Long-term priority effects among insects and fungi colonizing decaying wood

1. Priority effects have been hypothesized to have long-lasting impact on community structure in natural ecosystems. Long-term studies of priority effects in natural ecosystems are however sparse, especially in terrestrial ecosystems. 2. Wood decay is a slow process involving a high diversity of insect and fungus species. Species interactions that drive change in communities of insects and fungi during wood decay are poorly understood because of a lack of sufficient long-term studies. 3. In this paper, we followed the colonization and succession of wood-living insects and fungi on cut trees during 15 years, from tree death and onwards, in a boreal forest landscape. We test the long-term priority effects hypothesis that the identity and abundance of species that colonize first affect the colonization success of later-arriving species. We also hypothesize that species interact in both facilitative and inhibitory ways, which ultimately affect habitat quality for a red-listed late-succession beetle species. 4. Possible causal associations between species were explored by path analysis. The results indicate that one bark beetle species, Hylurgops palliatus, and one wood-borer species, Monochamus sutor, which colonized the wood during the first year after cutting, influenced the occurrence of a rare, wood-living beetle, Peltis grossa, that started to emerge from the stumps about 10 years later. The positive effects of Hylurgops palliatus and negative effects of M. sutor were largely mediated through the wood-decaying fungus species Fomitopsis pinicola. 5. The study shows that variable priority effects may have long-lasting impact on community assembly in decaying wood. The study also exemplifies new possibilities for managing populations of threatened species by exploring links between early, well-understood species guilds and late, more poorly understood species guilds.

Succession during decomposition of an ephemeral resource, such as a patch of wood, carrion or dung, so-called heterotrophic succession, is characterized by a particular sequence of species arrival. Decaying wood is typified by a particular sequence of insect and fungus species, many of which are broadly associated with a certain decay stage (Saalas 1917;Palm 1951;Swift 1977;Rayner & Boddy 1988;Boddy 2001;Ehnstro¨m & Axelsson 2002). The first species to colonize a newly dead or dying tree are beetles that feed on the nutrient-rich phloem and fungi that colonize broken or cut stems or scars in the bark. The sequence of arrival is best known among these pioneer species because they include those that kill trees or destroy timber with commercial value. Nevertheless, wood may take decades to decay completely, during which time the community structure of insects and fungi change as members undergo complex interactions with each other. Such patterns of interactions are only poorly understood because no study lasting a sufficiently long time has included both insects and fungi.
Both facilitative and inhibitory interactions are likely to occur between species inhabiting dead trees. Interspecific competition occurs both among insects (e.g. Schroeder & Weslien 1994) and fungi (e.g. Boddy 2000). Insects may disperse fungi and so facilitate their colonization (Rayner & Boddy 1988;Boddy & Jones 2008;Persson, Ihrmark & Stenlid 2011), and fungal mycelia are an important food sources for many insect species (Wilding et al. 1989). Several studies indicate that the insect community in decaying wood largely is formed by the type of rot in the wood, which is highly dependent on fungus species (Saalas 1917;Palm 1951Palm , 1959Jonsell, Schroeder & Weslien 2005).
Insect and fungus species dependent on dead trees in boreal forest ecosystems form a large group among which are included many endangered species (Dahlberg & Stokland 2004;Tikkanen et al. 2006;Ga¨rdenfors 2010). It is therefore essential to better understand how habitats emerge and develop in dead trees to preserve biodiversity in general and to protect these threatened and endangered species in particular.
In this paper, we present results from a 15-year study, during which we followed the colonization and succession of wood-living insects and fungi. We test the long-term priority effects hypothesis that the identity and abundance of species that colonize first affect the colonization success of later-arriving species. We also hypothesize that species interact in both facilitative and inhibitory ways, which ultimately affect habitat quality for a particular species of wood-beetle currently listed as 'vulnerable' in a boreal forest ecosystem in Sweden.

T H E S Y S T E M
The study system includes five beetle and two fungus species living on Norway spruce, Picea abies L. Karst. Four of the beetle species, the scolytids Ips typographus (L.), Pityogenes chalcographus (L.) and Hylurgops palliatus (Gyll.) and the cerambycid Monochamus sutor (L.), are among the most common early colonizers on dying or newly dead spruce in Northern Europe. The arrival sequence and approximate time for flight initiation in central Sweden are as follows: (i) H. palliatus in mid-April, (ii) I. typographus and P. chalcographus in mid-May and (iii) M. sutor in mid-June (Eidmann & Klingstro¨m 1990). The fifth beetle species, Peltis grossa (L.) (Trogossitidae), is rare and arrives several years after the other species. It is known to breed in standing stems with brown-rotted wood of different tree species (Saalas 1917;Palm 1951). A positive association between the bracket fungi Fomitopsis pinicola (Sw.) P. Karst and P. grossa has been proposed previously (Ehnstro¨m & Axelsson 2002;Krasutskii 2007). Because of the lack of suitable breeding substrate in today's managed forests, P. grossa has become rare (Fjellberg & Hansen 1997;Ehnstro¨m 2001) and is now classified as 'vulnerable' in the Swedish Red List according to the IUCN criteria (Ga¨rdenfors 2010). The adult P. grossa leaves a characteristic oval emergence hole about 5 · 10 mm (Ehnstro¨m & Axelsson 2002).
The bracket fungi F. pinicola and Trichaptum abietinum (Dicks.: Fr.) Ryv. are very common species and among the first fungi to colonize dead tree trunks (Ka¨a¨rik & Rennerfelt 1957;Lindhe, Å senblad & Toresson 2004). Insect communities in wood rotted by T. abietinum and wood rotted by of F. pinicola differ in their assemblage (Jonsell, Schroeder & Weslien 2005). Fomitopsis pinicola causes brown-rot and T. abietinum white-rot, and the two species grow spatially separated (Renvall 1995;Abrahamsson, Lindbladh & Ro¨nnberg 2008). The fruiting bodies of F. pinicola are perennial and sporulate from early spring to autumn (Nuss 1986), whereas the fruiting bodies of T. abietinum are annual and start to sporulate later during the growing season. Both species may be dispersed by wind and by bark beetles (Pettey & Shaw 1986, Persson et al. 2009Persson, Ihrmark & Stenlid 2011). Fomitopsis pinicola can occasionally live as a parasite on living trees of decreased vitality (Forestry Compendium, 2005 and references therein). In the present study, only vigorous trees were felled (see below), and thus the possibility for them to be pre-infected with the fungus is very unlikely.

F I E L D S T U D Y
To study the sequence in which beetles and fungi colonize standing, coarse dead-wood, high stumps of Norway spruce were created in 1994 and 1995 near Granga¨rde in central Sweden (60°16¢00¢¢N; 14°59¢00¢¢E). Prior to conventional clear cutting of six managed spruce stands, 45-100 dominant, apparently healthy spruce trees in each stand were selected to be high-cut. Trees were cut 1AE3-5AE3 m (mean = 2AE8, SD = 0AE7) above ground, and the remaining stump was left rooted in the ground. Diameter at breast height (1AE3 m above ground level) ranged from 18 to 59 cm (mean = 34AE7, SD = 7AE2). The total number of high-cut stumps was 425, of which 363 were followed during the entire study period. Cutting time varied between sites and within two of the sites, giving three cohorts of stumps (Table 1). Distances between sites ranged from 3 km to 5 km except for two sites (D and E), which were only 0AE7 km apart. On all sites except site D, most stumps were exposed to the sun during the warmest part of summer days (south and west). At site D, stumps were situated along a shaded forest edge facing northeast. More detailed information is given in Schroeder et al. (1999).
Stumps in the 1994 cohort were inspected in autumn 1994 and thereafter stumps in all cohorts were inspected in autumn 1995,1997,1998 and each year from 2003 to 2009. The areas of bark with galleries of H. palliatus, P. chalcographus and I. typographus were recorded as percentages in the year of felling (1994 or 1995) and transformed to absolute areas of colonized bark. The three species together covered an average of 59% of the bark area. An average of 30% of the bark area was not colonized by any bark beetle species during the first summer. The remaining c. 10% of the total bark area was occupied by Orthotomicus spp (3 possible species), which were not included in the analysis as these bark beetles only could be identified to genus. The reason for this was that their galleries are not very species-characteristic and the method used for estimating area of bark colonized by bark beetles had to be non-destructive, i.e. the bark was not peeled off. A more detailed description of the method, abundance of different bark beetle species and relationships between bark beetle species are given in Schroeder et al. (1999). Emergence holes of M. sutor were recorded in 1997 (3rd or 4th year after felling). The number of emergence holes of P. grossa, the percentage of stump surface area covered by bark, the number of living and dead fruiting bodies of F. pinicola and the occurrence of T. abietinum were recorded at each inspection. Trichaptum abietinum occurrence was assigned to one of three classes based on the frequency of fruiting bodies: absent, 0; < 25% coverage of stump surface area, 1; ‡ 25% coverage, 2.
The main start of P. grossa emergence was in 2003. This conclusion is based on field observations in July and September 2003. On a sample of 222 stumps, we found nine emergence holes on six stumps in July. All emergence holes except one were judged to be made the current summer. In September 2003, the number of emergence holes on the same 222 stumps had increased to 81 on 13 stumps.

S T A T I S T I C A L A N A L Y S I S
A hypothetical model of causal species associations was tested ( Fig. 1). The strongest organizing factor in this model is the arrival time of species. For instance, because H. palliatus arrives first, all paths connected to it have no alternative other than leading from it; similarly because P. grossa arrives last, all paths connected to it must lead towards it. The model was tested in two steps: first, the effects of 10 quantitative and categorical predictor variables on the number of P. grossa emergence holes (the latest colonizer) were analysed in general linear mixed models, PROCMIXED (SAS-inst. 2003); second, the continuous variables that had strong effects (P < 0AE001) were included in a path analysis in which possible causal links between variables were analysed.
The response variable in the PROCMIXED analyses was 'ln (1 + cumulative number of P. grossa emergence holes per stump in year 15)'. The log-transformation was made to achieve normality and avoid over-dispersion in the data. Each of the predictor variables was tested separately and the model controlled for random effects of site (categorical). The predictor variables for each stump were: (i) ln (1 + cumulative number of living and dead F. pinicola fruiting bodies in year 10); (ii) frequency class of T. abietinum in year 10; (iii) ln (1 + cumulative number of M. sutor emergence holes in year 3 or 4); (iv) area of bark colonized by I. typographus in year 1; (v) area of bark colonized by P. chalcographus in year 1; (vi) area of bark colonized by H. palliatus in year 1; (vii) ln (stump diameter); (viii) ln (stump height); (xi) cutting year (categorical); and (x) cutting season (categorical). For the early colonizing beetles, there was no choice of which year after cutting to use in the analyses. Estimates of variables relating to the two fungi and P. grossa were drawn from the data pertaining to the year that was thought might give the best possible estimate, i.e. a year in which numbers had increased since the previous inspection but did not increase later (year 10 for T. abietinum and F. pinicola, year 15 for P. grossa, see Fig. 2).
The path analyses of causal covariance structures were performed in the CALIS procedure in SAS, which provides goodnessof-fit tests for the whole path model using iterative maximum likelihood methods. Apart from the classical Chi-square statistics, the CALIS procedure also provides a goodness-of-fit index (GFI), (Jo¨reskog & So¨rbom 1985), which should be between zero and one. If GFI is negative or larger than one, the data probably do not fit the model. It is not possible to include categorical variables in path analysis (Petraitis, Dunham & Niewiarowski 1996); therefore, because the PROCMIXED analysis controlled for site effects, a site was excluded from the path analysis if any of the species was absent at that site. This was the case for site D where no emergence holes of M. sutor were found. We tested the path analysis also with site D included and, it produced similar results, the only differences being that the associations between M. sutor and other species became somewhat weaker and that the unexplained variance for F. pinicola and P. grossa became somewhat higher as compared to when site D was omitted.

Results
The number of living fruiting bodies of F. pinicola peaked between years 4 and 10 after which the number of dead fruiting bodies increased (Fig. 2a). The proportion of stumps with F. pinicola present increased from 0 to 50% during the first 4 years after cutting and peaked at c. 80% between years 4 and 10. The pattern for T. abietinum was similar but with lower frequency (Fig. 2b). No fruiting bodies of either species appeared year 1. In year 2, four stumps were found with F. pinicola and two with T. abietinum (1994 cohort, 1995 assessment, data not shown). The total number of P. grossa emergence holes increased linearly from year 10 (Fig. 2a), with the mean number per stump in year 15 varying from 3AE1 to 6AE9 at all sites except site D, which had a mean of only 0AE6.
In year 15, the proportion of stumps with P. grossa emergence holes varied over the six sites from 23% (site D) to 57% (site F) with an overall average of 36%, and 16% of stumps with more than 10 emergence holes. Bark coverage decreased linearly over time to an average of less than 40% by year 10 (Fig. 2c).

S P E C I E S A S S O C I A T I O N S
The PROCMIXED analysis revealed F. pinicola, M. sutor, T. abietinum, H. palliatus, stump diameter and stump height to have strong effects on numbers of P. grossa ( Table 2). The effects of M. sutor and T. abietinum were negative; those of all others were positive. Overall, the inclusion of data on preceding species decreased the unexplained variation in P. grossa numbers from 96% (only stump diameter and height as predictors) to 62%. All four species listed had significant total effects on P. grossa in the path analysis (total effect > 0AE15, P < 0AE01, Table 3). Fomitopsis pinicola had only direct effects on P. grossa, M. sutor acted both directly on P. grossa and indirectly via F. pinicola, and T. abietinum and H. palliatus acted mainly indirectly on P. grossa via F. pinicola (Fig. 3, Table 3). Trichaptum abietinum, M. sutor, H. palliatus, stump diameter and stump height explained 25% of the variation in the number of F. pinicola fruting bodies. All three of these species had mainly direct effects on F. pinicola. Stump diameter and height explained 7% of the variation in T. abietinum frequency and 2% of the variation in bark area covered by H. palliatus (Fig. 3). Stump diameter and height explained 0% and 2% of the variation in M. sutor, F. pinicola numbers, respectively (data not shown).
As hypothesized, there was a negative correlation between number of emergence holes by M. sutor and the percentage of remaining bark in year 10 (r = )0AE35; P < 0AE001) and a positive correlation between the percentage of remaining bark and number of F. pinicola fruiting bodies in year 10 (r = 0AE42; P < 0AE001). The mean (±SD) number of emergence holes of P. grossa on stumps with and without M. sutor was 1AE8 ± 5AE6 and 6AE5 ± 11, respectively. Corresponding values for stumps with and without F. pinicola were 5AE6 ± 11 and 0AE1 ± 0AE5, respectively. These correlation coefficients and mean values are based on 318 stumps, i.e. site D excluded.

Discussion
The  Olito & Fukami 2009;Fukami et al. 2007Fukami et al. , 2010Chase 2003Chase , 2010. This is in agreement with our main finding that the first-year colonizers, H. palliatus and M. sutor, affected most of the later-arriving species, and that community assembly was dependent on which of the two species that gained early dominance. Which stumps that were colonized by H. palliatus or M. sutor seems to be highly random, as no stump variable could explain their abundance, but after the colonization of these early species, we can verify two pathways of succession from early to late successional stages: facilitation; H. palliatus fi P. grossa and inhibition; M. sutor fi P. grossa (Fig. 3). In the classic paper on conceptual models for ecological succession, Connell & Slatyer (1977) predicted that their 'facilitation model' would apply to heterotrophic succession. This has been supported by studies on wood (Renvall 1995) and dung (Slade et al. 2007). Our results indicate that also the 'inhibition model' may be applicable for succession in wood, which also has support in a recent experimental study with wooddecaying fungi (Fukami et al. 2010). Another long-term study (Chase 2010), on experimental ponds, also indicated that stochastic priority effects may have long-lasting impact on community structure. Both ours and Chase's study were field studies that spanned over many years, during which time species inventories were repeatedly made on the same objects. In neither study, would an analysis based only on the final state have produced the data necessary for the conclusions made. For instance, in the present study, it would not have been possible to accurately assess the timely arrival of different species based only on stump Table 3. Total, direct and indirect effects of the suggested pathways between species and the effect of stump height and diameter on each of the species Hylurgops palliatus, Monochamus sutor, Fomitopsis pinicola, Trichaptum abietinum and Peltis grossa (see also Fig. 3.)

Paths
Total effect Causal effects Fig. 3. Suggested pathways during 15 years of wood decay, from Hylurgops palliatus colonization to Peltis grossa emergence. Coefficients indicate direct effects; for total and indirect effects, see Table 3. Stump diameter and height were included as independent variables for each species but are not illustrated here. Regressions performed with ln-transformed values for stump diameter and stump height and with ln(x + 1)-transformed values for number of P. grossa and Monochamus sutor emergence holes per stump, and number of Fomitopsis pinicola fruiting bodies per stump. Untransformed values for Trichaptum abietinum frequency and area of bark colonized by H. palliatus. U = unexplained variance (1 ) R 2 ). *P < 0AE05, ***P < 0AE001. Goodness-of-fit index (GFI) 0AE98, v 2 = 23, P < 0AE0001.

Species interactions during wood decay 1159
inspection year 15, nor to determine the bark area covered by galleries of the different bark beetle species so long time after their colonization. An alternative set-up with a space-fortime substitution, i.e. a snapshot assessment of stumps of different age, could not have lead to any conclusions regarding priority effects as stumps would differ in colonization history. This demonstrates the advantage of long-term studies over snapshot or chronosequence studies when studying ecological succession in natural ecosystems, as recently highlighted by Johnson & Miyanishi (2008) and Walker et al. (2010). Experimental short-term (weeks to months) studies on community assembly in microcosms (e.g. Ejrnaes, Bruun & Graae 2006;Fukami et al. 2007Fukami et al. , 2010Olito & Fukami 2009) have advantages but also limitations compared to our correlative field study. One of their main advantages is that all species can be added in all treatments, and the priority effect of each species can be systematically tested. In our study, we can only assume that all species had the opportunity to colonize the stumps. The fact that all species colonized several stumps at all sites (except D, which was excluded) indicates that this assumption holds true. An advantage of our study is that all colonizations were natural with regards to order and timing of species arrivals. Fukami et al. (2010) showed convincingly in a controlled experiment on wood discs that species richness of a fungal community is highly dependent on which species that was introduced first. Although the authors discuss their results in a context of natural colonization, they conclude that it is unclear to which extent the order of species occurs in the natural system. Another advantage with controlled experiments over our study is that the environment in which all interactions take place is homogeneous, yielding low statistical variance and enabling powerful hypothesis testing. On the other hand, heterogeneity is a major feature of natural ecosystems and important to consider when estimating the strength of species interactions (Hunter & Price 1992). The goal of our study was to test the hypothesis of long-term priority effects in a natural ecosystem and obviously the strength of these effects were large enough to appear also in the presence of variance caused by variables beyond our control.
Our study tests a hypothesis of causal relationships using path analysis, which is based on correlations. But positive or negative correlations between two species may be driven by similar or dissimilar habitat requirements, indicating a possible lack of causality in the association (Ovaskainen, Hottola & Siitonen 2010). Our data on correlations between stump variables and species do not indicate that similar or dissimilar habitat requirements were a cause for significant correlation between two species. For instance, stump height had a positive effect on both F. pinicola and T. abietinum (Table 3), but the correlation between these two species was negative. Also, bark cover (not included in the analysis) was positively correlated with both these species. Peltis grossa was positively correlated with diameter, but the four preceding species, all significantly correlated with P. grossa, were not correlated with diameter. Correlations should nevertheless be inter-preted with caution and therefore it is important that suggested relationships in a path analysis reflect rational causality, and that these hypothetical causal relationships be established before analysing data (Mitchell 1993;Petraitis, Dunham & Niewiarowski 1996). Both of these requirements were met in the present study.
The two fungus species T. abietinum and F. pinicola only occasionally occurred on the same stump but were then always spatially separated. These results are largely in accordance with those from another field study on high stumps (Abrahamsson, Lindbladh & Ro¨nnberg 2008). Ovaskainen, Hottola & Siitonen (2010) found that T. abietinum and F. pinicola often occurred together on the same logs, and they hypothesized that this was either due to higher competitive ability of T. abietinum, which made it possible to colonize logs already occupied by F. pinicola, or due to niche separation between the two species. We could not see any signs of niche separation in the present study, as either species could be found anywhere on a stump and the abundance of their fruiting bodies was negatively correlated. Instead, our results and observations suggest that competition occurred between T. abietinum and F. pinicola. In a laboratory study, Holmer, Renvall & Stenlid (1997) found that T. abietinum was the superior competitor of the two fungus species when introduced simultaneously and even so in combinations where T. abietinum was introduced in lower dose than F. pinicola. Thus, when colonizing dead wood objects simultaneously, T. abietinum is likely to reduce the potential substrate available to F. pinicola.
The final model differed from the hypothetical model in several aspects. First, the two bark beetle species I. typographus and P. chalcographus had no effect on P. grossa numbers. Second, the hypothesis that H. palliatus had an indirect effect on F. pinicola via M. sutor was rejected. Instead, almost all of its effect was direct. Third, there was an unpredicted direct effect of M. sutor on P. grossa. The direct positive association between H. palliatus and F. pinicola might be explained if H. palliatus carried F. pinicola spores or mycelia. Several bark beetle species that fly during F. pinicola sporulation have been shown to become contaminated with spores to a high degree during their flight (Pettey & Shaw 1986). Also, colonization by wind borne spores may have been facilitated by beetle holes in the bark. The reason why I. typographus and P. chalcographus did not facilitate F. pinicola colonization may be a result of differences in the flight periods of the bark beetles and in the sporulation periods of the fungi. Local meteorological data indicated that the interval between the main flights of H. palliatus with a flight threshold of 15°C, and of I. typographus and P. chalcographus, each with a flight threshold of 20°C, would have exceeded 6 weeks in both 1994 and 1995. In Scandinavia, F. pinicola sporulation starts in early spring, just after snow-melt (Ha˚gvar 1999), which coincides with the main flight of H. palliatus. In contrast, T. abietinum sporulates later in the summer as a result of having annual fruiting bodies. Thus, it may be that during the colonization of the stumps by I. typographus and P. chalcographus, F. pinicola had to compete with the superior competitor T. abietinum. The unpredicted direct effect of M. sutor on P. grossa is not easily explained but could be due to M. sutor larvae feeding deep within the sapwood and reducing habitat quality there for P. grossa.
Studies on beetle communities in dead trees have a long tradition, especially in Fennoscandia. The works by Saalas (1917) and Palm (1951Palm ( , 1959, in which detailed descriptions of beetle communities in various stages of wood decay are provided, deserve special mentioning. However, earlier studies on community ecology of wood-living beetles have not been strongly linked to succession or assembly theories. As such, this study is a novel contribution. Our study also exemplifies new options for species-oriented conservation measures. In the case of H. palliatus, M. sutor and P. grossa, cutting of stumps during fall and winter (after the flight of M. sutor and before the flight of H. palliatus) would be a good option to increase the probability for subsequent colonization by P. grossa, whereas cutting during spring and summer (after the flight of H. palliatus and before the flight of M. sutor) would have the opposite effect. Thus, exploring links between early, well-understood species guilds and later, more poorly understood species guilds may offer new possibilities for managing populations of threatened species.