How to promote spider diversity of heathlands: impact of management intensity

We performed this study on 15 heathland plots 20 m × 60 m in size in the Lüneburg Heath, Northern Germany. All plots were unmanaged for 10 years when the management practices were applied. They differed in previous land use with all but 4 plots being previously used by the British and Canadian militaries from 1963 to 1994, mainly by tanks and armored vehicles (Table S1). Mowing and scarification, two heathland management practices with varying intensity, were implemented at 5 plots each in March–April 2018, while the remaining 5 plots were left unmanaged as controls. In the plots subject to mowing, vegetation was mowed to a height of approximately 5 cm and the mowed material was removed as completely as possible. Scarified plots were mowed in the same way, but were subject to a subsequent mechanical removal of the dense cushions of moss that have developed under the canopy of heather, resulting in the creation of bare soil patches. Moss removal was accomplished with a rotary scarifier.

Spiders were sampled with pitfall traps a year after the management was implemented between 16.04.2019 and 02.04.2020. We set 5 pitfall traps at each plot in one linear transect, with 10 m distance between traps and the plot edge. Traps were transparent 500 ml plastic cups with a diameter at entry of 9.5 cm and a depth of 9.7 cm, placed flush to soil and covered with a 20 cm × 20 cm roof roughly 3–4 cm above the traps to avoid vertebrate bycatch. In each trap we added 150–200 ml of Renner solution (40% ethanol, 20% glycerine, 10% acetic acid and 30% water) as a killing agent. Traps were emptied in 2 week intervals with the exception of the winter period (02.10.2019–02.04.2020) when they were emptied once a month. All adult spiders were determined to species level using the identification key by Nentwig et al. (2024) following the nomenclature of the World Spider Catalog (2024). After species identification the fresh biomass of all spiders was estimated by using mean body lengths for both sexes separately (Nentwig et al. 2024), with the body length-fresh biomass equations of Penell et al. (2018). In addition to estimated biomass, we used spider functional guilds (sensing, sheet, space and orb weavers, specialists, ambush, ground, and other hunters) according to Cardoso et al. (2011) and the phenological length (in months) of adult spider activity (Nentwig et al. 2024) to calculate spider functional diversity. These traits have been previously shown to significantly affect spider resource use and are, therefore, important in determining their functional effect (Cardoso et al. 2011; Schuldt et al. 2014). Lastly, in October 2019 we visually estimated the percentage cover of heather, heather litter, grass, and bare soil in a 2 m radius around each trap, as well as the distance to the nearest tree in order to consider the fact that there was some heterogeneity in habitat structure between plots before treatment application.

We analysed all data at the trap level, with data from all periods pooled together. Spider abundance was calculated at a per trap/day basis to account for traps being open for different durations during different periods (13–37 days). Furthermore, species diversity was calculated using coverage-based rarefaction and extrapolation of species diversity (Hill numbers) with the iNEXT R package (Hsieh et al. 2016). Functional richness and divergence (Villéger et al. 2008) were calculated with the R package FD (Laliberté et al. 2014) using species biomass, guild and phenological length as traits.

Abundance per trap/day (total abundance and abundance of three threatened species collected with adequate abundance), Hill numbers (q = 0, 1, and 2), FRic and FDiv were used as response variables in linear mixed effect models with plot as a random effect. Abundance per trap/day and the Hill numbers were log(x + 1) transformed to improve modeling assumptions. We conducted the modeling in two successive steps to account for effects of management practice and previous military presence, as well as effects of habitat structure which can be independent of management practice. The first model only had management (unmanaged, mowed, scarified) and military presence/absence as fixed effects, while the second one replaced management and military presence with habitat structure (heather cover, heather litter cover, grass cover, bare soil cover, as well as the distance to the nearest tree). Additionally, we ran models for the effects of management and previous land history on habitat structure elements. Grass cover and heather litter cover were log (x + 1) transformed for these models to improve model fit. All models were estimated with the nlme R package (Pinheiro et al. 2020). After running the models, we confirmed that multicollinearity between variables was low using VIF (≤ 5), calculated with the R Package car (Fox and Weisberg 2018). For the first model, we used the Tukey’s honest significant difference (HSD) post hoc test with p-values adjusted with the Holm method using the multcomp R package (Hothorn et al. 2008) to test for significant differences between management practices and military presence/absence. For the second model we used a stepwise selection procedure based on AICc (Burnham et al. 2010) in order to acquire the most parsimonious models with the best model fit.

Furthermore, we analysed the similarity between spider assemblages at the plot level with non-metric multidimensional scaling (NMDS). Similarity was based on the Morisita-Horn index of square root-transformed abundance data (Jost et al. 2011). A stable solution was computed from multiple random starting points on the basis of two reduced dimensions. We tested for correlations of the ordination axes with habitat structure variables as environmental vectors, as well as management and military presence/absence as environmental factors utilizing permutational multivariate analysis of variance (PERMANOVA) using distance matrices (Anderson 2001) in the vegan R package. Moreover, we analysed the multivariate homogeneity of group dispersions as a measure of β diversity (Anderson 2006). Lastly, we analysed the ecological preference of species with the phi coefficient of association (Chytrý et al. 2002) allowing us to distinguish between species with a preference or aversion to a specific management. We used significant phi coefficient values > 0.25/ < −0.25 as a threshold. All analyses and figures were made in R version 4.3.3.

Habitat structure differed significantly between management practices with the exception of grass cover and distance from the nearest tree (Tables S2, S3). Bare soil was highest in scarified plots followed by mowed and then unmanaged plots (Fig. 1A). Heather cover was highest in unmanaged plots, followed by mowed and finally scarified plots (Fig. 1B), while heather litter cover was higher in mowed than unmanaged plots (Fig. 1C). Lastly, in terms of previous land history, plots with no previous military presence had higher heather litter cover and bare soil cover (Table S2).

We captured a total of 15 691 spiders, 12 419 of which were adults (79%) and were identified to 138 species (Table S4). Ground hunters from the Lycosidae family dominated, with the open habitat species Pardosa nigriceps (1213/9.8%) and Pardosa agrestis (1170/9.4%) being the most abundant. Of the recorded species only Eresus kollari and Oxyopes ramosus caught with 7 and 4 individuals each can be qualified as heathland specialists, with species being either eurytopic (associated with both forests and open habitats), or associated with a variety of open xerothermic or sandy habitats. Therefore, making a separate analysis on heathland specialists is not possible. However, we recorded several threatened species, of which five were critically endangered, seven endangered and seven near threatened according to the Red List of German spiders (Blick et al. 2016; Table S4). Three of these species (Psammitis sabulosus, Gnaphosa leporina, and Agroeca lusatica) with adequate abundance (20+) were further analysed for effects of management and habitat structure on threatened species.

In terms of species composition, the different management practices and land use histories (military presence/absence) supported significantly different spider communities (analysed with PERMANOVA), with a significantly higher β diversity in plots with previous military presence vs absence (analysis on multivariate homogeneity of group dispersions). Moreover, the clearest separation in community composition can be seen between the communities inhabiting scarified and unmanaged plots, while the community inhabiting mowed plots falls between the other two treatments with significant overlap with both (Fig. 2). Of the 75 species with a strong preference/aversion to one of the management practices, 27 species were strongly associated with mowed plots, 21 with unmanaged and 8 with scarified plots. On the other hand, 12 species avoided scarified plots, 7 avoided unmanaged plots and none avoided mowed plots (Table S5).

Management practices significantly impacted patterns of spider biodiversity with a greater spider abundance per trap/day in unmanaged and mowed plots than in scarified plots (Fig. 3A), as well as greater FDiv in scarified than unmanaged plots (Fig. 3B, Tables S6, S7). Spider taxonomic (Hill numbers q = 0–2) and FRic were not significantly different between management practices and previous military presence/absence didn’t significantly influence spider biodiversity (Tables S6, S7).

Habitat structure also strongly influenced spider biodiversity with lower abundance per trap/day (Fig. 4A) and FRic (Fig. 4B) in plots with greater bare soil cover, as well as lower taxonomic diversity (Hill numbers q = 0–2) in plots with greater heather cover (Fig. 4C; Table S8). Additionally, FDiv decreased with trap distance to the nearest tree (Table S8).

In terms of conservation of threatened species different management practices and habitat structures promoted the abundance of different species (Fig. 5 Tables S9, S10, S11). The endangered ground hunter Gnaphosa leporina preferred unmanaged over scarified plots (Fig. 5A) as well as plots with greater grass cover (Fig. 5B), the endangered ground hunter Agroeca lusatica preferred mowed plots (Fig. 5C) with greater heather litter cover (Fig. 5D) and lower bare soil cover (Table S11), while the critically endangered ambush hunter Psammitis sabulosa preffered scarified plots (Fig. 5E) with greater bare soil cover (Fig. 5F).

As expected, the different management practices support spider communities with different compositions, mainly as a result of differences in habitat structure (Uetz 1991). Bare soil cover and heather cover were the main factors structuring spider communities contributing to a significant difference in community structure between scarified and unmanaged plots which occupy the two extremes in terms of these two forms of cover. Mowed plots harbored an intermediate community and were not avoided by any of the recorded species as a result of having the most heterogeneous habitat structure that supported the ecological niches of the most species (Gibson et al. 1992). Conversely, scarified plots were avoided by the most species due to them harboring the most homogeneous habitat structure. This did not translate into a significant difference in α or β diversity between mowed and scarified plots with the only difference between these two management practices being lower abundance in scarified plots. It is important to consider that we registered these few significant differences in management effects on spider biodiversity only one-two years after the last management cycle, suggesting that they might disappear only several years after management implementation. However, the effects of habitat structure were stronger, with greater bare soil cover negatively impacting spider functional richness, highlighting the fact that increasing habitat homogenization results in the filtering of traits (Gámez-Virués et al. 2015). In terms of traits, loss of habitat structures would disproportionately affect web builders as they require attachment points for their webs (Uetz 1991). Considering that web builders are under-sampled by pitfall traps due to their lack of mobility, the trait filtering effect of increased bare soil cover in our study could be underestimated. Lastly, taxonomic diversity was lower in plots with greater heather cover likely as a consequence of lower habitat heterogeneity due to the dominance of one cover type. Overall, low intensity heathland management can promote species with a variety of niches due to increasing habitat heterogeneity, whilst reducing habitat heterogeneity with high intensity management can lead to trait filtering and general loss of spider density. Additionally, high intensity management can induce spider biodiversity loss through direct mortality, as spiders are vulnerable to mechanical soil disturbance (Thorbek and Bilde 2004). It should be noted that the negative effects of high intensity management on habitat heterogeneity is applicable on the small spatial scale of our study. In contrast, implementing high intensity management at small spatial scale within a landscape level context would likely lead to an increase in landscape level habitat heterogeneity.

If management impacts are assessed from a “threatened species” point of view, our take away message would be different. Due to spider species differing in their habitat requirements (Uetz 1991), different threatened species preferred different cover types including litter cover, grass cover and bare soil cover further emphasizing the importance of a heterogeneous habitat structure. However, one of the species of highest concern, the critically endangered ambush hunter Psammitis sabulosus was found almost exclusively in scarified plots as it requires patches of bare soil where its coloration serves as camouflage (Nentwig et al. 2024). Other species such as the critically endangered ladybird spider Eresus kollari also require heather dominated bare soil patches which are preserved in the Lüneburg Heath natural reserve (Krause et al. 2011). Furthermore, scarified plots harbored greater FDiv than unmanaged plots, underlining the fact that they support threatened specialists with extreme traits compared to the mean of the trait space. This is in line with recent research showing the benefits of the creation of bare soil patches through high intensity management on the diversity of a variety of heathland associated biota, especially of stenotopic and threatened species (Hawkes et al. 2021; Pedley et al. 2013). Information on the appropriate (or critical) size of bare soil patches is, however, limited. Cameron and Leather (2011) have studied the topic in carabids and have found that the perfect patch size is often dependent on a range of local environmental variables. In light of these results, frequent and small-scale high intensity management would likely have the biggest benefit for biodiversity conservation as there would always be some patches at the intermediate stages of succession when habitat heterogeneity is usually highest. Scarification is a good candidate for small-scale high intensity management as it removes extensive moss carpets and increases the cover of bare soil, without removing all the vegetation as is the case with choppering and sod-cutting. Therefore, it is an important management option that concomitantly tackles the issues of nitrogen deposition and biodiversity conservation by removing moss and creating bare soil patches while preserving as much habitat heterogeneity as possible.