Forest restoration to attract a putative umbrella species, the white-backed woodpecker, benefited saproxylic beetles

Umbrella species are often spatially demanding and have limited ability to adapt to environmental changes induced by human land-use. This makes them vulnerable to human encroachment. In Sweden, broadleaved trees are disadvantaged by forestry, and commercially managed forests are often deprived of dead wood. This has led to a situation where previously widespread top predators in saproxylic food webs, such as the white-backed woodpecker (Dendrocopos leucotos), have become species of conservation concern. The white-backed woodpecker is generally considered an umbrella species, and it has been linked to forests with large volumes of dead wood from broadleaved trees. In recent years, forest stands have been restored for the white-backed woodpecker, but post-treatment evaluations have rarely included other species that also occur in broad-leaved forests (co-occurring species). Many co-occurring species are saproxylic beetles. In this study, we collected saproxylic beetles and environmental data in restored and commercially managed forests to evaluate if habitat restoration for the white-backed woodpecker also benefited other species with similar habitat associations. We found that volumes of coarse woody debris were higher in restored than in commercially managed forests, and that a majority of man-made snags and downed logs were created from birch trees (Betula spp.). Most spruce trees (Picea abies) were extracted during forest restoration, and this opened up the forest canopy, and created stands dominated by broadleaved trees. Many saproxylic beetles were more common in restored forests, and there were significant differences in species composition between treatments. These differences were largely explained by species traits. Effects of sunexposure were particularly important, but many beneficiary species were also linked to dead wood from broadleaved trees. Red-listed saproxylic beetles showed a similar pattern with more species and individuals in restored sites. The white-backed woodpecker is still critically endangered in Sweden, but important prey species are already responding to forest restoration at the stand level. We recognize that landscape-level improvements will be required to bring the white-backed woodpecker back, but also that the umbrella species concept can provide a useful framework for successful forest restoration as many co-occurring saproxylic beetle species seemingly benefitted from restoration for the white-backed woodpecker.


INTRODUCTION
Human land-use has changed forest ecosystems world-wide.Forest management often favors production of commercially important tree species at the expense of less profitable but in many cases ecologically important tree species.In Sweden, stand-level volumes of timber have increased with 40-80% since the 1950s (SLU 2012).The explanation for this is an increased production of conifers at the expense of broadleaved trees that are disfavored by modern forestry, e.g., during thinning.Commercial forests are therefore denser and less permeable to sunlight.This has led to an impoverished fauna of species associated with sun-exposed conditions and broad-leaved trees (Ga ¨rdenfors 2010).Broadleaved trees are particularly important in the context of biodiversity management since most planted trees are coniferous and species associated with broadleaves are particularly disfavored (Bernes 2011).Broadleaved trees are also disadvantaged when natural disturbance regimes, such as recurrent wildfires in upland forests and seasonal floods in riparian environments, are suppressed or altered (Linder et al. 1997, Johansson and Nilsson 2002, Hellberg 2004).
One species harmed by practices favoring conifers over broadleaves is the white-backed woodpecker, which is critically endangered in Sweden with less than a handful of breeding pairs/year recorded during the last decade (Stigha ¨ll et al. 2011).The main reason for the dramatic decline of this species is forest management and changes in important disturbance regimes, such as flooding and fire.The habitat preferred by the white-backed woodpecker is relatively open forest with broad-leaved trees and large volumes of dead wood.The whitebacked woodpecker is known for being associated with edge habitats (Stigha ¨ll et al. 2011), postfire sites (Mild and Stigha ¨ll 2005), and areas of intermediate forest cover (Mikusiński and Angelstam 2004).The white-backed woodpecker feeds on invertebrates, mostly saproxylic (wood living) beetles in dead wood of broad-leaved trees (Aule ´n 1988).
In an effort to save the white-backed woodpecker in Sweden the largest forest restoration project in the country to date was launched in 2005 as part of the Swedish Environmental Protection Agency's species action plan for the bird (Mild and Stigha ¨ll 2005).To restore habitats for the white-backed woodpecker, forest managers in Sweden have created dead wood from broadleaved trees and selectively harvested spruce trees to open up forests, and to make deciduous trees more competitive (Mild andStigha ¨ll 2005, Blicharska et al. 2014).So far, more than 10000 ha of forested stands have been restored to aid white-backed woodpecker populations.In spite of these efforts, the white-backed woodpecker has failed to recover-only nine birds were identified in 2012 as living permanently in Sweden-which could lead to the conclusion that the restoration has been unsuccessful.
However, the white-backed woodpecker is not the only species that is disfavored by modern forestry.Species of conservation concern are often limited by substrates or structures that develop slowly, e.g., dead wood.On the Swedish red-list, 25% (1126 species) of all red-listed species are saproxylic (Dahlberg and Stokland 2004).Many wood-inhabiting beetles, including those that are woodpecker prey, are associated with sun-exposed substrates (Kaila et al. 1997, Martikainen 2000, Sverdrup-Thygeson and Ims 2002).Thus, although the white-backed woodpecker is yet to recover in Sweden, we could expect other species with similar habitat requirements to benefit from restoration for this species, and in fact the white backed woodpecker has been suggested as a putative umbrella species (Angelstam et al. 2003, Roberge et al. 2008, Stigha ¨ll et al. 2011).
Umbrella species are wide-ranging species whose requirements include those of many other species (Groom et al. 2006).Results from some studies suggest that the conservation of putative umbrella species might confer protection of cooccurring species (Fleishman et al. 2000, 2001, Suter et al. 2002, Caro 2003, Kerley et al. 2003) whereas other do not (Andelman and Fagan 2000, Caro 2001, Rubinoff 2001).It can be difficult to quantify restoration needs at the landscape level, but habitat requirements of umbrella species might provide a mechanism for determining tangible targets.In fact, recent findings suggest that umbrella species can guide management efforts (Branton andRichardson 2014, Sheehan et al. 2014).
Only a handful of studies have looked specifically at spillover effects of ecological restoration for an umbrella species.In the USA, pinegrassland communities have been restored for the red-cockaded woodpecker (Picoides borealis) with positive outcomes for other bird species associated with fire-induced environments (Wilson et al. 1995).Forest management for the cerulean warbler (Setophaga cerulean) has also been shown to benefit other disturbance-dependent bird species (Sheehan et al. 2014).Crosstaxon studies are uncommon, and to our knowledge we are among the first to evaluate the effects of ecological restoration for an umbrella species from the perspective of another taxonomic group.In fact, we are only aware of one previous study, i.e., Branton and Richardson (2014), and it presented results for an aquatic scenario.Branton and Richardson (2014) showed that efforts to restore floodplain ponds for coho salmon (Oncorhynchus kisutch) provided benefits for other fish species: abundance of coho salmon were, for instance, positively associated with biomass and abundance of other fish species and vertebrates of conservation concern.
Earlier studies have revealed positive relationships between the occurrence of white-backed woodpecker and species richness of other bird species (To ¨rnblom et al. 2007, Roberge et al. 2008), beetles, and cryptogams; many of which are threatened or red-listed (Martikainen et al. 1998, Roberge et al. 2008).Several studies show that avian top predators can be efficient umbrella species (Martikainen et al. 1998, Sergio et al. 2003, 2004, 2005, 2006, Roberge et al. 2008), but the evidence is not conclusive (Roth and Weber 2008).However, the usefulness of the whitebacked woodpecker's habitat requirements for guiding restoration efforts is yet to be properly evaluated.We therefore collected saproxylic beetles and environmental data in restored and commercially managed forests to evaluate spillover effects of stand restoration for the whitebacked woodpecker.Many wood-inhabiting beetles dwell in coarse woody debris from broadleaved trees.In fact, birch trees alone support more than 400 species (Ehnstro ¨m 2011).It is also clear that many saproxylic beetles are associated with sun-exposed substrates (Kaila et al. 1997, Martikainen 2000, Sverdrup-Thygeson and Ims 2002).This suggests that many saproxylic beetles could co-exist with the white-backed woodpecker in restored forests that resemble naturally disturbed environments.
For the white-backed woodpecker to be useful as an umbrella species, we would expect other organisms associated with similar forest types to benefit from forest restoration guided by habitat requirements of the white-backed woodpecker.To be more specific, we hypothesized that (1) forest restoration for the white-backed woodpecker would create open stands with more dead wood from broadleaved trees than commercially managed forests; (2) saproxylic beetles would respond positively, in terms of species richness and abundance, to an increase in sun-exposed dead wood from broadleaved trees; (3) forest restoration would cause changes in the species composition of saproxylic beetles towards communities favored by sun-exposure and dead wood from broadleaved trees; and (4) restored sites would support more saproxylic species and individuals of threatened beetles than commercially managed forests.

Study area
Nine restored and nine commercially managed reference sites were included in this study.All study sites were located between the latitudes 59.58 and 59.98 N, and the longitudes 12.08 and 13.78 E, in Va ¨rmland, southwestern Sweden, in the center of the historical distribution range of the white-backed woodpecker and one of the few areas in Sweden where the species still occasionally occurs (Stigha ¨ll et al. 2011).In elevation, the sites ranged from 67 to 246 m.a.s.l.Before treatment, restored sites resembled commercially managed sites.This was concluded from data provided by the Swedish Forest Agency and two forestry companies (Stora Enso and Bergvik Skog), but also from field measurements.Site selection was largely based on similarities in forest age and stand basal area (Table 1), but many more variables were considered (see Appendices A and B).Restored sites were selectively harvested to remove Norway spruce, between 2 and 12 years ago (mean ¼ 6.7 years).Dead wood was also created from broadleaved trees like birch and, rarely, European aspen (Populus tremula).Ring-barking or girdling killed many broadleaved trees, but harvesters were also used to create high-stumps or snags, i.e., remove tree tops approximately 4 m from the forest floor.All tree tops were deposited at restored sites to provide downed logs.On average, restored sites covered 5.6 ha (SE ¼ 0.77).This can be compared to an average stand size of 7.4 ha (SE ¼ 1.67) for commercially managed sites.Trees were, on average, 43.4 years old (SE ¼ 5.45) in restored sites and 41.8 years old (SE ¼ 4.85) in commercially managed sites.

Invertebrate sampling
Three flight-intercept traps (Polish IBL2-traps, CHEMIPAN, Warsaw, Poland) were positioned in the center of every study site, from early June until late October in 2012, to catch airborne saproxylic beetles.Flight-intercept traps were semi-transparent and shaped like downward-facing triangles (height ¼ 1 m, base ¼ 1 m).Collected invertebrates were conserved in bottles filled with propylene glycol and detergent, attached to the bottom of each trap.Sampling bottles were also equipped with vents to divert rain water.All flight-intercept traps were strung between two trees at breast height, and positioned at least 30 m and 120 degrees apart.The first trap at every site was positioned towards the north, and the remaining traps faced south-east and south-west, respectively.All saproxylic beetles were identified by expert taxonomists to the species-level except for specimens of the genus Acrotrichis (family: Ptiliidae).All species were also divided into categories based on their habitat association.Substrate (broadleaved or coniferous trees) and microclimatic (sun or shade) association were given special attention.Classifications were made from empirical data provided by other researchers (Lindhe et al. 2005; v www.esajournals.orgJ. Hja ¨lte ´n et al., unpublished data), but claims were also cross-referenced in scientific databases and in books on species traits (Koch 1989, Ehnstro ¨m andAxelsson 2002;www.beetlebase.com).When the scientific literature failed to provide answers, preferences were listed as unknown.Many species (39%) were also linked to dead wood from both broadleaved and coniferous trees.Others (12%) thrived in both open and closed forests.

Stand-level characteristics
Dead wood was surveyed in four random transects, 50 m long and 5 m wide, per study site.All coarse woody debris objects (dbh !0.1 m and length !1.3 m) were measured within 2.5 m on both sides of each transect.Objects partly outside of the surveyed area were measured to the outer limit, but not beyond the transect.If, by chance, transects continued outside the boundaries of the study site the outside area was withdrawn from the total area.Dead wood was divided into lying and standing coarse woody debris for each tree species.The volume of lying coarse woody debris was calculated with a formula for a truncated circular cone: Þ where h ¼ height or length, r 1 ¼ maximum radius, and r 2 ¼ minimum radius.For standing coarse woody debris, r 1 and r 2 were estimated from average changes in radius/m for lying coarse woody debris.
We counted pixels covered by vegetation in digital images to measure the canopy cover.All images were captured with a compact camera (Panasonic DMC-TZ20, resolution 14 Megapixel) placed 0.1 m above the forest floor.Forest stands were photographed between four and 10 times depending on their size to give an average estimate per study site.The percentages of visible sky on all the photos were calculated using the magic-wand and the pixel-count functions in Photoshop Elements (Ver.8.0, 2009).Angle count sampling was performed to determine the stand basal area (m 2 /ha) of all tree species in every study site.We used a relascope to assess the situation from fixed positions, randomly allocated across every study site, in scopes of 360 degrees.Study sites were sampled between four and 10 times depending on their size.
Understory plants of different species require varying amounts of nutrients.The forest productivity can be estimated by dividing forests into categories based on such differences.In this study, we used a method developed by the Swedish University of Agricultural Sciences and the Swedish Forest Agency (Ha ¨gglund and Lundmark 2004).The data were collected in three random survey plots per study site.Each circular plot (radius ¼ 10 m) covered an area of 315 m 2 , and all plant species typical for this forest type were identified within every plot.All surveys were conducted by the same field personnel, and within four days in August 2012.

Statistical analyses
Stand-level characteristics were assessed independently in pair-wise comparisons of restored and commercially managed forests.If requirements for normality and homoscedasticity were fulfilled, a Student's t-test was employed.Otherwise, an unequal variance t-test was used (Ruxton 2006).Alpha levels were sequentially Bonferroni adjusted due to repeated independent t-tests.We examined differences in forest age, stand basal area, coarse woody debris volume, and canopy cover.Equivalent analyses were also performed to examine differences in species richness and abundance for red-listed species, and saproxylic beetles of different substrate (broadleaved or coniferous trees) and microclimatic (sun or shade) preferences.All analyses were carried out in IBM SPSS Statistics (Version 21).
To explore differences in community structure between treatments, we performed multivariate analyses in PRIMER 6 (version 6.1.12)and PERMANOVAþ (version 1.0.2) by PRIMER-E Ltd.Prior to analysis, a Bray-Curtis similarity matrix was calculated for raw data and transformed ( presence-absence) data.Non-metric multidimensional scaling (nMDS) was used to illustrate differences in dominance structure (raw data) and species composition (presence-absence data) between restored and commercially managed sites.For conclusive statistical tests, however, we used PERMANOVA in PRIMER.Species contributions to the observed dissimilarity between treatments were calculated with SIMPER in PRIMER for all saproxylic beetles in transformed and untransformed datasets.
Distance-based linear modeling (DistLM) in PRIMER showed to what extent structural (canopy cover, elevation, coarse woody debris volume, productivity, aspect, tilt, and tree assemblage) and temporal (forest age and time since restoration) variables associated with the community composition of all saproxylic beetles.Variables with less than six observations were omitted from the analysis.To explore relationships for individual variables, marginal tests were performed.All variables were subsequently subjected to a step-wise selection procedure (selection criterion: AIC c ) in order to develop models.Prior to analysis, environmental variables were plotted against each other in a Draftsman plot to control for collinearity.Pearson correlation analysis was used to quantify the relationships.If the correlation coefficient was higher than 0.6 one variable was excluded from the analysis.Variables excluded were the basal area of Picea abies, Sorbus acuparia and Salix caprea, the volume of lying Betula coarse woody debris and the productivity class ''Vaccinium myrtillus type''.In all step-wise procedures and marginal tests, P-values were obtained with 999 permutations.Models with a DAIC c , 2 were considered equal.

Forest structure
Commercially managed sites were dominated by coniferous trees, and had a significantly higher stand basal area of Norway spruce than restored sites where spruce had been selectively harvested (Table 1).But after sequential Bonferroni correction of P-values due to multiple repeated independent tests, the basal area of all other tree species did not differ between restored and commercially managed sites (Table 1).Additionally, there were no significant differences in forest age between forest types (F ¼ 0.57, t ¼ À0.23, df ¼ 16, P ¼ 0.82), but restored sites had significantly less canopy cover than commercially managed sites (t ¼ 5.0, df ¼ 9.10, P ¼ 0.001).Dead wood was also created during forest restoration, and restored sites contained larger total volumes of lying coarse woody debris than commercially managed sites (Table 2).Most of the coarse woody debris in restored sites was derived from birch trees.Birch volumes were higher in restored sites for lying coarse woody debris (Table 2).In total, there were no significant differences between treatments in terms of standing coarse woody debris (Table 3).

Dominance structure and species composition
Species assemblages in restored forests were significantly different from species assemblages in commercially managed forests (Fig. 2), and results based on presence-absence data were in agreement with results based on counts.More than 30% of the variation in the raw data analysis was explained by 10 species (Table 4).Among them, only one, Tomoxia bucephala, was unique to restored forests.T. bucephala was also influential in the analysis based on presence-absence data (Table 4).Influential species were often linked, in terms of their ecology, to stand conditions associated with each treatment (Table 4).Sunexposure was particularly important.Abundant and widely distributed species in restored forests were generally associated with sun-exposure, in contrast to common species in commercially managed forests that were associated with sheltered conditions (shade).

Environmental drivers of community shifts
Distance-based linear modeling showed that several independent variables were significantly correlated with the species-derived, multivariate, data cloud (Appendices A and B), i.e., canopy cover, time since restoration, elevation and broadleaf grass vegetation type.Each of the mentioned variables single-handedly explained more than 10% of the variation.In subsequent, step-wise procedures to differentiate between models, canopy cover occurred in three out of four models for raw data and in four out of the eight best models for presence-absence data (models with DAIC c , 2; Table 5).Other important variables were time since restoration, elevation and vegetation type (presence-absence data only).The canopy cover was, however, negatively correlated with the volume of lying coarse woody debris from birch and with spruce basal area.

Red-listed species
Nineteen species on the Swedish red-list (Ga ¨rdenfors 2010) were caught in the flightintercept traps (Table 6).Analyses revealed significant differences in species richness (F ¼ 0.25, t ¼ À2.84, df ¼ 16, P ¼ 0.012; Fig. 3A) and abundance between treatments (t ¼ À2.53, df ¼ 16.0, P ¼ 0.022; Fig. 3B); with more species and individuals in restored sites.Hylis procerulus was particularly common (with 46 individuals unique to restored sites) and explained most of the difference in total abundance (Table 6).Altogether 10 species were unique to restored forests, compared to three species in commercially managed forests (Table 6).

DISCUSSION
Restoration efforts guided by the white-backed woodpecker created stands significantly different from those impacted by commercial forestry, as predicted in our first hypothesis.Stand-level v www.esajournals.orgcharacteristics showed that restored sites contained larger volumes of coarse woody debris than commercially managed forests.Restored forests were also more open, and permeable to sunlight.Saproxylic beetles with explicit substrate and microclimatic preferences were, as predicted in our second hypothesis, more common in restored sites (Appendix C).Taxonomic groups favored by dead wood from broadleaved trees were, for instance, more species-rich in restored sites.This is not surprising since large quantities of dead wood were created from birch trees.Selectively harvested forests had a higher species richness and abundance of saproxylic beetles favored by sun-exposure.In this study, we were able to anticipate the effects of sunexposure thanks to earlier assessments of forestry impacts.Several studies show that clear-felled forests, with retained dead wood from broadleaved trees, are utilized by different saproxylic beetles than mature forests with similar substrates (Kaila et al. 1997, Martikainen 2000, Sverdrup-Thygeson and Ims 2002).In this study, pair-wise comparisons showed that forest restoration can benefit saproxylic beetles favored by sun-exposure without affecting the species richness of shade-tolerant species.However, it is important to remember that 59 species were unique to commercially managed sites.
There were no differences in total species richness and abundance between restored and commercially managed forests, but in agreement with our third hypothesis there were significant differences in species composition, largely explained by canopy cover.A reason for this could be that many saproxylic beetles have developed adaptations to natural disturbances, such as fire (Wikars 2002, Boulanger and Sirois 2007, Hja ¨lte ´n et al. 2007, Hekkala et al. 2014).Wildfires, storms, floods, and pest outbreaks can create open habitats with large volumes of dead wood (Esseen et al. 1997, Kuuluvainen 2002), much like those restored for the white-backed woodpecker (Tables 2 and 3).Forest restoration for the white-backed woodpecker created favorable environmental conditions for warmth-demanding species currently disadvantaged by forestry practices.In addition, 91 species (31% of all species caught in the traps) were unique to restored sites.
Red-listed species were positively affected by forest restoration, both in terms of species richness and abundance.This provides support for our final hypothesis, but unfortunately the sample size was too small for us to analyze      v www.esajournals.orgdifferences in community composition.Two redlisted species, Hylis procerulus and Xylophilus corticalis, were particularly abundant in restored sites.In Sweden, H. procerulus has been associated with dead wood from Norway spruce (Baranowski et al. 1999), but observations in other countries (e.g., Ukraine and Russia) link it to both broadleaved and coniferous trees (Bu ¨che et al. 2010).X. corticalis will also use both substrates, but it is limited by other factors, such as the initial level of decomposition (Ehnstro ¨m 1999).It is also possible that both species benefit from sun-exposure, but we were unable to find support for this in the literature.
Canopy cover was identified as one of the most important environmental variables determining beetle community composition.However, canopy cover was correlated with spruce basal area and lying birch CWD making it difficult to distinguish between the effects of dead wood enrichment and those of spruce removal.Some wood-inhabiting beetles are likely to benefit from both measures, e.g., prey species targeted by the white-backed woodpecker in edge habitats, postfire sites, and areas of intermediate forest cover.Two examples of important prey species for the white-backed woodpecker, unique to restored sites in this study, were Scolytus ratzeburgii and Leptura quadrifasciata (Aule ´n 1988).Both species have been linked to sun-exposed dead wood from broadleaved trees (Ehnstro ¨m and Axelsson 2002, Dahlberg and Stokland 2004).They are also easily surveyed.Adult specimens of L. quadrifas-ciata are conspicuous and attracted to forbs like Filipendula ulmaria and Angelica sylvestris in early summer, and larval stages of S. ratzeburgi leave distinctive signs of foraging in dead wood (Ehnstro ¨m and Axelsson 2002).This might make them good sub-targets in landscape-level attempts to restore habitats for the white-backed woodpecker, as well as examples of restoration success at the stand-level.
Umbrella species have mainly been discussed as management shortcuts in environmental protection (Caro and O'Doherty 1999), but ecological restoration presents a different context.Cooccurring species will overlap at different spatial and temporal scales, and umbrella species with extensive habitat requirements will probably be among the last to recover.Under such circumstances, umbrella species become testimonies of ecosystem recovery rather than management shortcuts (Fig. 4).The white-backed woodpecker has not yet recovered in Sweden, but at this point in time habitat requirements have only been fulfilled at the stand-level.Forest restoration has created stands with coarse woody debris volumes similar to those found in typical whitebacked woodpecker habitats (10-20 m 3 /ha; Angelstam et al. 2003, Aule ´n et al. 2010), but whitebacked woodpecker territories cover much larger areas (150-650 ha;Aule ´n et al. 2010).
In conclusion, we argue that habitat requirements of umbrella species can guide successful restoration efforts, if there is congruence in the response of co-occurring species to manipulated v www.esajournals.org(environmental) attributes.Resource-and process-limited species with extensive area requirements, like the white-backed woodpecker, are particularly useful in this context (Lambeck 1997, Angelstam andAndersson 2001).It is not certain that white-backed woodpecker populations in Sweden will recover, validating successful restoration efforts at the landscape-level, but in this study there were many beneficiary species at the stand-level.Positively affected species were often linked to environments much like those inhabited by the white-backed woodpecker, and we argue that this shows that there is merit to the umbrella species concept in restoration design.Local improvements will not necessarily fulfill habitat requirements of top predators like the white-backed woodpecker, but less demanding species, like many saproxylic beetles, will probably respond more rapidly.In fact, forest restoration will most probably have bottom-up effects on top predators in saproxylic food webs.Under such circumstances, umbrella species become testimonies of ecosystem recovery rather than management shortcuts.v www.esajournals.org

Fig. 2 .
Fig. 2. The community structure of saproxylic beetles in restored sites (triangles) and commercially managed reference sites (squares) illustrated in two nMDS-plots.The left-hand plot (dominance structure) was composed from untransformed (raw) data.The right-hand plot was composed from transformed (presence-absence) data.Each symbol represents one study site.P-values denote significant (P 0.05) differences in dominance structure and species composition.

Fig. 3 .
Fig. 3.In A: number of species (saproxylic beetles, mean 6 SE) on the 2010 Swedish red-list in restored and commercially managed reference forests.In B: abundance of saproxylic beetles (mean 6 SE) on the 2010 Swedish red-list in restored and commercially managed reference forests.P-values denote significant (P 0.05) differences.

Fig. 4 .
Fig.4.Many species are resource-and processlimited, but at different spatial and temporal scales.Local improvements will not necessarily fulfill habitat requirements of top predators like the white-backed woodpecker, but less demanding species, like many saproxylic beetles, will probably respond more rapidly.In fact, forest restoration will most probably have bottom-up effects on top predators in saproxylic food webs.Under such circumstances, umbrella species become testimonies of ecosystem recovery rather than management shortcuts.

Table 1 .
All tree species and their stand basal area (m 2 /ha) in restored and commercially managed reference forests.
Notes: NS ¼ non-significant, Na ¼ no statistical test conducted due to low sample size.Ref ¼ reference, R ¼ restored, * denotes P-values significant after sequential Bonferroni adjustment due to repeated independent t-tests.

Table 2 .
Volume of lying coarse woody debris (m 3 /ha) per tree species in restored and commercially managed reference forests.
Notes: NS ¼ non-significant, Na ¼ no statistical test conducted due to low sample size, Ref ¼ Reference, R ¼ restored, * denotes P-values significant after sequential Bonferroni adjustment due to repeated independent t-tests.

Table 3 .
Volume of standing coarse woody debris (m 3 /ha) per tree species in restored and commercially managed reference forests.NS ¼ non-significant, Na ¼ no statistical test conducted due to low sample size, Ref ¼ Reference, R ¼ restored, * denotes P-values significant after sequential Bonferroni adjustment due to repeated independent t-tests.

Table 4 .
SIMPER results for raw data and presence-absence (P-A) data.Only the ten most influential species were included in this table.All listed species have been categorized in terms of their microclimatic (sun or shade) and substrate (broadleaved trees or coniferous trees) preferences.Notes: Contrib% ¼ contribution in percent, Cum% ¼ cumulative percentage, (?) ¼ no empirical evidence, but unique to restored forests, ?¼ no empirical evidence, x* ¼ weak empirical evidence.

Table 5 .
Best models from distance based linear modelling showing all models with DAIC c , 2 for raw data and presence-absence data respectively.df ¼ 16.Time ¼ time since disturbance, BRGR ¼ broadleaved grasses vegetation type.