Introduction

Forest ecosystems historically dominated the central European landscape (Küster 2000; Rackham 2006). The forest cover and its spatial structure have been dramatically altered (McNeely 2002). Most lowland forests were intensively exploited through traditional forest management, although the spatial structure has remained almost constant for a long time (Horak et al. 2012a). The forest cover in sub-mountainous and mountainous regions was exploited by humans considerably later than lowland areas. Nevertheless, the spatial and temporal continuity of forests at higher elevations was changed over a relatively short time span (i.e. one or two centuries; Fanta 1997; McNeely 2002). Except for a few areas which were traditionally managed (e.g. wood pasturing), most forests were clear-cut and the timber was floated on rivers to lowlands due to insufficient stocks of fuel wood and building timber in the latter (Fanta 1997, 2007). Therefore, the insufficiency of timber in lowlands caused a lack of dead wood (a crucial forest habitat) in mountains (e.g. Müller and Bütler 2010).

Sustainable forest management helps to maintain biodiversity. One of the main ambitions of conservation-oriented forest management is to maintain undisturbed forest areas for future generations (Jenkins and Joppa 2009). Thus, the implementation of a hands-off approach (i.e. unmanaged areas) to forest reserve management was, and still is, one of the main forest conservationists’ tools to protect forest biodiversity (Margules and Pressey 2000). From a biological point of view, most central European sub-mountainous and mountainous forests are heavily contaminated by commercial spruce stands (Slodičák et al. 2005; Svoboda et al. 2010). Such stands need to be carefully managed during their entire rotation period and foresters often face increased salvage cuttings caused by windstorms, snow-breakages, insect outbreaks, or fungal diseases (Kelty 2006; Slodičák and Novák 2006). For several reasons, it is difficult to leave naturally disturbed forest areas to their spontaneous development (e.g. due to insect outbreaks, imminent commercial losses, high game stocks). The most implicated management approach following disturbances is the above-mentioned salvage cutting (Jonasova et al. 2010). Nevertheless, it is well known that such management practices strongly affect biota (Lindenmayer et al. 2004; Müller et al. 2008; Vávrová et al. 2009; Pouska et al. 2010).

Forest reserves can serve as an opportunity for studying natural processes in the wild (Margules and Pressey 2000). One of the main problems is identifying an area large enough for worthwhile study, and creating sufficient buffer zones to prevent surrounding commercial forests from exerting potential negative influences (e.g. bark beetle attacks; Jakus et al. 2003). Even in the central European country (namely the Czech Republic), where the forest reserves have been established for almost two centuries (Vrška 2008), there is still a lack of knowledge of insect diversity. Old-growth forests are one of the most studied forest ecosystems, but there is very limited knowledge of the insect diversity at the tree level (e.g. Müller and Goßner 2010).

Decomposing wood, particularly tree trunks and other coarse woody debris, is important as a habitat and resource for many forest organisms (Dynesius et al. 2010). Remnants of forests with veteran trees provide many microhabitats (Winter and Möller 2008). Modern studies have shown that the quality of dead wood, which is associated mainly with tree species, decay stage and stem diameter, contribute to explaining species composition, population densities and species richness of saproxylic beetles (Jonsell et al. 1998; Lindhe and Lindelöw 2004; Toivanen and Kotiaho 2010).

The loss of habitats often leads to vanishing or even extinction, and subsequent red-listing, of formerly common species (Komonen et al. 2008). The recent IUCN Red List of saproxylic beetles (Nieto and Alexander 2010) shows that there is a sharp decline in the distribution of several species and, in certain cases, this led to recent relictual occurrence (Horak et al. 2012a).

Modelling species and environmental relationships is one of the main goals in ecological research (Frutos et al. 2007). Statistical techniques, such as multiple regression methods, are widely used (Guisan and Zimmermann 2000). However, some issues of robustness of these developed models have been detected (MacNally 2000). Techniques such as variance partitioning are used to solve these problems (Borcard et al. 1992; Økland 1999; Lepš and Šmilauer 2003; Svoboda et al. 2011). Through multivariate analyses of ecological data and variation partitioning, it is relatively easy to evaluate the relative importance of the species and their environmental relationships (Burrascano et al. 2011). Although variance partitioning is used in the study of flora and the environment (Aude and Lawesson 1998; Økland 1999), to our knowledge not so many studies have been undertaken on animals, including insects (Kitching et al. 1993; Crist et al. 2006; Horak et al. 2013).

The main goals of our study of undisturbed old-growth mountainous spruce-beech forest were to answer the following questions: (1) are tree level indicators, which are easy-to-see (i.e. simple to collect) in the field, able to explain the species composition of saproxylic beetles and, if so, to what extent and (2) how do red-listed species respond to the environmental variables included in our study?

Materials and methods

Study area

Our study sites were in Bukacka (Eastern Bohemia: Czech Republic), probably the oldest mountainous spruce–beech forest in the Orlicke Mts. (EN: Eagle Mts., DE: Adler Gebirge). This woodland area of more than 50 ha (960–1,025 m a. s. l.) is within the Protected Landscape Area (PLA) of the Orlicke hory and is located on the Czech–Polish border (50°20′N, 16°22′E). The area has recently been protected through hands-off management practices (i.e. unmanaged). However, salvage cutting has been carried out several times over the past decades due to bark beetle (mainly Ips typhographus) outbreaks (e.g. Belicová 1982). The European Beech (Fagus sylvatica) formerly dominated the forest, but is now declining due to several biotic (e.g. high palatability for game stocks) and abiotic (e.g. air pollution) factors, and the Norway Spruce (Picea abies) has been spreading through the area. Typical admixed trees in the tree species composition are rowan (Sorbus aucuparia), and the Sycamore Maple (Acer pseudoplatanus). A smaller part of Bukacka consists of orchid grasslands, formerly pastured and used for growing hay (Hájek and Kučera 2002).

Study groups

We studied saproxylic beetles (Coleoptera), a highly diverse and abundant ecological group. This ecological group is often used in studies on biological diversity and the natural condition of forest habitats (Goßner et al. 2008; Müller et al. 2010; Horak 2011).

Sampling method

We sampled the core area of Bukacka (≈20 ha). We used 20 (one per ha) passive (i.e. non-attractive) window traps, which were placed on tree trunks, and therefore worked as trunk traps. Each trap consisted of three transparent plastic panes (one pane 400 mm × 500 mm and two panes 200 mm × 500 mm), a protective top cover (d = 450 mm), and a funnel leading down into a container holding a solution of water and salt with a small amount of detergent to reduce the surface tension of the liquid. A detailed description of the traps can be found in Horak (2011). We sampled a priori standing high-dimension (i.e. veteran) trees (Table 1). Due to the problems with satellite signal reception in our study area, we were not able to use GPS to determine coordinates, and were therefore not able to provide a map of trap distribution. However, the traps did not have a clumped distribution, and were placed randomly with respect to the study variables (Table 1). Study trees with relatively similar surrounding environments were selected (medium dead wood accumulation, no gaps in canopy and high distance from stand walls). Due to this and the tree species distribution (see section on study area above), we presume that any significant spatial bias in the trap distribution used for our analyses was highly improbable. Each trap was placed at a height of 1.3 m facing the south. The traps were activated during the snow meltdown (i.e. at the end of April), and deactivated during the first frosts at the end of September 2009, which resulted in twelve routine sampling efforts. Thus, each trap was working for a period of 161 days (i.e. 3,220 days for our trapping design).

Table 1 Description of twenty study tree habitats sampled by window traps in the mountainous spruce–beech forest
Full size table

Study variables

We focused on six tree level characteristics grouped into three categories for the final analyses. We used (1) tree category with two categorical (dummy) variables sorted by tree species sampled (beech, rowan and maple as broadleaves and spruce as conifers); (2) diameter category with two quantitative (continuous) variables—diameter of stem at the breast height (DBH in cm measured at 1.3 m above ground) and height of stem (estimated in m); and (3) decay stage category with two categorical variables—fruiting bodies of wood-decaying fungi (presence and absence) and withering stage of stem (snag, dead tree, withered tree) and one semi-quantitative variable of bark coverage (estimated to the nearest 10 %).

Data analyses

For our analyses we used multivariate statistical methods provided by CANOCO for Windows version 4.5 (ter Braak and Šmilauer 2002). The length of gradient for species composition data was calculated with downweighting of rare species (Lepš and Šmilauer 2003), using detrended correspondence analysis (DCA), at 3.23. We therefore used a canonical correspondence analysis (CCA; ter Braak and Šmilauer 2002) with species composition matrix as dependent variable. The potential bias caused by multi colinearity was controlled by variance inflation factor (VIF < 5; Lepš and Šmilauer 2003). We used focused biplot scaling on inter-species correlations. Monte-Carlo permutation tests with significance of canonical axes (9,999 permutations) under the full model were used (Lepš and Šmilauer 2003). We employed test of explanatory variables for our tree level variables, using a Monte Carlo permutation test (9,999 permutations) under the full model to measure explained variance for each study variable. For the resulting ordination diagrams, we used biplots for both of the sample-environmental variables and species-environmental variables, created in CanoDraw 4.14 (ter Braak and Šmilauer 2002). Statistical significance of associations between particular species and variables of the tree category were tested using t value biplots with van Dobben circles (ter Braak and Šmilauer 2002). The process of variance partitioning is described in detail by many authors (see references in Peres-Neto et al. 2006). Two or three sets of environmental categories are commonly used. The measuring of each fraction was based on sets of multivariate analyses using combinations of species, variables and covariables (Lepš and Šmilauer 2003). We finally visualised the results of variance partitioning using a three circles Venn diagram.

Results

In total, we trapped 1,650 individual beetles (Coleoptera) from 179 species of which 83 species (900 individuals) were saproxylic beetles. One species was critically endangered, seven endangered, nine vulnerable and three near threatened (Table 2).

Table 2 Results of explained variance of tree level variables with species composition of saproxylic beetles (Coleoptera) as dependent variable in the old-growth mountainous spruce-beech forest
Full size table

Variance partitioning

We observed that the decay stage category had the highest explained variance for saproxylic beetles, while the diameter and tree had lower effect on explained variance (Fig. 1), which reflected the number of variables in particular category. The value of fractions with shared variance was not very high. The most shared variance was between the tree species and decay stage categories. On the other hand, the diameter category had a negative shared variation of joint fractions and acted as a suppressor.

Fig. 1
figure 1

Venn diagram showing results of variation partitioning among study categories for saproxylic beetles (Coleoptera), in old-growth mountainous spruce–beech forest (***P < 0.001)

Full size image

Visualisation of canonical correspondence analysis and explained variance of particular variables

We were able to explain nearly 50 % of the variance in the diversity of saproxylic beetles at the tree level (Fig. 1). The first axis explained more than 16 % of the variance and the cumulative percentage variance of species data of the first two shown axes was 26.56 % (Fig. 2). Most species-rich samples were distributed along the first ordination axis (Fig. 2). Broadleaves and conifers, presence and absence of fungi, bark coverage and DBH were significant variables. Broadleaves and conifers shared the most explained variance (>10 %) in saproxylic beetle composition (Table S1 of supplementary material) and six species indicated significant response to the tree category, while half of them preferred broadleaves and half of them conifers (Table 2).

Fig. 2
figure 2

Sample-environmental variables biplot of saproxylic beetles (Coleoptera) in old-growth mountainous spruce–beech forest. Continuous and semi-quantitative variables are shown with black arrows, categorical variables are shown with black squares and samples are shown with grey rings with their size corresponding to the number of species in the sample (***P < 0.001)

Full size image

Factors causing occurrence of red-listed species

Nearly the same number of red-listed species was attracted by broadleaves and conifers. Most trapped red-listed beetles seemed not to be responding to the withering stage of stem variables (Fig. 3). The critically endangered saproxylic click beetle Ampedus auripes seemed to prefer dead spruce stems without presence of wood-decaying fungi. The most abundant red-listed rove beetle, saproxylic Hapalaraea pygmaea (EN) seemed to prefer high diameter trees. Two near threatened flightless weevils were trapped in higher abundances—Acalles camelus was more associated with withered infested beech trees, while Trachodes hispidus occurred more on trees with low bark coverage.

Fig. 3
figure 3

Red-listed species-environmental biplot of the studied variables and species composition of saproxylic beetles (Coleoptera) in old-growth mountainous spruce–beech forest. Continuous and semi-quantitative variables are shown with black thick arrows, categorical variables are shown with black triangles, and red-listed species are shown with grey dots (Note species not included in red list were suppressed for visualisation). The projection of the dots of individual species in the right-angle direction on individual variable arrows shows the average value of species with respect to individual continuous or semi-quantitative variables (positive in the direction of an arrow and negative in the opposite direction). The distance of the individual centroid of variable to the individual centroid of species shows the average value of the species with respect to categorical variables

Full size image

Discussion

The tree species category was the only significant with respect to the saproxylic beetle composition. Significant response of saproxylics to broadleaves and conifers with respect to the conventionally studied categories of decay stage and diameter may seem to be surprising. Although this variable is often studied (Jonsell et al. 2007; Horak 2011), most tree-beetle species associations in the conditions of the central Europe do not seem to be so interconnected (e.g. Sláma 1998) as in Scandinavia (e.g. Jonsell et al. 1998). This may reflect that mountainous conditions are much closer to Boreal landscapes due to the continuity and the size of spruce stands in the study site. Although, explained variance and number of associates of conifers was the same as opposed to broadleaves. The coleopteran fauna was composed of species connected with sub-mountainous and mountainous beech and spruce forests of central Europe. The critically endangered click beetle Ampedus auripes, as an example, is a typical saproxylic beetle known to be associated with mountainous spruce forests, which are not negatively affected by intensive forest management (Jarzabek-Müller and Müller 2008). The significant response of saproxylics and nearly 9 % of explained variance to DBH was not surprising because veteran trees of high diameters are known to positively affect the species composition of saproxylic beetles (Alexander 2008). The same reason is most probably for presence and absence of fungi and bark coverage (Horak et al. 2012b).

We found that we were able to explain high levels of variance in our study group using easy-to-see tree level indicators through variance partitioning. In our case of saproxylic beetles, we were able to explain nearly one half of variance. This result highlighted the desirability of tree level indicators when using this design of trap.

The hands-off style of management in smaller areas (as is the case in most of the forest reserves in the Czech Republic), without large-scale disturbances, often results in decreasing heterogeneity of this forest habitat due to over-shading by the closed upper canopy (Konvicka et al. 2004; Horak et al. 2012b). Thus, in our study, in which we needed to combine quantitative, semi-quantitative and categorical variables, and where most of the study variables were not solely able to explain the relationships, possibly due to homogeneity of the study area, variance partitioning seems to be an appropriate method to help explain the interconnection of disparate groups and their habitat.

Conclusions

Our results showed that the beetle species composition and response to tree level indicators in a relatively homogenous environment of old-growth mountainous forest area could be explained reasonably well by variance partitioning. Species composition of saproxylic beetles did not show highly distinct levels of explained variance for categories of variables, but rather at different levels of significance. Some variables, even carefully selected, were not significant to explaining the relationships with the species composition themselves. This could be explained by an absence of spatial heterogeneity in the old-growth forest of our study area, which was subject to hands-off management. Our results highlighted the desirability of tree level indicators when using window traps in trunk tree trap design.