Edge effects of wildfire and riparian buffers along boreal forest streams

Authors


Correspondence author. E-mail: amallik@lakeheadu.ca

Summary

1. Clearcutting and wildfire are the two major edge-creating disturbances in boreal forests. While clearcutting retains at least a 30 m buffer, wildfires burn close to streams killing most of the trees and potentially creating a different edge structure. Different edge structures are likely to support different plant assemblages. The riparian buffer and fire edge structures and their effects on plant assemblages are unknown, but they have implications for forest management that aims to harvest trees and conserve biodiversity by emulating natural disturbance.

2. We hypothesized that environmental filters created by post-fire residual structures at the fire edge will support a different plant assemblage than the buffer edge. We further hypothesized that the fire edge with a weaker environmental filter because of post-fire residual structures and proximity to streams will result in lower depth of edge effects (DEEs) and magnitude of edge effects (MEEs) than buffer edges.

3. We determined the structure of canopy trees, understorey cover and the near-ground microclimate by sampling 576 (5 × 10 m) and 1820 (1 × 1 m) quadrats along 96 transects beside 24 streams near Thunder Bay, Canada. We determined DEEs and MEEs by comparing edge variables with reference forests. We used repeated measures/factorial anovas with Tukey’s post hoc tests to determine DEEs and manova for MEEs.

4. The average microclimatic DEEs extended 8 m into the buffer but only 2·5 m from the fire edge. Similarly, the DEEs for plant life-forms extended 20 m from the buffer edge and 5 m from the fire edge. At the fire edge, the structural MEEs were significantly higher, but the microclimatic MEEs were lower than the buffer edge. We found no significant difference in the overall life-form MEEs, but shrubs, shade-tolerant herbs and grasses were increased at the buffer edge and decreased at the fire edge.

5.Synthesis and applications. We discovered that the ecological structure at buffer and fire edges in boreal forests creates different environmental filters supporting different plant assemblages. Lower structural and microclimatic DEEs and MEEs at the fire edge result from (i) edge location, (ii) intact shrub layer and (iii) disturbance-resilient riparian vegetation. We suggest that replacing the conventional sharp edges of the riparian clear-cut buffer with ‘feathered’ edges by selective harvesting of trees will create wider edges mimicking wildfire legacy and will help to emulate natural disturbance for conserving biodiversity.

Introduction

The fundamental connection between environmental conditions and plant community characteristics was described in early days of plant ecology (Cowles 1899). More recently, abiotic conditions and resources have been recognized as environmental filters that may selectively allow plant colonization and persistence in a community (Keddy 1992; Grime 2002). This means that despite other assembly processes, in the presence of strong environmental filters, the regenerating community will consist of a subset of total species pool (Cornwell & Ackerly 2009). Disturbance (natural or anthropogenic) can alter the strength or magnitude of the environmental filters (Mayfield et al. 2010). For example, clearcutting may relax light restrictions (decreasing the strength of shade filter) or increase ambient temperate (increasing the strength of temperature filter) at the edge.

Habitat fragmentation and edge effects are major causes of biodiversity loss, which have been increasing exponentially over the last 30 years (Ries et al. 2004). Edge effects refer to abiotic and biotic processes at edges that result in a detectable difference in community structure, composition and function near an edge compared to the habitat on either side of the edge (Harper et al. 2005). Edge effects may create abiotic conditions different from nearby non-edge habitats by moderating light, temperature, moisture and wind (Chen, Saunders & Crow 1999; Stewart & Mallik 2006). Edge effects may also change species assemblages by creating conditions that enhance disease-spread and species invasion (Cadenasso & Pickett 2001). Edge-creating agents can be anthropogenic (e.g. clearcutting, road construction, agriculture etc.) or natural (wildfires, disease outbreaks etc.). Most studies on edge effects of riparian forests have focused on clear-cut edges. Less is known about edges created by wildfires, and very little is known about how these two types of edges compare in structure and species composition, which has forest management implications that aim for resource extraction and biodiversity conservation.

Treed buffers (hereafter called ‘buffers’) left beside water bodies are intended to protect riparian habitats from disturbance (Harper & Macdonald 2001). Such buffers control erosion and sedimentation (Steedman 2000), moderate temperature and light (Macdonald, MacIsaac & Herunter 2003; Wilkerson et al. 2006), reduce fine and large organic inputs (Kreutzweiser, Capell & Good 2005), filter and retain nutrients (Vought et al. 1994) and protect the structure and composition of riparian vegetation (Harper & Macdonald 2001).

Intuitively, edges created by fire and buffer edges will differ with respect to the location of the edge and the contrast between the edge and adjacent habitats. First, forest fires do not leave a consistent buffer zone along streams, unlike clear-cut edges. Wildfires typically burn close to the edge of the riparian zone leaving few if any live trees and other vegetation (Lamb, Mallik & Mackereth 2003). Generally, riparian zones have higher soil moisture than surrounding upland forests and consequently different understorey vegetation, fuel loads, ratio of live-to-dead plants and fuel moisture (Dwire & Kauffman 2003). Edge effects may be stronger where the contrast between edges and adjacent habitats is high (Ries et al. 2004). Buffer edges are more abrupt than fire edges with standing dead and charred trees.

It has been suggested that buffers are artificial anthropogenic structures that are functionally very different than the vegetation remnants after fires (Harper et al. 2004). Boreal forest managers are tasked to emulate wildfire patterns by varying the size, shape and rotation duration of clear-cuts in efforts to maintain biodiversity (OMNR 2001). However, little consideration is given for edge location and the structure and consequently the ecological processes driven by a combination of these two factors. Few studies have considered how riparian buffers fit into the paradigm of natural disturbance pattern emulation (Macdonald et al. 2004). Furthermore, the structural attributes of buffer and fire edges along streams and associated plant assemblages are not well documented. This knowledge is essential to develop effective forest management strategy that maintains biodiversity of riparian zones, as clearcutting with buffers is a common practice. It has been argued that the widespread implementation of fixed buffer widths fails to consider the inherent variability of fire and creates narrow strips of unnatural old growth vegetation along streams and lakes (Buttle 2002).

Our objective was to determine how different environmental filters affect plant assemblages at the riparian buffer and fire edges. This knowledge is essential for innovative forest management that aims for tree harvesting and biodiversity conservation through natural disturbance emulation. We hypothesized that the environmental filters created by the post-fire residual structures (standing dead trees) and the inherent high soil moisture (being close to streams) at the fire edge will result in lower depth and magnitude of edge effects (MEEs) and a different plant assemblage than those at the upland buffer edge.

Materials and methods

Study Area

We conducted this study in the boreal mixed-wood forests 30–70 km north-east of Thunder Bay, Ontario, Canada (48°22′N, 89°19′W: 199 m a.s.l). The area has rolling relief with a bedrock substrate overlain by glacial tills. January–July temperatures range from −20 to −26 and 22–25 °C, respectively, and annual precipitation ranges from 700 to 850 mm (Baldwin, Desloges & Band 2000). The dominant trees are black spruce Picea mariana inter-dispersed with white spruce Picea glauca, jack pine Pinus banksiana and trembling aspen Populus tremuloides (Lamb, Mallik & Mackereth 2003). The riparian vegetation is typically either a swamp thicket dominated by tall shrubs such as alders Alnus incana, Alnus viridis spp. crispa, willows Salix spp. or a grass- and sedge-dominated meadow marsh (Lamb, Mallik & Mackereth 2003). The common riparian species are shrubs, A. incana, Cornus stolonifera and Salix spp., Canada blue-joint grass Calamagrostis canadensis, fern Athyrium filix-femina, herbs Thalictrum dasycarpum and Mertensia paniculata, and the upland understorey consists of shrubs Aralia nudicaulis, Rhododendron groenlandicum and Acer spicatum, herb Aster macrophyllus and club-moss Lycopodium annotinum and Lycopodium dendroideum (Flora Ontario 2005).

We selected the study streams based on the following criteria: (i) c. 1 km2 catchment area, (ii) riparian bank slopes ≤15%, (iii) north to south flow direction and (iv) streamside forest dominated by P. mariana. We used a digital elevation model developed by the Ontario Ministry of Natural Resources (OMNR; resolution 25 m) to identify streams with 1 km2 catchment area. Buffer zones are an integral component of Ontario’s Timber Management Guidelines for the Protection of Fish Habitat (OMNR 1988). These guidelines ask forest managers to delineate riparian buffers along all streams visible on a 1 : 50 000 scale map. We chose streams with shoreline slopes ≤15% for consistency of buffer width. Edge orientation has a profound effect on stand structure (Harper et al. 2005) and vegetation composition (Hylander 2005). We selected streams with north to south flow direction to reconcile aspect effects. The selected forest stands were similar in tree species and stem density (Table S1, Supporting Information).

Sample Size and Sampling Protocol

We sampled 24 streams; eight in clear-cut with buffer, eight recovering from wildfires and eight in undisturbed mature forests. Buffers were 28–52 m wide retained on either side of the streams after clearcutting, 2–6 years prior to sampling. Stream sides recovering from fires were sampled 2–7 years post-fire (Fig. 1). The undisturbed mature forest (90- to 100-year-old fire originated) streams were considered ‘reference’ streams where no land use activity occurred within 500 m from the streams. Most boreal forest studies indicate that edge effects dissipate within 50 m of forest edges (Harper et al. 2005).

Figure 1.

 Schematic diagram and sampling design for streams within (a) reference forests, (b) buffers and (c) fires. Understorey vegetation surveys and microclimatic sampling were conducted in 1 × 1 m quadrats at 4-m intervals along 64-m transects laid twice (c. 50 m apart) on either side of the stream (four transects per stream). Canopy structure was sampled in 5 × 10 m contiguous quadrats along 60-m transects laid on either side of the stream. Locations along transects were divided into microhabitats (based on field observations) for multi-response permutation procedure.

Although 10 × 10 m plot size is standard for measuring structural properties of forests, to be sensitive to edge effects, we sampled canopy structure by laying 12 contiguous 5 × 10 m quadrats along the 10-m-wide lateral transects starting at the stream edge on either side of the stream (Fig. 1). For the structural properties, we sampled four transects per stream (two on either side). In each 5 × 10 m contiguous quadrat, we recorded the diameter at breast height (DBH c.1·3 m) of all trees >5 cm diameter. We recorded the number of trees fallen on the ground since logging as ‘downed’ trees. Dead standing trees were recorded as ‘snags’. Wind-throw was calculated as the ratio of downed to total stem. Total tree mortality was calculated by subtracting the number of all live standing trees with green and <50% scorched foliage from the total (standing, downed and wind-thrown stems). The basal area (BA) of each tree was calculated by multiplying the DBH by 7·854e−5 (Zhang, Peng & Dang 2004). We obtained canopy closure data in each quadrat using a convex spherical densiometer (Model A; Forest Densiometers, Rapid City, SD, USA) at three random locations in four cardinal directions. The three measurements were averaged to represent the canopy cover above the quadrat.

To determine the microclimatic and plant community response to edge effects, we studied four 64-m transects, two on either side of each stream starting at the stream edge. To avoid autocorrelation, we kept >50 m distance between the two transects. Edge effects are not monotonic; to capture finer-scale distance-dependent fluctuations of edge attributes, we placed 1 × 1 m quadrats at 4-m intervals along each transect. Where the 4-m interval did not match the ecotonal, fire or buffer edge locations, additional quadrats were placed across the identified edges (Fig. 1). Within the reference forest and buffer, the position of ecotonal edge was considered the point at which a shift in vegetation from riparian obligate species to those typical of the uplands was visually recognizable. The fire edge was considered to be the point at which there was minimal scorching of tall shrubs or the location of live canopy trees; in this ecosystem, fire edges are distinct and easy to delineate. The position of the buffer edge was considered to be the canopy dip line of mature uncut trees. We measured the distance of fire and clear-cut edges from the streams. Within each 1 × 1 m quadrat, we visually estimated percentage cover of all vascular plant and moss species. We grouped plants into the following life-forms: tall shrubs (>1 m in height), low shrubs (<1 m), shade-tolerant herbs, shade-intolerant herbs, grasses, sedges, ferns and allies (ferns, horsetails and club-mosses combined), acrocarpous mosses, pleurocarpous mosses, conifer and deciduous tree seedlings. In each quadrat, we also measured soil moisture and temperature using a HH2 Moisture Meter with a WET Sensor type WET-2 (Delta T Devices; Cambridge, UK), air temperature and relative humidity using a traceable hygrometer (Model 35519-050; Control Company, Friends-wood, TX, USA). All microclimatic measurements were taken three times per quadrat (averaged to represent the quadrat) on days with similar weather conditions between 11:00 and 13:00 (no rainfall within previous 24 h). Vapour pressure deficit (VPD) was calculated from the mean temperature and relative humidity as follows:

image

where RH is relative humidity (%) and vpsat(T) is saturated vapour pressure (kPa) at air temperature T (°C).

Data Analysis and Statistics

We avoided direct comparison of response variables at the buffer and fire edges because of two confounding factors; edge location (on average, fire and buffer edge distance from stream was 9·2 and 40·2 m, respectively; Table S2, Supporting Information) and, as a consequence, the ecosite immediately adjacent to the edges also differed (fire edges were adjacent to riparian areas, whereas buffer edges were bordering upland forests). Therefore, we determined the edge effect of each response variable by comparing measurements at the buffer or fire edges with reference forest conditions at analogous distances from the streams (Fig. 1). To accomplish these, we used two indices, depth of edge effect (DEE) and MEE.

Depth of edge effect is the distance from the edge into the adjacent community in which there is a statistically significant edge effect, also called ‘distance of edge influence’ and ‘edge width’. Repeated measures anova has become a widely used approach for determining DEE (Mascarúa-López, Harper & Drapeau 2006; Boudreault et al. 2008; Pohlman, Turton & Goosem 2009), and we used this method with treatment as the grouping variable and quadrat distance along transects as the repeated measure. For variables deemed significant by the repeated measures, a factorial anova with distance as the fixed factor was run with Tukey’s HSD post hoc test (SPSS 1999) to determine DEE. We considered DEE the point at which two consecutive plots showed significant difference ( 0·05) from the reference forests.

Magnitude of edge effect is a measure of the extent to which a given parameter differs at the edges compared to the reference forests. Following Harper et al. (2005), we calculated MEE as: (e − r)/(r), where e and r equal the value of the variable at the edge and reference forests, respectively. Thus, MEE varies between −1 and +1, and it equals 0 when there is no edge effect. Negative MEE values mean that lower values for a given response parameter were attained at the edge compared to the reference forest and vice versa. The variables were categorized into structural, microclimatic and understorey plant life-form, richness and cover. We calculated the total absolute values of MEE for each category at the buffer and fire edges by summing the negative and positive deviation reflecting the total change of the category.

We used multiple response permutation procedure (MRPP) to determine community-level differences along streamside microhabitat and vegetation gradients. We considered five microhabitats along the transects: riparian zone, ecotonal/fire edge, mid-slope, buffer edge and upland (Fig. 1). We used MRPP to test the null hypothesis of no species compositional difference between the treatments at the five microhabitats and to determine whether the edge effects penetrate into the riparian zones. The Sørensen distance measure and default weighting of groups (Ci nini, where ni is the number of items in group i and Ci is the weight applied to each item in group i) were used for MRPP tests (McCune & Grace 2002).

Results

Depth of Edge Effect

All structural variables measured at the burned sites were significantly different from reference forests along the entire transects (Fig. 2a–f). Conversely at the buffer edges, live tree and snag BA had a 15 m DEE (Fig. 2a,b). Downed tree BA and mortality had a DEE of 10 m. Canopy cover and wind-throw had 5 m DEE (Fig. 2d,e), which was the lowest of all structural parameters at the buffer edge.

Figure 2.

 Mean (±1 SE) values of (a) live tree basal area (BA) (m2 ha−1), (b) snag BA (m2 ha−1), (c) downed BA (m2 ha−1), (d) canopy cover (%), (e) wind-throw (%) and (f) mortality (%) along gradients starting at stream edge. Filled symbols indicate values that are significantly different from reference forest (see Materials and methods). Solid line at 10 m represents the average fire edge/ecotone location, and the dashed line at 40 m represents the average buffer edge location.

Soil moisture DEE was significantly higher at 2 m past the fire edge, while it was significantly lower up to 20 m into the buffer compared to reference forests (Fig. 3a). Within the buffer, soil temperature was significantly higher than that in reference forests up to 8 m past the edge, while no significant DEE was found at the fire edge (Fig. 3b). Soil organic matter depth did not differ at any locations within the buffer, but was significantly lower up to 6 m past the fire edge (Fig. 3c). VPD was significantly higher at both buffer and fire edges compared to the reference forests. VPD had a DEE of 24 and 2 m for the buffer and fire edges, respectively (Fig. 3d).

Figure 3.

 Mean (±1 SE) values of (a) soil moisture (m3 m−3), (b) soil temperature (°C), (c) organic matter depth (cm) and (d) vapour pressure deficit (kPa) along gradients starting at steam edge. Filled symbols indicate values that are significantly different than reference forest condition (see Materials and methods). Solid lines at 10 m represent the average fire/ecotone edge location, and the dashed lines at 40 m represent the average buffer edge location.

Overall, the DEE for plant life-forms was greater at the buffer compared to the fire edges (Fig. 4). Tall and low shrub cover was significantly greater for c. 15 m into the buffer compared to reference forests. However, only the low shrubs showed a DEE at the fire edge with less cover and richness for 7 m past the edge. Cover and richness of shade-intolerant herbs increased up to 11 and 15 m, respectively, into the buffer. Similarly, shade-intolerant herbs had greater cover and richness at the fire edge than the reference forest, but the changes were detected only at 3–4 m. At the fire edge, graminoid cover was greater than the reference forests at 2·5 m past the edge, but richness remained unchanged. Pleurocarpous mosses showed significantly less cover for c. 5 m into the buffer, while changes in species richness were detected at 10 m (Fig. 4).

Figure 4.

 Mean (±1 SE) depth of edge effect (DEE) of plant life-form cover and richness at (a) buffer edges and (b) fire edges. DEE was considered the point of two consecutive plots showing significant difference (P <0·05) from reference forests.

Magnitude of Edge Effect

Live tree and snag BA, tree mortality and canopy closure MEE significantly differed (< 0·05) between the fire and buffer edges and the degree of change was generally two times greater at the fire edge (Fig. 5a). Although there was variation in structural MEE values between the fire and buffer edges, the directionality (positive or negative MEEs) was similar for all variables (Fig. 5a).

Figure 5.

 Mean (±1 SE) magnitude of edge effect (MEE) of (a) structural and (b) microclimatic parameters at fire and buffer edges. MEE was calculated as: (e − r)/(r), where e and r equal the value of the variable at the edge and in the reference forest, respectively. MEE values range between −1 and +1 and is equal to 0 when there is no edge effect. Dissimilar letters (manova) identify significance of difference (α = 0·05) between fire and buffer MEE values. Abs. represents the total absolute MEE value, which includes total deviation both positive and negative and reflects the total change for the response group.

Magnitude of edge effect for soil moisture significantly differed between the fire and buffer edges (Fig. 5b) and showed a different directionality. The absolute soil moisture MEE at the buffer edge was more than double than that at the fire edge. Soil temperature MEE was more than three times higher at the fire edge than at the buffer edge (Fig. 5b). VPD was the only microclimatic parameter that did not differ significantly between the fire and buffer edges showing positive MEE at both edges (Fig. 5b).

Shade-intolerant herbs had the highest positive richness MEEs at both the fire and buffer edges (Fig. 6a). The lowest richness MEE values were obtained for mosses (acrocarpous at the buffer edge and pleurocarpous at the fire edge). The majority of the life-form richness had positive MEEs at the buffer edge, while MEE of most of the life-forms (eight of nine) was negative at the fire edge (Fig. 6a). Furthermore, the majority of the life-form group richness MEEs differed significantly (< 0·05) between the fire and buffer edges. In particular, tall shrubs, low shrubs, shade-tolerant herbs and grasses had opposite responses, with positive values at the buffer and negative values at the fire edges (Fig. 6a).

Figure 6.

 Mean (±1 SE) magnitude of edge effect (MEE) of (a) life-form richness and (b) life-form cover at fire and buffer edges. MEE was calculated as: (e − r)/(r), where e and r equal the value of the variable at the edge and in the reference forest, respectively. MEE values range between −1 and +1 and is equal to 0 when there is no edge effect. Dissimilar letters (manova) identify significance of difference (α = 0·05) between fire and buffer MEE values. Abs. represents the total absolute MEE value, which includes total deviation both positive and negative and reflects the total change for the response group.

Tall shrubs showed the highest positive MEE at the buffer edge (Fig. 6b). Similar to the richness MEE, the shade-intolerant herbs had the highest positive cover MEE at the fire edge (Fig. 6b). Most of the life-forms displayed corresponding trends for richness and cover MEEs (Fig. 6a,b). However, at the fire edge, acrocarpous mosses had a negative MEE for richness (Fig. 6a) but a positive MEE for cover (Fig. 6b). For pleurocarpous mosses, both richness and cover were negative at fire and buffer edges, but for acrocarpous mosses, MEE of cover was positive at the fire edge and negative at buffer edge (Fig. 6b).

Overall, the total absolute structural MEEs at the fire edge were almost two times greater (stronger MEE) than at the buffer edge (Fig. 5a). The total absolute microclimatic MEEs (Fig. 5b) were low, and the microclimatic variables showed the least deviation from the reference forest conditions. Fire and buffer edges had comparable total absolute richness MEE (Fig. 6a). Conversely, the total absolute life-form cover MEE of the fire edges (4·4) was much higher than that of the buffer edges (3·5; Fig. 6b).

Microhabitat Plant Communities

The MRPP analysis showed significant difference in understorey species composition across all microhabitats between the reference forest compared to the fire and buffer sites ( 0·05; Table 1). The plant communities of the riparian zone showed the least deviation (greatest T values; Table 1) from the reference forest. Overall, the plant communities of the burned sites showed a larger deviation (larger negative T-value; Table 1) from the reference forests than did the buffers across most of the microhabitats.

Table 1.   Multiple response permutation procedure pairwise comparisons of the plant composition of the reference forest understorey to buffer and fire sites at regions identified by field observation
RegionBufferFire
TAPTAP
  1. The T-statistic describes the separation among groups (the more negative T is, the stronger the separation is). The chance-corrected within-group agreement (A) indicates within-group homogeneity compared to random expectation. When all items are identical within groups, then = 1 (highest possible value). If heterogeneity within groups equals expectation by chance, then = 0. The P-value indicates the probability of having a more extreme observed delta and is significant at ≤ 0·05.

  2. *Within the fire sites, these locations were burnt (Fig. 1).

Riparian zone−6·75490·0602<0·0001−8·84990·0736<0·0001
Ecotonal/fire edge−6·72960·0701<0·0001−29·31950·3324<0·0001
Mid−slope*−8·90560·0864<0·0001−33·51050·3734<0·0001
Buffer edge*−24·36560·1542<0·0001−48·38190·3717<0·0001
Upland*−37·47250·2660<0·0001−49·03050·3761<0·0001

Discussion

We quantitatively characterized structure, near-ground microclimate and floristic composition of riparian buffer and fire edges in boreal forests. We demonstrate that the buffer and fire edges are significantly different with respect to edge structure (Fig. 2), near-ground microclimate (Fig. 3) and plant assemblage (Fig. 4). We argue that these differences are the direct product of differential environmental filters created at the fire and buffer edges. This is primarily driven by the difference in edge location and inherent difference in edge-creating agents (fire and clearcutting). Forest fires typically burn close to the edge of the riparian zone because of wet soil and insufficient fuel load in the riparian zone (Lamb, Mallik & Mackereth 2003). We found high soil moisture at the fire edge and comparable ecotonal locations at the reference forests, which continue to increase toward streams (Fig. 3a), supporting the notion that streams act as firebreaks owing to wet conditions. Higher values of soil moisture were not found at buffer edges because of the edge location. Being on average 40·8 m away from steams, the buffer edges were beyond the influence of riparian soil moisture. We attribute these differences in DEEs and MEEs to the structural legacy of the fire and buffer edges. Clearcutting at a buffer edge removes all standing timber creating a sharp microclimatic edge, but the post-burn structural legacy remaining at a fire edge creates moderating light and temperature conditions. Therefore, the two edge types produce two different environmental filters; a strong filter at the sharp buffer edge and a weak filter at the fire edge. These results support our hypothesis that the environmental filters created by the post-fire residual structures, and the inherent high soil moisture at the fire edge, form different plant assemblages than at the buffer edge. Our DEE and MEE results also support our hypothesis that the fire edge, with finer environmental filters created by post-fire residual structures, results in lower DEEs and MEEs than the buffer edge. However, the total absolute life-form richness MEE results did not support our assumption of less deviation in the floristic composition at the fire edge than at the buffer edge. We presume that greater deviation of floristic richness and cover at the fire edge is a direct effect of fire, which damaged portions of the plant community, whereas clear-cut harvesting at the buffer edge caused comparatively less physical damage to the plant community.

Harper et al. (2004) studied clear-cut and fire edges of upland (non-riparian) black spruce forests of north-western Quebec, Canada, and reported results similar to this study with lower canopy cover and higher density of snags/downed trees at the fire edge than at the clear-cut edge. Although it is well established that there are major differences in floristic composition between riparian and upland vegetation (Dwire & Kauffman 2003), our study overcomes these differences by focusing on the metrics of edge effects. We show that a structural DEE from the buffer edge extends 5–15 m but it is generally limited to 5 m from fire edges (Fig. 2). After clearcutting at the buffer edge, trees within the buffer are subjected to much higher wind exposure that increases wind-throw, mortality and canopy openings (Mascarúa-López, Harper & Drapeau 2006). The lower canopy cover DEE of the fire edge in our study can be attributed to thick bushes of alder and willow close to streams that remain unharmed by fire. In addition, absence of trees in the riparian zone is another factor for shorter structural DEEs at the fire edge.

We found that on average, the buffer edge affected near-ground microclimate up to 8 m, while the fire edge affected only up to 2·5 m. These values are lower than those reported for upland boreal forest edges, which range from 10 to 50 m (Gignac & Dale 2005). A critical difference between the riparian buffer and upland forest edge is that the former, by being close to streams, are greatly influenced by the natural ecotones (Stewart & Mallik 2006). Streams exert cooling and humid effects on the riparian zone and the upslope, mitigating the warm and dry conditions experienced at the buffer edge (Pohlman, Turton & Goosem 2009). Additionally, fire leaves standing dead trees, which provide shading and cool the forest floor (Dwire & Kauffman 2003). The combined effects of snag density and shrub cover can reduce evaporation and maintain greater relative humidity (Pettit & Naiman 2007). It is reasonable to assume that these differences in microclimatic conditions produced by the different post-disturbance structures (environmental filters, sensuKeddy 1992) at the fire and buffer edges would support floristically different communities. It is generally thought that environmental filters work simultaneously with biological filters, competition and facilitation (Grime 1998; Laakso, Kaitala & Ranta 2001; Lortie et al. 2004). However, in this study, the biological filtering effects would be minimal because of the short time (2–7 years) since edge creation.

Although we found significant difference in the plant assemblages of the buffer and fire edges, their riparian zone species composition showed very little difference from the reference forest (Table 1). This implies that the edge effects of the buffer and fire have not penetrated into the riparian zone causing little change in overall species composition. This may be due to the fact that many riparian species are long-lived and some are robust perennials such as alder, hazel and willow. These plants are rarely affected by fire, which usually stops at the terrestrial-wet ecotones. Even if these plants are scorched by fire, they generally regenerate rapidly by sprouting from underground organs (Lamb, Mallik & Mackereth 2003).

We found greater shade-intolerant herb cover and richness extending more than 15 and 4 m past the buffer and fire edges, respectively. The majority of shade-intolerant herbs that showed higher cover and richness compared to reference forests are heavily reliant on seed dispersal by wind (e.g. Aster ciliotaus, Epilobium angustifolium, Hieracium spp. and Taraxacum spp.). Biswas & Mallik (2010), working in the same general area, reported that buffers act as windbreaks where higher concentrations of wind-dispersed species occur. Janzen (1983) suggested that regenerating vegetation and patch edges often experience a ‘seed rain’ of weeds that are better adapted to disturbed ground and higher light conditions, which may account for the greater richness of shade-intolerant weeds in the buffers. The response of bryophytes to edge creation has been the focus of many studies in the boreal forests (Gignac & Dale 2005; Hylander 2005; Stewart & Mallik 2006), particularly because these species are sensitive to environmental conditions and can act as phytometers reflecting the near-ground microclimate. Our data showed comparably lower pleurocarpous moss cover in the clear-cut area and lower richness (data not shown) 5–10 m into the buffer. The fire edges supported greater cover of acrocarpous mosses such as Ceratodon purpureus and Polytrichum spp. classified as ‘invaders’, which are short-lived, easily dispersed pioneer species. Taken together, we demonstrate that the buffer and fire edges are structurally, microclimatically and compositionally different, and we argue that this new knowledge has implications for forest management in riparian zone.

Conclusions and Management Implications

Fire stops close to streams whereas buffer edges are deliberately set c. 30 m from stream banks. This, together with the inherently different characteristics of the edge-creating agents (clearcutting vs. fire), creates different environmental filters supporting different plant assemblages. We suggest that the structural legacy of the disturbances and their locations at the longitudinal profile of streams are mainly responsible for the difference in plant assemblages at the two edges.

It has been proposed that biodiversity in managed boreal forests can be maintained through management that mimics natural disturbance regimes (Macdonald et al. 2004). Guidelines have been designed to adapt forest harvesting patterns at the landscape and stand level to emulate natural disturbance (OMNR 2001). However, these guidelines rely on the unproven assumption that the size and shape of fires are primarily responsible for creating suitable conditions. Our data suggest that the sharp buffer edge created by clearcutting produces a very different microclimate, which in turn supports a different understorey to that found within the gradual structural features of the fire edge. In other words, the ecological structure and processes occurring at the buffer and fire edges along boreal streams are quite different and must be taken into account in forest management that aims to emulate natural disturbance to maintain biodiversity. Measures should be taken to reduce the contrast between the buffer and the fire edge structures. Harvesting too close to streams is unacceptable because of the severe negative effects on water quality, aquatic communities and the stability of stream banks (Macdonald, MacIsaac & Herunter 2003; Kreutzweiser, Capell & Good 2005; Wilkerson et al. 2006). Prescribed fire is also not a viable option because of challenging logistics, high costs and risks to human life and property. We recommend replacing the conventional sharp edge of riparian buffers created by clearcutting with a ‘feathered’ edge created by selective harvesting to produce a wider edge structure that better emulates the natural disturbance of wildfire. In areas with a wide riparian zone, the width of riparian buffers may also be reduced because riparian vegetation is resilient to disturbance.

Acknowledgements

We thank C. Spalvieri and S. Biswas for their assistance with the fieldwork and Drs K. Harper, E. Lamb, R. Mackereth, J. Carney, D. Morris and T. Hazenberg for comments on earlier drafts. We thank Dr C. Shahi for statistical advice during exploratory data analyses. Funding for this research was provided by Forestry Futures Trust Ontario and Living Legacy Trust, Lakehead University, Canada.

Ancillary