Forest Ecology and Management 131 (2000) 93±106

Ecological characteristics of mature forest

remnants left by wild®re
S. Craig DeLonga,*, Winifred B. Kesslerb
a
Ministry of Forests, 1011 4th Ave., Prince George, BC, Canada V2L 3H9
b
University of Northern British Columbia, 3333 University Way, Prince George, BC, Canada V2N 4Z9

Received 12 March 1999; accepted 6 July 1999

Abstract

Our study objective was to develop a better understanding of the ecological signi®cance of unburned forest remnants in
successional sub-boreal landscapes created by ®re. We characterized remnant forest patches and compared them to matrix
forest in young, mature and old age classes. Remnant patches could be discriminated from matrix forest types based on
variables relating to tree overstory and snag density. Some remnants displayed a unique uneven-aged pattern of lodgepole pine
regeneration. Differences between remnant patches and matrix forest stands, and high variation among patches, may re¯ect the
variable in¯uence of the wild®res through which the patches survived. Remnants share many ecological characteristics with
old forest and may provide `bridging habitats' in landscapes recovering from large-scale disturbance. Patches of mature forest
retained in logged landscapes may have potential to substitute for wild®re remnants. However, selection and management
criteria should be developed to guide the design of habitat retention and to monitor effects. The unusual regeneration dynamics
demonstrated by some remnants may suggest an alternative silvicultural model for regenerating stands dominated by
lodgepole pine within the boreal forest. # 2000 Elsevier Science B.V. All rights reserved.

Keywords: Wild®re disturbance; Forest remnants; Stand structure; Woody debris; Tree regeneration

1. Introduction landscapes dominated by lodgepole pine (Pinus con-

In many forest types, wild®re is instrumental in
controlling the distribution of habitat types across the
landscape and in shaping stand structure, composition
and biological diversity (Agee, 1993; Duchesne, 1994;
Whelan, 1995; DeLong and Tanner, 1996). Prior to ®re
suppression, stand replacing wild®res were the main
natural disturbance in sub-boreal and boreal plateau
*
Corresponding author. Fax: ‡1-250-565-3439.
E-mail address: craig.delong@gems1.gov.bc.ca (S.C. DeLong)
torta Dougl. ex Loud.), white spruce (Picea glauca
(Moench) Voss) and trembling aspen (Populus tremu-
loides Michx.) (Rowe and Scotter, 1973; Zackrisson,
1977; Van Wagner, 1978). Patterns of vegetation on
the landscape resulted primarily from infrequent,
large, highly intense ®res (Johnson, 1992). Individual,
stand-replacing wild®re often burned >10,000 ha
(Eberhart and Woodward, 1987; Andison, 1996)
and average ®re return interval ranged 80±125 years
(Johnson, 1992; Bunnell, 1995; Andison, 1996).
These natural ®res caused extensive mortality in

0378-1127/00/$ ± see front matter # 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 3 7 8 - 1 1 2 7 ( 9 9 ) 0 0 2 0 3 - 0
94 S.C. DeLong, W.B. Kessler / Forest Ecology and Management 131 (2000) 93±106
canopy and understory plants; and, because crown
®res were common, high levels of tree mortality were
usual (Johnson, 1992). Evidence of prompt natural
regeneration of burned stands is seen in the vigorous
and rapid natural stocking of lodgepole pine after
clearcutting, leading to the domination of sub-boreal
landscapes by even-age stands of lodgepole pine
(Eremko, 1990).
Although stand-replacing ®res dominate in sub-
boreal ecosystems, burned landscapes typically con-
tain patches of forest that did not burn. These patches
contain `biological legacies' ± such old large diameter
trees, snags, and coarse woody debris ± that are
important in the long term ecological functioning of
forest ecosystems (Amaranthus et al., 1994; Franklin,
1994; Hansen et al., 1991). Studies by DeLong and
Tanner (1996) in the sub-boreal forest and by Eberhart
and Woodward (1987) in the boreal forest found that
remnants comprised 3±15% of burned areas, and that
the proportion of island remnant to burned area
increased with the size of the wild®re. Carried over
from the pre-disturbance to the post-disturbance eco-
system, remnant forest patches may in¯uence the
structure, function, and composition of the recovering
ecosystem (Franklin, 1994). For example, they may
contribute to maintenance or re-establishment of
populations within disturbed landscapes
The occurrence of remnant forest patches within
burned area of large wild®res appears related to:
combinations of topography and soil moisture which
reduce ®re intensity; wind dynamics during the ®re;
timing of the burn relative to diurnal temperature and
wind patterns; and, horizontal continuity of the fuel
bed (Foster, 1983; Eberhart and Woodward, 1987).
Remnant forest patches in northern forests are gen-
erally small. DeLong and Tanner (1996) found that all
remnant patches were <10 ha within wild®res
<1000 ha in total area. On average, 50% of the rem-
nants were <2 ha. The small size of these remnant
forest patches likely in¯uences the microclimate
within them. Edge effects on microclimate and vege-
tation dynamics are reported to extend at least 2 tree
heights into forest patches (Oke, 1978; Chen et al.,
1992); hence the majority of remnant forest patches
consist mostly of edge habitat.
In a study of the ecological impacts of large ®res on
moose, Gasaway and DuBois (1985) found that 67%
of moose (Alces alces L.) observations within the burn
perimeter were in unburned patches that represented
only 15% of the total ®re area. Remnants may also act
as biological refugia for certain organisms. Amar-
anthus et al. (1994) found that hypogeous fungi were
more abundant in remnant Douglas-®r (Pseudotsuga
menziesii (Mirb.) Franco) stands than in the surround-
ing young forest. The role of remnant forest patches
may be especially important in landscapes where
residual habitat features such as individual live trees
or snags are uncommon. In a burned sub-boreal land-
scape, DeLong and Tanner (1996) reported an average
of only 0.75 live and 0.26 dead remnant trees per
hectare when island remnants were not considered.
Remnant forest patches, being older than the regen-
erating forest around them, typically contain large
diameter live and dead trees and other habitat features
lacking in the surrounding forest matrix.
Thus although remnant forest patches are small and
comprise a minor percentage of the landscape, they
may represent signi®cant and unique habitat oppor-
tunities that are not otherwise available in burned
successional landscapes. There are signi®cant forest
management implications as well. The current para-
digm in forest management in British Columbia
assumes that biodiversity can be maintained by
designing forest harvesting to mimic landscape pat-
terns created by wild®re. Management strategies
based on this idea (e.g., the Forest Practices Code
Biodiversity Guidebook [British Columbia Ministry
of Forests, 1995) include the retention of forest
patches within even-age cutting units to mimic the
remnant patches left by wild®re. The assumption is
that the retained forest patches serve the same habitat
roles as `natural' patches left by wild®re. We designed
this study to identify signi®cant ecological features of
remnant patches that would be desirable to emulate in
retained patches within harvested areas.
We examined stand structural attributes such as tree
density, basal area, coarse woody debris (CWD)
volumes, and tree regeneration to investigate the
ecological signi®cance of remnant patches within
surrounding contiguous forest (hereafter referred to
as `matrix' forest). We selected some variables that
previous studies have used to investigate structural and
functional relationships of old growth forests (Spies
and Franklin, 1991; Arsenault and Brad®eld, 1995).
Our purpose was to characterize those habitat features
contributed by remnant patches that are not otherwise
 S.C. DeLong, W.B. Kessler / Forest Ecology and Management 131 (2000) 93±106
found in matrix forest. With respect to forest manage-
ment, we sought to identify those habitat features of
natural, ®re-maintained landscapes that are lacking in
landscapes treated by conventional clearcutting. The
results have potential application in forest manage-
ment strategies that include habitat retention.
We postulate that remnant forest patches constitute
unique habitat elements within sub-boreal landscapes,
owing to their presence and to within-patch changes
induced by wild®re. We designed the study to answer
three speci®c questions relating to this thesis:
1. Can remnant forest patches be discriminated from
sucessional matrix forest on ecologically similar
sites based on differences in diameter class
distribution and density of live trees and snags
above certain thresholds?
2. Can remnant forest patches be discriminated from
successional matrix forest on ecologically similar
sites based on differences in volume of different
decay classes of coarse woody debris?
3. Can remnant forest patches be differentiated from
old matrix forest based on differences in occur-
rence and abundance of understory tree species
regeneration?
Rejection of the hypotheses will provide evidence
that remnant forest patches left by wild®re do not offer
habitat conditions or attributes that differ from those
within matrix forest. Results in the af®rmative will
support our thesis that remnant patches are unique
relative to matrix forest, thus adding to the biodiver-
sity of burned landscapes.
2. Methods and analysis
The study was conducted in the plateau portion of
the Salmon River variant of the moist cool Sub-boreal
Spruce (SBSmk1) biogeoclimatic unit (DeLong et al.,
1993), located 54±558N and 122.5±1248W. The range
in elevation is 750±1000 m. Mean annual precipitation
and temperature are 727 mm and 1.58C, respectively.
Of the total 782,213 ha, approximately 660,000 ha
are forested. Forest stands typically are dominated
by lodgepole pine or hybrid white spruce (Picea
glauca (Moench) Voss x engelmannii Parry ex
Englm.), with some local domination by trembling
aspen or black spruce (Picea mariana (P. Mill.)
95
P.S.B.). Scattered patches are dominated by Dou-
glas-®r or paper birch (Betula papyrifera Marsh.).
The study area was selected, in part, for relative
uniformity of topography.
We de®ne a remnant forest patch as a patch of older
forest 10 ha that is surrounded by younger forest,
based on the ®ndings of DeLong and Tanner (1996)
that within wild®res <1000 ha in total area, remnant
patches were <10 ha.
We used quadrat plot sampling techniques to com-
pare various ecological attributes between remnant
forest patches and matrix forest. We used maps of
stand age of the whole study area and more detailed
maps of individual wild®res to identify potential
sample sites. Potential sample locations within rem-
nant patches were identi®ed on air photos to assist
®eld location and inspection.
To minimize between-site variation, we selected
sites that met the following criteria: mesic moisture
regime; poor to medium nutrient regime; site series
classi®cation either PlSb±Feathermoss (SBSmk1/06)
or Sx±Huckleberry (SBSmk1/01) (DeLong et al.,
1993); location on morainal or lacustrine soils; slope
position level or mid-slope; slope gradient <20%;
without evidence of prior tree cutting; and, sample
plot placement >1 tree length (20±30 m) from obvious
clearings, stands of another age, or other stands not
meeting these criteria. Our method of plot location, an
approximation of the releve method, represented a
compromise between complete subjectivity and com-
plete randomness (Barbour et al., 1987). We estab-
lished 30 m  30 m plots at accessible sites that met
our selection criteria. Ten plots were established
within each age class of matrix forest (40±70, 71±
140, 141 ‡ years). Ten plots were established in forest
remnant patches surrounded by 40±70-year old matrix
forest. Sampling occurred May±September in 1994
and 1995.
We described site and soil conditions according to
the methods in Luttmerding et al. (1990). Site infor-
mation included slope gradient, aspect, and slope
position. At each plot we dug a pit at least 50 cm into
the mineral soil. For all soil layers with strongly
contrasting physical characteristics, we recorded per-
cent coarse fragment (>2 mm) content and soil texture
of the 2 mm fraction. We also estimated effective
rooting depth and identi®ed any root restricting layers
(Luttmerding et al., 1990). Humus layer depth was
96 S.C. DeLong, W.B. Kessler / Forest Ecology and Management 131 (2000) 93±106

estimated from ®ve random locations per plot. We also
recorded visible evidence of tree damage associated
with natural disturbance (e.g., windthrow or ®re
scars).
We assigned stands to four categories for analysis:
(1) remnant forest patches (left within burns 40±70
years old); and, matrix forest of stand age (2) 40±70,
(3) 71±140, and (4) >140 years. We tested each
variable used in Analysis of Variance (ANOVA) for
homogeneity of variance using the Fmax test (Sokal
and Rohlf, 1969). Where results indicate, we removed
young and mature stands prior to ANOVA testing to
meet the assumption of homogeneity of variance.
ANOVA residuals were saved and plotted against
expected values for a normal distribution to assess
normality (Wilkinson et al., 1996). These plots were
compared to those of 10 random variables plotted
against expected values for a normal distribution.
All reported ANOVA results met the assumptions of
homogeneity of variance and normality. We used
SYSTAT (Systat Inc. Evaston, Ill) to conduct all
statistical procedures
2.1. Overstory structure
We collected the following data for of all live and
dead standing trees 7.5 d.b.h. within each
30 m  30 m plot: diameter (cm) at breast height
(1.3 m); decay class of snags (intact, intact to partially
soft, hard large pieces, small soft blocky pieces, or soft
and powdery); and, presence of large (3 cm) and
small (<3 cm) diameter cavities.
For the species dominating the main canopy, gen-
erally lodgepole pine, we measured heights and ages
Table 1
List of attributes included in statistical analysis and reasons for inclusion
of six trees per plot. For other canopy species we
sampled 2 trees per plot. We measured trees that
were free of major defects (e.g., broken top) and that
represented the range of heights in the main canopy.
We also measured heights and ages of all canopy trees
that had survived earlier disturbances, as indicated by
®re scars or obvious larger size. We measured all
heights using a CriterionTM laser range®nder. Trees
were aged from cores extracted by an increment borer
at 1.3 m from the ground. For trees with ®re scars we
extracted cores from the scar area as well as a non
scarred area of the trunk.
We computed basal area and density of trees and
snags by species and 5 cm diameter classes. We con-
structed plots of density by diameter class for each plot
to test hypotheses relating to stand structure. We also
computed average tree diameter and number of trees
containing wildlife feeding cavities. Due to the large
number of variables that could be generated we devel-
oped criteria to select variables to include in analysis
of variance (ANOVA) and discriminant analysis
(Table 1), and computed basic statistics for the
selected variables. We used ANOVA and Tukey's
pairwise multiple comparisons to test for signi®cant
differences in those variables between stand categories
(Wilkinson et al., 1996). All selected variables were
included in the discriminant analysis between stand
types. Procedures for discriminant analysis followed
those outlined in Wilkinson et al. (1996).
2.2. Coarse woody debris
Coarse woody debris (CWD) was sampled along
90 m transects using a line intercept method adapted

Attribute Reasons for inclusion in discriminant analysis

>7.5 cm-d.b.h. tree density
>7.5 cm-d.b.h. basal area
>7.5 cm-d.b.h. snag density
>15 cm-d.b.h. tree density
>25-cm-d.b.h. tree density
>15 cm-d.b.h. snag density
>25 cm-d.b.h. snag density
a
Based on tree and snag requirements for birds in Thomas et al. (1979); Neitro et al. (1985).
likely to show sharp contrasts between stand types due to high initial densities in young stands and self-
thinning over time; commonly measured attribute during operational surveys
likely to show sharp contrasts between stand types due to changes of basal area over time; commonly
measured attribute during operational surveys
likely to show sharp contrasts between stand types due to high levels of tree mortality in intermediate-aged
stands; commonly measured attribute during operational surveys
represents trees large enough to meet the requirements of certain small cavity requiring bird speciesa
represents trees large enough to meet the requirements of certain large cavity requiring bird speciesa
represents snags large enough to meet the requirements of certain small cavity requiring bird speciesa
represents snags large enough to meet the requirements of certain large cavity requiring bird speciesa
 S.C. DeLong, W.B. Kessler / Forest Ecology and Management 131 (2000) 93±106 97

from Trowbridge et al. (1989). The ®rst 60 m of the
transect consisted of two perpendicular edges of the
30 m  30 m plot. The transect continued as a 30 m
diagonal line through the plot centre. We recorded
diameter (cm) at the point of interception and decay
class (intact (1), intact to partially soft (2), hard large
pieces (3), small soft blocky pieces (4), or soft and
powdery (5) for each piece of woody debris 7.5 cm
d.b.h. that intersected the tape: Diameter of soft and
powdery stems was measured both parallel and per-
pendicular to the ground, and an average calculated.
Volume by decay class and 5 cm diameter increment
were estimated using the formula described in Lofroth
(1992):
P stems were distributed over at least 3 diameter classes.
p2 d2
Vˆ
8L
within all plots by layer (dominant, co-dominant,
shrub (<10 m in height)).
3. Results
3.1. Overstory structure
In the young (40±70 years) forest matrix stands,
most of the lodgepole pine stems were of the two
smallest diameter classes, suggesting domination by a
single cohort that initiated following wild®re
(Fig. 1(a)). In the mature (41±140 years) and old
(>140 years) matrix stands and in remnant patches,
As well, black spruce, white spruce and subalpine ®r
(Abies lasciocarpa (Hook.) Nutt.) were generally

where V is volume in m3/ha, d is diameter (cm) of each

piece of woody debris, and L is the length (m) of the
transect. We conducted a 1-way ANOVA to test for
significant differences between stand types in total
woody debris volume and volume by decay class
(Wilkinson et al., 1996). This was followed by Tukey
multiple comparisons between remnant forest patches
and matrix forest types (Wilkinson et al., 1996). We
also conducted discriminant analysis to determine if
stand type could be discriminated on the basis of total
CWD volume or volume by decay class.

2.3. Tree regeneration

Tree regeneration by species was sampled at the rate

of ®ve samples per stand category (remnant patches,
40±70-year old matrix forest, and >140-year old
matrix forest). Mature (71±140-year old) matrix forest
was not sampled. Each regeneration sample was based
on nine circular plots of 3.99 m radius (50 m2 area)
established within the 30 m  30 m vegetation plots.
The circular plots were arranged in a grid pattern to
avoid overlap.
We recorded species, height (cm), and current and
previous years growth increments for all trees 60 cm
and <10 m in height. ANOVA tested for differences
between forest stand types in the number of regener-
Fig. 1. Density of live trees by diameter class for (a) a young
ating trees, by species. matrix forest stand and (b) an old matrix forest stand illustrating the
In addition, as part of a detailed vegetation survey larger spread in diameter class distribution in older stands relative
we recorded the percentage cover of all tree species to young stands.
98 S.C. DeLong, W.B. Kessler / Forest Ecology and Management 131 (2000) 93±106

more abundant in these categories than in young

matrix forest (Fig. 1(b)). Lodgepole pine density
exhibited a single, early peak in all but two stands.
The two exceptions, both remnants, exhibited a second
peak of small (7.5±12.5 cm) diameter lodgepole pine,
suggesting that a second cohort had established
(Fig. 2).
One±way ANOVA followed by Tukey's multiple
comparisons indicated signi®cant differences between
stand types for several overstory attributes. Tree den-
sity was highest in young matrix forest, intermediate
in mature matrix forest, and lowest in remnant patches
and old matrix forest (Table 2). Snag density was
higher in mature matrix forest than other stand types
(Tables 2 and 3). Average density of live stems >15 cm
but <25 cm in diameter, and therefore potentially
available to small cavity nesting species, was signi®-
cantly lower in young matrix forest than in other stand
types (Tables 2 and 3). Old matrix forest contained the
highest density of snags in this diameter class. Aver-
age density of live stems >25 cm was signi®cantly
higher in old matrix forest than in remnant patches
(p < 0.05). Average density of snags >25 cm d.b.h.
was not signi®cantly different between old matrix
forest and remnant patches. Trees or snags >25 cm
d.b.h. were rare or absent in young and mature matrix
forest but these stand types were excluded from the Fig. 2. Density of live trees by diameter class for two remnant
ANOVA because the assumption of homogeneity of forest patches exhibiting secondary peaks of lodgepole pine
density.
variance was violated. Average diameter of stems >7.5
d.b.h. was signi®cantly lower in young matrix forest
than other stand types and higher in old matrix forest
than in mature matrix forest (Tables 2 and 3). Greater

Table 2
Means and standard deviations of selected stand characteristics for young matrix forest, mature matrix forest, remnant patches and old matrix
forest and F ratio and p values for ANOVA testing (Df ˆ 3 for ANOVA where all stand types included and Df ˆ 1 where only remnant patches
and old matrix forest included)

Stand characteristics Stand type

Young Mature Remnant Old F p

Tree density 2597 (471) 1910 (780) 1165 (394) 984 (263) 24.44 0.000
Snag density 268 (198) 460 (193) 158 (78) 170 (72) 8.92 0.000
>15 cm-d.b.h. tree density 408 (273) 860 (189) 693 (171) 698 (215) 7.59 0.000
>25 cm-d.b.h. tree density 9 (11)a 59 (55)a 221 (90) 334 (99) 7.23 0.015
>15 cm-d.b.h. snag density 12 (17) 59 (52) 73 (45) 126 (54) 11.29 0.000
>25 cm-d.b.h. snag density 3 (5)1 2 (5)a 15 (17) 31 (27) 2.38 0.140
# of cavities/ha 13 (18) 9 (11) 30 (34) 18 (20) 1.68 0.189
Average tree d.b.h. (cm) 11.3 (1.2) 15.4 (1.9) 17.7 (3.7) 20.8 (3.3) 24.31 0.000
a
Stand type not included in ANOVA due to assumption of homogeneity of variance being violated when included.
 S.C. DeLong, W.B. Kessler / Forest Ecology and Management 131 (2000) 93±106 99

Table 3
p values for Tukey pairwise comparisons for stand characteristics where all stand types were included in the analysis and where ANOVA was
significant at <0.05

Stand characteristics Stand type pairwise combinations

Y±Ma Y±R Y±O M±R M±O R±O

Tree density 0.003 0.000 0.000 0.020 0.002 0.822
Snag density 0.030 0.358 0.461 0.000 0.001 0.998
>15 cm-d.b.h. tree density 0.000 0.027 0.024 0.325 0.348 1.000
>15 cm-d.b.h. snag density 0.105 0.020 0.000 0.885 0.008 0.051
Average tree d.b.h. (cm) 0.012 0.000 0.000 0.055 0.000 0.162
a
Stand types (Y ˆ young matrix forest, M ˆ mature matrix forest, R ˆ remnant patches, O ˆ old matrix forest).

variability in stand structure of remnant patches and
old matrix forest compared to young and mature
matrix forest is indicated by the larger standard devia-
tions for average tree diameter (Table 2).
Trees containing feeding cavities were found in all
stand types with no signi®cant differences in average
number detected. However, the mean number of trees
with cavities was highest in remnant patches and 8 of
10 remnant plots had cavity trees versus 7, 5, and 4 out
of 10 in old, mature and young matrix forest, respec-
tively.
Density of live trees >25 cm d.b.h. and overall tree
density were the most effective stand structure vari-
ables for discriminating among stand types. The dis-
criminant functions correctly classi®ed 34 of 40
stands, demonstrating an overall misclassi®cation rate
of 15% (Table 4). Seven of the 10 remnant patches
were classi®ed correctly while two were classi®ed as
old matrix forest and one as mature matrix forest. One
of the two remnant patches mis-classi®ed as old
matrix forest was the largest (10 ha) in the sample.
The other was the oldest (>200 years) patch sampled.
The remnant patch mis-classi®ed as mature matrix
forest was the youngest (90 years) patch sampled.
The discriminant analysis produced equations that can
be used to assign new stands to stand type based on the
stand structural criteria on which the equations are
based (Table 5).
Thus, remnant forest patches could be discrimi-
nated from young and mature matrix forest based
on a number of overstory stand structure variables.
ANOVA found only 1 variable, average density of live
stems >25 cm d.b.h., that was signi®cantly different
between remnant patches and old matrix forest. How-
ever, discriminant analysis indicated that remnant
patches could be discriminated from old matrix forest
based on overstory stand structure characteristics.
3.2. Coarse woody debris
No signi®cant differences in average total volume
of coarse woody debris (CWD) >7.5 cm in diameter
were found among stand types. ANOVA followed by
Tukey multiple comparisons indicated that average
volume of larger diameter (>22.5 cm) CWD in the
`intact' decay class was signi®cantly higher (n ˆ 10,
p < 0.05) in old matrix forest than in young and mature
matrix forest, but not signi®cantly higher in old matrix
forest than in remnant patches. Nine of 10 remnant
patches had CWD in piece sizes >22.5 cm in diameter

Table 4
Summary of discriminant classifications of young matrix forest, mature matrix forest, remnant patches, and old matrix forest based on stand
characteristics

Stand type # of plots classified into each stand type by discriminant function

Young matrix Mature matrix Remnant patch Old matrix

Young matrix (n ˆ 10) 8 2 0 0
Mature matrix (n ˆ 10) 0 10 0 0
Remnant patch (n ˆ 10) 0 1 7 2
Old matrix (n ˆ 10) 0 0 1 9
100 S.C. DeLong, W.B. Kessler / Forest Ecology and Management 131 (2000) 93±106

Table 5
Discriminant functions for predicting case membership within the young matrix, mature matrix, remnant patch, and old matrix forest
categories. New case is assigned to the category with the largest function value for the case

Category Functiona

Young matrix 0.013x1 ‡ 0.012x2 ‡ 0.014x3 ÿ 0.006x4 ÿ 0.031x5 ‡ 0.176x6 ÿ 23.288

Mature matrix 0.010x1 ‡ 0.019x2 ‡ 0.023x3 ‡ 0.002x4 ÿ 0.015x5 ‡ 0.193x6 ÿ 24.372
Remnant patch 0.006x1 ‡ 0.008x2 ‡ 0.018x3 ‡ 0.033x4 ÿ 0.006x5 ‡ 0.198x6 ÿ 24.372
Old matrix 0.005x1 ‡ 0.011x2 ‡ 0.016x3 ‡ 0.057x4 ÿ 0.018x5 ‡ 0.231x6 ÿ 24.847
a
Discriminant function where x1 ˆ live tree density, x2 ˆ snag density, x3 ˆ > 15 cm d.b.h. tree density, x4 ˆ > 25 cm d.b.h. tree density,
x5 ˆ > 15 cm d.b.h. snag density, and x6 ˆ > 25 cm d.b.h. snag density.

compared to 6, 5, and 4 out of 10 for young, old and

mature matrix forest, respectively. On average, 71% of
the volume of CWD in young matrix forest stands was
in diameter classes >17.5 cm, indicating that a high
proportion of the CWD originated from the pre-dis-
turbance stand (Table 6).
In young and mature matrix forest, CWD volume
was concentrated in advanced decay classes 4 and 5
(Fig. 3). CWD was more evenly distributed in old
matrix forest and remnant patches, with a higher
proportion of total CWD volume in decay classes
1±3 (Fig. 3).
Discriminant analysis using CWD variables found
that decay class was not effective in discriminating
between stand types (i.e., no decay class was highly
correlated with the discriminant factors). Discriminant
analysis correctly classi®ed only 24 of 40 stands
Fig. 3. Distribution of CWD volume in five decay classes for
(Table 7). Only 5 of 10 remnant patches were correctly young, mature, and old matrix forest and remnant forest patches.
classi®ed, a result that re¯ects the wide variation in
CWD volume within stand types.
Thus, remnant patches could be differentiated from 3.3. Tree regeneration
young and mature matrix forest on the basis of a
higher proportion of CWD in less decayed states. We found signi®cant differences between remnant
However, no detectable differences were found patches and old matrix forest with respect to the 0.6±
between remnant patches and old matrix forest. High 10 m high regeneration layer. Lodgepole pine regen-
variation in CWD attributes was found in all stand eration was present in three of ®ve remnant patches
types. sampled, but was not detected in old matrix forest.

Table 6
Summary statistics for CWD volume for young matrix, mature matrix, remnant patch, and old matrix forest categories. (n ˆ 10)

Category Total < 17.5 d.b.h >17.5 d.b.h.

Mean Range SD Mean SD Mean SD

Young matrix 261.8 5.6±590.3 201.3 76 54.7 188.5 177.3

Remnant patch 229.2 33±393.4 116.4 104.8 46.7 124.2 86.5
Mature matrix 174.4 23.4±283.3 90.5 71.8 37.5 84.1 74.6
Old matrix 192.6 38.6±286 78.7 82.5 37.6 112.8 61.5
 S.C. DeLong, W.B. Kessler / Forest Ecology and Management 131 (2000) 93±106 101

Table 7
Summary of discriminant classifications of stand type based on CWD attributes

Category # of plots classified into each stand type by discriminant function

Young matrix Mature matrix Remnant patch Old matrix

Young matrix (n ˆ 10) 5 3 2 0
Mature matrix (n ˆ 10) 1 9 0 0
Remnant patch (n ˆ 10) 1 1 5 3
Old matrix (n ˆ 10) 0 3 2 5

There were apparent differences in density of hybrid
white spruce and black spruce but they were not
signi®cant at the 0.05 level (Fig. 4 and Table 8). No
difference was apparent for subalpine ®r density (Fig. 4
and Table 8). Lodgepole pine was present in the shrub
layer of 7 of 10 remnant forest patches but in none of
the old matrix forest stands.
An average of 1600 (maximum 4000) stems/ha
occurred in remnant patches containing lodgepole
pine regeneration. Height of the regeneration
layer ranged 0.6±6.0 m, with an average height of
1.6 m. Much of the regeneration in remnant patches
occurred in smaller height classes with the distribution
approaching a negative exponential (Fig. 5). In
contrast, most of the ®re-initiated lodgepole pine in
surrounding young matrix forest was >10 m high,
and density of stems <10 m high averaged 600
stems/ha. Average height for stems <10 m high

Fig. 4. Density of live trees 0.6±10 m in height for different Fig. 5. Frequency of lodgepole pine regeneration within 5 height
species within remnant forest patches and old matrix forest. classes in remnant forest patches.

Table 8
Mean and standard error and results of nested ANOVA for density of regeneration >0.6±10 m in height in remnant patches and old matrix
forest

Species Remnant patches Old matrix forest F P

Mean density SE Mean density SE

(stems/ha) (stems/ha)

Lodgepole pine 449 145 0 na na

Black spruce 1116 214 160 47 4.73 0.061
Hybrid white spruce 240 45 49 20 4.16 0.076
Subalpine fir 596 155 800 193 0.107 0.752
102 S.C. DeLong, W.B. Kessler / Forest Ecology and Management 131 (2000) 93±106
Fig. 6. Frequency of lodgepole pine regeneration within 7 height
classes in young matrix forest.
was 7.5 m and most stems were in the taller classes
(Fig. 6).
No statistical differences in height growth of regen-
eration were found between stand types for any of the
tree species present.
4. Discussion
4.1. Uniqueness of remnant patches
High variability characterized all attributes mea-
sured within remnant patches and matrix forest alike.
However, some attributes differed enough that rem-
nant forest patches could be differentiated from the
matrix forest types. Strong differences were evident
for overstory and understory tree regeneration,
although some differences for CWD were also sig-
ni®cant. These results support our thesis that remnant
forest patches are a unique habitat component within
sub-boreal forests maintained by wild®re. Attributes
relating to overstory stand structure were generally
effective in differentiating the three age classes of
matrix forest from remnant forest patches. Overstory
attributes have been shown to have high discriminat-
ing power for separating age-classes (Spies and Frank-
lin, 1991). The trends observed ± decreased density
and increased tree size over time ± re¯ect the normal
processes of stand development and vegetation suc-
cession. In addition, as the stand opens up, the density
of late successional smaller-diameter, shade tolerant
trees such as subalpine ®r increase, leading to high
variability in tree diameter. The differences in overs-
tory stand attributes of remnant patches, which allow
their discrimination from matrix forest, may derive
from the wild®re events that created the patches. For
example, lower density of larger (>25 cm d.b.h.) trees
may re¯ect stem mortality that occurred during the
®re. Greater total stem density may relate to the in¯ux
of new regeneration, particularly lodgepole pine, after
the ®re. That remnant patches could be discriminated
from old matrix forest (albeit weakly) suggests that
stand structure becomes modi®ed even though most of
the overstory trees in the patch itself do not burn in the
course of the wild®re. Differences in pre-®re condi-
tions, ®re intensity, and post-disturbance factors such
as wind and disease could all contribute to the varia-
bility in overstory stand structure among remnant
patches. Remnant patches most affected by wild®re
likely exhibit the greatest differences from old matrix
stands.
Coarse woody debris volume was highly variable
and relatively ineffective for discriminating remnant
patches from matrix forest. Many factors in¯uence
CWD dynamics, including biotic agents such as
insects and disease and abiotic factors such as wind
and ®re. In younger stands, new CWD input is low, but
total volumes may be high if abundant CWD remains
from the previous stand (Spies et al., 1988).
A simple model for describing changes in CWD
biomass after disturbance has been proposed by Har-
mon et al. (1986) and validated for Douglas-®r stands
in the Paci®c Northwest (Spies et al., 1988; Agee and
Huff, 1987), eastern subalpine forests (Lambert et al.,
1980) and sub-boreal forests of west-central British
Columbia (Clark et al., 1998). A large initial pulse of
CWD from the pre-disturbance stand is followed by
low levels of CWD input as the stand develops.
Eventually, there is another increase associated with
individual tree mortality in maturing stands. This
pattern was somewhat evident from our ®ndings but
differences could not be demonstrated with the sample
size due to high variability. High variability, especially
in young stands, was likely a re¯ection of variable
stand conditions prior to disturbance. Andison's
(1996) studies in SBSmk1 forests found an average
®re return interval of 80±100 years, and similar risk for
burning across age classes. This means that stands
could burn repeatedly in a short period (i.e., 40 years),
 S.C. DeLong, W.B. Kessler / Forest Ecology and Management 131 (2000) 93±106
resulting in low levels of CWD. Andison (1996) also
reported stands >200 years of age which when burned
would result in relatively high levels of CWD. Differ-
ences in stand age at time of disturbance, disturbance
intensity, and other factors provide for a wide range of
possibilities with respect to CWD in young stands.
Concentration of CWD in later stages of decay (i.e.,
small soft blocky pieces, or soft and powdery)
indicated that most of the material was of pre-dis-
turbance origin and has been on the ground long
enough for decay to have progressed. It also indicates
a low level of recruitment subsequent to the initial pre-
disturbance pulse. A similar pattern was evident in
the mature matrix forest, but the peak was in the
most advanced (soft and powdery) decay class. Pre-
disturbance CWD could be partially undetectable in
mature stands, having been incorporated into the
humus. A more balanced distribution of CWD volume
across decay classes in remnant patches and old matrix
forest suggests a smoother pattern of CWD input
associated with the periodic death of individual large
trees.
Lodgepole pine regeneration in the understory of
some remnant patches was a key feature that distin-
guished them from old matrix forest. In general,
remnant patches exhibit higher amounts of early
successional species (i.e., lodgepole pine, white
spruce and black spruce) than does old matrix forest.
This may be a legacy of the disturbance event that
created the patches.
4.2. Effects of burning
The stand structure found in the young, mature, and
old matrix stands is typical of the pattern of develop-
ment in SBSmk1 landscapes following stand-replace-
ment wild®re (DeLong and Tanner, 1996). The
serotinous cones of lodgepole pine release seed only
under intense heat (Lotan, 1974). Lodgepole pine
regenerates readily after wild®re, resulting in a rela-
tively even-aged cohort of young lodgepole pine
throughout the burned area (Krajina et al., 1982).
Little lodgepole pine regeneration occurs subsequent
to the initial cohort, and the more shade±adapted
species (e.g., spruce and subalpine ®r) tend to dom-
inate the understory. This pattern of stand develop-
ment in sub-boreal forests has been documented by
Kneewshaw (1992); Clark (1994).
103
We observed an alternate pattern of stand develop-
ment in 7 of 10 remnant patches. Evidence of a second
cohort of lodgepole pine was indicated (in 2 plots) by a
second peak in the diameter class distribution of
lodgepole pine. A second cohort was also indicated
in ®ve other remnant patches by presence of lodgepole
pine regeneration and/or lodgepole pine presence in
the shrub layer. In three of these stands the presence of
a dense (>100 stems per ha) second cohort of lodge-
pole pine suggests that the wild®re may have in¯u-
enced the remnant patch enough to initiate lodgepole
pine regeneration. The presence of more lodgepole
pine and hybrid white spruce, both which prefer
mineral soil substrate for regeneration, suggests that
mineral soil may have been exposed by ground ®re
within some of the remnants. Further evidence of ®re
disturbance within these remnants was provided by the
presence of ®re scarred trees whose scar age was
similar to that of the surrounding young forest. Regen-
eration plots containing lodgepole pine also had thin-
ner forest ¯oors on average than all other plots (3.5 cm
versus 5.1 cm), suggesting forest ¯oor consumption at
the time of the disturbance.
Findings relating to size distribution of the second
cohort of lodgepole pine are also interesting, and
indicated signi®cant differences in timing and dura-
tion of lodgepole pine recruitment. The structure in the
young matrix stands is even-aged, with lodgepole pine
stems mostly restricted to the overstory. In contrast,
the structure in remnant patches indicated uneven-
aged lodgepole pine regeneration, concentrated in the
lowest height class and approximating a negative
exponential curve. Uneven-aged structure and a nega-
tive exponential distribution have been reported for
lodgepole pine in the Sierra Nevada of California
(Parker, 1986) and on xeric, nutrient-poor sties in
Yellowstone National Park, Wyoming (Despain,
1983). These stands are characterized by continuous
or intermittent understory lodgepole pine regeneration
and a lack of crown ®res. The uneven-aged regenera-
tion results from annual seedfall from non-serotinous
cones. Lotan (1967) determined that in such stands the
ratio of serotinous to non-serotinous cone type trees is
lower than in landscapes that experience stand repla-
cement wild®re. In sub-boreal landscapes, however, a
higher ratio of serotinous-coned trees would be
expected given the frequency of stand±replacing wild-
®res.
104 S.C. DeLong, W.B. Kessler / Forest Ecology and Management 131 (2000) 93±106
The two different regeneration patterns we
observed in remnant patches and surrounding young,
matrix forest suggest that lodgepole pine may exhibit
both serotinous and non-serotinous regeneration stra-
tegies regardless of long-term selection pressure for
cone serotiny. We did not ®nd the uneven-age regen-
eration pattern in the largest remnant patches, which
makes sense considering that effects of the ®re are
unlikely to have penetrated into the interior of large
remnants. Because most remnant patches in SBSmk1
forests are <10 ha in area, the uneven-aged regenera-
tion pattern may be fairly common. Additional
research is required to determine whether this regen-
eration pattern occurs in response to disturbance, or
whether it simply re¯ects the presence of open-cone
bearing trees.
4.3. Management implications
Our ®ndings have implications for current manage-
ment policies and practices in British Columbia. Many
forest harvest operations include the retention of wild-
life tree patches as recommended in the Forest Prac-
tices Code Biodiversity Guidebook (British Columbia
Ministry of Forests, 1995). These patches are assumed
to substitute for the remnant forest patches that occur
in SBS landscapes maintained by wild®re. The differ-
ences we found between remnant patches and matrix
forest call that assumption into question; remnant
patches that survived wild®re are unique relative to
matrix forest. At issue, however, is whether the differ-
ences are signi®cant ecologically. Do the differences
imply that `natural' remnants contain critical habitat
conditions or components that are not provided in
wildlife tree patches? These questions will require
additional research.
Meanwhile, the descriptive data and discriminant
equations we developed have potential application in
the evaluation and design of wildlife tree patches. As
well, certain ®ndings will be of interest to managers:
 Large diameter (> 25 cm) live trees and snags
were more abundant in remnant patches than in
surrounding mature matrix forest. These elements
of patches may be a significant `biological legacy'
for wildlife species such as the pine marten
(Stordeur, 1986; Koehler et al., 1990) and black-
backed woodpecker (Goggans et al., 1988;
Marshall, 1992) that require large diameter trees
for nesting.
 Our finding of abundant feeding cavities in remnant
patches, and in particular in those trees that sur-
vived the wildfire, provides additional evidence for
the importance of `biological legacies' provided in
remnant patches.
 Young (40±70 years) stands created by logging lack
CWD in the larger sizes and in the newer (i.e.,
intact and intact to partially soft) decay classes. The
lack of CWD recruitment is explained by past and
present utilization standards, and comprises one of
the major differences between forest landscapes
managed for timber production versus those main-
tained by natural processes. Leaving unharvested
patches is one option for maintaining a source of
CWD throughout the rotation in managed forest
landscapes.
 Windthrow is a major management concern for the
retention of wildlife tree patches. In the remnant
patches examined, we did not find evidence of
increased windthrow relative to matrix forest of
the same age classes. That the distribution of CWD
volume by decay class and the density of large
standing trees did not differ between remnants and
old forest suggested that remnant patches are at
least as stable as old forest matrix. Remnant
patches left by wildfire may provide useful infor-
mation for the design of windfirm wildlife tree
patches.
In conclusion, remnant forest patches left by wild-
®re comprise unique habitat conditions that are not
otherwise present in the landscape. They provide the
sharpest contrast to the developing young forest that
surrounds them and provide biological legacies, such
as large diameter trees and large diameter snags, from
one rotation to another. High variation among island
remnants likely re¯ects differences in conditions
before, during and after disturbance.
One ®re effect seen in most remnants ± the uneven-
age pattern of lodgepole pine regeneration ± is other-
wise very rare in sub-boreal landscapes. This inter-
esting pattern is not accounted for in prevailing
management schemes, which assume an even-age
pattern of regeneration for all lodgepole pine stands
in sub-boreal forests. Further study of these natural
examples of uneven-aged lodgepole pine regeneration
 S.C. DeLong, W.B. Kessler / Forest Ecology and Management 131 (2000) 93±106
may provide managers with additional silvicultural
options for managing lodgepole pine in interior British
Columbia forests. This result would be timely, given
public pressures and policy direction (e.g., the Forest
Practices Code) to reduce clearcutting in favor of other
regeneration methods.
Acknowledgements
We sincerely thank Kathi Zimmerman, Howard
DeLong, Jenny Marsden for assistance with ®eld
sampling and Dr. Ken Lertzman for critical comments
on the thesis document on which this paper is based.
This study was funded by the Sustainable Environ-
ment Fund (British Columbia Ministry of Environ-
ment) and British Columbia Ministry of Forests
References
Agee, J.K., 1993. Fire ecology of Pacific Northwest Forests. Island
Press, Washington, DC. 475 pp.
Agee, J.K., Huff, M.H., 1987. Fuel succession in a western
hemlock/Douglas-fir forest. Can. J. For. Res. 24, 697±704.
Amaranthus, M., Trappe, J.M., Badnar, L., Archer, D., 1994.
Hypogeous fungal production in mature Douglas-fir fragments
and surrounding plantations and its relation to coarse woody
debris and animal mycophagy. Can. J. For. Res. 24, 2157±2165.
Andison, D., 1996. Managing for landscape patterns in the sub-
boreal forests of British Columbia. Ph.D. Thesis, University of
British Columbia, 197 pp.
Arsenault, A., Bradfield, G.E., 1995. Structural-compositional
variation in three age-classes of temperate rainforests in
southern coastal British Columbia. Can. J. Bot. 73, 54±64.
Barbour, M.G., Burk, J.H., Pitts, W.D., 1987. Terrestrial Plant
Ecology. Benjamin/Cummings, Don Mills, Ontario.
British Columbia Ministry of Forests, 1995. Forest Practices Code
of British Columbia: Biodiversity Guidebook. Queens Printer,
Victoria, BC.
Bunnell, F.L., 1995. Forest-dwelling vertebrate faunas and natural
fire regimes in British Columbia. Cons. Biol. 9, 636±644.
Chen, J., Franklin, J.F., Spies, T.A., 1992. Vegetation responses to
edge environments in old-growth Douglas-fir forests. Ecol.
Appl. 2, 387±396.
Clark, D., 1994. Post-fire succession in the sub-boreal spruce
forests of the Nechako Plateau, central British Columbia. M.
Sc. thesis, University of Victoria, 124 pp.
Clark, D.F., Kneeshaw, D.D., Burton, P.J., Antos, J.A., 1998.
Coarse woody debris in sub-boreal spruce forests of west-
central British Columbia. Can. J. For. Res. 28, 284±290.
DeLong, S.C., Tanner, D., 1996. Managing the pattern of forest
harvest: lessons from wildfire. Biod. Cons. 5, 1191±1205.
105
DeLong, C., Tanner, D., Jull, M.J., 1993. A field guide for site
identification and interpretation for the southwest portion of the
Prince George Forest Region. BC Min. For., Victoria, B.C.
Land Management Handbook No. 20.
Despain, D.G., 1983. Nonpyrogenous lodgepole pine communities
in Yellowstone National Park. Ecology 64, 321±334.
Duchesne, L.C., 1994. Fire and diversity in Canadian ecosystems.
In: Boyle, T.J.B., Boyle, C.E.B. (Eds.), Biodiversity, Temperate
Ecosystems, and Global Change. Springer, Berlin. pp. 247±253.
Eberhart, K.E., Woodward, P.M., 1987. Distribution of residual
vegetation associated with large fires in Alberta. Can. J. For.
Res. 17, 1207±1212.
Eremko, R.D., 1990. A review of natural lodgepole pine
regeneration in the Chilcoltin. British Columbia: Min. For.
FRDA Rep. 126, Victoria.
Foster, D.R., 1983. The history and pattern of fire in the boreal
forest of southeastern Labrador. Can. J. Bot. 61, 2459±2471.
Franklin, J.F., 1994. Ecosystem management: An overview?
Ecosystem Management: Applications for sustainable forest
and wildlife resources. March 3±4, 1994. Stevens Point, WI.
Gasaway, W.C., DuBois, S.D., 1985. Initial response of moose to a
wildfire in interior Alaska. Can. Field-Nat. 99, 135±140.
Goggans, R., Dixon, R.D., Claire, L.C., 1988. Habitat use by three-
toed and black backed woodpeckers, Deschutes National
Forest, Oregon. Nongame Project No. 87±3±02. Oregon Dept.
of Fish and Wildlife, Portland, OR.
Hansen, A.J., Spies, T.A., Swanson, F.J., Ohmann, J.F., 1991.
Conserving biodiversity in managed forests: lessons from
natural forests. Bioscience 41, 382±392.
Harmon, M.E., Franklin, J.F., Swanson, F.J., Sollins, P., Gergory,
S.V., Lattin, J.D., Anderson, N.H., Cline, S.P., Aumen,
N.G., Sedell, J.R., Cummins, K.W., 1986. Ecology of coarse
woody debris in temperate ecosystems. Adv. Ecol. Res. 15,
133±302.
Johnson, E.A., 1992. Fire and Vegetation Dynamics: Studies from
the North American Boreal Forest. Cambridge University
Press, Cambridge.
Kneewshaw, D.D., 1992. Tree population dynamics of some old-
growth Sub-boreal Spruce stands. M. Sc. Thesis, University of
British Columbia.
Koehler, G.M., Blakesley, J.A., Koehler, T.W., 1990. Martin use of
successional forest stages during winter in north-central
Washington. Northwest Nat. 71, 1±4.
Krajina, V.J., Klinka, K., Worrall, J., 1982. Distribution and
ecological characteristics of trees and shrubs of British
Columbia. University of British Columbia, Vancouver, BC,
131 pp.
Lambert, R.L., Lang, G.E., Reiners, W.A., 1980. Loss of mass and
chemical change in decaying boles of a subalpine balsam fir
forest. Ecology 61, 1460±1473.
Lofroth, E., 1992. Measuring habitat elements at the stand level. In
Proc. Methodologies for monitoring wildlife diversity in BC
forests. Wildlife Branch. Ministry of Environment, Victoria.
Lotan, J.E., 1967. Cone serotiny of lodgepole pine near West
Yellowstone. Montana. For. Sci. 13, 55±59.
Lotan, J.E., 1974. Cone serotiny-fire relationships in lodgepole
pine. Proc. Ann. Tall Timbers Fire Ecol. Conf. 14, 267±278.
106 S.C. DeLong, W.B. Kessler / Forest Ecology and Management 131 (2000) 93±106
Luttmerding, H.A., Demarchi, D.A., Lea, E.C., Meidinger, D.V.,
Vold, T., 1990. Describing ecosystems in the field. Min. of Env.
MOE Manual 11. Victoria, BC.
Marshall, D.B., 1992. Status of the black-backed woodpecker in
Oregon and Washington. Audubon Soc. of Portland, May 26±29,
1992. Blue Mountains Biodiversity Conference, Portland, OR.
Neitro, W.A., Mannan, R.W., Taylor, D., Binkley, V.W., Marcot,
B.G., Wagner, F.F., Cline, S.P., 1985. Snags (Wildlife trees) In:
Brown, E.R. (Tech. Ed.) Management of Wildlife and Fish
Habitats in Forests of Western Oregon and Washington. USDA
Forest Service, Pacific Northwest Forest Service, Oregon,
pp. 129±170.
Oke, T.R., 1978. Boundary Layer Climates. Wiley, New York.
Parker, A.J., 1986. Persistence of lodgepole pine forests in the
central Sierra Nevada. Ecol. 67, 1560±1567.
Rowe, J.S., Scotter, G.W., 1973. Fire in the boreal forest. Quat.
Res. 3, 444±464.
Sokal, R.R., Rohlf, F.J., 1969. Biometry. Freeman, San Fransisco.
Spies, T.A., Franklin, J.F., Thomas, T.B., 1988. Coarse woody
debris in Douglas-fir forests of western Oregon and Washing-
ton. Ecology 69, 1689±1702.
Spies, T.A., Franklin J.F., 1991. The structure of natural young,
mature, and old-growth Douglas-fir forests in Oregon and
Washington. In: Ruggiero, L.F., Aubry, K.B., Carey, A.B., Huff,
M.H. (Tech. Coord.), Wildlife and Vegetation of Unmanaged
Douglas-fir forests. USDA For. Serv., Pacific Northwest
Research Station, Oregon, pp. 91±110.
Stordeur, L.A., 1986. Marten in British Columbia with implications
for forest management. BC Ministry of Forests, WHR-25.
Victoria, BC.
Thomas, J.W., Anderson, R.G., Maser, C., Bull, E.L., 1979. Snags
In: Thomas, J.W. (Tech. Ed.), Wildlife habitat in managed
forests, the Blue Mountains of Oregon and Washington.
USDA Forest Service Handbook. No. 553, Oregon, pp. 60±
77.
Trowbridge, R., Hawkes, B., Macadam, A., Parminter, J., 1989.
Field handbook for prescribed fire assessments in British
Columbia. FRDA Handbook. 001. Victoria, BC.
Van Wagner, C.E., 1978. Age-class distribution and the forest fire
cycle. Can. J. For. Res. 8, 220±227.
Whelan, R.J., 1995. Ecology of Fire. Cambridge University Press,
Cambridge.
Wilkinson, L., Blank, G., Gruber, C., 1996. Desktop Data Analysis
with SYSTAT. Prentice Hall, New Jersey.
Zackrisson, O., 1977. Influence of forest fires on the North Swedish
boreal forest. Oikos 29, 22±32.