Saproxylic beetle assemblages in artificially created high-stumps of spruce (Picea abies) and birch (Betula pendula/pubescens) – does the surrounding landscape matter?


Matts Lindbladh, Southern Swedish Forest Research Centre, SLU – Swedish University of Agricultural Sciences, P.O. Box 49, SE 230 53, Alnarp, Sweden. E-mail:


Abstract.  1. To create high-stumps (snags) is a common conservation action during final felling in Swedish production forests. However, many wood-living beetle species are only found in certain areas with higher overall biodiversity, so called hotspots. It has been argued that it is efficient to concentrate conservation efforts to hotspots.

2. The saproxylic beetle fauna was sampled on ten clearcuts inside hotspots and ten clearcuts outside the hotspots. They were collected with window traps mounted on 2- and 4-year-old spruce and birch high-stumps. We also used environmental data (e.g. tree species composition) to confirm differences between the surroundings of two, the clearcut types.

3. High-stumps on the hotspot clearcuts did not attract more saproxylic beetle species, or red-listed species, than high-stumps outside the hotspots. The environmental data showed that the clearcuts differed in several important aspects, for instance, were there a higher proportion of broadleaved trees around the hotspot compared with the clearcuts outside the hotspots. In a Canonical Correspondence Analysis, the proportion of coniferous and broadleaved forest was an important explanatory variables. The hotspot variable did contribute significantly in explaining the beetle composition on the birch high-stumps, but not on the spruce high-stumps.

4. In general, the study suggests that concentrating high-stumps to hotspot areas will not benefit more species. However, the result indicates birch high-stumps could be prioritised in a biologically rich landscape. The hotspot effect may be more noticeable in the future as the high-stumps decay and their importance for late successional species increase.


During the last decade, much focus in forest conservation has been on dead wood and species associated with dead wood. This substrate is a key feature in natural forests and constitutes a habitat where numerous saproxylic (wood-living) organisms live (Grove, 2002). The volumes of dead wood in Swedish production forests are at least one magnitude lower than that in natural forests (Fridman & Walheim, 2000). On the scale of single stands, studies have shown that lower amount of dead wood support a lower number of saproxylic species (Martikainen et al., 2000; Sippola et al., 2002; Hyvärinen et al., 2006a). However, stand variables cannot fully explain occurrences of these species because the regional species pool available for colonising a stand varies. These differences in species pool are created by the history of the forest (Nilsson & Baranowski, 1993; Lindbladh et al., 2003) and/or the amount of suitable habitat in the surrounding landscape (Franc et al., 2007). Thus, many rare wood-living beetles species are only found in certain areas, so called hotspots, which often have higher overall biodiversity as a result of lower previous management intensity (Ehnström & Waldén, 1986; Nilsson et al., 2005). It has been argued that it is more efficient to concentrate conservation efforts in hotspots rather than distributing the efforts throughout the landscape (Hanski, 2000; Franc et al., 2007). However, Swedish forestry policy is gauged for the latter. For instance, certification schemes like Programme for the Endorsement of Forest Certification (PEFC) schemes and Forest Stewardship Council (FSC) require that certain conservation actions are taken during silvicultural operations (Anon, 2000a,b). Examples of actions taken during final felling are the retention of green and dead trees, and the creation of high-stumps (snags) from living trees. For several decades, the making of snags is probably the most common means by which dead-wood habitats are created in Swedish production forests. There is evidence that high-stumps constitute a suitable substrate for many wood-living insects, e.g. beetles (Jonsell et al., 2004; Lindhe & Lindelöw, 2004; Abrahamsson & Lindbladh, 2006), even if some studies have shown that they are a major source of recruitment for only a few species (Schroeder et al., 2006).

If there is increased species diversity in high-stumps which are located within hotspots, it raises the question as to whether it would be more beneficial to concentrate high-stumps in these zones. In a recent study (Lindbladh et al., 2007), the saproxylic beetle faunas on spruce (Picea abies) and birch (Betula pendula/pubescens) were quantified for high-stumps located in clearcuts within and outside hotspot areas. The authors found only minor differences in saproxylic beetle diversity between the two landscape types. In their study, bark sieving was used for sampling the species. This method gives rather small samples and captures only a small subset of the saproxylic beetle community, namely species assemblages closely associated with the bark (Wikars et al., 2005; Alinvi et al., 2007). If the differences in beetle communities are limited to rare species, large quantities of individuals have to be collected to obtain sufficient statistical power to detect an effect (Martikainen & Kouki, 2003; Magurran, 2004).

Here, we replicate the study design of Lindbladh et al. (2007) with the exception of using window traps to maximise the proportion of rare beetles sampled (Hyvärinen et al., 2006b). Furthermore, the environmental settings around clearcuts will be compared by using two different GIS data sets with the information on landscape and forest configuration, e.g. area and volume of tree species. The aim is to test whether the lack of difference between beetle faunas in the previous study was because too few individuals had been caught and/or a result of minor differences between the landscape properties within and outside the hotspot areas.

The specific questions are:

  • 1 Are there differences in number of species or in beetle species compositions on clearcuts within and outside hotspot areas?
  • 2 Which differences are there in forest configuration in the surroundings of clearcuts within and outside hotspot areas?
  • 3 Are variables other than hotspot/non-hotspot more important for explaining patterns of species occurrence?

Materials and methods

Landscape settings and survey design

Forests occupy 67% and national parks and reserves ca. 2% of total land in the four counties where the study was performed (Nilsson et al., 2008). The forests are mainly coniferous monocultures of Picea abies (47% of total volume) and Pinus sylvestris (ca. 30% of total volume) with low biodiversity value (Nilsson, 1997). Temperate broadleaves, e.g. oak (Quercus spp) and beech (Fagus sylvatica), which dominated the landscape not more than ca. 500 years ago (Lindbladh et al., 2000), do not constitute more than ca. 4% of the total volume today. Based on both systematic and non-systematic inventories (e.g. Nilsson, 2001), the Swedish Environmental Protection Agency identified forest areas in southern Sweden that are valuable from a conservation point of view (Andersson & Lövgren, 2000). Even if coniferous production forests also dominate these hotspot areas, they are characterised by a relatively high proportion of temperate broadleaved trees, especially oak and beech, with high conservation value. The hotspot areas (100–300 km2) are often situated in hilly landscapes close to the ocean or to large lakes and in areas with large concentrations of old-cultural landscapes. Five hotspot areas were chosen that have a rich, well-documented fauna of saproxylic beetles, many of which are red-listed (Nilsson, 2001). They are located in an east-west band across southern Sweden with two clearcuts per area sampled (Fig. 1). One of the hotspot areas, Hornsö (easternmost region), has the highest number of red-listed saproxylic beetles recorded in Sweden (Nilsson & Huggert, 2001). For each hotspot area, two clearcuts were located within the hotspot and two in the surrounding landscape. The non-hotspot clearcuts were selected among available clearcuts of the same age and tree species composition as the corresponding hotspot clearcuts. We avoided clearcuts located close to nature reserves. The non-hotspot clearcuts were at least 11 km from the nearest hotspot clearcut, and hotspot clearcuts were at least 2.5 km apart. Although knowledge of beetle dispersal capacity is relatively scarce, we suggest that the sampling locations are sufficiently distant from each other to be considered independent. (Hedin & Ranius, 2002; Jonsell et al., 2003; Jonsson, 2003). Furthermore, some studies have shown that occupancy at a site of a specific beetle is more dependent on habitat availability than on dispersal capacity (Schroeder et al., 2006, 2007).

Figure 1.

 Map over southern Sweden showing the location of designated hotspot areas according to Andersson and Lövgren (2000) (shaded), and the location of the sampled 20 clearcuts. Hotspot clearcuts are white dots and non-hotspot clearcuts are black dots.

Sampling design

In the previous study, two high-stumps of spruce and birch respectively were randomly selected on each clearcut for sieving. The same high-stumps were sampled with window traps in the present study. In total, 80 high-stumps from 20 clearcuts were sampled, 40 birch and 40 spruce. During the data analysis, the two high-stumps of the same tree species and on the same clearcut were pooled giving a total of 40 samples. The clearcuts and high-stumps were made in the autumn/winter of 2001 (except for two clearcuts in the West region which were made in the autumn/winter of 2000). Window traps were attached as close as possible to the high-stumps. Trap dimensions were 40 by 60 cm and the lower edge of the window was placed 1 m above ground. To collect beetles, a funnel with an attached collector was placed under the window. The collector was filled with a 50% mixture of water and glycol to preserve the beetles. Several drops of detergent were added to break surface tension. Sampling was conducted in 2002 (first summer) and 2004 (third summer), and in both years, the traps were mounted in mid-April and dismounted in the end of September. The traps were emptied at least every month, with shorter intervals in June, July and August.

All wood-living beetle species were identified by Rickard Andersson and Markus Abrahamsson and the nomenclature follows Lundberg and Gustafsson (1995). A species was regarded as wood-living according to Dahlberg and Stokland (2004). Red-list categories follow Gärdenfors (2005). However, species red-listed in 2000 (Gärdenfors, 2000) were also included in the study and defined as ‘rare species’. The reason is that many saproxylic beetle species were removed in the revision because the criteria were more strictly applied. Although the discarded species do not reach the new level required for red-listing, they are of higher conservation interest than other non red-listed species.

Remote-sensing data

To investigate possible differences between the surroundings of the hotspot and non-hotspot clearcuts in forest composition, a number of environmental variables were compared using two kinds of GIS data sets: k-Nearest Neighbour algorithm (kNN), Geographical Swedish Data (GSD) land and vegetation cover (see Table 1 for variable explanation). The kNN-data is based on remote sensing from the Swedish National Forest Inventory where spectral images from survey plots are assigned to all pixels with a similar spectral image. From the kNN-data volume, tree species, age, etc. with a 25 × 25-m resolution can be obtained, even if it is recommended to be used in larger aggregates (Reese et al., 2002). The GSD-data from the National Land Survey of Sweden gives area-based information on vegetation and land-use types, such as broadleaved forest, pasture or coniferous forest. Moreover, Woodland Key Habitats (WKH – habitats harbouring, or expected to harbour, red-listed species) was obtained from the GSD-data. Around each clearcut, the kNN- and GSD-data were compiled for two different radii, 500 m (78.5 ha) and 2500 m (1963 ha).

Table 1.   A definition of the variables included in the forest composition analysis and in the CCA analyses.
  1. WKH, woodland key habitats; GSD, Geographical Swedish Data; kNN, k-nearest neighbour algorithm.

WKHArea of Woodland Key Habitats (from GSD-data)
NRArea of nature reserves (from GSD-data)
BroadArea of broadleaved forest (from GSD-data)
CultArea of grazing land and pastures (from GSD-data)
MixedArea of mixed coniferous/broadleaved forest (from GSD-data)
ConifArea of coniferous forest (from GSD-data)
TreespSpruce or birch high-stump sampled
YearSampling year (2002 or 2004)
LongitudeWest-east location of the sampled clearcut
BirchVolume of birch (from kNN-data)
SpruceVolume of spruce (from kNN-data)
PineVolume of pine (from kNN-data)
Broadl otherVolume of alder, aspen, rowan, ash, lime, elm etc. (from kNN-data)
OakVolume of oak (from kNN-data)
BeechVolume of beech (from kNN-data)
HotspotIf the sample were from a hotspot or non-hotspot clearcut

Statistical analysis

Paired t-tests were used to test for differences in the number of species, and the number of rare and red-listed species found within between hotspots and non-hotspot clearcuts. anova was used to test for differences in forest composition in the surroundings of the hotspot and non-hotspot clearcuts. Landscape type (hotspot or non-hotspot) and region (the five regions from west to east) were included as factors.

With the software CANOCO (ter Braak & Šmilauer, 2002), Canonical Correspondence Analysis (CCA) was used to examine which environmental variables best explained patterns in saproxylic beetle species composition (Table 1). This technique arranges the species to maximise the explanatory power of the included environmental variables (Quinn & Keough, 2002). To increase the explanatory strength, singletons were not included and the option downweighting rare species was used. First, a pre-selection with all variables were done and a Monte-Carlo simulation with 499 permutations was run. All variables that contributed significantly according to the Monte-Carlo test were included in the final model (see Table 2). Even if the variable Hotspot did not contribute significantly to the model, it was included in the final model to see with which variables it was correlated to. The variable Treesp (high-stump tree species) explained much of the variance, and therefore we also conducted the CCA ordination for each tree species separately. The separate analyses only included the beetle species associated with each respective tree species according to Dahlberg and Stokland (2004). The variable Year contributed most explanatory power in all runs because of the influence of early successional species in the first year. As the variable was of less interest in this context, it was excluded in the final analysis.

Table 2.   Eigenvalues and correlations from the three CCA analyses.
 Axis 1Axis 2Axis 3Axis 4
  1. *Of species-environment relation.

All samples together. Total inertia = 3.143
 Percent of variance explained*31.349.660.368.9
 WKH 5000.368−0.4410.059−0.393
 WKH 25000.171−0.327−0.389−0.143
 Broad 5000.457−0.4770.056−0.229
 Conif 25000.2240.013−0.3650.102
 Birch 5000.261−0.234−0.214−0.138
 Birch 25000.096−0.144−0.284−0.462
 Beech 5000.526−0.4240.0710.318
 Beech 25000.286−0.2700.145−0.057
Only spruce samples. Total inertia = 2.215
 Percent of variance explained*
 WKH 5000.722−0.3160.475−0.314
 Birch 5000.449−0.2980.073−0.708
 Beech 5000.8580.215−0.055−0.306
Only birch samples. Total inertia = 2.743
 Percent of variance explained*32.353.773.589.3
 NR 25000.141−0.1100.2420.087
 Beech 5000.681−0.3310.336−0.522
 Pine 25000.2300.911−0.053−0.293

The software EstimateS 8.0.0 (Colwell, 2006) was used to compute expected species accumulation curves (sample-based rarefaction), using the analytical formulas of Colwell et al. (2004). The samples from two high-stumps of the same tree species and on the same clearcut for both years were pooled. The sample-based rarefaction curves are not the estimators of species richness; rather it estimates the species richness for a sub-sample of the pooled total species richness, based on all species actually discovered (Gotelli & Colwell, 2001). Therefore, the term species density is used denoting the number of species per unit of habitat.


Species numbers

In total, 39 503 saproxylic beetle individuals belonging to 389 species were caught in this study (Table 3). Of these, 28 species (7%) are on the present red-listed (Gärdenfors, 2005) and an additional 31 species (8%) were also on the previous red-list (Gärdenfors, 2000), i.e. rare according to our definition (all species red-listed in 2005 were also red-listed in 2000). A total of 29 911 individuals were found in 2002 (63% bark beetles) and 9592 individuals in 2004 (15% bark beetles).

Table 3.   The number of saproxylic beetle individuals, species, rare species (red-listed 2000 but not 2005) and red-listed species caught in the study. Note all species red-listed in 2005 also were red-listed in 2000.
Individuals29 911959239 503
Rare species192131
Red-listed species182328

The total number of species caught in each region varied between 214 and 273 (Fig. 2). There were more species in the outside clearcuts compared with hotspots clearcuts in four out of the five regions, although the difference was not significant (paired t-test; t-value −0.5, P-value 0.32). The number of species caught on birch did not differ between hotspot clearcuts and outside clearcuts (t-value 0.26, P-value 0.4). Neither were there any differences for spruce (t-value −-0.66, P-value 0.26). The number of red-listed species did not show any significant difference between hotspots and outside clearcuts for spruce (t-value 1.5, P-value 0.1), birch (t-value 0.15, P-value 0.44), or if pooled together (t-value 1.3, P-value 0.13) (Fig. 3). However, even if not significant, there were more red-listed species caught in the hotspot clearcuts compared with the outside clearcuts in four of the five regions. The East region, the most species rich region, was the exception, and here more red-listed species were found in the outside clearcuts compared with the hotspot clearcuts (Fig. 3).

Figure 2.

 The number of species caught per region (West, Mid-West, Mid, Mid-East and East), landscape type [hotspot (HS) or non-hotspot (NHS) clearcut], tree-species (spruce or birch) and the total number of species caught in each region.

Figure 3.

 The number of red-listed species caught per region (West, Mid-West, Mid, Mid-East and East), landscape type (HS = Hotspot, NHS = Non-hotspot), tree species and the total number of red-listed species caught in each region.

Of the rare and red-listed species sampled, 18 species were found exclusively in hotspot clearcuts, 16 in the outside clearcuts, and 24 species were found in both landscape types. For red-listed species, 11 were found exclusively in hotspot clearcuts, 5 in outside clearcuts, and 12 occurred in both landscape types (Table 4).

Table 4.   Red-listed and rare beetle species individuals caught in the study Gärdenfors (2000, 2005). Rare denotes species red-listed 2000 but not 2005. Red-list category shown for red-listed species 2005 only. Dead wood associations and decay class for each species are according to Dahlberg and Stokland (2004).
SpeciesDead wood association*Decay class†HotspotNon-hotspotRed-list category‡
  1. *Abbreviations for tree species are Con-coniferous, Broad-broadleaves, Aln-alder, Bet-birch, Fag-beech, Pop-aspen, Pic-spruce, Que-oak.

  2. †Decay classes are given as L, living but weakened tree, 1–5 is how advanced the decay is from 1 (newly dead) to 5 (very decayed).

  3. ‡Red-list categories in the 2005 red-list: NT, near threatened; VU, vulnerable; EN, endangered.

Acanthoderes clavipesBet, Pop101 
Agathidium mandibulareCon, Broad2, 510NT
Agathidium nigrinumCon, Bet, Pop2, 510NT
Agrilus biguttatusQueL, 1, 210VU
Ampedus cinnabarinusBroad2, 555NT
Ampedus sanguinolentusAln2, 521NT
Anoplodera scutellataFag2, 3, 510VU
Anoplodera sexguttataQue410NT
Anthaxia similisConL, 1, 210 
Callidium aeneumPic321NT
Cis micansQue550NT
Cis rugulosus??04NT
Colydium elongatumPic, Fag2, 530EN
Corticeus unicolorFag2, 5146 
Cryptophagus longitarsis??86 
Cryptophagus micaceus??20 
Denticollis borealisBet2, 512NT
Dissoleucas niveirostrisBroad2, 310 
Dryocoetes villosusQue1, 212 
Enicmus planipennisCon, Bet, Fag2, 51  
Epuraea deubeliPic210NT
Euplectus brunneusAln, Fag, Que510 
Euryusa castanopteraBroad2, 3, 563 
Glischrochilus quadriguttatusBroadL, 1, 2, 510NT
Globicornis emarginataCon, BroadL, 3, 501 
Grynocharis oblongaCon, Broad4, 501 
Haploglossa gentilisQue501 
Hylis foveicollisPic, Broad3, 4, 561 
Hylis olexaiPic, Broad3, 4, 52411 
Hypoganus inunctusBroad2, 501 
Ipidia binotataCon2, 31422NT
Lucanus cervusQue2, 512NT
Melasis buprestoidesBet, Aln, Fag2, 3, 575 
Mycetina cruciataBroad520 
Mycetophagus piceusQue510 
Mycetophagus populiBroad2, 511 
Necydalis majorCon, Broad1, 2, 3, 511 
Nemadus colonoidesQue511 
Oplocephala haemorrhoidalisBet, Pop, Fag501 
Orchesia fasciataCon, Broad2, 3, 4, 501NT
Orchesia minorCon, Broad2, 4, 501NT
Oxypoda arborea  10 
Phloiotrya rufipesBroad2, 402NT
Platycerus capreaBet, Broad501 
Platyrhinus resinosusBet, Broad1, 2, 501 
Poecilium alniQue?10NT
Prionocyphon serricornisQue508 
Prionychus melanariusQue, Fag3, 4, 501VU
Pyrrhidium sanguineumQue1, 225NT
Rhizophagus picipesFagL, 1, 201 
Sphaeriestes stockmanniCon1, 201 
Stagetus borealisCon4, 511 
Strangalia attenuataQue, Bet, Aln2, 416VU
Uloma culinarisCon, Broad511NT
Velleius dilatatusBroad522 
Xyleborinus saxeseniiQue, FagL, 1, 25641NT
Xylophilus corticalisCon2, 310NT
Xylotrechus antilopeQue1, 2, 322NT
Zilora ferrugineaCon335NT

The rarefaction analysis estimated the total species density of each clear-cut type to be between 160 and 260 species, i.e. between 1/3 and 1/4 of total species pools were caught (Fig. 4). No large differences were found between the hotspot and outside clearcuts in regards to rarefaction curves or estimated total species density. This suggests that the lack of difference in species richness found between the landscape types would probably not have changed even if we had sampled more high-stumps.

Figure 4.

 Species accumulation curves (points connected with a line) and estimated species densities (single right points) according to the rarefaction analysis for the different tree species and landscape types.

Environmental data

Several statistically significant differences in landscape composition were detected between the hotspot and outside clearcuts from the kNN- and GSD-data. At 500 m radius, the kNN-data showed that the amount of spruce, oak and pine (Pinus sylvestris) differed between the two landscape types (Table 5). The volume of oak was twice as high around the hotspot clearcuts compared with outside clearcuts, whereas the opposite was true for spruce and pine. The GSD-data revealed a similar pattern as the kNN-data on the 500 m radius. The area of broadleaved forest was three times as high around hotspot clearcuts compared with outside clearcuts. Furthermore, the area of coniferous forest was double that found around the outside clearcuts compared with the hotspot clearcuts.

Table 5.   Area and volume for the analysed variables in the surroundings of the hotspot and non-hotspot clearcuts. First are the data for the 500 m radius and then for the 2500 m radius for both data sets. The kNN-data is estimated volume (m3/ha) and GSD-data is area (ha). See table 1 for definition of the variables.
 Landscape type
Hotspot meanNon-hotspot meanFP-value
  1. GSD, Geographical Swedish Data; kNN, k-nearest neighbour algorithm.

kNN 500 m
 Birch 5001740714 8240.660.436
 Beech 500666629974.100.710
 Oak 500861242905.010.049
 Spruce 50067 565100 3976.980.025
 Broadl other 50010 90064733.090.109
 Pine 50020 96845 64929.36<0.001
GSD 500 m
 WKH 5003.30.54.740.054
 Cult 5002.
 Broad 50015.04.316.750.002
 Conif 50025.649.227.86< 0.001
 Mixed 5008.
kNN 2500 m
 Birch 2500231 264316 2531.450.257
 Beech 250099 81880 3780.260.620
 Oak 2500117 31999 3130.320.586
 Spruce 25001 074 7402 167 7027.240.023
 Broadl other 2500146 937159 3530.080.778
 Pine 2500772 347617 3620.470.508
GSD 2500 m
 WKH 250032.814.85.100.048
 Cult 2500135.0122.00.130.722
 Broad 2500289.0161.013.390.004
 Conif 2500644.01141.048.20<0.001
 Mixed 2500133.0130.00.020.890

At 2500 m radius for the kNN-data, only spruce differed with twice the volume around the outside clearcuts compared with that found around hotspot clearcuts (Table 5). For the GSD-data, a similar pattern was found at 500 and 2500 m. At 2500 m, the area covered by broadleaves was twice as high around hotspot clearcuts compared with outside clearcuts, whereas the opposite was true for coniferous forest. The area of WKH was larger around the hotspots on the 500 m radius, but not significantly different (P = 0.054). It was, however, significantly different at the 2500 m radius, with about the double area of WKH around the hotspot compared with outside clearcuts.

Species, landscape and environmental variables

The eigenvalues for the three-first axes in the CCA analysis for all samples (both spruce and birch) were 0.223, 0.131 and 0.076 (Table 2). They explained 60.3% of the variance in the species/environmental data among the included variables. Treesp, Beech 500 and Broad 500 (for definition of the variables see Table 1) were correlated with the first axis explaining most of the variation (Fig. 5a). Longitud was negatively correlated to the same axis. Treesp and Longitud were positively correlated to Axis 2, and many variables negatively correlated to the same axis, for instance Broad 500 and WKH 500. The variable Hotspot did not contribute significantly to the model.

Figure 5.

 Axis 1 and 2 in CCA diagram with (a) all high-stumps included; (b) with only spruce high-stumps included and (c) with only birch high-stumps included. Black circles are hotspot sites, grey non-hotspot sites. Length of arrows is proportional to the explanatory power of the environmental variables.

In the CCA ordination for spruce, the eigenvalues were 0.154, 0.080 and 0.062. The first three axes together explained 84.7% of the variance in the species environmental data (Table 2). The variables Beech 500 and WKH 500 were positively correlated and Longitude was negatively correlated with axis 1 (Table 2, Fig. 5b). Longitude was positively correlated to axis 2. The variable Hotspot did not contribute significantly to the model, but was positively correlated to axis 1.

For birch, the eigenvalues were 0.154, 0.102 and 0.095 for the first three axes and they explained 73.5% of the variance in the species environmental data for the included variables (Table 2). Beech 500 was positively correlated and Longitude negatively correlated to axis 1. Pine 2500 was positively correlated with axis 2 (Fig. 5c). Contrary to the CCA analysis for all samples and for spruce Hotspot contributed significantly according to the Monte-Carlo test, and was included in the final model. It had a weak correlation to both axes.


Species number

The total material in this study contained 40 000 individuals and 389 saproxylic species, the latter is five times more than the 66 species found with sieving in Lindbladh et al. (2007). No significant difference in species numbers between the surroundings of the hotspot and outside clearcuts was detected – not for the total number of species (Fig. 2) and neither for the number of red-listed and rare species (Fig. 3). With nearly 400 saproxylic beetle species and almost 40 000 individuals sampled, a difference in beetle species composition and number ought to have been detected if there was one to find (Martikainen & Kouki, 2003). The rarefaction analysis supports this conclusion as the species accumulation curves flattened out in a similar way for both landscape types as more samples were added (Fig. 4). All together, these facts suggest that high-stumps in the hotspot areas in our study do not attract more species than high-stumps outside the hotspot do.

With window trapping, many species caught are not necessarily attempting to breed in the substrate where the trap was mounted (Wikars et al.,2005; Alinvi et al., 2007). These additional species caught (species not likely to be associated with the high-stumps at present) may be the species that could breed in the substrate in a later successional stage, species that are attracted to the high-stumps (probably by odours) but for which the attraction does not lead to a further colonisation or breeding in the stumps, or they may be pure tourists. There is, however, no reason to believe the tourists should be more frequent in any of the landscape types. The window traps provide extra information on the fauna on a larger scale which is desirable in this case as the aim was to compare the two landscape types.

The surroundings

The environmental data showed that hotspot and non-hotspot clearcuts differed in several important aspects. The largest difference was the higher proportion of broadleaved trees around the hotspot clearcuts, a difference that was more evident at 500 m than at 2500 m radius. Higher proportion of broadleaves is an interesting parameter in itself, but is also an indicator of a less intensively managed forest with a larger amount of old trees and more diverse tree flora (Ask & Nilsson, 2004). This indicated difference is supported by the significant higher proportion of WKH:s, in the surroundings of the hotspot clearcuts compared with the non-hotspot clearcuts on the larger scale (2500 m radius), and nearly significantly higher on the smaller scale also (500 m radius, Table 5). WKH:s usually contain more dead wood than the surrounding forests (Jönsson & Jonsson, 2007). Unfortunately, GIS data sets (kNN and GSD) do not measure or estimate the amount of dead wood, i.e. the key resource for saproxylic organisms. To measure dead wood relevant for the questions in this study, very large areas at and around the clearcuts would have to be sampled (Økland et al., 1996; Gibb et al., 2006), something that was not possible in this project. In summary, more broadleaved trees were found in the hotspot surroundings and several of the other variables also indicate that more dead wood is likely to be present in the hotpots, showing that there is really a difference in habitat quality between the hotspot and non-hotspot landscape types.

Species composition

Even if there were no major differences between the hotspot and non-hotspot clearcuts regarding beetle composition, some patterns can be found in the CCA. The proportion of coniferous and broadleaved forest in the surrounding of the sampled clearcuts were the important explanatory variables and, this aspect will be discussed first. Secondly, we continue with other important factors according to the analysis: the east-west location of the sampled clearcuts (variable Long) and the tree species of the sampled high-stumps. Thirdly, this section will be ended with a discussion on how scale dependant issues might influence the results.

The variable Hotspot could explain the species composition of saproxylic beetles to a significant degree for the birch high-stumps in the CCA, but not for spruce or for both. Thus, the hypothesis that birch high-stumps are more important in the hotspot areas was confirmed. However, there were several other forest composition variables associated with the hotspots which explained the beetle composition to a higher degree. Broadleaved forest variables, for instance Broad 500 and Beech 500, contributed significantly in explaining beetle species composition in all three CCA analyses (Table 2, Fig. 5a). These variables give a better description of the landscape qualities than the variable Hotspot, which only denotes that the site is within a biologically richer area than the surroundings. Most of the species, which were positively correlated with broadleaved tree variables in our study, e.g. Trypodendron domesticum, Euryusa castanoptera and Saperda scalaris (data not shown), are also according to the literature associated with broadleaved trees (Palm, 1959; Dahlberg & Stokland, 2004).

Secondly, Longitude was a very important variable in all the three CCA analyses in this study. This is probably related to climatic factors – dryer and warmer to the east – that influence the beetle communities both directly and indirectly. The latter is probably mediated through the distribution of tree species. Beech 500 for instance, was negatively correlated with longitude in all three analyses, which makes sense as beech has a western distribution in southern Sweden. There are a number of species in the study which were only found in the western regions (e.g. Ipidia binotata, Euryusa castanoptera and Cis nitidus). They are, however, probably not associated to beech but to some other factor, for instance precipitation, and the latter species is known to favour moist environments. Other studies have also found a correlation between longitude and beetle fauna which they interpret as connected to differences in precipitation or mean summer temperature between the west and east parts of Southern Sweden (Økland et al., 2005; Franc et al., 2007; Lindbladh et al., 2007). Also the distribution of pine may add to the importance of longitude as the proportion of pine forest is higher to the east Nilsson, and Pine 2500 was an important variable for birch high-stumps.

Thirdly, the scale that we sample and how well it corresponds to the scale that the beetles act upon is important to acknowledge. Some of the sampled clearcuts in hotspot areas had a large proportion of coniferous production forests in the close surroundings, and therefore probably have a beetle fauna similar to the non-hotspot clearcuts. The opposite is probably also true, i.e. that the surroundings of a non-hotspot clearcut may hold unusually many broadleaved trees and dead wood. The size of some of the designated hotspot areas could have an effect on the species composition as a sampled clearcut might have been too distant from the biologically rich source area within the hotspot. It is not probable, however, that the distance between the hotspot and non-hotspot clearcuts could explain the small differences in species composition in our study between the landscape categories. The distance was at least 11 km, and it should be large enough for most species to separate them into different populations. This argument has some support from the spatial variables in the CCA, as the smaller scale (500 m radius, ∼0.8 km2) was more important than the larger scale (2500 m radius, ∼20 km2, Table 2). The result is consistent with other studies on saproxylic beetles that have looked at what scale environmental factors become important. Økland et al. (1996) for instance, found that factors like amount of dead wood and related variables on 1 km2 and 4 km2 were important for saproxylic beetles. Another study by Franc et al. (2007) found that the proportion of oak-dominated woodland within 1 km radius (∼3 km2) was important for saproxylic beetles associated with oak.

Conclusions and consequences for forest conservation

In general, the study suggests that concentrating high-stumps to hotspot areas (as defined in this study) will not benefit more species as high-stumps of spruce and birch are used by a large part of the wood-living beetle fauna regardless of their location. The created high-stumps in this study only are four summers old at the most, and they predominantly attract early successional species. These species are often agile and good disperses (Nilssen, 1984; Forsse & Solbreck, 1985), and they are mostly associated with fresh cambium. Therefore, from these beetle species’ point of view are non-hotspot clearcuts probably not that different from the hotspot clearcuts. In this sense, high-stumps outside hotspots could improve the overall situation for a large number of early successional wood-living species. Making the normal forests a less hostile environment is argued to be important to enhance the dispersal facilities in a landscape (Fahrig, 2001). Furthermore, high-stumps are a rather cost effective conservation action (Jonsson et al., 2006), and the fact that they can be created everywhere further supports the idea of spreading high-stumps instead of concentrating them to hotspots. It is, however, possible other regions in Fennoscandia have true hotspots for the spruce and birch beetle fauna, in particular in northern regions with more biodiversity associated with these boreal tree species (Nilsson, 1997).

However, there are some factors which argue in favour of concentrating efforts to hotspot also in southern Sweden. The fact that hotspot variable did contribute significantly in explaining the beetle composition on the birch high-stumps, but not on the spruce high-stumps could indicate that birch high-stumps should be prioritised in a biologically rich landscape. Moreover, nearly half of the red-listed species were found at hotspot clearcuts only. Many of these species, e.g. Anoplodera scutellata and Anoplodera sexguttata, are associated with old dead wood of broadleaved trees. This implies that hotspots areas have qualities which the outside clearcuts lack, for instance dead wood continuity in the surroundings which is important for many species (Nilsson & Baranowski, 1997; Siitonen & Saaristo, 2000). The hotspot effect (more species, red-listed and rare species) would perhaps have been larger if older wood important for late successional species were sampled; something future studies can give insights in.


We are grateful to the personnel at Sveaskog, Södra skogsägarna, Skogsstyrelsen, Asa and Tönnersjöhedens experimental forests for assistance during the site selection. Thanks to all the landowners who let us use their high-stumps. Jonas Hedin and Sven G Nilsson contributed during the project’s planning phase, thanks. We also thank Rickard Andersson for helping with the beetle identification. The manuscript was significantly improved by comments by Jari Kouki, Bengt-Gunnar Jonsson, Joakim Hjältén, Fredrik Schlyter, Henrik Smith and Sven G Nilsson. Adam Felton and Richard Bradshaw corrected linguistic errors, thanks. The study was funded by Granprogrammet at the Southern Swedish Forest Research Centre.