New Forests

, Volume 45, Issue 2, pp 215–234 | Cite as

Variation in post-wildfire regeneration of boreal mixedwood forests: underlying factors and implications for natural disturbance-based management

  • Stefanie M. Gärtner
  • Mike Bokalo
  • S. Ellen Macdonald
  • Ken Stadt


To test the direct regeneration hypothesis and support natural disturbance-based forest management we characterized the structure and composition of boreal mixedwood forests regenerating after large wildfires and examined the influence of pre-fire stand composition and post-fire competing vegetation. In stands which had been deciduous (Populus sp.)-dominated, conifer (white spruce)-dominated, or mixed pre-fire we measured regeneration stocking (presence in 10 m2 plots), density and height 10–20 years post-burn in five wildfires in Alberta, Canada. Most plots regenerated to the deciduous or mixed stocking types; plots in the older fire and in stands that were pure conifer pre-fire had higher amounts of conifer regeneration. Surprisingly, regeneration in pre-fire ‘pure’ white spruce stands was most often to pine, although these had not been recorded in the pre-fire inventory. Pre-fire deciduous stands were the most resilient in that poplar species dominated their post-fire regeneration in terms of stocking, density and height. These stands also had the highest diversity of regenerating tree species and the most unstocked plots. High grass cover negatively affected regeneration density of both deciduous and conifer trees. Our results demonstrate the natural occurrence of regeneration gaps, pre- to post-fire changes in forest composition, and high variation in post-fire regeneration composition. These should be taken into consideration when developing goals for post-harvest regeneration mimicking natural disturbance.


Direct regeneration hypothesis Forest resilience Boreal mixedwood stand dynamics Picea glauca Populus tremuloides Post-fire regeneration Wildfire 


Wildfire is the predominant natural disturbance in many forests around the world, including the boreal forest of western Canada (Johnson et al. 1998). Post-fire regeneration is affected by the characteristics of the fire including fire interactions with the landscape topography (e.g., Albani et al. 2005), site conditions (e.g., Bridge and Johnson 2000), and the regeneration ecology of species (e.g., Chen et al. 2009). The direct regeneration hypothesis predicts that forests will recover to the pre-disturbance composition within a short time period after disturbance (Yih et al. 1991); some studies in the boreal have supported this (e.g., Greene and Johnson 1999). Other studies, however, found that this does not hold true in the boreal because, while broadleaf species and conifers with serotinous cones regenerate directly after a disturbance this may not be true for non-serotinous conifers (Lavoie and Sirois 1998; Ilisson and Chen 2009b).

To adopt a natural disturbance-based approach to forest management (Bergeron and Harvey 1997) we need to understand the range of variation in natural regeneration following wildfire. Pre- to post-fire differences in composition and structure are of interest because they can tell us about the resilience of different pre-fire cover types, and indicate the extent to which pre-harvest composition should be considered in planning for regeneration. There is a need for data with which to benchmark post-harvest regenerating stands relative to those establishing after natural disturbance, because data from regeneration surveys are being used to predict future yields (ASRD 2007). Further, such information can be used to develop approaches for post-harvest regeneration that more closely resemble the natural situation and do not require expensive management inputs (Lieffers et al. 2008). As our landscapes become increasing dominated by managed forests and post-fire stands are widely subjected to salvage harvesting, we are rapidly losing the opportunity to create the quantitative benchmarks needed to inform approaches and standards for natural disturbance-based management.

Most of the boreal mixedwood stands harvested in western Canada today are of fire origin and their regeneration has been structured by large wildfires of variable severity (Burton et al. 2008). We have a good understanding of regeneration processes in boreal forests (e.g., Simard et al. 1998; Greene et al. 1999, 2004; Johnstone et al. 2004) and a general understanding of post-fire boreal mixedwood stand dynamics (Andison and Kimmins 1999; Chen and Popadiouk 2002; Peters et al. 2006; Taylor and Chen 2011; Bergeron et al. 2014). However, we have a poorer understanding of the relationship between pre- and post-fire forest composition in the western boreal (but see Greene and Johnson 1999; Lavoie and Sirois 1998; Bergeron 2000; Chen et al. 2009). Most importantly, we lack quantitative information on the range of variation in structure and composition of post-fire regeneration—particularly for the diversity of boreal mixedwood forest types in the western Canadian boreal and for stands of intermediate age (past the early post-fire regeneration stage but <20 years). Such information is urgently needed for the development of approaches and standards for natural-disturbance based management at the stand and landscape scale (Andison and Kimmins 1999).

The dominant tree species in the boreal forest are generally well adapted to disturbance by fire (Greene et al. 1999). The broadleaved species [trembling aspen (Populus tremuloides Michx.); balsam poplar (Populus balsamifera L.) and white birch (Betula papyrifera Marsh.)] have strong vegetative reproduction (Greene et al. 1999). The boreal pine species [Jack pine (Pinus banksiana Lamb.), lodgepole pine (Pinus contorta Dougl.)] and black spruce [Picea mariana (Mill)] have long lived aerial seed banks held in serotinous or semi-serotinous cones. Only white spruce [Picea glauca (Moensch) Voss] and balsam fir [Abies balsamea (L.) Mill.] do not have these advantages; their post-fire establishment therefore depends on seed dispersal from surviving individuals, patches or from unburned edges (Greene et al. 1999). These two species are often described as ‘late-successional’ because of their ability to grow under lower light conditions and establish under the canopy of the broadleaf or pine pioneers (Greene et al. 1999). However, white spruce can establish directly after fire during the small window of opportunity when competition is low and mineral soil seedbeds are available; white spruce regeneration is particularly abundant when these ideal conditions coincide with a heavy cone crop year in this masting species (Peters et al. 2005, 2006). Given the short fire cycle in the boreal forest of western Canada (<5 % of the area is older than 150 years; Weir et al. 2000), immediate post-fire regeneration of white spruce must be an important process explaining the prevalence of mature white spruce forest on the landscape.

Herein we present data on forest regeneration 10–20 years after large wildfires on upland sites within the boreal mixedwood natural region in Alberta, Canada. Our objective was to document the range of variation in composition of regenerating forests and to examine the influence of pre-fire cover type and post-fire competing vegetation on: post-fire stocking type, presence of P. glauca regeneration, the density of conifer and deciduous regeneration and the height of Populus sp. regeneration. We hypothesized that, even if post-fire relative abundance is not proportional to pre-fire, the pre-fire stand composition will have an important influence on post-fire composition of these forests.

Materials and methods

Study area

We sampled stands in five wildfires located in the Lower Foothills (three fires) and Central Mixedwoods (two fires) natural subregions of central and northeastern Alberta (Table 1). These represent a gradient of climatic conditions across which white spruce-dominated mixedwood forests are found in Alberta (Natural Regions Committee 2006). The Lower Foothills is at higher elevation than the Central Mixedwood (650–1,625 m vs. 200–1,050 m. a.s.l.) and is cooler in summer (mean temperatures 12.8 vs. 13.7 °C), warmer in the winter (mean temperatures −7.8 vs. −11.9 °C) and has higher annual (464 vs. 389 mm) and summer (295 vs. 238 mm) precipitation. Soils on upland sites in both natural subregions developed on morainal and lacustrine deposits and are predominantly Luvisols and Brunisols (Beckingham and Archibald 1996; Beckingham et al. 1996). In both regions, mesic sites have varying dominance by broadleaf (trembling aspen, balsam poplar, paper birch) and conifer (lodgepole pine, Jack pine, white spruce, balsam fir) trees.
Table 1

Characteristics of the fires sampled in the study

Fire Name

Fire type

Year of burn

Ecological subregionb



Size (ha)

Chip Lake (CL-10)a



Lower Foothills (LF)




O′Chiese (OC-20)



Lower Foothills




Virginia Hills (VH-10)



Lower Foothills




Mariana Lakes (ML-13)



Central Mixedwood (CM)




Mitsue (MI-10)



Central Mixedwood




aAbbreviations in brackets: fire name, including age in years at time of sampling; 10- and 13-year fires were considered ‘younger’ while the 20-year fire is referred to as ‘older’

bNatural Regions Committee (2006)

Fires 10–20 years old (1988–1998) were selected with the aid of provincial fire maps and the provincial fire database (ASRD 2008). This age range was chosen because it is similar to the age range at which postharvest regeneration performance surveys are conducted in Alberta (nominally 8–14 years although most surveys are done towards the higher end of this range; ASRD 2003, 2007). We selected fires that burned predominantly on upland (modal) ecosites with a representative range of pre-fire forest composition. We selected only early season fires because these are historically dominant in this part of the boreal (Johnson et al. 1999). Salvaged areas were deemed unsuitable for sampling and finding fires that had sufficient representative unsalvaged area severely constrained our selection of fires. We were only able to find five suitable and accessible fires, four at the younger end of the chosen range (10–13 years) and one older (20 years) (Table 1). Each of these fires was large and encompassed considerable variation within its boundaries, as is typical in wildfires in this region.

Stratification and sampling design

Within each fire we selected sample units on mesic upland sites which were generally flat. Forest inventories for the study region delineate stands from aerial stereo photos (1:15,000–1:40,000 scale) as areas of similar species composition, height, crown cover, and landscape position subject to a minimum size of 2 ha (Alberta Forest Service 1985). We designated sample units as areas >2 ha which were homogenous in pre-fire species composition, structure, and site conditions. Sampling units were selected using the “Phase 3” forest inventory maps (Alberta Forest Service 1985), which provided information on pre-fire stand structure and species composition; unfortunately, information on pre-disturbance forest composition is not retained in the more recent Alberta Vegetation Inventory. Suitable sample units were those with a pre-fire composition dominated by white spruce, aspen, or a mixture of the two species. Thus, the sampled areas were assumed to be suitable for post-fire regeneration to white spruce and aspen. All areas were checked in the field before sampling to ensure they had been burned (charred snags, few residual trees remaining) and had not been salvage harvested. The sample units were stratified, based on the pre-fire species composition, into categories according to the Alberta Forest Management Planning Standard (ASRD 2006) as follows: conifer dominated (C), mixedwood with conifer species leading (CD), mixedwood with deciduous species leading (DC), and deciduous dominated (D) (Table 2). In all cases the conifer species was white spruce and the deciduous was aspen or balsam poplar. Here we use the term ‘deciduous’ to refer to broadleaved species as is the convention in forest management in the region.
Table 2

Number of sampled plots (503 in total) in the five fires by pre-fire cover typea: ‘pure’ conifer (C), conifer-dominated (CD), deciduous-dominated (DC) and ‘pure’ deciduous (D)

Fire name





Chip Lake (CL-10)





O’Chiese (OC-20)





Virginia Hills (VH-10)





Mariana Lakes (ML-13)





Mitsue (MI-10)










aPre-fire cover type was determined using Alberta Phase 3 inventory (Alberta Forest Service 1985), the inventory is based on 1:15,000 scale aerial photographs, the mapped forested stands are areas of uniform canopy attributes; Canopy species composition is expressed as proportions of estimated merchantable volume for stands >12 m in height, and otherwise by proportion of canopy cover

Data collection

A systematic grid (30 m spacing) was laid out in each sampling unit and circular plots (10 m2; 1.78 m radius) were centered on each point. We had planned on sampling at least 30 plots per sampling unit (if the sampling units were at least 2.7 ha) but this was not always possible due to plots falling in unsuitable locations (rock, water, unburned, salvage logged). In total we sampled 503 plots across the four pre-fire forest types and the five fires (Table 2).

At the time we sampled there were several different ground survey protocols for sampling post-harvest tree regeneration in Alberta. We followed one of the more detailed protocols, the WAS Regeneration Survey Manual (WAS 2008), with several additional measurements. In each 10 m2 circular plot, the number of trees taller than 0.3 m was counted per species. We further categorized the abundance of white spruce smaller than 0.3 m in three classes: none, few (≤5) and many (>5). For the tallest conifer and deciduous tree (over 0.3 m) in each plot we measured the total height, the diameter at stump (0.3 m) height and recorded the species. Crown closure of tall shrubs (i.e., alder and willow) was estimated in classes (<5; 5.1–10; and >10 % in 10 % classes). The canopy height of tall shrubs within the plot was estimated. Additionally, percent cover (<5; 5.1–10; and >10 % in 10 % classes) was visually estimated for eight layers: mosses and lichens, graminoids, forbs, shrubs, trees (above 10 m in height; i.e., pre-fire residuals), exposed mineral soil, exposed rock, and fallen dead wood. We estimated the distance to a potential seed tree in four categories: <30, 30–100, 100–500 and >500 m. The data were collected in the summer of 2008.

Data analysis

In Alberta, “stocking” is defined as the presence of an acceptable (meeting regional quality, height and species criteria) tree in a 10 m2 plot. Our criteria were simply the presence of deciduous species or conifer species >0.3 m height. We summarized stocking by time since fire, i.e., 10 years [Chip Lake (CL-10), Mitsue (MI-10), Virginia Hills (VH-10); 13 years: Mariana Lakes (ML-13); 20 years: O’Chiese (OC-20) and also for fires grouped as: “younger” (y) (10 and 13 years old: CL-10, MI-10, VH-10, ML-13] versus “older” (o) (OC-20) and by the four pre-fire cover types (Supplement 1). Plots were also categorized as to their stocking type: conifer (‘C’) plots had at least one conifer taller than 0.3 m and no deciduous >0.3 m, deciduous (‘D’) plots had at least one deciduous >0.3 m and no conifer >0.3 m, mixed (‘MX’) plots had at least one conifer >0.3 m and one deciduous >0.3 m, while not sufficiently restocked (‘NSR’) plots had no trees >0.3 m.

Species composition of regeneration

To examine variation in species composition of the regeneration we used unconstrained ordination on the dataset of number of trees (>0.3 m height) per species for each plot. Plots with no regeneration were removed, leaving a total of 439 plots. We used non-metric multidimensional scaling (NMDS, Oksanen et al. 2010) based on chord distance and 20 random starts. A step-across procedure was used to replace the dissimilarities for plots which had no species in common, and zero dissimilarities were changed to 0.0012 (a randomly selected number lower than all non-zero dissimilarities).

To evaluate the influence of pre-fire cover type, fire and their interaction on composition of the regeneration we conducted a permutational multivariate analysis of variance (PERMANOVA) (Anderson 2001). The analysis is partitioning sums of squares of a multivariate data set analogous to the multivariate analysis of variance but uses semi-metric and metric distance matrices. Significance tests (α = 0.05) were done using F tests based on sequential sums of squares from permutations of the raw data (Oksanen et al. 2010).We used the multiple-response permutation procedure (MRPP) to test for pairwise differences in regeneration composition between the two subregions and between pre-fire cover types, applying the chord distance and 999 permutations with group size (number of plots) as a weighting factor. All vegetation analyses were done in R using the “vegan” package 1.17-3 (Oksanen et al. 2010).

Factors influencing regeneration

To determine which factors were influencing regeneration composition we used classification and regression trees. Recursive partitioning was used to derive classification trees for categorical response variables: stocking type and presence/absence of white spruce regeneration. We used regression trees for continuous response variables (density of conifer, density of deciduous trees >0.3 m height, height of Populus sp. >0.3 m) within a conditional inference framework to explain variation in the response variables as a function of the explanatory variables. We accomplished this with the non-parametric conditional inferences tree using the “ctree” function of the r package “party” (Hothorn et al. 2010). At each step of the analysis, one explanatory variable was selected from all the available variables, based on the best separation of two homogenous groups using a permutation test; this point is determined by a numerical value (threshold) of the explanatory variable (Hothorn et al. 2006a, b). The relationships between the response variable and explanatory variables are presented by a dichotomous tree diagram with nodes that represent split points, branches that connect nodes, and leaves or terminal nodes that represent the final groups.

Explanatory variables included in these analyses were: the percent cover of the different ground layers [grass, forb, shrub, tree, moss, lichen, mineral, rock, organic, downed wood (CWD)]; crown closure of alder (Alder %) and willow (Willow %); average height of alder (Alder-ah) and willow (Willow-ah); fire, pre-fire cover type, time since fire, and time to first white spruce mast crop after the fire (based on Peters et al. 2005; Martin-DeMoor et al. 2010). For the analyses of density of conifer and deciduous trees >0.3 m height, as well as the height of Populus sp. >0.3 m we included the number of trees of other species within the plot as an explanatory “competition” variable.


Range of variation in regeneration composition

There was considerable variation in regeneration composition within and between fires, stand ages and pre-fire cover types (Supplementary 1).

We next examined the regeneration composition based on which species were ‘potential crop trees’; i.e., the tallest conifer and/or deciduous tree above 0.3 m height (Supplementary 1). The two Populus species were dominant in this regard, being the only potential crop tree in 60 % of the plots in the younger fires and in 47 % of plots in the older fire. Only stands that were ‘pure’ conifer pre-fire had white spruce as the only potential crop tree, and there were very few of these (2.2–7.7 % of plots). In the younger pre-fire conifer stands, 13 % of plots had only white birch as the crop tree.

Despite the fact that we avoided sampling areas that had pine pre-fire, many plots regenerated to pine. In stands that were ‘pure’ white spruce pre-fire 8 % of the plots in the younger fires had pine as the only crop tree (either lodgepole or jack pine, depending on the region). For plots that regenerated to a mixture (i.e., had both deciduous and conifer trees >0.3 m) the proportion that included some pine was even higher; overall for stands that were ‘pure’ white spruce pre-fire, 34 % of plots regenerated to a pine-deciduous mixture. Averaged over all pre-fire cover types, a deciduous—pine mixture was found in 10 % of the plots in the younger stands and 32 % of plots in the older fire. In contrast, white spruce–deciduous (aspen, poplar or birch) mixtures were found in only 5 % of the plots in the younger fires and 11 % of plots in the older fire.

The regeneration composition was highly variable and the NMDS ordination showed no clear differentiation among the four pre-fire stand types. The most obvious pattern in the ordination diagram was the association of plots that were ‘pure’ white spruce pre-fire with tree species that did not regenerate in the other three cover types (e.g., tamarack, black spruce, balsam fir) along with pine and white spruce (Fig. 1). The linear pattern of plots from the pre-fire mixed and pure deciduous stands on the right side of the first axis (Fig. 1) is a result of aspen dominance in terms of frequency and density, which defines the upper limit of the first NMDS axis. Aspen regeneration was present in most of the plots which had contained a deciduous component pre-fire (D, DC, CD); thus, there was little differentiation among these three pre-fire cover types in terms of regeneration composition (see also results of MRPP below).
Fig. 1

Results of NMDS ordination examining regeneration composition. The composition dataset was based on number of individuals (>0.3 m height) per plot by species (the ordination used Chord distance, 20 random starts, was iterated for two dimensions, Stress: 19.04, non-metric fit r2 0.97, linear fit r2 0.91). Symbols represent individual plots coded by pre-fire cover type [‘pure’ conifer (C), conifer dominated (CD), deciduous dominated (DC), ‘pure’ deciduous (D)], 95 % ellipses around pre-fire cover type centroids are given (goodness of fit: r2 0.13, p = 0.001). Centroids are given for the fires: Chip Lake (CL-10), Mariana Lakes (ML-13), Mitsue Lake (MI-10), O′Chiese (OC-20), Virginia Hills (VH-10), (goodness of fit: r2 0.11, p = 0.001) and ecological subregions: Central Mixedwoods (CM), Lower Foothills (LF) (goodness of fit: r2 0.06, p = 0.001). Species codes show their location in ordination space [Potr trembling aspen (Populus tremuloides), Poba balsam poplar (Populus balsamifera), Bepa white birch (Betula papyrifera), Piba jack pine (Pinus banksiana), Pico lodgepole pine (Pinus contorta), Pima black spruce (Picea mariana), Pigl white spruce (Picea glauca), Abba balsam fir (Abies balsamea), Lala tamarack (Larix laricina)]. Species were included in the ordination diagram by weighted averages of the plot scores. Environmental factors (pre-fire cover type, fire, ecological subregion) show the averages of factor levels (class centroids). Confidence limits (0.95) for the cover types were included in the ordination diagram based on the standard error of the (weighted) averages of the plots

The PERMANOVA showed a significant effect of fire, pre-fire cover type and their interaction on regeneration composition (Table 3). The subsequent MRPP showed significant differences in post-fire composition between stands that were ‘pure’ conifer pre-fire versus the other three cover types, which did not differ from one another (Table 4). There were also significant differences between the Central Mixedwoods (CM) and the Lower Foothills (LF) in terms of regeneration composition (by MRPP, Table 4). In the Lower Foothills fires (CL-10, OC-20, VH-10, upper part of the ordination diagram in Fig. 1) there was more regeneration of balsam poplar, tamarack and lodgepole pine. In contrast, more regeneration of balsam fir, paper birch, jack pine and white spruce was observed in the Central Mixedwood fires (MI-10, ML-13, lower part of the ordination diagram in Fig. 1).
Table 3

Results of PERMANOVA testing for the effects of pre-fire cover type, fire and their interaction on regeneration composition






Pre-fire cover type










Pre-fire cover type x fire





Table 4

Results of pairwise comparisons of regeneration composition by means of multi-response permutation procedure tests (MRPP) between ecological subregions and pre-fire cover types




Ecological subregionsa

 CM versus LF



Pre-fire cover typesb

 C versus CD



 C versus DC



 C versus D



 CD versus DC



 CD versus D



 DC versus D



aEcological Subregions: Central Mixedwoods (CM) and Lower Foothills (LF)

bpre-fire cover type: ‘pure’ conifer (C), conifer dominated (CD), deciduous dominated (DC) and ‘pure’ deciduous (D)

cA: chance-corrected within-group agreement which ranges from 0 to 1 where 1 indicates that the composition of all plots are identical within groups. With significant separation of groups, A values <0.1 are common with community data (McCune et al. 2002)

Significant differences are indicated in bold

Factors influencing regeneration


The notable difference between pre-fire ‘pure’ conifer stands and the other cover types, as described above, was further emphasized by the results of the classification tree; the most important factor determining the post-fire stocking was whether the pre-fire stand was ‘pure’ conifer or not (Fig. 2). Within the pre-fire conifer stands, the percent grass cover (mostly Calamagrostis canadensis), was the most influential factor determining whether the plots regenerated to the mixed (MX) versus the deciduous (D) stocking type Fig. 2). When grass cover was higher than 50 %, 49 % of the plots were stocked to D and about 35 % were NSR. In contrast, when grass cover was lower than 50 % there was a higher probability of plots regenerating to the MX stocking type. The subsequent influential factor for pre-fire ‘pure’ conifer stands with low grass cover was the cover of alder. When alder cover was <20 %, 62 % of the plots regenerated to the MX stocking type while 22 % regenerated to the C stocking type. Of the 8 plots with alder cover >20 %, none regenerated to the C stocking type, while there were fairly similar proportions of plots regenerating to the MX or D stocking type or being identified as NSR (~25–37 %).
Fig. 2

Classification tree to predict post-fire stocking type based on the conditional inference tree (cTree) model. The encircled explanatory variables are those showing the strongest association to the response variable. Values on lines connecting explanatory variables indicate splitting criteria; for example, the first split separated plots in the pre-fire ‘pure’ conifer (C) cover type (left split) from those in the other three cover types (right side of the split). Numbers in boxes above the explanatory variable indicate the node number. “n =” next to terminal nodes indicate the number of plots classified in that node. Bar graphs illustrate the proportion of plots in that node that regenerated to a stocking type. C = conifer, MX = mixed, D = deciduous, NSR = not sufficiently restocked. See Methods for detailed definitions of the stocking types. Explanatory variables: pre-fire CoverType: ‘pure’ conifer (C), conifer dominated (CD), deciduous dominated (DC), ‘pure’ deciduous (D), “Grass %”: percent cover of graminoid species in the plot, “Alder %”: percent cover of alder in the plot, “Fire”: Chip Lake (CL), Mariana Lakes (ML), Mitsue Lake (MI), O′Chiese (OC), Virginia Hills (VH)

In the other three pre-fire cover types (D, DC, CD; right split of Node 1 in Fig. 2) plots most often regenerated to the D stocking type but there were differences among the fires in terms of the proportions of MX versus NSR plots (Node 7 in Fig. 2). The 10 year old fires separated from the two other fires due to the latter having a much higher proportion of plots regenerating to the MX stocking type (right split of Node 7 in Fig. 2). Among the 10 year-old fires one (VH) had a much higher proportion of NSR plots (split at Node 9, Fig. 2).

White spruce presence and density

Since white spruce regeneration was so infrequent in our plots, we used another classification tree (Fig. 3) to examine factors influencing its presence/absence in a plot regardless of whether it was the tallest conifer. The most important factor influencing presence of white spruce trees taller than 0.3 m was the mast year following the fire which separated the older fire (O’Chiese, 20 years, mast in 1991)—which had a higher percentage of plots containing a white spruce (>0.3 m height)—from the four younger fires (10 or 13 years; mast in 1998). The next most important factor in the older (O’Chiese) fire was the pre-fire cover type; pre-fire ‘pure’ deciduous stands had a very low frequency of plots with a white spruce (<5 %) compared to pre-fire ‘pure’ conifer or mixedwood stands in this fire (>50 % frequency of plots with a white spruce). For the four younger fires (right split of Node 1; Fig. 3) the next mast year was in 1998—either in the same year as (CL-10, VH-10, MI-10) or 3 years after the fire (ML-13). For these, grass cover was the most important factor determining the presence of spruce regeneration >0.3 m height. If grass cover was ≤60 % the chance of finding a white spruce was 15 %; with higher grass cover, this dropped to about 3 %.
Fig. 3

Classification tree to predict the presence (dark proportion of bar) or absence (light grey proportion of the bar) of white spruce regeneration (>0.3 m height) based on the conditional inference tree model. See Fig. 2. Explanatory variables: “first mast” after the fire: 1991 (3 years after for the 1988 fire) or 1998 (3 years after for the 1995 fire and in the same year for 1998 fires); pre-fire “CoverType” ‘pure’ conifer (C), conifer dominated (CD), deciduous dominated (DC), ‘pure’ deciduous (D), “Grass %”: cover of graminoid species in the plot

Even considering presence of white spruce of any size (including the semi-quantitative data on spruce <0.3 m) our results showed that 79–91 % of the plots had no white spruce in the older and younger fires, respectively (Supplement 2). Most of the small white spruce trees were found in the stands that were pure white spruce pre-fire. We noticed very few seed trees left standing in any of the fires; most of the established white spruce would have come from seed sources 500 m or further away—or from seed trees that had fallen before the time of our data collection.

In concordance with the results for stocking, pre-fire stand type was the most important factor influencing conifer regeneration density (Fig. 4). Pre-fire conifer stands had higher densities of conifer regeneration than did the other three cover types. Within the pre-fire conifer stands, the amount of grass cover was the most influential factor. There was little or no conifer regeneration in plots where grass cover was >30 %. In plots with lower grass cover, shrub cover had an additional influence; when shrub cover was ≤5 %, the median conifer density was about 25 trees per 10 m2 plot, whereas with higher shrub cover this dropped to 5 per plot.
Fig. 4

Regression tree to predict conifer regeneration density (trees per 10 m2 plot, height > 0.3 m) within the conditional inference framework. See Fig. 2. The p values listed at each node represent the significance of the influence of the listed independent variable on the response variable. Box plots at the terminal nodes show the distribution of regeneration density (number of trees per plot) for plots within that branch. Boxes represent the inner-quartile range (25th–75th percentiles) of the data, dark horizontal lines within the boxes represent the median, while whiskers represent the extent of data within the 1.5 × inner-quartile range. Circles above and below the whiskers represent data points outside of this range. The number of plots that fall within each branch (n) is listed above the box plots. Explanatory variables: pre-fire “CoverType”: ‘pure’ conifer (C), conifer dominated (CD), deciduous dominated (DC), ‘pure’ deciduous (D); “Grass %”: percent cover of graminoid species in the plot; “Shrub %”: percent cover of shrubs in the plot; “firstmast”: in years after the fire

In the pre-fire cover types with a deciduous component (D, DC, CD) the timing of white spruce mast after the fire was an important factor determining the density of conifer regeneration. In the three youngest (10 years old) fires, for which the mast was in the same year as the fire (1998), conifer regeneration density was very low (right split at Node 7 in Fig. 4). In the 13 and 20 year older fires, for which the mast occurred 3 years after the burn conifer regeneration densities were higher. For these fires, shrub cover was a further controlling factor; with shrub cover ≤15 % (left split at Node 8; Fig. 4) the median conifer density was about 20 trees/plot whereas higher shrub cover resulted in lower conifer regeneration densities (right split at Node 8, Fig. 4).

Deciduous density and height

The first factor influencing density of deciduous trees was the fire (Fig. 5). Deciduous regeneration density was, overall, considerably greater in CL-10 and ML-13 than in the other three fires (split at Node 1; Fig. 5). In these two fires the next influential factor was pre-fire cover type; stands with a greater deciduous component pre-fire (D, DC) had higher densities of deciduous regeneration post-fire than did stands with more conifer pre-fire (C, DC) (split at Node 2; Fig. 5). For stands of the D or DC pre-fire cover type, the final influential factor was grass cover (split at node 3; Fig. 5). The median deciduous regeneration density was 28 stems per plot when grass cover was ≤10 % but about half that (15 trees per plot) with higher grass cover. For stands that had a greater conifer component pre-fire (right split of Node 2; Fig. 5), the next important factor was whether their pre-fire cover type was CD or C (split at Node 6; Fig. 5). The CD stands had higher densities of deciduous regeneration (median ~10 per plot) than did the ‘pure’ conifer stands (median of 3).
Fig. 5

Regression tree to predict the density of juvenile deciduous (trembling aspen, balsam poplar and white birch) regeneration (per 10 m2 plot, height > 0.3 m) within the conditional inference framework. See Figs. 2, 4. Box plots at the terminal nodes show the distribution of regeneration density (number of trees per plot) for plots within that branch. Explanatory variables: “Fire”: Chip Lake (CL-10), Mariana Lakes (ML-13), Mitsue Lake (MI-10), O′Chiese (OC-20), Virginia Hills (VH-10); pre fire “CoverType”: ‘pure’ conifer (C), conifer dominated (CD), deciduous dominated (DC), ‘pure’ deciduous (D); “Grass %”: percent cover of graminoid species

For the other three fires (right split at Node 1; Fig. 5), which had generally lower deciduous regeneration density, grass cover was also an important factor. Plots with ≤30 % grass cover had higher deciduous regeneration density (median 8 per plot) than plots with greater grass cover (split at Node 9, Fig. 5). In plots with grass cover >30 % there was a significant influence of pre-fire cover type; pre-fire D and CD stands had higher deciduous regeneration densities than did DC and C stands (split at Node 11; Fig. 5).

As mentioned above, aspen was not only the most frequent and abundant regenerating species, but also the tallest. For other species we did not have enough trees to examine factors impacting height. We examined factors influencing aspen height only in the four younger fires because 20 years post-fire, the trees were much taller. Aspen regeneration in the Virginia Hills (VH-10) fire was shorter (median 2.8 m) than in the other three fires (split at Node 1; Fig. 6), for which there was evidence of a significant influence of competition from other regeneration or from tall shrubs. Aspen were, overall, shorter when there were ≤10 stems per plot of deciduous regeneration (left split at Node 2; Fig. 6). In this case aspen were taller when willow cover was ≤10 % than with higher willow cover (median height of 4.5 vs. 3 m respectively). In more densely stocked stands (>10 stems per plot of deciduous regeneration), aspen were taller (right split at Node 2; Fig. 6) and also showed evidence of a significant influence of alder presence/absence. Median aspen regeneration height was greater when alder was present than absent (6.8 vs. 5 m respectively).
Fig. 6

Regression tree to predict the height of aspen (m) within the conditional inference framework. See Figs. 2, 4. Box plots at the terminal nodes show the distribution of average regeneration height (m) for plots within that branch. Explanatory variables: “Fire”: Chip Lake (CL-10), Mariana Lakes (ML-13), Mitsue Lake (MI-10), O′Chiese (OC-20), Virginia Hills (VH-10); “allD”: number of additional deciduous individuals in the plot; “Willow %”: percent cover of willow in the plot; “Alder-ah”: average height of alder in the plot


Post-fire regeneration composition

Our results provide some support for the direct regeneration hypothesis (Yih et al. 1991) and are in accordance with previous predictions for the boreal forest (Greene and Johnson 1999; Chen et al. 2009; Ilisson and Chen 2009a, b) of North America. The composition of post-fire stands tended to correspond to their pre-fire cover type—but there were notable exceptions. The composition of post-fire regeneration was dominated by Populus, mostly aspen; this was expected because of their strong vegetative reproduction ability and fast early growth following fire (Frey et al. 2003; Chen et al. 2009). Plots in which Populus spp. were a part of the pre-fire cover (D, DC, CD cover types) were almost always stocked with Populus spp. post-fire—either exclusively or as a mixture with conifers. More conifer regeneration was observed in pre-fire stands that were conifer-dominated or ‘pure’ conifer pre-fire (CD or C cover types) and ‘pure’ white spruce stands pre-fire had the highest occurrence of restocking to conifer. However, both pre-fire CD and C cover types were most often stocked, post-fire, to the deciduous or mixed type, rather than to conifer only.

There was a wide variety of conifer species regenerating in the pre-fire conifer stands and—surprisingly—the leading (tallest) conifer was most often a pine. Even when we considered white spruce regeneration under 0.3 m, there was a very high percentage of plots that had no white spruce. The amount of pine regeneration was variable among fires but was, overall, quite surprising considering that they were not included in the pre-fire stand attribute data (Phase 3 forest inventory cover type call) as even a minor species. Since pine were not detected in the pre-fire inventory because of low cover, the relatively high densities of pine regeneration found in four of our fires (OC, CL, VH, ML) must be attributable to a few mature pre-fire pine. Jack pine and Lodgepole pine are shade intolerant and are described as “obligate seeders” because of their serotinous cones (Chen et al. 2009). As such, they are adapted to regenerate after fire (Ilisson and Chen 2009b). Lavoie and Sirois (1998) reported a change in species composition from black spruce to jack pine post-fire in the eastern boreal forest and suggested that an increase in fire frequency could trigger the expansion of jack pine in the region.

In general, white spruce recruitment was poor in the five fires we studied even when fires coincided with a spruce mast (the CL, MI, VH fires). Stands that were ‘pure’ white spruce pre-fire, regenerated to a diversity of conifer species—including pine, tamarack, balsam fir, and black spruce—and to a mixedwood (conifer–deciduous) type. Thus management regimes that aim to regenerate ‘pure’ white spruce stands to ‘pure’ white spruce through planting and selective herbicides do not represent the typical post-fire regeneration composition and could cause a loss in tree species diversity on these upland sites, relative to what occurs following wildfire.

An important finding was the relatively high percentage of plots that did not meet any regeneration standard and that this varied among pre-fire cover types. Heterogeneity in fire intensity and severity (consumption of organic matter) can influence post-fire regeneration (Simard et al. 1998; Charron and Greene 2002; Greene et al. 2005; Burton et al. 2008). Spruce regenerate best in mineral soil microsites which are more available following higher severity fires (Charron and Greene 2002; Purdy et al. 2002; Greene et al. 2005). We found the highest occurrence of unstocked plots (regeneration gaps) in pre-fire ‘pure’ white spruce stands in the younger fires (23 % of plots) and in the 20-year old fire in the pre-fire ‘pure’ deciduous and deciduous dominated mixedwoods (11–12 % of plots). These results suggest that some areas within fires likely have stocking gaps which could persist over the long term. It should be noted that the regeneration tree height limit we used (0.3 m for both conifer and deciduous) was lower than for past (~1 m thresholds; ASRD 2007; WAS 2008) or current (0.3 m for conifers, 1.3 m for deciduous; ASRD 2012) performance surveys in Alberta. If we had used a 1 m height limit, the proportion of not sufficiently stocked plots would have been 40 % in the pre-fire ‘pure’ spruce stands and 9–29 % in stands with a pre-fire deciduous component. Similarly, MacIsaac et al. (2006) found unstocked gaps covered 29 % of the stand area in five 14 year-old aspen stands that were regenerating naturally after harvesting in northern Alberta.

Factors affecting conifer regeneration

Post-fire conifer regeneration (stocking and density) was affected by grass and shrub cover. Bluejoint (C. canadensis) grass is known to be a serious competitor to regeneration and growth of both white spruce and aspen (Lieffers et al. 1993; Lieffers and Stadt 1994; Greenway and Lieffers 1997; Landhäusser and Lieffers 1998; Shropshire et al. 2001; MacIsaac et al. 2006). The stronger influence of grass competition in the pre-fire conifer stands may be explained by the fact that there was less vegetative regeneration of Populus spp., which otherwise shades out the competing grass and shrubs (MacIsaac et al. 2006; Man et al. 2008). Our results suggest a threshold of 50 % grass cover, above which white spruce regeneration is very low and the proportion of unstocked plots is as high as 35 %. In plots with lower grass cover, alder had a strong inhibitory effect on regeneration; in plots with >20 % alder cover there were no plots stocked to conifer and 37 % of plots were unstocked. Bartemucci et al. (2002) found shrub-dominated gaps in boreal forests of British Columbia occupied 5.2 % of the area and persisted over a long period.

Timing of the mast year relative to the year of the fire is known to be key to successful recruitment of white spruce (Purdy et al. 2002; Peters et al. 2005, 2006; Martin-DeMoor et al. 2010). We expected good regeneration of white spruce in the three fires that burned in 1998, since this was reported to be a widespread mast year in northern Alberta, and previous studies had found good white spruce regeneration in stands burned in 1998 (Purdy et al. 2002). However, we found poor white spruce regeneration in these fires, particularly for stands which had a pre-fire deciduous component. Our results do conform to those of Purdy et al. (2002) who found low densities of regenerating white spruce in the Virginia Hills fire (28 trees/ha). The poor regeneration we observed could be due to a lack of seed sources. Most of the potential seed trees that were still standing when we sampled were more than 500 m away from our plots; too far for adequate seed rain (Stewart et al. 1998, 2001; Martin-DeMoor et al. 2010). However, seed trees could have died and fallen since they released their seed. It is also likely that post-fire salvage logging in areas adjacent to our study sites removed white spruce seed trees.

Surprisingly, we found higher densities of white spruce in the 13 and 20 year old burns, which had a 3 year delay before the first mast year, than in the 1998 burns. These higher densities could be due to an accumulation of recruitment from the first mast year and subsequent high seed years, despite the rapid deterioration of seedbeds following a fire (Purdy et al. 2002). This is supported by the fact that the oldest (20 years) burn had the highest density of white spruce regeneration and the most small (<0.3 m) seedlings. Previous studies showed peak white spruce sapling densities 5–20 years post-fire (Galipeau et al. 1997; Awada et al. 2004; Johnstone et al. 2004). Together, the evidence supports the importance of delayed or on-going recruitment of white spruce (Lieffers et al. 1996; Peters et al. 2006).

The poor white spruce regeneration we observed could also be due to the fact that spring or early summer fires consume less organic matter and thus have lower availability of the mineral soil microsites that are critical for white spruce recruitment (Moore and Wein 1977; Charron and Greene 2002; Miyanishi and Johnson 2002; Purdy et al. 2002; Peters et al. 2006; Gärtner et al. 2011). However, we chose these early season fires because they are historically predominant in this part of the boreal forest (Johnson et al. 1999) and this is expected to continue under future climates (Albert-Green et al. 2013).

Factors promoting deciduous regeneration

Broadleaved dominance not only increased in frequency following fire but deciduous species also had much higher densities than coniferous species. This confirms earlier findings that broadleaved tree species are promoted by wildfires in the boreal forest (Bartos and Mueggler 1981; Brown and Debyle 1987; Johnson et al. 1998; Frey et al. 2003; Greene et al. 2004; Brassard et al. 2008; Chen et al. 2009; Ilisson and Chen 2009a, b). Density of the deciduous regeneration and height of the aspen regeneration were highly variable among fires. Our two fires with the most deciduous regeneration had densities (~14,000–19,000 trees/ha) similar to those reported 5 years after a fire in Wyoming (14,000–20,000 trees/ha; Bartos and Mueggler 1981). Our other fires had densities about half that. The lower density in the 20 year old fire could indicate that self-thinning had begun (Chen et al. 2009). As for white spruce, fire severity likely also influence aspen regeneration by removal of too much or too little of the organic layer (Fraser et al. 2004).

The condition of the aspen root system likely also influenced density and height of regeneration. Higher vitality and density of the parental root system in stands that were deciduous-dominated pre-fire (D or DC) had higher deciduous regeneration density than other pre-fire cover types (DesRochers and Lieffers 2001; Frey et al. 2003). The variability in deciduous regeneration density among fires could also be partly due to the impact of grass cover, as described above.

The tallest aspen (~6.5 m) were found 10–13 years post-fire, in plots that had more than ten other deciduous trees, and in which alder was present. This apparent beneficial effect of alder could be attributed to its effects on soil nitrogen availability (Landhäusser et al. 2010). The negative effect of willow on aspen height could be due to competition or might suggest that willow abundance is higher on wetter and cooler sites, on which aspen growth would be poorer (Landhäusser et al. 2003).


We found wide variation in post-fire regeneration outcomes—in terms of stocking, composition, densities and heights of deciduous and conifer. However, stands with a pre-fire deciduous component were strongly deciduous-dominated post-fire, demonstrating the highest resilience and conforming to the direct regeneration hypothesis. In contrast, relatively ‘pure’ white spruce stands appear to have poorer resilience to fire, with a surprising proportion of plots that regenerated to pine or deciduous species, or which were unstocked. Both the obligate resprouting species, like Populus spp., and the fire-adapted pine species seem to take advantage of fire by increasing their dominance in the early post-fire landscape. This is in contrast to many studies which document that if the right conditions coincide, white spruce has a chance to regenerate well after fire. It’s worth noting that these results were obtained at a rather early stage in succession. Stands that were pure deciduous post-fire could still possibly develop into mixedwoods with on-going recruitment of white spruce over time, and succeed to spruce dominance when the deciduous species senesce.

Overall our results suggest that regeneration gaps and pre- to post-fire switches in composition occur naturally. ‘Pure’ pre-fire white spruce stands rarely regenerate just to white spruce. Rather, they seem to provide an opportunity for establishment of natural tree species mixtures during early successional stages in the post-fire western boreal mixedwood landscape. The high natural variability in regeneration should be taken into consideration when developing approaches to post-harvest regeneration according to natural-disturbance based management.



We thank the Forest Resource Improvement Association of Alberta (FRIAA) and the Mixedwood Management Association (MWMA) for financial support. We appreciated the assistance of Vern Peters and Lowell Lyseng in fire selection, and Farrah Gilchrist for her help in the field. We are grateful to Alberta Pacific Forest Industries Ltd., Alberta Plywood, Blue Ridge Lumber, Millar Western Forest Products, Weyerhaeuser, and the Western Boreal Growth and Yield Association (WESBOGY) for their assistance.

Supplementary material

11056_2013_9404_MOESM1_ESM.docx (18 kb)
Supplementary material 1 (DOCX 17 kb)


  1. Albani M, Andison DW, Kimmins JP (2005) Boreal mixedwood species composition in relationship to topography and white spruce seed dispersal constraint. For Ecol Manag 209:167–180CrossRefGoogle Scholar
  2. Alberta Forest Service (1985) Alberta phase 3 forest inventory: an overview. ENR report 1/86. Alberta Energy and Natural Resources, EdmontonGoogle Scholar
  3. Alberta Sustainable Resource Development (ASRD) (2003) Alberta regeneration survey manual. Forest Management Branch, EdmontonGoogle Scholar
  4. Alberta Sustainable Resource Development (ASRD) (2006) Alberta forest management planning standard (Version 4.1, April 2006). Alberta Sustainable Resource Development. EdmontonGoogle Scholar
  5. Alberta Sustainable Resource Development (ASRD) (2007) Alberta regeneration survey manual. Alberta Sustainable Resource Development, EdmontonGoogle Scholar
  6. Alberta Sustainable Resource Development (ASRD) (2008) Historical wildfire database. Accessed 1 March 2008
  7. Alberta Sustainable Resource Development (ASRD) (2012) Regeneration standard of Alberta, May 2012. Pub. No. T/268. Accessed 1 Sept 2012
  8. Albert-Green A, Dean CB, Martell DL, Woolford DG (2013) A methodology for investigating trends in changes in the timing of the fire season with applications to lightning-caused forest fires in Alberta and Ontario, Canada. Can J For Res 43:39–45CrossRefGoogle Scholar
  9. Anderson MJ (2001) A new method for non-parametric multivariate analysis of variance. Austral Ecol 26:32–46Google Scholar
  10. Andison DW, Kimmins JP (1999) Scaling up to understand British Columbia’s boreal mixedwoods. Environ Rev 7:19–30CrossRefGoogle Scholar
  11. Awada T, Henebry GM, Redmann RE, Sulistiyowati H (2004) Picea glauca dynamics and spatial pattern of seedlings regeneration along a chronosequence in the mixedwood section of the boreal forest. Ann For Sci 61:789–794CrossRefGoogle Scholar
  12. Bartemucci P, Coates KD, Harper KA, Wright EF (2002) Gap disturbances in northern old-growth forests of British Columbia, Canada. J Veg Sci 13:685–696CrossRefGoogle Scholar
  13. Bartos DL, Mueggler WF (1981) Early succession in aspen communities following fire in western Wyoming. J Range Manag 34:315–318CrossRefGoogle Scholar
  14. Beckingham JD, Archibald JH (1996) Field guide to the ecosites of northern Alberta. Canadian Forest Service, Northern Forestry Centre, special report 5, EdmontonGoogle Scholar
  15. Beckingham JD, Corns IGW, Archibald JH (1996) Field guide to ecosites of west-central Alberta. Canadian Forest Service, Northern Forestry Centre, special report 9, EdmontonGoogle Scholar
  16. Bergeron Y (2000) Species and stand dynamics in the mixed woods of Quebec’s southern boreal forest. Ecology 81(6):1500–1516CrossRefGoogle Scholar
  17. Bergeron Y, Harvey B (1997) Basing silviculture on natural ecosystem dynamics: an approach applied to the southern boreal mixedwood forest of Quebec. For Ecol Manag 92:235–242CrossRefGoogle Scholar
  18. Bergeron Y, Chen HYH, Kenkel NC, Leduc AL, Macdonald SE (2014) Boreal mixedwood stand dynamics: ecological processes underlying multiple successional pathways. For Chronic (in press)Google Scholar
  19. Bokalo M, Comeau PG, Titus ST (2007) Early development of tended mixtures of aspen and spruce in western Canadian boreal forests. For Ecol Manag 242:175–184CrossRefGoogle Scholar
  20. Brassard BW, Chen HYH, Wang JR, Duinker PN (2008) Effects of time since stand-replacing fire and overstory composition on live-tree structural diversity in the boreal forest of central Canada. Can J For Res 38:52–62CrossRefGoogle Scholar
  21. Bridge SRJ, Johnson EA (2000) Geomorphic principles of terrain organization and vegetation gradients. J Veg Sci 11:57–70CrossRefGoogle Scholar
  22. Brown JK, Debyle NV (1987) Fire damage, mortality, and suckering in aspen. Can J For Res 17:1100–1109CrossRefGoogle Scholar
  23. Burton PJ, Parisien M, Hicke JA, Hall RJ, Freeburn JT (2008) Large fires as agents of ecological diversity in the North American boreal forest. Int J Wildland Fire 17:754–767CrossRefGoogle Scholar
  24. Charron I, Greene DF (2002) Post-wildfire seedbeds and tree establishment in the southern mixedwood boreal forest. Can J For Res 32:1607–1615CrossRefGoogle Scholar
  25. Chen HYH, Popadiouk RV (2002) Dynamics of North American boreal mixedwoods. Environ Rev 10:137–166CrossRefGoogle Scholar
  26. Chen HYH, Vasiliauskas S, Kayahara GJ, Ilisson T (2009) Wildfire promotes broadleaves and species mixture in boreal forest. For Ecol Manag 257:343–350CrossRefGoogle Scholar
  27. DesRochers A, Lieffers VJ (2001) Root biomass of regenerating aspen (Populus tremuloides) stands of different densities in Alberta. Can J For Res 31:1012–1018CrossRefGoogle Scholar
  28. Fraser E, Landhäusser S, Lieffers V (2004) The effect of fire severity and salvage logging traffic on regeneration and early growth of aspen suckers in north-central Alberta. For Chron 80:251–256Google Scholar
  29. Frey BR, Lieffers VJ, Landhäusser SM, Comeau PG, Greenway KJ (2003) An analysis of sucker regeneration of trembling aspen. Can J For Res 33:1169–1179CrossRefGoogle Scholar
  30. Galipeau C, Kneeshaw D, Bergeron Y (1997) White spruce and balsam fir colonization of a site in the southeastern boreal forest as observed 68 years after fire. Can J For Res 27:139–147CrossRefGoogle Scholar
  31. Gärtner SM, Lieffers VJ, Macdonald SE (2011) Ecology and management of natural regeneration of white spruce in the boreal forest. Environ Rev 19:461–478CrossRefGoogle Scholar
  32. Greene DF, Johnson EA (1999) Modelling recruitment of Populus tremuloides, Pinus banksiana, and Picea mariana following fire in the mixedwood boreal forest. Can J For Res 29:462–473CrossRefGoogle Scholar
  33. Greene DF, Zasada JC, Sirois L, Kneeshaw D, Morin H, Charron I, Simard M (1999) A review of the regeneration dynamics of North American boreal forest tree species. Can J For Res 29:824–839CrossRefGoogle Scholar
  34. Greene DF, Noël J, Bergeron Y, Rousseau M, Gauthier S (2004) Recruitment of Picea mariana, Pinus banksiana, and Populus tremuloides across a burn severity gradient following wildfire in the southern boreal forest of Quebec. Can J For Res 34:1845–1857CrossRefGoogle Scholar
  35. Greene DF, Macdonald SE, Cumming S, Swift L (2005) Seedbed variation from the interior through the edge of a large wildfire in Alberta. Can J For Res 35:1640–1647CrossRefGoogle Scholar
  36. Greenway KJ, Lieffers VJ (1997) A boreal forest grass with an open meadow photosynthetic strategy. Can J Bot 75:562–567CrossRefGoogle Scholar
  37. Hothorn T, Hornik K, Zeileis A (2006a) Unbiased recursive partitioning: a conditional inference framework. J Comput Graph Stat 15:651–674CrossRefGoogle Scholar
  38. Hothorn T, Hornik K, Van De Wiel MA, Zeileis A (2006b) A lego system for conditional inference. Am Stat 60:257–263CrossRefGoogle Scholar
  39. Hothorn T, Hornik K, Strobl C, Zeileis A (2010) Party: a laboratory for recursive part(y)itioning. R package version 0.9-9995 Accessed 1 April 2011
  40. Ilisson T, Chen HYH (2009a) Response of six boreal tree species to stand replacing fire and clearcutting. Ecosystems 12:820–829CrossRefGoogle Scholar
  41. Ilisson T, Chen HYH (2009b) The direct regeneration hypothesis in northern forests. J Veg Sci 20:735–744CrossRefGoogle Scholar
  42. Johnson EA, Miyanishi K, Weir JMH (1998) Wildfires in the western Canadian boreal forest: landscape patterns and ecosystem management. J Veg Sci 9:603–610CrossRefGoogle Scholar
  43. Johnson EA, Miyanishi K, O’Brien N (1999) Long-term reconstruction of the fire season in the mixedwood boreal forest of western Canada. Can J Bot 77:1185–1188Google Scholar
  44. Johnstone JF, Chapin FS III, Foote J, Kemmett S, Price K, Viereck L (2004) Decadal observations of tree regeneration following fire in boreal forests. Can J For Res 34:267–273CrossRefGoogle Scholar
  45. Landhäusser SM, Lieffers VJ (1998) Growth of Populus tremuloides in association with Calamagrostis canadensis. Can J For Res 28:396–401CrossRefGoogle Scholar
  46. Landhäusser SM, Silins U, Lieffers VJ, Liu W (2003) Response of Populus tremuloides, Populus balsamifera, Betula papyrifera and Picea glauca seedlings to low soil temperature and water-logged soil conditions. Scand J For Res 18:391–400CrossRefGoogle Scholar
  47. Landhäusser SM, Wan X, Lieffers VJ, Chow PS (2010) Nitrate stimulates root suckering in trembling aspen (Populus tremuloides). Can J For Res 40:1962–1969CrossRefGoogle Scholar
  48. Lavoie L, Sirois L (1998) Vegetation changes caused by recent fires in the northern boreal forest of eastern Canada. J Veg Sci 9:483–492. doi:10.2307/3237263 CrossRefGoogle Scholar
  49. Lieffers VJ, Stadt KJ (1994) Growth of understory Picea glauca, Calamagrostis canadensis and Epilobium angustifolium in relation to overstory light transmission. Can J For Res 24:1193–1198CrossRefGoogle Scholar
  50. Lieffers VJ, Macdonald SE, Hogg EH (1993) Ecology of and control strategies for Calamagrostis canadensis in boreal forest sites. Can J For Res 23:2070–2077CrossRefGoogle Scholar
  51. Lieffers VJ, Stadt KJ, Navratil S (1996) Age structure and growth of understory white spruce under aspen. Can J For Res 26:1002–1007CrossRefGoogle Scholar
  52. Lieffers VJ, Armstrong GW, Stadt KJ, Marenholtz EH (2008) Forest regeneration standards: are they limiting management options for Alberta’s boreal mixedwoods? For Chron 84:76–82Google Scholar
  53. MacIsaac DA, Comeau PG, Macdonald SE (2006) Dynamics of regeneration gaps following harvest of aspen stands. Can J For Res 36:1818–1833CrossRefGoogle Scholar
  54. Man CD, Comeau PG, Pitt DG (2008) Competitive effects of woody and herbaceous vegetation in a young boreal mixedwood stand. Can J For Res 38:1817–1828CrossRefGoogle Scholar
  55. Martin-DeMoor J, Lieffers VJ, Macdonald SE (2010) Natural regeneration of white spruce in aspen dominated boreal mixedwoods following harvesting. Can J For Res 40:585–594CrossRefGoogle Scholar
  56. McCune B, Grace JB, Urban DL (2002) Analysis of ecological communities. MjM Software Design, Gleneden BeachGoogle Scholar
  57. Miyanishi K, Johnson EA (2002) Process and patterns of duff consumption in the mixedwood boreal forest. Can J For Res 32:1285–1295CrossRefGoogle Scholar
  58. Moore JM, Wein RW (1977) Viable seed populations by soil depth and potential site recolonization after disturbance. Can J Bot 55:2408–2412CrossRefGoogle Scholar
  59. Natural Regions Committee (2006) Natural regions and subregions of Alberta. Compiled by D. J. Downing and W. W. Pettapiece. Government of Alberta. Pub. No. T/852Google Scholar
  60. Oksanen J, Blanchet GF, Kindt R, Legendre P, O’Hara RB, Simpson GL, Solymos P, Henry M, Steven H, Wagner H (2010) Vegan: community ecology package. R package version 1.17-3. Accessed 1 April 2011
  61. Peters VS, Macdonald SE, Dale MRT (2005) The interaction between masting and fire is key to white spruce regeneration. Ecology 86:1744–1750CrossRefGoogle Scholar
  62. Peters VS, Macdonald SE, Dale MRT (2006) Patterns of initial versus delayed regeneration of white spruce in boreal mixedwood succession. Can J For Res 36:1597–1609CrossRefGoogle Scholar
  63. Purdy BG, Macdonald SE, Dale MRT (2002) The regeneration niche of white spruce following fire in the mixedwood boreal forest. Silva Fennica 36:289–306Google Scholar
  64. Shropshire C, Wagner RG, Bell WF, Swanton CJ (2001) Light attenuation by early successional plants of the boreal forest. Can J For Res 31:812–823CrossRefGoogle Scholar
  65. Simard M, Bergeron Y, Sirois L (1998) Conifer seedling recruitment in a southeastern Canadian boreal forest: the importance of substrate. J Veg Sci 9:575–582CrossRefGoogle Scholar
  66. Stewart JD, Hogg EH, Hurdle PA, Stadt KJ, Tollestrup P, Lieffers VJ (1998) Dispersal of white spruce seed in mature aspen stands. Can J Bot 76:181–188Google Scholar
  67. Stewart JD, Landhäusser SM, Stadt KJ, Lieffers VJ (2001) Predicting natural regeneration of white spruce in boreal mixedwood understories. For Chron 77:1006–1013Google Scholar
  68. Taylor AR, Chen HYH (2011) Multiple successional pathways of boreal forest stands in central Canada. Ecography 34:208–219CrossRefGoogle Scholar
  69. WAS—West Fraser Mills Ltd, ANC Timber Ltd, Sundance Forest Industries, Alberta-Pacific Forest Industries (2008) Regeneration Survey Manual. Version 2.0, May 2008Google Scholar
  70. Weir JMH, Johnson EA, Miyanishi K (2000) Fire frequency and the spatial age mosaic of the mixed-wood boreal forest in western Canada. Ecol Appl 10:1162–1177CrossRefGoogle Scholar
  71. Yih K, Boucher DH, Vandermeer JH, Zamora N (1991) Recovery of the rain forest of Southeastern Nicaragua after destruction by Hurricane Joan. Biotropica 23:106–113CrossRefGoogle Scholar

Copyright information

© Springer Science+Business Media Dordrecht 2014

Authors and Affiliations

  • Stefanie M. Gärtner
    • 1
  • Mike Bokalo
    • 2
  • S. Ellen Macdonald
    • 2
  • Ken Stadt
    • 3
  1. 1.Chair of Site Classification and Vegetation Sciences, Faculty of Environment and Natural ResourcesUniversity of FreiburgFreiburgGermany
  2. 2.Department of Renewable ResourcesUniversity of AlbertaEdmontonCanada
  3. 3.Alberta Environment and Sustainable Resource DevelopmentEdmontonCanada

Personalised recommendations