Beyond skepticism: uncovering cryptic refugia using multiple lines of evidence

The increasingly popular opinion that temperate plant and animal species persisted in many small refugia in Europe north of the main southern peninsular refugia (Tzedakis et al., 2013) directly challenges a classical paradigm (Huntley & Birks, 1983; Bennett et al., 1991). As indirect evidence for the existence of such glacial microrefugia readily accumulates (reviewed in Tzedakis et al., 2013 for plants; Schmitt & Varga, 2012 for animals), it is interesting to note that the few direct in situ occurrences of fossils testifying their local existence are almost always debated (Kullman, 2002; Birks et al., 2005, 2012; Parducci et al., 2012). Our recent study on European beech (Fagus sylvatica) refugia in south-western France was no exception to this trend (de Lafontaine et al., 2014; Huntley, 2014, in this issue of New Phytologist, pp. 715–729 and 447–449, respectively). Yet, as mentioned by the philosopher Bertrand Russell (1959), it is not possible ‘to get anywhere if we start from skepticism. We must start from a broad acceptance of whatever seems to be knowledge and is not rejected for some specific reason.’ We therefore believe that no a priori preference should be given to one of the two scenarios of migration from a distant macro-refugium or of survival in situ in small refugia. Microrefugia are by definition spatially restricted and include few and possibly less fit individuals, so that a direct assessment for their existence at the stand-scale is extremely challenging (Gavin et al., 2014). In fact, without prior information on the possible location of these refugia, finding fossils testifying to their existence is akin to looking for a needle in a haystack. We believe that to uncover glacial micro-refugia, the fossil record needs not to be abundant, as long as the claims are reasonable and the dataset includes both indirect and robust direct field evidence for local presence during the most severe climatic conditions. The criticisms of Huntley (2014) extend beyond our own study to other recently published studies on so-called cryptic glacial refugia, that is putative microrefugia inferred indirectly from genetic analyses of extant populations (Cruzan & Templeton, 2000), species distribution modeling (SDM) or regional-scale pollen analyses (Tzedakis et al., 2013). As we will now show, considerable progress has been made recently in these fields that deserve to be more widely appreciated (Gavin et al., 2014). We infer beech survival in microrefugia located between 43°54′N and 44°24′N in south-western France, merely c. 100 km north of an established beech refugium in the Pyrenees (Magri et al., 2006). In the most recent review of the distribution of such refugia in Europe, Tzedakis et al. (2013) also suggest the possible survival of temperate trees close to this location (see their Fig. 1). European beech can be extremely hardy, surviving at higher altitudes than most other temperate broadleaved tree species. In particular, it is known to endure extreme conditions above the alpine timberline thanks to clonal reproduction through layering despite a near total lack of generative reproduction (Vacek & Hejcman, 2012). Crucial for such survival under extreme conditions is the microclimate as modulated by local topography. In this respect, the two locations where we found the oldest beech charcoal particles in south-western France are moist ravines along permanent rivers (Fig. 1), a habitat shown to have acted as a microrefugium for the species (Papageorgiou et al., 2014). Climatic niche models also support the view that beech persisted in south-western France (Svenning et al., 2008), although Huntley (2014) criticized the supposed propensity of the Maxent model to overestimate species distribution. In a recent study, Saltré et al. (2013) used another modeling approach relying on process-based SDM to assess whether beech postglacial colonization was in equilibrium with climate during the last 12 kyr. According to their model, beech colonization in western Europe lagged behind favorable climate, a finding consistent with the postglacial migration lag hypothesis proposed by Svenning et al. (2008). Our finding that beech was already present in south-western France 12 kyr ago thus contributes to solve the discrepancy between beech presence and climate detected by both SDM studies in western Europe. Yet, as pointed out by Huntley (2014) and as discussed by Gavin et al. (2014), scale is still an issue when addressing glacial microrefugia using SDM approaches given the lack of high-resolution climate datasets. Well dated and precisely geolocalized fossils such as our soil charcoals will ultimately help progress on such issues (Maiorano et al., 2013). Finally, high-resolution pollen analysis of marine sediments core MD04-2845 from the Bay of Biscay suggest the persistence of broadleaved tree species (including beech) between 25 and 23.8 cal kyr before present (bp; i.e. during Heinrich stadial-2 (HS2)) in the region (M. F. Sánchez-Goñi, unpublished). Although a precise location of the corresponding Pleistocene populations is not possible with such an approach, the results imply the existence of glacial temperate forest refugia somewhere in the Loire and Garonne Valleys. Huntley (2014) specifically questioned the relevance of the reference period that we used to determine the existence of a refugium. Following Tzedakis et al. (2013), we used the Late Pleniglacial (LPG) interval (24.0–14.6 cal kyr bp), defined as the interval of most extreme glacial conditions devoid of major oscillations (in the North Atlantic), as reference period for detecting glacial refugia, whereas the last glacial maximum (LGM; 23–19 cal kyr bp), defined in terms of the most recent interval when global ice sheets reached their maximum integrated volume (Mix et al., 2001), was considered less relevant. The difference is not only semantic: the LPG interval ends with the onset of the abrupt warming of the Lateglacial whereas the LGM reference period excludes c. 4400 yr with extremely cold and arid conditions, including Heinrich stadial-1 (HS1) that was colder and drier than the LGM itself in western Europe. Heinrich events sensu stricto (as invoked by Huntley, 2014) are included in a cycle of recurrent global climate degradations named Heinrich stadials that can be longer than Heinrich events (Sánchez-Goñi & Harrison, 2010). The terminology used by Huntley (2014) ambiguously confuses processes (i.e. Heinrich event sensu stricto) and their climatic impacts (i.e. Heinrich stadial). In our study, we rather focused on the climatically relevant stadial, HS1. A large body of recent research shows that the period post-dating LGM was in fact colder or drier than the LGM in Europe. A thorough (quantitative-based) review on the topic would be highly desirable to formally disentangle this issue. Here we simply provide a preliminary summary of these studies, showing that from 18.1 to 14.6 cal kyr bp the climate was particularly cold and dry, arboreal taxa were at a minimum while semi-desert taxa were at a maximum (Supporting Information Table S1; Fig. 2). This period spans the calibrated age ranges of one beech and one Scots pine macrofossils that we reported in our study. So we conclude that we have indeed found (scattered) fossils during one of the coldest and driest interval in the region. Specifically, given that drought rather than temperature has been shown to limit beech (Jump et al., 2006; Piovesan et al., 2008), a beech charcoal macrofossil found during one of the driest phase of the LPG (HS1, see Fig. 2b) suggests local persistence of the species in a climatically unusual microrefugium. While Huntley (2014) argues that some records and a vegetation model suggest that HS2 was colder than HS1, it is still not clear whether HS1 or HS2 was colder. Indeed, either scenario can be inferred depending on the climatic tracer used, the precise geographic region or the period of the year considered. For example, the coldest interval recorded at the Galicia Margin was HS2 according to alkenone-based annual sea surface temperature and HS1 according to foraminifera-based winter sea surface temperature (Naughton et al., 2009). What is clear however is that both intervals (HS1 and HS2) were climatically harsher (colder and dryer) than the LGM in Europe (e.g. Eynaud et al., 2009). Given that the nearest marine pollen record suggests regional beech presence during HS2 (M. F. Sánchez-Goñi, unpublished) and that our charcoal record indicates local beech presence during HS1, we conclude that beech was present during the two coldest episodes of the last ice age (immediately before and after the LGM). The scenarios of rapid response to millennial climatic fluctuation through colonization from a distant macrorefugium vs persistence in a microrefugium cannot be rigorously distinguished on the basis of our discontinuous macrofossil charcoal record, as pointed out by Huntley (2014) and as acknowledged in our paper. Instead, we must consider the most parsimonious between these two hypotheses. First, we note that quantitative reconstructions of postglacial beech spread in Europe suggest that colonization started in multiple regions from a mosaic of sparse stands, at densities too low to be detectable by pollen analysis (Magri, 2008), as expected if beech populations persisted in several microrefugia. Beech is also known for its particularly low dispersal ability (Bradshaw & Lindbladh, 2005; Feurdean et al., 2013; Saltré et al., 2013), so hypotheses involving multiple episodes of invasions within short periods at times when conditions were still very harsh for the species appear unlikely. In fact, while significant expansions of the Atlantic flora occurred during interstadials 12 and 14 (Sánchez-Goñi et al., 2008; Fletcher et al., 2010), our charcoal record in south-western France corresponds to completely distinct periods (interstadials 10–11, 5–6, and HS1), implying that its presence was not strictly a consequence of those warm episodes in response to millennial climatic fluctuations. Another reason why we consider the microrefugium hypothesis more parsimonious comes from our previous genetic study of beech across 65 populations (2510 individuals typed at 16 highly polymorphic microsatellite loci) covering the species range in France and northern Spain, with a particular focus on south-western France (de Lafontaine et al., 2013). In this region, we detected a complex genetic structure bearing the signature of glacial refugia, that is a lack of isolation by distance, high genetic differentiation between populations and the presence of many outlier populations (Fig. 3). The presence of outlier populations in this region was interpreted as being caused by strong genetic drift caused by small population size and long isolation in unsuitable regional habitat matrix at the rear edge, suggesting the ancient origin of present-day range fragmentation. This interpretation was supported by the lack of evidence for recent bottlenecks in these populations. This pattern is characteristic of other well-established refugial areas, whereas newly colonized regions in Europe show a clear pattern of isolation by distance and little differentiation between populations, with no outlier populations. A review of the existing literature shows that these predictions are robust both in plants and in animals (de Lafontaine et al., 2013). Huntley (2014) only takes note of the fact that in the study of de Lafontaine et al. (2013) a few genetically divergent populations in south-western France characterized by reduced genetic diversity were spuriously assigned to other parts of the range, depending on the assignment method used. In fact, many populations are not only unique to this region but also differ from the other nearby populations, as expected if these populations had been isolated for extended periods of time, given the lack of signal of recent bottleneck (Fig. 3). This is especially true for the Ciron site, where beech charcoal dated from before and after the LPG was found (Fig. 1). This evidence supports the idea of glacial survival of beech in situ rather than of early postglacial colonization of the Landes de Gascogne from the Pyrenees or from south-eastern France. Given that a microrefugial scenario for temperate trees in the Landes de Gascogne area is plausible and is supported by genetic data (de Lafontaine et al., 2013), paleoclimatic models (e.g. Svenning et al., 2008) and high-resolution pollen record from nearby marine sediments core (M. F. Sánchez-Goñi, unpublished), we were interested in assessing whether it was possible to detect the presence of such putative cryptic microrefugia from direct in situ fossil evidence. Our approach to target macrofossils isolated from mineral soil samples allowed us to specifically pinpoint those few unusual sites in the region that still harbor beech. The same reasons that explain beech survival today in these climate refugia within an inhospitable matrix (increased humidity caused by local edaphic and topographic conditions) could explain the species persistence at the exact same location during the last ice age. This might also explain why these microrefugia were not a major source of Holocene expansion (Edwards et al., 2014). The 140 dated charcoal particles reported by de Lafontaine et al. (2014) represent hard-won macrofossils providing robust direct in situ field evidence for the presence of at least two forest tree species (beech and Scots pine) before, during, and after the LPG interval in multiple stands within the study region (for details on the numerous precautions taken to ensure reliable charcoal dating, see de Lafontaine et al., 2014). The microrefugia scenario therefore stands out as the most parsimonious one to explain the existing data. While the questions raised by Huntley (2014) are legitimate, a balanced view should consider both alternatives on the basis of all available evidence. Please note: Wiley Blackwell are not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.