Confounding legacies of land uses and land-form pattern on the regional vegetation structure and diversity of Mediterranean montane forests

Confounding legacies of land uses and land-form pattern on the regional vegetation structure and diversity of Mediterranean montane forests

Forest Ecology and Management 384 (2017) 268–278 Contents lists available at ScienceDirect Forest Ecology and Management journal homepage: www.elsev...

2MB Sizes 1 Downloads 11 Views

Forest Ecology and Management 384 (2017) 268–278

Contents lists available at ScienceDirect

Forest Ecology and Management journal homepage:

Confounding legacies of land uses and land-form pattern on the regional vegetation structure and diversity of Mediterranean montane forests Xavier Fortuny a, Sandrine Chauchard b,c, Christopher Carcaillet a,d,⇑ a

Ecole Pratique des Hautes Etudes (EPHE) & Paris Research University, 75014 Paris, France Université de Lorraine, Ecologie et Ecophysiologie Forestières (UMR INRA 1137), Faculté des Sciences et Technologies, BP70239, F54506 Vandoeuvre les Nancy, France c INRA, Ecologie et Ecophysiologie Forestières (UMR INRA 1137), F54280 Champenoux, France d Université Lyon 1, Laboratoire d’Écologie des Hydrosystèmes Naturels et Anthropisés (UMR 5023 CNRS), 69622 Villeurbanne, France b

a r t i c l e

i n f o

Article history: Received 10 July 2016 Received in revised form 31 October 2016 Accepted 1 November 2016

Keywords: Historical ecology Large herbivores Forest management Pastoralism Grazing Forest uses

a b s t r a c t Grazing by livestock and logging are among the main widespread forest uses in mountain forests. Centuries of these anthropogenic disturbances have shaped forests in Europe, especially in the Mediterranean region. Since the 19th century, pastoralism and forestry have experienced deep changes. The present study aims to assess and quantify the effects of these changes in forest uses along with environmental variables on forest vegetation structure and diversity in order to disentangle the role of landform versus land uses on ecological diversity in Mediterranean Pyrenees forests. Forest uses history was inferred from historical records and census whereas environmental variables, forest structure and diversity proxies were collected and measured based on 42 plots within 5 forests situated on the French-Spanish boundary. Both variables of environment and forest uses explained forest structure and diversity pattern. The most important part of the forest structure and the diversity pattern were due to the joined effect of forest uses, environmental and spatial variables. As expected, altitude, slope, temperature and precipitation were the main environmental drivers of forest structures and diversity. Among forest uses, the historical forestry treatment and the historical dominant livestock also strongly controlled the forest structure and diversity. Forest uses and environmental variables are closely related since forest use pattern is constrained by physical features. Covariation among forest uses and environmental variables produce confounding effects that make difficult the determination of the exact causal relationships. The contributions of forest uses, environmental and spatial variables were not disentangled due to large covariance between each variable. Ó 2016 Elsevier B.V. All rights reserved.

1. Introduction Land uses have globally changed owing to diverse driving forces (Lambin et al., 2001). Such changes can be divided into changes in land cover, e.g. forest transition following land use abandonment, (Rudel et al., 2005) or changes in practices in an extant cover, e.g. forest use changes, (Gimmi et al., 2008). The ecological consequences of land cover changes, notably of land abandonment, have been well studied (Chauchard et al., 2007; Niedrist et al., 2009). However, consequences of changes in traditional management practices in extant covers have been less investigated.

⇑ Corresponding author at: Ecole Pratique des Hautes Etudes (EPHE) & Paris Research University, 75014 Paris, France. E-mail addresses: [email protected] (X. Fortuny), [email protected] (S. Chauchard), [email protected] (C. Carcaillet). 0378-1127/Ó 2016 Elsevier B.V. All rights reserved.

In the Mediterranean region, forests are exploited for centuries, at least, by local societies to meet their needs. Grazing and logging are one of the main widespread forest uses in Mediterranean forests (Barbero et al., 1990; Fabbio et al., 2003) and these practices are gathered under the term ‘‘sylvo-pastoralism” when they take place in the same forest (Rapey et al., 2001). Past and present anthropogenic disturbances are powerful drivers shaping forest ecosystems (Dupouey et al., 2002; Foster et al., 2003; Chauchard et al., 2013), although it appears that natural disturbances such as wildfires can be a component of the ecosystem for long time (Leys et al., 2014). Since the 19th century, pastoralism, as traditional non-timber forest use, has experienced a net decrease in the northern Mediterranean basin, linked to rural exodus and abandonment of traditional practices (Barbero et al., 1990). Logging is still used but management practices have changed, notably due to the increasing use of fossil fuel during the second half of the 20th century that

X. Fortuny et al. / Forest Ecology and Management 384 (2017) 268–278

resulted in a coppice regime change (Barbero et al., 1990), i.e. decrease of firewood harvesting and charcoal production. It is expected that these forestry changes may result in transformation in vegetation features (Cierjacks and Hensen, 2004; Onaindia et al., 2004; Torras and Saura, 2008) including forest structure and species diversity or distribution (Bodin et al., 2013; Fortuny et al., 2014). These changes can affect resilience and stability of forest ecosystems notably facing changes in disturbance regimes such as fires. Indeed, structural and species diversity levels are important factors of resilience and stability of forest ecosystems (Lähde et al., 1999; Neumann and Starlinger, 2001; Pommerening, 2002). It is well known that environmental variables, e.g. soil or topography, have a key role on the organisation of vegetation pattern. But, environmental variables, as forest uses variables, are generally scale dependent (Levin, 1992) and their values are not randomly distributed in space, likely linked to the properties of natural landscape. Even if forest uses and environmental factors can each separately have an effect on ecosystems, their interactions are important in determining the structure, the composition and the dynamics of many systems (Paine et al., 1998; Kulakowski et al., 2011). To assess the function of forest uses on the ecosystems, it is essential to reconstruct the history of social disturbances (Axelson et al., 2002; Veski et al., 2005; Gimmi et al., 2008), while taking into account the variation in environmental factors and the contribution of the spatial structures (Schwarz et al., 2003; Gazol and Ibanez, 2009). This study aims to quantify and disentangle the functional properties of changes in pastoralism and forestry practices on the present-day vegetation structure and vascular plants diversity while taking into account environmental variables and spatial structures. The investigations were carried out in forests from the Mediterranean Pyrenees, which showed different histories of local uses, allowing precise comparisons in the same geographical and environmental context. The present paper (1) defines the forest uses and environmental variables that strongly influence current vegetation structure and plant diversity, (2) describes the effects of past and present forest use changes and environmental variables, and (3) determines covariation between these explanatory variables and to attempt to disentangle the relative contribution of historical forest uses, environmental variables and spatial structure. A previous paper highlighted that understorey plant assemblages characterized by their species list (presence/absence) and their life traits were selected according to both past land uses and landscape slopes (Fortuny et al., 2014). The present investigation is more complex by taking into account as response variable the whole plant community (under- and overstorey) and woody debris, characterized by quantitative values of diversity (e.g. richness), biomass (e.g. basal area), leaf area index and growth form. 2. Methods 2.1. Study area The study area consisted of five montane forests located in the eastern French Pyrenees on the northern slopes of a range along the French-Spanish boundary corresponding to the crests up to 1256 m above sea level (asl) at the Puig Neulós in the east and, 1450 m asl at the Roc de France in the west (Fig. 1). The two forests in the western part are forests of the municipalities of Maureillaslas-Illas and Céret. The two forests of Sorède and Laroquedes-Albères in the east are French state forests. Both municipal and state forests are managed by the French national forest administration (Office national des forêts, ONF), but the stakeholder is the city for municipal forests, and the ONF for state forests. The eastern forest of the study area is a Natural Reserve (La Massane NR) located in the municipality forest of Argelès-sur-Mer. This forest


was designated as Nature Reserve in 1973, but was already protected since the early 1950s. Approximately 10 ha of the inner forest zone were fenced-off by the RN managers since 1954 to exclude livestock grazing. The fence consists of a wire netting of 1.10 m height without fastening in the soil. No forestry intervention was conducted in the Massane forest during the studied period except black pine plantation for soil protection after the wildfire of 1880 (Chauchard et al., 2006). From the Middle Age to the beginning of the 20th century, the forests were widely used for grazing (cows, sheep, goats, horses, and pigs), for charcoal production, for firewood and wood for shipbuilding. During the 20th century, these forests were managed for timber production, protection against erosion, landscape preservation and also biodiversity conservation. Except in the Massane forest (Argelès-sur-Mer) where forestry has been abandoned notably for ecological conservation during the 20th, timber is the current forestry treatment in the study area. During the first half of the 19th century, sheep dominated livestock herds (Fig. SM.1). At the end of the 19th, livestock composition changed since sheep husbandry was largely substituted by cattle in the eastern part of the study area. Generally, livestock decreased since the 18–19th century transition (Fig. SM.1). However, each forest has its own grazing and forestry history that is analysed in the Supplementary Material, from the eastern to the western forest. Generally, the livestock of municipalities (i.e., the overall livestock from the municipality inhabitants) exhibited the same trend than municipal forest livestock authorized to graze in the forest, supporting the general trend observed at the forest scale (Fig. SM.1). On the contrary, no direct link can be made between trends of inhabitants and livestock. Indeed, the number of local inhabitants has increased in the study area during the study period independently from agricultural processes (Fig. SM.1). In this study, we focused on the lower montane belt of five selected forests. The mean altitude of sampling stands is about 800 ± 185 m asl, and the mean slope is about 26 ± 6° (Table 1). Bedrock is composed of shale and gneiss and soils belong to the acid (pH 4.3 ± 0.4) brunisol category (Servant, 1970). The lower montane vegetation belongs to the acidic beech forest type (Dupias, 1985). The woody community is dominated by beech (Fagus sylvatica L.) mixed with other broad-leaved trees, mainly white oak (Quercus pubescens Willd.), chestnut (Castanea sativa Mill.) and whitebeam (Sorbus aria L.). Individuals of maples (Acer campestre L.), lime tree (Tilia cordata Mill.), wild cherry-tree (Prunus avium L.) and holm oak (Q. ilex L.), are scattered throughout the forests. In the understory, beech is the main species along with European holly (Ilex aquifolium L.) in some places and, individuals of chestnut, whitebeam, common hazel (Corylus avellana L.), briar-root (Erica arborea L.) and Italian maple (Acer opalus Mill.) are also locally abundant. The herbaceous vegetation is dominated by Deschampsia flexuosa (L.) Trin., Galium pumilum Murray, Hieracium murorum L., Conopodium majus (Gouan) Loret, Luzula forsteri (Sm.) DC. and Veronica officinalis L. The climate is typical of the Mediterranean mountains, with warm and dry summers, and cool winters. Precipitations are mainly concentrated in spring and autumn, and summer precipitations are provided by thunderstorms. Climate data are not available in forests nor at their altitude, but simulated data based on geographic coordinates x, y, z (long., lat., alt.) can be inferred for each study site from the AUREHLY meteorological model (Benichou and Le Breton, 1987). Between 1961 and 1990, the mean annual precipitation is 1038 ± 70 mm, and mean annual temperature is 12.1 ± 0.7 °C, with the coldest month in February (5.6 ± 0.8 °C) and the warmest in July (20.4 ± 0.8 °C). The forest components of the study area are along an east-west gradient from the Mediterranean coast to the inner Pyrenees, which create a gentle climatic gradient with a decrease in temperatures (about


X. Fortuny et al. / Forest Ecology and Management 384 (2017) 268–278

Fig. 1. (a) Study area and, (b) sampling design.

0.4 °C on mean annual temperatures) and an increase in precipitations (about +60 mm on mean annual precipitations) from eastern to western part of the study area. The same climatic gradient is observed between the lowest and the highest parts of the study area. The mean annual temperatures decrease along this altitudinal gradient (about 0.7 °C) while the mean annual precipitations increase (about +80 mm). 2.2. Sampling design For this study, we used the same sites as in Fortuny et al. (2014). But the analyses are not based on the same set of

explanatory variables and, the response variables are here qualitative including both the over- and the understorey. Historical land use maps (Fortuny et al., 2014) were used to select stands covered by woodlands in the 18th to 19th centuries. Forty-two circular plots of 314 m2 were distributed in old stands of the five forests (Fig. 1a). Eight plots were located in Sorède, Laroque-desAlbères, Maureillas-las-Illas and Céret forests. Ten plots were located in the natural reserve: five in the fenced area and five in the unfenced area. A 2  5 m2 sub-plot was placed against the northern border of the plot (Fig. 1b) to measure shrub biomass. Four 1  1 m2 quadrats were located at the cardinal points of the plots to measure the herb biomass. In the front of


X. Fortuny et al. / Forest Ecology and Management 384 (2017) 268–278

Table 1 Summary of the environmental, grazing and logging explanatory variables and significant variables from forward selection (results of the permutation test): *p 6 0.05, **p 6 0.01, *** p 6 0.001. The variables excluded in the preselection analysis appear in bold. Explanatory variables


Mean value (± sd)


Permutation test

Environmental Altitude (m asl) Slope (°) Aspect Rocky cover (%) Soil pH Humus depth (cm) Humus type Soil type Organic matter (%) Mean Annual precipitation (mm) Mean annual temperature (°C) Precipitation during the driest month (mm) Emberger’s precipitation index Precipitation continentality index

Quantitative Quantitative Qualitative Quantitative Quantitative Quantitative Qualitative Qualitative Quantitative Quantitative Quantitative Quantitative Quantitative Quantitative

800 ± 185 26 ± 6

493–1127 15–37

F = 5.1323, ⁄⁄⁄ F = 2.2926, ⁄

28 ± 20 4.3 ± 0.4 3.0 ± 4.1

3–63 3.7–5.3 0.0–23.0

6.6 ± 1.9 1038.0 ± 70.1 12.1 ± 0.7 41.9 ± 10.5 135.2 ± 9.1 0.84 ± 0.07

2.5–10.4 936–1184 10.9–13.0 32–66 122–154 0.76–0.94

Grazing Trend in livestock density (whole period) Trend in livestock density (last period) Present livestock density (LU ha1) Trend intensity (whole period) Trend intensity (last period) Length of the last trend (year) Trend intensity progressiveness (whole period) Trend intensity progressiveness (last period) Dominant livestock species (whole period) Dominant livestock sp. (last period) Modern dominant livestock species

Qualitative Qualitative Quantitative Quantitative Quantitative Quantitative Quantitative Quantitative Qualitative Qualitative Qualitative

0.08 ± 0.11 0.89 ± 0.09 0.98 ± 0.11 34.76 ± 13.30 0.010 ± 0.002 0.033 ± 0.013

0.00–0.38 0.73–1.00 0.74–1.10 20–60 0.010–0.012 0.012–0.051

Forest management Time since the last logging (year) Historical forest management regime Last forest management regime Length of the last forest management regime

Quantitative Qualitative Qualitative Quantitative

each quadrat, a 0.25  0.25 m2 sub-quadrat was placed to sample soil. Eleven parameters of forest structure and diversity were recorded (Table 2). (1) The woody debris (m3 ha1) are the cumulated volumes of large (Motta et al., 2006) and fine woody debris (Genries et al., 2009). Large woody debris (ø > 7 cm) includes logs (fallen stems or branches) and snags (dead standing trees and

Table 2 Summary of the response variables. Forest structure and richness indicators

Mean value (±sd)



Woody debris (m3 ha1)

57.93 ± 43.80

14.72– 203.98 18.90– 73.40 0–7849 2.50e4– 480.00 0.28–2.46 2.30–5.80 9–49 0–5 1–6 0–7 0–5 3–27 1–11 0–7 1–4 0–9 0–9 0–4 0–4 0–4



Basal area (m ha



Shrub biomass (kg ha1) Herbaceous biomass (kg ha1) Structural diversity index Leaf area index Herb richness (Sherb) Shrub richness (Sshrub) Tree richness (Stree) Chamephytes Geophytes Hemicryptophytes Phanerophytes Therophytes Competitor Competitor-Ruderal Competitor-Stress-tolerant Ruderal Ruderal-Stress-tolerant Stress-tolerant

38.00 ± 12.34 444 ± 1295.76 67.35 ± 86.54 1.70 ± 0.43 4.11 ± 0.79 23.93 ± 9.43 1.76 ± 1.43 2.05 ± 1.15 1.64 ± 1.50 1.31 ± 1.18 12.88 ± 5.91 6.21 ± 2.34 1.76 ± 1.65 2.79 ± 0,90 3.98 ± 2.29 2.68 ± 1.93 1.14 ± 0.93 1.19 ± 1.21 1.71 ± 1.11

log(x) log(x + 1) p x (x1.95  1)/1.95 None log(x) None log(x) X2/3 X2/3 p x None log(x + 1) None X2/3 X2/3 log(x + 1) X2/3 X2/3

F = 2.2003, ⁄ F = 4.0649, ⁄⁄⁄ F = 9.0399, ⁄⁄⁄

F = 4.4146, ⁄⁄⁄

50 ± 45

9–127 F = 4.0658, ⁄⁄⁄

46 ± 46


branches) present in the 314-m2 plot. Fine woody debris (ø < 7 cm) were recorded using intersect-lines method: a series of five intersect lines, from 1.5 m to 7.5 m, were used for each plot. (2) Basal area (m2 ha1) was measured through circumference at 1.30 m above ground for all woody stems rooted in the plots. (3) Aboveground biomass of shrubs (dry weight, kg ha1) was measured on plant rooted in the 10-m2 sub-plots. (4) Aboveground biomass of herbs (dry weight, kg ha1) was measured in the four 1-m2 quadrats. (5) Leaf area index (LAI) was assessed by sixty-four measurements of light transmittance with an optical sensor across each plot (Cournac et al., 2002). (6–8) The species richness of herb, shrub and tree per plot is based on the species number of herbs, shrubs (ø < 5 cm at 30 cm above-ground) and trees (ø > 5 cm at 30 cm above-ground) rooted in the plot. (9) Shannon index, measured on woody plants, is a structural diversity index of woody diversity based on the basal area per species (Lindgren and Sullivan, 2001; Lenière and Houle, 2006). (10–11) Botanical and ecological traits of plant species, growth forms (Raunkiaer, 1934) and adaptive strategies (Grime, 1977), were obtained from the BASECO database (Gachet et al., 2005) completed by information from French and Spanish Floras (Bolòs et al., 1998; Rameau et al., 2008). These variables correspond, at the plot level, to the species number per modality of each growth forms and adaptive strategies traits. Vascular plants were listed three times (April, May and June 2009) in each plot in order to record maximum species, including vernal species.

2.3. Explanatory variables 2.3.1. Environmental variables Fourteen environmental variables were recorded (Table 1). (1) Altitude (m asl) was measured by altimeter (precision 15 m). (2)


X. Fortuny et al. / Forest Ecology and Management 384 (2017) 268–278

Slope inclination (degree) was measured by clinometer along the steepest diameter of each plot. (3) Aspect (compass) categorised into five classes, i.e. east, northeast, north, northwest and west (no south-facing exposure). (4) Rocky cover (%) was visually estimated for each plot. (5) Soil pH (KCl solution, De Schrijver et al., 2006) and (6) organic matter (%, loss on ignition method, (Konen et al., 2002) were measured at the laboratory in the organomineral horizon (Baize and Jabiol, 1995) of each sub-quadrat, i.e. 168 samples. In the northern sub-quadrat of each plot, (7) humus thickness (cm), (8) humus type (Baize and Jabiol, 1995), (9) soil type (AFES, 1995) were measured or identified from a soil profile. (10) Annual precipitation (mm), (11) mean annual temperature (°C), (12) precipitation during the driest month (mm), (13) Embergers precipitation index (Daget, 1977), (14) precipitation continentality index (Daget, 1977) were inferred from the AUREHLY meteorological model (Benichou and Le Breton, 1987). 2.3.2. Land use history Grazing history. Past and present census on inhabitant populations and livestock densities are used to assess the agricultural and silvicultural history (Chauchard et al., 2010). In the present study, various written sources were used to assess the land uses on varying scales. The grazing histories in the five forest components of the study area were assessed from the second half of the 19th to the end of the 20th century using administrative archives from ‘‘Archives départementales des Pyrénées-Orientales” at Perpignan, France: livestock municipality taxes and grazing authorisations were analysed. The information about present livestock densities in forests results from the current forest management plans (Office National des Forêts). Both sources provide information on the livestock number and species that grazed in the forests. For numerical analyses, the grazing history at the forest scale was used. Unfortunately, all livestock time series do not start at the same time. The oldest (Laroque-des-Albères) starts during the 1840s, while the most recent (Maureillas-las-Illas) starts in the 1900s. For analyses, the choice was made to use only the livestock time series from the 19–20th century transition in order to cover the same period for each forest. The livestock trends were assessed following two steps: (1) the identification of significant tipping points in the livestock time series using the ‘‘turnpoints” function of the ‘‘pastecs” package in the R software (Ibanez and Etienne, 2006), which brings out different periods characterized by an increase or a decrease of the livestock density. The last period highlighted by this method, i.e. the period ending at the present day, is hereafter called ‘‘last period”. (2) The significance of the trend for each period determined is tested using the MannKendall test (Chauchard et al., 2010). Ten proxies were used to describe present and historical grazing pressure (Table SM.1): (1) trend in livestock density over the last period (increase or decrease); (2) present livestock density in livestock unit (LU ha1) based on the equivalent of one adult dairy cow; (3) trend intensity over the whole period assessed by the livestock density ratio between the present and the 19th-20th century transition, (4) trend intensity during the last period assessed by the livestock density ratio between the present and the beginning of the last period, (5) length of the last trend (years) since the last change in livestock density trend, (6) trend intensity progressiveness over the whole period estimated by the ratio between the intensity of the trend over the whole period and the duration of the whole period; (7) trend intensity progressiveness during the last period assessed by the ratio between the intensity of the trend over the last period and the duration of the last period, (8) dominant livestock species in the forest over the whole period; (9) dominant livestock species in the forest during the last period, (10) present-day dominant livestock species in the forest. Forestry history. The past and current forest management plannings (Archives départementales des Pyrénées-Orientales, Office National des Forêts) were used to investigate the forestry history since the second half of the 19th century. These archives include both the reports on the previous management and the future guidelines. Others information sources, mainly reports and local historical literatures, were consulted to understand the wider context of the timber exploitation. Totally, five forestry treatments were identified in the forests under study: (1) no management, (2) coppice-with-standards, (3) timber, (4) coppice-withstandards and timber in equivalent proportion and, (5) coppice selection system which is defined as a cut which removes only larger shoots (Coppini and Hermanin, 2007). For numerical analyses, four proxies were used to describe past and present forestry (Table 1): (1) the time since the last logging (years), (2) the historical forestry treatment since the second half of the 19th century (3) the last forestry treatment applied and, (4) the duration of the last treatment (years). 2.3.3. Modelling of spatial variables Principal coordinate of neighbour matrices (PCNM) was used to create spatial explanatory variables usable in numerical analysis from the x- and y-coordinates of the plots centroids (Borcard and Legendre, 2002). This method provides a set of vectors that can be used to model spatial structure at several scales (Dray et al., 2006), notably gradients (Borcard and Legendre, 2002). The spatial vectors inferred by this method are ordered from broadest-scale, encompassing the whole sampled area, to finest-scale. The truncation threshold used was the maximum value of the minimum spanning tree calculated on the Euclidian distance matrix of the plots (Dray et al., 2006). The x- and y-coordinates of the plots centroids were added to the PCNM dataset to model linear trends (Gazol and Ibanez, 2010). 2.4. Data analyses A preselection of non-correlated variables was conducted. To assess correlation between variables, two Correspondence Analyses (CA) were carried out, first on the environmental variables and, second, on the land uses variables. Only preselected variables were used in the following analysis. The proxies of forest structure and diversity were included in a response variable dataset. Previously, all the proxies were assessed for normality (Shapiro-Wilk test) and transformed to meet this assumption when necessary (Table 2). The forest structure and diversity data set was standardized prior to statistical analyses to homogenize their range values and make each proxy equal weighted in the analyses. To assess the influence of land uses and environmental variables on forest structure and diversity pattern, two redundancy analyses (RDA) were carried out with the forest structure and diversity dataset as response variables and, separately, environmental and land uses datasets as explanatory variables (Legendre and Legendre, 1998). To provide an estimate of the best sets of explanatory variables for predicting forest structure and diversity, a forward selection of the explanatory variables (Blanchet et al., 2008) was performed on each RDA result. Two other RDAs were performed with the selected significant variables of each explanatory dataset. Permutation tests were used to assess the significance of the relationships between response and explanatory variables (Legendre and Legendre, 1998). The significance of each axis was assessed using permutation tests (Legendre et al., 2011). A third RDA analysis, with the forest structure and diversity dataset as response variables and spatial explanatory variables followed by a forward selection was used to determine the significant modelled spatial variables.

X. Fortuny et al. / Forest Ecology and Management 384 (2017) 268–278

To compare the relative contribution of each explanatory variable (modelled spatial, environmental and forest uses) on forest structure and diversity, a variation partitioning was conducted (Borcard et al., 1992; Peres-Neto et al., 2006). Only the explanatory variables selected in the previous steps were used in the variation partitioning. All the statistical analyses were performed with R 2.12.1 software (R-Development-Core-Team, 2010), together with the package ‘‘vegan” (Oksanen et al., 2010). 3. Results 3.1. Variable selection The two first axes of the CA of the environmental variables matrix capture 88.9% of the total inertia (Fig. 2a). The aspect characterizes the first, axis and the second axis opposes the rocky cover and the humus thickness. No strong correlation between environmental variables can be made from this analysis. All environmental variables were kept in the following analysis (Table 1). The two first axes of the CA of the forest use variables matrix capture 84.0% of the total inertia (Fig. 2b). Both the dominant species of livestock grazing in the forests over the whole period (spe) and the dominant species of livestock grazing in the forests during the last period (spe.lp) provide similar scores. Consequently, the rest of the analysis is based only on the dominant species of livestock grazing in the forests over the whole period ( The CA exhibits a correlation between the intensity of the trend both over the whole period (tri) and during the last period (tri.lp). Then, only the intensity of the trend over the whole period (tri) has been used in the rest of the analysis (Table 1). The CA on the matrix of forest use variables did not show clear correlation between the four forest management variables. Consequently, all these four variables were kept in the rest of the analysis (Table 1). 3.2. Effects of environmental, forest use and spatial variables on forest structure and diversity The three RDAs performed on preselected variables showed that environment (F = 2.2380, p = 0.001), forest use explanatory


variables (F = 3.2577, p = 0.001) and spatial explanatory variables (F = 1.6438, p = 0.001) significantly explained forest structure and diversity. 3.2.1. RDA with environmental variables Five variables were forward-selected and then included in the final model: altitude, slope, annual precipitation, mean annual temperature and precipitation during the driest month (Table 1). The first canonical axis of the RDA with forward-selected environmental variables accounted for 16.0% of the total variance, the second for 9.8%. The two first axis of the RDA were mainly defined by precipitation and temperature gradients as a consequence of altitudinal variation (Fig. 3a). According to these gradients, the down-left part of the RDA biplots corresponds to low elevation with relatively higher temperature and lower precipitation. Two series of forest structure and diversity parameters were grouped in this part of the diagram. The first one included three Raunkiaer’s growth form trait (chamephytes, geophytes and phanerophytes) and the Grime’s competitive strategies trait. The second included basal area, herb species richness, ruderal stress-tolerant Grime’s trait and therophytes Raunkiaer’s growth form trait. A slope gradient was also depicted along axis 1 and 2. Shrubs biomass, LAI, shrubs and trees species richness were positively correlated to steeper slopes. Another series of forest structure and diversity parameters were correlated to each other, and correlated to lower slope along with lowest precipitation during the dry season: ruderal and stress-tolerant Grime’s traits, hemicryptophyte Raunkiaer’s growth form trait and woody debris volume. 3.2.2. RDA with forest use variables The two forest use variables forward-selected were the dominant species of livestock over the whole period and the historical forestry treatment (Table 1). The first canonical axis resulting from the RDA conducted with the selected forest use variables explained 13.7% of the variation and the second axis 7.3%. Shrubs biomass, LAI, shrubs and trees species richness were positively correlated to sheep as historical dominant livestock (Fig. 3b). A second group of proxies included herb species richness, three Raunkiaer’s growth form traits (chamephytes, therophytes, phanerophytes and geophytes) and three Grime’s strategies traits

Fig. 2. Ordination scores of axes 1 and 2 of the correspondence analysis: (a) scatter diagram of environmental variables and (b) scatter diagram of forest-use variables. Environmental variables: slo: slope, alt: altitude, asp: aspect, roc: rocky cover, om: organic matter, hut: humus type, sot: soil type, hud: humus thickness, tem: mean annual temperature, con: precipitation continentality index, prd: precipitation during the driest month, pre: annual precipitation, emb: emberger precipitation index. Grazing variables: tre.lp: trend in livestock density over the last period, pld: present livestock density, tri: trend intensity over the whole period, tri.lp: trend intensity during last period, llt: length of the last trend, tip: trend intensity progressiveness over the whole period, tip.lp: trend intensity progressiveness during the last period, spe: dominant species of livestock grazing in forest over the whole period, spe.lp: dominant species of livestock grazing in forest during last period, actual dominant species of livestock grazing in forest. Forestry variables: tll: time since the last logging, hft: historical forestry treatment, lft: last forestry treatment, dlf: duration of the last forestry treatment.


X. Fortuny et al. / Forest Ecology and Management 384 (2017) 268–278

Fig. 3. Redundancy analysis biplots (first and second axes) of forest structure and diversity indicators data (black) constrained by (gray): (a) environmental variables and (b) forest-use variables. The first canonical axis accounts for 16.0% of the total variance, the second for 9.8%.

(competitive, competitive stress-tolerant and ruderal stresstolerant). This group was correlated to coppice-with-standards as the historical forestry treatment. Stress-tolerant Grime’s trait, hemicryptophytes and herb biomass was positively related to the forest abandonment status and to cattle as the historical dominant livestock. The structural diversity index reached its highest values in the forests where the historical forestry treatment was coppice selection system. 3.2.3. RDA with spatial variables Four of the twenty-six spatial variables were forward selected and included as spatial variables in the variation partitioning analysis. The spatial selected variables were, according to their explanatory power, the x-coordinates of the plots centroids, the second vector produced by the PCNM method (PCNM2), the sixteenth vector produced by the PCNM method (PCNM16) and the first vector produced by the PCNM method (PCNM1). The xcoordinates accounted for an east-west linear trend, the PCNM1 and 2 variables represent broad scale spatial patterns and the PCNM16 variable represent fine scale spatial pattern. 3.3. Contribution of explanatory variables on forest structure and diversity variation The total variation in forest structure and diversity patterns explained by variation partitioning of the entire explanatory selected variables was 33.6% (Fig. 4). The contribution of each explanatory subset variables was: 34.3% for the environmental variables, 30.7% for the spatial variables and 27.7% for the forest uses variables. Each of these contributions take into account the variation explained jointly with the other explanatory subset variables. When considering only the pure contribution of each explanatory subset variables, the ranking according to the amount of contribution is the same although scores drop: 3.7% for the environmental, 2.0% for the spatial and, interestingly, no contribution for the forest uses variables. The shared effect of the entire set of explanatory variables represented 25.9% of the total variation (Fig. 4).

Fig. 4. Partitioning of the variation in forest structure and diversity pattern using three subsets of explanatory variables: spatial, environmental and forest-uses. The variation percentages explained by the explanatory variables subsets and their interactions are indicated in the diagram. The total percentage variation explained by each explanatory variables subset is indicated in the caption at the bottom of the diagram.

histories (eleven variables related to grazing activities and four to forestry activities) and their landscape structures (fourteen environmental variables). The main finding of this study is that land use proxies and environmental variables depict a clear confounding effect on the forest structure and diversity (Fig. 4). The rest of the discussion will attempt to analyse and to disentangle the effect of land uses and spatial drivers on the Mediterranean forest vegetation.

4. Discussion

4.1. Forest uses and environmental drivers vs. forest structure and diversity

The stand structure and the plant diversity of the five forests (eleven response variables) were investigated in light of their land use

Precipitation (annual precipitation and precipitation during the driest month), temperature (mean annual temperature), altitude

X. Fortuny et al. / Forest Ecology and Management 384 (2017) 268–278

and slope were the main environmental determinants of the forest structure and diversity in our dataset (Fig. 2a), whereas the chief determinants of forest uses are the historical forestry treatment and the historical dominant livestock, i.e. cattle or sheep (Fig. 2b). These observations underline the legacies of both land uses and landscape pattern on vegetation composition (Davasse and Galop, 1994; Linares et al., 2011; Chauchard et al., 2013). The largest number of stress tolerant and ruderal species was encountered together in the same forests (Argeles-sur-Mer, data not shown). The coexistence of these two adaptive strategies reflects harsh conditions along with a regime of disturbance (Grime, 1977). The summer drought, characteristic of Mediterranean climate, would act as a stress factor for vegetation (Grime, 1977). Indeed, the precipitations during the dry season were the lowest in this part of the study area. Further, these forests were also historically grazed by cattle, while grazing is a wellknown disturbance for vegetation and, the level of disturbance is dependent on the size and behaviour of the grazing animals (Papachristou and Platis, 2011). For instance, cattle may cause greater trampling damage than sheep (Adams, 1975). In this sense, according to the Stress Gradient Hypothesis, the frequency of positive interactions between plants will growth with increasing abiotic stress or grazing pressure (Bertness and Callaway, 1994). The coexistence of different adaptive strategies (stress tolerant and ruderal species) may thus have resulted from this positive interaction. Herb richness was dependent on altitude (Fig. 3a), which is a well-known confounding variable of precipitation and temperature gradients and land uses. An increase in altitude resulted in a decrease in herb richness, which is a reported pattern (Bruun et al., 2006), although present-day distribution can depict sometimes inverse relations compared to paleo-data due to centuries of land uses that alter the natural species richness along the altitude gradient (Carcaillet and Brun, 2000). In the richest plots, ruderal stress-tolerant species and therophytes were more abundant and reflect a regime of disturbance (Grime, 1977; Kahmen and Poschlod, 2008). This regime of disturbance may be caused by coppice-with-standards treatment. Indeed, coppicewith-standards had the shorter rotation period compare to the other treatment applied in the forests of the study area. Forestry treatment also had an influence on forest structure. Notably, the structural diversity index was the highest in forests where the coppice selection system was historically conducted (Fig. 3b). Highest tree and shrub richness were found in forests historically grazed by sheep. Tree and shrub richness were also positively correlated to slope. The correlation between tree and shrub richness and slope may reflect a confounding effect between the use of sheep and the physical stress within the landscape (Fig. 5). Indeed, slope is an important factor controlling the distribution of large herbivore, with sheep favoured in the steepest forests rather than cattle that generally avoid steep slopes (Bailey et al., 2006). These results emphasized the confounding effects of the variables selected in this study that make difficult the precise determination of causal relationships. Confounding effects are inherent in the empirical field data (Vandvik and Birks, 2002) but are also related to the nature of the explanatory variables. This later point will be discussed in the following section. 4.2. Regional-scale linkages between forest uses, environment and spatial structure Both the environmental variables and the forest uses drive the observed pattern of forest structure and diversity (Fig. 4). The most important part of this variance was explained by the joined effect of environmental, forest uses and spatial variables (Fig. 4). Forest


Fig. 5. Relation between the slope of the forests and the dominant livestock species over the whole period. Significance of the livestock effect (multiple comparisons of means using the Tukey’s contrasts).

uses variables depend on environmental factors according to a landscape pattern. As mentioned above, sheep would have been favoured in steeper forests, while cattle were preferred on gentle slopes. Such relationships can partly explain the rather important shared effect between environmental, uses and spatial variables (25.9%, Fig. 4). This observation supports the fact that uses depend on environmental variables acting as dominant drivers (Taillefumier and Piégay, 2003), i.e. land-owners and land users adapted their practices to land-form, altitude and other regional landscape pattern. Land use gradients are frequently superimposed on an underlying environmental gradients, such as topography. Anthropogenic and natural factors covary because the latter influences the suitability of locations for human activities (Iverson, 1988; Pan et al., 1999), e.g. agriculture or forestry. Such relation has also been reported for protohistoric societies in the Pyrenees (Galop et al., 2007). Covariation among forest uses and environmental variables can make the decomposition of variation and thus the contribution of these variables difficult or impossible to demonstrate. This does not imply that these groups of variables are redundant or unnecessary variables. Rather, it highlights the fact that regional land uses pattern is constrained by physical factors, especially in rugged landscapes like mountains. The spatial structure of the forest uses variables indicates that usages were dependent on locality, i.e. their specific local physical properties (Fig. 5). On the eastern part of the study area, cattle grazed predominantly the forests since the 19th-20th century transition, while sheep always grazed the western part. Forestry treatments were also spatially structured since the end of the 19th century, since almost each forest was under its own silvicultural treatment within a regional pattern, likely due to different owners, i.e. the municipalities or the state. But before the 19–20th transition, forestry treatments were likely the same in the different sites. This large scale spatial structure is explained by the spatial variables selected that mirror landscape pattern at broad-scale and east-west trend. The coarse-grained nature of the forest uses and environmental variables may also explain the large mixed variance fractions (Svenning et al., 2004). Indeed, the historical dominant livestock, the historical forestry treatment and the topography (altitude, slope) exhibit large-scale patchiness. This large-scale patchiness makes intercorrelation within the set of uses, environmental and


X. Fortuny et al. / Forest Ecology and Management 384 (2017) 268–278

spatial variables unavoidable. Furthermore, livestock types are selective in their feeding and resting behaviour (Papachristou et al., 2005a, 2005b), especially cattle. This selectiveness creates fine-scale patterns of vegetation that vary considerably in space and time (Kohler et al., 2009) even under the same grazing regime, i.e. species and stocking rate. Therefore, the coarse-grained characteristic of the grazing history variables used in this study makes this variable less powerful to explain such fine-scale patterns. A minor variance of forest structure and diversity was solely explained by pure spatial effect, i.e. 2% (Fig. 4). This unexplained spatially structured variance can notably be attributed to nonmeasured variables (Jones et al., 2008), such as natural disturbances (e.g. fires) or non archived land uses. Some variations in the regional forest structure and diversity were explained by pure environmental contribution (3.7%), while no variation was explained by pure contribution of forest uses variables. Furthermore, the pure contribution of environmental variables was almost more than two fold higher than the pure contribution of spatial variables (3.7 vs 2.0%, Fig. 4) but remains low on the whole since it explains only 3.7% of the variance. The pure explanatory power of the forest uses may have been weakened because of the relatively recent historical period taken into account in our study. Indeed, land uses can have very long-lasting effect on ecosystems (Tatoni et al., 1994; Verheyen et al., 1999). For example, Dupouey et al. (2002) have highlighted the legacies of Roman land uses on present forest soil and biodiversity. Therefore, historical land uses such as forest continuity before the 19th century may explain the present pattern of forest structure and diversity. Several studies aimed to disentangle the contribution of environmental variables and spatial structures on forest attributes (e.g. Jones et al., 2008; Gazol and Ibanez, 2009) or the contribution of environment and land use (e.g. Vandvik and Birks, 2002; Palo et al., 2008). But few studies have simultaneously considered land use, environmental and spatial drivers (Svenning et al., 2004, 2009). The present study helps to understand the respective role of these variables in explaining forest structure and diversity patterns. It particularly highlights the confounding effects of uses and environment at the regional scale. This covariance is of crucial importance when studying the effect of uses on ecosystems. Indeed, because land uses and environmental drivers covary, land use is overestimated when assessed alone to explain regional ecological patterns. The total unexplained variance is ca. 66%, which is a rather high score. Factors acting on a smaller scale than those modelled by the PCNM method can explain a fraction of this unexplained variation. Indeed, the spatial structure resolution described by the PCNM variables is limited by the minimum distance between sample plots (Borcard and Legendre, 2002) and does not include the habitat heterogeneity that could be captured by the inter-plot variance (Fournier et al., 2012). Seed dispersal and seedling establishment are likely to produce spatial vascular plant community structure at relatively fine scales (Jones et al., 2008) and to control diversity (Gazol and Ibanez, 2010). Therefore, this biotic factor may play a role in the unexplained variation. Moreover, forest structure and diversity can be affected by other environmental/land uses unknown factors not taken into account in our study, which may create variance that acts as noise when we try to extract the effect of a given data set of explanatory variables. Land use factors in this study come from administrative sources that probably do not reveal all the practices (Davasse et al., 1997; Bürgi et al., 2013). Among the factors not taken into account and that may have an effect on forest structure and diversity, wild large herbivores would be of major importance, notably the wild boar those land uses is generally patchy. Further the developing regional roe deer populations might have contribute to select species. Indeed, wild large herbivores are determinant factors in forest ecological

functioning (Kuiters et al., 1996; Putman, 1996). For instance, wild boar generally present patch usage of the landscape and mostly impact processes occurring on the ground (tree regeneration, herbaceous layer richness, etc.), while roe deer are dispersed throughout the landscape and mostly acts by browsing of woody plants. 4.3. Functional versus taxonomic based studies This observation results from a quantitative approach of the whole vascular plant communities (under- and overstory) expressed in terms of diversity or biomass, and of organic debris. Interestingly, a study based on the same sampling design and plots (Fortuny et al., 2014) indicated that both land uses and site variables were less confounding when only understory plant species assemblages were characterized with a qualitative approach (presence/absence). This clearly demonstrates that the final conclusion of a study closely depends on the measured variables of experimental design and notably the forest compartments studied. The present study is more integrative and functional than the Fortuny et al. (2014) study that was more detailed in terms of taxonomic resolution. We thus observe that the qualitative approach based on species composition (previous study) is more powerful to split the effects of drivers on community, likely because some scattered or low biomass species are very sensible to the diverse land uses. 5. Conclusion Past forest uses have played an important role in shaping the present regional forest structure and diversity. Within the montane belt chiefly covered by beech in the eastern Pyrenees, the pastoralism associated to coppice was the main land uses form, but local differences in physical landscape pattern controlled local specificities on livestock, forestry and timing of land uses abandonment. The complexity of such interrelations makes difficult the possibility to isolate the contribution of each one. We cannot rule out that physical landscape drivers determined the historical land uses, and finally, both jointly controlled the present-day forest and diversity pattern. However, a different conclusion could be enhanced if understory composition (presence/absence of species) were taken into account (cf. Fortuny et al., 2014). However, with a functional based protocol investigating biomass and diversity, we show that land use legacies are difficult to disentangle from the function of landform pattern on the plant communities. This is likely due to the fact that land use depends on unavoidably from landform, except where landform results from land use (terraces, e.g. Tatoni et al., 1994). Our study emphasized the importance of taking into account environmental variables when analysing the effect of past and present land uses on forest ecosystems. Indeed, the covariance between land use, environmental and spatial variables is meaningful in studies aiming to describe anthropogenic disturbances and their effects on ecosystems. We also highlight the importance of using fine-scale land use data to assess the role of these variables in shaping vegetation patterns. Acknowledgments Financial support was partially provided by the FIREMAN program (ANR/ERA-net BiodivERsA) to CC. We thank Loic Birker and Benoit Brossier for their valuable help with fieldwork. We also thank the non-profit nature association called Les Amis de la Massane, who allowed us to work in the Massane Nature Reserve. We are grateful to Joseph Garrigue and Jean-André Magdalou, the

X. Fortuny et al. / Forest Ecology and Management 384 (2017) 268–278

Nature Reserve rangers, for providing valuable information about the reserve surveys and management. We finally thank Caroline Crawford who helped improved the English phrasing.

Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at 002. References Adams, S.N., 1975. Sheep and cattle grazing in forests: a review. J. Appl. Ecol. 12, 143–152. AFES, 1995. Référentiel Pédologique 1995. In: Baize, D., Girard, M.C. (Ed.). INRA Editions, Paris, 332p. Axelson, A., Östlund, L., Hellberg, E., 2002. Changes in mixed deciduous forests of boreal Sweden 1866–1999 based on interpretation of historical records. Landscape Ecol. 17, 403–418. Bailey, D.W., Gross, J.E., Laca, E.A., Rittenhouse, L.R., Coughenour, M.B., Swift, D.M., Sims, P.L., 2006. Mechanisms that result in large herbivore grazing distribution patterns. J. Range Manage. 49, 386–400. Baize, D., Jabiol, B., 1995. Guide pour la description des sols. INRA Editions, Paris. Barbero, M., Bonin, G., Loisel, R., Quézel, P., 1990. Changes and disturbances of forest ecosystems caused by human activities in the western part of the Mediterranean basin. Plant Ecol. 87, 151–173. Benichou, P., Le Breton, O., 1987. Prise en compte de la topographie pour la cartographie de champs pluviométriques statistiques: la méthode AURELHY. Agrométéorologie des régions de moyenne montagne. In: Les Colloques de l’INRA, vol. 39, pp. 51–69. Bertness, M.D., Callaway, R.M., 1994. Positive interactions in communities. Trends Ecol. Evol. 9, 191–193. Blanchet, F.G., Legendre, P., Borcard, D., 2008. Forward selection of explanatory variables. Ecology 89, 2623–2632. Bodin, J., Badeau, V., Bruno, E., Cluzeau, C., Moisselin, J.M., Walther, G.R., Dupouey, J. L., 2013. Shifts of forest species along an elevational gradient in Southeast France: climate change or stand maturation? J. Veg. Sci. 24, 269–283. Bolòs, O., Vigo, J., Masalles, R.M., Ninot, J.M., 1998. Flora Manual dels Països Catalans. Editorial Portic, Barcelona. Borcard, D., Legendre, P., 2002. All-scale spatial analysis of ecological data by means of principal coordinates of neighbour matrices. Ecol. Model. 153, 51–68. Borcard, D., Legendre, P., Drapeau, P., 1992. Partialling out the spatial component of ecological variation. Ecology 73, 1045–1055. Bruun, H.H., Moen, J., Virtanen, R., Grytnes, J.A., Oksanen, L., Angerbjörn, A., 2006. Effects of altitude and topography on species richness of vascular plants, bryophytes and lichens in alpine communities. J. Veg. Sci. 17, 37–46. Bürgi, M., Gimmi, U., Stuber, M., 2013. Assessing traditional knowledge on forest uses to understand forest ecosystem dynamics. For. Ecol. Manage. 289, 115– 122. Carcaillet, C., Brun, J.J., 2000. Changes in landscape structure in the northwestern Alps over the last 7000 years: lessons from soil charcoal. J. Veg. Sci. 11, 705–714. Chauchard, S., Pille, G., Carcaillet, C., 2006. Large herbivores control the invasive potential of nonnative Austrian black pine in a mixed deciduous Mediterranean forest. Can. J. For. Res. 36, 1047–1053. Chauchard, S., Carcaillet, C., Guibal, F., 2007. Patterns of land use abandonment control tree-recruitment and forest dynamics in Mediterranean mountains. Ecosystems 10, 936–948. Chauchard, S., Beilhe, F., Denis, N., Carcaillet, C., 2010. An increase in the upper treelimit of silver fir (Abies alba Mill.) in the Alps since the mid-20th century: a land use change phenomenon. For. Ecol. Manage. 259, 1406–1415. Chauchard, S., Guibal, F., Carcaillet, C., 2013. Land use legacies: multi-centuries years-old management control between-stands variability at the landscapescale in Mediterranean mountain forests, France. J. For. Sci. 59, 1–7. Cierjacks, A., Hensen, I., 2004. Variation of stand structure and regeneration of Mediterranean holm oak along a grazing intensity gradient. Plant Ecol. 173, 215–223. Coppini, M., Hermanin, L., 2007. Restoration of selective beech coppices: a case study in the Apennines (Italy). For. Ecol. Manage. 249, 18–27. Cournac, L., Dubois, M.A., Chave, J., Riéra, B., 2002. Fast determination of light availability and leaf area index in tropical forests. J. Trop. Ecol. 18, 295–302. Daget, P., 1977. Le bioclimat méditerranéen: caractères généraux, modes de caractérisation. Vegetatio 34, 1–20. Davasse, B., Galop, D., 1994. Impact des activités pastorales et métallurgiques sur les forêts d’altitude dans les Pyrénées ariégeoises (France). Rev. Valdotaine Hist. Nat. 48, 151–160. Davasse, B., Galop, D., Rendu, C., 1997. Paysages du Néolithique à nos jours dans les Pyrénées de l’Est d’après l’écologie historique et l’archéologie pastorale. In: La dynamique des paysages protohistoriques, antiques, médiévaux et modernes. APDCA Edition, pp. 577–599. De Schrijver, A., Mertens, J., Geudens, G., Staelens, J., Campforts, E., Luyssaert, S., De Temmerman, L., De Keersmaeker, L., De Neve, S., Verheyen, K., 2006.


Acidification of forested podzols in North Belgium during the period 1950– 2000. Sci. Total Environ. 361, 189–195. Dray, S., Legendre, P., Peres-Neto, P.R., 2006. Spatial modelling: a comprehensive framework for principal coordinate analysis of neighbour matrices (PCNM). Ecol. Model. 196, 483–493. Dupias, G., 1985. Végétation des Pyrénées. Centre National de la Recherche Scientifique, Paris. Dupouey, J.L., Dambrine, E., Laffite, J., Moares, C., 2002. Irreversible impact of past land use on forest soils and biodiversity. Ecology 83, 2978–2984. Fabbio, G., Merlo, M., Tosi, V., 2003. Silvicultural management in maintaining biodiversity and resistance of forests in Europe - the Mediterranean region. J. Environ. Manage. 67, 67–76. Fortuny, X., Carcaillet, C., Chauchard, S., 2014. Land use legacies and site variables control the understorey plant communities in Mediterranean broadleaved forests. Agri. Ecosyst. Environ. 189, 53–59. Foster, D.R., Swanson, F., Aber, J., Burke, I., Brokaw, N., Tilman, D., Knapp, A., 2003. The importance of land use legacies to ecology and conservation. Bioscience 53, 77–88. Fournier, T., Brossier, B., Gauthier, O., Carcaillet, C., 2012. Post-fire dynamics and spatial heterogeneity of woody debris associated with Pinus halepensis in an oak-pine Mediterranean ecosystem. Ecoscience 19, 356–363. Gachet, S., Vela, E., Tatoni, T., 2005. BASECO: a floristic and ecological database of Mediterranean French flora. Biodiv. Cons. 14, 1023–1034. Galop, D., Carozza, L., Marembert, F., Bal, M.C., 2007. Activités agropastorales et climat durant l’âge du Bronze dans les Pyrénées: l’état de la question à la lumière des données environnementales et archéologiques. In: Richard, H., Magny, M., Mordant, C. (Eds.), Environnements et cultures à l’âge du Bronze en Europe occidentale, pp. 107–119. Gazol, A., Ibanez, R., 2009. Different response to environmental factors and spatial variables of two attributes (cover and diversity) of the understorey layers. For. Ecol. Manage. 258, 1267–1274. Gazol, A., Ibanez, R., 2010. Variation of plant diversity in a temperate unmanaged forest in northern Spain: behind the environmental and spatial explanation. Plant Ecol. 207, 1–11. Genries, A., Morin, X., Chauchard, S., Carcaillet, C., 2009. The function of surface fires in the dynamics and structure of a formerly grazed old subalpine forest. J. Ecol. 97, 728–741. Gimmi, U., Bürgi, M., Stuber, M., 2008. Reconstructing anthropogenic disturbance regimes in forest ecosystems: a case study from Swiss Rhone Valley. Ecosystems 11, 113–124. Grime, J.P., 1977. Evidence for the existence of three primary strategies in plants and its relevance to ecological and evolutionary theory. Am. Nat. 111, 1169–1194. Ibanez, F., Etienne, P., 2006. Pastecs: package for analysis of space-time ecological series. R packag. version, vol. 1, pp. 3–4. Iverson, L.R., 1988. Land use changes in Illinois, ASA: the influence of landscape attributes on current and historic land use. Landscape Ecol. 2, 45–61. Jones, M.M., Tuomisto, H., Borcard, D., Legendre, P., Clark, D.B., Olivas, P.C., 2008. Explaining variation in tropical plant community composition: influence of environmental and spatial data quality. Oecologia 155, 593–604. Kahmen, S., Poschlod, P., 2008. Effects of grassland management on plant functional trait composition. Agr. Ecosyst. Environ. 128, 137–145. Kohler, F., Gillet, F., Gobat, J.M., Buttler, A., 2009. Seasonal vegetation changes in mountain pastures due to simulated effects of cattle grazing. J. Veg. Sci. 15, 143–150. Konen, M.E., Jacobs, P.M., Burras, C.L., Talaga, B.J., Mason, J.A., 2002. Equations for predicting soil organic carbon using loss-on-ignition for north central US soils. Soil Sci. Soc. Am. J. 66, 1878–1881. Kuiters, A.T., Mohren, G.M.J., Van Wieren, S.E., 1996. Ungulates in temperate forest ecosystems. For. Ecol. Manage. 88, 1–5. Kulakowski, D., Bebi, P., Rixen, C., 2011. The interacting effects of land use change, climate change and suppression of natural disturbances on landscape forest structure in the Swiss Alps. Oikos 120, 216–225. Lähde, E., Laiho, O., Norokorpi, Y., Saksa, T., 1999. Stand structure as the basis of diversity index. For. Ecol. Manage. 115, 213–220. Lambin, E.F., Turner, B.L., Geist, H.J., Agbola, S.B., Angelsen, A., Bruce, J.W., Coomes, O.T., Dirzo, R., Fischer, G., Folke, C., 2001. The causes of land use and land-cover change: moving beyond the myths. Global Environ. Chang. 11, 261–269. Legendre, P., Legendre, L., 1998. Numerical Ecology. Elsevier Science, Amsterdam. Legendre, P., Oksanen, J., terBraak, C.J.F., 2011. Testing the significance of canonical axes in redundancy analysis. Meth. Ecol. Evol. 2, 269–277. Lenière, A., Houle, G., 2006. Response of herbaceous plant diversity to reduced structural diversity in maple-dominated (Acer saccharum Marsh.) forests managed for sap extraction. For. Ecol. Manage. 231, 94–104. Levin, S.A., 1992. The problem of pattern and scale in ecology. Ecology 73, 1943– 1967. Leys, B., Finsinger, W., Carcaillet, C., 2014. Historical range of fire frequency is not the Achilles’ heel of the Corsican black pine ecosystem. J. Ecol. 102, 381–395. Linares, J.C., Carreira, J.A., Ochoa, V., 2011. Human impacts drive forest structure and diversity. Insights from Mediterranean mountain forest dominated by Abies pinsapo (Boiss.). Eur. J. For. Res. 130, 533–542. Lindgren, P.M.F., Sullivan, T.P., 2001. Influence of alternative vegetation management treatments on conifer plantation attributes: abundance, species diversity, and structural diversity. For. Ecol. Manage. 142, 163–182. Motta, R., Berretti, R., Lingua, E., Piussi, P., 2006. Coarse woody debris, forest structure and regeneration in the Valbona Forest Reserve, Paneveggio, Italian Alps. For. Ecol, Manage. 235, 155–163.


X. Fortuny et al. / Forest Ecology and Management 384 (2017) 268–278

Neumann, M., Starlinger, F., 2001. The significance of different indices for stand structure and diversity in forests. For. Ecol. Manage. 145, 91–106. Niedrist, G., Tasser, E., Luth, C., Dalla Via, J., Tappeiner, U., 2009. Plant diversity declines with recent land use changes in European Alps. Plant Ecol. 202, 195–210. Oksanen, J., Blanchet, F.G., Kindt, R., Legendre, P., O’Hara, R.G., Simpson, G.L., Solymos, P., Henry, M., Stevens, H., Wagner, H., 2010. Package ‘‘vegan” Community Ecology Package. In R-Development-Core-Team (Ed.), The-RFoundation-for-Statistical-Computing, Vienna, Austria. Onaindia, M., Dominguez, I., Albizu, I., Garbisu, C., Amezaga, I., 2004. Vegetation diversity and vertical structure as indicators of forest disturbance. For. Ecol. Manage. 195, 341–354. Paine, R.T., Tegner, M.J., Johnson, E.A., 1998. Compounded perturbations yield ecological surprises. Ecosystems 1, 535–545. Palo, A., Linder, M., Truu, J., Mander, U., 2008. The influence of biophysical factors and former land use on forest floristic variability on Saaremaa and Muhu islands, Estonia. J. Nat. Conserv. 16, 123–134. Pan, D.Y., Domon, G., de Blois, S., Bouchard, A., 1999. Temporal (1958–1993) and spatial patterns of land use changes in Haut-Saint-Laurent (Quebec, Canada) and their relation to landscape physical attributes. Landscape Ecol. 14, 35–52. Papachristou, T.G., Dziba, L.E., Provenza, F.D., 2005a. Foraging ecology of goats and sheep on wooded rangelands. Small Ruminant Res. 59, 141–156. Papachristou, T.G., Platis, P.D., 2011. The impact of cattle and goats grazing on vegetation in oak stands of varying coppicing age. Acta Oecol. 37, 16–22. Papachristou, T.G., Platis, P.D., Nastis, A.S., 2005b. Foraging behaviour of cattle and goats in oak forest stands of varying coppicing age in northern Greece. Small Ruminant Res. 59, 181–189. Peres-Neto, P.R., Legendre, P., Dray, S., Borcard, D., 2006. Variation partitioning of species data matrices: estimation and comparison of fractions. Ecology 87, 2614–2625. Pommerening, A., 2002. Approaches to quantifying forest structures. Forestry 75, 305–324. Putman, R.J., 1996. Ungulates in temperate forest ecosystems: perspectives and recommendations for future research. For. Ecol. Manage. 88, 205–214. R-Development-Core-Team, 2010. R: a language and environment for statistical computing. In: R-Development-Core-Team (Ed.), The-R-Foundation-forStatistical-Computing, Vienna, Austria. Rameau, J.C., Mansion, D., Dumé, G., Gauberville, C., Bardat, J., Bruno, E., Keller, R., 2008. Flore forestière française - Guide écologique illustré - Tome 3 Région Méditerranéenne. Institut pour le Développement Forestier, Paris.

Rapey, H., Lifran, R., Valadier, A., 2001. Identifying social, economic and technical determinants of sylvopastoral practices in temperate uplands: results of a survey in the Massif Central region of France. Agr. Syst. 69, 119–135. Raunkiaer, C., 1934. The Life Forms of Plants and Statistical Plant Geography. Clarendon Press, Oxford, UK. Rudel, Y.K., Coomes, O.T., Moran, E., Achard, F., Angelsen, A., Xu, J., Lambin, E., 2005. Forest transitions: towards a global understanding of land use change. Global Environ. Change 15, 23–31. Schwarz, P.A., Fahey, T.J., McCulloch, C.E., 2003. Factors controlling spatial variation of tree species abundance in a forested landscape. Ecology 84, 1862–1878. Servant, J., 1970. Carte pédologique de France - Argelès-sur-Mer, Perpignan. In: Agronomique, I.N.d.l.R. (Ed.), INRA - Service d’Etude des Sols, Montpellier. Svenning, J., Baktoft, K., Balslev, H., 2009. Land use history affects understorey plant species distributions in a large temperate-forest complex, Denmark. Plant Ecol. 201, 221–234. Svenning, J.C., Kinner, D., Stallard, R.F., Engelbrecht, B., Wright, S.J., 2004. Ecological determinism in plant community structure across a tropical forest landscape. Ecology 85, 2526–2538. Taillefumier, F., Piégay, H., 2003. Contemporary land use changes in prealpine Mediterranean mountains: a multivariate GIS-based approach applied to two municipalities in the southern French Prealps. Catena 51, 267–296. Tatoni, T., Magnin, F., Bonin, G., Vaudour, J., 1994. Secondary successions on abondoned cultivation terraces in calcareous Provence. 1. Vegetation and soil. Acta Oecol. 15, 432–447. Torras, O., Saura, S., 2008. Effects of silvicultural treatments on forest biodiversity indicators in the Mediterranean. For. Ecol. Manage. 255, 3322–3330. Vandvik, V., Birks, H.J.B., 2002. Partitioning floristic variance in Norwegian upland grasslands into within-site and between-site components: are the patterns determined by environment or by land use? Plant Ecol. 162, 233–245. Verheyen, K., Bossuyt, B., Hermy, M., Tack, G., 1999. The land use history (1278– 1990) of a mixed hardwood forest in western Belgium and its relationship with chemical soil characteristics. J. Biogeogr. 26, 1115–1128. Veski, S., Koppel, K., Poska, A., 2005. Integrated palaeoecological and historical data in the service of fine-resolution land use and ecological change assessment during the last 1000 years in Rouge, southern Estonia. J. Biogeogr. 32, 1473– 1488.