Content uploaded by Mathieu Bouchard
Author content
All content in this area was uploaded by Mathieu Bouchard on Apr 15, 2019
Content may be subject to copyright.
Forest Ecology and Management 261 (2011) 811–820
Contents lists available at ScienceDirect
Forest Ecology and Management
journal homepage: www.elsevier.com/locate/foreco
Long-term influence of fire and harvesting on boreal forest age structure and
forest composition in eastern Québec
Mathieu Bouchard∗, David Pothier
Center for Forest Research (CEF), Department of Forest Sciences, Laval University, Québec, QC, Canada G1K 7P4
article info
Article history:
Received 3 August 2010
Received in revised form
17 November 2010
Accepted 19 November 2010
Available online 3 January 2011
Keywords:
Natural disturbances
Clearcutting
Forest composition
Natural range of variability
Boreal forests
abstract
In boreal forests, historical variations in the area disturbed by natural disturbances or harvesting have
rarely been compared. We measured temporal and spatial variations in areas affected by severe fires and
clearcutting throughout the 20th century in a 57, 332 km2section of the eastern Canadian boreal forest.
We examined the effects of these disturbances on spatio-temporal variations in the abundance of forests
>60 years. Natural variability for the abundance of forests >60 years was estimated from simulations of
natural disturbance regimes. We also measured compositional and structural differences between three
categories of stands originating from relatively recent disturbances (∼50 years; clearcutting, fires, and
clearcutting followed by fires), and one category of stands that were undisturbed for at least 200 years.
At the regional level, we observed that forests >60 years gradually became scarcer throughout the 20th
century due to a gradual expansion of harvested areas, an effect most pronounced in the southern part
of the region, where mature and old forest abundance was clearly outside the range of natural variability
at the end of the studied period. At the stand level, forest composition and structure differed between
stand-origin categories: clearcutting-origin stands contained more balsam fir (Abies balsamea), fire-origin
stands more black spruce (Picea mariana), and fire/clearcutting-origin stands more hardwoods (Betula
papyrifera and Populus tremuloides). Overall, we estimate that strict forest management targets based on
natural disturbance regimes will be difficult to achieve in eastern North-American boreal forests, most
notably because contemporary disturbance rates, including both clearcutting and fire, have gradually
become higher than the fire rates observed during the preindustrial period.
© 2010 Elsevier B.V. All rights reserved.
1. Introduction
The gradual conversion of virgin boreal forests to managed
forests is still an ongoing process in much of the North Ameri-
can boreal forest (Kurz and Apps, 1999). During this conversion,
regenerating stands originating from clearcutting increase in abun-
dance, and older forest habitats decline in abundance and tend to
become increasingly fragmented (Östlund et al., 1997; Schroeder
and Perera, 2002; Fahrig, 2003). Dominant forest composition has
also been reported to change during this transition in boreal forests,
with highly competitive species well adapted to disturbed environ-
ments generally becoming more abundant (Östlund et al., 1997;
Haeussler and Kneeshaw, 2003; Boucher et al., 2009).
The amount of mature and old forests is a useful regional-scale
indicator of changes in forest age class structure (Didion et al.,
2007). In boreal forests, these older forests have attributes such as
large diameter trees and coarse woody debris, which can be impor-
∗Corresponding author.
E-mail addresses: mathieu bouchard@yahoo.ca (M. Bouchard),
David.Pothier@sbf.ulaval.ca (D. Pothier).
tant for many kinds of specialized organisms. For a given region, the
area covered by older forests is directly and negatively related to the
frequency of severe natural disturbances, and particularly stand-
replacing fires (Johnson, 1992; Wimberly et al., 2004; Bergeron
et al., 2006). In regions where frequency of stand-replacing fires
is high, the amount of old forests is low, and it is thus realistic to
assume that forest management can maintain a forest age-class
structure that is relatively close to what occurs in nature, particu-
larly if fire suppression is efficient (Bergeron et al., 2006). However,
in regions where severe fire return intervals are relatively long, for
instance in humid coastal areas, it will be harder to implement for-
est management policies based on the use of even-aged silvicultural
systems, while at the same time maintaining a relatively natural
forest age class structure (Cissel et al., 1999; Wimberly et al., 2004;
Nonaka and Spies, 2005). This difficulty can be compounded by
the fact that, in many regions, natural disturbances are not effi-
ciently suppressed, and continue to occur in juxtaposition with
forest management practices such as clearcutting (Kurz and Apps,
1999; Armstrong, 1999; Bergeron et al., 2006; Girard et al., 2009).
The forest age class structure, measured at the regional or land-
scape level, is not the only forest attribute that changes during
the transition from natural to managed forests. At the stand level,
0378-1127/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.foreco.2010.11.020
812 M. Bouchard, D. Pothier / Forest Ecology and Management 261 (2011) 811–820
forest characteristics for a given age class may vary depending
on the type of disturbance from which they originate. In boreal
forests, it has been observed that stands originating from clearcut-
ting are different in terms of composition and structure compared
with stands originating from natural disturbances (reviewed by
Haeussler and Kneeshaw, 2003). Different compositional or struc-
tural attributes in post-clearcutting stands compared with post-fire
stands can accumulate at the regional scale, and eventually have
impacts on biodiversity (Carey and Johnson, 1995; Drapeau et al.,
2000), ecosystem processes such as carbon dynamics (Luyssaert
et al., 2008; Rhemtulla et al., 2009) or disturbances agents such as
diseases or insect outbreaks (McCullough et al., 1998).
In regions where natural forest characteristics, including forest
age class structure or forest composition, have not been eliminated
yet, the gap between current conditions and pre-industrial condi-
tions can be evaluated, in order to propose corrective measures if
necessary. In the present study, we use a 57,332 km2area located in
the eastern Canadian boreal forest to look at the cumulative effect
of natural and anthropogenic disturbances on forest structure and
composition throughout the 20th century. We also used simulation
modelling based on natural disturbance regimes to quantify the
range of natural variability (RNV; Landres et al., 1999) for mature
and old forests, an approach that was often advocated in previous
studies in boreal forests (Keane et al., 2004). For the purpose of
this research, we focused on stand-replacing disturbances, which
we define operationally as disturbances that cause >75% canopy
mortality at the stand level. The three main objectives of this study
are: (1) to use simulation studies to define a RNV for the abundance
of mature and old forests at the regional level, (2) to examine, by
using mature and old forest abundance as an indicator, whether
the interaction between the two most significant disturbances
— fire and clearcutting — moved the ecosystem outside the RNV
during the 20th century, and (3) to examine the effect of different
disturbance types on forest composition at the stand level.
2. Methods
2.1. Study area
Our study area is a 57,332 km2portion of the boreal forest
located in the North Shore region of eastern Québec (Fig. 1). For-
est composition in the region is dominated by black spruce (Picea
mariana) and, to a lesser extent, by balsam fir (Abies balsamea).
Species such as jack pine (Pinus banksiana), trembling aspen (Popu-
lus tremuloides) and white birch (Betula papyrifera) are also present
as relatively minor species in the landscape, but sometimes dom-
inate in recently disturbed environments. In natural conditions,
balsam fir tends to be more abundant in stands that have not been
affected by fire for more than 100 years (Bouchard et al., 2008).
Fires are by far the most frequent severe natural disturbance in
this region (cycle of 250–600 years; Bouchard et al., 2008). Fire fre-
quency varies according to an east-west gradient, with the eastern
parts having mean fire return intervals estimated at 600 years, and
the eastern part 250 years. The other types of natural disturbances
are less important. Stand-replacing windthrows (severe mortal-
ity over areas of more than 5 ha) affected a mean of 0.0255% of
the area per year, corresponding to a cycle of about 3900 years
(Bouchard et al., 2009). Severe spruce budworm (SBW) outbreaks,
the major defoliator insect in the region, are not very frequent or
extensive when examined from a regional perspective: one severe
outbreak occurred in the southern part of the territory at the end of
the 20th century, corresponding to a recurrence of stand-replacing
mortality of about 9200 years based on the last century (Bouchard
and Pothier, 2010). Fire is the only natural disturbance that has
the ability to create large mortality patches (several tens of km2)
throughout the region, as stands killed by the last SBW outbreaks
were mostly confined to the southern part of the territory and
windthrow patches are much smaller (Fig. 2). For this reason, we
chose to focus on fire for our comparison with clearcutting and for
the landscape simulations establishing the RNV.
Although some partial cuts (defined as removing less than 75%
of the original dominant canopy cover) have occurred in the study
area since before 1900, the impact of forest harvesting increased
drastically after the advent in the early 1920s of clearcutting, a
technique that was implemented in order to provide wood for the
booming pulp and paper industry (Frenette, 1996). Early clearcut-
ting mostly occurred in the southern part of the territory, especially
along the rivers used for floating logs to the mills, and gradually
expanded northwards and away from the main rivers with the pro-
gressive development of the road network (especially after 1950).
The limit of forest management units was also moved progressively
northwards during the same period, but as of 2000, the northern
Fig. 1. Location of the study area and large-scale distribution of clearcuttings during the 20th century.
M. Bouchard, D. Pothier / Forest Ecology and Management 261 (2011) 811–820 813
Fig. 2. Spatial pattern of stand-replacing mortality associated with the three types of natural disturbance that are most common in the Quebec North Shore region, for
the 1900–2000 period. (a) For fires, shaded areas represent fires for the 1900–2000 period. (b) For windthrows, shaded areas represent major windstorm tracks during the
1900–1970 period (stand mortality >20%), and dots represent stand-replacing windthrows during the 1971–2000 period. (c) For spruce budworm outbreaks, dots and black
areas represent severe mortality during the 1974–1990 outbreak (earlier outbreaks were not severe).
and eastern fringes of the study area still had not been included in
any active forest management units.
For the purpose of this study, all the clearcutting that occurred
between 1920 and 2000 was located using available historical
information, which includes harvesting maps found in forest com-
pany archives, aerial photographs taken at various points in time
(between 1930 and 2000; details available in Bouchard et al.
(2008)), and dendrochronological surveys performed in the field
to confirm the year at which harvesting took place. We consid-
ered a stand to be clearcut when the harvesting operation left
merchantable trees (>9 cm DBH) providing less than an estimated
25% of the canopy cover after the cut. In all subsequent analyses,
all stands originating from clearcutting were grouped into 10-year
categories according to the date of harvest.
Similarly, all stands originating from fires that occurred between
1800 and 2000 were identified by using the fire map developed
by Bouchard et al. (2008) for this region. This maps contains all
fires >10 km2, which are estimated to represent about 98% of the
burnt area in this territory (Bouchard et al., 2008). A significant
proportion of the stands logged in the early 20th century were
affected by low to moderate severity fires in the subsequent years.
This phenomenon was probably due mostly to the abundance of
dry combustibles left behind after harvesting, an increased igni-
tion probability associated with the logging operations carried
out nearby, and a limited technical capacity to efficiently con-
trol these fires (Boucher et al., 2006). Areas in which clearcutting
was followed by fires were identified as such in the cartographic
database.
2.2. Range of natural variation
We used simulation analyses to estimate an RNV based on key
characteristics of the 1800–2000 fire regime. Direct (ignition, sup-
pression) or indirect (abundance and spatial distribution of forest
fuels) human influence on the fire regime may have been present
in the southern part of the territory towards the later part of the
period, but because these impacts are uncertain, and because tem-
poral variations in fire frequencies appeared to be driven mainly
by climatic variations during this period (Girardin, 2010), we con-
sider this regime as “preindustrial”. Details about the rationale for
the fire model used in this study, which was developed specifically
for the territory, are contained in Bouchard and Pothier (2008). The
important features of this model are summarized in the present
section.
The fire model is based on a statistical analysis of the severe
fires that occurred in the region after 1800, and thus the RNV
reported in this study essentially reflects conditions during that
period. This model is based on two major parameters of the fire
regime: a fire size distribution and a mean fire return interval.
The fire size distribution is derived empirically from the dataset
of Bouchard et al. (2008), and corresponds to a reverse Weibull dis-
tribution (Bouchard and Pothier, 2008). For each fire event, fire size
was determined from a reverse Weibull distribution according to
the following equation:
x=m+b(−ln(U)) 1
c(1)
where xrepresents fire size (in km2), and Uis a number drawn from
a uniform random distribution between 0 and 1. Parameters band
cwere obtained by fitting the function to the observed fire size
distribution, and correspond respectively to b= 25.9 and c= 0.7357.
Parameter mcorresponds to the minimal fire size considered in the
simulations, 10 km2. Fires larger than 2100 km2(maximal fire size
observed in the study area) were eliminated from the simulations;
truncating the fire size distribution in this manner is important in
order to generate realistic fire size distributions (Cumming, 2001;
Bouchard and Pothier, 2008).
All simulations were realized on a landscape with the spatial
characteristics of the study area in terms of size, shape and local-
ization of water bodies. This landscape comprised 31,864 square
cells (1 km2) in the western section of the territory, and 25,468
cells in the eastern section (W and E of 68◦W, respectively). Due
to different fire frequencies in these two sections (250 years for
the west, and 600 years for the east; Bouchard et al., 2008), the
number of fires per year necessary to obtain the desired fire return
intervals was determined by iteration, with fires ignited in either
section allowed to cross the boundary between the two sections.
Within each of the two regions, we assumed conservatively that
fire ignition and spreading occurred randomly: information from
North American boreal forests suggests that most of the burned
area is caused by extreme climatic conditions (droughts, lightning
occurence, etc.) and not local-scale factors (Johnson, 1992; Girardin
and Mudelsee, 2008). The only exception were large water bodies,
which could not be crossed directly by simulated fires. The number
814 M. Bouchard, D. Pothier / Forest Ecology and Management 261 (2011) 811–820
of fires per year was picked randomly from a Poisson distribution.
The number of fires per year necessary to achieve the target mean
fire return intervals was 0.28 and 0.83 for the eastern and western
sections, respectively.
All simulations were conducted with the SELES platform (Fall
and Fall, 2001). For each scenario, 500 simulations were run, and the
quantity of mature forests present at the last year of each simulation
was used to construct a 95% confidence envelope for the proportion
of forests >60 years. The 60-year value was taken from studies indi-
cating that key attributes of old-growth forests, such as large living
and dead trees, start to develop after this age in eastern balsam fir
dominated forests (Sturtevant et al., 1997; Desponts et al., 2004).
Because the study area is very large and clearcuttings tended to be
concentrated in the south, two RNVs were compilated on two spa-
tial units: the first for the whole study area of 57,332 km2, and the
second for a subsection of 12,147 km2located south of 50◦N.
2.3. Abundance of forest-interior habitat throughout the 20th
century
We reconstructed the cumulative effect of clearcutting and fires
on the quantity of stands with forest-interior characteristics for the
20th century. We defined a forest-interior habitat as a cell, occu-
pied by forests >60 years, that is located more than 1 km from any
cell with forests <60 years. The 1 km distance is convenient con-
sidering the precision of the data (notably the fire map) and the
considerable size of the study area, and should be considered as an
intermediate distance for edge effects observed for various organ-
isms in eastern Canadian boreal forests. For example, while some
old-growth species needing particular microclimatic conditions are
affected by edge effects on the order of 0–100 m (Harper et al.,
2005; Gagné et al., 2007; Boudreault et al., 2008), species such as
the woodland caribou that are affected by generalist predators (or
humans), which are more abundant in clearcutting-origin environ-
ments, may need much larger buffer zones around their core habitat
(several km’s; Courtois et al., 2008). The 95% confidence intervals
for interior forest habitat were constructed using the same simula-
tion analyses that were used to calculate the abundance of mature
and old forests, but by subtracting edge habitats (all cells adjacent
to recently disturbed cells).
2.4. Stand characteristics 50 years after disturbance
We examined the effect of different disturbance types on two
forest attributes: species composition and the abundance of large
stems (>21 cm DBH in the context of this region; Aakala et al.,
2007). These two components are of major importance for wildlife
and biodiversity in boreal forests, including birds such as wood-
peckers, mustelids, and saproxylic organisms including insects and
fungi (Nilsson et al., 2001; Higdon et al., 2006; Gagné et al., 2007).
The presence of large stems also suggests that conditions are ade-
quate for a continuous temporal recruitment of large dead trees, an
important feature of old-growth stands (Aakala et al., 2007).
The basic dataset consisted of circular (400 m2) plots that were
established across the area for forest inventory purposes by the
Quebec Ministry of Natural Resources between 1971 and 2000. The
advantage of this dataset is that it comprises a very high number of
plots, which are well distributed across the territory according to a
random sampling design. In every plot, all trees (>9 cm DBH) were
measured by species and diameter (see Bouchard et al., 2008 for
further details). Basic information about topography and surficial
deposits for each plot were retrieved from soil maps of the study
area.
The plots were overlaid on the clearcutting and fire maps of
this territory to examine forest composition in four stand origin
categories. These stand origins were fire only, clearcutting only,
clearcutting immediately followed by fire, and forests that had not
burned for more than 200 years (old-growth unaffected by partial
cutting or clearcutting). For the first three categories, we restricted
our sampling to disturbances that occurred between 1920 and
1950, so that all sample plots came from the same age-class (50
years). Forests in this age group constituted the best compromise
in terms of sample size in the different stand-origin categories, and
in terms of representativeness in a managed forest mosaic where
stands are usually harvested between age 60 and 100 under cur-
rent forest management practices. Plots belonging to the >200 years
category were included in the analysis to verify how different the
three previous stand categories were from old-growth conditions.
In order to minimize potential differences due to the north-south
climatic gradient, we limited sampling of all stand-origin categories
to a sub-region located south of 50◦30N, where all stand-origin
categories were present. Nonetheless, the few fires that occurred
during the 1920–1950 period tended to be located more to the
north, and as a result of the fire-only plots are in average 94 km to
the north compared with clearcuttings, 80 km compared with fire
superimposed with clearcutting, and 75 km compared with undis-
turbed stands. The potential implications of this latitudinal gradient
are dealt with in Section 4.
To minimize the effect of abiotic characteristics and simplify
the interpretation of compositional differences between stand ori-
gin categories, only plots that were located on relatively deep
tills (>50 cm) with mesic drainage conditions were kept in the
final dataset. Other types of materials such as organic soils, coarse
textured fluvio-glacial deposits, or relatively shallow soils over
bedrock were more rare in the study area, and were excluded
from the analyses because they were not represented by a suffi-
cient number of plots to realize proper statistical analyses. Overall,
these relatively deep tills represent 67% of the study area. For each
selected plot, time since disturbance was determined as the time
elapsed between disturbance occurrence and plot measurement.
Stand characteristics before disturbance could not be estimated
precisely at the plot level, but based on the available historical infor-
mation (including the fire information for the 1800–1900 period
[Bouchard et al., 2008]), we assume that the vast majority of these
stands contained forests more than 100 years old when the distur-
bance occurred.
Overall, once plots located in stands of unknown origin or stands
with deposits other than relatively deep tills were removed from
the dataset, there remained 180 forest inventory plots that were
located in stands originating from fires, 63 plots in stands origi-
nating from clearcutting, and 79 plots in stands originating from
clearcutting followed by fires. In addition, a total of 128 plots
located in stands that had not been disturbed for more than 200
years either by fire or harvesting were used as controls to examine
stand characteristics in undisturbed conditions (Table 1).
The structure and composition of the different stand-origin cat-
egories were examined with several statistical analyses. An ANOVA
was performed to examine differences between stand-origin cat-
egories in terms of structure (total basal area and density of
stems >21 cm DBH). When the F-test of the ANOVA was significant
(P< 0.05), least squares means (LS-means) tests were used to deter-
mine significant differences (P< 0.05) between categories. Because
the data are unbalanced, we used the type II method to calculate
sums of squares in the ANOVA (Shaw and Mitchell-Olds, 1993).
Because the residuals of large stem density were not normally dis-
tributed, we performed the ANOVA on ranks for this variable.
We used multivariate analyses to examine whether stand-origin
categories were associated with different tree species compo-
sition. For each of the 450 plots, basal area was chosen as a
measure of species abundance. First, nonmetric multidimensional
scaling (NMDS) was used to represent the plots from different
stand-origin categories in ordination space. NMDS is a robust
M. Bouchard, D. Pothier / Forest Ecology and Management 261 (2011) 811–820 815
Table 1
Composition (in basal area and large stem density) of stands originating from clearcutting alone, severe crown fires alone, or clearcutting followed by fires of variable severity,
assessed from temporary inventory plots measured between 1971 and 1999.
Clearcutting alone
(n= 63)
Fire alone (n= 180) Clearcutting + fire
(n= 79)
Undisturbed
(n= 128)
Mean SD Mean SD Mean SD Mean SD
Time since last disturbance (years) 50.2 5.9 55.8 8.9 46.9a3.2 >200 –
Basal area White birch 3.4 4.3 2.3 3.6 2.1 3.5 1.5 2.8
White spruce 1.3 2.1 0.1 0.3 0.4 0.7 1.0 2.3
Black spruce 3.1 5.5 11.4 8.5 4.3 5.7 10.8 9.0
Trembling aspen 0.4 1.7 1.5 3.7 5.6 6.1 0.5 2.8
Jack pine 0.1 0.7 0.7 2.9 0.1 0.7 0.0 0.3
Balsam fir 11.2 10.1 1.3 3.8 2.5 4.8 10.4 8.3
Other species 0.3 0.6 0.2 0.7 0.1 0.3 0.0 0.2
Total 19.8 8.8 17.5 9.6 15.0 7.2 24.2 8.1
Stems/ha > 21 cm DBH White birch 14 36 5 19 5 20 17 41
White spruce 16 31 0 2 3 11 12 29
Black spruce 17 33 22 43 12 49 139 141
Trembling aspen 3 10 12 37 23 47 6 32
Jack pine 1 5 4 21 1 6 0 3
Balsam fir 78 96 7 29 16 46 107 116
Other species 0 0 0 2 0 3 0 2
Total 128 106 50 77 60 84 281 151
aIn the case of the clearcutting +fire category, time since last disturbance corresponds to the fire date. Post-clearcutting fires were generally not very severe, and trees
have often survived through them.
nonparametric dimension reduction technique commonly used in
ecology (Legendre and Legendre, 1998). We used the isoMDS func-
tion in the MASS package (Venables and Ripley, 2002) with R (R
Development Core Team, 2008). The ordination was applied on a
dissimilarity matrix built with the Bray–Curtis distance metric. We
performed this analysis by using two ordination axes (k= 2; stress
factor = 17.12).
We used the multi-response permutation procedure (MRPP) in
the vegan package (Oksanen et al., 2007) with R to test whether
there is a significant difference between stand-origin categories
based on their species composition. The test provides a P-value,
which is a probability measure of significant differences between
groups, and an A-value for within-group agreement, ranking
from 1 (total within-group agreement) to 0 (heterogeneity within
groups equals expectations from chance) (McCune and Mefford,
1999).
Finally, we used indicator species analysis (duleg function in
the labdsv package; Roberts, 2007) to identify tree species that
are significantly associated with one stand-origin category or
another. This test compares the relative frequency of occurrence
and abundance of tree species in different stand-origin categories
and identifies tree species that vary more between stand-origin
categories than would be expected by chance (Legendre and
Legendre, 1998). The statistical significance of each indicator
value was determined by a Monte Carlo test using 999 permuta-
tions.
3. Results
3.1. Harvesting history
Decadal rates of clearcutting progressively increased through-
out the 20th century (Fig. 3). Spatially, clearcutting had a relatively
contagious spatial pattern at the regional scale, with a gradual
expansion from south to north (Fig. 1). This contrasts with the spa-
tial arrangement of fires, which tended to be relatively dispersed
throughout the entire study area (Fig. 2).
200019801960194019201900
Disturbed area (ha)
0
50000
100000
150000
200000
250000
Year of origin
200019801960194019201900
Area disturbed by clearcutting Area disturbed by fire
ba
Fig. 3. Temporal patterns of occurrence for clearcutting and fire during the 20th century. Area disturbed by clearcutting (a) and fire (b).
816 M. Bouchard, D. Pothier / Forest Ecology and Management 261 (2011) 811–820
Southern part of study area
200019801960194019201900
0
20
40
60
80
100
200019801960194019201900
Abundance of forests > 60 years (%)
0
20
40
60
80
100
Entire study area
Fig. 4. Observed (line) abundance of stands that have not been affected by fire or clearcutting for >60 years during the 20th century. Grey areas indicate the 95% confidence
interval under a natural fire regime.
3.2. Departure from the RNV throughout the 20th century
The 95% confidence interval (RNV) obtained from simulations
of the fire regime is indicated in Fig. 4 (shaded area), with the
black line corresponding to the actual amount of forests >60 years.
Results of the simulations indicate that the 95% C.I. for the amount
of forests >60 years is between 78% and 91% for the entire territory,
and between 64% and 96% when only the southern sub-region is
considered (Fig. 4). When only forests >60 years with forest-interior
conditions are considered, the range is between 74% and 90% for the
entire territory, and between 59% and 94% for the southern portion
(Fig. 5). The confidence intervals are wider for the southern sub-
region because the size of the reference area is smaller relative to
maximal fire size (Turner and Romme, 1994).
The observed amount of forests >60 years with or without
forest-interior conditions was measured for every decade of the
1900–2000 period (Figs. 4 and 5). For this analysis, stands affected
by disturbances other than clearcutting and fire, such as windthrow
and SBW outbreak, were considered undisturbed (see Section 2.1
for the rationale). The amount of forests that were not recently
disturbed by either clearcutting or fire drops below the RNV
towards the latter part of the 20th century, only slightly when
the whole territory is considered (Fig. 4), but much more when
the southern portion is considered (Fig. 4). The trends are sim-
ilar for forests >60 years with forest-interior conditions, which
represent 20% of the area by the last decade of the 20th cen-
tury – below the 95% confidence interval of the RNV (59–94%)
(Fig. 5).
3.3. Effect on forest characteristics at the stand scale
Composition of the different stand-origin categories, both
in terms of basal area and large stem density, are shown in
Table 1. ANOVA results indicate significant differences in terms
of basal area and abundance of large stems between stand-origin
categories (Fig. 6). Basal area and large stem density are gen-
erally higher in undisturbed stands (>200 years), followed by
stands of clearcutting origin (Fig. 6). Stands originating from
clearcutting + fire and from fire only exhibit similar basal areas
(Fig. 6). Large stems were more abundant in stands originating
from clearcutting than in those originating from the other two
types of recent disturbance (fire and clearcutting + fire) (Table 1).
Whereas stands originating from clearcutting are dominated by
balsam fir, stands originating from fire tend to be dominated
by black spruce, and stands originating from clearcutting + fire
are dominated by hardwoods and black spruce (Table 1). Undis-
turbed stands are co-dominated by balsam fir and black spruce
(Table 1).
The ordination plot of the results of the NMDS analysis indicates
significant heterogeneity in the dataset (Fig. 7), but stand-origin
categories are nonetheless loosely associated with particular
regions of the ordination space, as shown by the 50% ellipses
(Fig. 7). The MRPP analysis confirms that species composition varies
significantly among stand-origin categories when all species are
considered simultaneously (P< 0.0001, A= 0.1552). The indicator
species analysis indicates that several tree species are associated
with particular stand-origin categories (Table 2). Latitude and lon-
Southern part of study area
200019801960194019201900
0
20
40
60
80
100
Entire study area
200019801960194019201900
Forest-interior habitat abundance (%)
0
20
40
60
80
100
Fig. 5. Observed (line) abundance of mature (>60 years) forest stands with forest-interior conditions during the 20th century. Grey areas indicate the 95% confidence interval
under a natural fire regime.
M. Bouchard, D. Pothier / Forest Ecology and Management 261 (2011) 811–820 817
Stand origin category
UndisturbedCC+FFireClearcut
Basal area (m2/ha)
0
10
20
30
40
50
60
Stand origin category
UndisturbedCC+FFireClearcut
Stems > 21 cm dbh (n/ha)
0
200
400
600
800
a
c
bc
ba
c
c
b
Fig. 6. Boxplots showing mean basal area and mean large stem (>21 cm DBH) density for four stand-origin categories. The horizontal line in each box is the median. Boxes
enclose the 75th and 25th percentiles, and error bars enclose the 90th and 10th percentiles. Significant differences (P< 0.05) between stand-origin categories are indicated
by different letters above the boxes.
Table 2
Results of the indicator species analysis for tree species in four stand-origin
categories. Values >0.3 are interpreted as species highly associated with the cor-
responding category (Legendre and Legendre, 1998).
Tree species Clearcutting Fire CC + F Undisturbed
White birch 0.270 0.180 0.146 0.082
Black spruce 0.059 0.379 0.117 0.329
Balsam fir 0.420 0.025 0.084 0.376
Trembling aspen 0.004 0.073 0.553 0.003
Jack pine 0.007 0.120 0.002 0.000
White spruce 0.291 0.001 0.059 0.122
gitude were correlated with the first and second axes of the NMDS
using Pearson correlations, and latitude is strongly correlated with
the first axis of the NMDS (R2= 0.258; P< 0.0001).
4. Discussion
4.1. Changes in forest age class structure at the regional level
Several methods have been proposed in the scientific litera-
ture to model forest dynamics at the landscape level in reaction
to large-scale forest disturbances, whether anthropogenic or nat-
ural. Many such models were reviewed by Keane et al. (2004)
and He (2008), and several of them concern boreal or near-boreal
forests. The approach that was used in the present study could
be described as relatively simple compared with most of these
models (e.g., it does not incorporate forest succession), but it
nonetheless provided a robust first approximation of the RNV
based on two parameters obtained from empirical studies of fire
regimes in the study area: a mean fire return interval and a fire
size distribution (Bouchard and Pothier, 2008). This approach was
useful to calculate RNVs that could be used as targets for for-
est management in territories where severe fires are a dominant
disturbance type. Such a model can also be used to calculate a
RNV coherent with the size of the reference area, as the range
tends to be wider for smaller reference areas (provided that
the fire size distribution remains constant; Turner and Romme,
1994).
Under natural conditions, regions with long fire return inter-
vals such as the Quebec North Shore are characterized by a high
abundance of old-growth forests. During the 20th century, forest
management activities, combined with co-occurring fires, altered
this pattern of dominance by older forests, particularly in the
1st axis
0.80.60.40.20.0-0.2-0.4-0.6-0.8
2nd axis
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Clearcutting
Undisturbed
Clearcutting + fire
Fire Clearcutting + fire
Undisturbed
Clearcutting
Fire
Fig. 7. Scatter plot of the scores obtained from the NMDS analysis performed on tree species assemblages in forest inventory plots. 50% confidence ellipses are shown for
the four stand-origin categories.
818 M. Bouchard, D. Pothier / Forest Ecology and Management 261 (2011) 811–820
southern part of the territory, an effect that is more pronounced
when a 500-m edge effect is taken into account. Even though
these young forest types probably remain suitable habitats for a
range of species in boreal forests (Drapeau et al., 2000; Higdon
et al., 2006), the reduced abundance of older forests can be detri-
mental for species that depend specifically on attributes such
as large trees (dead and living) or heterogeneous canopy struc-
tures (Nilsson et al., 2001; Gagné et al., 2007). The different tree
species composition that was observed in these young forests
can also have an impact on biodiversity and various ecosystem
processes at the landscape scale. In the southern part of the ter-
ritory, for example, where these changes were more pronounced,
dwindling woodland caribou populations (Courtois et al., 2003)
may be partly attributed to an increased abundance of hard-
woods and fir trees that creates a more favourable habitat for
moose (Alces alces), which will in turn favour an increased abun-
dance of predators such as wolf (Canis lupus)(Courtois et al.,
2008). Increased road access to this territory is also associated
with increased hunting and predation pressures (Courtois et al.,
2008).
4.2. Usefulness of the RNV concept in regions with long fire return
intervals
The transition from pre-industrial to managed forest struc-
tures is still an ongoing process in much of the North American
boreal forest (Kurz and Apps, 1999). Our study clearly illustrates
that the additive impacts of different disturbance types regu-
late the abundance of old-growth stands during this phase of
the transition in the Quebec North Shore. Although the gradual
increase in clearcutting rates, caused by the progressive north-
wards expansion of government leases and forest management
activities (Frenette, 1996) is the most obvious explanation for
reduction in abundance of mature forest habitats in this region,
the departure from the RNV was accentuated by two additional
factors. First, large fires continued to occur up to the end of the
20th century despite fire protection efforts, meaning that the level
of “disturbance substitution” was relatively low. Second, the fire
cycle is naturally long in this region, and old-growth abundance
is thus naturally high, which means that even if fire suppression
was very efficient, current disturbance levels would be likely to
remain above pre-industrial levels of natural disturbances. For
example, whereas mean annual clearcutting rates are generally
around 1% in boreal forests under sustainable yield constraints, a
fire cycle of 250–600 years corresponds to mean annual distur-
bance rates of only 0.17–0.4%. Moreover, even if the abundance of
forests >60 years remains close to the RNV when the territory is
considered as a whole (Fig. 4, left pane), at least half of the ter-
ritory was not actively managed until the 1990s, and was thus
excluded from the annual allowable cut determination process. In
other words, the downward trend in the abundance of forests >60
years for the whole territory could continue during the post-2000
period because a larger proportion of the territory is now actively
managed.
Maintaining forests within the RNV can be much more difficult
to achieve in regions where natural disturbance return inter-
vals are long, a fact that has already been underlined for forest
ecosystems of the Pacific Northwest (Cissel et al., 1999; Wimberly
et al., 2004; Nonaka and Spies, 2005). In this context, the tools
usually proposed for the maintenance or restoration of forests
with pre-industrial characteristics, such as the modification of sil-
vicultural practices (Franklin et al., 2002; Harvey et al., 2002),
or the emulation of the spatial pattern of disturbance patches
(Mladenoff et al., 1993; Delong and Tanner, 1996) may be insuf-
ficient, unless they are accompanied by a decrease in the annual
rates of severe disturbance (i.e., clearcutting) compared with the
levels currently determined using standard annual allowable cut
calculations (Didion et al., 2007).
4.3. Effect of disturbance type on forest composition at the stand
level
Overall, the different compositions observed in the 50-year age-
class for different stand-origin categories were relatively consistent
with differences previously observed elsewhere in eastern boreal
forests (see for example Greene et al., 1999). In their review, Greene
et al. (1999) report that numerous studies showed that a rela-
tively high black spruce abundance after fire (Table 1 and Fig. 7)
can be explained by the destruction of advance regeneration by
fire, and by post-fire seeding from the aerial seed bank contained
in the semi-serotinous cones of this species. In our study, the
destruction of the seedling bank (advance regeneration) was also
probably much less extensive following clearcutting, given that in
the 1920–1950 period, harvesting operations generally took place
during winter time when the soils were frozen and covered with
snow. Because the seedling bank is very often dominated by bal-
sam fir in eastern boreal forests, including in the Quebec North
Shore region (De Grandpré et al., 2000; Pham et al., 2004), the pro-
tection of advance regeneration during logging is the most likely
explanation for relatively high balsam fir abundance in stands of
clearcutting origin. Finally, the high abundance of light seeded colo-
nizing species (trembling aspen, white birch) in stands successively
affected by clearcutting and fire might be explained by the additive
effects of a severe soil disturbance by fire combined with the pre-
vious removal of seed trees (Boucher et al., 2006; Etheridge et al.,
2006).
Separating the influence of latitude and disturbance type on
forest composition is difficult because the two factors are entan-
gled: fire-origin stands are located more to the north compared
with the other stand-origin categories. It is possible that climatic
differences associated with latitude accentuated the response of
some tree species following disturbance, specifically by inhibiting
the recruitment potential of balsam fir, white birch or aspen after
fire (Messaoud et al., 2007). However, because these tree species
are relatively widespread in the study area and up to 200 km north
from the sample plots used in this study, we estimate that the lat-
itudinal gradient is unlikely to be the primary factor behind the
observed differences in stand composition among stand-origin cat-
egories. Future studies should examine1—ifthere is indeed a
latitudinal climatic gradient, and 2 — if this gradient is sufficient
to explain major changes in post-disturbance forest composition.
The disentanglement of disturbance and climatic factors to explain
forest composition is still a major research problem in boreal forests
(Bergeron et al., 2004; Messaoud et al., 2007).
Stands originating from clearcutting had more large trees (diam-
eter >21 cm) than those in the other stand-origin categories,
with the exception of undisturbed stands (Table 1). This suggests
that some old-growth attributes could recover more rapidly after
clearcutting than after a natural disturbance such as fire, a fact
that should be taken into account when designing forest ecosys-
tem management strategies. This relatively rapid recovery may
have been caused in part by the retention of a significant propor-
tion of small trees when the original clearcuttings were performed
(M. Bouchard, unpublished results), whereas severe fires tend to
kill most trees within a stand. However, the abundance of large
trees is only one indicator of stand structural diversity that is not
necessarily directly related to other old-growth attributes, such as
dead wood abundance. Even if post-clearcutting recovery of old-
growth structural characteristics appears to be relatively quick in
some respects, these second-growth stands remain different from
stands that were not affected by severe disturbances for more than
200 years.
M. Bouchard, D. Pothier / Forest Ecology and Management 261 (2011) 811–820 819
5. Conclusion
Our study identified two important problems associated with
the transition phase from pre-industrial to actively managed land-
scapes in eastern Canadian boreal forests. The first consequence is
the change in forest composition observed at the stand level when
clearcutting is compared with fire (Table 1 and Fig. 6), a problem
that could be addressed partly by modifying silvicultural practices
at this scale (Harvey et al., 2002; Haeussler and Kneeshaw, 2003),
and by leaving a proportion of these stands to age sufficiently
such that old-growth attributes will develop. The second prob-
lem, concerns the declining abundance of mature and old forests
throughout the 20th century. Potential solutions to this second
problem include: improving the spatio-temporal distribution of
clearcuttings across the region to maintain a proportion of forests
with forest-interior characteristics, and adjusting rates of clearcut-
ting, to take into account the risk of occurrence of uncontrolled
disturbances, and to provide some leeway for the maintenance
of a sufficient old forest component at the regional scale. These
measures could be considered as a precautionary approach while
research continues to document and monitor the actual effects
of a reduced abundance of old-growth forests on regional-scale
populations of potentially threatened species, and to determine to
what degree second-growth forests can fulfill the same functions
(Higdon et al., 2006).
Acknowledgements
We thank D. Cyr, Y. Bergeron, S. Gauthier and Y. Boucher for their
comments on the manuscript. Financing for this study was provided
by the Natural Sciences and Engineering Research Council (NSERC)
- Laval University Research Chair in Silviculture and Wildlife, and
from a NSERC postdoctoral fellowship to M. Bouchard.
References
Aakala, T., Kuuluvainen, T., De Grandpré, L., Gauthier, S., 2007. Trees dying
standing in the northeastern boreal old-growth forests of Quebec: spatial pat-
terns, rates, and temporal variation. Canadian Journal of Forest Research 37,
50–61.
Armstrong, G.W., 1999. A stochastic characterisation of the natural disturbance
regime of the boreal mixedwood forest with implications for sustainable forest
management. Canadian Journal of Forest Research 29, 424–433.
Bergeron, Y., Gauthier, S., Flannigan, M., Kafka, V., 2004. Fire regimes at the transi-
tion between mixedwood and coniferous boreal forest in northwestern Quebec.
Ecology 85, 1916–1932.
Bergeron, Y., Cyr, D., Drever, C.R., Flannigan, M., Gauthier, S., Kneeshaw, D., Lauzon,
È., Leduc, A., Le Goff, H., Lesieur, D., Logan, K., 2006. Past, current, and future fire
frequencies in Quebec’s commercial forests: implications for the cumulative
effects of harvesting and fire on age-class structure and natural disturbance-
based management. Canadian Journal of Forest Research 36, 2737–2744.
Bouchard, M., Pothier, D., Ruel, J.-C., 2009. Stand-replacing windthrow in the boreal
forests of eastern Quebec. Canadian Journal of Forest Research 39, 481–487.
Bouchard, M., Pothier, D., 2008. Simulations of the effects of changes in mean fire
return intervals on balsam fir abundance, and implications for spruce budworm
outbreaks. Ecological Modelling 218, 207–218.
Bouchard, M., Pothier, D., 2010. Spruce budworm outbreak effects over a large
spatio-temporal gradient in boreal forests of eastern Québec. Canadian Journal
of Forest Research 40, 86–94.
Bouchard, M., Pothier, D., Gauthier, S., 2008. Fire frequency and tree species succes-
sion along large-scale geographic gradients in the North Shore region of eastern
Quebec. Canadian Journal of Forest Research 38, 1621–1633.
Boucher, Y., Arseneault, D., Sirois, L., 2006. Logging-induced change (1930–2002) of
a preindustrial landscape at the northern range limit of northern hardwoods,
eastern Canada. Canadian Journal of Forest Research 36, 505–517.
Boucher, Y., Arseneault, D., Sirois, L., Blais, L., 2009. Logging pattern and landscape
changes over the last century at the boreal and deciduous forest transition in
Eastern Canada. Landscape Ecology 24, 171–184.
Boudreault, C., Bergeron, Y., Drapeau, P., Mascarua-Lopez, L., 2008. Edge influence
on epiphytic lichens in remnant forests (cutblock separators, riparian buffers,
and large forest remnants) of the eastern boreal forest in Canada. Forest Ecology
and Management 255, 1461–1471.
Carey, A.B., Johnson, M.L., 1995. Small mammals in managed, naturally young, and
old-growth forests. Ecological Applications 5, 336–352.
Cissel, J.H., Swanson, F.J., Weisberg, P.J., 1999. Landscape management using histor-
ical fire regimes: Blue River, Oregon. Ecological Applications 9, 1217–1231.
Courtois, R., Gingras, A., Fortin, D., Sebbane, A., Rochette, B., Breton, L., 2008. Demo-
graphic and behavioural response of woodland caribou to forest harvesting.
Canadian Journal of Forest Research 38, 2837–2849.
Courtois, R., Ouellet, J.-P., Gingras, A., Dussault, C., Breton, L., Maltais, J., 2003. Histor-
ical changes and current distribution of caribou Rangifer tarandus, in Quebec.
Canadian Field Naturalist 117, 399–414.
Cumming, S.G., 2001. A parametric model of the fire-size distribution. Canadian
Journal of Forest Research 31, 1297–1303.
De Grandpré, L., Morissette, J., Gauthier, S., 2000. Long-term post-fire changes in the
northeastern boreal forest of Québec. Journal of Vegetation Science 11, 791–800.
Delong, S.C., Tanner, D., 1996. Managing the pattern of forest harvest: lessons from
wildfire. Biodiversity and Conservation 5, 1191–1205.
Desponts, M., Brunet, G., Bélanger, L., Bouchard, M., 2004. The eastern boreal old-
growth balsam fir forest: a distinct ecosystem. Canadian Journal of Botany 82,
830–849.
Didion, M., Fortin, M.-J., Fall, A., 2007. Forest age structure as indicator of boreal forest
sustainability under alternative management and fire regimes: a landscape level
sensitivity analysis. Ecological Modelling 200, 45–58.
Drapeau, P., Leduc, A., Giroux, J.-F., Savard, J.-P, Bergeron, Y., Vickery, W.L., 2000.
Landscape-scale disturbances and changes in bird communities of boreal mixed-
wood forests. Ecological Monographs 70, 423–444.
Etheridge, D.A., MacLean, D.A., Wagner, R.G., Wilson, J.S., 2006. Effects of intensive
forest management on stand and landscape characteristics in Northern New
Brunswick Canada (1945–2027). Landscape Ecology 21, 509–524.
Fahrig, L., 2003. Effects of habitat fragmentation on biodiversity. Annual Review of
Ecology, Evolution, and Systematics 34, 487–515.
Fall, A., Fall, J., 2001. A domain-specific language for models of landscape dynamics.
Ecological Modelling 141, 1–18.
Franklin, J.F., Spies, T.A., Van Pelt, R., et al., 2002. Disturbances and structural
development of natural forest ecosystems with silvicultural implications using
douglas-fir forests as an example. Forest Ecology and Management 155,
399–423.
Frenette, P., 1996. Histoire de la Côte-Nord. Institut québécois de recherche sur la
culture, Québec, Qc Canada.
Gagné, C., Imbeau, L., Drapeau, P., 2007. Anthropogenic edges: their influence on
the American Three-Toed Woodpecker (Picoides dorsalis) foraging behaviour
in managed boreal forests of Québec. Forest Ecology and Management 252,
191–200.
Girard, F., Payette, S., Gagnon, R., 2009. Origin of the lichen–spruce woodland in the
closed-crown forest zone of eastern Canada. Global Ecology and Biogeography
18, 291–303.
Girardin, M.P., 2010. Wildfire risk inferred from tree rings in the Central Laurentians
of boreal Quebec Canada. Dendrochronologia 28, 187–206.
Girardin, M.P., Mudelsee, M., 2008. Past and future changes in Canadian boreal
wildfire activity. Ecological Applications 18, 391–406.
Greene, D.F., Zasada, J.C., Sirois, L., Kneeshaw, D., Morin, H., Charron, I., Simard, M.-J.,
1999. A review of the regeneration dynamics of North American boreal forest
tree species. Canadian Journal of Forest Research 29, 824–839.
Haeussler, S., Kneeshaw, D., 2003. Comparing forest management to natural pro-
cesses. In: Burtont, et al. (Eds.), Towards Sustainable Management of the Boreal
Forest. NRC Research Press, Ottawa, pp. 307–368.
Harper, K.A., MacDonald, S.E., Burton, P.J., Chen, J., Brosofske, K.D., Saunders, S.C.,
Euskirchen, E.S., Roberts, D., Jaiteh, M.S., Esseen, P.-A., 2005. Edge influence
on forest structure and composition in fragmented landscapes. Conservation
Biology 19, 768–782.
Harvey, B., Leduc, A., Gauthier, S., Bergeron, Y., 2002. Stand-landscape integration
in natural disturbance-based management of the southern boreal forest. Forest
Ecology and Management 155, 371–388.
He, H.S., 2008. Forest landscape models: definitions, characterization, and classifi-
cation. Forest Ecology and Management 254, 484–498.
Higdon, J.W., MacLean, D.A., Hagan, J.M., Reed, J.M., 2006. Risk of extirpation for
vertebrate species on an industrial forest in New Brunswick Canada: 1945, 2002,
and 2027. Canadian Journal of Forest Research 36, 467–481.
Johnson, E.A., 1992. Fire and Vegetation Dynamics – Studies From the North Amer-
ican Boreal Forest. Cambridge Studies in Ecology. Cambridge University Press,
Cambridge, UK.
Keane, R.E., Cary, G.J., Davies, I.D., Flannigan, M.D., Gardner, R.H., Lavorel, S., Lenihan,
J.M., Li, C., Rupp, T.S., 2004. A classification of landscape fire succession models:
spatial simulations of fire and vegetation dynamics. Ecological Modelling 179,
3–27.
Kurz, W.A., Apps, M.J., 1999. A 70-year retrospective analysis of carbon fluxes in the
Canadian forest sector. Ecological Applications 9, 526–547.
Landres, P.B., Morgan, P., Swanson, F.J., 1999. Overview of the use of natural vari-
ability concepts in managing ecological systems. Ecological Applications 9,
1179–1188.
Legendre, P., Legendre, L., 1998. Numerical Ecology, 2nd English ed. Elsevier Science,
Amsterdam.
Luyssaert, S., Schulze, E.-D., Börner, A., Knohl, A., Hessenmö ller, D., Law, B.E., Ciais, P.,
Grace, J., 2008. Old growth forests as global carbon sinks. Nature 455, 213–215.
McCullough, D., Werner, R.A., Neumann, D., 1998. Fire and insects in northern and
boreal forest ecosystems of North-America. Annual Review of Entomology 43,
107–127.
McCune, B., Mefford, M.J., 1999. PC-ORD Multivariate Analysis of Ecological Data.
Version 4. MjM Software Design, Gleneden Beach, OR, USA.
820 M. Bouchard, D. Pothier / Forest Ecology and Management 261 (2011) 811–820
Messaoud, Y., Bergeron, Y., Leduc, A., 2007. Ecological factors explaining the location
of the boundary between the mixedwood and coniferous bioclimatic zones in
the boreal biome of eastern North America. Global Ecology and Biogeography
16, 90–102.
Mladenoff, D.J., White, M.A., Pastor, J., Crow, T.R., 1993. Comparing spatial pattern in
unaltered old-growth and disturbed forest landscapes. Ecological Applications
3, 294–306.
Nilsson, S.G., Hedin, J., Niklasson, M., 2001. Biodiversity and its assessment in
boreal and nemoral forests. Scandinavian Journal of Forest Research 16,
10–26.
Nonaka, E., Spies, T.A., 2005. Historical range of variability in landscape
structure: a simulation study in Oregon USA. Ecological Applications 15,
1727–1746.
Oksanen, J., Kindt, R., Legendre, P., O’Hara, R.B., 2007. Vegan: Community Ecology
Package, Version 1.8-6. http://cran.rproject.org/.
Östlund, L., Zackrisson, O., Axelsson, A.-L., 1997. The history and transformation of a
Scandinavian boreal forest landscape since the 19th century. Canadian Journal
of Forest Research 27, 1198–1206.
Pham, A.T., De Grandpré, L., Gauthier, S., Bergeron, Y., 2004. Gap dynamics and
replacement patterns in gaps of the northeastern boreal forest of Quebec. Cana-
dian Journal of Forest Research 34, 353–364.
Rhemtulla, J.M., Mladenoff, D.J., Clayton, M.K., 2009. Historical forest baselines reveal
potential for continued carbon sequestration. PNAS 106, 6082–6087.
Roberts, D.W., 2007. labdsv: Ordination and Multivariate Analysis for Ecology. R
package version 1.3-1. URL http://CRAN.R-project.org/package=labdsv.
Schroeder, D., Perera, A.H., 2002. A comparison of large-scale spatial vegetation pat-
terns following clearcuts and fires in Ontario’s boreal forests. Forest Ecology and
Management 159, 217–230.
Shaw, R.G., Mitchell-Olds, T., 1993. ANOVA for unbalanced data: an overview. Ecol-
ogy 74, 1638–1645.
Sturtevant, B.R., Bissonette, J.A., Lond, J.N., Roberts, D.W., 1997. Coarse woody debris
as a function of age, stand structure, and disturbance in boreal Newfoundland.
Ecological Applications 7, 702–712.
Turner, M.G., Romme, W.H., 1994. Landscape dynamics in crown fire ecosystems.
Landscape Ecology 9, 59–77.
Venables, W.N., Ripley, B.D., 2002. Modern Applied Statistics with S, 4th ed. Springer-
Verlag, New York.
Wimberly, M.C., Spies, T.A., Nonaka, E., 2004. Using natural fire regime-based crite-
ria to evaluate forest management in the Oregon Coast Range. In: Perera, A.H.,
Buse, L.J., Weber, M.G (Eds.), Emulating Natural Forest Landscape Disturbances:
Concepts and Applications. Columbia University Press, New York, NY, USA, pp.
146–157.