Responses of ground-dwelling spider assemblages to changes in vegetation from wet oligotrophic habitats of Western France

While many arthropod species are known to depend, directly or indirectly, on certain plant species or communities, it remains unclear to what extent vegetation shapes spider assemblages. In this study, we tested whether the activity-density, composition, and diversity of ground-dwelling spiders were driven by changes in vegetation structure. Field sampling was conducted using pitfall traps in bogs, heathlands, and grasslands of Brittany (Western France) in 2013. A total of 8576 spider individuals were identified up to the species level (for a total of 141 species), as well as all plant species in more than 300 phytosociological relevés. A generalised linear model showed that spider activity-density was negatively influenced by mean vegetation height and mean Ellenberg value for moisture. Indices of diversity (ɑ, β, and functional diversities) increased with increasing vegetation height and shrub cover. Variables driving spider composition were mean vegetation height, dwarf shrub cover, and low shrub cover (results from a redundancy analysis). Spiders, some of the most abundant arthropod predators, are thus strongly influenced by vegetation structure, including ground-dwelling species. Although later successional states are usually seen as detrimental to local biodiversity in Europe, our results suggest that allowing controlled development of the shrub layer could have a positive impact on the diversity of ground-dwelling spiders.


Introduction
Globally, increases in plant species diversity or structural heterogeneity are often correlated with an increase in species richness of animals (Southwood et al. 1979;Madden and Fox 1997). The architectural or structural heterogeneity of plants, which is likely correlated with both plant-species diversity and productivity, can be an important determinant of arthropod diversity and abundance at different trophic levels (Lawton 1983). Many arthropod species depend, directly or not, on vegetation, and it consequently shapes their assemblages. This is especially obvious for phytophagous taxa, but has also been shown for other groups using vegetation as shelter or, in the case of spiders, for building their webs. Spider assemblages of vegetation-dwelling and web-building guilds are known to be shaped by vegetation structure, but ground-dwelling spiders are ideal models to test whether vegetation also drives community structure and composition of ground-active predators. While strong relationships have been reported previously between web-building spiders and vegetation (e.g. Ávila et al. 2017), other studies reported a weak effect of vegetation structure on spider diversity (Rodrigues et al. 2014) and no bottom-up effect of vegetation biomass on spiders (Lassau and Hochuli 2008;Lafage et al. 2014;Sousa-Souto et al. 2014). Ground-dwelling spiders are known to react to several local, abiotic factors, such as pH, disturbance, soil structure or moisture level (Schaefer 1990;Andersen 1995;Paquin and Coderre 1997;Pétillon et al. 2008). In this study, we tested whether their assemblages are shaped by changes in vegetation structure along a successional gradient. We took advantage of the monitoring systems of heathlands and grasslands (i.e. fairly stable habitats where landscape factors are likely less determinant than local factors for spiders: Horváth et al. 2015) in natural reserves to test whether a set of variables derived from the surrounding vegetation was able to predict changes in spider structure and composition at ground level.
Numerous studies have tried to understand the determinants of assemblages' composition and local species richness, i.e. α-diversity (Jiménez-Valverde et al. 2010). There are fewer studies dealing with β-diversity and functional diversity (McKnight et al. 2007), but their number has increased in recent years (e.g. Hendrickx et al. 2007;Boieiro et al. 2013;Braaker et al. 2013;Woodcock et al. 2013). We therefore chose to investigate assemblage composition, α-diversity, β-diversity, and functional diversity abundance. Spider abundance was expected to be positively influenced by vegetation complexity due to a higher abundance of prey. Species composition and α-diversity were expected to be positively influenced by local abiotic factors. Indeed, α-diversity describes within-habitat diversity (MacArthur 1965) and is mainly driven by local processes while β-diversity is generally thought to be driven by both local and landscape factors, the latter being the predominant factor for spiders and carabids (Lafage et al. 2015). We expected a weak or null link between vegetation structure and β-diversity.
Finally, functional diversity was expected to be positively linked to vegetation complexity as this would allow more guilds to coexist (Cardoso et al. 2011).

Study sites and habitats
Samples were taken in three Special Areas of Conservation (SAC) in the inner part of Brittany (Western France) at the head of drainage basins, including two natural reserves: the bogs of Langazel (LG) and the fens and heathlands of Lann Bern and Magoar-Penn Vern (LB) (Fig. 1). Both sites comprise colluvial and peaty plains crossed by streams and alluvial 'streaks'. They are composed of wet oligotrophic habitats including large areas of Ulicion minoris heathlands (EUR 28 4020) and Juncion acutiflori rush pastures (EUR 28 6410) sometimes in a mosaic with small patches of blanket bog communities (Oxycocco palustris-Ericion tetralicis).
Each set of sampling stations had a wide heterogeneity of structural forms resulting from several factors (de Foucault 1984;Clément and Aidoud 2009). While these vegetation forms have close spatial links, they nevertheless belong to different dynamical series mainly determined by edaphic conditions, notably water and trophic level. The diversity of past and current management practices (mostly mowing and grazing at different pressure levels) also explains the diversity of forms as encroachment or regressive vegetation stages.
Wet heaths have a progressive dynamic, ranging from dwarf shrub communities (Ulex gallii-Erica

Sampling design
Sampling of spiders took place in May and June 2013 over 30 consecutive days using pitfall traps. To compensate for this short sampling duration, we chose to increase spatial effort, as advised by Lövei & Magura (2011). Thus, forty-five plots were sampled (23 in LB site with 15 grasslands and 8 heathlands and 22 in LG site with 13 grasslands and 12 heathlands), with four traps per plot (100 mm diameter, filled with preservative solution (50% ethylene-glycol, 50% water) (Schmidt et al. 2006). Traps were placed 10 m away from each other to avoid interference between traps (Topping and Sunderland 1992).
Phytosociological surveys of 25 m 2 plots were conducted at each site to identify vegetation growthform types following Cristea, Gafta, and Pedrotti (2015) and Westhoff and van der Maarel (1978). In addition, five square plots of 0.25 m 2 each were set within this area. In each square plot, vascular species cover (%), mean and maximum vegetation height, and litter depth were measured.

Statistical analyses
Spider activity-density was defined as the mean number of individuals caught per trap. Spider -ɑ, diversity was estimated as mean species richness per plot. Spider β-diversity was estimated using a dissimilarity matrix (corresponding to Sørensen pair-wise dissimilarity) partitioned into its two components -species turnover (β t ) and nestedness (β n ) -following Baselga (2010)  included in the analyses despite its importance for ground spiders (Uetz 1979a).
Responses of spider abundance, α-diversity and functional diversity were tested using generalised linear models (GLMs) with binomial negative distributions and a stepwise model selection by AIC (Akaike 1974). Predictors were the same as for RDA. Linear, logarithmic, inverse, quadratic, cubic, power, compound, growth, and exponential regression were compared and the model with the highest R² was selected.
To identify the variables significantly influencing spider β-diversity, we performed a multiple regression analysis on the distance matrix of predictors (MRM) following the methods outlined in Legendre et al. (1994) using the ecodist R package (Goslee and Urban 2007). Predictors were the same as for RDA and GLMs. All statistical analyses were performed using R 3.2.3 (R Core Team 2015).

Results
A total of 8576 spider individuals belonging to 141 species (see taxonomic list in Table S1) were caught, with a high (93.38%) ratio of adult spiders. Assemblages were dominated by two lycosids (Pirata latitans and Pardosa pullata representing 24% and 23% of adult individuals respectively).
Spider activity-density was significantly and negatively influenced by mean vegetation height and the mean Ellenberg value for moisture (Table 3). Spider -diversity was significantly and positively ɑ, influenced by dwarf shrub and forb cover (Table 3). It was also positively influenced by Ellenberg index values for light and negatively by Ellenberg index values for moisture (Table 3). Spider functional diversity was positively influenced by dwarf shrub cover and Ellenberg index values for conductivity (Table 3). MRM was significant (P = 0.007, R² = 0.13). Spider β-diversity was significantly and positively influenced by mean vegetation height and by Ellenberg index value for pH, and negatively influenced by Ellenberg index value for nitrogen (Table 3).

Discussion
Our results suggest that the spider assemblages studied were strongly influenced by vegetation structure, even when considering ground-dwelling species. We found spider activity-density to be negatively influenced by vegetation height. This is in opposition with previous studies dealing with Spider activity-density was also negatively influenced by moisture, which is in accordance with previous findings (Uetz 1979b). In a literature review, Wise (1995) suggested that the abundance of spiders depends on three variables: wind, moisture, and temperature. More recently, Entling et al. Spider assemblages were best explained by variables reflecting vegetation structure, and more specifically vegetation closure and complexity (mean vegetation height, dwarf shrub cover, and low shrub cover). The importance of these variables is confirmed by the fact that we found dwarf shrub and forb cover to be positively related to -diversity and mean vegetation height to increasing β-diversity. ɑ, According to Entling et al. (2007), spider assemblages are mainly related to habitat type and depend on the shading, as well as the moisture, of habitats. Shading is obviously related to the development of shrubs and we logically found spider -diversity to be positively influenced by light. Thus, changes in ɑ, vegetation structure and shading may explain our results. Indeed, the role of habitat structure in itself has repeatedly been shown to determine the species richness of spiders (more so than the age of habitats, for example: Gibson, Hambler & Brown 1992;Hurd & Fagan 1992;Pétillon 2014). Greater vegetation complexity is likely to allow more species to co-exist by reducing inter-specific competition (Marshall and Rypstra 1999;Wise 2006). Vegetation closure is also positively linked to litter depth (e.g. Pétillon et al. 2008), which has been identified as a key variable explaining spider assemblages (Uetz 1979).
We found spider β-diversity was positively influenced by mean vegetation height and, therefore, vegetation closure. Spider β-diversity is considered higher in open habitats than in forests (Entling et al.

2007)
, and shading has long been identified as a major driver of β-diversity among habitats (MacArthur 1965). Nevertheless, at the beginning of the succession toward forested stages, the development of tall grasses and shrubs may increase β-diversity by providing new habitats for spiders. This is especially true for web-building species, but our results suggest this could also be true for ground-dwelling spiders. Studies dealing with small-scale drivers of arthropod β-diversity are scarce but compositional heterogeneity of spiders between samples is higher in young forest stands (Niemala et al. 1996). More recently, Sobek et al. (2009) found that habitat heterogeneity induced by tree diversity increases the β- Spider β-diversity was also influenced positively by pH and negatively by nutrient level. Forestation of moorland is often characterised by an increase in pH, nutrient level, and conductivity (Kampichler and Platen 2004). Thus, spider β-diversity seems to be positively influenced by the abiotic consequences of shrub and tree development.
Functional diversity was also positively influenced by dwarf shrub cover and conductivity, indicating that the impact of vegetation closure is linked to modification of vegetation structure and abiotic changes induced by it. This result is not surprising as functional diversity is considered more sensitive to environmental change than taxonomic diversity (Cadotte et al. 2009;Schirmel et al. 2012;Woodcock et al. 2014). This is also in accordance with Schirmel et al. (2016), who found functional diversity to be higher in woody than herbaceous sites.

Conclusion
Spider community assemblages appear to be driven by factors clearly linked to characteristics of vegetation and edaphic conditions such as vegetation height, shrub cover, pH, and soil richness. These parameters vary not only with vegetation dynamics but also according to vegetation management. In Europe, and especially in France, management strategies considered encroachment-and more generally natural dynamics-as negative trends for the conservation of agro-pastoral habitats leading to a loss of diversity and to the regression of their specific components. Management of wet heathlands and Juncus acutiflorus fens responds to this logic by increasing dwarf shrubby and herbaceous structures of vegetal communities through cutting or grazing operations. Conversely, our results suggest that allowing a controlled development of the shrub layer could have a positive impact on the diversity of certain groups such as ground-dwelling spiders. This clearly illustrate that "ecological value" of habitats and resulting management choices should be made using a pluritaxonomic approach.
However, these results do not allow us to reach a conclusion on the value of a particular management strategy according to the type of habitat. Further analyses are consequently needed to test for the effect of management modalities and habitat types.      Table S2).