Deadwood and saproxylic beetle diversity in naturally disturbed and managed spruce forests in Nova Scotia

Even-age industrial forestry practices may alter communities of native species. Th us, identifying coarse patterns of species diversity in industrial forests and understanding how and why these patterns diff er from those in naturally disturbed forests can play an essential role in attempts to modify forestry practices to minimize their impacts on native species. Th is study compares diversity patterns of deadwood habitat structure and saproxylic beetle species in spruce forests with natural disturbance histories (wind and fi re) and human disturbance histories (clearcutting and clearcutting with thinning). We specifi cally examine how beetle diversity diff ers in relation to disturbance history and how beetle variation is linked to the diversity of deadwood habitats. Beetle and deadwood data were collected from thirty spruce forests in Nova Scotia and analyzed under three related diversity perspectives: alpha (diversity within local forests); beta (heterogeneity among local forests within disturbance classes); and gamma (cumulative species richness within disturbance classes). Few data support a prediction of lower alpha deadwood and beetle diversity in managed forests, or a prediction of lower gamma species richness in managed forests. Th e beta scale analysis yielded support for the following two hypotheses: (1) beetle assemblages are diff erent in forests with diff erent disturbance histories; (2) turnover of beetle assemblages is higher among naturally disturbed forests than among managed forests. Th e prediction of lower gamma diversity of saproxylic beetle species in managed forests compared to naturally disturbed forests was not supported. Th e lack of diff erences between naturally disturbed and industrial forests in structures that are characteristic of older forests (e.g., large-diameter deadwood) may relate to the presence of residual deadwood in second growth forests lingering from before clearcut harvesting. However, such residual deadwood is only an artifact that will soon decay and not be replaced. Th is suggests that the continuity of deadwood microhabitats for species that depend on old-forest structures is only short-term. ZooKeys 22: 309–340 (2009) doi: 10.3897/zookeys.22.144 www.pensoftonline.net/zookeys Copyright DeLancey J. Bishop et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Launched to accelerate biodiversity research A peer-reviewed open-access journal


Introduction
Disturbance is one of the main determinants of ecosystem structure, composition, and function (Pickett and White 1985;Attiwill 1994;Abel and Stepp 2003).A major concern with respect to biodiversity loss in forest ecosystems is driven by the threat that disturbance imposed by modern forestry has on species diversity through altering the habitat structure (Haila and Kouki 1994;Haila et al. 1994;Siitonen 2001; Langor et al. 2006).Deadwood substrates serve as microhabitats that host an immense diversity of forest species (called saproxylics), primarily invertebrates (Speight 1989;Warren and Key 1991;Siitonen 1994Siitonen , 2001Siitonen , 2003;;Grove 2002).Unfortunately, the physical diversity of deadwood in forests has a high potential for being altered by modern forestry practices.Many saproxylic species are threatened due to loss of habitat (Siitonen 2001(Siitonen , 2003;;Jonsson et al. 2005;Larsson et al. 2006, Langor et al. 2006).Th us, considering the fate of saproxylic species and their deadwood microhabitats in industrial forests is an essential part of any forest management initiative that values native species conservation and ecologically sustainable resource use (e.g., Canadian Council of Forest Ministers 1995;National Board of Forestry Sweden 1996;Canadian Forest Service 1997;Siitonen 2001;Langor et al. 2006).
Th e diversity of deadwood substrates (abundance and richness of physical states) is often changed when natural disturbances are replaced by forestry disturbances (Gore and Patterson 1986;Hagan and Grove 1996;Sturtevant et al. 1997;Siitonen 2001;Langor et al. 2006).Whereas natural disturbances occur stochastically, forestry is deterministic (Attiwill 1994) and aimed toward specifi c economic ends, namely, the maximization of sustainable fi ber productivity.Over recent decades, industry and government in Canada and across the globe have favored even-age management practices of clearcut-replanting or clearcut-regeneration with thinning.Natural disturbances have therefore been largely replaced by even-age forestry.
We predicted several changes to the local diversity of deadwood substrates in human disturbed forests.Even-aged harvesting practices may drive (1) the loss of large diameter deadwood due to short time-spans of stand rotations; (2) a decreased local diversity of deadwood decay structures due to all deadwood being in a single decay cohort which senesces through even-aged stand cycles; (3) the loss of standing deadwood due to the view that standing deadwood is a source of economic ineffi ciency and a safety hazard; and (4) the loss of deadwood from the many native tree species that are selected against because they are not considered economically valuable.
We examined saproxylic beetles (Order Coleoptera) diversity patterns and relationships with deadwood diversity patterns in managed and naturally disturbed spruce (Picea rubens Sarg.or Picea mariana Mill., Pinaceae) forests in Nova Scotia, Canada.Many species have highly specialized saproxylic niches (Speight 1989;Warren and Key 1991;Siitonen 2001;Langor et al. 2006), suggesting that species assemblages should be sensitive to variation in deadwood substrates.Th at most saproxylic beetles can be easily sampled using relatively inexpensive methods (Økland 1996) favors the group logistically as well as ecologically.
Th is study takes an exploratory look at a system that has only recently begun to receive detailed attention in North America (Hammond et al. 2001(Hammond et al. , 2004;;Langor and Spence 2006;Langor et al. 2006).In the Maritime Provinces of Canada only Kehler et al. (2004) and Dollin et al. (2008) have previously examined saproxylic beetle communities in detail.Our basic objective is to identify whether modern forestry is generating invertebrate diversity patterns that are diff erent from those generated by natural disturbances.Th is task can be approached on many scales and from many perspectives on species diversity.Specifi cally, we adopt a three-part view on species diversity, similar to the alpha, beta, and gamma diversity perspectives proposed by Whittaker (1972).Alpha diversity pertains to local diversity or, in our case, diversity measured at the scale of a single forest stand.Beta diversity pertains to the turnover of alpha diversity between or among localities; in other words, the degree to which alpha diversity differs from locality to locality.At the beta scale, we compare the turnover of species and deadwood diversity among local forests within four disturbance history classes (see methods).We use gamma diversity to refer to the total number of species cumulatively sampled within all forests in each disturbance history class.
We conceptualized and tested the following set of specifi c predictions for the alpha, beta, and gamma scales: Alpha scale: (1) naturally disturbed forests exhibit higher local diversity of deadwood (habitat) structures than forests that are under even-age management regimes; and (2) local diversity of saproxylic beetle assemblages within forests is proportional to the heterogeneity of deadwood habitat structures within forests; and thus (3) naturally disturbed forests exhibit higher local diversity of saproxylic beetle species than forests that are under even-age management regimes.
Beta scale: (1) beetle assemblages and forest structures are diff erent in forests with diff erent disturbance histories, which is to say that species and deadwood compositions are more similar within forest disturbance classes than among forest disturbance history classes (see methods); (2) heterogeneity in deadwood habitat structures is higher among naturally disturbed forests than among forests that are under even-age management regimes; and (3) turnover of saproxylic beetle species among forests is proportional to the heterogeneity of deadwood habitat structures among forests; and thus (4) turnover of saproxylic beetle species is higher among naturally disturbed forests than among forests that are under even-age management regimes.
Gamma scale: (1) the cumulative beetle species diversity across all naturally disturbed forests is higher than the cumulative beetle species diversity across all forests under even-age management regimes.

Methods
Study forests.Th irty spruce forests were selected for beetle sampling in central Nova Scotia, Canada, concentrated in two regions: the Liscomb Game Sanctuary (45° 09΄ N, 62° 30΄ W) and an area north of St. Margaret's Bay (44° 44΄ N, 63 ° 54΄ W).Th e forests were chosen to represent one of four general disturbance histories, two anthropogenic and two natural: (1) clearcut origin (CC), (2) clearcut origin followed by pre-commercial or commercial thinning (CC+), (3) fi re disturbance origin (F), and (4) wind disturbance origin (W).We set several criteria for the experimental forests.Canopy trees had to be numerically dominated (70 % minimum) by spruce (Picea spp.).Stands had to be large enough to contain a 40 × 120 m trap-sampling grid with a surrounding 100 m buff er (minimum of 7.7ha in total size) to limit edge eff ects from riparian areas, forests of diff erent tree species composition, and any non-forested areas, including recent clearcuts.Th e closest two centers within forest were 800 m apart while the most distant two were separated by 190 km.Stands were in addition selected to reduce as much as possible geographic spatial dependence.Time since last intense disturbance (clearcut, fi re, severe hurricane, etc.) was limited to a minimum of 30 years, as indicated by mean age of dominant canopy trees.Of the selected forests, mean stand ages ranged from 30 to over 250 years.A concentration of thinned forests in the young age range refl ects the recent implementation of thinning as a management practice.Forest management in Nova Scotia was not strictly regulated and private woodlots constituting 75% of the province could be managed as the owners saw fi t when the present study stands experienced their last disturbance.Th us, rotation ages vary greatly and the selection of stands could not be controlled with regard to ages of stands or specifi c management regimes applied.Clearcutting, however, is the standard practice but subsequent silvicultural practices vary (Nova Scotia DNR 2000).
Th e concentration of wind disturbed forests in the older age range refl ects the fact that wind disturbance in Nova Scotia is generally of low intensity, allowing trees to become much older than does short-rotation forestry disturbance.Managed forests were common and thus chosen randomly from a larger subset of forests; the rarity of naturally disturbed spruce forests, on the other hand, allowed no room for random selection.Table 1 shows the distribution of the study forests among disturbance history classes and age classes.It is evident that a rigorous sampling design with regard to stand age and disturbance was not possible to achieve and has been addressed in the analysis (see Analysis).
Th e remaining three forests were fi re origin and dominated by black spruce (Picea mariana), a species dependent on fi re for regeneration when not in lowland or bog soils (Fowells 1965).Black spruce and red spruce are known to hybridize and they are often diffi cult or impossible to distinguish (Roland and Smith 1969).
Forest habitat structures.In each study forest, deadwood volume and crown closure were sub-sampled in six 15 × 15 m plots located at 40 m intervals along two parallel transects spaced 40 m apart.For each piece of deadwood, the length and diameter at each end were measured (or estimated) to calculate volume based on the frustum of a cone.Tree species and decay states (see Table 2) were recorded for each piece of deadwood to allow independent tallying of deadwood in diff erent classes.To estimate percent crown closure, two methods were used.In most forests, a fl at mirror with 20 equally spaced dots was held horizontal at chest level to determine the proportion of dots occurring in crown-covered space.In remaining forests, a short cardboard tube was used as a scope to visually estimate proportions of crown closure.Applying both methods in one forest to test comparability returned similar results.Th ree crown closure readings were taken at random distances (10-20 m) from each of the six plot centers through late July to early August (during peak deciduous foliage).Within each plot, a healthy dominant canopy tree was cored to estimate tree age for a total of six trees per forest.Descriptions of all variables measured are given in Table 2.
Beetle sampling and identifi cation.To sample saproxylic beetles systematically at each forest, a window fl ight intercept trap (FIT) design was employed.Th e FIT is passive in that it does not attract beetles, thus providing random samples of the fauna that actively fl y in the local environment.Økland (1996) showed that window FITs were eff ective for providing large samples of saproxylic beetles and Wikars et al. (2005) showed that window FITs compared with other methods caught the greatest number of species.In a single season Muona (1999) found that FITs collected 50.6% of the species of beetles (and 48.3% of forest species) in a study in the Oulanka region of Finland.Emergence traps where beetles are sampled as they emerge from experimental logs (Gibb et al. 2006) are logistically diffi cult.Th is was not a popular method for sampling saproxylic beetles in the fi eld when this study was conducted in 1997 and thus, was not considered an option for this study.
We employed a design consisting of two bisecting 30 × 30 cm transparent plastic (Lexan 0.030) panes, covered with a white plastic (Styrene 0.040) conical roof for rain shelter, and attached to a Styrene collecting funnel below.Removable plastic sample jars attached at the base of the collecting funnel held a 50 % diluted ethylene glycol killing/preserving solution.In each forest site, six FITs were hung from existing tree branches, with the bottom of the trapping surface approximately 1 m above the forest fl oor.FITs were located 40 m apart along two parallel transects separated by 40 m.
All FITs were set up during 13-19 May 1997 and run until 13-15 August 1997.Catches were collected every two weeks within 3 days of each other, and the ethylene All beetles were identifi ed to species although only species considered to be saproxylic were employed in subsequent analyses.Saproxylic (sensu lato) is defi ned here in accordance with Speight (1989).Th e inclusion of species in this category was made on a specifi c or, more commonly, generic basis, consulting a wide variety of published sources (commencing with Arnett and Th omas (2000) and Arnett et al. (2002), followed by family-specifi c treatments), or if such information was not readily available in the literature, consulting with specialists of diff erent Coleoptera families listed in the acknowledgments.Appendix 1 gives the full list of saproxylic species caught.
Species identifi cations were initially made by D.J. Bishop employing the reference collection of the Canadian National Collection of Insects, Arachnids, and Nematodes.Subsequently the process of identifi cation was continued by C.G. Majka and employing the reference collection of the Nova Scotia Museum.At both stages the process was greatly assisted by Coleoptera specialists (listed in the acknowledgments) who were able to identify diffi cult species or confi rm determinations done by Bishop or Majka.Th e general systematics and taxonomy follow Arnett and Th omas (2000) and Arnett et al. (2002).
Statistical analyses.Th e aim of the initial analysis was to test whether habitat structures diff ered according to the four disturbance history classes: (1) clearcut, (2) thinned, (3) fi re, and (4) wind.For this, distribution-free Kruskal-Wallis tests (nonparametric analogs to one way analysis of variance) were used, followed by the Numenyi a posteriori multiple comparison tests in cases where the initial tests suggested a signifi cant (p<0.05)diff erence (described in Zar 1999).
To test predictions at the alpha scale, the initial procedure was to reduce the structural information and the beetle assemblage from each forest to meaningful diversity attributes.Th e Shannon diversity index (Brower et al. 1990) was selected to index both entities.
Th e Shannon diversity value was calculated independently for four aspects of deadwood habitat diversity for each forest: (1) decay classes, (2) size classes, (3) standing/ fallen classes, and (4) conifer/deciduous classes.Th e Shannon diversity for measures of beetle assemblage was calculated without refi nement (Table 2).
Th e habitat diversity and beetle assemblage attributes were independently tested for any diff erences with respect to disturbance history using parametric, one-way ANOVA tests, with Tukey HSD multiple comparison post hoc tests to detect the location of any signifi cant (p<0.05)diff erences between any two disturbance history classes.Taking the measures of the beetle assemblage to be the dependent variables and the forest habitat diversity variables as the independent variables, stepwise multiple linear regression analysis (Weisberg 2005) was used to test if and how aspects of the beetle assemblage were related to deadwood habitat heterogeneity.Assumptions of independence and homogeneity of variance were checked with plots of residuals.Th e assumption of normality of residuals was checked with histograms of residuals for each test.
Analysis at the beta scale comparing beetle community turnover and deadwood structure turnover among forest sites in relation to disturbance history class was carried out by non-metric multidimensional scaling (NMS) (see Clark 1993) and SIMPER and ANOSIM analyses in Primer 5 2002 (PRIMER-E Ltd. 6 Hedingham Gardens, Roborough, Plymouth PL6 7DX, United Kingdom).
Secondly, multi-response permutation procedures (MRPP) were used to test whether beetle-and habitat-defi ned forests diff ered across disturbance history classes.Th is non-parametric analog to discriminate analysis circumvents assumptions of normality and homogeneity of variance that could not be met with the current data set.Th e MRPP method also returns the mean distance among sample points within each disturbance history class, which describes the relative heterogeneity among forests in each class.
For the distance measure in MRPP, we used the Sørenson coeffi cient (also known as the Czekanowski or Bray-Curtis coeffi cient).It measures percent dissimilarity (PD) between two samples, calculated as PD = 1-2W/A+B, where W is the sum of shared abundances and A and B are the sums of abundances in individual sample units.
For analysis at the gamma scale, species-sample curves were constructed to test the prediction that cumulative species richness for all naturally disturbed forests (fi re and wind disturbed forests inclusive) was higher than that for all forests under even-age management (clearcut and thinned forests inclusive).Species-sample curves were also constructed individually for each disturbance history class to allow a visual evaluation of the gamma richness trends for each.

Results
A total of 12,151 beetles of 389 species were collected.Of these, 296 species comprising 10,488 specimens in 45 families were determined to be saproxylic species (Appendix 1).Two species, Anapsis rufa Say and Isomira quadristriata (Cooper) were caught at all 30 sites and were the most abundant with 3,880 and 1,129 specimens respectively.Eighty-fi ve species were represented by single captures and 216 species were represented by less than 10 specimens.
Forest structure generally showed high variation within disturbance history classes.Signifi cant disturbance-class diff erences tested by a Kruskal-Wallis analysis were identifi ed for 2 (CROWN X 2 =10.54, p= 0.015, SIZE1 X 2 =8.36, p=0.039, and marginally DECAY2 X 2 =6.34, p=0.07) of the 12 variables tested.Crown closure (CROWN) was highest in wind disturbed and clearcut forests and lowest in fi re disturbed forests.Small-diameter deadwood (SIZE1) showed signifi cantly higher volumes in thinned forests than in either clearcut or wind disturbed forests.DECAY2 was greatest in WIND disturbed sites.
Stand age was not a function of any of the deadwood measures among stands.Alpha diversity.Th e ANOVA analysis to test whether naturally disturbed forests exhibited higher diversity of deadwood structures off ered evidence for one of the fi ve attributes tested.SHAN-DECAY (diversity of deadwood decay classes in each forest) was signifi cantly higher in both fi re (p=0.040) and wind (p=0.004)disturbed forests than in thinned forests (F=4.75, p=0.01).
Th e general linear regression models constructed to determine whether the beetle diversity could be explained by habitat diversity attributes with disturbance class and age as factors resulted in four models (Table 3).
In interpreting these results, it should be borne in mind that the dependant variable is a function of the explanatory variables.If the coeffi cients are negative, the dependant variable is a negative function of the explanatory variable, and if positive a positive function.Th e F statistic represents the full model and the T/P are the univariate results.
Beta diversity.Th e forest sites plotted in the space of the beetle assemblage show two important patterns: diff erences in the inter-site distances within disturbance classes and actual segregation of forest disturbance classes (Figure 1, Table 4).Clearcut forests show the tightest aggregation (mean distance = 0.421), followed by fi re disturbed forests (mean distance = 0.571), then thinned forests (mean distance = 0.581), with wind disturbed forests showing the greatest heterogeneity (mean distance = 0.603) which was statistically signifi cant (Table 4).
Diff erent beta patterns were observed in forests with regard to habitat structures.Figure 1 shows that, unlike the species-defi ned plot, fi re disturbed forests show the  greatest heterogeneity (mean distance = 0.353), followed by thinned forests (mean distance = 0.296) and then clearcut forests (mean distance = 0.217), and wind disturbed forests with the lowest heterogeneity (mean distance = 0.175) but the statistical diff erence was only signifi cant at the 0.1 probability level (Table 4B).However, one must note that an outlying wind disturbed site (number 25) had an extremely high volume of deadwood and was removed due to diffi culty in interpreting the NMS diagram otherwise.Th e MRPP for wind-disturbed forests, when the outlier was included, ranked the wind-disturbed class as second most heterogeneous, behind fi re-disturbed forests.
Th e ANOSIM analysis testing for pair-wise similarities among disturbance classes gave a global R statistic of 0.174 with p=0.09 but none of the pair-wise R values were signifi cant even at p=0.10.Results of the SIMPER analysis on average dissimilarity among pairs of sites by disturbance are presented in Appendix 2.
Gamma diversity.Of the 296 native saproxylic species sampled, 227 (77.2%) were sampled in the naturally disturbed forests and 228 (77.6%) were sampled in the managed forests.Th e fi rst and second order Jackknife estimates of total species richness were 301.7 and 337.2 for naturally disturbed forests and 308.3 and 351.4 for managed forests.Th is does not support the prediction that the total fauna richness would be higher in naturally disturbed forests; in fact, the opposite appears more likely to be true.Th e species-sample curves constructed for the four disturbance classes (Figure 2) suggest the following relationship of species richness across disturbance history classes: thinned>wind>fi re>clearcut.

Discussion
4.1 Alpha diversity.Th e analysis of deadwood structure and diversity at the alpha level revealed two insights about thinned forests: there is both a higher volume of small-diameter deadwood and a lower decay-class diversity in thinned forests compared to wind and fi re disturbed forests.Both pre-commercial and commercial thinning treatments were represented in the thinned forest class (fi ve and three forests, respectively).While thinning treatments are applied to stands of varying ages, the results are typically an immediate increase in the volume of similar-diameter downed deadwood.One would predict that this would lead to local homogeneity of deadwood size-structure, though this fi nding was not supported in our results.Th e senescence of the downed deadwood in a single cohort predictably results in decreased decay variation in space, as supported by the lower decay-class diversity in thinned forests.Th e higher volume of small-diameter deadwood in thinned forests probably refl ects the dominance of pre-commercial thinning in the thinned class, because pre-commercial treatments are applied in forests with young, small trees.Th us, our results suggest that lower diversity and increased volumes of deadwood may be generally associated with the thinning practice.Saproxylic beetle abundance was a positive function of small diameter deadwood volume and the diversity of total deadwood volume.Diversity of saproxylic beetles measured as richness divided by abundance, however, was a negative function of small diameter deadwood and deciduous deadwood but a positive function of intermediate size deadwood.Shannon diversity of saproxylic beetles was a negative function of the diversity deadwood (deciduous or coniferous) volume.Th us the relationships are not clear.Rare beetles were a negative function of crown closure and a positive function of deciduous deadwood volume.
One post hoc observation is worthy of mention: deadwood diversity levels in the managed forests might largely be an artifact of pre-harvest forest environments.Nearly all managed forests in this study were in their fi rst or second even-age rotations.Hence, most of these forests likely contained large amounts of residual deadwood from deciduous species and large diameter trees, patterns one would not expect to endure through continued successive short-term rotations.Since second growth forests will usually be harvested before they can reproduce their own old-forest characteristic structures, we can only anticipate their eventual disappearance.In so much as these decomposing artifacts are microhabitats and resources, we can of course also anticipate the disappearance of the species that depend on them.
One explanation for the lack of clear results may be that deadwood may not have been measured over an area that was adequate to refl ect any deterministic infl uences on local beetle assemblages.One cannot necessarily assume that a sampled array of fl ying beetles will refl ect only the microhabitats available within a few hundred meters of the sample points.In theory, the deadwood sampling area that is ideal for discerning relationships emerges from a combination of 'best΄ areas for individual species and individual forests.In recognizing this quandary of FIT sampling, Økland (1996) tested the habitat relationships for individual beetle species in Norway and found that habitat measurements collected over an area of 32 hectares produced the highest number of relationships.To assume that the measurements in this study represented the surrounding forest environment over this large an area may be unreasonable.
In a study comparing FIT sampling to bark-peeling as methods of determining saproxylic faunas, Siitonen (1994) found that the numbers of species and individuals of aerially-dispersing beetles captured by FITs were unaff ected by the quantities of deadwood in the local area of the trap.He interpreted this to mean that the distribution of such aerially dispersing species was related to larger forest scales and not the immediate local conditions.Since the present study employed FITs, the beetle assemblages sampled may have been refl ective of such larger forest scales.Gibb et al. (2006) found that saproxylic insects trapped using emergence traps showed a poor relationship with deadwood at local scales.Th us, in the future a combination of FIT and emergence trapping should probably be employed and larger stands, if available, should be sampled.Other techniques such as baited Lindgren funnel traps, bark peeling, car nets, hand collecting, and light-trapping may also yield beetles, particularly rare species, not sampled by conventional techniques (Muona 1999).
Beta diversity.While the alpha and gamma analyses focus on the unqualifi ed diversities of habitats and beetle assemblages, within forests and within disturbance classes, the beta analyses examine the relative composition of the beetle assemblages and deadwood structures.Th e beta analyses revealed that actual species compositions of the assemblages were not clearly diff erent among classes, and that higher turnover among naturally disturbed forests was evident.Observations of high turnover of saproxylic beetle species without changes in species richness were made in central Finland by Kaila et al. (1997) when comparing recent clearcuts with mature forests; by Väisänen et al. (1993) when comparing the fauna of birch trunks in old-growth forests with those in managed forests; and by Sippola et al. (2002) in comparing saproxylic beetle diversity in old growth and clearcut regeneration stands in Finnish Lapland.While the current study does not support the prediction of reduced species diversity related to reduced deadwood habitat diversity in managed forests, it does support the claim that managed forests support diff erent faunal assemblages (Niemelä 1997).
Th e larger heterogeneity of fi re disturbed forests is in keeping with studies that indicate that many saproxylic species are adapted to fi re disturbances and subsequent forest succession (Granström 2001;Buddle et al. 2006).Siitonen (2003) identifi ed the virtual elimination of fi re disturbances in Fennoscandian boreal forests as one of two principal conservation concerns for saproxylic species (the other being the low proportion of old growth forests).Similarly, in relation to wind-disturbed forests, Duelli et al. (2002) documented dramatic short-term (over the span of a decade) increases in the species richness and biodiversity of invertebrates, reptiles, and small mammals in severe windthrow sites in Switzerland following a 1990 storm.At these same sites Wermelinger et al. (2002) found profound changes in saproxylic beetles with certain groups (Cerambycidae and Buprestidae) becoming 30 to 500 times more abundant than in adjacent intact forests.Although the dramatic increases in numbers of individuals were relatively short-lived, the composition of the saproxylic beetle fauna became progressively more dissimilar from the control plots.
Gamma diversity.Th e fi nal prediction of lower gamma diversity of saproxylic beetle species in managed forests compared to naturally disturbed forests was not supported.Th e prediction was partly underpinned by an assumption that forest habitat structures diff ered between managed and naturally disturbed forests at the alpha scale, which was not highly supported.If anything, the higher species-sample curve in thinned forests seems to suggest that thinning disturbance generates more species than either wind of fi re disturbance.Otherwise, the non-thinned clearcut forests appeared to have a lower diversity of species than both wind and fi re origin forests.
In Swedish forests Larsson et al. (2006) found that 'new forestry΄ management practices that left large volumes of deadwood in stands improve conditions for some species of saproxylic beetles.Furthermore saproxylic beetles as a group are a composite, refl ective of diff erent stages of wood decay.Sippola et al. (2002) examined saproxylic and non-saproxylic beetle diversity in old-growth and regeneration forests in Finland.Although no signifi cant diff erences were detected in the rarefi ed number of species between old-growth and regeneration stands, the species composition of both saproxylic and non-saproxylic beetles diff ered.Species colonizing recently dead trees, soil-dwelling open-habitat species, and some polypore-dwelling ciids were more abundant in recently cut and regenerating sites, whereas many mycetophagous beetle families were almost completely absent.Overall saproxylic species richness or even species diversity does not discriminate between these diff erent saproxylic components and may mask important diff erences between forest beetle communities in forest stands of diff erent age, composition, or stand history.
Conservation requirements.Th e general claim that modern forestry in northern forests has altered invertebrates by altering forest habitat structure is well supported in the literature (Niemelä, 1997;Martikainen et al. 1999Martikainen et al. , 2000;;Siitonen 2001;Grove 2002;Jonsson et al. 2005).In Nova Scotia, Dollin et al. (2008) found that harvested forest stands had lower Coleoptera species richness and were host to a signifi cantly diff erent suite of species than unharvested stands.Two tentative conclusions from the current study help extend this claim in Nova Scotia forests: 1) the species assemblage of saproxylic beetle species diff ered between managed and naturally disturbed forests, and 2) these beetle assemblages are more variable among naturally disturbed forests than among managed forests.If management is to decrease the altering eff ect of forestry practices, the next requirement is to identify or confi rm the particular aspects of forest structure that are being altered and that are in turn altering biodiversity (Essen et al. 1992;Swanson and Franklin 1992;Noss and Cooperrider 1994;Franklin 1995;Lubchenco 1995;Niemelä 1997;Siitonen 2001Siitonen , 2003;;Grove 2002;Langor et al. 2006).
Research from countries with longer histories of intensive forestry has revealed several cases of invertebrate endangerment due to loss of microhabitats found in deadwood or moribund trees.Some of these microhabitats are lost because they are found in tree species that are not economically desirable, like aspen (Siitonen and Martikainen 1994) and beech (Nilsson and Baranowski 1997).Others are threatened because they are part of trees with large diameters, which are not maintained amid short-lasting even-age stand rotations (Økland et al. 1995;Nilsson et al. 1995;Siitonen and Saaristo 2000).Th ese and other factors have lead to concern about the diversity of deadwood microhabitats in general, as defi ned by decay states, sizes, and tree species (Økland et al. 1995).
If the goal in managing forests in Nova Scotia is to maintain patterns of heterogeneity similar to those that arise after natural disturbances, the indication off ered here by saproxylic beetles is that this goal is not being met.Majka (2007) examined 14 families, subfamilies, and tribes of saproxylic beetles in the Maritime Provinces of Canada and found 59 apparently rare species (representing ≤ 0.005% of specimens from the region) that comprise 33% of the 178 species within these groups.Majka (2007) proposed that this apparent scarcity of a large proportion of the saproxylic fauna might be due to the history of forest management practices in the region.If Nova Scotia has not yet suff ered biodepletion to the extent of European forests, this may only refl ect the fact that not enough time has passed for the deadwood lingering from old-forests in second-growth forests to fully return to soil.

Figure 1 .Forests
Figure 1.Non-metric multidimensional scaling ordination diagrams of forest sites in two-dimensional space defi ned by (A) beetle assemblage and (B) habitat structures.Symbols signify disturbance history class of forests: = CLEARCUT, = THINNED, = FIRE, and = WIND.Numbers identify the specifi c forest.Forest number 25 was excluded as an outlier in (B) due to much higher deadwood volumes than all other sites.

Figure 2 .
Figure 2. Species-sample curves for each forest disturbance class.

Table 1 .
Distribution of forests delineated by age and disturbance history classes.

Table 2 .
Description (A) forest structure variables measured in each forest, (B) deadwood diversity and (C) beetle diversity attributes.

Table 3 .
Signifi cant multiple linear regression models of beetle assemblage attributes (dependent variables) regressed against habitat structure diversity attributes (explanatory variables).Construction by forward and backward stepwise inclusion and exclusion of variables, based on two-tailed p-value of 0.1.T is the test statistic; P is the probability value.

Table 4 .
Results of multi-response permutation procedures (MRPP) analysis, testing for separation of disturbance history classes based on (A) beetle assemblage and (B) forest habitat structures.Mean distance within class can be read as a relative measure of beta heterogeneity within disturbance history classes.P-values associated with multiple comparisons denote signifi cance of group separation.Adjusted multiple comparison a level for p-value signifi cance at 0.05 = 0.009, and for signifi cance at 0.10 = 0.017, based on adjusted a = 1-(1-α) 1/ number °f c °mparis °ns List of saproxylic beetles found in the present study (Av.S, average species richness; S, species richness).