The aim of the study was to assess the impact of a pastoral management chosen to limit the recent expansion of woodland on a Mediterranean mountain on land snail diversity. An additional aim was to acquire quantitative data that could be used to identify pasture environments from Holocene molluscan assemblages. The work was undertaken at the Luberon mountain, Provence, south of France. We used a stratified quantitative sampling scheme according to altitude and vegetation structure. A total of 80 sites were studied. Large species were collected within a 5 × 5-m plot. Small species were extracted from litter and surface soil. A standard procedure for site description was used based on 35 environmental variables. Grazing pressure was estimated according to the impact of grazing on the herb layer. Correspondence analysis and canonical correspondence analysis were performed using canoco 4.0 software. The distribution of land snails is related to altitude and grazing intensity. Large patches of grazed grassland harbour open country and mountain snail species. Thermophilic open ground species are located in grazed grasslands at lower altitude. Shade-loving species are present in ungrazed scrublands or in small clearings on the upper slopes. The lowest species richness, diversity and equitability are associated with large patches of grazed grassland, the presence of a continuous cover of short grass reinforcing this negative impact on snail diversity. Our study is consistent with similar works on land snails or other invertebrates but discordant with vegetation studies. A homogeneous grazed herb layer significantly reduces snail diversity and abundance. Heterogeneity seems to favour snail diversity both at the local and landscape scales. However, sheep grazing contributes to the expansion of suitable habitats for rare snail species.
Land abandonment, which has been ongoing in Europe since 1950 (Meeus et al., 1991), is a crucial problem in Mediterranean France, and chiefly in mountain areas traditionally grazed by sheep and goats during the dry season. As in other parts of the northern rim of the Mediterranean basin, the livestock population has been drastically reduced, resulting in undergrazing or even complete elimination of grazing (Papanastasis, 1997). This recent trend contrasts with several centuries of overgrazing, which was partly responsible for the deforestation of the Mediterranean region. As a consequence, settlement of scrublands and woodlands follows abandonment and leads to a decrease in open ground habitats. The aim of most management policies is now to maintain different kinds of open environments using grazing (and even mechanical cutting of vegetation) in order to protect plant and animal diversity (Bakker, 1989).
The Luberon mountain, in Provence (Fig. 1a), is an ideal site well representing the effect of such landscape changes on flora and fauna. At the beginning of the nineteenth century, Pazzis (1808) described a totally deforested mountain, covered with poor dry pastures, with some rare woodland refuges. This can be confirmed by an examination of the ‘Napoleonic’ survey established 20 years later during the maximum of rural population. The first sign of a reversal in this trend took place after World War I. Land abandonment increased between 1945 and 1950. Woodlands and scrublands became dominant in 1950 (Hostein, 1990), but aerial photographs at this date show a still-open vegetation with active gullies. The present-day landscape is both the result of this long environmental history and of recent land-use conflicts, i.e. forestry vs. grazing and biological conservation: the north-facing and south-facing slopes are highly wooded, open spaces being limited to small clearings, while a narrow band of grazed grassland is maintained on the ridges between 900 and 1125 m (Fig. 1b). These recent changes are dramatic, since 51% of grazed grasslands were lost between 1950 and 1991 (Garde, 1992).
Ecological consequences of both land abandonment and pastoral management must be studied at several levels in the ecological hierarchy (Allen & Hoekstra, 1982; O’Neill et al., 1986). Does land abandonment lead to a local decrease in biodiversity as can be predicted for many arthropods (Brown & Kalff, 1986) and the loss of some rare open habitat species (Noy-Meir et al., 1989)? Alternatively, land abandonment could lead to an increase in species diversity (EGPN, 1987) through increased heterogeneity in the landscape mosaic. Further complications can include the expansion of forest fires (Barbero et al., 1990). Is pastoral management an effective measure for the preservation of rare species and for biodiversity augmentation?
Here we present a study which focuses on the response of land snail communities to a pastoral management regime used to thwart the effects of the recent reforestation. Although land snails do not appear to be dependent on the presence of particular plant species, they are very dependent on the structure of their habitat (Boycott, 1934; Grime & Blythe, 1969; Cameron & Morgan-Huws, 1975; Labaune & Magnin, 2001) and have a low dispersal ability. Little is known on these themes for the Mediterranean area, although related research has been undertaken on the impact of grazing on lowland snails in Britain (Cameron & Morgan-Huws, 1975; Cameron, 1978; Ruesink, 1995; Davies & Grimes, 1999). A second, subsidiary goal was to acquire quantitative data on the land snail communities of dry pastures to enable detection of ‘pastoral signatures’ from Holocene assemblages in the same geographical area.
The Luberon is a limestone mountain approximately 40 km long, running from east to west, and located in Provence (France), 50 km from the Mediterranean sea (Fig. 1a). Our study focused on the eastern part, known as the ‘Grand Luberon’, which reaches an altitude of 1125 m. Precipitation exceeds values for other Provence mountains and temperatures are cooler (Livet, 1965). Mean annual temperatures for the Grand Luberon range between 11 °C and 12.5 °C at the foot and between 7 °C and 9 °C at the summit. There is only one dry month and 4–5 cold months when temperatures fall below 7 °C (CNRS, 1975). Rainfall was estimated at 802 mm per year at the summit (Silvestre, 1977). The harsh climate is accentuated on the ridges and northern slopes by a fierce, north-westerly wind (‘Mistral’).
The Grand Luberon’s ridges are covered by a thin band of short grasslands between 900 and 1125 m (Fig. 1b). There is a continuous gradient between the typical Genistetum villarssii and the Festuco-Brometum, according to soil thickness (Garde, 1992). The Parc Naturel Régional du Luberon maintains these grasslands by encouraging grazing with three main goals: to create a firebreak between the two slopes, to preserve a remarkably rich biological area and to protect an original pastoral landcape in Provence. Grasslands on the summit ridges are grazed in summer by flocks of 1250–1800 sheep. The slopes themselves are highly wooded (primarily Quercus pubescens and Quercus ilex stands) and open spaces are limited to variably sized clearings. Soils are superficial and gravely on steep slopes and ridges, with a pH between 6.8 and 7.6. CaCO3 content varies between 0 and 2%.
A stratified sampling scheme was chosen according to altitude and vegetation structure. Sites were sampled throughout the mountain. Only open environments were studied. Different vegetation structure types were analysed, from low grasslands to Buxus sempervirens scrublands, including high ungrazed grasslands with or without low, woody vegetation. Sites were distributed equally between eight altitude categories, from 700 m to the summit. When possible, each different vegetation structure type was equally represented for each altitude category. A total of 80 sites were studied from 13 March to 6 May, 1997.
A 5 × 5-m plot was examined at each site for 15 min. All living or fresh dead snails were sampled, then identified and counted in the laboratory. A method based on the works of Evans (1972), Puisségur (1976) and André (1981, 1982) was used to collect shells of less than 5 mm diameter: vegetation, litter and soil surface down to 5 cm were sampled over a 25 × 25-cm plot. Five samples were taken within the 5 × 5-m square already sampled for large species. Samples were dried in a drying oven, then immersed in water. Floating particles were collected in a 0.5-mm mesh sieve and shells were then separated from plant material under magnifying glass for the smallest portion.
A standard site description procedure was used, based on Godron et al. (1968). Thirty-five environmental variables were noted for each site during fieldwork. In the present study only 23 variables were kept (list in Appendix 1). They describe (i) topography and climate, (ii) vegetation structure and type, (iii) habitat structure at the soil surface and (iv) grazing pressure. Grazing pressure was estimated using the following chart formulated by the CERPAM (1996) according to the impact of grazing on the herb layer:
1Rapid crossing of the flock of sheep; grass more or less flattened.
2Low level of consumption; the best plant species are eaten; strong wasting.
3Medium consumption; only the less palatable species neglected.
4Large consumption; the whole grassland is grazed; only some neglected tufts.
5Very large consumption; the whole herb layer is grazed short; earth sometimes apparent.
Statistical analyses were performed using canoco 4.0 (ter Braak & Smilauer, 1998). The species–sites matrix was first studied by correspondence analysis (CA). Abundance data were transformed logarithmically to make their distribution correspond to a normal law (Legendre & Legendre, 1984). Thirty-two snail species were taken into account, but rare species abundance was down-weighted by an algorithm available in canoco.
From the 23 environmental variables we have retained the nine that explained most of the variance after a step-by-step procedure of selection. The effect of grazing pressure and environmental variables on the species matrix was then studied using canonical correspondence analysis (CCA) (ter Braak, 1986). The significance of each variable selected and of the ordination axes was verified by a permutation test available in canoco (Monte Carlo test). The significance level was set at P < 0.05.
Organization of land snail communities within the landscape
The 80 samples yielded 26 642 individuals of 32 species (Table 1). The composition of communities is homogeneous: nine species are found in 40% of the sites and 18 species in more than 10%. Fourteen species are found in less than 10% of the sites. Most of these species are shade-loving ones collected in small clearings (edges effect) or in scrublands. Other species are mountain or Mediterranean species, which can be found in some sites at the two ends of the altitudinal gradient.
Table 1. Occurrence and number of land snails collected in the 80 sites of the Grand Luberon
Species and abbreviations
CUN Candidula unifasciata (Poiret 1801)
TGE Trochoidea geyeri (Soós 1926)
PTR Pupillatriplicata (Studer 1820)
TCA Truncatellinacallicratis (Scacchi 1833)
CAC Cecilioidesacicula (Müller 1774)
GVA Granariavariabilis (Draparnaud 1801)
MCA Monachacantiana (Montagu 1803)
CRP Clausiliarugosaparvula (Férussac 1807)
JQU Jaminiaquadridens (Müller 1774)
TCY Truncatellinacylindrica (Férussac 1807)
XCE Xerosectacespitum (Draparnaud 1801)
ASE Abidasecale (Draparnaud 1801)
CSE Cochlostomaseptemspirale (Razoumowsky 1789)
AAC Acanthinulaaculeata (Müller 1774)
PPY Punctumpygmaeum (Draparnaud 1801)
VCN Vitreacontracta (Westerlund 1871)
VNA Vitreanarbonensis (Clessin 1877)
PPA Pagodulinapagodula (Des Moulins 1830)
UGL Urticicolaglabellus (Draparnaud 1801)
CNE Cepaeanemoralis (Linnaeus 1758)
PEL Pomatiaselegans (Müller 1774)
MOB Merdigeraobscura (Müller 1774)
LCY Lauriacylindracea (Da Costa)
VCO Valloniacostata (Müller 1774)
CGI Candidulagigaxii (Pfeiffer 1850)
PMA Phenacolimaxmajor (Férussac 1807)
ZDE Zebrinadetrita (Müller 1774)
SDO Sphyradiumdoliolum (Bruguière 1792)
ZAL Zonitesalgirus (Linnaeus 1758)
APO Abidapolyodon (Draparnaud 1801)
CAV Chondrinaavenacea (Bruguière 1792)
SSI Solatopupasimilis (Bruguière 1792)
A correspondence analysis was performed using the 32 species. Axes 1 and 2 of the CA (Fig. 2a,b) explain 21.4% and 13.5%, respectively, of between-species variation. The cumulative eigenvalue total is 1.872. Axis 1 clearly represents an altitude gradient (Fig. 2b). It separates low altitude and high altitude sites. Sites at the foot of the sampled gradient at less than 800 m are at the far positive end of axis 1 (e.g. L66 = 769 m, L67 = 710 m). The negative end of axis 1 is characterized by sites at an altitude of more than 1050 m (e.g. L52: 1064 m), i.e. near the summit. There is a significant correlation between site altitudes and their rank on axis 1 of the CA (r = 0.684, P = 0.01). Axis 2 captures a gradient of habitat structure. The sites with the strongest contributions on the negative side (L40, L38, L41, L79 and L35) show the highest complexity in vegetation stratification and soil cover. These are Buxus scrublands where the different layers are relatively equally represented with an extremely heterogeneous soil cover. On the other hand, the positive end shows the sites (e.g. L52, L71) containing grazed grasslands where the lowest herbaceous layer (0–5 cm) has a cover of more than 90%.
The distribution of snail species is as follows for the two gradients (Fig. 2a). The highest contributions to axis 1 are T. geyeri (Soós) on the negative side and C. unifasciata (Poiret) on the positive side. These two hygromiid species are mainly responsible for the horizontal shape of the plot. They are the most common in the Grand Luberon (Table 1) and differ in their altitudinal range: T. geyeri occupies the highest sites and is replaced by C. unifasciata at sites lower than 900 or 1000 m. Typical mountain species of southern France can be found towards the negative end, near to T. geyeri, i.e. P. triplicata (Studer), C. rugosa (Férussac) and A. secale (Draparnaud). On the positive side, C. unifasciata is associated with southern species less frequent at high elevation, i.e. C. gigaxii (Pfeiffer), J. quadridens (Müller) and A. polyodon (Draparnaud). The negative side of axis 2 contains species such as Z. detrita (Müller), which prefers mountain scrublands, P. elegans (Müller), which lives on loose soils and eats dead leaves, V. costata (Müller) and U. glabellus (Draparnaud) which are also common species in preforest habitats. Species from very open environments can be found on the positive side, such as T. cylindrica (Férussac), V. narbonensis (Clessin) and T. geyeri.
Thus the 1–2 factorial plane reveals three land snail community types: (i) mountain assemblages of low grasslands; (ii) open ground Mediterranean assemblages at low altitude; and (iii) shade-loving assemblages of a more complex habitat structure (from the point of view of vegetation layers and soil cover) or landscape structure. Thus, two of the most important factors found to explain species distribution seem to be the altitudinal gradient and a gradient of complexity in habitat structure.
Habitat relationships and impact of grazing pressure
A canonical correspondence analysis was carried out in order to clarify the relationships between habitat structure, grazing pressure and species assemblages. Nine of the initial 23 environmental variables were kept after a step-by-step selection (Table 2). The sum of the canonical eigenvalues of this CCA is 0.655, i.e. 35% (0.655/1.870) of species variation is explained by these variables. The two first ordination axes explain 21.9% of the variation observed in the land snail assemblages (Fig. 3).
Table 2. Correlations between environmental variables and ordination axes 1 and 2 (NS = nonsignificant) of the CCA. Only variables with some significant relation to an ordination axis are listed
51–100 cm layer
0–5 cm layer
Grassland patch size
The first axis (12% of variance of species data) discriminates between species of low altitude at the positive end (C. gigaxii, V. narbonensis, S. similis (Bruguière), Z. algirus (Linnaeus), A. polyodon) and species of mid- and high altitude (T. geyeri, T. cylindrica, P. elegans, Z. detrita). The projection of environmental factors shows that the first axis can be interpreted as an altitudinal gradient. This is confirmed by the location of all the sites from the summit ridge at the negative end of the first axis (Fig. 3b). There is a significant correlation between the altitude variable and axis 1 (Table 2). The 0–5 cm vegetation layer and the cryptogams contribute to the positive end of this axis.
The second axis (9.5% of variance of species data) opposes open ground species of high altitude at its positive end (T. geyeri, T. cylindrica) to shade-loving species (P. elegans, P. pagodula (Des Moulins), V. contracta (Westerlund), U. glabella, C. septemspirale (Razoumowsky), P. pygmaeum (Draparnaud), A. polyodon) or species from scrubland habitats (X. cespitum (Draparnaud), C. gigaxii, Z. detrita, S. similis, C. avenacea (Bruguière)) on its negative end. The significant correlations between the variables ‘grazing pressure’, ‘0–5 cm vegetation layer’, ‘grassland patch size’ and axis 2 show that large patches of heavily grazed grassland contribute strongly to the positive part of this axis. The variable ‘stones’ is also correlated with axis 2, but on its negative end, for scrublands are frequently established on screes or stone paved grounds.
Overall, this CCA discriminates between four kinds of habitats and species assemblages. These are: (i) large patches of grazed grassland on the summit, with mountain species; (ii) grazed grasslands at low altitude, having an important cover of the 0–5 cm vegetation layer, with more thermophilic species; (iii) ungrazed scrublands on screes or small grassland patches of the upper slopes, with shade-loving species; and (iv) ungrazed scrublands at low altitude, having a strong cover of cryptogams (mainly lichens of the genus Cladonia) and boulders, with both shade-loving and saxicolous species.
The effect of grazing on species diversity
We have plotted three diversity indices of land snail assemblages on the previous canonical plane: species richness, Shannon index of diversity and evenness. We give only the trend of Shannon entropy (Fig. 4) because these three measures of diversity show the same gradient pattern according to environmental variables and grazing pressure. The highest diversity values correspond to mid-altitude scrublands and screes or to small clearings. The lowest diversity scores are related to large patches of grazed grasslands, with a dense, short herb layer. The presence of a continuous cover of short grass seems to reinforce the strong, negative impact on snail diversity.
Although the altitudinal gradient is not large (425 m), it explains most of the variation observed in the data matrix. One could be tempted to interpret this altitudinal gradient only in terms of a climatic gradient (i.e. −0.72 °C/100 m). In fact we have shown that the altitudinal gradient is more complex and results from a combination of several factors such as climate, competitive interaction between species, history and landscape management styles (Magnin, 1991; Magnin, 1993; Labaune & Magnin, 2001), grazing by sheep having considerably changed vegetation structure on the summit ridges, and thus climate near the ground. So, the distinctive altitudinal limits of species do not correspond to true climatic ones, for these limits are often different from one gradient to another. Classical studies (e.g. Terborgh, 1971; MacArthur, 1972) had already pointed out the actual complexity of altitudinal gradients.
Grazing pressure is the variable best correlated with axis 2 of the CCA and is associated with large grassland patches and a high herb cover between 0 and 5 cm. This result highlights the role of sheep grazing in shaping both the structure of the landscape and local snail habitat.
Land snail communities from grazed grassland are poorly diversified and this is independent of altitude. CCA underlines the negative effect of grassland patch size, and above all of a short and continuous herb cover. Correlation between structural heterogeneity and species richness has been demonstrated in a number of classical studies (MacArthur & MacArthur, 1961; Pianka, 1967; Cody, 1975; Southwood et al., 1979; Burel & Baudry, 1999). In our case, structural heterogeneity acts on diversity at two levels. At the site level, a homogeneous and short herb layer offers a very small number of niches since such a simple habitat excludes shade-loving, litter-feeding and saxicolous species. On the contrary, ungrazed grasslands and scrublands have a more heterogeneous horizontal and vertical structure and thus present a larger number of niches for different functional groups of land snails. Our results are consistent with those of previous studies in Britain. Morris (1969) found that, in grassland subjected to constant grazing, snails were half as abundant as in grasslands left ungrazed for several years, indicating that high grazing pressure may actually be detrimental to snails. Cameron & Morgan-Huws (1975) have shown that grazing pressure does not have to be relaxed for very long before a massive change in fauna occurs resulting in a reduction of xerophilic elements and a colonization by widespread and shade-loving snails. In a further study, Cameron (1978) found that grazing decreases the proportion of woodland species, soil and litter species being generally more affected than rocky-habitat species. In this case study, one unique site harboured a fauna as diverse as that of an ancient woodland but it was a mosaic of grazed and ungrazed habitats with damp patches. In the same way Ruesink (1995) found that open-ground snails tend to be relatively more common at sites consistently grazed, although heavy grazing causes low overall snail abundance. At the landscape level, small grassland patches benefit from the surrounding woodlands. The community is enriched with the dispersal of some shade-loving snails and, if the patches are not too small, open ground species can survive (Magnin et al., 1995). On the other hand, a successful colonization by shade-loving species is not possible in the larger patches of grazed grassland because suitable habitats are lacking and distances to sources of immigration are too great (Cameron et al., 1980).
If consistent grazing over large surfaces does not contribute to the diversity of land snail communities, it could be argued that it maintains suitable habitats for rare species such as Trochoidea geyeri, a species which has had a very discontinuous range since the beginning of the Holocene (Magnin, 1989). Sheep grazing above 900 m asl has probably contributed to the expansion of open grasslands favourable for T. geyeri populations, while such habitats were formerly restricted to some rocky sites on the summit ridge. Is this habitat expansion necessary for species conservation? Pfenninger (Pfenninger et al., 1996; Pfenninger, 1997) found from computer-simulations that with the average survival-probability of T. geyeri being independent of habitat size, survival in small refuges is possible. Moreover, it must be noted that T. geyeri has probably survived in small rocky habitats of the mountain during the most wooded phases of the Holocene.
Another impact of pastoral management could be the passive dispersal of snails since sheep might play an important role in such dispersal as was coincidentally shown in a study of the dispersal of plant seeds (Fischer et al., 1996). The comparison of the intrapopulation genetic structures for different populations of T. geyeri shows a very weak degree of population subdivision in grazed habitats of the Luberon, which is attributed to passive dispersal by sheep (Pfenninger, 1997). The patchy distribution of several species (e.g. T. geyeri and C. unifasciata) in similar grassland habitats along flock trails also could suggest their passive dispersal by sheep.
Our present study is consistent with the other rare observations about the impact of grazing pressure on land snail communities. The number of xerophilic open-ground snails decreases when the grassland remains ungrazed, but a homogeneous grazed herb layer significantly reduces snail diversity and abundance.
A low richness and diversity of land snail communities is associated with large patches of grazed grassland, mainly with a continuous herb layer 5-cm high. On the other hand, the highest diversity is observed for communities living in scrublands or in smaller patches of grassland. Thus, heterogeneity seems to favour snail diversity both at the local and landscape scales. At the local level, the heterogeneity of vegetation (horizontal and vertical) and a complex cover of the soil surface enable more species to co-exist. At the landscape level, heterogeneity has an effect on land snail dispersal and on microclimate.
Sheep grazing contributes to the maintenance (or expansion) of suitable habitat for rare snail species, which otherwise could be restricted to very limited habitats. However, a restriction of their habitat may not threaten populations of such species since there is some genetic evidence that survival probability is independent of habitat size.
This work was performed through financial support from the CNRS Ecological Systems and Human Activities Programme. We would particularly like to thank Laurent Garde (CERPAM, Manosque) and Hervé Magnin (PNR du Luberon) for valuable information about pastoral management, Léon Sotgia for useful discussions during field-work, Frédéric Heurgué for his help in sample treatment, Erol Vela for assistance with plant identification and Carey Suehs for improving our English. This manuscript benefited from comments by M.V. Lomolino and two anonymous reviewers.
Corinne Labaune studied biochemistry and ecology at the Université de Bourgogne (Dijon, France) before preparing in 1997 a Diplôme d’Etudes Approfondies at the Université d’Aix-Marseille 3, Marseille. This work has provided the field data for this paper. She has completed (2001) a doctoral thesis on the ecology and parasitology of the invasive land snail Xeropicta derbentina in South-East France.
Frédéric Magnin is at the Institut Méditerranéen d’Ecologie et de Paléoécologie (C.N.R.S., Marseille) working on the ecology of both Recent and Quaternary land snails within the Mediterranean basin.
Table Appendix 1. Environmental variables used in the present study. Nine of them were selected for CCA analysis after a step-by-step procedure