Wood-Inhabiting Beetles in Low Stumps, High Stumps and Logs on Boreal Clear-Cuts: Implications for Dead Wood Management

The increasing demand for biofuels from logging residues require serious attention on the importance of dead wood substrates on clear-cuts for the many forestry-intolerant saproxylic (wood-inhabiting) species. In particular, the emerging harvest of low stumps motivates further study of these substrates. On ten clear-cuts we compared the species richness, abundance and species composition of saproxylic beetles hatching from four to nine year old low stumps, high stumps and logs of Norway spruce. By using emergence traps we collected a total of 2,670 saproxylic beetles among 195 species during the summers of 2006, 2007 and 2009. We found that the species assemblages differed significantly between high stumps and logs all three years. The species assemblages of low stumps, on the other hand, were intermediate to those found in logs and high stumps. There were also significant difference in species richness between the three examined years, and we found significant effect of substrate type on richness of predators and fungivores. As shown in previous studies of low stumps on clear-cuts they can sustain large numbers of different saproxylic beetles, including red-listed species. Our study does, in addition to this fact, highlight a possible problem in creating just one type of substrate as a tool for conservation in forestry. Species assemblages in high stumps did not differ significantly from those found in low stumps. Instead logs, which constitute a scarcer substrate type on clear-cuts, provided habitat for a more distinct assemblage of saproxylic species than high stumps. It can therefore be questioned whether high stumps are an optimal tool for nature conservation in clear-cutting forestry. Our results also indicate that low stumps constitute an equally important substrate as high stumps and logs, and we therefore suggest that stump harvesting is done after carefully evaluating measures to provide habitat for saproxylic organisms.


Introduction
Dead wood is a key factor in the preservation of biodiversity in forests [1][2][3][4]. In Fennoscandia, were forests are intensively managed, there has been a strong decline in the volumes of dead mainly constituted of managed stands dominated by Norway spruce and Scots pine (Pinus sylvestris L.) in various age classes.
At each of the ten clear-cuts we created three high stumps of spruce (approximately 3 m in height and 21 cm in diameter at breast height) and we introduced three logs of spruce (4 m in length and approximately 21 cm in diameter). At each clear-cut we also used nine low stumps (three per sampling year) of spruce left after clear-cutting. These low stumps were located close to the other two substrate types. The low stumps and the high stumps were created at clear-cutting between 2000 and early 2001. All logs were introduced in early 2002 except in one clearcut, where the logs were introduced in early 2001 [see 14 for details]. All ten clear-cuts were sampled in 2006, 2007 and 2009, approximately five to ten years after clear-cutting. Most of the substrates had some bark still remaining the last year of the study although in many cases detached from the wood. According to the description of successional phases of saproxylic beetles by Esseen et al. [27], our study would be placed in the transition between successional phase II and III. Phase II is characterized by species associated with early fungi growing under the loose bark, and secondary phloem-consuming species. These two groups do also have their own predators, different from the predators feeding on early phloem-consumers common in successional phase I [28]. The subsequent, phase III coincides with a peak in the occurrence of wood fungi and is dominated by fungivores and their associated predators [27,28].
We sampled each substrate unit separately by using emergence traps. This trap type is useful when investigating substrate associations of saproxylic insects [29] because they almost exclusively trap individuals hatching from the sampled substrate. Furthermore, when the aim is to resample the same substrate several times the emergence trap is superior to e.g. sieving or debarking, which is likely to destroy sections of the substrate. The traps were made of black polypropylene weed barrier cloths which were wrapped around a defined section (approximately 20 cm for the low stumps, depending on height, and 30 cm for logs and high stumps) of each substrate type. The cut surface of the low stumps was covered by the trap (Fig. 2). At the top of each cloth we attached a white 250 ml plastic vessel filled to 1/3 with 50% propylene glycol and some detergent to break the surface tension. The diameter of the low stumps was often slightly wider than on logs and high stumps. On the other hand, low stumps were shorter than the enclosed parts of the logs and of the high stumps [see also 18]. To reduce effects of the trapping from previous years, we randomly moved the traps on logs and high stumps so that the same section was only used once. While moving the traps up and down the high stumps, no traps were put lower than 50 cm above ground and never higher than 50 cm below the top. For the same reason that we moved the traps on logs and high stumps, the same low stump was only sampled once during the three years. The sampling period ran from May to September in 2006, 2007 and 2009. The traps were emptied only once per year, in September. Observe that the cut surface was covered by the trap only on low stumps. Note that in drawings B and C, the proportional length of the substrate covered by the trap is longer than in reality.

Species determination and classification
The beetles were counted and identified to species level by experts. We defined saproxylic species according to Speight [30] and we used the saproxylic database [31,32] to get species specific information on connection to dead wood. The species were also classified into nutritional subgroups (S1 Table, including the list of references used to classify the species). The two dominating subgroups were: fungivores (mycetophagous species) and predators (predators and ectoparasitoids). Other nutritional subgroups, represented in smaller numbers, were cambivores (phloem and cortex feeders), detritrivores (necrophagous species) and wood borers (xylophagous species). The nutritional subgroups were partially overlapping because of multiple classifications for individual species (S1 Table). A small number of the species were not saproxylic and were neither included in the Supplementary material (S1 Table) nor in any analysis. Red-listed species were classified according to the 2010 Red List of Swedish Species [33]. Taxonomy and nomenclature of the beetles follows [34].

Statistical analysis
We used a general linear mixed effect model (GLMM) for poisson distributed response to analyze the effects of substrate type (high stumps, logs and low stumps) and year (2006, 2007 and 2009) and their possible interaction on the species richness of all saproxylic beetles, predatory, fungivorous and wood boring saproxylic beetles. Saproxylic cambivores and detritrivores were left out of this analysis because of few observations and a large number of zeroes. Instead of making a separate analysis on the relation between beetle abundance, years and substrate type, we made a univariate analysis on species level (see description below). Low stumps were in contrast to high stumps and logs not resampled. Instead new low stumps were sampled in 2007 and in 2009. Also, some traps were destroyed every year. After deleting data from destroyed and non-working traps (five high stumps, six logs and one low stump; two traps year 1, four traps year 2 and six traps year 4) we remained with three repeated measurements of 30 high stumps, 30 logs and 89 measurements of low stumps (see Table 1). Only in one case was the same substrate affected more than once (one log year 2 and year 4). In the GLMM analysis of species richness, locality and substrate identity were treated as random factors while substrate type and time were treated as fixed factors. Significant (α < 0.05) results were examined with pair-wise Tukey tests. Mixed-effect models do, in contrast to standard repeated measures ANOVA, not require constant correlation structure between time points [35]. Thus, no test for sphericity was done. The residuals of each model were graphically studied and no obvious patterns could be seen.
To analyze effects from substrate type on community composition of saproxylic beetles we used multivariate generalized linear models (ManyGLM). By using corrections for poisson or negative binomial responses this relatively new method can handle zero inflated multiple species count data, which often do not fulfill assumptions of multivariate normal distribution [36,37]. ManyGLM is also better than multivariate methods like PERMANOVA in detecting between group effects in less variable taxa [37]. However, as this method cannot yet handle mixed effect models we analyzed the effect of substrate type for each year separately. We created non-orthogonal contrasts to examine all pair-wise substrate combinations of significant (α < 0.05) results in the main test. To analyze the effect of substrate type on individual species, the pair-wise comparisons were followed by the univariate test procedure implemented in ManyGLM. Because the number of comparisons was high and hence the detectability after adjustment for multiple comparisons very low we report unadjusted p-values for all tests, but interpret the results with caution [38][39][40]. To visualize the community composition in our substrate types we explored the data with non-metric multidimensional scaling (NMDS) [41]. We used the full assemblage in the NMDS ordination. All statistical analyses were carried out in the open source software R [42]. The GLMMs were analyzed in the packages lme4 [43] and multcomp [44], the ManyGLM in the package mvabund [45] and the NMDS ordination was done in the package vegan [46].

Results
We trapped 2,670 saproxylic beetle individuals belonging to 195 species. Total saproxylic beetle abundance varied among substrate types (813 individuals in low stumps, 795 in high stumps and 1,062 in logs) and over years (872 in 2006, 1,105 in 2007 and 693 in 2009). Also, species richness varied among substrate types (62 species in low stumps, 72 in high stumps and 50 in logs). Among these specimens, 1,028 were classified as fungivores, 1,132 as predators, 615 as wood borers, 378 as cambivores and 24 as detritrivores. Seven species were not classified and for 25 species the classification was uncertain (S1 Table). High stumps produced 48 unique species (species only trapped on high stumps), while logs and low stumps produced 34 and 27 unique species respectively. The most abundant species was Ampedus tristis (15% of total abundance) followed by Scaphisoma agaricinum (9.4%) and Crypturgus pusillus (8.9%).

Abundance and species richness
Substrate type did not have a significant effect on total saproxylic species richness, but there were more saproxylic predators in high stumps and more saproxylic fungivores in logs and low stumps ( Table 2). Sampling year had a significant effect on all groups except saproxylic fungivores and in all cases did the second sampling year stand out as a year with higher species richness ( Table 2). No significant interaction between substrate type and year was found. We trapped 60 individuals of 20 nationally red-listed species; 10 species (17 individuals) from high stumps, 9 species (34 individuals) from logs, and 7 species (9 individuals) from low stumps. Four of these species (six individuals) are currently classified as vulnerable, 15 (53) as near threatened (NT) and one (one) as data deficient (DD) (S1 Table). The most common red-listed species was the predator Lacon fasciatus (NT) with 18 individuals, of which 11 hatched from logs (S1 Table). The number of red-listed species and individuals was, however, too low to allow for statistical analysis.

Species assemblages in different substrate types
The community composition of saproxylic beetles differed significantly between high stumps and logs all three years. However, the pair-wise comparisons showed that the species composition in low stumps was not different from high stumps, and different from logs only the second year ( Table 3). The two-dimensional solution of the Bray-Curtis dissimilarity in the NMDS ordinations illustrates that the composition in stumps is intermediate to those in logs and high stumps (Fig. 3). Also, the NMDS plots indicate that assemblages in logs became more distinct in the two later years (Fig. 3).
There were 28 species that, according to the univariate analysis, responded significantly to the effect of substrate type one, at least one study year (Table 4). These species were in many cases also among the most common (S1 Table). Three of these species were caught all three years and all three were significantly more common on logs ( Table 4). Five of the 28 species occurred all three years but was only significantly affected by substrate type two of these three years. Among these five species, three were significantly more common on one substrate type (one on each type), one was more common on logs and low stumps than on high stumps, and one showed inconsistent results. Two red listed species, Corticaria interstitialis (NT) and Lacon fasciatus (NT), were significantly more common on logs than on low stumps (Table 4). Among the remaining 18 species that were significantly affected by substrate type, four were more common on high stumps; six were more common on logs and seven more common on low stumps. The overall result from the analysis of individual species responses to substrate type suggests that high stumps were preferred by fewest species whereas logs were preferred by most. Low stumps were intermediate in this respect with fewer species preferences than logs but richer than high stumps.

Nutritional subgroups
In consistence with previous results [47] we did not find significant differences between logs and low stumps when all species were included in the analysis. However, after dividing species into nutritional subgroups we found that fungivores and predators had different relationships to the studied substrate types. Fungivores were more species-rich on logs and low stumps and predators were more species-rich on high stumps. Taking one step down in the food chain could shed light on the observed pattern. Berglund et al. [48] found that although most of the investigated species of wood decaying fungi were not restricted to logs, they clearly favored this substrate type. It is possible that higher moisture and lower temperature in logs is responsible for this fungal pattern. Also, Jonsell and Weslien [22] (with support from studies by Thunes and Willassen [49] and Jonsell et al. [50]), suggested that the difference between substrate types they found in the abundance of single beetle species could be explained by differences in water content. In our study, both logs and low stumps were sampled close to the ground while high stumps were sampled higher up from the ground. As a consequence, low stumps and logs may have been more similar in terms of moisture level and supply of fungus, hence attracting more fungivorous beetles than high stumps. Yet, moisture level alone does not necessarily cause higher presence of saproxylic beetles [51]. Since the assemblage of beetles in high stumps is different at breast height than in the moister base [23], it is likely that the patterns in our study would have looked different if all high stumps had been sampled also close to the ground. The predator group were in contrast to fungivores more species rich on high stumps. We have no good explanation for this result.

Beetle assemblages and substrate types
Our results, that low stumps, snags and logs on clear-cuts are inhabited by differing assemblages of saproxylic beetles, conforms with the results from previous studies [18,23,47]. Adding the data from 2007 and 2009 to the 2006 data already published in Hjältén et al. [18] revealed (i) that assemblages in low stumps are similar to those in high stumps (Fig. 3, Table 3), (ii) that the species assemblages of logs is more distinct in relation to the ones found in high and low stumps, and (iii) that individual species responses to substrate type are consistent over time  Abbreviations for nutritional subgroups: C = cambivore, D = detritrivore, F = fungivore, P = predator, and W = wood borer. "?" before a classification means that the classification is uncertain; and for substrate types: HS = high stumps, L = logs and LS for = low stumps. HS > L denotes higher abundance in high stumps than in logs. "i" and "ns" denotes inconclusive result and non-significant result respectively and "-"indicate absence of the species. Species that, according to the univariate pair-wise tests, significantly (α < 0.05) responded to the different substrate types ordered from top to bottom in respect to their respective substrate preference and according to the number of years the response was significant (all years, two years or one year).
doi:10.1371/journal.pone.0118896.t004 (Table 4). In earlier comparisons of high stumps and low stumps, results have shown that species assemblages, although widely overlapping, differ significantly in these two substrate types.
Our new insights, in combination with the previous ones, lead us to the suggestion that high stumps and low stumps may lose their specific features with time. The result in the first year, which was also analyzed in Hjältén et al. [18], was different in this new study; no significant difference between high and low stumps. The divergence is hard to explain, but we used a different analysis which might be more sensitive than the PERMANOVA used in Hjältén et al. [18]. After comparing the ManyGLM result with the NMDS plots (Table 3, Fig. 3) it is clear that the results are coinciding; that assemblages in low stumps and high stumps are more alike and that assemblages in logs are more distinct. Also, our new result is similar all three years which further suggests that this pattern is correct.

Responses by individual species
The repeated inference of substrate preference, as the one made for A. bella, C. maculosa and S. agaricinum (significantly affected all three years) and A. nigrinus, A. tristis, A. axillaris and A. striatum (significantly affected two of three years) suggests that substrate preferences change little over time. If this is true for other less common species remains unknown. As with the nutritional subgroups it might be useful to look for answers elsewhere in the food chain. For example, the obligate saproxylic fungivore, Curtimorda maculosa was one of the two species that were significantly more common on logs all three years. The species was during the same period never found on high stumps. According to Dahlberg and Stokland [32] C. maculosa is strictly associated with the bracket fungi Gloeophyllum sepiarium which is commonly found on dry, sun exposed dead wood on clear-cuts. Furthermore, this bracket fungi is found more often on logs than on low stumps and standing dead wood [52]. Another example of a species found significantly more often on logs was the fungivore S. agaricinum. In an earlier study conducted with same methods and on the same clear-cuts as the ones in the study presented here, it was found that this species favored logs before high stumps [14]. Eight species were significantly more common on low stumps. The most common of these was the saproxylic fungivore E. rugosus. This species have been found to favor top boles of spruce before high stumps in one year old substrates [53], high stumps at ground level before low stumps in one to three year old substrates [23] and low stumps before logs in four to five year old substrates [47]. The result from studying our four to nine year old substrates strongly agrees with that of Jonsell and Hansson [47] that low stumps is the favored substrate type in intermediate stages of decay. Anisotoma axillaris, was significantly more common on low stumps than on high stumps two of these three years. Here we can make an interesting comparison between our data on A. axillaris and results published on a large data set from the same area: In Hjältén et al. [14] logs and high stumps with a variety of features were placed in three different forest types and sampled four different years over a period of six years (2001 to 2006). They caught 20 individuals of A. axillaris on logs and high stumps. In the study presented here, 59 individuals of this species were caught; all but five on low stumps (S1 Table). Hence, low stumps appear to be a very important substrate for this species. Enicmus rugosus, on the other hand, is very common and also reproduce on other substrate types than low stumps, it is unlikely that this species would perish if stump harvesting becomes widespread.
The species Ampedus nigrinus, Anaspis rufilabris, Cis comptus, Corticaria rubripes and Nepachys cardiace were all significantly more common on high stumps (Table 4). Fäldt et al. [54] evaluated the attraction of different beetle species to the chemical compounds found in the volatiles released from two commonly occurring species of bracket fungi; Fomes fomentarius and Fomitopsis pinicola. A. rufilabris was significantly attracted to the chemical compounds released from F. pinicola. The bracket fungi F. pinicola is very common brown rotting species which is abundantly occurring on high stumps as well as other substrate types on clear-cuts [22,48,55,56] and a possible positive correlation between stump height and this species have been suggested [56]. This species may therefore be more common on high sumps than on logs. However, we did not investigate the occurrence or the possible substrate preference of any bracket fungi; hence we cannot tell if the substrate preference by A. rufilabris was due to higher occurrence of F. pinicola on our high stumps. The larva of the very common Elaterid species A. nigrinus is a predator living in brown rotted wood while feeding on the larvae of other saproxylic insect species. Also in this case, F. pinicola may have caused the pattern. Yet another species which have been shown to favor high stumps is the fungivore Hadreule elongatula [57]. We only caught three individuals of this species, all on high stumps.
Low stumps are difficult to sample due to the complex structures in this part of the tree. For the same reason, it is also difficult to ensure that the same mantle area or volume is sampled in low stumps compared to the more simply shaped high stumps and logs. Therefore, the comparison between stumps and other substrates is not a straight-forward one, neither in the study presented here nor in previous studies. We judge, however, that at least the results concerning species composition mirrors the real differences between the aboveground parts of low stumps, high stumps and logs.

Implications for sustainable forest management
Our study shows that, in varying degree, beetle assemblages differ between different substrate types. This implies that some species cannot use one substrate type as a substitute for another; a fact which has been carefully noted in previous studies [53,58,59]. We suggest that forest management should create a variety of different substrate types, more than is done today. High stumps of conifers, currently common in the present forestry in this region, are important, but according to our results they should be complemented with logs (see also [58]) and dead wood from other tree species or with varying diameters [20,60]. Low stumps represent a large proportion of the coarse dead wood found on clear-cuts [19,61,62] and they might constitute the dominating substrate for saproxylic species associated with dry and warm dead wood on landscape level. Red listed species does also reproduce in low stumps. In our study, nine beetle individuals of eight red-listed species hatched from 89 low stumps during three growing seasons (S1 Table), which might sound like a minor issue. But it translates into one red-listed beetle per ten low stumps and potentially thousands of red listed beetles on large clear-cuts. In areas with very low amounts of coarse dead wood like the forests in Sweden, clear-cuts may be the last outpost for some warmth demanding saproxylic species. We therefore recommend that stump harvesting should be used with absolute caution until a more thorough evaluation of the importance of low stumps for red listed species is done. We also recommend that high stumps are complemented with logs.