This dataset contains data on vascular plant diversity and community composition of the understory in mature broadleaf and conifer forest stands in the Sologne natural region, Central France. The objectif of the project was to study the effects of large wild ungulates on forest biodiversity using a natural and anthropogenic gradient of ungulate pressure.

Study area The study area was located in the Sologne Natural Region in the center of France. Sologne is characterized by flat topography and poor sandy soils on top of an impermeable clay layer, and the area is thus subject to frequent winter floodings and summer droughts are recurrent. The dominate land use is forest (53%), of which the majority is recent forests stemming from spontaneous colonization of abandoned land, but also the result of massive afforestation programs during the second half of the 19th century. Deciduous trees represent approximately 77% of the forest cover (Quercus robur: 39%, Quercus petraea: 14%, Betula pendula 9%), while the remaining 23% is made up of coniferous tree species, mainly Pinus sylvestris (13%). The three most common forest stand compositions are monospecific stands of Q. robur (23%), Q. petraea (9%) and P. sylvestris (9%).

In Sologne, population densities of large ungulates (red deer, Cervus elaphus, roe deer, Capreaolus capreolus, and wild boar, Sus scrofa) have shown a marked increase in number during the past decades, as elsewhere in France and Europe. No absolute estimates of ungulate densities are available for the study area, but hunting statistics for the three species are among the highest in France, and hunting bags for the 2004/2005 hunting season for red deer, roe deer and wild boar averaged 0.4 ± 0.5 (mean ± SD), 1.9 ± 1.4 and 3.7 ± 3.7 individuals per km², respectively (source: the French national agency for wildlife, ONCFS). No wild large predators were present in the study area.

Sampling design We made use of four fenced, five partially fenced and ten unfenced private forest properties to set up an experimental gradient of wild ungulate densities to test their effects on the structure, composition and diversity of plant communities. We selected private properties with at least 100 ha land and where forest was the dominant land use (60-100% forest). Average area of forest was 295 ± 165 ha (mean ± SD; range = 91-703) per land property (Appendix 1: Table S1). A preselection of private properties all over the study area was established by contacting the Centre régional de la propriété forestière (CRPF) d’Ile-de-France et du Centre (Regional public organizations for private forest owners) who helped out to suggest potential land owners willing to participate to the study. We then set up a list of equal number of fenced and unfenced properties in different parts of the study area. Private land owners were then contacted by telephone in order to obtain their permission to carry out field observations on their land property. We stopped contacting landowners once we had obtained the permission from ten unfenced private properties and that were well spread over the study area. Due to difficulties in obtaining permission from land owners with fenced properties, we did not reach a completely balanced design. A completely random sample of private properties would not have been possible due to the high degree of reluctance among private landowners to give their permission to carry out observations on their land property. For each land property, we randomly assigned five study plots stratified according to the proportion of area of deciduous and coniferous forest stands. A buffer zone of at least 50 m was applied to each forest stand nearby roads and open areas. A field visit was made before final selection to assure that the study plot was not situated in recently harvested forest stands or nearby forest edges (<30 m). We also rejected coppice forest stands.

Ungulate pressure data We used the observed intensity of major foraging activities by ungulates (browsing for deer and rooting for wild boar) to situate sampling points along a gradient of increasing ungulate densities. As mark-recapture data was not available for our study sites, we could not base our gradient on absolute ungulate densities, but situate the study sites on a relative scale based on the above-mentioned indices (and detailed below) of ungulate activities.

Deer browsing pressure was quantified at each sampling point by comparing forage use and availability based on resource selection theory. Forage use and availability were estimated on winter browse (woody and semi-woody vegetation) accessible to deer (0-2 m) in late winter (March) before the start of the growing season. Forage use and availability were estimated on three 40 m2-circular subplots per study plot, each situated at a distance of 14 m from the center of the study plot. For each species, forage availability was quantified by estimating the percentage of plant cover (i.e. the horizontal projection of shoots, twigs and branches and thus a proxy of the total number of “bites” available), while forage use was quantified by estimating the percentage of available shoots browsed (i.e. the percentage of actual “bites”). Visual estimates of forage use and availability were then attributed to one of six classes (0-1%, 1-5%, 5-20%, 20-50%, 50-75%, 75-100%), converted to mid-point values for statistical analyses. For each circular plot (40 m2), we then calculated a browsing pressure index, B, based on the sum of the forage consumed weighted by forage availability. We then used the mean value of B for the three subplots as a representative measure of browsing pressure at study plot.

Wild boar rooting was quantified at the sampling points by visually estimating the percentage of soil disturbed by wild boar rooting behavior. Observations of wild boar rooting were carried out in late winter at the same time as observations of deer browsing and were estimated on the same three circular plots used for estimating deer browsing pressure (40 m2). The mean percentage of wild boar rooting for the three subplots was used as a representative measure of wild boar rooting at the study plot level.

Vegetation data At each sampling point, we recorded all vascular plant species according to their presence in two vertical understory vegetation layers. We defined the two vegetation layers in relation to their accessibility to one or both of the two deer species present in our study area: low understory layer accessible both to roe and red deer (up to 130 cm in height) and high understory layer accessible only to red deer (from 130 cm to 200 cm in height). However, the data for the two vegetation layers were merged (see below). We attributed plant cover values, to each species and for each vegetation layer, based on visual estimates to the nearest percent for common species (plant cover >1%) and to the nearest promille for rare species (plant cover <1%). Vegetation sampling was carried out by five experienced botanists (nBot1 = 75, nBot2 = 59, nBot2 = 26, nBot2 = 20, nBot2 = 10) that formed mixed teams composed of two observers (A and B) in order to minimize observer effects. A team was composed of either botanists A (n = 36) or B (n = 20) and any of the other botanists, or both of them (n = 39). In order to harmonize the sampling effort among study plots, teams spent at least 30 minutes of actively searching new species, excluding extra time that was added for species identification problems and estimations of plant cover values. We used a relatively large sample plot size (1 000 m2) as we were interested in capturing not only common but also rare species, while limiting the size in order to include only one forest habitat type.

Plant functional traits We used a trait-based approach to determine any correlations between ungulate activities and understory plant community structure, composition and diversity. From the vegetation data, we derived three families of response variables: (i) species density and (ii) plant cover for qualitative traits (including plant functional groups and categorical habitat preferences), and (iii) community-weighted means (CWM) for quantitative trait values (including quantitative habitat preferences). Data on response variables were calculated at each sampling point for the overall plant community, and separately for four plant functional types (trees, shrubs, forbs and graminoids).

Data on plant functional traits were extracted from four main sources: the LEDA and BiolFlor plant trait data bases, and the floras “Flore Forestière Française”and “Nouvelle flore de la Belgique, du Grand-Duché de Luxembourg, du Nord de la France et des Régions voisines”. Missing data was added by consulting the scientific literature. We used plant functional trait data of categorical traits of plants (plant life span, plant leaf vertical distribution and spiny plants) and seeds (seed life span, frugivory seed characteristics and seed appendages), quantitative traits of plants (specific leaf area, canopy height, plant leaf vertical distribution, plant life span) and seeds (seed mass, seed releasing height, seed longevity, seed shape), as well as qualitative and quantitative plant habitat preferences (forest history, successional stage, EUNIS main habitats, Ellenberg’s indicator values and Grime’s CSR-scheme). Observed differences at the community level of these plant and seed characteristics among study plots are likely to inform about the plant community’s response to various ungulate activities related to trophic interactions (e.g. direct effects of grazing, browsing and frugivory) and engineering effects (e.g. direct effects of trampling, rooting, seed dispersal).

Site characteristics In order to take into account possible confounding factors, known to be strong determinants of vegetation composition, we made a forest stand description and took soil samples at each sampling point. Forest stands were described by measuring the dominant tree height within a radius of 18 m from the sampling point, the basal area at 1.3 m (breast height, BAbh) and canopy openness. BAbh was estimated using point sampling methodology and by separately estimating BAbh for broadleaves (BAbroadleaves), conifers (BAconifers) and coppice (BAcoppice), which allowed us to calculate overall BAbh and canopy mixture (varying from 0%, pure stand, to 50% , equally mixed stand of broadleaves and coniferous tree species). We visually estimated canopy openness at sampling points along three radial transects (one measurement every two meter along the 16 m-long transects, a total of 27 measurements per sampling point). We also determined forest history at sampling points distinguishing between recent and ancient forests. Forest history was derived from three times series of historical maps (Carte d'État-Major) drawn between (i) 1820 and 1866 and aerial photographs for the periods (ii) 1947-1950 and (iii) 1975-1980. We classified forest stands at sampling points as ancient forests whenever continuous forest cover was observed for all three time series (i.e. forest as land use since at least 1820-1866), while stands were classified as recent forest whenever any other form of land use was described at the sampling points for any of the three time series.

Soil samples of about 500 g were taken of the mineral soil at 20 cm depth at a distance of 10 m from the point center and in three different directions (0°, 120° et 240°). Soil samples were sent to the Soil Analysis Laboratory of INRA, Arras France. Soil samples were analyzed for soil texture (particle-size fractions in percentage of sand, silt and clay), cation exchange capacity (CEC, cmol+/kg), organic carbon (C, g/kg), total nitrogen (N, g/kg), and extractable soil phosphorus (P2O5, g/kg). Total organic carbon and total nitrogen content in the soil was measured after dry combustion (ISO 10694, ISO 13878), and the cation exchange capacity (CEC) was determined by extracting exchangeable cations (Al3+, Ca2+, Fe2+, K+, Mg2+, Mn2+, Na+) using a hexamminecobalt trichloride solution (ISO 23470). Extractable soil phosphorus was determined using Duchaufour’s method, which is a method appropriate for acidic forest soils. Soil pH was measured at our own laboratory using a pH-meter (Eutech Instruments Eco Scan 6+) in a 1:5 (volume fraction) suspension of soil in 1 mol/l potassium chloride solution (pHKCl) following the ISO 10390 standard.

The majority of forest stands were recent forests (n = 76), while the remaining fifth were classified as ancient forests (n = 19). Two thirds of sampling points were situated in high stands dominanted by broadleafs (n = 63) composed of oak trees (Quercus petraea, Q. robur), while one third were in high stands of coniferous trees (n = 32) composed of pine trees (Pinus sylvestris, P. nigra subsp. laricio). Overall mean basal area was 22.6 ± 0.9 m2/ha and the mean dominant tree height was 23.4 ± 0.4 m. Coppice stools of Betula sp., Carpinus betulus, Castanea sativa, Corylus avellana and Quercus sp. were present in the understory at about one third (n = 35) of the sampling points with a mean basal area of 6.5 ± 1.2 m2/ha. Soils were representative of the region characterized by low soil fertility of N (0.4 ± 0.04 g/kg) and P2O5 (0.03 ± 0.004 g/kg), low CEC (1.8 ± 0.2), and high acidity (4.2 ± 0.03).

R, 4.3.0