Epiphytic bryophytes near forest edges and on retention trees: reduced growth and reproduction especially in old-growth-forest indicator species

Authors


Correspondence author. E-mail: swantje.lobel@ebc.uu.se

Summary

  1. Epiphytes are an important component in many forest ecosystems. The proportion of threatened epiphyte species is high, and the impact of clearcuts on key demographic processes via edge-influence is still poorly understood. There are few studies on epiphyte growth, and even less is known about how reproduction is affected by proximity to forest edges. For retention trees, demographic studies are even scarcer.
  2. Based on the results from a 6-month transplant experiment and a 3-year study of natural colonies, we modelled growth and reproduction of epiphytic bryophytes used as indicators of old-growth forests and widespread epiphytes in relation to distance from the forest edge. We also modelled growth and reproduction on retention trees within the clearcut. Species responses were linked to variation in canopy openness.
  3. Unlike the widespread species, the old-growth-forest indicators grew exponentially with distance from the edge, and this response was more pronounced at the south-east than north-west forest edge. In one red-listed species, reproduction was thoroughly inhibited near the edge, whereas the reproductive rate of the widespread species tended to increase. However, the widespread species also showed reduced shoot lengths on the retention trees.
  4. Reduced growth and inhibited reproduction of sensitive epiphytes near edges decrease the number of dispersing diaspores and may, in combination with lower local connectivity and increased tree fall rates close to edges, increase the risk of metapopulation extinction.
  5. Synthesis and applications. Two general management implications for boreal forests are drawn. First, retention trees may not have the capacity to act as a ‘lifeboat’ for epiphytic bryophytes and support their populations during the regeneration phase. Second, the creation of buffer zones is a useful conservation strategy for bryophytes. The exact width of zones depends on the forest structure and should be orientated in relation to the requirements of the most sensitive species. For the rather dense experimental forest, a width of at least 30 m was required for the south-facing buffer, whereas for the north-facing buffer 10 m was sufficient.

Introduction

Forest ecosystems worldwide have been subjected to intensive exploitation, and in current landscapes many forest species are restricted to remnants of a former, mostly contiguous forest cover. A large proportion of the remnant forest fragments experience edge-influence (Harper et al. 2005), that is, physical or biological processes originating from the human-induced boundary between the forest and the clearcut. For many sessile forest species, such edge-influence may be even more detrimental than the effects of habitat loss and isolation (Moen & Jonsson 2003). Gradients occur in some microclimatic variables, for example, light and humidity, from the clearcut, via the forest edge to the interior (Chen, Franklin & Spies 1995). Some species are favoured by the increased availability of light, but others are negatively affected by desiccation (Murcia 1995). The incorporation of buffer zones around protected areas attempts to ensure long-term species persistence, but the distance of edge-influence has been rarely quantified (Roberge et al. 2011), and there is also an edge-orientation effect (e.g. Hylander 2005). Another conservation strategy applied worldwide is the retention of some living trees within the clearcut to act as a ‘lifeboat’ for species over the regeneration phase and to increase connectivity in the managed landscape. However, for many forest species it is very uncertain whether solitary retention trees or grouped trees (retention patches henceforth) actually have the capacity to do so (Rosenvald & Lõhmus 2008; Perhans et al. 2009).

Epiphytes are an important component in many forest ecosystems: over 10% of all terrestrial plant species grow epiphytically, typically on trees (e.g. Burns & Zotz 2010). In the temporal and boreal regions, bryophytes and lichens are the main epiphytes (Barkman 1958). Since they lack water-regulating mechanisms and obtain most of their water from rain and atmospheric humidity, they are especially prone to altered microclimates (Barkman 1958). Here, we focus on bryophytes because most studies on retention trees have used lichens (Gustafsson, Kouki & Sverdrup-Thygeson 2010). In Europe, only 110 out of 1719 bryophyte species grow epiphytically (6·4%, Dierßen 2001; M. Breuer, personal communication), but 25% of them are red-listed (32 species, ECCB 1995). Several epiphytic bryophytes are used as indicators of forest continuity and high probability of occurrence of red-listed species (Nitare 2005).

Altered microclimates on retention trees in clearcuts or at forest edges affect epiphytes, but rather little is known about the underlying ecological mechanisms, for example, the alteration of recruitment, growth, reproduction or mortality (Murcia 1995). The few studies on epiphytic bryophytes and lichens have focused on growth (Renhorn et al. 1997; Roberge et al. 2011) or vitality (Hazell & Gustafsson 1999), and more seldom on colonization (Hilmo, Holien & Hytteborn 2005; Roberge et al. 2011) or reproduction (Jairus, Lõhmus & Lõhmus 2009). In the northern hemisphere, epiphyte vitality and abundance is often higher on the north than on the south sides of retention trees in clearcuts (Hazell & Gustafsson 1999; Hedenås & Hedström 2007). Roberge et al. (2011) observed for one red-listed epiphytic bryophyte that the abundance of local populations decreased closer to the forest edge and that deterministic local extinctions due to tree falls increased. So far, there are no studies on bryophyte reproduction in relation to proximity to the forest edge. Neither it is known whether widespread epiphytic bryophytes are also influenced by forest edges.

The central aim of our study was to shed light on the mechanisms behind forest edge-influence on epiphytic bryophytes. We quantified the magnitude and distance of edge-influence on growth and reproduction, and investigated whether these processes are altered on solitary retention trees. We combined an experimental approach with transplantations and monitoring of colony growth and reproduction over 3 years. Epiphyte growth and reproduction were modelled in relation to distance from the south-east and north-west facing forest edges and to light conditions. We tested whether the edge-influence differed among tree-trunk orientations. We hypothesized that the old-growth-forest indicator species are more sensitive to altered microclimates at forest edges and on clearcuts than widespread species.

Materials and methods

Study species

We studied six obligate epiphytic bryophytes confined to broad-leaved deciduous trees (Table S1, Supporting information). The mosses Neckera pennata Hedw. and Orthotrichum speciosum Nees were included both in the transplantation experiment and in the study of colony growth and reproduction. The mosses Homalia trichomanoides Hedw. B.S.G and Isothecium alopecuroides Dubois Isov. could only be used for the transplantation experiment, and the liverworts Frullania dilatata L. Dum. and Radula complanata L. Dum. only for the study of colony growth.

The studied species vary in frequency and local abundance within the study forest (Table S1, Supporting information). Neckera pennata is red-listed (‘Near Threatened’, Gärdenfors 2010) and, together with H. trichomanoides, used as an old-growth-forest indicator (Nitare 2005). The other species are widespread within the region, that is, they occur on broad-leaved deciduous trees in more open forests, in parks or along avenues.

Study site

The studies were conducted in a forest stand in the boreo-nemoral region in Eastern Sweden (2·0 ha, Fig. 1). We previously studied the distribution of the study species within 135 forest stands having broad-leaved deciduous trees in the region (Löbel, Snäll & Rydin 2006). Most stands had to be excluded as they did not border a clearcut or because N. pennata occurred only as isolated individuals. The selected stand was the only site that allowed us to study epiphytic performance (i) on retention trees, (ii) along a gradient into the stand, (iii) on different sides of the trunk, (iv) in different edge orientations and (v) for different categories of species.

Figure 1.

The study forest and adjacent clearcuts with the Fraxinus excelsior trees used for the transplantation experiment, sample trees used for the study of colony growth and reproduction, and all other F. excelsior trees.

The north-west and south-east forest borders are formed by a large clearcut (c. 30 ha and 300–500 m wide) since 1987. Today, the clearcut is mostly covered by shrubs, tall herbs as well as young birches and aspens 1–2 m in height. The forest lies in a slight, level depression. The soil is mesic to moist; the ground vegetation indicates nutrient richness.

The stand consists of Fraxinus excelsior (basal area of trees per ha: mean = 11·2, SD = 3·9), Alnus glutinosa (12·3, 4·8) and Picea abies (11·3, 3·9). The basal area was recorded at the intersections of a randomly placed regular grid with 20-m divisions and 45 plots in total. Fraxinus excelsior is the only suitable host tree for the study species, and we recorded a total of 492 trees, with a mean diameter at breast height (1·3 m, ‘DBH’) of 24·3 cm (SD = 8·9 cm). The stand age is about 80 years. We mapped all potential host trees using the established, regular grid system, measurement tapes and sighting compass. The forest edge was defined as the line formed by the outermost standing trees. Host tree positions and forest edges were digitized in a GIS (Fig. 1).

Transplantation experiment

For the transplantation experiment, we chose eight solitary F. excelsior trees on the clearcut, 16 trees at the forest edge (0–5 m), 16 at 5–10 m distance, 16 at 10–20 m distance and 16 at >20 m distance from the forest edge (totalling 72 trees). Trees were selected with the criteria that they should be evenly distributed over the stand, not leaning in any direction and not in contact with spruce branches. The trees were equally distributed along the south-east (SE) and north-west (NW) forest edges (Fig. 1). The mean DBH of the experimental trees was 28·9 cm (SD = 6·9 cm). The distance to the nearest forest edge was calculated for each sample tree using the GIS.

We collected vital colonies (c. 8 cm in diameter) of the study mosses attached to bark from the interior of the study forest and an adjacent stand. Colonies were sometimes extracted as parts of larger colonies. One colony of each species was glued onto a wooden plate. We marked five shoots on each colony with small threads of different colours and measured their length under a dissecting microscope. During 4–6 June 2005, two wooden plates were placed on each experimental tree at 1·3 m height – one facing NW, and one SE. This height was chosen because all species occurred at this height within the study forest, although H. trichomanoides typically has a lower main growth position and O. speciosum is typically higher on the tree trunks. A further concern was to place the plates above the field layer. After 6 months (December 7), we collected the plates and re-measured the length of the shoots in the laboratory. We recorded colony vitality on a 4-point scale: (1) whole colony alive: green and vital, (2) most parts alive: green with some brownish or bleached parts, (3) most parts dead: large brownish or bleached patches, (4) fully dead: completely brown or bleached. The monthly mean temperature for July through October 2005 was 1–2·7 °C higher than the 1961–1990 average, and the precipitation was higher than the long-term average for June to August but lower in September and October (SMHI 2012, weather station Uppsala).

We recorded the DBH of each experimental tree. Light conditions were assessed by measuring canopy openness (%) using digital hemispherical photography (see Appendix S1 in Supporting Information). We estimated the cover (%) of the field layer and shrubs within a 2 m radius around each tree. The NW and SE forest edges were similar with respect to the measured environmental variables. Tree diameter did not vary among trees at different distances from the forest edge.

Colony growth and reproduction

We restricted our study of colony growth and reproduction to the SE forest edge, where we expected the largest variation. We chose 20–30 colonies of each species growing on host trees at the following distances from the edge: 0–5, 5–10, 10–20 and >20 m (Fig. 1). The total number of colonies was 91 in N. pennata, 98 in F. dilatata, 99 in R. complanata and 94 in O. speciosum. The distance to the SE forest edge was calculated for each sample tree using GIS. Growth position was 30–225 cm above ground. We chose colonies that were smaller than 100 cm2 (larger colonies often fuse), healthy and separated from other bryophyte colonies by at least 10 cm.

The colonies were marked and their peripheries drawn on transparent plastic sheets in August 2005 and 2008. At both occasions, we noted the occurrence of sporophytes for all colonies. All colonies that had sporophytes in 2005 also had sporophytes in 2008, except two colonies of O. speciosum. Thirteen colonies of N. pennata, 19 of F. dilatata, 23 of R. complanata and 22 of O. speciosum lacked sporophytes in 2005, but had sporophytes in 2008. When analysing whether reproduction is suppressed closer to the forest edge, we used the more conservative measure of occurrence of sporophytes in 2008. We digitized the colony drawings using a scanner (300 dpi). Colony area was calculated by counting pixels inside the perimeter (using Image, Karlsson 2007).

For each colony, we recorded height above ground (cm), orientation, DBH, depth of bark crevices and stem inclination. For orientation, we used the contrasts between ‘SE-facing’ (S, SE, SW, E) and ‘NW-facing’ (N, NW, NE, W). Since rainwater percolating through coniferous tree branches might alter host tree bark chemistry, we noted if a branch from a spruce touched the host tree. We estimated the cover (%) of the field layer and shrubs in circles of 2 m radius around each tree. We measured canopy openness (%) using digital hemispherical photography at the intersections of a randomly placed regular grid with divisions of 10 m (Appendix S1, Supporting information). Canopy openness was interpolated to the whole stand by kriging based on a geostatistical semi-variance model (Appendix S2, Supporting information).

Statistical analyses

Transplantation experiment

The mean of the five measurements of shoot length from each colony was used in the statistical analyses. There was no bias in initial shoot length at different distances from the edge. Length increase (mm) was not related to initial shoot size in any of the species.

We tested whether shoot growth differed among retention trees, trees at the forest edge (0–5 m distance) and trees in the forest interior (>20 m) using linear mixed-effect models with ‘sample tree’ as the random factor, and ‘distance class’ and ‘plate orientation’ as fixed factors. Separate analyses were carried out for the SE and NW edges. Model selection was based on Akaike Information Criterion (AIC). We report the difference in AIC (∆AIC) between the final model and the null model (including the random factor only). For ∆AIC ≤ −2, the level of empirical support for the full model is higher than for the null model (Burnham & Anderson 2002).

We also specifically investigated how shoot growth changed with distance from the forest edge using nonlinear regression models. Models were fitted for each species separately. We applied the two-parameter exponential model

display math(eqn 1)

where SG is the shoot growth at distance d (m) from the forest edge, SG0 is the shoot growth at the forest edge, α is the asymptote, that is, the maximum shoot growth, and γ scales the increase in shoot growth with distance from the edge. We also fitted a null model containing the intercept only, which was used to calculate ∆AIC and an approximate R² value. The 95% confidence intervals were based on nonparametric bootstrapping (Ritz & Streibig 2008). Data were resampled 1000 times. As suggested by Hylander (2005), we calculated the distance from the forest edge where 90% of the maximum growth was reached as an estimate for the distance of edge-influence. Models were fitted for the NW and SE edges separately, and data were grouped by plate orientations (SE, NW). Neither DBH nor cover of ground vegetation was correlated with shoot growth, and we therefore used only ‘distance from forest edge’ and ‘canopy openness’ as explanatory variables.

To explore the importance of increased canopy openness in explaining the observed distance effects, we analysed how shoot growth changed with canopy openness. This meant replacing distance (d) in Equation 1 with canopy openness. In this case, models were fitted for trees from both forest edges and the clearcut together, and the data were grouped by the plate orientation. In I. alopecuroides and O. speciosum, the relationship was linear rather than nonlinear, and a linear model was applied.

To test whether the biological edge-influence covered longer distances than the light effect, we analysed how canopy openness decreased with distance from the forest edge. In the exponential model (Equation 1), SG was replaced by canopy openness.

We tested for spatial autocorrelation in residuals of the regression models (Appendix S3, Supporting information), reporting only cases where residuals showed spatial autocorrelation.

Colony growth and reproduction

We used the relative rather than the absolute colony growth rate (mm² year−1) as the dependent variable in our analyses, because the colony growth rate was related to initial colony area in 2005, except in O. speciosum (Fig. S1, Supporting information). There was no bias in initial colony area at different distances from the edge. The relative growth rate was calculated for each colony as

display math(eqn 2)

A0 is the initial colony area (2005), At is the colony area at time t (2008) and ∆t = 3 years.

We analysed how relative growth rate (RGR) and canopy openness changed with distance from the edge. We again applied the three-parameter exponential model (Equation 1); SG was replaced by RGR and canopy openness, respectively.

To test the effects of canopy openness, orientation (SE, NW) and the measured environmental variables on RGR, we fitted multiple linear regression models. The relationship between RGR and canopy openness was better described by a linear than nonlinear model. We first tested the individual effects of single variables and selected those which did not result in an increase in AIC. Multiple starting models were built using these variables. We included biologically plausible two-way interactions and squared terms. Starting models were simplified using stepwise variable selection minimizing AIC.

We used logistic regression models to analyse how the probability of occurrence of sporophytes in 2008 changed with distance from the edge. Colony area in 2008 was included as a covariate because, for some species, there is a threshold in colony area before they start to reproduce (Löbel & Rydin 2009). We similarly tested the effects of canopy openness and the measured environmental habitat factors on sporophyte occurrence. The multiple model building strategy was as described above. We tested the residuals of the different regression models for spatial autocorrelation (Appendix S3, Supporting information).

We used the software r 2.12.0 (R Development Core Team 2010).

Results

Transplant experiment: shoot growth and vitality

During the 6 months, shoot lengths decreased on retention trees, and this response was most pronounced in the old-growth-forest indicators N. pennata and H. trichomanoides (Fig. 2). Shoot growth was lower at the SE than NW edge and was lower on the SE-facing than NW-facing plates in both the clearcut and at the SE edge (Table S2, Supporting information).

Figure 2.

Mean, 6-month shoot-growth length in colonies transplanted to trees in the clearcut, at the forest edge (0–5 m distance) and in the forest interior (>20 m). Error bars indicate ±SE; sample sizes are also given.

Vitality of all species was low on retention trees, and especially low on SE-facing plates (Fig. S2, Supporting information). The percentage of heavily damaged colonies was already lower at the SE edge, and even lower at the NW edge. In the forest interior, between 80 and 100% of the colonies were fully or mostly alive, except in H. trichomanoides.

The distance of edge-influence was longest for the old-growth-forest indicators. In N. pennata, shoot growth increased exponentially with distance from the edge (Fig. 3a,b, Table S3, Supporting information). The distance of edge-influence was longer at the SE edge (23 m for SE-facing and 17 m for NW-facing plates) than at the NW edge (SE-facing = 8 m, NW-facing = 12 m). For H. trichomanoides and I. alopecuroides, there was empirical support for the exponential model only at the SE edge (Fig. 3c–f, Table S3, Supporting information). In H. trichomanoides, the distance of edge-influence reached 10 m, independent of plate orientation, but in I. alopecuroides it reached 8 m for SE-facing and 4 m for NW-facing plates. Growth of O. speciosum was affected by neither distance nor plate orientation (Fig. 3g,h).

Figure 3.

Six-month shoot-growth length of colonies transplanted to trees in relation to distance from the SE (left) and NW (right) forest edges as described by the three-parameter exponential model. The vertical arrows indicate the distance from the edge where the estimated maximum growth is reduced by 10%. Data were grouped by plate orientations. Models with ∆AIC ≤ 0 are specified in Table S3 (Supporting information).

Shoot growth in all species decreased with increasing canopy openness (Fig. S3, Supporting information). The relationship was nonlinear in N. pennata (∆AIC = −85·7, R² = 0·69) and H. trichomanoides (∆AIC = −42·9, R² = 0·49), but linear in I. alopecuroides (∆AIC = −29·4, R² = 0·57) and O. speciosum (∆AIC = −37·5, R² = 0·46). Canopy openness differed among trees on the clearcut (mean = 46·7%, SE = 1·07), at the edge (mean = 17·6%, SE = 1·02) and in the interior (mean = 10·8%, SE = 0·34). Canopy openness declined rapidly with distance from the edge; at 3·5 m, 90% of the maximum decline was reached (Fig. S4a, Supporting information).

Colony growth

Relative growth rate of N. pennata and F. dilatata decreased in proximity to the forest edge. The distance of edge-influence was much longer in N. pennata (29 m, Fig. 4a, Table S4, Supporting information) than in F.dilatata (3 m, Fig. 4b, Table S4, Supporting information). In R. complanata and O. speciosum, RGR was not related to the distance from the edge (Fig. 4c,d).

Figure 4.

Relative growth rate (RGR) (August 2005 and 2008) of colonies in relation to distance from the SE forest edge as described by the three-parameter exponential model. The dotted vertical arrows indicate the distance from the edge where the estimated maximum growth is reduced by 10%. Models with ∆AIC ≤ 0 are specified in Table S4 (Supporting information).

Canopy openness explained less of the variation in RGR of N. pennata (∆AIC = −14·1, R² = 0·16) and F. dilatata (∆AIC = −1·8, R² = 0·04) than the distance from the edge, and the relationships were linear (Fig. S5, Supporting information). Canopy openness was not related to RGR in R. complanata (∆AIC = 0·8), whereas a slight quadratic relationship was observed in O. speciosum (∆AIC = −1·5, R² = 0·06). Neither orientation nor any of the other measured habitat variables had a significant effect on RGR of the study species. At the sample trees, 90% of the maximum decline in canopy openness was reached at a distance of 11·4 m from the edge (Fig. S4b, Supporting information).

Colony reproduction

The proportion of fertile colonies of N. pennata was zero at the edge compared with 56% in the interior (Fig. 5), and the logistic model showed positive effects of distance from the edge on the presence of sporophytes (Table S5, Supporting information). The first fertile colony occurred at a distance of 14 m from the edge. In F. dilatata, reproduction was reduced only within 5 m from the edge (19% vs. 40% at 5–10 m distance, Fig. 5). The pattern was the same, but less pronounced in R. complanata and O. speciosum. However, when accounting for colony area, we did not detect any (F. dilatata), or negative (R. complanata, O. speciosum) effects of distance (Table S5, Supporting information). Colony area had a positive effect on sporophyte production in all species.

Figure 5.

Proportion of the colonies that were fertile (%) in 2008 at different distances from the SE forest edge. Sample sizes are given.

In N. pennata, canopy openness had a negative effect, and the depth of bark crevices had a positive effect on sporophyte production (∆AIC = −31·6, R² = 0·43; colony area included as covariate). Sporophyte production of the other species was not related to any of the measured habitat variables.

Discussion

Our results indicate that growth and reproduction of the old-growth-forest indicator species are substantially lower at forest edges than in the forest interior, whereas the widespread species remain largely unaffected. In sensitive species, reproduction can be totally inhibited under edge conditions. Widespread species also seem to be negatively affected by the harsh microclimate on retention trees.

Shoot growth and vitality on retention trees

Our study is the first to show that shoot growth of both old-growth-forest indicators and widespread species can be negatively affected on solitary retention trees. This, together with the low vitality of all transplants on retention trees, suggests that epiphytic bryophytes do not survive the time between logging and forest regeneration, a process that takes many decades; almost 20 years after harvesting, the clearcut in our study remained open. Reduced shoot lengths and high mortality of bryophytes also provide possible mechanistic explanations for the reported decrease of red-listed and old-growth-forest indicator species in retention patches (0·01–0·5 ha, Perhans et al. 2009). The only other experimental study on the performance of an epiphytic bryophyte on retention trees reported higher survival (Hazell & Gustafsson 1999), probably because the study species grows in more open forests. In contrast, three of our study species are found mainly in shady, moist forests (Nitare 2005). Lõhmus, Rosenvald & Lõhmus (2006) observed reduced bryophyte abundance and vitality on retention trees during the first 2 years after clearcutting, although species richness stabilized after 5 years (Lõhmus & Lõhmus 2010). However, the focus of that study was at the community level and mostly included common species, among them many generalists. In addition, the clearcuts used in that study averaged 2·8 ha (range 0·4–6·6 ha), which is distinctly smaller than here. Low survival abilities on retention trees have also been observed for tropical vascular epiphytes (Werner 2011). Lichens, in contrast, showed varying responses: some declined, whereas others appeared to persist or even increase following clearcutting (e.g. Hedenås & Hedström 2007; Perhans et al. 2009). Jairus, Lõhmus & Lõhmus (2009) have suggested that lichens with a thick upper cortex, in particular, have a good ability to adapt to high irradiation.

Transplants were placed on the retention trees in early June, and shoot growth might have been higher had they already been transferred in spring. However, that particular spring was extremely dry with <5 mm precipitation in April, whereas June to August was very wet (SMHI 2012). The growing season lasted until December 2005, with a very mild November. We are convinced, therefore, that the starting date of the experiment did not have a major impact on the outcome. If anything, we may have underestimated the negative effects of sun exposure.

Edge-influence on shoot and colony growth

Our study indicates that old-growth-forest indicators and widespread epiphytic bryophytes differ in their demographic responses to altered microclimates near forest edges. Although widespread species were severely damaged in the clearcut, it was only in the old-growth-forest indicators where growth differed among trees at the forest edge and in the interior. In the red-listed N. pennata, the distance of edge-influence on growth was 23–29 m, whereas in H. trichomanoides it was just 10 m. This could explain why the number of red-listed species showed the strongest decline after logging in 10-m-wide riparian buffer stripes (Hylander et al. 2005). In two ground-living bryophytes, the distance of edge-influence was longer in the species specialized to moist forests than in the widespread species (Hylander 2005).

The distance of edge-influence was ≤30 m, similar to the earlier reported distance for N. pennata (Roberge et al. 2011); closer than 30 m from the edge, the local abundance on host trees decreased or increased only slightly, whereas all larger increases were observed at distances >35 m. The distance of edge-influence was also within the same range as observed by Hylander (2005) for ground-living bryophytes, and within the commonly assumed 50-m penetration zone of edge-influence in plants (Murcia 1995; Ries et al. 2004). In the sensitive species, the biological edge-influence penetrated distinctly further than the light effect, most probably due to wind effects that usually penetrate much deeper into forest-clearcut edges compared with light effects (Hylander 2005). Roberge et al. (2011) found that tree fall rates can be accelerated up to 70–80 m into the forest. This may apply to open stands with south-west facing edges, but our study forest seemed unaffected: no trees fell during the study period, nor we could not detect any recently fallen trees.

The exponential increase in growth with distance from the edge can be explained by the nonlinear decline of irradiation and wind speed from the forest edge to the interior (Zheng & Chen 2000). Bryophytes are poikilohydric organisms and water loss rises steeply with increasing wind speed and irradiation, reducing the duration of wet conditions and thus the period of growth (Proctor 1982). The fact that our widespread species did not show any net growth and a low vitality on retention trees suggests that there is an irradiation threshold above which the bryophyte is severely damaged by desiccation. Shrinking and breaking of dead or desiccated shoots are a probable cause of the reduced shoot length.

This northern-hemisphere study showed that both the distance and magnitude of edge-influence are larger at the SE than NW edge. In contrast, Hylander (2005) observed an impact in magnitude only. We further showed that for epiphytes orientation on the tree also matters; even close to the SE edge, there can be a rather sheltered microclimate at the north-facing side of a tree trunk (cf. Hazell & Gustafsson 1999).

The differential sensitivity between the old-growth-forest indicator and widespread species is linked to growth form (Mägdefrau 1982). Neckera pennata and H. trichomanoides can be classified as fans: as they grow in height, shoots may dry out fast. In contrast, the liverworts R. complanata and F. dilatata grow tightly attached to the bark surface making them less prone to desiccation. This finding contrasts with earlier conclusions concerning a proposed higher sensitivity of liverworts than mosses to forestry operations (Hylander et al. 2005; Perhans et al. 2009).

The chosen height might have influenced the results of the transplant experiment. Homalia trichomanoides preferably grows at the tree base, and this may partly explain the overall low vitality of transplants in this species. Placing the transplants at a lower height might have decreased the depth of the edge-influence. However, no colonies of this species were detected closer than 10 m from the edge.

Edge-influence on reproduction

Our study showed that reproduction in the old-growth-forest indicator N. pennata can be totally inhibited at the forest edge. In contrast, the reproductive rate of the widespread species was even slightly enhanced. Roberge et al. (2011) did not find a lower colonization rate of N. pennata near the edge, but Hilmo, Holien & Hytteborn (2005) did for some foliose epiphytic lichens. Reduced colonization rates close to the edge, however, can be caused by lower connectivity to dispersal sources and/or lower habitat quality near edges. Our study provides the first evidence that altered microclimates near edges can directly inhibit reproduction: N. pennata is an autoecious species and reproduction can occur via self-fertilization.

No colonies ceased to reproduce during the three study years. This is not surprising given that the clearcut was already 20 years old. Instead, a considerable number of colonies lacking sporophytes in 2005 reproduced in 2008. Most likely, these colonies had now reached their reproductive size and age (Löbel & Rydin 2009). It has been suggested that the age at first reproduction increases in the order O. speciosum < R. complanata < F. dilatata < N. pennata (Löbel & Rydin 2009). We can now add that the sensitivity to edge effects increases in this order as well. Reduced growth and inhibited reproduction at the edge reduce the number of dispersing diaspores and can then have negative synergistic effects in combination with lower local connectivity and increased tree fall rates close to forest edges (Roberge et al. 2011). In addition, low humidity may reduce establishment probabilities (Wiklund & Rydin 2004). Hence, the paths by which epiphytic bryophytes are affected by forest edges are complex and their co-action may increase the risk of metapopulation extinction.

Implications for management

We conclude with two general management implications for boreal forests. First, retention trees are unlikely to have the capacity to act as a ‘lifeboat’ for epiphytic bryophytes over the regeneration phase (Rosenvald & Lõhmus 2008; Perhans et al. 2009). Second, the creation of buffer zones is a useful conservation strategy for bryophytes. The positive effects of retention trees increase, and the distance of edge-influence decreases, from red-listed, old-growth-forest indicators to widespread epiphytes, that is, focusing on the requirements of the most sensitive epiphytes should give robust guidelines for nature conservation.

Epiphyte performance on the retention trees was poor, particularly in red-listed and old-growth-forest species. This has wide implications: today, tree retention at clearcutting is included in the national legislation of many countries (e.g. in Northern Europe, Northern America and Australia, Gustafsson, Kouki & Sverdrup-Thygeson 2010), imposing substantial costs for forest owners. The required retention levels vary, and so does the spatial configuration of retention trees. Limited resources require choosing the most efficient conservation strategies. At least for bryophytes, the benefits of solitary retention trees seem to be small. Leaving retention trees in larger groups or close to northern edges may be more beneficial (Rudolphi & Gustafsson 2011).

The minimum required width of the buffer zone depends on the structure of the forest. For our study forest, a width of at least 30 m is required for the south-facing buffer, whereas for the north-facing buffer 10 m can be sufficient. Many forests are less dense, and a 50-m buffer has been recommended (e.g. Murcia 1995; Ries et al. 2004; Hylander 2005; Roberge et al. 2011). However, given the exponential nature of edge-influence, even narrower buffers may, if not eliminate, at least markedly decrease the edge-influence. Buffers may be established by leaving an uncut zone between the focal forest and the clearcut. Moderate thinning in a wider buffer strip is an alternative. This could not only diminish the negative impacts of clearcutting on the focal forest, but may also achieve a positive edge-influence on the clearcut (Caruso, Rudolphi & Rydin 2011) by increasing the probability of establishment of old-growth-forest species in the new developing stands.

Acknowledgements

The study was financially supported by a grant from Bjurzon's fund to S.L., a grant from the Swedish research council Formas to T.S. (2005-933) and grants from the councils VR and Formas to H.R. We thank Scott Spellerberg for revising the English.

Ancillary