This work has been supported by the European Community - Marie Curie Grant (MOIF-CT-2004-509918). The author thanks Tom Meijer, who provided a more general reference list about European Quaternary and Tertiary malacology (Meijer, 2006) that was enriched by the bibliographical compilation done during this work. The author also thanks Jean-Pierre Lautridou, Jean-Laurent Monnier, and Pierre Antoine for stratigraphic information concerning several French sites, Pierre Antoine once again who helped with the drawing of Figure 2. The manuscript benefited the careful reviews of Pierre Antoine and Nicole Limondin-Lozouet, and of William McCoy for the English. Finally, I thank all librarians of Bayreuth University who contributed to the efficiency and the quickness of the bibliographical compilation.

1 - Introduction

1For the last twenty years, the reconstruction of European palaeoenvironmental changes of the last climatic cycle rose in importance, especially since the first characterisation of Heinrich Events in the North Atlantic Ocean sediments (Heinrich, 1988) and Dansgaard Oeschger cycles in Greenland ice‑cores (Dansgaard et al., 1993). The linkage of these phenomena with millennial-timescale global climatic changes led to the study of numerous high-resolution records in polar, oceanic and continental archives (See reviews by Voelker (2002) and Rohling et al. (2003) to understand their dynamics and mechanisms and to reconstruct associated millennial-timescale environmental changes on the continents. Palaeoenvironmental reconstructions or paleodatasets are required (1) by modellers to test the robustness of both functioning and outputs of climatic and vegetation models for conditions different than modern ones, like during the Last Glacial Maximum, before attempting the simulation of future changes (Prentice et al., 1998; Alfano et al., 2003; Huntley et al., 2003; Jost et al., 2005; Kageyama et al., 2006; Braconnot et al., 2007a; Braconnot et al., 2007b; Hargreaves et al., 2007; Roche et al., 2007; Wu et al., 2007), and (2) by archaeologists to study the distribution dynamics of Upper Pleistocene human populations on the continent (Bocquet-Appel & Demars, 2000a, 2000b; Demars, 2002; Bocquet-Appel et al., 2005; Demars, 2006).

2The European Great Plain is particularly interesting, as it was, during the last glaciation, a transition area between permanently occupied southern regions and almost unoccupied northern territories. In western Europe, the temporary presence of human settlements has been noted by compiling and dating open air sites (Haesaerts, 1978; Otte, 1979; Monnier, 1980; Otte, 1984; Haesaerts, 1994; Locht, 2004; Floss, 2006; Lautridou & Cliquet, 2006; Otte & Noiret, 2006). It suggests that human populations might have been forced by inhospitable conditions to move temporarily southwards, despite their ability to adapt to harsh climates and environments. However, the relative sparsity of large mammals (Antoine, 1991), and thus of game, suggests a poor vegetation cover and might also explain the difficult subsistence of human populations, especially during the Weichselian Upper Pleniglacial (Antoine et al., 2003).

3Most palaeoenvironmental reconstructions in Europe come from palynological sequences. Indeed, compared with other biological environmental proxies, i.e. molluscs, ostracods, diatoms, insects, small and large mammals, phytoliths and plant macro-fossils, pollen is preserved in various deposits, and often provides long, continuous or sub-continuous records (See review studies by Guiot (1993), Watts et al. (1996), Peyron et al. (1998), Allen & Huntley (2000), Caspers & Freund (2001), Guiter et al. (2003), Müller et al. (2003), Preusser (2004) and Sánchez Goñi (2006)). Concerning the last climatic cycle, many palynological sequences precisely recorded vegetation changes during the Eemian interglacial. On the contrary, pollen records of the last glaciation are often incomplete or show a lower resolution, especially for the Weichselian Upper Pleniglacial (~ Oxygen Isotope Stage 2 (OIS 2) (tab. 1). Most of them are also located outside the former periglacial domain of the European Great Plain, where eolian sedimentation was dominant (fig. 1). The same limitations affect the few available isotopic records from Belgian and French speleothems (Quinif et al., 1994; Genty et al., 2003).

4In Benelux countries, England and the Alpine area, palynological records are discontinuous and come from organic levels occurring in otherwise pollen-sterile deposits (fig. 1; tab. 1). The reconstruction of the regional vegetation evolution was based on pollen samples dated with ¹⁴C and reported on a chronological scale (Zagwin, 1961, 1974; Behre & van der Plicht, 1992). However, some gaps remain and the chronology of environmental changes is still approximate in levels older than the technical limits of the ¹⁴C dating method (Allen et al., 2000; Müller et al., 2003). In central France and in the Mediterranean domain, long and almost complete records have been obtained due to the nearly continuous sedimentation of drilled sites (seas, lakes and peat-bogs) (fig. 1; tab. 1). However, only the Füramoos record and several others from the Mediterranean domain show well-characterised millennialtimescale changes correlated with global climatic fluctuations recorded in Greenland ice-cores (Allen et al., 1999, 2000; Sánchez Goñi et al., 2000, 2002; Müller et al., 2003; Sánchez Goñi, 2006). In loess deposits, between both groups, pollen-based palaeoenvironmental reconstructions are limited or biased by both quantity and reliability of preserved material (Zelikson, 1995; Tomescu, 2000). As most other biological proxies mentioned above rarely provide continuous records throughout loess sequences, sedimentological, pedological and mineralogical indicators are preferentially used to reconstruct palaeoenvironmental changes (Huijzer & Vandenberghe, 1998).

5Consequently, after the successful characterisation of millennial-timescale environmental changes in the Nussloch loess section (Rhine Valley, Germany) based on the study of the terrestrial mollusc fauna (Moine et al., 2002; Moine, 2003; Moine et al., 2005) in the frame of a multi-proxy approach (Antoine et al., 2001; Hatté et al., 2001; Rousseau et al., 2002; Hatté & Guiot, 2005), the aim is now to evaluate whether or not molluscan records may fill in the lack of palaeoenvironmental data and allow the analysis of rapid environmental changes during the last glaciation in the loess domain. A first compilation of molluscan records of the last glaciation was performed by Rousseau et al. (1990) to strengthen the palaeoclimatic interpretation of a comparison between long loess sequences and associated molluscan records of Achenheim (France) and 0ervený Kopec (Czech Republic). The goal of the present work is to create a dataset of Weichselian Upper Pleniglacial molluscan records from Western Europe, as this time span is characterised by the thickest loess deposits (Haesaerts, 1985; Frechen, 1999; Antoine et al., 2001; Frechen et al., 2003). This compilation and appraisal of the data have thus been performed to evaluate the potential of available molluscan records.

2 - West-european chronostratigraphic framework of the molluscan data

6This is a compilation of about one hundred publications from the last 120 years dealing with Quaternary malacology in loessic environments, or including malacological analyses.

7The quality of molluscan data relies on good identifications but also on the precision of their chronostratigraphic framework. In this study, I conduct a detailed checking of the Weichselian Upper Pleniglacial age of selected samples on the basis of absolute datings and stratigraphy. All ages are given as originally mentioned in the quoted publications.

8In the chosen reference sequence, i.e. Nussloch, the Weichselian Upper Pleniglacial is represented by about 10 m of loess alternating with gleys (tundra soils) deposited above the Lohne soil (Antoine et al., 2001). The Lohne soil marks the end of the Middle Pleniglacial in loess deposits from South Germany (Frechen, 1999), and has been dated at 27,2-31,3 ka cor. BP (top) and 28,4-31,4 ka cor. BP (base) by ¹⁴C dating (Hatté et al., 2001) and at 34 ± 3,2 ka BP (top) and 33,1 ± 5,2 ka BP (base) by OSL (Lang et al., 2003). Antoine et al. (2001) thus proposed to correlate the Lohne soil with both the Denekamp and GRIP IS-8 interstadials on the basis of absolute datings. The correlation between the Nussloch loess/gley alternation and GRIP interstadials IS-2 to IS-8 has been strengthened by the correlation between the loess grain size index in Nussloch and the dust concentration in the GRIP ice-core (Rousseau et al., 2002) and by the synchronicity between the molluscan abundance in Nussloch and δ¹⁸O record in the GRIP ice-core (Moine, 2003).

9In northwestern France, the main units constituting the Weichselian Upper Pleniglacial stratigraphy are from the base to the top: the Mesnil‑Esnard/Bonsecourt soil, the Lower Recent Loess, the Nagelbeek horizon and the Upper Recent Loess (Lautridou, 1985; Antoine et al., 1998). They overlie the Saint-Acheul soil complex, which represents the Weichselian Middle Pleniglacial in this area. According to the framework of stratigraphic correlations across northwestern Europe (fig. 2), the Lohne soil would be synchronous with the Les Vaux soil dated at 33050 ± 850 yr BP in Wezembeek Oppen (Belgium) (Haesaerts, 1985) and with the upper part of the Saint-Acheul soil complex (Antoine et al., 2001). The Nagelbeek horizon yields ¹⁴C ages of 22,270 ± 380 years BP in Kesselt (Belgium), and 22,190 ± 130 years BP in Lixhe (Belgium) (Haesaerts, 1985), and a TL age of 24 ± 2,7 ka BP in Achenheim (France) (Rousseau et al., 1998). The top of the Saint Acheul soil complex yields a TL age of 35,45 ± 3,7 ka BP in Villers‑Adam (France) (Locht et al., 2003). In Achenheim, a TL age of 34,1 ± 4,3 ka BP has also been obtained 0.7 m above a level of large ice‑wedges (Rousseau et al., 1998) correlated with the Middle/Upper Pleniglacial transition (Rousseau & Puisségur, 1990). Consequently, I selected all molluscan records located above the Saint‑Acheul, Les Vaux and Lohne soils, or which have an age between ca. 34‑35 and 1516 ka BP, i.e. the age of the youngest Weichselian Upper Pleniglacial loess (Frechen, 1999).

3 - Results and discussion

10The compilation presented here (tab. 2; fig. 3) is representative of available well-dated molluscan data for the Weichselian Upper Pleniglacial. Weichselian Lower and Middle Pleniglacial samples have been removed. Weichselian/Würmian samples from old studies, without more precise chronological and stratigraphic information have also been removed. Except for samples analyzed by Mazenot and by Dubar in several studies performed in the Rhône Valley and in southeastern France, many of the removed sites were located in the same geographical areas than selected sites. Furthermore, the removed data often include only a single sample, which thus reduces the importance of their absence.

11The geographical distribution of selected sites is mainly dependent on the position of both natural exposures and brickearth quarries exploiting loess, which are often located along large fluvial systems, where deposits are particularly thick as they are closed to

3.1 - Age validation of selected molluscan records

121: Port-Morvan (Brittany, France) (Monnier & Van Vliet-Lanoë, 1986)

13The molluscan samples come from level 1 of the Port-Morvan sequence (Monnier & Van Vliet-Lanoë, 1986), which became later level 3 of the Sables‑d’Or–Les-Pins formation in the regional stratigraphy (Monnier & Bigot, 1987). The uppermost calcareous level of this formation, close to level 3 of Port Morvan, yields a TL age of 18,2 ± 2,5 ka BP in Nantois (France) (Loyer et al., 1995).

142: Ot-Ar-Villiec (Brittany, France) (Monnier, 1980)

15As in Port-Morvan, the molluscan sample comes from level 3 of the Sables‑d’Or‑Les-Pins formation (Monnier, pers. comm.).

163: Sangatte (Nord Pas-de-Calais, France) (Antoine, 1989)

17A Weichselian age was attributed to the molluscan sample of unit X on stratigraphic criteria (Antoine, 1989). I confirm this age with the comparison of later amino acid analyses (Bates, 1993) with those of the database of Moore (2002). According to a recent review of the regional stratigraphy (Antoine et al., 1998), this molluscan sample probably belongs to the Upper Pleniglacial (Antoine, pers. comm.).

184: Achenheim (Alsace, France) (Puisségur, 1978; Rousseau, 1987)

19According to the correlation of the Achenheim loess sequence (Heim et al., 1982) with marine isotopic stages (Rousseau & Puisségur, 1990), I selected all samples located above the level of large ice-wedges correlated with the Middle/Upper Pleniglacial transition. A focus on the last glaciation is presented in Locht et al. (2003). The molluscan data are those of Rousseau (1987), which includes data from Puisségur (1978).

205: Schiltigheim (Alsace, France) (Puisségur, 1976)

21According to Wernert (1957) who helped with the molluscan sampling, and to most of the people who worked in this area, this sequence was deposited during the Würm (Puisségur, 1965). As the only other chronological information is the discovery of Upper Palaeolithic artefacts at the base of the Upper Younger Loess in unit 11 (Puisségur, 1976), I selected only two samples in the overlying loess layer.

226: Reculver (South-East, England) (Preece, 1990)

23Luminescence dating performed on the loess that yields the molluscan sample provide an age range of 17-23 ka BP (Parks & Rendell, 1988; Preece, 1990).

247: Portelet (Channel Islands) (Rousseau & Keen, 1989)

25According to the correlation established between the geliflucted soil overlying the molluscan samples and the Nagelbeek horizon recorded in Normandy, Rousseau and Keen (1989) suggested that the age of the molluscan assemblages should be close to the Middle/Upper Weichselian boundary. The latest synthesis of the stratigraphy of Jersey Island was established after the re-examination of two long sedimentary records (Keen et al., 1996). Its comparison with the stratigraphy of Normandy supports the attribution of the loess formation that yielded molluscan assemblages to the Weichselian Upper Pleniglacial.

268: Sandy (East, England) (Gao et al., 1998)

27The two samples come from lenses of silt and organic deposits located in a formation of sand and gravels. A ¹⁴C analysis performed on plant detritus yielded an age of 29,250 +460 –420 years BP for the lowest lens (Gao et al., 1998). The Weichselian Upper Pleniglacial age of this sequence is also supported by stratigraphic correlations and biostratigraphical indicators (Gao et al., 1998).

289: Woolpack Farm (East, England) (Gao et al., 2000)

29The molluscan sample comes from a silt level bracketed by two gravel formations, i.e. Hemingford Member and St. Ives Member (Gao et al., 2000), which are respectively dated at 43-18 ka BP and 18-10 ka BP by ¹⁴C (Gao, 1997).

3010: Halling (South-East, England) (Kerney, 1971)

31According to Kerney (1971), molluscan assemblages from Halling cannot have an Early Weichselian age as they show a “purely glacial” character and lack of thermophilous species having survived from the previous interglacial period unlike other records from the Cambridgeshire area. An age between 13 and 28 ka BP does not seem possible to the author as the poverty of organic remains in northwestern Europe for this interval suggest harsh climatic and environmental conditions that does not fit with the development of enough vegetation to support such molluscan populations. Hence, he proposed an age between 55 and 28 ka BP and argued that the central abundance increase suggests an interstadial period similar in magnitude to that of Denekamp or Hengelo. However, it is not possible to give a more precise age to these samples due to the lack of absolute ages and to the disappearance of the site. Consequently, as it is the most precise record of a part of the last glaciation available in England and that at least the upper part of it may be of Weichselian Upper Pleniglacial age, all samples have been selected. Moreover, the composition of all samples being almost the same, the selection of the totality or just a part of them will not strongly influence eventual statistic calculation or interpretations.

3211: Petit Saule/Villiers-Adam; 12: Chamesson/Villiers‑Adam (Ile de France, France) (Limondin-Lozouet & Gauthier, 2003)

33According to the precise knowledge of the regional stratigraphic framework (Antoine et al., 1998) and to luminescence analyses that yield TL ages of 35,45 ± 3,7 ka BP at the top of the Saint‑Acheul soil complex and 23,5 ± 2,1 and 23,1 ± 2,1 ka BP for the uppermost loess unit above the Nagelbeek horizon (Locht et al., 2003), I select all samples located in the loess above the Saint-Acheul soil complex.

3413: Achenheim; 14: Riedisheim; 15: Sierentz; 16: Pfastatt; 17: Eguisheim; 60: Brunstatt; 61: Wittenheim; 62: Blaesheim; 63: Mothern; 64: Schleithal (Alsace, France) (Mazenot, 1963)

35The chronostratigraphy of all these sites is based on Briquet (1930), Franc de Ferrière (1937) and Wernert (1957). Franc de Ferrière and Wernert helped Mazenot to sample and to locate samples in the exposures. Mazenot (1963) mentions that these three authors almost agree on sedimentological units that were deposited during the last glaciation. In their synthesis about the stratigraphy of the Achenheim sequence, Heim et al (1982) mention that the uppermost loess layer, named Last Loess by Wernert and Upper Younger Loess by Franc de Ferrière, represents the same stratigraphic unit and they attribute it to the Upper Pleniglacial, as well as the upper part of the Lower Younger Loess. The lower part of this latter unit represents both Lower and Middle Pleniglacial (Heim et al., 1982). Consequently, I select only samples from the Upper Younger Loess as those from the Lower Younger Loess may be of Lower or Middle Pleniglacial age.

3618: Gräselberg/Wiesbaden; 19: Hainerberg/Wiesbaden (Hesse, Germany); 20: Michelsberg; 21: Wallertheim (Rhineland-Palatinate, Germany) (Remy, 1969)

37In the Gräselberg and Hainerberg profiles, the stratigraphy was described by Semmel (1968). In both sequences of Wiesbaden, I selected samples located above the Hainerberg soil, which is the local equivalent of the Lohne soil. In Wallertheim, the stratigraphy of Remy (1969) and those of previous studies (Schmidtgen & Wagner, 1929; Fauler, 1938) does not help to identify precisely Weichselian Upper Pleniglacial deposits. In the middle of the loess unit, Remy (1969) identified a volcanic tuff that he named “Kärlicher Tuff”. However, Zöller (1995) mentioned two volcanic tuffs in the Upper Würm, i.e. the Rambacher and the Eltviller tuffs, the true Kärlicher Tuff being much older, i.e. around 396 ± 20 ka BP. Given the absence of tuff in Middle and Lower Pleniglacial deposits of Wallertheim and the position above the Lohne soil of both tuffs, I selected only four samples located around the single represented tuff. In Michelsberg, only one sample has been selected, the six others being attributed to the Late Pleniglacial (Remy, 1969).

3822 : Brühl ; 23 : Hennef in Sieg ; 24 : Bonn ; 29 : Römlinghoven ; 30 : Rodderberg ; 31 : Mehlem (North Rhine-Westphalia, Germany) ; 25: Rhens ; 26 : Remagen ; 27 : Moselweiss ; 28 : Weitersburg ; 36 : Biebelnheim ; 37 : Arzheim (Rhineland‑Palatinate, Germany) ; 32 : Schiersteinerstrasse in Wiesbaden; 33 : Kriftel quarry in Wiesbaden; 34 : Klüter quarry in Eltville ; 35 : Schick quarry in Wallertheim (Hesse, Germany); 38 : Grossostheim (Bavaria, Germany); 39 : Kesselt in Western Maastricht (Flanders, Belgium) (Remy, 1968)

39The chronostratigraphy of each site (Gullentops, 1954; Paas, 1962; Remy & Paas, 1959; Remy, 1960; Remy, 1968; Juvigné, 1996) is in agreement with modern concepts. All samples attributed to the Upper Würmian have thus been selected.

4040: Bötzingen (Baden-Württemberg, Germany) (Münzing, 1969)

41The upper loess formation was attributed to the Weichselian on a stratigraphic basis (Khodary Eissa, 1968). In the absence of more precise chronological information and according to the three-fold structure of this loess, I select only samples from the uppermost loess unit.

4241: Maisières-Canal (Wallonia, Belgium) (de Coninck, 1973)

43The stratigraphy of this site has been precisely described in Haesaerts (1978). A ¹⁴C age of 27,965 years BP has been obtained for the archaeological level including molluscan sample DC, and a ¹⁴C age of 30,780 years BP has been obtained for the basal humic soil of the sequence. According to Haesaerts (1985), this sequence is younger than the Les Vaux soil. All samples have thus been selected.

4442: Am Biedensteg/Bad Wildungen (Hesse, Germany) (Jacobshagen, 1963)

45The whole fauna indicates a Weichselian age for this two-fold loess sequence. In the absence of more precise chronological information, I selected only two samples from the uppermost loess unit attributed to the Upper Weichselian on the basis of the stratigraphy.

4643: Bad Kösen (Thuringia, Germany) (Mania & Altermann, 1970)

47The upper loess formation has been attributed to the Weichselian on the basis of a synthesis of the regional stratigraphy (Ruske & Wünsche, 1961). The occurrence of a weak soil formation distinguishes a lower loess unit with stratified and reworked structures and an upper eolian loess unit slightly or not perturbed (Mania & Altermann, 1970). A parallel made by the authors between this weak soil and the Stillfried B soil synchronous with the Lohne soil led me to select only samples from the upper loess unit.

4846: Böckingen/Heilbronn (Baden-Württemberg, Germany) (Rähle, 2004)

49According to the precise stratigraphy established by Bibus (2002), I selected all samples located above the Lohne soil.

5047: Schwalbenberg in Remagen (Rhineland-Palatinate, Germany) (Schiermeyer, 2000)

51The stratigraphic framework was established by Schirmer (1995). Schirmer (2000) then proposed a correlation between the Sinzig soil complex and interstadials IS-5 to IS-8 of Greenland ice-cores. This correlation is supported by ¹⁴C ages from the upper part of the Sinzig complex and which yielded ages of 27,890 ± 440 and 28,080 ± 530 years BP (App et al., 1987). Furthermore, unpublished TL ages from the Sinzig complex are around 30 ka BP (Schirmer, 2000). The upper part of the Sinzig soil complex also provided archaeological artefacts attributed to the Middle-to‑Upper‑Palaeolithic transition (Schirmer, 2000). Schirmer (2000) stated that the correlation between one of the three soils of the Sinzing complex with the Lohne soil upstream in the Rhine valley cannot be directly established. However, he mentioned that the lowest one, i.e. Sinzig soil 1, is also the most developed. According to the correlation established between the Lohne soil in Nussloch and the interstade 8 in GRIP (Antoine et al., 2001; Moine, 2003; Rousseau et al., 2002), the Sinzig soil 1 would be synchronous with the Lohne soil. Consequently, I selected molluscan samples located above the Sinzig soil 1.

5248: Nauberg; 49: Jahna-Aue; 50: Klosterbuch; 51: Golzern (Saxony, Germany) (Fuhrmann, 1973)

53In these Late Weichselian-Holocene sequences, the base of molluscan records show a continuous transition from glacial to interglacial conditions characterised by the progressive appearance of new species. The absence of significant gaps suggests that underlying deposits belong to the Upper Weichselian rather than to a previous glaciation or another glacial period. Consequently, I selected only the basal sample in each of these sequences.

5452: Nördlingen (Bavaria, Germany) (Hollaus, 1969)

55According to the stratigraphic framework of Würmian deposits in the area (Brunnacker, 1957), I selected only samples from the uppermost loess units in absence of other chronological information.

5653: Gönnersdorf, Neuwied (Rhineland-Palatinate, Germany) (Brunnacker et al., 1978)

57At this archaeological site, selected samples bracket the Eltviller Tuff, which was dated in Nussloch at 19,5 ± 2,3 and 19,2 ± 1,7 ka BP by OSL (Lang et al., 2003) and at 21,131-22,258 years cor. BP by ¹⁴C (Hatté et al., 2001).

5854: Wischlburg; 55: Schambach; 56: Niederwinkling; 57: Strasskirchen; 58: Pfelling (Bavaria, Germany) (Brunnacker & Brunnacker, 1956)

59The regional stratigraphy and the different facies of the Würmian loess in the area are well described by Brunnacker (1954; 1956). Full glacial deposits are generally made of two loess units separated by a tundra soil. In absence of more precise chronological information for this tundra soil, I selected only samples from the upper loess.

6059: Beckford (South-West, England) (Briggs etal., 1975)

61Selected molluscan samples come from a black silt layer including wood remains that yielded a ¹⁴C age of 27,650 ± 250 years BP.

6265: Wallstadt; 66: Grenzhof (Baden-Württemberg, Germany) (Münzing, 1985)

63In Grenzhof, wood remains located below the two upper most molluscan samples yielded ¹⁴C ages between 37 and 45 ka BP. I selected only the uppermost sample to be sure of its Weichselian Upper Pleniglacial age. In Wallstadt, I selected samples located between ¹⁴C ages of 27,080 ± 510 and 30,860 ± 500 years BP obtained on wood remains (Löscher et al., 1980).

6445: Plinz; 67: Mücheln; 68: Ascherslebener See (SaxonyAnhalt, Germany) (Mania & Stechemesser, 1969)

65The basal sample of the Late Weichselian-Holocene record of Ascherslebener See reflects full glacial conditions. According to ¹⁴C analyses (Mania & Stechemesser, 1970), it probably belongs to the Weichselian Upper Pleniglacial, and has thus been selected. In absence of other chronological information, the presence of an ice-cap in the region of Plinz and Mücheln during the Saalian glaciation (Grahmann, 1932) suggests a Weichselian age for loess formations underlying these Late Weichselian-Holocene sequences. As in Ascherslebener See, I thus selected the basal sample, characterizing full glacial conditions.

6644: Krumpa; 69: Möckerling (Saxony-Anhalt, Germany) (Mania et al., 1993)

67For these two Late Weichselian-Holocene sequences located in the same valley as Plinz and Mücheln, only the presence of the ice cap in the area during the Saalian glaciation suggests that underlying loess has a Weichselian Upper Pleniglacial age. I thus selected the basal sample of each sequence.

6870: Nussloch (Baden-Württemberg, Germany) (Moine, 2003)

69The Upper Weichselian molluscan record in Nussloch (Moine et al., submitted) is clearly identified at this site on the basis of precise stratigraphic data and numerous TL, OSL and ¹⁴C ages (Antoine et al., 2001; Hatté et al., 2001; Lang et al., 2003) as previously described in Data and Methods.

7071: Ascherslebener See (Saxony-Anhalt, Germany) (Mania, 1967)

71In this sequence, ¹⁴C ages on wood remains are : 11,300 ± 280, 10,940  ± 190 and 10,570 ± 180 years BP from the basis to the top of unit VII, 23,050 ± 750 years BP; for unit V, and 30,550 ± 2,600 years BP for unit IV (Mania & Stechemesser, 1970). According to these ages, I selected only samples from units V and VI that should belong to the Weichselian Upper Pleniglacial.

7272: Tancarville; 73: Gonfreville (Upper Normandy, France) (Lautridou, 1968, 1985)

73The single selected sample of each site is located above the Nagelbeek horizon.

7474: Rouelles (Lower Normandy, France) (Lechevalier, 1975)

75The single sample mentioned by Lechevalier (1975) belongs to the Lower Younger Loess according to Lautridou (1985).

7675: Mesnil-Esnard; 76: Roumare; 77: Epouville (Upper Normandy, France); 78 : Bréau ; 79 : Sergines (Burgundy, France) (Lautridou, 1985)

77At Mesnil-Esnard, Roumare and Epouville, the single selected sample or each site is located in and above the Mesnil-Esnard soil. At Bréau and Sergines, I selected samples from the Recent Loess.

7880: Luc-sur-Mer (Lower Normandy, France) (Coutard et al., 1970)

79The selected sample is located in a level of calcareous silt equivalent to the Lower Recent Loess (Lautridou, pers. comm).

3.2 - Representativeness and potential of molluscan data

80The second map (fig. 4) shows the distribution of molluscan records and the number of samples per site. The structure of the 80-site database (number of site number of samples per site) is as follows: 46-1; 14-2; 63; 3-4; 2-5; 1-7; 7-10/30; 1-96. With only eight sites including more than ten samples and only two 4-m thick records, the potential of terrestrial mollusc fauna is definitely underexploited in the reconstruction of millennial-timescale palaeoenvironmental changes. Indeed, according to both extent and thickness of loess deposits (Grahmann, 1932; Fink, 1969; Catt, 1985, Haesaerts, 1985; Frechen et al., 2003; Haase et al., 2007), if 2-to-9-m-thick loess sequences are sampled continuously every 5 cm for the 35-15 ka BP time span, molluscan records would have an average time resolution of 500-to-111-years per sample, which is enough to characterise the impact of millennial-timescale environmental changes. Among all compiled Weichselian Upper Pleniglacial records, only the Nussloch record fits these requirements. The study of more sequences is thus needed to fill in geographical gaps in loess areas (fig. 5) and to develop a network of high resolution molluscan sequences to tackle millennial-timescale environmental changes in the European Great Plain. For the Lower and Middle Weichselian Pleniglacial, even fewer molluscan data are presently available or can be precisely dated, but several loess sequences could also be sampled for mollusc shells at high-resolution, such as at Mainz-Weisenau (Bibus et al., 2002) and Nussloch (Moine et al., 2005).

3.3 - Palaeoenvironmental information

81As a first approximation, the Weichselian Upper Pleniglacial was considered as a single entity without subdivisions due to the heterogeneity of available data for this time interval. The composition of West-European molluscan populations shows a distinct opposition between a “western domain” surrounding the Channel and extending to eastern Belgium, which includes only one to five taxa of the following group (Columella columella, Pupilla muscorum, Succinea oblonga, Trichia hispida, slugs), and a “central domain” located around the Rhine Valley and extending to the East of Germany, which includes a larger pool of 39 taxa (fig. 6).

82In a study of the relationship between loess formation and associated molluscan populations from glacial times, including less data but spread all over the Great Plain, Ložek (1965) mentioned a general homogeneity of glacial malacofaunas from France to Ukraine, but even then noted regional differences in their composition. He differentiated three types of assemblages based on taxa presence/absence and relative abundances (see Ložek (1990) for their precise description): the Columella-fauna, the Pupilla-fauna and the Helicopsis striata-fauna. In this order, they reflect a gradient from cold-humid to cold-dry conditions. The composition of the malacofauna can shift gradually in space or time from one type of fauna to another depending on regional and continental climatic and environmental changes.

83According to assemblage types defined by Ložek (1990), the “western domain” is characterised by poor Columella-faunas and poor malacofaunas including only core species of loess assemblages, whereas Columella-faunas, Pupilla-faunas, and intermediate transitional faunas, mainly occur in the “central domain”. The more contrasted topography of the “central domain” has probably benefited the persistence of diversified and more vegetated microenvironments and of a richer diversity of taxa besides the core species Pupilla muscorum, Succinea oblonga and Trichia hispida, which are common in the whole studied area, and live now over most of the temperate zone of Eurasia.

84Ložek characterised several large biogeographical domains at the scale of Europe, and at the regional scale for Czech Republic, Slovakia and surroundings areas. However, geographical taxonomic distributions are rarely discussed for given time intervals, except for several taxa presenting a particular biostratigraphical interest. Ložek (1972) only mentioned a certain resemblance between poor malacofaunas from France and those from humid loess areas in Central Europe. Later, Rousseau et al. (1990) analysed a set of molluscan data from the last glaciation for Western and Central Europe, including newer data for France, and compared them with both periglacial features of loess deposits and climatic modelling results. They concluded (1) that the “western domain” was characterised by thick permafrost and increased cloudiness, atmospheric moisture and precipitation, and (2) that the composition of molluscan populations reflects a climatic gradient between oceanic conditions around the Channel and continental conditions in Central Europe.

85First, the four species identified for Weichselian Upper Pleniglacial loess deposits of the “western domain” are characteristic of cold, humid and poorly vegetated environments with either a mosaic aspect or a seasonal shift of environmental conditions, according to slightly different ecological preferences (compiled from several atlases) and optimal life temperature ranges (Sümegi, 2005) of the taxa: Columella columella (cold, humid, vegetated, 5-15^oC); Pupilla muscorum (open, dry-humid, sunny, 10-22^oC); Succinea oblonga (open, humid-swampy, 13-19^oC); and Trichia hispida (everywhere except too dry, 10-20^oC). Nevertheless, optimal life temperature ranges from Russian material and data must be considered with caution since geographical distributions of species in the former USSR (Likharev & Rammel’meier, 1962) are still not well constrained due to sampling gaps and to the use, for a long time, of a different taxonomic system than used in Europe. Optimal life temperature ranges given in Sümegi (2005) are originally based on Russian material identified with the European taxonomic system (Sümegi, 1989).

86Second, according to Murton & Lautridou (2003), the mapping of permafrost distribution during the Weichselian is limited by the precise dating of periglacial structures, the climatic instability, the lack of data, and the palaeoclimatic significance of many periglacial structures. This led to the realisation of many different permafrost distribution maps in the last fifty years (Poser, 1948; Kaiser, 1960; Maarleveld, 1976; Velichko, 1982 (all four redrawn in Vandenberghe & Pissart (1993), Van Vliet-Lanoë, 1989; Huijzer & Vandenberghe, 1998; Van Vliet-Lanoë & Hallégouët, 1998; Renssen & Vandenberghe, 2003). A comparison has thus been attempted with the geographical distribution of Last Glacial Maximum ice-wedge casts, which are used as periglacial indicators of the presence of permafrost (Karte, 1987; Van Vliet-Lanoë & Hallégouët, 1998). Most of the malacological “western domain” is included within the area of highest density of icewedge casts and ice-rich loess deposits (fig. 7). The correspondence might have been even better if the site 42, located north-eastward of the Middle Rhine Valley (fig. 3), and including a lower sample with Pupilla muscorum, Succinea oblonga and slugs, and an upper sample with only Pupilla loessica, had also been considered as representative of the “western domain”. Furthermore, modern observations show that the seasonal thawing of an ice-rich permafrost causes large water releases that cannot always be used by the vegetation due to low temperature and high amounts of organic solutes (Pruitt, 1978). The thawing also disturbs the uppermost soil layer and perturbs vegetation by shifting the root zone (Shur et al., 2003). Moreover, if runoff is weak, waterlogging may take place, which limits vegetation development and enhance the thawing itself by warming the permafrost.

87The link between poor malacofaunas and the presence of permafrost in the “western domain” proposed by Rousseau et al. (1990) is thus confirmed for the Weichselian Upper Pleniglacial, as well as cold, poorly vegetated and more humid environmental conditions than in the “central domain”. Our new dataset, enriched in German molluscan records, also indicates the presence of poor molluscan samples in Bavaria (sites 53 to 58), in Saxony (sites 49 to 50), and also in the Rhine Valley, which are generally ice-poor loess areas, which poses the question of the limits between both biogeographical domains. Indeed, whereas rich malacofauna are absent from the “western domain”, poor westernlike malacofauna occurred in the “central domain”, mainly in low-elevation regions and at the northern edge of the German plateau region. On the one hand, the survival of the same most tolerant species as in the “western domain” may be due either to the predominant influence of the regionally flat topography on the loess ice content or to a locally higher loess ice content due to regional climatic conditions. Damp loesses from marginal zones of loess deposition are indeed often characterised by poor Columella-faunas, especially during the Weichselian Upper Pleniglacial (Ložek, 1990). On the other hand, these conditions may have also been temporary, but as most of the records are only made of a few samples, composition changes of the malacofauna during the Weichselian Upper

88Pleniglacial are poorly documented at many sites. Indeed, Rousseau et al. (1990) remarks that for each of the last four glaciations, molluscan populations at the loess section of Achenheim (Alsace, France) resembles either the poor malacofauna of the “western domain”, or the rich ones of the “central domain”. They concluded for the loessic domain (1) that changes in the position of the glacial climatic gradient probably accompanied changes in the extent of both Alpine and Scandinavian ice-sheets, and (2) that if the occurrence of pleniglacial malacofaunas was limited by low temperatures, their regional composition changes mainly resulted from different precipitation regimes.

89According to this result, it can be envisaged that occurrences of poor malacofauna in the “central domain” might also result from changes in the climatic gradient, and particularly from an eastward expansion of oceanic climatic conditions. Whereas Rousseau et al. (1990) highlighted such changes between successive glaciations, our future objectives will be to look for changes occurring within the last glaciation, and even, within the Weichselian Upper Pleniglacial. The position of ice-wedge casts in regional stratigraphic frameworks for the Weichselian Upper Pleniglacial also reflects temporary occurrences rather a permanent presence during this period, even in the “western domain”, which implies associated climatic changes (Lautridou & Sommé, 1981; Karte, 1987). However, even if the precise stratigraphic position of these ice-wedge casts is known, the reconstruction of the dynamics of environmental domains in response to Weichselian Upper Pleniglacial millennial-timescale climatic changes is still precluded by the quantity of data available per site. Only the Nussloch molluscan record has indeed a sufficient resolution to look for millennial-timescale environmental changes (Moine, 2003; Moine et al., 2002, 2005, accepted). Consequently, new high-resolution molluscan records are required in Western and Central Europe and must be analysed in parallel with sedimentological, lithological and periglacial features to better constrain the composition, the distribution, and the dynamics of biogeographical domains, and to determine and reconstruct the evolution of the most influential environmental and climatic factors.

White-coloured sites include taxa of the first group. Most of these sites are located around the Channel, except several ones located in the Rhine and Danube valleys, and in Eastern Germany. Grey‑coloured sites include taxa of the second group. These sites are mainly located in Germany and in the East of France, except a few ones located in the Parisian region and in Burgundy. The more contrasted topography of the central domain may explain the persistence of differentiated microenvironments and of a larger pool of taxa.

Annotations about loess ice content after Brunnacker (1958) for Franconia and Bavaria, Lieberoth (1963) for Saxony and Semmel (1968) for Hesse. The poor molluscan assemblages of the “western domain” (dashed line) are mainly located in areas of high ice content and ice-wedge density in loess (solid line).

4 - Conclusion

90The compilation of a regional database of molluscan records has been initiated to study the distribution of the vegetation cover in the European Great Plain and to fill in gaps of palynological records. This first step was focused on the Weichselian Upper Pleniglacial of Western Europe for which literature was easier to find and to use. With the selection phase accomplished, the analysis of the selected dataset shows that less than 10% of the 80 compiled molluscan records may be useful to study millennial-timescale environmental changes. According to both extent and thickness of shell-bearing loess deposits, the potential of terrestrial malacofauna as a palaeoenvironmental proxy appears to be underexploited. The extension of the molluscan records data base to Central and Eastern Europe and the analysis at high resolution of new molluscan records, as in Nussloch (Moine, 2003), will thus improve our knowledge of millennial-timescale environmental changes in the European Great Plain.

91This first analysis of the dataset already shows the existence of two domains. The first one located in northwestern Europe is characterised by poor malacofauna reflecting a cold, humid and poorly vegetated environment developed on an ice-rich permafrost in a flat lowland region, whereas the second one located in western Germany is characterised by richer malacofauna reflecting a more vegetated environment developed on an ice-poor permafrost in a topographycally more contrasted region. The future analysis of new records will allow a more precise definition of environmental domains in the European Great Plain and the study of their evolution and dynamics through time as already suggested by Ložek (1972). A better understanding of the local and regional ecological values of the taxa is also expected from such a geographical approach of European glacial malacofauna. It will be help ful for the palaeoenvironmental interpretation of molluscan records and provide new reconstructions to be discussed with LGM modelling results and the distribution of Palaeolithic human populations through comparisons with archaeological data and the remains of large mammal faunas.