Prescribed fires and retention trees help to conserve beetle diversity in managed boreal forests despite their transient negative effects on some beetle groups


Jari Kouki. Faculty of Forest Sciences, University of Joensuu, PO Box 111, FIN–80101 Joensuu, Finland. E-mail:



  • 1Controlled burning and green-tree retention have been suggested to alleviate the negative effects of forestry on species diversity in boreal forests, but the ecological impacts of these measures are poorly known.
  • 2We studied experimentally the response of four ecological groups of beetles – saproxylics, herbivores, species on ephemeral resources, and litter-dwelling species – to different harvesting intensities and controlled burning in Scots pine-dominated forests. The study included four levels of green-tree retention (0, 10, 50 m3 ha−1, and no harvesting) with burning on 12 of the 24 study sites, covering ~4 ha each. A beetle data of 153 334 individuals representing 1142 species were collected during one pre-treatment (2000) and two post-treatment years (2001–2002), using window traps.
  • 3Species richness increased in all four groups after harvesting, with and without burning, and there were major community-level changes. The species richness of saproxylics and herbivores continued to increase in the second post-treatment year on burned sites, whereas it decreased on many unburned sites. The assemblages were strongly affected by the treatments, but higher volumes of green-tree retention maintained them closer to the pre-treatment structure.
  • 4Although some ecological groups, such as species on ephemeral resources, experienced substantial turnover as a result of burning, populations of species that initially declined recovered. Since the increase in the saproxylics was evident, and the population reductions of other species were transient, we recommend the controlled burning with reasonable volumes of green-tree retention to reduce negative effects of forestry on insect diversity.


Disturbances in forests vary from small-scale (e.g. gap dynamics, small-scale flooding) to large-scale disturbances and affect the forest structure at the stand and landscape level (Angelstam, 1996). Natural large-scale disturbances, of which forest fires and windstorms are the most intensive in northern Europe (Zackrisson, 1977; Esseen et al., 1997; Ryan, 2002), form habitats with extensive biological legacies from the pre-disturbance forest, such as large living or dead trees, but also initiate a natural forest succession at the same time (Lindenmayer & Franklin, 2002).

During the 20th century in northern Europe, natural disturbances have been largely replaced by stand-replacing disturbances of human origin, such as clear-cuttings applied in intensive forestry (Linder & Östlund, 1998). The change in the disturbance regime has led to habitat loss, fragmentation and a decrease in habitat quality. In particular, decrease in the amount of dead wood has been dramatic (Siitonen, 2001). Such changes have had substantial consequences for the biota of forests, and often they have led to loss of species and declines in population sizes (Berg et al., 1995; Hanski & Hammond, 1995; Esseen et al., 1997; Linder & Östlund, 1998; Siitonen, 2001).

An adequate reserve network has often been regarded as the principal method for alleviating the consequences of habitat loss to species diversity. The role of the matrix outside and between reserves in the conservation of forest biota has also been emphasised recently (Mönkkönen & Reunanen, 1999; Kouki et al., 2001; Lindenmayer & Franklin, 2002; Lindenmayer et al., 2006; Tikkanen et al., 2007). These areas may provide essential habitats for a number of species, if their requirements are taken into account during management operations (Martikainen, 2000; Kouki et al., 2001). The conservation of biodiversity in production forests seems to be essential, since the current coverage of the reserve network may not be able to include all species or maintain the forest biodiversity in the long term (Niemelä, 1997; Hanski, 2000). Furthermore, in some cases it may be more cost-effective to create substrates important for red-listed species (such as dead wood) in managed forests than to set aside new reserves (Jonsson et al., 2006; Tikkanen et al., 2007).

It seems obvious that natural disturbance regimes cannot and will not be restored on a large scale in north European forests as long as timber harvesting remains the main form of land use in these ecosystems (Brown et al., 2004; Kauffman, 2004). Consequently, conservation activities and techniques applied on a smaller scale are more realistic, but their consequences and effectiveness have so far remained poorly explored. The use of controlled burning and green-tree retention has often been suggested as a means of improving the quality of managed boreal forests (Franklin et al., 1997; Niemelä, 1997; Kouki et al., 2001; Siitonen, 2001; Ehnström & Axelsson, 2002; Penttiläet al., 2004). These measures potentially improve living conditions of saproxylic species considerably, i.e. species that are dependent on dead wood (Speight, 1989). Saproxylic species are the ones that suffer most from modern forestry and it has previously been observed that living conditions for saproxylic species can be improved in managed Fennoscandian forests by these measures (Hyvärinen et al., 2005; Hyvärinen et al., 2006b).

Additionally, the measures are also likely to have significant effects on other groups of forest-dwelling beetle species, too, such as herbivores and epigaeic species. Although some groups – in particular the ground beetles, Carabidae – have received much attention (e.g. Niemelä, 2001; Martikainen et al., 2006a; Matveinen-Huju et al., 2006; Nunes et al., 2006), studies that would simultaneously compare the responses of different groups of beetles are lacking. Since forest ecosystems are inhabited by a wide array of taxa with very variable ecological properties (2000 species of beetles in Finland, for example), extensive studies are needed to reveal the effects of different management methods on them. In principle, a management activity aimed at enhancing survival of a particular group may be harmful to another group. For example, intensive fire may reduce populations of many species (Paquin & Coderre, 1997; Wikars & Schimmel, 2001), although being beneficial for the pyrophilous species. Thus, different taxa must be addressed separately to understand the consequences of conservation and management actions, and this was the main purpose of the present study.

We created a large-scale field experiment to study the effects of different volumes of green-tree retention and controlled burning on the biota of boreal forest. In this study, we specifically test the following null hypothesis:

Different ecological groups of beetles – saproxylic species, herbivores, species dependent on ephemeral resources, and litter-dwelling species – respond in a similar way to forest management actions that aim at maintaining beetle diversity in managed forests. The management actions tested include the use of prescribed (controlled) fire after timber harvests and the reduction of harvest intensity (i.e. the amount of retention trees).

In addition to this null hypothesis, we also examine if the positive responses – in terms of species richness – observed previously (Hyvärinen et al., 2005) among the saproxylic species are maintained over the longer time period of two post-treatment years, or are the effects restricted to a very short term?

Material and methods

Study area

The study area is situated in eastern Finland (municipalities Lieksa and Ilomantsi, approx. 63°N, 30°E), close to the border with Russia in the transition zone between the south and middle boreal vegetation zones (Ahti et al., 1968). The mean annual temperature in the area is +2.0 °C, and July is the warmest month, with a mean temperature of +15.8 °C. Mean annual precipitation is 600 mm, of which about 40% falls as snow. The duration of the growing season is 150 days (Ilmatieteenlaitos, 1991). The forest landscape in the area consists mostly of managed forests dominated by Scots pine (Pinus sylvestris L.), although sites dominated by Norway spruce (Picea abies [L.] Karst.) are relatively common, too. The forest landscape is fragmented by mires, clear-cut areas, roads and small water bodies. Although most of the forests are managed for timber production, some reserves do exist.

Forest sites and experimental design

The 24 sites assigned to the experiment were situated within an area of 20 × 30 km and varied in size from 3 to 5 ha. At the beginning of the experiment, they contained 150-year-old forests dominated by Scots pine (72% of the growing stock), with Norway spruce comprising 22% of the growing stock on average, birches (Betula spp.) 5% and other deciduous trees such as aspen (Populus tremula L.) and white alder (Alnus incana [L.]) 1%. The mean volume of living trees was 287.9 m3 ha−1 (SD = 71.1) and that of decaying wood 40.8 m3 ha−1 (SD = 17.5), of which 36% was contained in downed logs. Selective harvesting had taken place at all of the sites, dating back to 1950s and before, but intensive modern forestry with extensive timber harvesting was not yet being practised anywhere. Signs of previous forest fires were found at all of the sites (Kaipainen, 2001). Eighteen of the sites had been burned during the 19th century, when slash-and-burn cultivation was common in the area (Lehtonen et al., 1996).

The factorial experimental design followed the before-after-control-impact (BACI) principle (Green, 1979) and focused on two factors, fire and the volume of green-tree retention. ‘Green-tree’ is a tree that is left unharvested on the site. The experiment had eight treatment combinations, each of which was replicated three times (see Hyvärinen et al., 2006b). The volume of tree retention had four levels: 0, 10 and 50 m3 ha−1 and the whole standing volume, the logging being implemented during winter 2000/2001. The trees were retained primarily in small groups, forming 3 tree groups ha−1 on the 10-m3 ha−1 sites and 5 tree groups on 50-m3 ha−1 sites, but also a few solitary trees were left. Twelve of the 24 sites were burned on 27–28 June 2001, i.e. in the summer following the harvests. For a detailed description of the burning procedure, see Hyvärinen et al. (2005). The treatments were assigned to the sites at random, except for the unharvested sites, which were situated within the Patvinsuo National Park. The forests on these sites were nevertheless similar to those outside the park despite their national park status. After the experimental treatments, study sites were left to regenerate naturally.

Changes in the depth of the humus layer and average flame height were measured to record possible differences in the severity and intensity of fire between different harvesting intensities. The humus layer became 27% thinner on average at the harvested sites and 8% thinner at the unharvested sites as a result of burning, but there was considerable small-scale variation within sites (Laamanen, 2002). The scorch height was measured by estimating the height of the charred bark on the retained trees (Sidoroff, 2001). The mean heights of the flames were 2.2 m on the unharvested sites, 3.9 m on the sites with 50-m3 retention trees ha−1 and 5.8 m on the sites with 10-m3 retention trees ha−1. This method could not obviously be applied to the sites where no trees were retained.

Sampling of beetles

The beetles were sampled in one pre-treatment year (2000) and two post-treatment years (2001–2002) using 10 freely hanging flight-interception traps on each site, giving a total of 240 traps each year. Each trap consisted of two perpendicular transparent plastic panes (40 × 60 cm) attached to a funnel leading to a 1-L container which held salt water and few drops of detergent to preserve the insects. No attractants were used. The traps were suspended from strings between two poles, or trees if available.

Traps were set up in 1-ha quadrats established in the middle of each site. The traps were placed in a U-formation with distance of 20 m between neighbouring traps. After the treatments (in 2001 and 2002), the U-formation was changed a little, so that five traps on the 50-m3 sites and three on the 10-m3 sites were placed within the tree groups, corresponding to the number of tree groups per hectare. Sampling continued almost throughout the growing season, and the traps were emptied once a month. The sampling period was 16 May–1 September in 2000. During the burning year 2001, the traps could be set up only after the burnings in those sites where burning was applied (to prevent fire damages to traps). Thus, in 2001, the sampling was carried out 28 June–7 September on the burned sites and 14 May–7 September on the unburned sites. In 2002, sampling continued from 13 May to 12 September.

Identification and classification of beetles

The beetles could be identified to species level with only a few exceptions. Three species of Staphylinidae, Amischa nigrofusca (Stephens), A. analis (Gravenhorst) and A. bifoveolata (Mannerheim), were treated as a triplet due to their high abundance (3967 specimens) and laborious identification, while Atheta crassicornis (Fabricius)–A. paracrassicornis Brundin (44 specimens), Stenus nanus Stephens–S. assequens Rey (one specimen), Stenus tarsalis Ljungh–S. bohemicus Machulka (six specimens) and Scymnus frontalis (Fabricius)–S. mimulus Capra & Fürsch (one specimen) were treated as species pairs, since their females are almost impossible to identify to species. Each triplet or pair was counted as one species in the analyses. One hundred and eight specimens could not be identified to species level, mostly due to deterioration in condition, and were not included in the data. Aquatic species (115 individuals of 18 species) were also excluded from the analyses. The nomenclature follows (Silfverberg, 2004).

The remaining 153 334 beetles represented 1142 species (Table 1), which were classified into four ecologically distinct groups: (i) obligatory saproxylic species, (ii) herbivores, (iii) species living on ephemeral resources such as fungi (Agaricales, Boletales etc.), dung and carrion, and (iv) litter-dwelling species, which also included many facultatively saproxylic species that primarily dwell on other rotten plant material than wood. As many facultative saproxylics often live as predators in litter or on various forms of decaying matter, these were not placed in a separate group. Hence, four groups of species were used in the analyses. The classification was based on several published sources (Saalas, 1917, 1923; Palm, 1948–1972; Palm, 1951, 1959; Koch, 1989–1992; Ehnström & Axelsson, 2002) and our own experience.

Table 1.  Numbers of species and individuals of saproxylics, herbivores, species dependent on ephemeral resources (ephemeral) and litter-dwelling species (litter dwelling) in the pre-treatment (2000) and two post-treatment years (2001, 2002). Corresponding figures with the first sampling period in each year excluded are given in parentheses.
Species group200020012002Total
No. of species
 Saproxylics   254 (232)  330 (297)  346 (301)   402 (379)
 Herbivores  41 (38)   75 (57)   96 (64)   117 (90)
 Ephemeral  120 (108)  174 (143)  131 (108)   214 (193)
 Litter dwelling  157 (139)  289 (249)  328 (240)   409 (337)
 Total  572 (521)  868 (750)  901 (717)  1142 (999)
No. of individuals
 Saproxylics10106 (6222)29793 (17652)28408 (11424) 68307 (35298)
 Herbivores  820 (290) 2099 (1003)20773 (1697) 23692 (2990)
 Ephemeral 3279 (2680) 3486 (1922) 3111 (1505)  9876 (6107)
 Litter dwelling 3890 (2608)32423 (29842)15146 (6877) 51459 (39327)
Total18095 (11923)67801 (50491)67438 (21648)153334 (83722)

Statistical analyses

The 10 subsamples obtained from the traps in each site were pooled to form one sample, as also were the samples from different periods of the year. This guaranteed the use of proper error term and error d.f.'s in the analyses, and avoided pseudoreplication of the treatments. Due to the shorter sampling time period on the burned sites in the first post-treatment year, however, the samples from the first periods of each year were excluded from all the analyses to facilitate comparison with the burned sites in 2001. Without the first period, the data included 83 722 individuals of 999 species (Table 1). Thus, there were always 24 samples for each year, one per site, used in the analyses. The effects of burning, tree retention level and sampling years on the species richness were tested with a factorial repeated measures analysis of variance (anova) applied to the four groups of species using spss 13.0 statistical software.

For the non-metric multidimensional scaling (NMDS) and similarity analyses of the species assemblages, species recorded at only one site were excluded. The NMDS was performed using pc-ord (McCune & Mefford, 1999) and the data were log(x + 1) transformed before ordination. The autopilot mode with ‘slow and thorough’ settings was applied, in which a six-dimensional ordination is first performed in order to evaluate the optimal number of axes for the final solution. A three-dimensional solution was recommended by the program and was used for all the groups of species studied. Bray–Curtis (Sørensen) distance was used as the distance measure. Bray–Curtis similarity indices between the assemblages recorded before and after the treatments were calculated with the estimates program (Colwell, 2004).

Methodological aspects

To be able to make reliable comparisons between treatments or areas in diverse species groups such as beetles, large samples are often required (Muona, 1999; Martikainen & Kouki, 2003; Hyvärinen et al., 2006a). Window traps are reasonable for studying beetle communities, and representative samples can be collected with them (Hyvärinen et al., 2006a), but some methodological aspects should be considered when interpreting the results.

Our sampling was standardised so that equal amount of trapping effort (traps and trapping days) was performed in each area. The performance of the traps used here can be expected to have been rather similar on all the harvested sites, but lower on the more shady unharvested ones. The beetles may also have been more active on burned sites than on unburned ones due to the higher temperature caused by the charred environment, which is efficient in absorbing solar radiation. This bias is sometimes corrected by re-sampling the data, or by calculating species accumulation curves. These approaches tend to underestimate the species richness, however, if there is even one very abundant species in the sample. In this study, there were a few overabundant species such as Orithales serraticornis (Paykull) and Corticaria ferruginea Marsham in the burned areas that also occurred on the unburned sites but with a much lower abundance. We can thus assume that the results regarding species richness based on unstandardised samples can be considered reasonably reliable and justified in current case.


Species richness

In the pre-treatment year (2000, all sampling periods included), the 24 study sites showed no differences in the richness of saproxylics (one-way anova: d.f. = 7, 16, F = 2.378, P = 0.072), herbivores (d.f. = 7, 16, F = 0.886, P = 0.539), species dependent on ephemeral resources (d.f. = 7, 16, F = 1.020, P = 0.454), or litter-dwelling species (d.f. = 7, 16, F = 1.510, P = 0.233).

The treatments applied in 2000/2001 had variable effects on the species groups, and statistical interactions between the effects and time of sampling were also observed (Table 2). Thus, interpretation of the results becomes somewhat complex. In all of the groups, there were clear effects of time (see the within subjects analyses), indicating the remarkable changes in assemblages during the three study years. One general pattern emerged: an increase in species richness in all groups on the harvested sites with or without burning in the first post-treatment year (Fig. 1). In the second post-treatment year after the harvests, the numbers of species in several ecological groups (14 out of 24, Fig. 1) began to decrease. During the two post-treatment years, the species richness in several groups changed considerably (Table 2, Fig. 1). For the group of saproxylics, species richness increased in most of the burned sites, but started to decline in unburned sites in 2002 (Table 2, Fig. 1). Also for the herbivores, the species richness increased clearly on harvested and burned sites, but on the unburned sites the changes were less dramatic (Fig. 1). For the species on ephemeral resources, the effects of burning were not clear and unambiguous, although there was a tendency for an initial increase followed by decrease (Fig. 1). Finally, the litter-dwelling species showed a very steep increase first, especially after the burnings in 2001, and subsequent decline. Unharvested sites behaved differently in burned and unburned sites, probably causing the significant three-way interaction in the results. For saproxylics, herbivores and species dependent on ephemeral resources, the effects of burning and tree retention level were independent of each other (when time was included in the analyses), but for litter-dwelling species these factors were interdependent (Table 2). The between-subject effects (that ignore time effect) were not considered informative or easily interpretable because of the numerous within-subject effects and their interactions.

Table 2.  Repeated measures factorial anova results for the effects of sampling years (time), burning and tree retention level (TRL) on the richness of saproxylics, herbivores, species dependent on ephemeral resources (ephemeral), and litter-dwelling species (litter dwelling). The sphericity condition was not met in the case of herbivores, and thus Greenhouse–Geisser adjustment for degrees of freedom was applied.
Species groupSourced.f.MSFP
SaproxylicsBetween subjects
 Burning 12.720.0150.904
 TRL 33744.3220.770< 0.001
 Burning*TRL 31087.946.0350.006
Within subjects
 Time 27994.7647.621< 0.001
 Time*burning 24238.8525.249< 0.001
 Time*TRL 6876.585.2210.001
 Time*burning*TRL 6333.851.9890.097
HerbivoresBetween subjects
 Burning 1203.3526.005< 0.001
 TRL 373.469.3940.001
 Burning*TRL 325.273.2320.050
Within subjects
 Time 1.409186.0219.390< 0.001
 Time*burning 1.40972.327.5380.006
 Time*TRL 4.22671.887.492< 0.001
 Time*burning*TRL 4.22610.291.0720.396
EphemeralBetween subjects
 Burning 180.221.3120.269
 TRL 326.130.4270.736
 Burning*TRL 366.821.0930.381
Within subjects
 Time 2283.9310.265< 0.001
 Time*burning 243.601.5760.222
 Time*TRL 685.953.1070.016
 Time*burning*TRL 658.132.1020.081
Litter dwellingBetween subjects
 Burning 13280.5037.109< 0.001
 TRL 31472.8516.661< 0.001
 Burning*TRL 370.830.8010.511
Within subjects
 Time 25854.63118.809< 0.001
 Time*burning 2560.3811.372< 0.001
 Time*TRL 6444.269.015< 0.001
 Time*burning*TRL 6159.603.2390.013
Figure 1.

Species richness of saproxylics, herbivores, species dependent on ephemeral resources and litter-dwelling species on burned and unburned sites in one pre-treatment year (2000) and two post-treatment (2001, 2002) years. The data are means and standard errors per study plot in each treatment combination (n = 3 for each bar in the figure). Note the different scales on the y-axes.

Composition of assemblages

The assemblages of saproxylic, ephemeral and litter-dwelling species were clearly affected by the harvesting and burning treatments (based on NMDS, Fig. 2), but the assemblages of herbivores were originally more heterogeneous between the sites and did not show such a marked change in response to the treatments.

Figure 2.

The first two dimensions of the three dimensional solutions to the NMDS ordinations for (A) saproxylics, (B) herbivores, (C) species dependent on ephemeral resources, and (D) litter-dwelling species. The final stresses for the ordinations are 24.69, 20.91, 25.37, and 25.70, respectively. Black symbols represent burned sites and white symbols represent unburned sites. The size of a symbol indicates the volume of green-tree retention, smallest for 0 m3 ha−1, slightly larger for 10 m3 ha−1, larger still for 50 m3 ha−1, and the largest for unharvested sites. The third dimensions of the NMDS ordinations contained little relevant information for the interpretation of results and are thus omitted.

Harvesting and burning had distinct effects on the composition of the saproxylic species, and the assemblages also differed between the two post-treatment years (Fig. 2a). In the first year, the burned and unburned sites had somewhat different assemblages, particularly when tree retention levels were also taken into account. The volume of green-tree retention increases from lower right to the upper left in the ordination space. In the second post-treatment year, the difference between the burned and unburned sites remained fairly similar, but the assemblages showed a clear change from the first post-treatment year. The burned unharvested sites seemed to have similar assemblages to the control sites in the second post-treatment year with respect to the first two dimensions of the ordination, but these sites fell into a group of their own regarding the third dimension (not shown in the graphs).

Herbivores did not show such a clear grouping according to treatments and years as did the saproxylic species (Fig. 2b), but the assemblage composition of burned and unburned sites did differ slightly from each other within the years, just as the harvested sites differed from the unharvested ones.

The assemblages of species that depend on ephemeral resources were affected by the burn treatment in the first post-treatment year (Fig. 2c). By the second post-treatment year, the assemblages had already moved closer to the pre-treatment structure, but differences in assemblage composition between burned and unburned areas were still apparent. The effects of tree retention were less obvious in both years than the effects of burning.

The assemblages of litter-dwelling species differed distinctly between the burned and unburned sites in both post-treatment years (Fig. 2d). The tree retention level seemed to affect the composition of the assemblages in burned sites, the change compared to pre-treatment assemblages was the greater the higher the harvesting intensity. In unburned harvested sites, tree retention level did not seem to affect the composition of the assemblages. In the first post-treatment year, the burned unharvested sites differed from the harvested ones and the pre-treatment samples. In the second post-treatment year, the assemblages on the burned unharvested sites differed from those on the control sites with respect to the third dimension (not shown).

The site-specific similarities (Bray–Curtis) between the pre-treatment assemblages and the assemblages for the post-treatment years were always smaller on the harvested sites than on the unharvested ones (Fig. 3), causing the significant main effect of tree retention level in all species groups (Table 3). This difference was much smaller in the second post-treatment year than in the first, however (Fig. 3). Burning reduced the similarities very significantly only among the saproxylics and litter-dwelling species in the first post-treatment year, and the similarity remained at a low level among the saproxylics in the second post-treatment year as well (Table 3, Fig. 3). The litter-dwelling species exhibited the opposite pattern in the second post-treatment year, as their similarities to the pre-treatment catches were higher on the burned sites than on the unburned ones. For the herbivores, the differences in similarity values were not as clear: only the retention level had consistent effect on both years (Table 3, Fig. 3). Burning had no obvious effect on the similarities of the species that are dependent on ephemeral resources. The effects of burning and tree retention level were independent of each other in all cases except for the litter-dwelling species in 2000 vs. 2001 (Table 3).

Figure 3.

Site-specific Bray–Curtis similarity indices between the pre-treatment and post-treatment years for the assemblages of saproxylics, herbivores, species dependent on ephemeral resources and litter-dwelling species. The data are means and standard errors. Note the different scale of the y-axis for the litter-dwelling species.

Table 3. anova results for site-specific similarities (Bray–Curtis) between pre-treatment assemblages (2000) and post-treatment assemblages (2001, 2002) for saproxylics (saproxylic), herbivores (herbivore), species dependent on ephemeral resources (ephemeral), and litter-dwelling species (litter).
Species groupSourced.f.MSFP
  • *

    The assumpions of anova were not met and thus the result is not entirely reliable.

Saproxylic 2000 vs. 2001Corrected model 70.06130.081< 0.001
Burning 10.267132.524< 0.001
TRL 30.04823.792< 0.001
Burning × TRL 30.0042.2220.125
Saproxylic 2000 vs. 2002Corrected model 70.0147.976< 0.001
Burning 10.02715.1520.001
TRL 30.02212.270< 0.001
Burning × TRL 30.0021.2890.312
Herbivore 2000 vs. 2001Corrected model 70.0432.2660.083
Burning 10.0934.8290.043
TRL 30.0633.2700.049
Burning × TRL 30.0080.4080.749
Herbivore 2000 vs. 2002Corrected model 70.0211.9290.131
Burning 10.0030.2340.635
TRL 30.0474.1790.023
Burning × TRL 30.0030.2440.865
Ephemeral 2000 vs. 2001Corrected model 70.0244.4050.007
Burning 10.0050.8380.374
TRL 30.0519.4060.001
Burning × TRL 30.0030.5940.628
Ephemeral 2000 vs. 2002Corrected model 70.0182.3760.072
Burning 10.0172.1400.163
TRL 30.0324.1220.024
Burning × TRL 30.0050.7080.561
Litter 2000 vs. 2001*Corrected model 70.11027.637< 0.001
Burning 10.599150.567< 0.001
TRL 30.0379.4210.001
Burning × TRL 30.0194.8760.014
Litter 2000 vs. 2002Corrected model 70.0375.1770.003
Burning 10.12817.8990.001
TRL 30.0334.5670.017
Burning × TRL 30.0111.5450.241


Effects of burning

Considerable differences were found among the responses of the ecological groups of beetles – saproxylics, herbivores, species dependent on ephemeral resources, and litter-dwelling species – to controlled burning with different levels of green-tree retention. In terms of species richness, all four groups seemed to increase in numbers after harvest with or without burning, but further analyses suggested that the observed increase in species number was a result of more complicated responses, including major community level changes in beetle assemblages that remain hidden if only the number of species is analysed (see also Niemelä, 1997).

The number of saproxylic species clearly increased immediately after cuttings. However, in the second post-treatment year this increase was evident only on the burned sites (Fig. 1). This pattern may be because logging residue, which provides a substrate for many saproxylic species living on fine woody debris (Jonsell et al., 2007), was available on unburned sites and provided a short-term resource for species. The richness of saproxylic species on the burned sites continued to increase in the second post-treatment year, indicating on-going colonisation, successful reproduction and/or longer-term availability of dead wood. A similar pattern has previously been observed for red-listed and rare saproxylic species (Hyvärinen et al., 2006b). Observations by Saint-Germain et al. (2004) revealed a similar trend in the abundance of xylophagous species: unlike total catch and Elaterid abundance, the abundance of xylophagous beetles did not drop from first to second post-fire year in burned black spruce forests in eastern Canada. The richness of saproxylic species on the unburned sites decreased in the second post-treatment year, suggesting that, although the number of saproxylics increased at first, many of the species initially arriving at these sites were unable to find suitable resources for reproduction. Burning had killed or weakened many of the trees that had been retained, and thus resources for saproxylic species were created rapidly, whereas on the unburned sites tree deaths were naturally less common during the 2-year post-treatment period, and were postponed to the following years and decades. The rather high occurrence of saproxylic species on the unburned harvested sites in the first post-treatment year can be explained by the olfactory stimuli from recently cut stumps and logging waste, which will have attracted beetles to the areas (Brattli et al., 1998).

Burning seemed to be detrimental to many litter-dwelling species in the short-term. Although the species numbers increased in response to the treatments, the similarities between the first post-treatment year assemblages and the pre-treatment assemblages indicated a very high species turnover, as burning radically altered the litter habitat. It is likely that burning caused direct mortality among the epigaeic species (Paquin & Coderre, 1997; Wikars & Schimmel, 2001). Although the assemblages seemed to recover fairly quickly (see also Moretti et al., 2002; Nunes et al., 2006), as indicated by higher similarity indices in the second than in the first post-treatment year with the pre-treatment assemblages, the entire recovery of original epigaeic assemblages may take decades (Niemeläet al., 1993; Buddle et al., 2006; Pohl et al., 2007). Species living in the soil surface in boreal forests, such as most of the Carabids and many Staphylinids, are probably well placed to colonise disturbed areas rapidly from the surroundings (Niemelä, 2001). The same pattern was also seen in the unburned harvested areas, which indicates that also harvesting operations without burning likewise had a strong but transient impact on these species.

Fire kills most of the deciduous trees on which many forest-dwelling herbivores are dependent, but it also creates favourable conditions for the regeneration of deciduous trees (Esseen et al., 1997). This may explain the rapid recovery of herbivores, particularly on the burned sites, as indicated by the increased similarities in the second post-treatment year. According to the ordination, however (Fig. 2b), the assemblages of herbivores were fairly heterogeneous before the treatments, which makes interpretation of the results more difficult.

The species that are dependent on ephemeral resources were only negligibly affected by the treatments, presumably because of their high mobility in the forest landscape. Suitable resources such as elk dung and rotten fungi were also readily and commonly available on the sites and in their vicinity, and also the species pool of ephemerals is reasonably large in our study region to provide colonisers.

Did tree retention level matter?

Our results indicated that higher tree-retention levels maintained assemblages closer to the pre-treatment structure on both the burned and unburned sites. It has been observed previously that the assemblages of canopy arthropods tolerate quite substantial harvesting of trees in the short-term (Schowalter et al., 2005), but our results showed clear and immediate changes in the composition of beetle assemblages due to harvesting. Groups of trees may provide refugia for some species from the effects of harvesting and burning (Gandhi et al., 2001; Martikainen et al., 2006b), although there are indications that small groups cannot maintain the original assemblages of forest ground beetle species (Koivula, 2002; Matveinen-Huju et al., 2006).

It is obvious that retained trees affect different species through different mechanisms. For saproxylic species, the addition of dead wood resources is relevant and they colonise the newly formed substrate after the trees have died as a result of burning or some other factor. For many other species, such as litter-dwelling ones, the volume of tree retention may not be important per se, but can affect the size of the area not disturbed by harvesting, for example, or the burning result (e.g. fire intensity), thus having indirect consequences for them.

The different levels of tree retention also affected the impact of the burning treatment, as sites with a higher tree retention level had lower flame height and consequently lower fire intensity. Fire intensity affects beetle assemblages in three obvious ways in relation to the tree-retention level. A higher fire intensity (i) will increase tree mortality – creating more resources for saproxylics but less for herbivores, (ii) will increase burning of the ground layer (Schimmel & Granström, 1996) – creating fewer refuges for litter-dwelling species, and (iii) will increase mortality among individuals that are unable to escape (Paquin & Coderre, 1997; Wikars & Schimmel, 2001). The differences in species richness and assemblage composition within the harvested sites were generally rather small, however, whereas unharvested sites differed markedly from them. It may be concluded that although there were differences in fire intensity among the harvested sites, the fires were intense enough everywhere to cause rather similar environmental impacts on the ecosystems. The lack of logging waste on the unharvested sites resulted in a lower fire intensity, and thus the mortality of beetle individuals of pre-treatment beetle assemblages remained lower. There was nevertheless obvious increase in the richness of saproxylic species in the second post-treatment year in these sites. A noteworthy pattern is that unharvested burned sites – that had likely the closest resemblance to natural fires – had more or less different assemblages compared to any other treatment combinations, which indicates that effects of natural forest fire on forest beetles is hard to mimic in harvested areas.

The eventual effects of different levels of green-tree retention on saproxylic species at unburned sites can be evaluated only after the retained trees begin to die or grow old enough to create hollows and other decaying parts. At this point of time, we may conclude only that harvesting operations had marked effects on the beetle assemblages, as observed in several previous studies dealing with litter-dwelling beetles (Niemeläet al., 1993; Spence et al., 1996; Koivula, 2002), and that there were major differences in response between the ecological groups.


The present results can be used to evaluate the consequences of controlled burning and different levels of green-tree retention for one of the most diverse taxa in forest ecosystems – beetles. One important aim for using green-tree retention and controlled burning in forest management is creating habitats and substrates for declined species that are dependent on dead wood, but current results show that they also have both direct and indirect effects on other forest-dwelling species. The groups of beetles studied here showed variable responses, although there were many similarities, too. All of them experienced profound changes in composition of assemblages due to the harvesting and burning treatments. Management of lower intensity with higher volumes of tree retention maintained the composition of assemblages closer to the pre-treatment ones.

If the purpose of controlled burning as a post-harvest operation is to mimic natural disturbance, it seems to fail. Burned unharvested sites are actually the only ones that resemble the conditions following natural forest fires (lower intensity of fire, more patchy burning result, uneven mortality of trees), and they support rather different species assemblages from harvested sites, as shown here. Important structural properties which are currently almost completely lacking in the managed forests in Fennoscandia, such as dead wood, are nevertheless produced by the burning of harvested sites with retention trees. Although some ecological groups experience profound habitat destruction and species turnover due to the burning, e.g. species living in the ground layer, they seem to recover quickly. Since positive effects can be induced on the focal group – saproxylics – by controlled burning and the negative effects on other, often more generalist species seem to be transient, the use of controlled burning should be promoted.

Improvement in the quality of managed forests outside reserves is essential for the conservation of many species currently restricted to reserves. Although the populations in the matrix between the reserves may be more or less temporary, they can still serve as source populations rather than sinks if suitable habitats are actively and frequently created as a part of forest management. Populations of this kind may be very important for the long-term persistence of species in the landscape, e.g. in that individuals originating from managed forests may reverse local extinctions occurring in the reserves (Lindenmayer & Franklin, 2002).

Finally, our findings reveal that the ecological consequences of habitat modifications show quite distinct patterns depending on the ecological group of species studied. This sounds quite obvious, but there are many studies that focus only on a specific taxonomic or ecological group, such as the litter-dwelling species. In the light of our results, we suggest that much more attention should be paid to delineating appropriate ecological response groups when carrying out environmental impact studies or assessing the prerequisites for ecological sustainability in the use of natural resources. The multitude of responses among thousands of ecologically different insect species provide many opportunities for good impact assessments if the species studied are chosen carefully to represent ecologically different life-history characteristics and habitat associations of species.


We would like to thank the Metsähallitus (formerly the Finnish Forest and Park Service) and especially I. Heikkinen, M. Ikonen, P. Leppänen, A. Tervonen and K. Tuhkalainen and their staff for preparing the sites for the experiment and carrying out the harvesting and burning treatments. Forestry students from the University of Joensuu helped in the burning of the sites. We also thank I. Rutanen, H. Lappalainen and S. Karjalainen for their participation in identifying the beetles, and the staff of the Mekrijärvi Research Station (University of Joensuu) for the laboratory work and facilities. Bengt-Gunnar Jonsson and two anonymous reviewers gave valuable comments on the manuscript. The research was funded by the Academy of Finland, the Ministry of Agriculture and Forestry, the Ministry of the Environment, Metsähallitus, the Finnish Forest Industries’ Association and the Finnish Forest Research Institute (all through grants to the author J.K.). E.H. was supported by the Finnish Graduate School in Forest Sciences. The experiments used in this study comply with the current laws of the country in which they were performed.