leroy Arpe2007JBI

Journal of Biogeography (J. Biogeogr.) (2007) 34, 2115–2128 ORIGINAL ARTICLE Glacial refugia for summer-green trees in Europe and south-west Asia as proposed by ECHAM3 time-slice atmospheric model simulations Suzanne A. G. Leroy1* and Klaus Arpe2 1 Institute for the Environment, Brunel University, Uxbridge UB8 3PH, UK, 2Max Planck Institute for Meteorology, Hamburg, Germany ABSTRACT Aim To generate maps of potential refugia for summer-green trees during the Last Glacial Maximum (LGM). Locations Southern Europe and south-western Asia. Methods Time-slice simulations of the atmospheric climate with the ECHAM3 model are used for the LGM. Limiting factors beyond which cool and warm groups of deciduous trees cannot grow (such as temperature in growing degree days, minimum monthly temperature and precipitation in summer) are chosen. A limited validation by fossil pollen and charcoal records from LGM sites was done. Results Two sets of maps extending from Europe to the Caspian region for cool and warm summer-green trees are presented. Three criteria are combined using contour lines to indicate confidence levels. Small areas within the three southern peninsulas of Europe (Spain, Italy and Greece) are highlighted as possible refugia for summer-green trees. Further, areas that have remained poorly known are now proposed as refugia, including the Sakarya–Kerempe region in northern Turkey, the east coast of the Black Sea and the area south of the Caspian Sea. Main conclusions The maps produced in this study could be used to facilitate better long-term management for the protection of European and south-western Asian biodiversity. *Correspondence: Suzanne A. G. Leroy, Institute for the Environment, Brunel University, Uxbridge, UB8 3PH, UK. E-mail: [email protected] Keywords Biodiversity, climatic model, Europe, glacial refugia, LGM climate, palynology, summer-green trees. During the Quaternary, Europe lost some species of warm and cool deciduous trees (Bertini, 2003). These losses were greater than those in other continents (Leroy, 2007). During glaciations, most of these two groups of species survived in glacial refugia, called long-term refugia by Bennett et al. (1991). It is believed that most refugia were located in cool, moist midaltitude mountain belts or in drought-prone, but warmer, valley bottoms such as the Danube (Leroy & Roiron, 1996; Svenning, 2003). Some boreal deciduous trees (e.g. Betula) and conifers had refugia at higher latitudes (Willis & van Andel, 2004). It is likely that extinctions of trees took place when populations were restricted to refugia during glacial periods, as the smaller populations were more easily affected by disease, competition and conditions too extreme for the ecological range of the species (Svenning, 2003; Leroy, 2007). It is essential to locate as many long-term refugia as possible, as they are often hot spots of biodiversity worth preserving for the future. Three approaches are possible: fossil pollen and charcoal sites dating from the Last Glacial Maximum (LGM), which are extremely rare, phylogeography (based on modern DNA) and climatic modelling, which is the approach developed here. ª 2007 The Authors Journal compilation ª 2007 Blackwell Publishing Ltd www.blackwellpublishing.com/jbi doi:10.1111/j.1365-2699.2007.01754.x INTRODUCTION 2115 S. A. G. Leroy and K. Arpe It has been shown that there is a direct link between the time when an area was deglaciated and the biodiversity of the area (Montoya et al., 2007). Moreover, it is generally accepted in phylogeography that refugia contain the most diverse gene pools (e.g. Comes & Kadereit, 1998). Molecular phylogeography combined with palaeoecology has established that the recolonization of Europe at the beginning of each interglacial period depended solely on those refugia (Bennett et al., 1991; Hewitt, 1999). Palynology (Brewer et al., 2002; Cheddadi et al., 2006) and the analysis of macrocharcoals (Carcaillet & Vernet, 2001; Cheddadi et al., 2006) have long been the main techniques used to locate refugia. However, each has its own limitations. Palynological assemblages actually reflect the regional vegetation, as pollen grains are easily transported over distances that could reach hundreds or thousands of kilometres, e.g. Cedrus pollen were found in lakes in Italy originating from trees in North Africa (Leroy et al., 1996). Pollen productivity, on the other hand, might have decreased during glacial periods due to a low concentration of CO2 (Ziska & Caulfield, 2000; Leroy, 2007). Macrocharcoal sites are also relatively rare, but their contribution is essential as their taxonomic resolution is often better than for pollen assemblages. The contribution of ancient DNA analysis remains very limited (see ‘Previous analyses’). Refugia are generally located in southern European peninsulas (Iberia, Italy, the Balkans). An appreciation of the importance of refugia in the Black and Caspian Sea regions is now emerging. In general, a westward invasion is more likely for most tree genera, given the relatively isolated position of Iberia at the western edge of Eurasia and the Asian origin of many tree genera (Petit et al., 2005). The south and south-east of the Black Sea (Euxinian vegetation) as well as the south of the Caspian Sea (Hyrcanian vegetation) are regions where Quaternary relicts are most likely to have survived, as can be seen in the modern flora. Zelkova carpinifolia, Pterocarya fraxinifolia and Parrotia persica are examples of this survival at the boundaries of Europe (Leroy & Roiron, 1996; Leroy, 2007). Here we use a third approach to locate potential glacial refugia and hot spots of biodiversity, namely climate modelling. This method is based on the reconstructed climate of the LGM, which identifies all potential areas that have climatic conditions sufficient for the survival of plants and animals, including humans, with common climatic requirements. Palynological, macrocharcoal and phylogeographical studies are all limited by the number of sites or species investigated, but climate modelling does not have such limitations and therefore has the potential to produce more detailed maps than those previously available and to identify potential refugial areas that have not yet been detected by other methods. PREVIOUS ANALYSES Huntley & Birks (1983) and Huntley (1990) presented a series of maps, principally for forest trees, that established the changing distribution of pollen types in space and time across 2116 Europe since 13,000 yr bp. These maps depended heavily on the existence of only a few sites with pollen diagrams. These locations are not distributed evenly over Europe, with large gaps over the continent. However, from the map of 13,000 yr bp it is possible to determine those areas where trees were already present at the beginning of the most recent interglacial period (Huntley & Birks, 1983). Using this method, seventeen possible broad areas of glacial refugia were proposed. The forest was restricted to southern Europe, with the richest refugia in the Balkans, the Alps, the Carpathians and along the Apennines. The analysis by Bennett et al. (1991) confirmed the previous results but it also suggested that the Iberian Peninsula was not so important for the colonization of the rest of Europe. Birks & Line (1993) further discussed the results of Huntley & Birks (1983) and found that the small number of taxa at 13,000 yr bp in areas such as north Spain and the Black Sea is an artefact caused by the small number of available pollen diagrams. A more recent analysis based on the larger number of sites in the European Pollen Database (EPD; Brewer et al., 2002), which starts at 15,000 yr bp, suggested three main regions that acted as primary refugia (refugia for the entire glacial period) with slightly better geographical precision than earlier: the southern Iberian Peninsula (e.g. Sierra Nevada), southern Italy (the southern Apennines chain) and the southern Balkans (the Pindos in Greece). In other words they considered the primary European refugial zone to be limited to the extreme south of the continent (Brewer et al., 2002). A vegetation map of Europe for the LGM has recently been proposed by Jalut et al. (2005). However, extrapolations were made from very few LGM pollen diagrams and from younger pollen sequences. It is important to remember that pollen and charcoal diagrams from glacial refugia themselves are very scarce. Some authors have claimed to have found LGM refugia. Bennett et al. (1991) suggested that sites from Ljubljana (Slovenia) have tree pollen frequencies high enough to indicate the presence of trees through the last cold stage. Brewer et al. (2002) suggested an additional three sites: Padul, Spain; Lagho di Monticchio, Italy; and Ioannina, Greece. All of these sites are characterized by the presence of arboreal pollen during glacial times. The Crimea has also been proposed as a glacial refugium (Cordova, 2006), and Tsereteli et al. (1982) found pollen of warm and cool summer-green trees for the LGM in sufficient numbers to suggest that they were growing locally in Apiancha, Georgia (P. Tarasov, personal communication, 2007). On the basis of well-dated charcoal, Willis & van Andel (2004) found glacial refugia in central and eastern Europe for a range of coniferous trees, including some deciduous species. These sites are discussed further below. A detailed analysis of pollen diagrams for Castanea sativa (Krebs et al., 2004) and the construction of a refugium probability index (Krebs et al., 2004) highlight the likelihood of the existence of more refugia than proposed by the available pollen diagrams alone. Temperate refugial regions in Europe show relatively deep DNA divergence for many taxa, indicating their presence over Journal of Biogeography 34, 2115–2128 ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd Summer-green trees in the LGM CLIMATE MODEL The ECHAM3 T42 model was used for the simulations of the LGM, 18 14C ka bp (thousands of radiocarbon years before present) or 21 ka cal. bp (thousands of calibrated years before present). This model has been presented in detail by Roeckner et al. (1992). Lorenz et al. (1996) described the set-up for these simulations and only a few points are given here. All processes in the atmosphere are simulated in the model, though only partly in a parameterized form. The model was one of the most advanced models available when these simulations were carried out (Boer et al., 1992). T42 refers to the horizontal resolution and means that no waves on any great circle with wavelengths shorter than (360°/42) = 8.6° can be resolved. A few calculations are carried out in the grid domain and for that a grid resolution of 2.8125° is used, i.e. three grid points represent the shortest possible wave. The main quantities of interest, surface temperature, latent heat flux (evaporation) and precipitation, are provided on this grid. The model resolution has an effect on the orography used in the simulations, as it had to be adjusted to the T42 resolution. Its effect is demonstrated in Fig. 1, showing a north–south section of the model orography at 11.25°E, i.e. from Scandinavia, via Germany, across the Alps to Italy. For comparison the orography of a T106 model (1.125° grid) is shown. This is the resolution of the ECMWF (European Centre for Medium-Range Weather Forecast) reanalysis data (Simmons & Gibson, 2000) that will also be used in this study. The original orography, available on a 0.5¢ grid and averaged to a 0.1° grid is shown (0.1). The maximum height of the Alps in T42 is reduced to 834 m compared with 1510 m in a T106 model and 2500 m in a 0.1° grid. The valleys of the Danube (49°N) and Po (45°N) rivers completely disappear in T42 whereas they are still indicated in the T106 resolution, though in T106 the Po valley is too far south. These are deficiencies that severely hamper the usefulness of the model, and a method to reduce their effect is proposed below. The ECHAM3 model had further deficiencies which are documented by Arpe et al. (1994) who concentrated on the 2000 11.25E 1500 Elevation m several glaciations and suggesting speciation by repeated allopatry (Hewitt, 2004). The latter paper adds that DNA evidence for temperate animal species in Europe indicates the existence of different patterns of post-glacial colonization across the same area (a grasshopper dominated by the Balkans source, hedgehogs with three parallel south to north tracks, brown bears with the west–east embrace, the chub along Black Sea rivers) and different patterns in previous climate oscillations. Both the deciduous oaks and humans have a pattern similar to that of hedgehogs (Hewitt, 2004). This also indicates that some refugia did not contribute to the recolonization of Europe; these are called interglacial refugia or silent refugia (Cheddadi et al., 2006). Studies are available for an ever-increasing number of species: for animals see Hewitt (1999), Seddon et al. (2001), Schmitt & Hewitt (2004) and Culling et al. (2006), and for plants see Hewitt (1999), Soranzo et al. (2000), Gugerli et al. (2001) and Petit et al. (2005). All these investigations limit themselves to identifying broad geographical areas. Some genetic studies postulate formerly unknown refugial areas by pointing to locations with a high genetic diversity, such as south-eastern France, Calabria, Crimea, southern Spain and Turkey (Comes & Kadereit, 1998). The specificity of these genetic analyses enables the establishment of colonization routes and suture zones on a map of Europe. Only some of the trees and shrubs of interest in the present study have been previously analysed. For the white oaks (Quercus robur), colonization in three parallel routes from the south to the north occurred (‘pattern of the hedgehog’). Recent investigations have shown that the Italian populations of these trees expanded north of the Alps, which is contrary to what pollen studies had suggested (Ferris et al., 1998; Petit et al., 2002). Hampe et al. (2003) recognized a Balkan refugium for Frangula alnus. This area is a source for most of Europe, complemented by the well-defined haplotypes in Iberia and Anatolia. The phylogeography of Alnus glutinosa (King & Ferris, 1998) indicates that this plant also followed the ‘grasshopper pattern’ in combination with the Balkans haplotype. The latter colonized most of Europe. In addition, refugia in Iberia and Turkey have been found. Both species of Carpinus (Carpinus betulus and Carpinus orientalis) have refugia in southern Italy and the Balkans (Grivet & Petit, 2003). Refugia of Fraxinus excelsior (Heuertz et al., 2004) have been found in Iberia, Italy, the eastern Alps and the Balkan Peninsula. A partial phylogeography of Corylus avellana (Persson et al., 2004) along a north–south transect in Europe is available. In brief, there is still very little known about the locations of tree refugia during the LGM, as most palaeostudies use data that are younger, often starting a few thousand years after the LGM. 1000 500 0 40 45 0.1 50 Latitude LGM 55 T42 60 T106 Figure 1 North–south section of orographic height along the 11th meridian in Europe. The original orography averaged to a 0.1° grid (0.1) is compared with those in the model for the present day (T42), a higher-resolution model (T106) and for the one used for the LGM simulation (LGM). Journal of Biogeography 34, 2115–2128 ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd 2117 S. A. G. Leroy and K. Arpe hydrological cycle, one of the most difficult quantities to simulate but most important for the present study. One problem that affects the present investigation is a simulated annual cycle of precipitation that predicts too much precipitation in winter and too little in summer for the modern climate of mid-latitude and polar regions. To remedy these problems the Max Planck Institute for Meteorology has improved many aspects of their model and generated new versions for general use; at present the best is ECHAM5, but there are no LGM simulations currently available. Figure 1 also shows the orography during the LGM, i.e. the orography topped by the ice sheet. Due to the smoothing effect when generating a T42 resolution for the LGM orography some areas not covered by ice are different from the presentday T42 orography. This creates another problem – for the surface or 2 m (above the ground) temperatures. The temperatures at 2 m above the ground (T2m) in the LGM simulations should surely be colder than those for the presentday climate. A cooler temperature of, for example, 10°C in southern Germany during the LGM does not result only from the cooler climate but also from an elevation that is up to 300 m higher in the LGM orography, which would cause an adiabatic cooling of 3°C. For Greece the difference in orography could mean a warming of perhaps 0.7°C. This problem will be discussed further below. The low resolution of the model is further highlighted in Fig. 2, which shows the land–sea mask used in the model. Squares indicate grid points that are land in the simulation for the present. The area with land points is enlarged during the LGM, indicated by crosses, and dots denote the remaining sea points. The impact of this change in the land–sea mask will become obvious below. The change from sea to land points in the Caspian Sea was probably an error, as the Caspian Sea is too deep to have dried out during the LGM, and there are suggestions that the Caspian Sea was even larger during the LGM than at present (e.g. Crowley & North, 1991). Boundary forcing The model was run with the present-day conditions for sea surface temperatures (SSTs), orography, solar radiation, ice Figure 2 Land–sea mask in the ECHAM3 simulations. Squares indicate grid points that are land in the simulation for the present day. The area with land points is enlarged during the LGM, indicated by crosses, and dots denote the remaining sea points. 2118 cover and CO2 and also under LGM conditions for these parameters (CO2 200 ppm) as reconstructed by the Climate/ Long-Range Investigation, Mapping Prediction project (CLIMAP, 1981). Three estimates are available for the SST. One LGM simulation uses the SST suggested by CLIMAP, and this is the one used mainly in this study. Another simulation uses the same SST except for a tropical SST that is 3°C cooler. A third one uses a more recent estimate by GLAMAP (Sarnthein et al., 2003) with a warmer northern Atlantic and slightly cooler tropical ocean SST than CLIMAP. The SST did not vary from year to year in these simulations so that one can expect a year-by-year variability of atmospheric variables that is too low. Therefore it is reasonable to restrict this study to long-term means only. The availability of a run under present conditions allows for the evaluation of the changes between the two climatic scenarios, which is likely to reduce the impact of systematic errors of the model and help in the investigation of scales that are not covered by the model (downscaling). Model results: LGM minus present climate The model results have been discussed comprehensively by Lorenz et al. (1996). Therefore only a few points that are of interest for possible tree growth in Europe are discussed here. It was assumed that the wind at the surface over Europe was much stronger than today (e.g. Crowley & North, 1991). The surface winds in the two simulations during two seasons were investigated (not shown). For summer (June, July, August) the LGM simulation had only slightly stronger winds of mostly less than 1 ms)1 over large areas of Europe, with a more northerly component. In winter (December, January, February) the increase was larger, often up to 3 ms)1, with a more southerly component. Over the Black Sea differences of 4 ms)1 were reached and over the Mediterranean Sea and the Atlantic 6 ms)1. In winter this difference in wind speed might cause an additional Föhn effect on the northern slopes of the Alps, in Hungary and in the north of Romania. This effect will probably not be represented in the model due to the smoothed orography. It has been stated (e.g. by Crowley & North, 1991) that the surface temperature was cooler by 10°C away from the edge of the glacier. Figure 3 confirms this by showing the difference between the LGM and present-day simulations, with lower temperatures during the LGM for most of Europe, north of the Alps and the Pyrenees. In winter the cooling was more than 10°C and in summer more than 5°C. In Mediterranean countries the cooling was moderate, mostly less than 5°C. Over the ice itself, the cooling was more than 20°C. A few points stand out, especially during summer with warm spots in the LGM run, i.e. points where there is ocean in the present-day run and land in the LGM run (crosses in Fig. 2). Therefore, in summer the surface temperature can warm up much more in the LGM run than for the present. Here the role of the Caspian Sea is a matter of controversy (Grosswald, 1980) but the simulation is probably wrong in assuming that there was no Caspian Sea at all during LGM. Journal of Biogeography 34, 2115–2128 ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd Summer-green trees in the LGM Figure 3 Temperature differences 2 m above ground (T2m) in the simulations for LGM minus the present day in summer and winter. Contours at ± 0, 5, 10, 15, 20°C, shading for < )10°C and > 0°C. During the LGM, overall precipitation is said to have been much lower than now, resulting in polar desert conditions in some parts of Europe (e.g. Crowley & North, 1991). A decrease of precipitation was also likely because a cooler atmosphere can carry less water vapour, although there is no simple relationship between air temperature and precipitation. In Fig. 4 this is examined from a broader perspective. The simulation for the present climate compares reasonably well with estimates of the true precipitation (not shown), for example the data from the Global Precipitation Climate Project (GPCP; Huffman et al., 1996). The simulated values are only slightly lower than the observed data. In Fig. 4, the LGM simulation shows a clear decrease of precipitation over the Mediterranean region in summer and winter. Over central Europe a decrease in precipitation can be found only for summer, while there is an increase in winter. For the annual mean the difference is small; an increase over western Europe, e.g. England, and a decrease over eastern and southern Europe. The decrease is strongest in the simulation using the GLAMAP SSTs, which are warmer over the northern Atlantic. The reason for the increase in precipitation over central and western Europe in the LGM simulations for winter lies in a change of the cyclone tracks. In Fig. 4, the cyclone track is made visible by bands with larger precipitation values. In the present climate, during winter the cyclones move from the north American coast in an east-north-east direction over the Atlantic, and then shortly before reaching Ireland they change to a north-north-east direction towards the coast of Norway. During the LGM, the cyclones do not change direction west of Ireland but continue almost straight to the east towards central Europe where they finally decay. This brings more precipitation into western and central Europe but less for Scandinavia. In winter one finds a weak secondary precipitation maximum over the Mediterranean Sea for the present-day climate due to the occurrence of blockings of the atmospheric flow. This is absent in the LGM simulations. The summer precipitation is Figure 4 Precipitation in the simulations for the present day and the LGM during summer and winter. Contours at 30, 60, 100, 200, 400, 600 mm per season, shading for < 100 and > 400 mm per season. governed more by convective precipitation that decreases in LGM conditions. The precipitation minus evaporation (P ) E) is more important for life. Evaporation rates should have been higher during the LGM due to stronger surface winds but lower due to lower temperatures. The combined effect of these two processes in the model is an overall decrease in evaporation under LGM conditions. Another parameter for evaporation is the amount of solar radiation reaching the ground, a variable which was not available to us. P ) E in the LGM simulation is enhanced, i.e. there is a higher water availability, compared with the present-day simulation in all seasons except spring (not shown), and especially strongly so over north-western Europe, e.g. northern France, England and Denmark. Again, areas that are ‘sea points’ in the simulation for the present day but land during LGM stand out with positive P ) E values (or less E during LGM) for all seasons. Annual P ) E means are most similar to the winter values. Perhaps the reduced concentration of CO2 has to be considered, meaning that plants would have to open their stomata more under LGM than under present conditions, which would require a higher availability of water in the ground, an effect that cannot be quantified here (Gedney et al., 2006). Journal of Biogeography 34, 2115–2128 ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd 2119 S. A. G. Leroy and K. Arpe Another quantity used in this respect is the ratio between annual means of evaporation and precipitation (E/P). This can only take values between zero and one over land. The smaller the value the more water is available for plants and river discharge. Over land this ratio decreases in the LGM run compared with the present-day run (not shown), i.e. the LGMsimulated climate is more favourable for tree growth than the simulated present climate as already shown above for P ) E. Over the Mediterranean Sea itself this ratio is considerably enhanced under LGM conditions that provide more water vapour for the adjacent continents. The impact of annual mean water availability on tree growth therefore seems to be unimportant for areas where there are trees now, but precipitation in summer or the first growing season (defined below) will be used as a limiting factor (see further discussion). Requirements for tree growth The main focus of this study is temperate deciduous forests. For these the limits beyond which the trees cannot grow were examined. Possible factors limiting the growth of trees are the temperature, water in the ground, nutrients and wind. These factors were included in a sophisticated way in the BIOME model (Kaplan et al., 2003) used by Harrison & Prentice (2003). For the present study simpler definitions of such limits were preferred because the focus is not on the dominant vegetation but whether trees could have survived in specific locations. This approach also provides some insight into how much uncertainties in the precipitation and temperature estimates during the LGM impact upon the results. For the separation of temperate deciduous trees into cool and warm species, the definitions in Prentice et al. (1996) and in Laurent et al. (2004) were used (Table 1). Those lists were combined for this study because limits for possible tree growth from both publications had to be used to generate a complete set. The limits for tree growth were not identical in both papers (see Table 2). The values selected from these are given Table 3 and will be discussed below. It is not obvious why GDD5 (the sum of growing degree days multiplied by the temperature above 5°C) for cool summer-green trees should be higher than that for warm Table 1 List of warm and cool summer-green trees according to P (Prentice et al., 1996) and L (Laurent et al., 2004). Source Warm Cool P Castanea, Juglans, Ostrya, Platanus, Rhamnus, Fraxinus ornus, Vitis Castanea, Juglans, Ostrya, Quercus pubescens Carpinus, Corylus, Fagus, Tilia, Ulmus, Frangula L 2120 Carpinus, Corylus, Fagus, Tilia, Ulmus, Acer, Populus, Fraxinus excelsior, Alnus, Quercus robur Table 2 Minimum requirements for summer-green tree growth according to P (Prentice et al., 1996), L (Laurent et al., 2004)* and K (Kaplan et al., 2003). Parameter Mean temperature of coldest month (°C) GDD5 (day °C) Ratio (alpha) of actual to equilibrium evapotranspiration Temperature of coldest month (°C) GDD5 (day °C) Temperature of coldest month (°C) GDD5 (day °C) Summer precipitation (mm per season) Cooltemp. Warmtemp. Source )15 )2 P 900 0.65 800 0.65 P P )19 ) K 900 )7 to )14 ) )2.5 K L 560 to 1250 45 to 75 1380 60 L L *They provide ranges that are simplified here. GDD5 = growing degree days, i.e. sum of days times the temperature above 5°C. Table 3 Minimum requirements for summer-green tree growth simplified from Table 2 and used in this study. Parameter Cool-temp. Warm-temp. Mean temperature of coldest month (°C) GDD5 (day °C) Summer precipitation (mm per season) )15 )2.5 800 50 1000 60 summer-green trees as given by Prentice et al. (1996), but for the latter we found that the limiting factor is mainly the coldest monthly mean temperature. So the minimum GDD5 does not matter as much for this group of trees. It is also doubtful that only summer precipitation is important since growing starts for large areas during the LGM in April or May (mean temperature > 5°C). When using precipitation over 3 months after reaching the 5°C threshold, instead of summer (June, July, August), a much larger area of possible tree growth is found, especially for Mediterranean regions. The ratio alpha (approximated by E/P) has already been discussed above for the simulation data and was found to be more favourable for tree growth under LGM conditions. This is not used as a limiting factor in the further discussion. Also Prentice et al. (1996) did not use it explicitly in their BIOME4 model (http://www-lsce.cea.fr/pmip2/synth/biome4.shtml). In fact they set evaporation to zero and left evapotranspiration to the parameterization of transpiration by plants. In the LGM simulations there are much stronger wind speeds than at the present, especially in winter. Increases of 100% are found over large areas, but the limits beyond which tree growth is no longer possible are not known. However, Journal of Biogeography 34, 2115–2128 ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd Summer-green trees in the LGM wind speeds calculated from monthly mean wind components of more than 2.5 ms)1 over large areas of Europe, reaching maximum values of 8 ms)1 are greater than those, for example, in Fireland (Ushuaia) under present conditions, with seasonal mean wind speeds of 3 ms)1. There wind prevents tree growth except in sheltered valleys. If one uses this as a limit, much of Europe would have been excluded from tree growth during the LGM. Part of Spain, northern Italy, the Balkans and the southern Caucasus would have been the only suitable areas. In Ushuaia strong winds prevail throughout the year, while for the LGM model in Europe strong winds are not found in summer, which is the main growing season; so it might not be a limiting factor. Also, on a daily scale, 3 ms)1 is not a strong wind. This question was not investigated further because the limiting factors of winds are not known for the present climate. The minimum size required for an area to be able to act as a refugium for trees is unknown. On the one hand foresters believe that there is enough biodiversity in a forest of 100 ha for renewal of the forest, but this is on the short time-scale of timber production. On the other hand, tress would have had to survive in glacial refugia for glacial periods that may have lasted several tens of millennia; hence the refugia would need to be larger. In the Amazonian forest, the ‘minimum critical size of ecosystems’ has been studied in the field (Laurance et al., 2002) with sites up to 100 ha. The results are not conclusive, but these authors believe that 100 ha is too little and suggest sizes of 10,000 to 100,000 km2. The size of the grid cell used in this downscaling model is between 2500 and 2700 km2. This is perhaps near to the minimum area required for a glacial refugium. The climatic conditions of the LGM were probably much more extreme than those of previous Quaternary glacial periods (EPICA Community Members, 2004), hence the LGM refugia were smaller than those during other glacial periods. Therefore it is likely that the results using the LGM represent the maximum constraint on the size of the refugia. Downscaling using present-day climate data As mentioned above, the available simulations were carried out with a relatively low-resolution model and are therefore unable to reproduce the relatively higher temperatures in broad valleys like the Danube in southern Romania or the upper Rhine valley. For downscaling the results of the model to the higher resolution needed for further investigation, a simple method is used similar to that used by Harrison & Prentice (2003) and others: a high-resolution present-day climatology of the temperature at 2 m above ground and precipitation is taken, and the difference between the LGM simulation and present-day simulation is added to that. The ECMWF reanalysis data are available on a T106 resolution (1.125° grid). Another possible data set is the 0.5° grid climatology of Cramer and Leemans (http://www.pik-potsdam.de/~cramer/ climate.html). Finally, emphasis is placed on the latter because it has a higher resolution and shows clearly the effects of mountain ranges. It is also the data set used by Harrison & Prentice (2003). The resolution in this climatology (0.5°) is in many areas higher than is justified by the available observational data, a fact that has to be kept in mind when interpreting the results. On the other hand it is still too coarse to represent climatically favoured regions like the upper Rhine Valley in Germany, now known for its wine growing. Similarly it is still too coarse to fully represent the orographic impacts on scales that are important for Italy and Greece. This method also serves to compensate for the systematic error of the model. For example, the model has a tendency to simulate too little summer precipitation in mid-latitudes. Assuming that this systematic error is the same in the simulations for the present-day and for the LGM, it would not affect the results in the present approach at all. However, some aspects of the systematic errors are connected with the cyclone tracks, and it has been shown above that the cyclone tracks in the LGM runs during winter are noticeably different from the present-day runs (discussed in connection with Fig. 4). The method is likely to enhance the precipitation over the Mediterranean regions in winter. This might mean a falsification of the LGM runs, although it is hard to be certain. The Caspian Sea is another point of concern because it does not exist in the LGM simulations. Its impact on the temperature at 2 m above ground has been shown above. This error cannot be repaired by the technique used here. On the whole, however, the method works very well, giving for example the detailed structures connected with the orography; and the positive points of the selected approach appear to outweigh the negative ones. When comparing the surface temperatures in the presentday and the LGM simulations, one has to bear in mind that these differences are due in part to the different altitudes in the different simulations that are shown in Fig. 1. The vertical temperature lapse rate is on average about 0.6°C per 100 m and the adiabatic one is 1°C per 100 m. In the lee of the main mountain ranges, the latter might be the best lapse rate to choose for compensating the differences in the elevation, and in other cases the former lapse rate might be chosen. As a compromise a lapse rate of 0.7°C was used to correct for the height differences. The effect of this temperature correction is largest over the ice sheet and near its edge, but in the area of interest the corrections are small, mostly a warming of less than 0.3°C, except for Hungary and Romania with had values of up to 1.5°C. For northern Spain and Greece this correction leads to a cooling of up to 1°C. It was shown above that the surface winds not only change their strength but also their direction. This could have a profound impact on orographic precipitation, mostly on smaller scales. This problem can only be solved by using a nested high-resolution model. RESULTS There are three primary limiting factors for tree growth: temperature expressed in growing degree days (GDD5), minimum monthly mean temperature and soil moisture Journal of Biogeography 34, 2115–2128 ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd 2121 S. A. G. Leroy and K. Arpe expressed as summer precipitation. These factors are presented in Fig. 5 for the LGM simulation using the CLIMAP SSTs. The former two provide, as would be expected, the northern boundaries for tree growth and the latter provides the southern boundary. From the range of limiting factors given in Table 2 we chose those given in Table 3. The minimum temperature contour of )15°C (dark shading for )14°C in Fig. 5a), which is the limit for cool summer-green trees (Table 3), however, lies so far north that it is mostly beyond the limit provided by GDD5. It may play a role in France and north of the Caspian Sea. For warm summer-green trees, the )2.5°C contour eliminates most of Europe for these trees, and only areas (light shading) in Spain, Italy, Greece, Turkey and spots on the eastern coast of the Black Sea and the southern coast of the Caspian Sea remain (Fig. 5a). In Fig. 5b, the GDD5 map shows a very strong gradient from 500 to 1000 units between 45 and 50°N and along the main mountain ranges. This means that the exact limit of the minimum GDD5 for tree growth seems to be less certain for most areas except France and Romania. The minimum summer precipitation (Fig. 5c) of > 60 mm per season (shaded) is fulfilled over most of Europe except for parts of the Mediterranean region. This criterion is also fulfilled in the Caucasus, around the Black Sea, and on the western and southern coasts of the Caspian Sea. Figure 5 Limiting factors for possible growth of summer-green trees. (a) Monthly mean minimum 2 m temperature in the downscaled LGM simulation. Contours at ± 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 25°C, shading for < )14°C and > )2°C. (b) Growing degree days, GDD5, in the downscaled LGM simulation. Contours at 300, 600, 1000, 1400, 2000, 3000 GDD5, shading for > 1400 and < 600 GDD5. (c) Precipitation during summer in the downscaled LGM simulation. Contours at 30, 60, 100, 200, 400, 600 mm per season, shading for > 60 and > 200 mm per season. 2122 It is interesting to compare Fig. 5c with Fig. 4b as it shows the effect of the downscaling method applied here. It compensates for the systematic error of the model by increasing the summer precipitation over most of Europe. It also introduces the effect of mountains, which results in more precipitation for the Pyrenees, the Alps, the Carpathians and the Caucasus that might be important for the following discussion. Also, the bands of enhanced precipitation along the southern and eastern coast of the Black Sea and the southern coast of the Caspian Sea are important, as these areas are not too far away to have an influence on the European gene pool. The use of summer precipitation as a limiting factor was questioned above, because for some trees it is during the early growing season that most water is needed for producing their leaves. Therefore, precipitation in the early growing period might be more important than summer rainfall. The beginning of the growing season is the month when the temperature at 2 m above ground (T2m) exceeds 5°C. In all areas south of 45°N this occurs no later than May. Precipitation in the 3 months after T2m > 5°C has been reached is much greater in the Mediterranean regions than in summer. This increases the area of possible tree growth, even if one increases the minimum precipitation requirement to the higher value of 100 mm per season. In Fig. 6, the three limiting factors have been combined for summer-green trees. To show the increasing likelihood of tree growth in a warmer and wetter climate, different combinations of GDD5 and summer precipitation were used, ranging from a number 1 to 7. For cool summer-green trees, i.e. minimum temperature > )15°C in Fig. 6a, the number 1 represents the combination GDD5 > 800 and precipitation > 50 and the number 6.5 represents GDD5 > 1500 and precipitation > 150. The best conditions for cool summer-green trees are in the Balkans and around the Black Sea (especially the north and east). Some smaller areas can be found in the Iberian Peninsula, the Pyrenees, southern France, the upper Rhine Valley, Italy, the Sakarya–Kerempe–Sinop section of the northern Turkish coast and the southern coast of the Caspian Sea. When discussing water availability (P ) E) it was stated that this limiting factor could be neglected for the LGM in areas where there are trees growing now. At present there is steppe north of Crimea, and this was probably steppe during the LGM, contrary to what is shown in Fig. 6a. In Fig. 6b, the limiting factors have been combined for warm summer-green trees, i.e. minimum temperature > )2.5°C. For this group the values can be 2 (GDD5 > 1000, precipitation > 60), 4 (GDD5 > 1600, precipitation > 80) or 6 (GDD5 > 2000, precipitation > 100). Only a few places remain for possible tree growth – about 10 regions in the Iberian Peninsula, the Riviera, large parts of the Apennines and Ancona, Dalmatia, isolated spots in Greece and Thracia, the Sakarya–Kerempe region of Turkey and those with the highest values on the east coast of the Black Sea and the southern coast of the Caspian Sea. The refugial areas along the Caspian Sea should probably be enlarged as the LGM model simulations lacked the influence of Journal of Biogeography 34, 2115–2128 ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd Summer-green trees in the LGM tree growth, mostly in connection with individual mountains. For example, at the points in Tunisia and southern Turkey, summer precipitation just surpasses 60 mm per season and in the Caucasus the minimum monthly mean temperature is warmer than )2.5°C in a single valley. These relative extremes result from the climatological data of the present and cannot be attributed to the model simulations. These points show up so strongly because the value for the lowest threshold for tree growth is set at 2 for warm summer-green trees while the lowest contour is set at 1. From these it is easy to imagine that more small-scale areas might emerge if a higher-resolution data set was available. Figure 6 Limits for the growth of summer-green trees. Contours at 1, 3, 5, higher values mean higher likeliness. Shading for > 1 and > 5. Squares and crosses (white in black) indicate positions of findings of trees. Crosses less certain. (a) Cool summergreen trees: < 1 = precipitation < 50 mm per season or GDD5 < 800 or minimum temperature < )15°C; 4 = precipitation > 100 mm per season and GDD5 > 1200 and minimum temperature > )15°C. (b) Warm summer-green trees: < 2 = precipitation < 60 mm per season or GDD5 < 1000 or minimum temperature < )2.5°C; 4 = precipitation > 80 mm per season and GDD5 > 1600 and minimum temperature > )2.5°C. the sea itself, which would moisten the air and reduce the extremes of surface temperature in summer and winter. The presence of refugia along coastlines has often been proposed (Leroy et al., 1994; Leroy, 2007). In the present study coastal refugia were found frequently in close proximity to mountains. This is probably the result of increased precipitation due to the abrupt change in altitude from sea level. The areas of possible tree growth are considerably enlarged when using spring instead of summer precipitation as a limiting factor (not shown). Wide areas of the Iberian Peninsula, Turkey and northern Africa are likely to have had trees during the LGM. This contradicts the negative sites given by Krebs et al. (2004). This idea was not examined further, because even if there is enough precipitation in spring the trees would have to survive summer droughts, a limiting factor not fully included in the present approach. Comparing the present results with Fig. 7 of Harrison & Prentice (2003) for ECHAM3, one should not be surprised to find many similarities, as the same data and similar methods were used. However, they had a different aim in their paper and therefore only undetailed global maps were shown. The present two maps [cool summer-green (Fig. 6a) and warm summer-green (Fig. 6b)] portray the distribution of these groups of trees from western Europe to the east of the Caspian Sea with contour lines showing confidence intervals for the likelihood of their presence. It is even possible to see for any site (Fig. 5) which of the three parameters is the limiting factor (see following section). In Fig. 6b, there are a few very small areas that might look like plotting errors because they cannot be seen in Fig. 6a. In fact they are single grid points that just surpass the limits for Observational sites for validation Pollen sites used as validation points are those that have evidence of local tree growth and which span the LGM. Only five sites (the first five in Table 4) fulfil this requirement despite a concerted search of the literature and data bases (Peyron et al., 1998; Tarasov et al., 1999; Elenga et al., 2000). The positions of these sites are marked in the figures as squares. Additional sites have low values for temperate deciduous pollen, which suggest that the trees might not have grown locally. Therefore these were not included in the validation. However, this low level of tree pollen may be due to low CO2 at that time. The latter has been demonstrated in greenhouses to have a direct effect on reduced pollen production (Ziska & Caulfield, 2000). To expand the discussion, four other sites (the last in Table 4), that have been considered to be very likely refugia for other reasons, have also been included, marked in the figures as white in black crosses. Charcoal of Castanea, Corylus, Fraxinus and deciduous Quercus have been found in the caves of Altamira and El Buxu in Cantabria (Uzquiano, 1992). The stratigraphy of the first of these caves indicates a possible charcoal age of 17–15 14C ka bp. This site fails the requirement for summer precipitation to be greater than 50 mm per season. In Padul, deciduous species of Quercus seem to survive the LGM well (Pons & Reille, 1988). Padul is south-west of the Sierra Nevada near Grenada in southern Spain, an area for which the requirements for warm and cool summer-green trees are also met (see Fig. 6), although Padul itself fails the requirement of precipitation of more than 50 mm in summer. The tree pollen could have been transported from the nearby Sierra Nevada. The pollen site of Siles, at 1320 m a.s.l., is a lake in an intramontane valley of the Sierra de Segura in Jaén (Carrión, 2002). The pollen spectra at c. 17 14C ka bp indicate the presence of Corylus, Fraxinus, Acer, Ulmus and other trees. This site easily fulfils the requirements for warm and cool summer-green trees in this study. In Ioannina, western Greece (Tzedakis, 1994), some warm summer-green trees survived some glacial periods (e.g. Quercus during the LGM) but during other glacials there are often very low levels of temperate deciduous pollen. In this study, this site fails the requirements for both the warm and cool summergreen tree growth because of the summer precipitation Journal of Biogeography 34, 2115–2128 ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd 2123 S. A. G. Leroy and K. Arpe Table 4 Pollen sites with evidence of tree growth during the LGM. Site Latitude Longitude Altitude (m a.s.l.) Source Cantabria Padul Siles Ioannina Apiancha Istria 43.4°N 37.0°N 38.24°N 39.65°N 42.97°N 44.9°N 4.10°W 3.67°W 2.30°W 20.91°E 41.25°E 13.9°E 70 785 1320 470 450 – Ljubljana region L. di Monticchio 46.0°N 40.94°N 14.5°E 15.6°E 300 1326 Crimea 44°N 34°E 0–1000 m Uzquiano (1992) Pons & Reille (1988) Carrión (2002) Tzedakis (1994) Tsereteli et al. (1982) Culiberg & Sercelj (1995) in Willis & van Andel (2004) Sercelj (1962) Watts et al. (1996); coordinates corrected Cordova (2006) requirement. An adjacent grid point has precipitation of 45 mm in summer and other grid points a little further north do fulfil the requirement for cool trees, as can be seen in Fig. 6a. In Apiancha in Georgia, Tsereteli et al. (1982) found pollen of warm and cool summer-green trees for the LGM. In Fig. 6 the area around this site is marked as a possible refugium with a high likelihood for both taxa, though the site itself fails the precipitation criterion. Fagus sylvatica and Rhamnus cathartica charcoals have been found in Istria, Croatia, for 25,000–20,000 cal. yr bp (Willis & van Andel, 2004). According to the present study this site has possible cool tree growth (Fig. 6a). In the same publication several sites were given where charcoal had been analysed, especially in Hungary. None of those revealed the existence of summer-green trees during the period 25,000–15,000 cal. yr bp whereas the present study suggests that such trees could have been present. Pollen diagrams of Slovenia (Sercelj, 1962) contain many arboreal taxa in the early Holocene and may be as old as the LGM, but the sequences are not dated. In Fig. 6 the region near Ljubljana (Slovenia) has a clear potential for cool summergreen trees but not for warm summer-green trees. The lowest temperature in the year is simulated to be )8.1°C, whereas warm summer-green trees require a minimum temperature of no lower than )2.5°C. This agrees with the data of Sercelj (1962), who found pollen only from cool summer-green trees in this area for the LGM. This finding is not changed by using the warmer Atlantic SSTs suggested by GLAMAP. Bennett et al. (1991) considered the area of Lagho di Monticchio (Monte Vulture, east of Napoli in Italy) as possible refugia for deciduous trees, but Watts et al. (1996) found no tree pollen for the full glacial period except for Betula. This site fulfils the requirements for warm and cool summer-green trees. The minimum temperature is, however, at the borderline and the use of the colder tropical SSTs or warmer Atlantic SSTs in the simulations brings the minimum temperature to values < )2.5°C, thus excluding warm summer-green trees. So the present analysis does not solve the controversy concerning this site. Cordova (2006) and Cordova & Lehman (2006) suggested that the Crimean coast was a refugium for Alnus, Carpinus, 2124 Corylus, Quercus and Ulmus, i.e. cool summer-green trees. Their pollen data did not go as far back as the LGM but as their earliest data at 12,000 14C yr bp showed pollen from these trees, it is likely that these trees survived the LGM locally. In Fig. 6a it can be seen that this is a site where cool trees might have survived the LGM according to the present study. Two other regions with likely tree growth during the LGM have been deduced from modern observations to the south and south-east of the Black Sea as well as the south of the Caspian Sea. In these regions Quaternary relicts are most likely to have survived because of the modern flora: Zelkova carpinifolia, Pterocarya fraxinifolia and Parrotia persica are examples of this survival at the boundaries of Europe (Leroy & Roiron, 1996; Leroy, 2007). Several sites which were investigated above failed the requirement of at least 50 mm precipitation during the summer season. The common factors in these areas are that they are dominated by strong small-scale orographic structures, which suggests that the impact on local precipitation might not be represented by the climatological estimate on the 0.5° grid used here. However, within the same mountain range possible tree growth can be found in Fig. 6, which suggests that the method used needs improvements. This is discussed further below. Results using other SST estimates during the LGM The discussion above has concentrated on the simulations that use the SSTs suggested by CLIMAP (1981). A simulation with SSTs 3°C cooler in the tropics is also available, and Lohmann & Lorenz (2000) regard this as more realistic in some respects. This SST generally leads to a shrinking of the area of possible tree growth, but the differences are small. The loss of possible warm summer-green tree growth along the southern coast of the Caspian Sea due to a drop of the minimum winter temperature by 4°C is important. The estimates by Sarnthein et al. (2003) (GLAMAP2000) provide much warmer SSTs over the northern Atlantic, and one can also expect increased temperatures at 2 m above ground over Europe. Accordingly there is a much larger area with possible tree growth than in the LGM simulation using the CLIMAP SSTs. The whole of western Europe including Journal of Biogeography 34, 2115–2128 ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd Summer-green trees in the LGM Belgium could have had cool summer-green trees. For warm summer-green trees an expansion of the area of possible growth is strongest for the Iberian Peninsula. cooperation with the German Climate Computer Centre (DKRZ). Mike Turner (Brunel University) and Ken Wallace of IGCP 521 WG12 have kindly improved the English in this paper. Outlook The atmospheric model data were generated with the ECHAM3 model. This investigation should be repeated when simulations with a more recent model such as ECHAM5 and with a higher resolution become available. There are quite a few uncertainties in the present approach, such as the spatial resolution; nevertheless these results are a valuable first step in finding possible areas for refugia and hot spots of biodiversity that have not yet been identified by observational data. In the LGM simulations the Caspian Sea is shown as dry land. It is most likely, however, that at least the south basin (south of which refugia are very likely) never dried out completely. Further simulations should be made with a full Caspian Sea. CONCLUSION This study used climate modelling to provide a geographical distribution of possible tree growth during the LGM, filling the huge gaps between sites with observational data. Warm summer-green trees could have survived in several areas of the Iberian Peninsula, Italy, the Balkans, in the Pindos, east of the Rhodopes, at some places along the south and east coasts of the Black Sea and along the south coast of the Caspian Sea. For cool summer-green trees the area is much larger, including much of southern France, the Balkans, Hungary and large areas north of the Black Sea. There is a great scarcity of pollen and charcoal sites dated to the LGM with summer-green trees (only five sites have been found). These and four further sites were used for validation. This study has allowed areas of potential high biodiversity to be highlighted not only for plants but also for animals and humans, which depend on each other. On the one hand these areas should serve as guides for further palaeoecological studies as well as ancient and modern DNA analyses; on the other hand these areas, which served as long-term glacial refugia (and some also as interglacial refugia), will most probably be refugia again in the future. Therefore, these regions require protection and this study may serve as a guide for the establishment of new nature reserves. The inclusion of the Black and the Caspian Sea regions, which are often excluded from European maps, is especially critical as these regions no doubt have had an influence on the European genotypes of the temperate summer-green trees, especially the warm group. ACKNOWLEDGEMENTS The authors are grateful to their colleagues (especially Stephan Lorenz) at the Max-Planck Institute for Meteorology who prepared and made available the LGM simulations in REFERENCES Arpe, K., Bengtsson, L., Dümenil, L. & Roeckner, E. (1994) The hydrological cycle in the ECHAM3 simulations of the atmospheric circulation. Global precipitations and climate change (ed. by M. Desbois and F. Desalmand), NATO, 361–377. Bennett, K.D., Tzedakis, P.C. & Willis, K.J. (1991) Quaternary refugia of North European trees. Journal of Biogeography, 18, 103–115. Bertini, A. (2003) Early to Middle Pleistocene changes of the Italian flora and vegetation in the light of a chronostratigraphic framework. Il Quaternario, 16, 19–36. Birks, H.J.B. & Line, J.M. (1993) Glacial refugia of European trees–a matter of chance? Dissertationes Botanicae, 196, 283– 291. Boer, G.J., Arpe, K., Blackburn, M., Deque, M., Gates, W.L., Hart, T.L., LeTreut, H., Roeckner, E., Sheinin, D.A., Simmonds, I., Smith, R.N.B., Tokioka, T., Wetherald, R.T. & Williamson, D. (1992) Some results from an intercomparison of the climates simulated by 14 atmospheric general circulation models. Journal of Geophysical Research, 97(D12), 12771–12786. Brewer, S., Cheddadi, R., de Beaulieu, J.-L., Reille, M. & Data Contributors (2002) The spread of deciduous Quercus throughout Europe since the last glacial period. Forest Ecology and Management, 16, 27–48. Carcaillet, C. & Vernet, J.-L. (2001) Comments on ‘The fullglacial forests of central and southern Europe’ by Willis et al.. Quaternary Research, 55, 385–387. Carrión, J.S. (2002) Patterns and process of late Quaternary environmental change in a montane region of southwestern Europe. Quaternary Science Reviews, 21, 2047–2066. Cheddadi, R., Vendramin, G.G., Litt, T., François, L., Kageyama, M., Lorentz, S., Laurent, J.-M., de Beaulieu, J.-L., Sadori, L., Jost, A. & Lunt, D. (2006) Imprints of glacial refugia in the modern genetic diversity of Pinus sylvestris. Global Ecology and Biogeography, 15, 271–282. CLIMAP (1981) Seasonal reconstructions of the Earth’s surface at the last glacial maximum. Geological Society of America Map Chart Series, MC-36. Geological Society of America, Boulder, CO. Comes, H.P. & Kadereit, J.W. (1998) The effect of Quaternary climatic changes on plant distribution and evolution. Trends in Plant Sciences, 3, 432–438. Cordova, C.E. (2006) Holocene mediterranization of the southern Crimean vegetation: Paleoecological records, regional climate change, and possible non-climatic influences. The Black Sea flood question: changes in coastline, climate and human settlement (ed. by V. YankoHombach, A. Gilbert, N. Panin and P. Dolukhanov), pp. 319–344. Springer, Berlin. Journal of Biogeography 34, 2115–2128 ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd 2125 S. A. G. Leroy and K. Arpe Cordova, C.E. & Lehman, P.H. (2006) Mediterranean agriculture in southwestern Crimea: paleoenvironments and early adaptations. Beyond the Steppe and the Sown. Proceedings of the 2002 University of Chicago Conference on Eurasian Archaeology (ed. by D.L. Peterson, L.M. Popova and A.T. Smith), pp. 425–447. Colloquio Pontica Series 13. Leiden, Brill. Crowley, T.J. & North, G.R. (1991) Paleoclimatology. Oxford University Press, New York. Culiberg, M. & Sercelj, A. (1995) Anthracological and palynological research in the palaeolithic site Sandalja II (Istria, Croatia). Razprave IV, Razreda Sazu, 3, 49–57. Culling, M.A., Janko, K., Boron, A., Vasil’ev, V.P., Cote, I.M. & Hewitt, G.M. (2006) European colonization by the spined loach (Cobitis taenia) from Ponto-Caspian refugia based on mitochondrial DNA variation. Molecular Ecology, 15, 173–190. Elenga, H., Peyron, O., Bonnefille, R., Jolly, D., Cheddadi, R., Guiot, J., Andrieu, V., Bottema, S., Buchet, G., de Beaulieu, J.-L., Hamilton, A.C., Maley, J., Marchant, R., Perez-Obiol, R., Reille, M., Riollet, G., Scott, L., Straka, H., Taylor, D., Van Campo, E., Vincns, A., Laarif, F. & Jonson, H. (2000) Pollen-based biome reconstruction for southern Europe and Africa 18,000 yr BP. Journal of Biogeography, 27, 621–634. EPICA Community Members (2004) Eight glacial cycles from an Antarctic ice core. Nature, 429, 623–628. Ferris, C., King, R.A., Vainola, R. & Hewitt, G.M. (1998) Chloroplast DNA recognizes three refugial sources of European oaks and suggests independent eastern and western immigrations to Finland. Heredity, 5, 584–593. Gedney, N., Cox, P.M., Betts, R.A., Boucher, O., Huntingford, C. & Stott, P.A. (2006) Detection of a direct carbon dioxide effect in continental river runoff records. Nature, 439, 835–838. Grivet, D. & Petit, R. (2003) Chloroplast DNA phylogeography of the hornbeam in Europe: evidence for a bottleneck at the outset of postglacial colonization. Conservation Genetics, 4, 47–56. Grosswald, M. (1980) Late Weichselian ice sheet of northern Eurasia. Quaternary Research, 13, 1–32. Gugerli, F., Anzidei, M., Madaghiele, A., Büchler, U, Sperisen, C., Senn, J. & Vendramin, G.G. (2001) Chloroplast microsatellites and mitochondrial sequences indicate phylogeographic relationship of Swiss stone pine (Pinus cembra), Siberian stone pine (P. sibirica), and Siberian dwarf pine (P. pumila). Molecular Ecology, 10, 1489–1497. Hampe, A., Arroyo, J., Jordano, P. & Petit, R.J. (2003) Rangewide phylogeography of a bird-dispersed Eurasian shrub: contrasting Mediterranean and temperate glacial refugia. Molecular Ecology, 12, 3415–3426. Harrison, S.P. & Prentice, C.I. (2003) Climate and CO2 controls on global vegetation distribution at the last glacial maximum: analysis based on palaeovegetation data, biome modeling and palaeoclimate simulations. Global Change Biology, 9, 983–1004. Heuertz, M., Fineschi, S., Anzidei, M., Pastorelli, R., Salvini, D., Paule, L., Frascaria-Lacoste, N., Hardy, O.J., Vekemans, X. & 2126 Vendramin, G.G. (2004) Chloroplast DNA variation and postglacial recolonisation of common ash (Fraxinus excelsior L.) in Europe. Molecular Ecology, 13, 3437–3452. Hewitt, G. (1999) Postglacial re-colonization of European biota. Biological Journal of the Linnean Society, 68, 1–2. Hewitt, G.M. (2004) Genetic consequences of climatic changes in the Quaternary. Philosophical Transactions of the Royal Society B: Biological Sciences, 359, 183–195. Huffman, G.J., Adler, R.F., Arkin, P.A., Chang, A., Ferraro, R., Gruber, A., Janowiak, J., Joyce, R.J., McNab, A., Rudolf, B., Schneider, U. & Xie, P. (1996) The Global Precipitation Climatology Project (GPCP) combined precipitation data set. Bulletin of the American Meteorological Society, 78, 5–20. Huntley, B. (1990) Dissimilarity mapping between fossil and contemporary pollen spea in Europe for the past 13,000 years. Quaternary Research, 33, 360–376. Huntley, B. & Birks, H.J.B. (1983) An atlas of past and present pollen maps for Europe: 0–13,000 yrs ago. Cambridge University Press, Cambridge. Jalut, G., Carrión, J., David, F., Gonzalez Sampériz, P., Sanchez Goñi, M F., Tonkov, S. & Willis, K. (2005) The vegetation around the Mediterranean basin during the last Glacial Maximum and the Holocene climatic optimum. The Mediterranean basin: the last two climatic extremes (ed. by N. Petit-Maire and B. Vrielinck), pp. 37–57. ANDRA, MMSH, Châtenay-Malabry. Kaplan, J.O., Bigelow, N.H., Prentice, I.C., Harrison, S.P., Bartlein, P.J., Christensen, T.R., Cramer, W., Matveyeva, N.V., McGuire, A.D., Murray, D.F., Razzhivin, V.Y., Smith, B., Walker, D.A., Anderson, P.M., Andreev, A.A., Brubaker, L.B., Edwards, M.E. & Lozhkin, A.V. (2003) Climate change and Arctic ecosystems: 2. Modeling, paleodata-model comparisons, and future projects. Journal of Geophysical Research, 108(D19), 8171. King, R.A. & Ferris, C. (1998) Chloroplast DNA phylogeography of Alnus glutinosa (L.) Gaertn. Molecular Ecology, 7, 1151–1161. Krebs, P., Conedera, M., Pradella, M., Torriani, D., Felber, M. & Tinner, W. (2004) Quaternary refugia of the sweet chestnut (Castanea sativa Mill.): an extended palynological approach. Vegetation History and Archaeobotany, 13, 145– 160. Laurance, W.F., Lovejoy, T.E., Vasconcelos, H.L., Bruna, E.M., Didham, R.K., Stouffer, P.C., Gascon, C., Bierregaard, R.O., Laurance, S.G. & Samapaio, E. (2002) Ecosystem decay of Amazonian forest fragments: a 22-year investigation. Conservation Biology, 16, 605–618. Laurent, J.-M., Bar-Hen, A., François, L., Ghislain, M. & Cheddadi, R. (2004) Refining vegetation simulation models: From plant functional types to bioclimatic affinity groups of plants. Journal of Vegetation Science, 15, 739–746. Leroy, S.A.G. (2007) Progress in palynology of the GelasianCalabrian Stages in Europe: ten messages. Revue de Micropaléontologie, doi: 10.1016/j.revmic.2006.08.001. Leroy, S.A.G. & Roiron, P. (1996) Final Pliocene macro and micro floras of the paleovalley of Bernasso (Escandorgue, Journal of Biogeography 34, 2115–2128 ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd Summer-green trees in the LGM France). Review of Palaeobotany and Palynology, 94, 295– 328. Leroy, S.A.G., Ambert, P. & Suc, J-P. (1994) Pollen record of the Saint-Macaire maar (Hérault, southern France): a Lower Pleistocene glacial phase in the Languedoc coastal plain. Review of Palaeobotany and Palynology, 80, 149–157. Leroy, S.A.G., Giralt, S., Francus, P. & Seret, G. (1996) The high sensitivity of the palynological record in the Vico maar lacustrine sequence (Latium, Italy) highlights the climatic gradient through Europe for the last 90 ka. Quaternary Science Reviews, 16, 189–201. Lohmann, G. & Lorenz, S. (2000) On the hydrological cycle under paleoclimatic conditions as derived from AGCM simulations. Journal of Geophysical Research, 105(D13), 17417–17436. Lorenz, S., Grieger, B., Helbig, P. & Herterich, K. (1996) Investigating the sensitivity of the Atmospheric Global circulation Model ECHAM3 to paleoclimatic boundary conditions. Geologische Rundschau, 85, 513–524. Montoya, D., Rodrı́guez, M.A., Zavala, M.A. & Hawkins, B.A. (2007) Contemporary richness of holarctic trees and the historical pattern of glacial retreat. Ecography, 30, 173–182. Persson, H., Widen, B., Andersson, S. & Svensson, L. (2004) Allozyme diversity and genetic structure of marginal and central populations of Corylus avellana L. (Betulaceae) in Europe. Plant Systematics and Evolution, 244, 157–179. Petit, R.J., Brewer, S., Bordács, S. et al. (2002) Identification of refugia and post-glacial colonisation routes of European white oaks based on chloroplast DNA and fossil pollen evidence. Forest Ecology and Management, 156, 49–74. Petit, RJ., Hampe, A. & Cheddadi, R. (2005) Climate changes and tree phylogeography in the Mediterranean. Taxon, 54, 877–885. Peyron, O., Guiot, J., Cheddadi, R., Tarasov, P., Reille, M., de Beaulieu, J.-L., Bottema, S. & Andrieu, V. (1998) Climatic reconstruction in Europe for 18,000 yr BP from pollen data. Quaternary Research, 49, 183–196. Pons, A. & Reille, M. (1988) The Holocene and Upper Pleistocene pollen record from Padul (Granada, Spain): a new study. Palaeogeography, Palaeoclimatology, Palaeoecology, 66, 243–263. Prentice, C., Guiot, J., Huntley, B., Jolly, D. & Cheddadi, R. (1996) Reconstructing biomes from palaeoecological data: a general method and its application to European pollen data at 0 and 6Ka. Climate Dynamics, 12, 185–194. Roeckner, E., Arpe, K., Bengtsson, L., Brinkop, S., Dümenil, L., Esch, M., Kirk, E., Lunkeit, F., Ponater, M., Rockel, B., Sausen, R., Schlese, U., Schubert, S. & Windelband, M. (1992) Simulation of the present-day climate with the ECHAM model: impact of model physics and resolution. MaxPlanck-Institut für Meteorologie Report 93. Max-PlanckInstitut für Meteorologie, Hamburg. Sarnthein, M., Gersonde, R., Niebler, S., Pflaumann, U., Spielhagen, R., Thiede, J., Wefer, G. & Weinelt, M. (2003) Overview of Glacial Atlantic Ocean Mapping (GLAMAP 2000). Paleoceanography, 18, 1030. Schmitt, T. & Hewitt, G. (2004) The genetic pattern of population threat and loss: a case study of butterflies. Molecular Ecology, 13, 21–31. Seddon, JM, Santucci, F., Reeve, N.J. & Hewitt, G.M. (2001) DNA footprints of European hedgehogs, Erinaceus europaeus and E. concolor: Pleistocene refugia, postglacial expansion and colonization routes. Molecular Ecology, 10, 2187–2198. Sercelj, A. (1962) Razvoj Würmske in Holocenske gozdne vegetacije v Sloveniji. Rasprave SAZU, IV, 363–418. Simmons, A.J. & Gibson, J.K. (2000) The ERA-40 project plan, ERA-40 Project Report Series1. ECMWF, Shinfield Park, Reading. Soranzo, N., Alia, R., Provan, J. & Powell, W. (2000) Postglacial history of European Pinus sylvestris populations revealed by patterns of mitochondrial variation at sequence-tagged site loci. Molecular Ecology, 9, 1205– 1211. Svenning, J.-C. (2003) Deterministic Plio-Pleistocene extinctions in the European cool-temperate tree flora. Ecology Letters, 6, 646–653. Tarasov, P.E., Peyron, O., Guiot, J., Brewer, S., Volkova, V.S., Bezusko, L.G., Dorofeyuk, N.I., Kvavadze, E.V., Osipova, I.M. & Panova, N.K. (1999) Last Glacial Maximum climate of the former Soviet Union and Mongolia reconstructed from pollen and plant macrofossil data. Climate Dynamics, 15, 227–240. Tsereteli, L.D., Klopotovskaya, N.B. & Kurenkova, E.L. (1982) Mnogosloinaya arheologicheskaya stoyanka Apiancha (Abkhazia). Chetvertichnaya sistema Gruzii (ed. by A.L. Tsagareli), pp. 198–212. Metsniereba, Tbilisi (in Russian). Tzedakis, P.C. (1994) Vegetation change through glacialinterglacial cycles; a long pollen sequence perspective. Philosophical Transactions of the Royal Society B: Biological Sciences, 345, 403–432. Uzquiano, P. (1992) L’Homme et le bois au Paléolithique en région Cantabrique, Espagne. Exemples d’Altamira et d’El Buxu. Bulletin de la Société Botanique de France, 139, 361– 372. Watts, W.A., Allen, J.R.M., Huntley, B. & Fritz, S.C. (1996) Vegetation history and climate of the last 15,000 years at Lagho di Monticchio, southern Italy. Quaternary Science Reviews, 15, 113–132. Willis, K.J. & van Andel, T. (2004) Trees or no trees? The fullglacial environments of central and eastern Europe Quaternary Science Reviews, 23, 2369–2387. Ziska, L.H. & Caulfield, F.A. (2000) Rising CO2 and pollen production of common ragweed (Ambrosia artemisiifolia), a known allergy-inducing species: implications for public health. Australian Journal of Plant Physiology, 27, 893–898. Journal of Biogeography 34, 2115–2128 ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd 2127 S. A. G. Leroy and K. Arpe BIOSKETCHES Suzanne Leroy, Professor at Brunel University, London, UK, is a palynologist. She has analysed sequences covering the Pliocene to the present in north-west Africa, Europe and south-west Asia. One of her interests is vegetation succession at the scale of Milankovitch forcing. Klaus Arpe is a retired meteorologist from the Max-Planck Institute of Meteorology, Hamburg. His interests are the hydrological cycle of the atmosphere and diagnostic studies of atmospheric processes using observational and model data. Editor: Bradford Hawkins 2128 Journal of Biogeography 34, 2115–2128 ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd