most of Central Europe was covered by a steppe-tundra. The continuous deglaciation of the ice shield began c. 18,000 years ago (Fig. 1) when environmental conditions were still fully glacial (for a precise spatio-temporal reconstruction of temperatures and precipitation see Barron et al., 2003). During GI 1 (Fig. 1), a widespread warming epoch became established. Annual mean temperatures rose by more than 10°C within a few decades, and during the thermal maximum of GI 1 the mean temperatures of the warmest month reached 16–18°C in Britain. Early in the interstadial Central Europe was increasingly covered by birch, but late in GI 1 a mixed forest of pine and birch developed. An abrupt cooling occurred during Greenland Stadial 1, also known as the Younger Dryas, leading to an advance of the ice shield in Scandinavia and an apparent extinction of the birch/pine forests in the central European lowlands. Discontinuous permafrost extended across southern Britain, Belgium and northern Germany, suggesting mean annual temperatures of between −8°C and −4°C (Isarin, 1997). In the early Holocene, some 11,000 years ago, temperatures reached the present level (for temporal definition of the mentioned climatic epochs see Appendix S1).

Figure 1.

 Chronology of climatic events according to the GISP2 ice core (after Sommer et al., 2008, and Svensson et al., 2006). Ordinate: δ18O difference in ‰ of a specific standard value that is similar to the present state. Abscissa: time in calendar years before present (cal. bp), meaning before 1950. Upper numbers: Greenland Interstadials 4–1 (warmings). Greenland Interstadial 1 is subdivided into warmer epochs (e,c,a) and two short cold spells (d,b) (after Jöris & Weninger, 2000). Lower numbers: Greenland Stadials 1–5 (coolings). Maximum glaciation and culmination point of the Last Glacial Maximum (LGM sensu stricto) is at c. 24,000 cal. bp.

Phylogeography and post-glacial recolonization

Phylogeography, the study of the phylogeny of genetic lineages combined with their geographic distribution, has been an enormously active field of research over the past two decades (for reviews see Avise, 2000, 2009). Many of the studies carried out deal with the identification of refugial areas during the last glacial and the reconstruction of the post-glacial recolonization routes. For Europe, it is commonly assumed that temperate species retreated to lower latitudes as a reaction to the shift of climatic zones during Quaternary cold pulses. Consequently, the three south European peninsulas of Iberia, Italy and the Balkans are believed to have played a major role as glacial refugia (Hewitt, 2000). In the wake of climatic amelioration within interstadials, temperate species expanded their distribution range to the north, following the expansion of favourable habitats (‘habitat tracking’, Provan & Bennett, 2008).

In many cases, only a fraction of the genetic diversity present in the refugia will also be found in recolonized areas, leading to a relative genetic depletion in northern and Central Europe (the ‘southern richness versus northern purity’ paradigm, Hewitt, 2000). If these northern areas have been recolonized from distinct refugia, however, the admixture of different genetic lineages will lead to high levels of genetic diversity, which sometimes makes it difficult to decide whether an area constitutes an admixture zone or a former northern refugium (Provan & Bennett, 2008). In such cases, the occurrence of private alleles may be interpreted as evidence of a refugium. In addition, refugia are more likely to show low levels of nucleotide diversity compared with admixture zones, because they usually represent a single genetic lineage made up of many different, but relatively closely related, alleles. Admixture zones, on the other hand, by definition harbour alleles from different lineages, which results in high nucleotide diversities (as this parameter of genetic diversity takes into account the amount of pairwise differences among alleles) (Provan & Bennett, 2008). Alternatively, narrow hybrid zones (not large-scale admixture) acting as barriers to the further expansion of refugial genomes may form. Matters are further complicated by the different modes of recolonization: slower expansion involving shorter dispersal and larger effective population sizes (the ‘Phalanx’ mode that retains more genetic diversity) or fast expansion of fewer individuals via long-distance dispersal, producing pockets of distinct genomes and less diversity (‘Pioneer’ mode, see Hewitt, 2004, and references therein).

In any case, the use of genetic data alone might be misleading, and a multi-disciplinary approach including palaeontological and palaeoclimatological data seems likely to prove more fruitful. Notably, the alleles of the most frequently used molecular marker, mitochondrial DNA, often coalesce deeper in the past than the onset of the last glaciation; that is, many alleles did not originate in the Pleistocene but had already been present in pre-Quaternary ancestral gene pools or at least date back to times earlier than the last ice age (Taberlet et al., 1998; Hewitt, 2000). Therefore, differences in the genetic composition among glacial refugia and between phylogeographic lineages may be a result of the sorting process of mutations that occurred in the more distant past (Taberlet et al., 1998; Hewitt, 2000).

Comparative analyses have shown that the recolonization routes from the peninsular refugia vary greatly among species. Even very closely related taxa may show conspicuous differences: whereas the yellow-necked fieldmouse (Apodemus flavicollis) recolonized much of Europe from the Italo-Balkan region, its congener, the woodmouse (Apodemus sylvaticus), expanded northwards from Iberia (Michaux et al., 2005). Generally speaking, three main patterns of recolonization were suggested originally (Hewitt, 1999): the grasshopper pattern (after Chorthippus parallelus) refers to a post-glacial expansion from the Balkans, the bear pattern (after Ursus arctos) implies recolonization from Iberia and an eastern refuge (different from the Balkans), whereas in the hedgehog pattern (after Erinaceus europaeus/roumanicus/concolor) all three southern peninsulas (Iberia, Italy and the Balkans) contributed to the present populations in northern and Central Europe. A fourth major paradigm was subsequently formulated for freshwater fish species (recolonization from the Black Sea via rivers such as the Danube or the Dneiper; Hewitt, 2004), and yet another paradigm, the expansion of Italian and Balkan lineages, has more recently been proposed on the basis of data derived from some butterfly species (Habel et al., 2005). A number of suture zones, clusters of hybrid zones where different genetic lineages meet in various species, have been identified in Europe (see Hewitt, 1999, in particular Figure 5): the Alps, the Pyrenees, the region between the western coast of the Baltic Sea and the Alps (where lineages from western and eastern refugia are often found) and, interestingly, Scandinavia, where genetically distinct lineages have been localized, for example in brown bears and moor frogs (Knopp & Merilä, 2009). This latter pattern suggests that many species recolonized Fennoscandia via both western and eastern recolonization routes.

It is important to note that most phylogeographic analyses rely exclusively on data derived from extant populations. These data are valuable in the inference of post-glacial recolonization patterns and the identification of the large-scale refugial region (Iberia, Italy, Balkans etc.) during the LGM, but they will usually not contain much information about distribution ranges deeper in the past. Because the LGM was the coldest epoch since the Wolstonian/Saalian glaciations of the Middle Pleistocene, the distribution ranges of temperate species were most confined during the LGM. Consequently, distribution ranges during earlier cold pulses probably did not contribute significantly to the present genetic pattern of these species. Because range contractions often resulted in lineage sorting, only a fraction of the former genetic diversity was represented in LGM refugia and available as a basis for recolonization. This scenario is exemplified by brown bear mitochondrial DNA data. Today, the western and the eastern bear lineages are clearly separated geographically (except for Romania; Zachos et al., 2008), with even the Balkan populations belonging to the western clade. However, ancient DNA analyses have revealed that the eastern lineage occurred as far west as Iberia before the LGM (Valdiosera et al., 2008).

Late Quaternary dynamics of temperate species using high-resolution fossil evidence and genetic data

For Central Europe there is a large body of radiocarbon-dated fossil evidence of temperate species available from the Devensian/Weichselian Glacial periods (e.g. Brunnacker et al., 1977;Feustel, 1980; Albrecht et al., 1983; Delpech, 1983; Woodman et al., 1997; Aldhouse-Green & Pettitt, 1998; Münzel et al., 2001; Baales et al., 2002; Pacher, 2003; Sommer & Benecke, 2005a,b; Benecke et al., 2006). Recently, there have been attempts to demonstrate the effects of climatic changes on the range dynamics of mammal species during the Last Glacial. These reconstructions clearly show that temperate species were present in Central Europe during warmer periods of the Last Glacial (Baales et al., 2002; Sommer & Benecke, 2005a,b; Sommer et al., 2008, 2009).

To depict the typical biogeographical history of temperate species we will give a short overview of the distribution dynamics of red and roe deer in the following section. These two ungulates are particularly well suited as model species because they were common prey animals of Neanderthals and anatomically modern humans during the Pleistocene, thus leaving a comprehensive fossil record. The high-resolution pattern of fossil evidence of both species (Sommer et al., 2008, 2009) shows a distribution range comprising remarkably large parts of southern Central Europe during marine isotope stage 3 (60,000–25,000 years ago). The red deer occurred in France as well as in southern England and Ireland and also in southern Germany and the Carpathian region (Fig. 2a). Between 25,000 and 18,000 years ago a change in the fossil pattern of both species is evident as records become generally restricted to areas in southern Europe. The fossil records tell us that the red deer and roe deer not only occurred in the ‘classical’ refuge areas of Iberia, Italy and the Balkans, but also survived in areas of southern France (Dordogne and Rhone Valley) and around the Carpathians. During the early Late Glacial (18,000–14,700 years ago), that is, during the initial deglaciation, no fluctuation is evident in the distribution range of either species. This situation completely changed with the beginning of GI 1 (Fig. 1). During that period, a sudden range expansion into northern regions has been confirmed for both species, most remarkably so in the red deer, which was already present in southern England almost 300 14C years before the start of the GI 1. Fossils of red deer are known from northern France and Germany within the earliest phase (1e) of the GI 1 (Fig. 1). In the course of the GI 1 a nearly complete recolonization of western Central Europe, including parts of Ireland and England, can be inferred for the red deer (Fig. 2g). The roe deer shows a similar distribution pattern in GI 1, but it did not spread as far north as the red deer (Fig. 2h). During the following (last) cold stage, the Greenland Stadial 1 (Younger Dryas), both species again lost some parts of their northern range, but the space of only 1000 years is too short to obtain a representative fossil pattern (not displayed in Fig. 2; for details see Sommer et al., 2008, 2009). The recolonization process of Central Europe was completed during the early Holocene, when the fossil pattern of both species included the northern European Lowlands (Fig. 2i,j). Genetic data show that Central and north European red deer largely originate from an Iberian immigration (Fig. 2e), but in the case of the roe deer the standard refugia do not seem to have played a major role in the recolonization of Europe (Fig. 2f). Rather, one or several refugia further east seem to explain the genetic pattern of extant populations best. In line with this, it has been suggested that, as a result of a bias in available data, the importance of refugia in eastern Europe and/or Asia may have been underestimated (Taberlet et al., 1998; see also Sommer & Nadachowski, 2006, and Provan & Bennett, 2008, for discussion on the Carpathian region as a glacial refugium, and Hewitt, 1996, 1999, for reviews including eastern refugia).

Figure 2.

Figure 2.

 Distribution dynamics of (left column: a,c,g,i) red deer (Cervus elaphus) and (right column: b,d,h,j) roe deer (Capreolus capreolus) between 60,000 and 9000 years ago (after high-resolution fossil evidence, see Sommer et al., 2008, 2009) and inferred recolonization routes of (e) red deer and (f) roe deer after the Last Glacial Maximum. The lines denote the northernmost distribution of fossil records during each climatic epoch. Dotted lines indicate that the northern range border is inferred on the basis of only a few records and the known environmental conditions to the south of the line. The dotted black lines in (e) and (f) indicate uncertain recolonization routes. Time periods: (a,b) 60,000–25,000 years ago; (c,d) 25,000–14,700 years ago; (g,h) 14,700–12,650 years ago; (i,j) 11,600–9,000 years ago. (e,f) Recolonization routes on the basis of mitochondrial DNA and, for roe deer, also of nuclear microsatellite loci; glacial refugia inferred from the fossil record.

Figure 2.

Figure 2.

 Distribution dynamics of (left column: a,c,g,i) red deer (Cervus elaphus) and (right column: b,d,h,j) roe deer (Capreolus capreolus) between 60,000 and 9000 years ago (after high-resolution fossil evidence, see Sommer et al., 2008, 2009) and inferred recolonization routes of (e) red deer and (f) roe deer after the Last Glacial Maximum. The lines denote the northernmost distribution of fossil records during each climatic epoch. Dotted lines indicate that the northern range border is inferred on the basis of only a few records and the known environmental conditions to the south of the line. The dotted black lines in (e) and (f) indicate uncertain recolonization routes. Time periods: (a,b) 60,000–25,000 years ago; (c,d) 25,000–14,700 years ago; (g,h) 14,700–12,650 years ago; (i,j) 11,600–9,000 years ago. (e,f) Recolonization routes on the basis of mitochondrial DNA and, for roe deer, also of nuclear microsatellite loci; glacial refugia inferred from the fossil record.

Conclusions

The high-resolution pattern of fossil evidence of two temperate large mammal species shows repeated shifts in their northernmost distribution range in Central Europe. Our red and roe deer data suggest that during the last 60,000 years of the Late Quaternary there have been four main biogeographical stages: (1) a relatively wide distribution in Central Europe during marine isotope stage 3 (during an epoch with numerous longer climate warmings); (2) a strong restriction of nearly 10,000 years to areas in southern Europe during the LGM sensu lato and the early Late Glacial; (3) a range expansion into Central European regions during GI 1; and (4) a distribution range throughout most of Central Europe, including the northern European lowlands, in the early Holocene. If our findings based on two mammal species with an exceptionally rich fossil record can be generalized, this has an important bearing on our understanding of glacial refugia and post-glacial recolonization histories of temperate species.

Only those regions where temperate species survived the LGM are relevant as the potential geographical origin of their extant distributions and diversity in Europe, because the distribution ranges of these species were more contracted than ever after this time.

The discovery of so-called ‘cryptic refugia’ is fascinating insofar as the biogeographical history of the biota during the whole Quaternary is of interest, but these refugia are in some cases irrelevant with regard to the post-glacial recolonization process. Hitherto postulated cryptic refugia of the red deer in northern Europe (Stewart & Lister, 2001) were only transient refugia in the sense that the species was found to have occurred there during the ice age. Red deer disappeared from these refugia during the LGM and recolonized Central and northern Europe from other refugia in the south. For this process, GI 1 (also known as Bølling/Allerød warming) was one of the most important climatic events because much of the recolonized area was conquered during this time.

Although the term ‘glacial refugium’ for the transient ‘cryptic’ refugia is, of course, linguistically correct (in the meaning of a refugium occurring some time during the glacial), its use can be misleading, because in particular geneticists interested in phylogeography and post-glacial recolonization tend to equate ‘glacial refugium’ with the regions where species lived during the LGM and, consequently, from which higher latitudes were recolonized during the Holocene. This is not to dismiss the role of ‘cryptic refugia’ altogether, because in ecoregions other than the temperate zone and for particular taxa it might not be temperature that is the limiting factor but, for example, precipitation. The existence and identification of cryptic refugia promise important insights into biogeography, but with respect to LGM-driven distribution ranges and post-glacial recolonization (as in the case of mammals) the term might be misleading if not further specified. At the very least, therefore, it should be made absolutely clear in each context which phase of the Quaternary is being referred to when the term glacial refugium is used (see also McLachlan & Clark, 2004; Holderegger & Thiel-Egenter, 2009).

Acknowledgements

The authors thank M. B. Bush, G. Hewitt and an anonymous referee for valuable comments on an earlier version of this manuscript, and R. J. Whittaker for corrections and comments that improved the paper.

Biosketches

Robert S. Sommer is interested in animal ecology and Quaternary ecology. He uses fossil and molecular evidence as well as climatic records to analyse the influence of climatic changes on the distribution and extinction dynamics of animal species.

Frank E. Zachos is an evolutionary zoologist with a particular interest in population genetics, phylogeography and conservation.

Editor: Mark Bush

Ancillary