We would like to thank M.-F. Diot (Centre National de la Préhistoire - UMR 5808 CNRS, Institut de Préhistoire et de Géologie du Quaternaire, Périgueux, France), B. Valadas (GEOLAB, UMR 6042 CNRS, Université de Limoges, France) and E. Roussel (Maison des Sciences de l’Homme, Clermont-Ferrand, France). We are also grateful to A. Ejarque (ICAC, Tarragona, Spain) for her valuable comments on the manuscript and also to J. Franklin (East Tennessee State University, USA) for his assistance in improving the English. This study received financial support from the CNRS, the Conseil Régional de Franche-Comté, the Crédit Agricole Centre-France and the FNADT (Fonds National d’Aménagement du Territoire).

1 - Introduction

1Recent archaeological and palaeoenvironmental studies carried out in different western and central European mountains provide a new concept of mountainous areas since they revealed significant evidence of upland human occupation and landscape management that can be traced back to the Mesolithic period (for example, in Pyrenees (Galop, 1998; Rendu, 2003; Miras et al., 2007, 2010; Palet et al., 2007; Ejarque et al., 2009, 2010; Mazier et al., 2009), in Jura (Gauthier, 2004; Richard & Ruffaldi, 2005), and in Alps (Oeggl & Wahlmüller, 1993; Moe & Hjelle, 1999; Walsh & Richer, 2006; Court-Picon, 2007)). However, such interdisciplinary studies are still scarce in the whole French Massif Central (Miras et al., 2003, 2004a; Argant & Cubizolle, 2005; Stebich et al., 2005; Prat, 2006; Pulido Avil, 2006; Jouffroy-Bapicot et al., 2007; Trément et al., 2007; Surmely et al., 2009) and do not concern the Limousin area located in the north-western part of the French Massif Central at all. Moreover, despite two isolated case studies (Diot in Allée et al., 1997; Valadas & Marambat, 1999), previous pollen analyses carried out in the Limousin were focused on Holocene vegetation history (Denèfle et al., 1980; Guenet, 1993) and provided pollen diagrams with a low spatio-temporal resolution preventing a detailed land-use history reconstruction.

2This paper presents the results of the first pollen analyses carried out in the Limousin, and more particularly in the upper part of this area called the “plateau de Millevaches”, with the following overall objectives:

• to more accurately determine the regional Holocene vegetation changes, and

• to draw the first rhythms in the human impact history from the Neolithic to the Roman period on a regional scale.

3The pollen data presented come from 4 peat sequences - one of them studied at a high temporal resolution – situated in the “plateau de Millevaches” (French Massif Central). They are also compared with one peat sequence situated at a lower altitude in the northern Limousin uplands called “Monts d’Ambazac” (fig. 1 & tab. 1).

2 - Study area

4The “plateau de Millevaches” (pdM) is a granite plateau (150 km long and 40 km wide) located in the north-west of the French Massif Central. Ranging in altitude from 700 to 900 m a.s.l., it is the highest part of the “Montagne Limousine”. This region is characterized by an oceanic-montane climate. The mean annual temperature is about 7.5°C and the mean annual rainfall varies from 1600 to 1700 mm (Valadas & Allée, 2000). The main part of the pdM lies in the montane Fagion belt with forests dominated by Fagus sylvatica, but vegetation communities belonging to the Quercion belt are also very developed (Vilks, 2000). The present day pdM landscape is largely reforested (Picea abies, Pseudotsuga menziesii, etc.) and is mainly characterized by heath vegetation, meadows, pastures and isolated arable lands.

3 - Material and methods

5Cores were extracted between 2000 and 2003 from different peat bogs and fens using a 50 x 5 cm “Russian” corer. Very large mires (over 100 ha) and large ones (about 50 ha) were chosen to take into account relevant pollen source areas on a regional scale (Sugita, 1994).

6Samples were taken at various intervals (1, 5 and 10 cm; tab. 1). They were prepared using standard procedures for pollen analysis (Faegri & Iversen, 1989). Pollen counting was performed using a Zeiss microscope at 500x magnification. Pollen and spores identification followed published illustrations and keys (Faegri & Iversen, 1989; Reille, 1992-1999). Minimum pollen counts of 450 dry land pollen grains per sample were made. Pollen values were calculated as a percentage of total land pollen excluding Cyperaceae, fern spores and aquatic plants. The Anthropogenic Pollen Indicators (API) summary curve shows pollen taxa indicative of human impact (Behre, 1981), including those underlined by studies on modern pollen assemblages performed in the study area (Miras, 2004, 2009) and in other mountainous environments (Mazier et al., 2009; Ejarque 2010). Diagrams were plotted using the GpalWin program (Goeury, 1997).

7A total of 16 AMS radiocarbon dates on bulk sediment were performed by the Institüt für Isotopenforschung und Kernphysik (Wien, Austria) and the Vienna Radium Institute (Austria). Dates were calibrated using CALIB 5.0.2 (Reimer et al., 2004).

4 - Results and interpretations

4.1 - Dating

8Radiocarbon dates are shown in table 2. Several radiocarbon dates were refused due to their inconsistency with similar well radiocarbon dated regional palaeobotanical events (Denèfle et al., 1980; Guenet, 1993). An age-depth model for the Longeyroux peat sequence was performed using Microsoft Excel 2003. This model was established by plotting accepted calibrated radiocarbon dates versus depth and assuming linear sedimentation between dates (fig 2).

Tab. 2: AMS-radiocarbon results.

4.2 - Pollen results

9Profiles of the 5 pollen diagrams presented are divided into pollen assemblage zones (PAZ) defined by the most important regional taxa. These PAZ are organized in 8 units (U1 to U8) summarizing the major palynological events (fig. 3, 4, 5, 6, 7 & 8; tab. 3):

• U1 (D-a): this unit records the end of the Late Glacial period and its transition to the Preboreal. This period is characterized by a low AP/NAP ratio at around 30 %. Non-arboreal pollen types present a decreasing trend but they still dominate Cyperaceae (20 %) and Poaceae (30 %) in particular, and, to a lesser extent, Apiaceae and Artemisia. Tree pollen is also present such as Juniperus (20 %) and Betula whose values progress to about 15 %;

• U2 (L₂-a; R-a; D-b): the increase of the AP/NAP quotient (to 70 %) is explained by the progression of Betula (to 40 %) and Pinus (10 %). Quercus and Corylus start simultaneously. Salix and Alnus are locally present;

• U3 (L₂-b_1/2; M-a; R-b; D-c): Corylus frequencies culminate (> 50 %) which explains the increase of the AP/NAP ratio to 90 %. Quercus values also progress to 15 % whereas those of Betula and Pinus decline to 10 % and 5 %, respectively. Ulmus spreads and herbaceous pollen taxa are essentially represented by Poaceae (≤ 5 %). Two isolated Cerealia pollen types are registered without any significant rise of API (L₂-b_1/2);

• U4 (C-a; D-d; R-c): the AP/NAP ratio shows maximum values for the entire unit (around 80-90 %) despite a drop in Corylus frequencies (to 25 %). Quercus percentages increase gradually (to 20 %) while the first regular occurrences of Tilia are evidenced;

• U5 is mainly defined by the variations of Quercus, Corylus and Tilia rates. Two sub-units have been identified:

  • U5-a (L₂-c_1/2; C-b; R-d): The Tilia pollen curve starts and the AP/NAP ratio remains stable at around 80 %. The arboreal pollen types are mainly represented by Quercus (20 %) and Corylus whose frequencies show a gradual decreasing trend. The first regular occurrences of Acer and Fraxinus are observed while Poaceae percentages increase appreciably (to 20 %);

  • U5-b (L₂-c₃; C-b; M-b; R-d): Quercus, Betula and Corylus rates, at about 25 %, 20 % and 20 % respectively, explain a high AP/NAP ratio (about 60 %). Tilia, Fraxinus and Ulmus reach their peak. Poaceae dominates the non-arboreal pollen assemblage with average values at around 20 % and sporadic increases up to 40 %. Calluna frequencies spread. Irregular observations of Cerealia pollen-type and different API (Plantago sp., Rumex, Galium, Urticaceae) pollen types are registered;

• U6 is mainly defined by the variations of Fagus percentages. Four sub-units have been identified:

  • U6-a (L₂-d₁; C-c; M-c; R-e): arboreal taxa (Quercus: 20 %; Corylus: 15 %; Tilia: ≤ 5 %) remain the major component in the assemblage whereas Poaceae values remain stable (about 5 %). The AP/NAP quotient oscillates and the first regular occurrences of Fagus and Abies are recorded. Punctual occurrences of Cerealia pollen-type are registered while API rates show an overall increasing trend (Plantago lanceolata, Plantago sp., Chenopodiaceae and punctual record of Centaurea cyanus, Papaver rhoeas, Cirsium and Urticaceae pollen types).

  • U6-b (L₂-d_2/3; C-d; M-d; R-f₁): arboreal taxa show an overall tendency to increase, particularly Fagus (around 5 %) and Quercus which still dominate the arboreal pollen assemblage. Sporadic occurrences of Cerealia pollen-type as well as noticeable percentages of the API group (mainly Galium, Plantago sp., Plantago lanceolata, Rumex, Urtica, Cirsium, Caryophyllaceae and Chenopodiaceae) which remain stable at around 2 % evidence the existence of agropastoral activities;

  • U6-c (L₂-d₄; C-d; M-e; R-f₂): maximum values (around 80 %) of the AP/NAP ratio are mainly explained by high frequencies of Fagus comprised between 20 and 40 %. Non-arboreal pollen types are mainly represented by Poaceae and Calluna, which comprise 20 % and 15 % of the total pollen sum, respectively;

  • U6-d (L₂-d_5/6; C-d; M-e; R-g_1/2): the AP/NAP quotient fluctuates throughout the unit from 80 to 40 %. This is mainly explained by an initial decline in Fagus values. This event is particularly evidenced in Longeyroux sequence where the percentages of Fagus fall down to 10 %. Poaceae (about 20 %) dominates the herbaceous pollen spectra, and an increasing trend of grazing indicators (Plantago sp., Plantago lanceolata, Rumex, Urticaceae, Galium, Cirsium and Asteroideae) is evidenced. Sphagnum is locally abundant (40 %). A later renewal of Fagus percentages is linked to a new increase of the AP/NAP ratio. Quercus (20 %), Betula (10 %) and Alnus (5 %) are also present in the whole unit and Corylus rates remain stable at around 10 %;

• U7 (L₂-e_1/3; C-e; M-f; R-h; D-e_1/5): the gradual decrease of the AP/NAP quotient (from 70 to 20 %) is explained by the decline of Fagus and Quercus percentages down to 10 %. Poaceae and Calluna values progress up to 40 % while an increase in cereal cultivation (continuous curve of Cerealia pollen type, regular occurrences of Secale) and grazing indicators (Rumex: 10 %, Galium, Urticaceae, Plantago sp., Plantago lanceolata, Caryophyllaceae, Chenopodiaceae etc) is recorded. These taxa reach maximum values in the second part of the unit (percentages of Cerealia and Secale are together toward 5 % are and API values oscillate around 8-10 %, etc.) in which Castanea and Juglans are regularly recorded and Fagopyrum appears;

• U8 (L₂-e₄; C-f; R-h; D-e₆): the AP/NAP ratio progresses from 20 to 50 %. This result is explained by the overall increasing trend of tree pollen taxa such as Betula (10 %) and Pinus (10 %). Picea is also evidenced and the API regresses (mainly Cerealia pollen types).

• Tab. 3: Correlation of the pollen assemblage zones evidenced in the studied peat sequences and overview of the main palaeobotanical descriptions.

  

  Agrandir Original (jpeg, 132k)

5 - Discussion

5.1 - Long-term changes in regional vegetation and woodland composition: an original history

5.1.1 - Late Glacial and Early Holocene

10One sequence located in the northern studied area (Mont d’Ambazac) registers the Late Glacial / Early Holocene transition (U1: [D-a], fig. 1 & 3). Pollen data show a local vegetation dominated by extensive populations of Cyperaceae and Poaceae at around 10800 ± 130 BP (ca. [13049-12399] cal. BP). They probably hide the steppic grasslands which were regionally developed on drier soils and scattered by Juniperus. The regression of steppic vegetation at around 10800 ± 130 BP (ca. 11700-12100 cal. BP) is synchronously documented in the rest of the French Massif Central (Reille et al., 1992). Quercus and Corylus pollen evidence may derive from long-distance transport. Rapid warming and soil accumulation in the early Holocene provided favourable growing conditions for the extension of Betula and Pinus. Although Pinus and Betula are both strong pollen producers, the moderate values of the AP/NAP ratio suggest that the regional woodland cover was probably not important (U2: [R-a], [L₂-a], [D-b], fig. 3, 4 & 5). This period is thus characterized by a mosaic landscape of open grasslands and clear woodland of pioneering trees.

11Corylus and Quercus (end of U2) spread at the same time on the pdM ([R-a]) and in the whole Limousin ([D-b]) towards 9120 ± 90 BP (ca. [10566-10152] cal. BP). Similar data have been revealed in the Madeleine Mounts around 10185 cal. BP (Argant & Cubizolle, 2005). This differs from other areas of the centre of the Massif Central where just Quercus develops between ca. 9500 and 9000 BP (ca. 10800-10100 cal. BP) (Reille et al., 1992). The regional woodland is subsequently dominated by Corylus. This data, thus, underline the existence of different Quercus and Corylus dynamics between the West and East boarder of the Massif central and its central part. This process culminates towards 9500 cal. BP (U3: [L₂-b₁], levels 164-163 cm depth, fig. 2 & 3) while the former Betula and Pinus forests regress and disappear. A third mesophilous taxon – Ulmus- begins to develop at this period even if its Holocene expansion remains moderate (U3: [R-a], [D-b], [M-a]).

12Finally, the regression of Corylus populations towards 7510 ± 70 BP (ca. [8417-8183] cal. BP, U4: [R-c]) appears to be progressive, moderate and synchronous with the arrival of Tilia. Contrary to the rest of the Massif Central, where spaces created by the removal of Corylus are filled by an expansion of Quercus (Reille et al., 1992), no significant expansion of the oak woodlands is evidenced on the pdM and Corylus remains the main arboreal taxa during the early Atlantic.

5.1.2 - Mid-Holocene

13The woodland composition notably changed around 7040 ± 110 BP (ca. [8051-7657] cal. BP, U5-a: [L₂-c_1/2]) with the outstanding progression of a diversified deciduous oak forest and a second diminution of hazelnut populations. Tilia starts to expand on the pdM between 7040 ± 110 BP and 6850 ± 140 BP (ca. [7944-7433] cal. BP) (Guenet, 1993) while it tends to be dated around 7240 ± 170 BP (ca. [8327-7794] cal. BP) in the rest of the Massif Central (Reille et al., 1992).

14A second phase of vegetation change occurs from 6850 ± 140 BP (ca. 7900-7400 cal. BP) (Guenet, 1993) and 6500 BP (ca. 7300-7200 cal. BP, U5-b: levels 115-113 cm depth in [L₂-c₃], [C-b], [C-M], fig. 2, 4, 6 & 7), with the remarkable expansion of a diversified oak woodland. The contemporaneous expansion of Tilia, Fraxinus and Alnus underlines the installation of more mesophilous and humid climatic conditions. The development of Fraxinus is lower in the Limousin (at every altitudinal range) than in neighbouring Artense (Vergne, 1989; Guenet, 1993). These new pollen data thus confirm that Fraxinus expansion seems to describe an East-West gradient rather than an altitudinal one in the northern French Massif Central.

15The diversified oak forests regress from ca. 5500 cal. BP with a gradual decline of Tilia which begins with the spread of Fagus (U6-a: 4800 ± 90 BP, ca [5664-5319] cal. BP, [M-c]; 4820 ± 70 BP, ca. [5663-5446] cal. BP, [L₂-d₁], fig. 4 & 7) and which is accentuated with the first oak beech woodlands ([M-c], [C-c]). A singular expansion of Fagus in this period is suggested by pollen data on the pdM. Fagus then starts to spread on the pdM ca. 700 years later than in neighbouring Artense where its arrival is dated towards 5470 ± 70 BP (ca. [6405-6003] cal. BP) (Guenet, 1993). This chronological discrepancy is accentuated when the pdM is compared with the lower altitude western Loire valley where the first regular occurrences of Fagus are dated between ca. 5100 and 4500 cal. BP (Planchais, 1971; Beaulieu et al., 1991). The pdM thus appears to be an intermediate zone in the Fagus Holocene expansion between the centre of the mountainous Massif Central and the western valleys. Climatic factors may be an important driving factor behind these asynchronisms in the expansion of Fagus which describes an East-West altitudinal gradient. It can be assumed that the climate reversal between 5600 and 5300 cal. BP (Magny et al., 2006) caused cooler conditions and/or changes in moister conditions which were not sufficient enough to cause a real decline of the diversified deciduous oak forests and their rapid substitution by beech woodlands. Other possible factors – mainly anthropogenic – will be considered below (see 5.2).

16Pollen data also evidence a gradual expansion of Fagus on the pdM. Even if the first oak-beech forests start to spread from 4080 ± 60 BP (ca. [4826-4418] cal. BP) (Guenet, 1993), beech populations completely filled the oak woodlands only towards 3340 ± 70 BP (U6-b: ca. [3724-3402] cal. BP, levels 75-70 cm depth [C-d]). A similar radiocarbon date, 3650 ± 95 BP (ca. [4242-3694] cal. BP), had previously been obtained at Longeyroux (Denèfle et al., 1980). After 3450 ± 70 BP (ca. [3889-3482] cal. BP, levels 56-54 cm depth [L₂-d₄]), beech and oak-beech forests progressively dominate the regional landscape. Finally, as previously described (Guenet, 1993), these new pollen data confirm a total absence of fir populations on the pdM despite an important rainfall which was probably not sufficient enough for its expansion.

5.2 - Overview of the human impact history

17The earliest Anthropogenic Pollen Indicators (API) of grazing, agricultural practices and woodland clearances are recorded on the pdM during the early Neolithic from ca. 6600-6500 cal. BP (U5-b: levels 100-98 cm depth [L₂-c₃], fig. 2, 4 & 8; levels 125-120 cm depth [C-b], fig. 6 & 8). Even if regular occurrences of agricultural and grazing pollen indicators could be related to an increasing trend of anthropogenic pressure particularly during the Middle Neolithic, it is the late Neolithic / early Bronze Age transition which constitutes the first threshold in the land-use history of the pdM. From 4820 ± 70 BP (U6-a: ca. [5663-5446] cal. BP in Longeyroux-2, [L₂-d₁]; 4800 ± 90 BP (ca. [5664-5319] cal. BP in Malsagne, [M-c], fig. 4, 7 & 8), the landscape starts to be regionally dominated by oak forests progressively filled by Fagus with small cleared areas as suggested by the extension of heathlands and grasslands (e.g. [C-c] and [L₂-d₁]). The progressive spread of Fagus is synchronous with a more significant human activity as expressed by the overall increasing trend of several farming and grazing pollen indicators (Cerealia-pollen-type and Centaurea cyanus in [L₂-d₁] and [C-c], Papaver rhoeas-type in [C-c], regular occurrences of trampling indicators such as Plantago lanceolata, Plantago sp., and ruderal plants like Rumex in [L₂-d₁], [C-c] and [R-e], first notations of nitrophilous plants - Urticaceae in [C-c] and [M-c] - and pastoral indicators evidenced recently in montane areas such as Cirsium (Mazier et al., 2009) in [R-e] and [M-c] or Galium (Miras, 2009) in [L₂-d₁], [R-e] and [M-c]). Even if climatic factors may have played an important role in European Fagus dispersal (Magri, 2008), these pollen data suggest a possible anthropogenic trigger of Fagus expansion as it has been previously demonstrated in different western and central European mountain areas (e.g. Galop & Jalut, 1994; Tinner & Lotter, 2006).

18Since the early Bronze Age, the continuous and stable presence of grazing and farming pollen indicators (U6-b: occurrences of Cerealia pollen-type, higher diversity of pastoral indicators such as Plantago sp., Plantago lanceolata, Galium, Rumex, etc.) suggest the permanence of a regional grazing pressure accompanied by small-scale agriculture which may have taken place in cleared areas. Relevant pollen indicators reveal an intensification of forests clearances: pollen data evidence an abrupt interruption of the Fagus curve dated to 3450 ± 70 BP (U6-b: ca. [3896-3556] cal. BP, levels 60-57 cm in [L₂-d₃]; levels 115-110 depth in [R-f₁]; levels 90-80 cm in [M-d], fig. 4, 5, 7 & 8). These pollen data thus provide the first evidence of a possible anthropogenic driving force in the differed and uneven installation of beech on the pdM. Early Bronze societies may have had an impact on the beech populations through woodland clearances which hampered beech settling.

19Oak-beech woodlands dominate the landscape after ca. 3500-3200 cal. BP (U6-c: levels 50-54 cm depth [L₂-d₄], [R-f₂], fig. 2, 4 & 8). Subsequently, a drastic decrease of the Fagus rates recorded in several study sites (U6-d: [L₂-d₅], [R-g₁] and middle of [C-d] and also Prugnolas [2H] in Guenet, 1993) from ca. 3200-3000 cal. BP (levels 48-50 depth in [L₂-d₅], fig.  2, 4 & 8) suggests a second beech and oak-beech forest clearance on a regional scale which could be related to the final Bronze. A pastoral purpose can be suggested as an increase in grasses and different grazing indicators is recorded (Plantago sp., Plantago lanceolata, Rumex, Galium, Cirsium and Chenopodiaceae).

20Pollen data record a subsequent regeneration of the woodland cover with the maximum extension of the beech and beech-oak forests (U6-d: end of [C-d], [R- g₂], [l₂-d₆]). According to the pollen data, the composition of the forest cover remained largely unchanged until La Tène, and more particularly the 4^th-2^nd centuries BC, which represent a threshold in the long-term land-use history of the pdM (U7: 2230 ± 60 BP, ca. [2350-2112] cal. BP in Chabannes and 2210 ± 60 BP, ca. [2345-2098] cal. BP in Ribière Nègre). Important forest openings are recorded at a large scale in the Limousin in both mountain areas (U7: [R- g₂/h], [C-d/e], [L₂-d₆/e₁]) and at lower altitudes (Miras, 2004). These deforestations seem to be devoted more to pastoral purposes since grazing indicators increase more significantly than farming ones (fig. 8). Forests are rapidly replaced by open pasture areas such as grasslands and Calluna heathlands. These deforestations become more extensive at the La Tène / Gallo-Roman transition (around 2070 ± 80 BP, ca. [2307-1870 cal. BP], in Denèfle et al. (1980) and 2180 ± 90 BP, ca. [2351-1933 cal. BP] in Miras (2004)). Cleared areas are probably devoted to the establishment of the first mixed exploitation system insofar as both pastoral and arable pollen indicators increase together (beginning of U7). Castanea and Juglans appear for the first time on the pdM subsequent to the start of these deforestations (U7: [M-f], [C-e], [R-h] and Prugnolas [2I] in Guenet, 1993). This is also observed at a lower altitude (Miras, 2004). All this argue in favour of the Roman introduction of these tree taxa in the Limousin.

21Unfortunately, the low sedimentation rate recorded at the top of the peat sequences presented prevents us from describing further palaeobotanical events with accuracy. As this paper aims to deal with human/environment interactions from Prehistory to the beginning of the Roman period, pollen assemblages from U7 and U8 will not be discussed here with more precision. Pollen analyses of new peat sequences have been undertaken to trace roman, medieval and modern human actions related to woodland clearances and land-uses on the pdM. The high sedimentation rate found in these new sequences for the late Holocene will allow us to focus on detailed historical human activities.

6 - Conclusions

22This palynological study underlined the singular Holocene vegetation history of the pdM, and it also evidenced the first steps of the long-term land-use history since the early Neolithic. Several environmental phases related to upland human activities were evidenced.

1. Pollen data suggest a first potential episode of human activity towards ca. 4600-4500 cal. BC (ca. 6600-6500 cal. BP). Similar activities were earlier documented in the neighbouring Auvergne region in both mountain and valley areas. For instance, in the Cantal Mountains, repeated local forest fires are concomitant at the beginning of the 6^th millennium BC with relevant grazing pollen indicators, and the first cereal pollen-types are recorded as early as ca. 5800 cal. BC (Surmely et al., 2009). Moreover, in the Sarliève valley, archaeological and pollen data evidence a phase of human activity and woodland clearance between ca. 5800-5600 cal. BC (Trément et al., 2005). Multidisciplinary research carried out in other western European mountains also reflects early occupations of high altitude spaces, mainly between 5500 and 4300 cal. BC (e.g. for Alps (Galop & Vaquer, 2004; Walsh & Richer, 2006), for Pyrenees (Ejarque et al., 2010), for Jura massif (Gauthier, 2004; Richard, 2005) etc.). Further research will be necessary to explain this early Neolithic land-use dissimilarity evidenced between the pdM and the Auvergne.

2. Even if proper radiocarbon dated studies are further needed, the pollen data presented here allow for the first time to depict this medium mountain of the pdM as a complex landscape shaped over a long-term land-use history marked by important thresholds.

   1. During the middle to the late Neolithic, from ca. 3700 to ca. 3300 cal. BC (ca. 5700-5300 cal. BP), pollen data document higher agropastoral pressure which enhanced the expansion of grasslands and woodland clearance. Archaeological data also reveal a greater human occupation of the pdM from this period (Lintz, 1992). This phase was also recorded in Auvergne at ca. 3650-3348 cal. BC in the Chain of Puys (Michelin et al., 2001; Miras et al., 2004a) or in the Madeleine Mounts (Argant & Cubizolle, 2005) and after ca. 3300 cal. BC in the Sarliève valley where a higher intensity of erosion and human settlement evidence an important phase of human impact (Macaire et al., 2010). This period seems thus to constitute a real threshold in the landscape shaping of mountains and valleys areas in the whole Massif Central.

   2. During the early Bronze Age, from ca. 2000 to ca. 1500 cal. BC (ca. 4000-3700 cal. BP), the existence of woodland clearances related to agricultural and grazing activities in all the studied sequences indicated a phase of coincident human pressure and landscape change. Our data also suggest that human-induced forest clearances, in addition to climatic factors, could have played a role in the delayed development of the beech forests in this area. Similar data have been recently advanced in the Morvan massif, which is situated in the North-Eastern boarder of the Massif Central (Jouffroy-Bapicot, 2010). Besides, in the Madeleine Mounts, located in its eastern boarder (Argant & Cubizolle, 2005), clearances have also been shown at the beginning of the installation of beech woodlands between ca. 2150 and 1600 cal. BC.

3. Between ca. the 4^th and the 2^nd centuries BC (ca. 2350-2100 cal. BP), large beech and beech-oak woodland clearances are related to an important agropastoral extension. This induced a rapid change towards a human-dominated cultural landscape. Archaeological data argue in favour of a more important Lemovices occupation (Lintz, 1992) and suggest that these landscape changes could also be connected to gold exploitation dated from ca. the 5^th and the 1^st centuries BC in the mines located in the Limousin (Corrèze, Haute-Vienne, Dordogne) (Cauuet, 2000). Moreover, in the lower plateaux of Tulle, close to the pdM, a maximum of gold mining is dated to ca. the 3^rd-2^nd centuries BC, and this phase is concomitant with the appearance of buried mines (Boussicault, 2004). This technological change implied an important wood supply. The Iron Age / Gallo-Roman transition in the French Massif Central thus appears to be a complex threshold in the shaping of this medium mountain landscape, suggesting important variability in mountain area human management. For instance, similar important landscape openings are evidenced at the La Tène / Gallo-Roman transition at both the Chain of Puys (Miras et al., 2004a) and the Madeleine Mounts (Cubizolle et al., 2004), a process which culminates in the Cantal around the 3^rd-4^th centuries AD (Surmely et al., 2009).

4. A new phase of woodland clearance during the Roman period is accompanied by the development of a diversified land-use system which combined grazing with mixed cereal cultivation. Pollen data testify to the regional cultivation of Secale (rye) as well as the first records of Fagopyrum pollen (Diot et al., 2006). Further archaeological and palaeoecological studies with systematic archaeobotanical analyses at high spatial and temporal resolution will be necessary to clarify this matter since carpological studies underline its medieval introduction (Ruas, 1992). Nevertheless, pollen evidence of Fagopyrum is often registered in peat sequences or archaeological sites from different socio-environmental contexts (Marguerie et al., 2009), putting into question an earlier introduction of this crop.

23This paper demonstrates firstly that Holocene landscape variability evidenced on the pdM responds to the existence of changing human pressure during specific periods. Nevertheless, further palaeoenvironmental multi-proxy analyses combined with archaeological landscape investigations are needed to accurately reconstruct Limousin land-use history from the Neolithic onward. Secondly, these upland land-use thresholds evidenced for the first time in this region are largely in accordance with those underlined in neighbouring Auvergne and other mountains such as those found in the middle to the late Neolithic (from ca. 3700 to ca. 3300 cal. BC) or in the early Bronze Age (from ca. 2000 to ca. 1500 cal. BC) (Miras et al., 2004b). However, some variability in this land-use model begins to be detected on a regional scale. The chronology of the early Neolithic occupation shown by pollen data in the Limousin, for example, is different from that in Auvergne or the important landscape openings registered in the whole French Massif Central (e.g. Reille et al., 1992; Cubizolle et al., 2004; Pulido Avil, 2006) which occur with different chronologies, suggesting the development of complex and diversified land-use systems on a regional scale. Multi-proxy palaeoenvironmental and archaeological research performed in the southern Pyrenean slope also evidences a temporal and spatial fragmentation, and structuration of upland land-uses on a microregional scale (Ejarque et al., 2010). Moreover, the study presented here corroborates the importance of agriculture and grazing in causing a long-term and progressive environmental impact on the pdM landscape. Nevertheless, it also questions the hypothesis of Iron Age mining activities. A more complete understanding of mountain land-use systems needs to be based on the existence of a wide range of human practices which were not merely focused on grazing, but also included the exploitation of multiple upland resources. Finally, these results point to the necessity of carrying out integrated palaeoenvironmental multiproxy research and archaeological investigations with high spatial resolution on a microregional scale in the French Massif Central. This appears to be the most suitable research strategy for the study of human mountain management and landscape change during the Holocene (Davies, 2007; Ejarque al., 2009, 2010; Palet et al., 2007). This would also provide the basis for re-assessing how interactions between environmental (mainly climatic) and cultural, social, demographic and economic factors affected land-use dynamics reconstructed in the French Massif Central.