Salvage logging after windthrow
alters microsite diversity,
abundance and environment,
but not vegetation
CHRIS J. PETERSON1* and ANDREA D. LEACH2
1 Department
of Plant Biology, University of Georgia, 2502 Plant Science Building, Athens, GA 30602, USA
of Cellular and Molecular Biology, University of Georgia, Athens, GA 30602, USA
*Corresponding author. [email protected]
2 Department
Summary
An increasing number of researchers propose that disturbance effects in forests are mediated through
‘legacies’, which are organisms or organically derived structures that persist after a disturbance. Much
controversy currently surrounds the potential impact of post-disturbance salvage logging, in part,
because of the potential for such actions to alter post-disturbance legacies and thus forest regeneration.
Microsites (e.g. downed tree crowns, boles, stumps, treefall pits and mounds) created by natural
disturbances are a subset of the broader concept of legacies, but the effect of windthrow + salvage
logging on microsites and their environment and vegetation has not been previously examined. In a
wind-damaged forest in western Tennessee, USA, we documented microsite diversity and abundance,
environmental conditions, and initial vegetation regeneration in salvaged and unsalvaged areas. We
found that salvaged areas had significantly greater variety of microsites, altered microsite abundance,
higher soil temperature and greater canopy openness relative to unsalvaged areas. However, 2 years
after the storm, herbaceous cover and species richness and tree seedling density and species richness
did not differ between salvaged and unsalvaged areas. Soil moisture also was unaffected by salvaging.
In contrast, environmental conditions and vegetation characteristics differed significantly among
microsite types, with treefall mounds being warmer and drier than other microsites. This intermediateseverity wind disturbance, followed by moderate intensity of salvaging, created microsites that differed
in environment and vegetation, and although the salvaging altered microsite diversity, abundances
and conditions, the initial vegetation did not show detrimental effects of the salvage operations. We
suggest that primary determinants of the consequences of salvaging after natural disturbance are the
severity of the natural disturbance, and intensity of salvage operations. Detrimental effects of salvaging
may accrue only if some combined severity threshold is exceeded.
Introduction
Recent natural disturbances (e.g. extremely large
wildfires in numerous locations worldwide) and
controversial research findings (Donato et al.,
© Institute of Chartered Foresters, 2008. All rights reserved.
For Permissions, please email: [email protected]
2006; Newton et al., 2006; Baird, 2006) have
brought the discussion of post-disturbance salvage logging to a heated intensity (Floyd, 2006).
Currently, most forest areas subject to large natural disturbances (e.g. fire or wind) subsequently
Forestry, Vol. 81, No. 3, 2008. doi:10.1093/forestry/cpn007
Advance Access publication date 19 March 2008
362
FORESTRY
have salvage harvesting activities that aim to recoup financial losses from the natural disturbance
(Lindenmayer et al., 2004). These salvage logging
operations are widespread and common, but we
know very little about how such management actions influence forest regeneration, biodiversity
and risk of later disturbance.
One of the potential influences of salvage logging on forest recovery and diversity might occur
via alterations in the variety and abundance of
microsites (Beatty, 1984; Peterson et al., 1990;
Kuuluvainen and Juntunen, 1998), which are
one component of the broader concept of biotic
legacies (Franklin et al., 2000; Lindenmayer,
2006). Lindenmayer et al. (2004, see also Foster
and Orwig, 2006; Lindenmayer and Noss, 2006)
suggested that one way disturbances maintain
ecosystem diversity and function is through the
variety of legacies resulting from disturbances,
and that salvaging reduces legacy diversity. It is
indeed well known that differing microsites in
wind-disturbed forests have distinct environments (Beatty, 1984; Beatty and Sholes, 1988;
Carlton and Bazzaz, 1998a), and differentially
influence tree seedling germination, growth and
vulnerability to herbivores (Long et al., 1998;
Peterson and Pickett, 1995, 2000; Krueger and
Peterson, 2006). These effects often result in
vegetation variation among microsites, often
with greater abundance of pioneer species in the
more severely disrupted microsites (e.g. Peterson et al., 1990; Peterson and Campbell, 1993),
although such among-microsite variation is not
always found (Webb and Scanga, 2001). Thus,
decreases in the variety or abundance of microsites might decrease forest regeneration potential or species diversity in the regenerating area.
However, we know of no studies of microsite
abundances, composition or diversity after natural disturbance and salvaging.
The existing information about consequences
of post-fire or post-windthrow salvaging in North
America are mostly restricted to the north-west
(e.g. Donato et al., 2006) or boreal sites in Canada
(Greene et al., 2006; Macdonald, 2007). Donato
et al. (2006) found reduced tree seedling density
and diversity in salvaged areas of the 2002 Biscuit
Fire in Oregon, and Greene et al. (2006) report
that although asexual reproduction by Populus
was unaffected by salvaging in central Quebec,
the regeneration from seed by conifers was much
reduced in salvaged areas. To our knowledge,
Elliott et al. (2002) and Rumbaitis Del Rio (2006)
represent the only studies available on the effect
of windthrow + salvaging on regeneration. Consequently, there is yet exceedingly little information about forest response to salvaging after wind
disturbance, particularly in eastern North America. Therefore, we report here on the microsite
abundances and very early regeneration following moderate windthrow in hardwood forests of
central Tennessee, USA. Following Lindenmayer
et al. (2004), we hypothesized that salvaging
might destroy treefall mounds and remove tree
trunks and crowns in addition to exposing bare
soil in skid trails and depositing pruned branches
in slash piles. Therefore we expected that salvaging reduces microsite diversity and alters
microsite abundance (hypothesis 1). We further
expected that since salvaging removes downed
stems and crowns potentially capable of sprouting and surviving, it would increase canopy openness at ground level (hypothesis 2). Finally, we
expected that microsites with exposed mineral
soil (pits, mounds, bare soil patches) and more
open conditions (salvaged areas) would contain
a higher proportion of shade-intolerant, earlysuccessional species (hypothesis 3) and higher
species richness (hypothesis 4).
Materials and methods
Study site
This research was conducted at Natchez Trace
State Park and Forest (NTSF), an 18 245-ha
natural area located in west-central Tennessee
(35º N 88º W). The area lies within the East Gulf
Coastal Plain section of the Coastal Plain physiographic province (Braun, 1950). Braun (1950)
classified the vegetation of western Tennessee as
belonging to the Mississippi Embayment section
of the Western Mesophytic Forest Region. The
topography consists of gently rolling uplands
separated by broad floodplains; elevation averages 165 m above sea level, ranging from 137 to
182 m (Smalley, 1991). Soils are derived from the
McNairy sands geologic formation and consist of
well-drained fine sandy loams and silt loams with
occasional fragipans and a discontinuous loess
cap (Kupfer and Franklin, 2000). The climate is
humid continental, with short, mild, wet winters
SALVAGING AFTER WIND DISTURBANCE
and long, hot, dry summers and an average growing season of 202 days. Mean annual precipitation is 1240 mm, mostly rain falling in the late
winter and early spring (Flowers et al., 1960).
Much of the park was settled and cleared for
agriculture in the 1800s and early 1900s, but the
land was abandoned in 1935 and was deeded to
the state of Tennessee in 1955 (Smalley, 1991).
The second-growth forests consist of Quercus
species (oaks), Carya species (hickories) and other
mixed mesophytic species, including Pinus taeda
L. (loblolly pine), in the uplands. The understory
consists primarily of Cornus florida L. (flowering
dogwood), Nyssa sylvatica Marsh. (blackgum)
and Sassafras albidum (Nutt.) Nees. (sassafras).
On 5 May 1999, a downburst (straight-line
thunderstorm) damaged ~3000 ha at NTSF (Roy
Ward, State Forest Supervisor, personal communication). As recorded in nearby Lexington,
Tennessee, USA, the thunderstorm lasted over
1.5 h with sustained wind speeds over 90 km
h−1 and produced wind gusts exceeding 145 km
h−1 (NOAA, 1999). Forest damage was patchy,
and basal area losses ranged from 26.6 per cent
to 49.5 per cent in our study area. All of the
‘downed’ trees were harvested with the exception
of two areas (6 ha each, ~2.5 km apart) that were
set-aside for research purposes and left unsalvaged. Because two pairs of salvaged/unsalvaged
locations were available at NTSF, we sampled at
both locations and refer to them as Site A and
Site B. Salvage operations removed 105 stems (73
stems ha−1) and 10.88 m2 (7.55 m2 ha−1) of basal
area in the areas that we sampled.
Sampling design
During June–August 2001, sampling plots were
established within salvaged (S) and unsalvaged
(U) portions of wind-damaged forest at Sites A
and B (‘stands’ refers to the four site × treatment
combinations, i.e. unsalvaged portion of Site
A = ‘UA’, etc.). Eight large (30 m × 30 m) plots
were used to quantify forest damage in each of
the four stands. In the strictest sense, plots may
not qualify as replicates because they are nested
within stands, and salvage treatments were assigned haphazardly rather than randomly to
stands. However, we perform our statistical analyses below as if plots were true replicates; at the
scale of this study (e.g. stands cover many tens of
363
hectares), perfect randomization and replication
would be difficult to achieve.
To quantify relative abundances of microsite
types, we sampled along four parallel 30-m lines
10 m apart in each large plot. At 0.5-m intervals along these lines, we dropped a pin and recorded microsite information, yielding a total of
244 points per plot (244 points × 32 plots; n =
7,808 sampling points). We classified points into
13 microsite categories on the basis of the substrate conditions and the nature of overlying materials; this classification is the basis for Figure
2, but does not utilize vegetation in the microsite
definition (see ‘focal microsites’ below). Substrate
was recorded as organic or mineral, and as intact
or disrupted. If disrupted, the disruption was recorded as natural (treefall pits or mounds, other
natural disruption) or anthropogenic (skid trails,
other anthropogenic). Any material overlaying
the substrate was recorded as leaves, needles, fine
woody debris (includes the distal portion of tree
crowns and fine debris piles in salvaged areas),
coarse woody debris (includes tree trunks), moss
or nothing (bare). We also recorded the identity
of any vascular plants touching the pin below
1.5 m (ferns, sedges, grasses, to genus for shrubs
and to species for trees). From this inventory,
we calculated the total absolute area covered by
each of the following microsite types: intact bare
mineral soil, intact bare organic soil, treefall root
pits, treefall mounds, other naturally disturbed
soil (e.g. eroded stream bank), intact leaf-covered
soil, intact needle-covered soil, fine woody debris,
coarse woody debris, stumps, moss, skid trails
and other anthropogenically disturbed soil.
We investigated the vegetation and environmental conditions in detail within the five most
common wind-related microsite types (focal microsites): treefall pits, treefall mounds, downed
tree crowns, patches of bare soil (intact soil,
sparse litter, no vegetation) and Vitis rotundifolia patches. Note that in contrast to the general
microsite abundance inventory above, here we
include a vegetation-defined microsite: Vitis rotundifolia was the most locally abundant groundlayer vegetation and therefore used as a vegetated
microsite category. Our aim was to use microsite types similar to those used in other studies
(Nakashizuka, 1989; Carlton and Bazzaz, 1998b;
Kuuluvainen and Juntunen, 1998). Because most
crowns were removed from salvaged areas, we
364
FORESTRY
substituted slash piles for tree crowns in salvaged
areas when necessary since they were the closest
analog to downed tree crowns.
To examine groundlayer (<2 m tall) vegetation
and environmental characteristics of microsites,
we established 0.5-m2 circular quadrats in the
five microsite types within each large plot. Quadrats were placed in up to three replicates of each
of the five microsite types in a stratified random
manner, although three replicates were not always
available for all five microsite types in each plot.
The total number of microsite quadrats was 369
(see Table 1, for microsite quadrat frequencies).
Within each quadrat, we recorded the species
identity and measured the height of all woody
stems up to 2 m tall, and visually estimated per
cent cover of herbs and vines. Per cent cover was
noted for all plant material overhanging the quadrat, whether or not the stem was rooted in the
plot (thus total cover could exceed 100 per cent).
When species identities were unknown, voucher
specimens were collected for identification. Species nomenclature follows the USDA Plants list
(USDA and NRCS, 2002).
We characterized canopy openness using hemispherical photography (Robison and McCarthy,
1999). Photographs were taken at 1.25 m above
the center of each quadrat in late June 2002 using
a Nikon 885 Coolpix digital camera fitted with
an FC-E8 fisheye converter lens (Nikon Corporation, Tokyo, Japan). We used a gimbal mounted
on a monopod to level the camera to a zero zenith
angle and oriented it to true north prior to taking
photos. All photographs were taken in the early
morning and early evening hours to minimize
the solar glare that distorts the estimates of true
canopy cover. Per cent canopy openness was determined from the digital images using Gap Light
Analyzer software (Frazer et al., 1999), which
divides pixels into sky and non-sky classes and
calculates canopy openness based on per cent sky
pixels relative to non-sky (canopy) pixels (Frazer
et al., 1999).
We measured soil temperature and soil moisture simultaneously on two occasions in June
2002. Within each quadrat, we measured temperature of the upper 15 cm of soil at three points
using Reotemp stem thermometers that were allowed to equilibrate for 45 s. Per cent soil moisture (volumetric water content) of the top 12
cm of soil was determined using a Hydrosense
time-domain reflectometer (Decagon Devices,
Table 1: MANOVA and univariate ANOVAs for treatment and microsite effects on environmental variables
Source of variation
MANOVA
Treatment
Microsite
Treatment × microsite
ANOVA
Temperature
Treatment
Microsite
Treatment × microsite
Error
Moisture
Treatment
Microsite
Treatment × microsite
Error
Per cent openness
Treatment
Microsite
Treatment × microsite
Error
Num df
Den df
Pillai’s trace
F
P
3
12
12
357
1077
1077
0.368
0.384
0.054
Type III SS
69.41
13.19
1.65
<0.0001
<0.0001
0.0729
1
4
4
359
0.012
0.020
0.001
0.049
84.40
35.77
1.61
<0.0001
<0.0001
0.1705
1
4
4
359
0.000
2.178
0.048
11.133
0.00
17.55
0.39
0.9880
<0.0001
0.817
1
4
4
359
6.441
0.296
0.606
14.579
158.61
1.82
3.73
<0.0001
0.1243
0.006
df = Degrees of freedom; Num = numerator; Den = denominator.
SALVAGING AFTER WIND DISTURBANCE
Inc., Pullman, Washington, USA) at three points
within each microsite quadrat. All measurements
were taken within 2 h of solar noon. Because soil
temperature and moisture can vary dramatically
temporally and spatially, readings were averaged
within and between sampling periods.
Statistical analysis
Microsite diversity and abundance Because most
plots were represented by 244 pin-drop points
classified into the 13 microsite categories, we first
calculated richness (number of microsite categories) and diversity (Shannon’s H’) of microsites
per plot. These values were then compared among
site/treatment combinations to test for a treatment
(salvaging) effect, using site (A or B) as a blocking variable and n = 8 plots as replicates. Because
natural windthrow levels were not equal across all
plots, and severity of natural disturbance is likely
to influence microsite characteristics, we used the
severity of natural disturbance (proportion of predisturbance trees fallen) as a covariate.
We examined microsite ‘composition’ of plots
in two ways: in an exploratory step, we performed detrended correspondence analysis ordination (McCune and Mefford, 1999) to reveal
similarity and differences among plots of various
treatments in terms of their microsites. To test for
differences in relative abundances, we performed
G-tests of homogeneity among treatments, using
data from all plots pooled within a treatment.
Abiotic conditions To determine whether environmental conditions (canopy openness, soil
temperature and soil moisture) varied predictably
among microsite types and/or treatments, we used
multivariate analysis of variance (MANOVA).
This was done because when dependent variables are correlated, it is more appropriate to use
MANOVA than multiple univariate ANOVAs;
MANOVA controls for Type I error (which is
inflated when conducting multiple univariate
ANOVAs) and adjusts for correlations between
dependent variables. Pillai’s trace test statistic is
extremely robust and its use is suggested when designs are unbalanced, sample sizes are small and
covariances are non-homogeneous. We analyzed
light, temperature and moisture simultaneously
using a 2 (treatment) × 5 (microsite type) facto-
365
rial MANOVA. Environmental data were logtransformed to meet assumptions of normality
and homogeneity of variances. Significant main
effects or interactions from MANOVA were examined more closely with univariate two-factor
ANOVAs.
Vegetation We conducted separate two-way
(five microsite types × two treatments) ANOVAs on herbaceous cover and species richness,
and woody seedling density and richness. Data
were log10(x + 1) or square root transformed
to meet normality and equal variance assumptions. Significant ANOVAs were followed by
Tukey’s studentized range test (HSD) multiple
comparison procedure to determine which
means differed significantly from one another.
Correlations, MANOVA and ANOVAs were
performed using SAS statistical software (SAS
Institute, 1995).
We conducted G-tests of independence to
test for differences among treatments in relative
abundances of microsite types. We also assigned
all tree species to one of three shade-tolerance
classes (tolerant, intermediate or intolerant) based
on information in Burns and Honkala (1990) and
Brown and Kirkman (1990) and conducted Gtests of independence to determine whether successional status (based on proportions of tolerant,
intermediate or intolerant) of the seedling layer
or abundance of the five most common seedling
species differed between salvaged and unsalvaged
areas or among the five microsite types (Sokal and
Rohlf, 1995). Species without published tolerance
values were excluded from analysis, but because
these were rare, their exclusion is unlikely to have
influenced conclusions.
We used canonical correspondence analysis
(CCA) to examine similarities in species composition among quadrats and correlations between
compositional variation and resource levels
(Palmer, 1993). Herbs and woody seedlings were
analyzed separately because we used different
measures of abundance (percent cover for herbaceous vegetation and height for woody seedlings). Quadrats not containing any herbaceous
or woody seedling species or lacking complete
environmental information were removed from
analysis. We included 344 of the 369 quadrats in
the herbaceous analysis and 208 in the seedling
analysis. Environmental variables used in both
366
FORESTRY
analyses include soil temperature, soil moisture
and per cent canopy openness; all environmental variables were log-transformed to reduce the
influence of outliers (Palmer, 1993). We used
PC-Ord software (McCune and Mefford, 1999)
to conduct the CCAs, and all environmental variables (soil temperature, soil moisture and per
cent canopy openness) were log-transformed to
reduce the influence of outliers. In the following, all reports of significant tests are significant
at P < 0.05.
Results
Microsite diversity
The richness of microsite types per plot ranged
from 3 to 11 and was significantly greater in salvaged than in unsalvaged areas (Figure 1) when
controlling for blowdown severity (ANCOVA,
site was not significant). Evenness of microsites
was significantly greater in salvaged than unsalvaged areas (ANCOVA, block was again not
significant), and microsite diversity measured as
H′ was marginally greater in salvaged areas than
in unsalvaged. Considering all 32 plots, simple
linear regression of severity of wind damage vs
richness of microsites produced a significant
positive relationship (R2 = 0.260, n = 32).
Microsite abundance
Intact soil with leaf litter cover was the most
abundant of the microsites, followed by fine
woody debris/tree crown microsites (Figure 2,
abundances based on soil and debris microsites,
and thus do not consider vegetation microsites).
Among the five focal microsites that were studied in more detail (and which include vegetationdefined microsites), patches of Vitis and downed
tree crown microsites were the most abundant,
accounting for 58.1 per cent and 22.8 per cent
of 2450 sampling points falling within one of
the five focal microsites. Bare soil patches (9
per cent), mounds (6 per cent) and pits (4.1 per
cent) combined accounted for 19.1 per cent of
total coverage. The relative abundances of the
five focal microsite types differed significantly
between treatments (G-test). Although Vitis microsites accounted for the greatest overall cover-
Figure 1. Richness of microsites across four stand
types. Mean ± standard deviation of number of microsites encountered in n = 8 plots per stand type.
Sal = salvaged; Uns = unsalvaged.
age, there was more coverage by Vitis microsites
in unsalvaged areas than in salvaged areas (data
not shown). In salvaged areas, rank order of
microsite abundance was Vitis, bare, crown,
mound and pit microsite types (in descending
order), whereas in unsalvaged areas, it was Vitis,
crowns, mounds and pits, and bare soil microsites. Salvaged areas contained a much higher
percentage of bare soil microsites (24.8 per cent
vs 0.7 per cent) and mound microsites (12.2
per cent vs 2.7 per cent) than unsalvaged areas.
However, unsalvaged areas contained more
crown coverage (27 per cent vs 14.7 per cent)
than salvaged areas.
Microsite environmental conditions
Soil temperature and soil moisture were negatively correlated (r = −0.324, n = 369) while temperature and canopy openness were significantly
positively correlated (r = 0.397). Although moisture and canopy openness were not correlated
(r = 0.032), a MANOVA was appropriate because both were correlated with temperature.
Though the sampling design included two sites
(statistical blocks), conducting separate MANOVAs for each site gave similar results. Therefore, we
pooled data from the two sites to increase statistical power (n = 369 quadrats). The two-treatment
(salvaged vs unsalvaged) × five microsite type
SALVAGING AFTER WIND DISTURBANCE
367
Figure 2. Relative abundance of soil + debris—defined microsites across four stand types. Abundance based
on 244 pin drops in each of eight 30 m × 30 m plots per stand type (1952 points); a few points were inaccessible and therefore no data were recorded. Sal = salvaged; Uns = unsalvaged; min = mineral; CWD = coarse
woody debris; FWD = fine woody debris.
MANOVA revealed a significant multivariate
main effect for both treatment and microsite
type on environmental variables, but no effect
for treatment × microsite interaction (results for
MANOVA and ANOVAs are summarized in
Table 1). Thus, environmental differences among
microsites are consistent between the salvaged
and unsalvaged treatments.
Because the MANOVA revealed significant
multivariate main effects, we examined the responses of the three environmental variables separately in univariate ANOVAs. Soil temperature
varied significantly among both treatment and
microsite type (Table 1). Soil temperature was
higher in salvaged areas than in unsalvaged areas
(Figure 3). Mounds were warmer than all other
microsite types, bare soil patches were intermediate in temperature and pits, crowns, and Vitis
patches were coolest (Figure 4). Soil moisture did
not differ between treatments (Figure 3) but did
significantly differ among microsite types, with
pits having the wettest soils, mounds having the
driest soils and bare, Vitis, and crown microsites
having intermediate soil moisture levels (Figure
4). While canopy openness was higher in salvaged
areas than in unsalvaged areas (Figure 3), it did
not differ between microsite types (Figure 4).
Microsite vegetation
Groundlayer vegetation Total amount of herbaceous cover differed between microsite types
but did not differ between treatments (Figure 5);
Vitis microsites had the highest cover, and crown,
open and pit microsites had the lowest. However,
because ANOVA revealed a significant treatment
× microsite type interaction, differences in herbaceous cover among microsite types were inspected
within each treatment separately. In the salvaged
areas, Vitis (as expected) had the highest per cent
cover followed by the other microsite types, which
did not differ in amount of cover. However, in unsalvaged areas, herbaceous cover was significantly
higher in mounds and Vitis patches, compared
with crowns, pits and bare microsites, which did
not differ (Figure 5). Herbaceous species richness
differed between microsite types but did not differ between treatments (Figure 6), nor was there
a significant treatment × microsite interaction.
Species richness of herbaceous vegetation was
highest in bare soil microsites and lowest in
crown microsites (Figure 6).
Seventy-five groundlayer species were present
in the microsite quadrats. Species with the highest
overall cover included V. rotundifolia (41.3 per
cent), Parthenocissus quinquefolia (10.8 per cent),
368
FORESTRY
Figure 3. Resource levels by treatment. (A) Soil
temperature. (B) Soil moisture. (C) Per cent canopy
openness. Mean ± standard deviation. Different letters denote significant differences at level α = 0.05.
Smilax spp. (5.3 per cent; includes Smilax glauca
and Smilax rotundifolia) and Lonicera japonica
(4.1 per cent). Disturbance specialists or pioneer
species—Rubus spp. (includes Rubus argutus and
Rubus flagellaris), L. japonica, S. glauca, Toxicodendron radicans, and Pueraria montana—were
more abundant in salvaged areas while P. quinquefolia and V. rotundifolia were more abundant
in unsalvaged areas.
The CCA of environmental conditions and
groundlayer vegetation revealed substantial sepa-
Figure 4. Resource levels by focal microsite. (A) Soil
temperature. (B) Soil moisture. (C) Per cent canopy
openness. Mean ± standard deviation. Different letters denote significant differences at level α = 0.05.
ration of salvaged and unsalvaged quadrats along
Axis 1 (data not shown), but the eigenvalue for
the first axis was only 0.257, accounting for a
meager 1.1 per cent of the variance in species
composition.
Tree seedlings Total tree seedling density (all
species pooled) was highest in bare soil and pit
microsites and lowest on mounds (Figure 7)
while crown and Vitis microsites exhibited intermediate density levels; these differences among
SALVAGING AFTER WIND DISTURBANCE
369
Figure 6. Groundlayer species richness by (A) treatment and (B) focal microsite. Least square mean ±
standard error. Different letters denote significant
differences at level α = 0.05.
Figure 5. Per cent cover of groundlayer vegetation
by (A) treatment, (B) focal microsite in salvaged
areas and (C) focal microsite in unsalvaged areas.
Least square mean ± standard error. Different letters
denote significant differences at level α = 0.05.
microsites were highly significant. Seedling density did not differ between treatments, and there
was no treatment × microsite interaction. Tree
seedling species richness was highest in bare soil
and pit microsites and lowest on mounds (Figure
8), and crown and Vitis having intermediate species richness; these among-microsite differences
were again highly significant. Tree seedling rich-
ness did not differ between treatments (Figure 8),
nor was there a significant treatment × microsite
interaction.
The microsite quadrats contained seedlings of
28 tree species. There were higher percentages of
shade-tolerant seedlings (29.6 per cent vs 23.8
per cent) and intolerant seedlings (60.9 per cent
vs 53.9 per cent) in salvaged areas (G-test; Table
2) than in unsalvaged areas, whereas there were
more intermediates in unsalvaged areas (22.6 per
cent vs 9.4 per cent). Several shade-intolerant
species such as P. taeda, Quercus stellata, Betula
nigra and Ulmus alata were entirely absent from
microsites in unsalvaged areas. Oaks, particularly Quercus alba and Quercus velutina, accounted for most mid-tolerant species (Table 2),
but since oaks accounted for a higher proportion of pre-disturbance canopy in the unsalvaged
370
FORESTRY
Figure 7. Tree seedling density by (A) treatment and
(B) focal microsite. Least square mean ± standard
error. Different letters denote significant differences
at level α = 0.05.
Figure 8. Tree seedling species richness by (A) treatment and (B) focal microsite. Least square mean ±
standard error. Different letters denote significant
differences at level α = 0.05.
areas, we cannot attribute this finding solely to
salvaging.
G-tests of shade-tolerance categories revealed
that proportions of seedlings in these categories
differed significantly among microsites. Mounds
had the highest percentage of shade-intolerant
seedlings (65.5 per cent) followed by bare (63 per
cent) and pit (53.5 per cent) microsites (Table 2).
Crowns had the highest percentage of seedlings
intermediate in shade tolerance (29.2 per cent)
followed by bare soil microsites (21.2 per cent).
Pits (44.6 per cent) and mounds (29.1 per cent)
had the highest percentage cover by shade-tolerant
seedlings.
Liriodendron tulipifera accounted for 16.1
per cent of total seedling density (Table 3) followed by Q. alba (13.4 per cent), S. albidum
(12.2 per cent), Acer rubrum (12.0 per cent)
and Prunus serotina (11.2 per cent). Of these
common species, Q. alba and S. albidum were
most abundant in unsalvaged areas, whereas
A. rubrum and L. tulipifera were most abundant
in salvaged areas (significant G-test; Table 3).
Pinus taeda, Q. stellata, L. tulipifera and Liquidambar styraciflua were often found in skid trails
(A. D Leach, personal observation).
In pit microsites, L. tulipifera (51.7 per cent of
all seedlings) and A. rubrum (42.5 per cent) seedlings were the most abundant while Q. alba (42.5
per cent) was the most abundant within crown
microsites, S. albidum (41.9 per cent) within Vitis
microsites and P. serotina (27.6 per cent) within
bare soil microsites. The abundances of these five
individual species differed significantly among
microsites (G-test, Table 3).
The CCA of seedling and environmental data
(data not shown) revealed strong separation of
quadrats by treatment, but the total variance
SALVAGING AFTER WIND DISTURBANCE
371
Table 2: Frequencies of tree seedlings in different shade-tolerance categories, within treatments and microsite
types
Tolerant
Intermediate
Intolerant
Total
(A) Treatment
Unsalvaged
Salvaged
Total
60
69
129
57
22
79
135
142
277
252
233
485
(B) Microsite
Bare
Crown
Mound
Pit
Vitis
Total
29
18
16
45
21
129
39
21
3
2
14
79
116
33
36
54
38
277
184
72
55
101
73
485
G-tests
G2 = 16.1, df = 2,
P = 0.0003
G2 = 58.62, df = 8,
P = < 0.0001
df = Degrees of freedom.
Table 3: Frequencies of the five most common tree seedling species within microsite quadrats
LITU
QUAL
SAAL
ACRU
PRSE
Total
(A) Treatment
Unsalvaged
Salvaged
Total
36
43
79
49
17
66
46
14
60
14
45
59
45
10
55
190
129
319
(B) Microsite
Bare
Crown
Mound
Pit
Vitis
Total
32
3
12
31
1
79
33
17
3
1
12
66
23
10
8
1
18
60
17
3
6
24
9
59
40
7
2
3
3
55
145
40
31
60
43
319
G-tests
G2 = 64.3, df = 4,
P < 0.0001
G2 = 132.16, df = 16,
P = < 0.0001
Abbreviations are first two letters of genus and species name (Liriodendron tulipifera, Quercus alba, Sassafras
albidum, Acer rubrum, and Prunus serotina); df = degrees of freedom.
explained was again very small. Axis 1 was heavily influenced by canopy openness (r = 0.926).
Quadrats in salvaged areas containing lightdemanding species, such as Quercus coccinea,
Q. stellata and P. taeda, were separated from unsalvaged quadrats with lower canopy openness
and dominance by more shade-tolerant species,
such as C. florida, Fagus grandifolia and N. sylvatica. Nevertheless, the eigenvalue for the first
axis was again small at only 0.295, accounting for
1.8 per cent of the variance in the species data, so
we make no further interpretation of this result.
Discussion and conclusions
In this study, we documented greater microsite richness in salvaged areas compared with unsalvaged
windthrow areas, along with significant changes in
microsite relative abundances and microsite environments among treatments. However, vegetation
did not significantly respond to salvage activities
in this study 2 years after the disturbance. Indeed,
we found greater variation among microsites than
between treatments for both environmental conditions and regeneration vegetation. We are not
372
FORESTRY
aware of any previous studies that have reported
(unconfounded) effects of salvage logging in winddamaged forests, although a number of previous
studies document microsite effects (Nakashizuka,
1989; Peterson et al., 1990; Clinton and Baker,
2000; Harrington and Bluhm, 2001). In a related
paper (Peterson and Leach, 2008), we report what
we believe to be the first attempt to quantitatively
relate combined (natural blowdown + anthropogenic salvaging) disturbance severity to vegetation
response at the plot level.
It is not surprising that environmental conditions within individual microsites differ, due to
their unique modes of formation and topography.
Peterson et al. (1990) found that soil moisture
and light levels differed between intact soil and
pit microsites. We found significant microsite effects for soil temperature and moisture; treefall
mounds were significantly warmer and drier than
pits, which were cooler and moister (Figure 4).
In contrast to the present study, though, Peterson
and Pickett (1995) found that light and temperature were greatest in open (=bare: intact soil, no
litter or vegetation) and mound microsites but that
moisture did not differ. We suspect that the lesser
wind damage severity and therefore lesser canopy
openness in the present study, compared with
Peterson and Pickett (1995), may have limited the
among-microsite differences that have been documented in more severely wind-damaged sites.
Most studies of wind disturbance that consider microsites have focussed on differential
seedling establishment and survival between pits
and mounds. The pits and mounds created by
uprooted trees often increase or maintain tree
species diversity by supporting species that are
rare in undamaged forest (Peterson et al., 1990)
and may be important in determining future species composition. We found that pits had greater
tree seedling density (Figure 7) and species richness (Figure 8) than mounds, but in unsalvaged
areas, mounds had higher herbaceous cover (Figure 5). Herbaceous species richness did not differ between pits and mounds (Figure 6). Peterson
and Pickett (1990) found that pits had greater
species richness and diversity, total biomass and
stem abundance per square metre than mounds
because mounds are more susceptible to erosion and are subject to more extreme temperatures (Beatty, 1984; Carlton and Bazzaz, 1998a).
Peterson and Pickett (1990) proposed that the
primary factor affecting differential establishment was higher moisture levels in pits, and this
may also be true for pits at NTSF.
In contrast to our findings, Nakashizuka (1989)
found that seedling richness and diversity were
highest on mounds in old-growth temperate forests in Japan because mounds provided competition-free growing space for tree species with small,
wind-dispersed seeds. The raised, mineral seedbed
condition of mounds is suitable for germination
of these light-seeded species (Denny and Goodlett, 1956; Lyford and MacLean, 1966; Peterson
and Campbell, 1993; Oliver and Larson, 1996)
and may provide protection from mammalian herbivory (Long et al., 1998; Krueger and Peterson
2006). Additionally, pits may be subject to litter
and water accumulation as well as soil deposition
from mound erosion (Putz, 1983). In cooler and/or
wetter climate zones, harsh environmental conditions on mounds may be ameliorated. However,
the harsh environmental conditions coupled with
high herbaceous cover on mounds may have led
to lowered seedling density and species richness at
NTSF.
Although none of the vegetation characteristics
differed significantly between treatments, herbaceous cover and richness and tree seedling density and richness differed between microsite types
(Figures 5–8). Bare soil and pit microsites exhibited high herbaceous richness and tree seedling
density and richness but low herbaceous cover.
Therefore, bare soil and pit microsites may be
instrumental to maintaining species diversity of
both the seedling and herbaceous layers.
Vitis microsites supported high levels of herbaceous cover but low tree seedling density, perhaps
because V. rotundifolia shaded or outcompeted
tree seedlings. Like Vitis patches, mounds in
unsalvaged areas also had high herb cover but
low seedling density. In parallel to our findings,
Nakashizuka (1989) found that undisturbed soil
patches beneath gaps (equivalent to ‘bare’ in this
study) did not contribute to maintenance of pioneer species, possibly due to inhibition by dwarf
bamboo. Similarly, the fern understory in New
England forests has been shown to alter biotic and
abiotic conditions, thereby decreasing tree seedling emergence and/or survival (Horsley, 1993;
George and Bazzaz, 1999). Similar mechanisms
may be operating within Vitis and mound microsites at NTSF.
SALVAGING AFTER WIND DISTURBANCE
Thickets of Prunus spp. and Rubus spp. are
often common following wind disturbance in
eastern North America, due to buried pools of
viable seeds, and can act to accelerate succession if shade-intolerant species are unable to establish (Peterson and Carson, 1996). However,
high cover by herbaceous plants did not lead to
higher proportions of shade-tolerant species in
Vitis patches and mounds. Instead, while shadeintolerant seedlings (Table 2) were most abundant within all microsite types, pits had the most
shade-tolerant seedlings (44.6 per cent). Because
pits and mounds contain high proportions of both
shade-tolerant and shade-intolerant species, they
may be important for maintaining or increasing
species diversity.
Treefall pits and mounds contained very few
intermediate-tolerance seedlings (2 per cent and
5.5 per cent, respectively) because few oaks were
present (Tables 2 and 3). Mounds may have
low retention of heavy and/or round seeds (i.e.
acorns), perhaps because they roll off the elevated
microsites (Peterson et al., 1990; Carlton and
Bazzaz, 1998a). Because crown and bare microsites contained the highest proportion of seedlings intermediate in shade tolerance (primarily
Q. alba, Q. rubra and Q. velutina), crowns and
bare soil patches may be important for maintaining oak populations. However, whether this is
due to decreased mammalian herbivory, moderate environmental conditions or increased seed
input from downed and dying oak crowns is uncertain. Oaks are not represented in the seedling
layer in proportion to canopy abundance (data
not shown) and removing crowns by salvaging
may reduce already low oak regeneration.
Differences in resource levels among microsites
may be important if they directly affect vegetation characteristics. That the first three axes of
the CCAs accounted for only a miniscule portion of the variation in species composition of the
both herbaceous (Axis 1 = 1.1 per cent, Axis 2 =
0.5 per cent, Axis 3 = 0.2 per cent) and seedling
(Axis 1 = 1.8 per cent, Axis 2 = 1.3 per cent, Axis
3 = 0.4 per cent) layers may suggest that soil temperature, soil moisture and canopy openness did
not strongly influence species composition. However, our characterization of soil temperature and
moisture was coarse and may have missed extremes that could influence vegetation; thus, the
apparent limited influence of these environmental
373
factors should be interpreted cautiously. Other
factors that were not measured, such as seed
availability, nutrient availability or soil organic
matter, may be more important.
A number of studies (reviewed in McIver and
Starr, 2001) led us to expect higher soil temperatures, lower soil moisture and higher light levels
in the more open, salvaged stands since selective
logging and the use of heavy machinery removes
woody biomass from disturbed areas. Indeed, we
found that the temperature and canopy openness
were significantly higher in salvaged areas but
treatment did not affect soil moisture. Most of
the studies reviewed in McIver and Starr (2001)
appear to have had more severe disturbance from
fire than in our study, in addition to more thorough salvaging. It is likely that such differences
in combined disturbance severity explain some
of the discrepancies between the reviewed studies
and our findings.
Overall, microsite type was far more important
than treatment in influencing vegetation parameters. It is likely that the wind disturbance at NTSF
(with or without salvaging) increased diversity of
the seedling layer by allowing germination and
survival of shade-intolerant species within disturbance-created microsites. The importance of
post-windthrow salvaging derives from potential
alteration of environmental conditions (particularly soil temperature and canopy openness) and
microsite diversity, abundance and composition.
Contrary to our expectations, pits and mounds
were not destroyed by salvaging operations and
more existed in salvaged areas due to higher levels of uprooting in the natural windthrow (A.D.
Leach and C.J. Peterson, unpublished data).
However, salvaging operations were of very low
intensity at NTSF and this generalization may not
hold following higher severity disturbance and/
or higher intensity logging operations. This leads
us to suggest that the impacts of salvage logging
on forest regeneration potential and diversity are
likely to vary along a combined severity gradient—one that sums the combined severity of the
natural and the anthropogenic disturbances. This
is consistent with some recent conceptual treatments that predict forest response to combined
natural and anthropogenic impacts by summing
the effects if the impacts occur in quick succession (Frelich and Reich 1999; Roberts, 2004). In
this perspective, the combined disturbances are
374
FORESTRY
unlikely to cause drastic shifts in composition
and diversity if summed severity is low, but the
risk of such deleterious changes will increase as
severity increases (Roberts, 2004), perhaps with a
threshold (e.g. Frelich and Reich, 1999). In some
cases, unexpected and undesirable consequences
may result (Paine et al., 1998).
Research documenting microsite and vegetation characteristics after wind disturbance and
both before and after salvage logging, across
a range of combined disturbance severities,
would allow determination of whether and how
salvaging truly affects microsite abundance,
environmental conditions and vegetation characteristics.
Funding
USDA Forest Service Southern Research Station and by
grants from the University of Georgia Department of
Plant Biology.
Acknowledgements
We thank Chris Goetz and Roy Ward of Natchez Trace
State Park and Forest for allowing use of the field sites.
Paul Stufel, Lisa Krueger and Charles Cowden helped
with fieldwork. Lisa Kruse helped with plant identification. Rebecca Sharitz, Bruce Haines and Eleanor Pardini provided useful comments on the manuscript.
Conflict of Interest Statement
None declared.
References
Baird, B.N. 2006 Comment on “Post-wildfire logging
hinders regeneration and increases fire risk”. Science.
313, 615b.
Beatty, S.W. 1984 Influence of microtopography and
canopy species on spatial patterns of forest understory plants. Ecology. 65, 1406–1419.
Beatty, S.W. and Sholes, O.D.V. 1988 Leaf litter effect on plant species composition of deciduous forest
treefall pits. Can. J. For. Res. 18, 553–559.
Braun, E.L. 1950 Deciduous Forests of Eastern North
America. The Blakiston Company, Philadelphia, PA.
Brown, C.L. and Kirkman, L.K. 1990 Trees of Georgia
and Adjacent States. Timber Press, Portland, OR.
Burns, R.M. and Honkala, B.H. 1990 Silvics of North
America: 1. Conifers; 2. Hardwoods Agriculture
Handbook 654. U.S. Department of Agriculture,
Forest Service, Washington, D.C.
Carlton, G.C. and Bazzaz, F.A. 1998a Regeneration
of three sympatric birch species on experimental
hurricane blowdown microsites. Ecol. Monogr. 68,
99–120.
Carlton, G.C. and Bazzaz, F.A. 1998b Resource
congruence and forest regeneration following an
experimental hurricane blowdown. Ecology. 79,
1305–1319.
Clinton, B.D. and Baker, C.R. 2000 Catastrophic
windthrow in the southern Appalachians: characteristics of pits and mounds and initial vegetation responses. For. Ecol. Manage. 126, 51–60.
Denny, C.S. and Goodlett, J.C. 1956 Microrelief resulting from fallen trees. United States Geological Survey Prof Paper. 288, 59–66.
Donato, D.C., Fontaine, J.B., Campbell, J.L., Robinson,
W.D., Kaufman, J.B. and Law, B.E. 2006 Postwildfire logging hinders regeneration and increases
fire risk. Science. 311, 352.
Elliott, K.J., Hitchcock, S.L. and Krueger, L. 2002 Vegetation response to large scale disturbance in a southern Appalachian forest: Hurricane Opal and salvage
logging. J. Torrey Bot. Soc. 129, 48–59.
Flowers, R.L., Wiliams, L.D., Walker, D.D. and Phillips,
J.A. 1960 Soil Survey of Henderson County, Tennessee. Series 1954. No. 9. U.S. Department of Agriculture, Soil Conservation Service, Washington, D.C.
Floyd, D. 2006 Science, censure, and the reignited debate over the 2002 Biscuit fire. J. For. 104, 57.
Foster, D.R. and Orwig, D.A. 2006 Preemptive and
salvage harvesting of New England forests: when
doing nothing is a viable alternative. Conserv. Biol.
20, 959–970.
Franklin, J.F., Lindenmayer, D.B., Macmahon, J.A.,
Mckee, A., Magnusson, J. and Perry, D.A. 2000
Threads of continuity: ecosystem disturbances, biological legacies and ecosystem recovery. Conserv.
Biol. Pract. 1, 8–16.
Frazer, G.W., Canham, C.D. and Lertzman, K.P. 1999
Gap Light Analyzer (GLA): Imaging Software to
Extract Canopy Structure and Gap Light Transmission Indices from True-Colour Fisheye Photographs,
Users Manual and Program Documentation. Simon
Fraser University, Burnaby, B.C.
Frelich, L.E. and Reich, P.B. 1999 Neighborhood effects, disturbance severity, and community stability
in forests. Ecosystems. 2, 151–166.
SALVAGING AFTER WIND DISTURBANCE
George, L.O. and Bazzaz, F.A. 1999 The fern understory as an ecological filter: emergence and establishment of canopy-tree seedlings. Ecology. 80,
833–845.
Greene, D.F., Gauthier, S., Noel, J., Rousseau, M. and
Bergeron, Y. 2006 A field experiment to determine
the effect of post-fire salvage on seedbeds and tree
regeneration. Front. Ecol. Environ. 4, 69–74.
Harrington, T.B. and Bluhm, A.A. 2001 Tree regeneration responses to microsite characteristics following
a severe tornado in the Georgia Piedmont, USA. For.
Ecol. Manage. 140, 265–275.
Horsley, S.B. 1993 Mechanisms of interference between
hay-scented fern and black cherry. Can. J. For. Res.
23, 2059–2069.
Krueger, L.M. and Peterson, C.J. 2006 Effects of
white-tailed deer on Tsuga canadensis regeneration:
evidence of microsites as refugia from browsing. Am.
Midl. Nat. 89, 151–159.
Kupfer, J.A. and Franklin, S.B. 2000 Evaluation of an
ecological land type classification system, Natchez
Trace State Forest, western Tennessee USA. Landsc.
Urban Plan. 49, 179–190.
Kuuluvainen, T. and Juntunen, P. 1998 Seedling establishment in relation to microhabitat variation in a
windthrow gap in a boreal Pinus sylvestris forest. J.
Veg. Sci. 9, 551–562.
Lindenmayer, D.B. 2006 Salvage harvesting—past lessons and future issues. For. Chron. 82, 48–53.
Lindenmayer, D.B., Foster, D.R., Franklin, J.F., Hunter,
M.L., Noss, R.F. and Schmiegelow, F.A. 2004 Salvage harvesting policies after natural disturbance.
Science. 303, 1303.
Lindenmayer, D.B. and Noss, R.F. 2006 Salvage logging, ecosystem processes, and biodiversity conservation. Conserv. Biol. 20, 949–958.
Long, Z.T., Carson, W.P. and Peterson, C.J. 1998 Can
disturbance create refugia from herbivores: an example with hemlock regeneration on treefall mounds?.
J. Torrey Bot. Soc. 125, 165–168.
Lyford, W.H. and MacLean, D.W. 1966 Mound and
pit microrelief in relation to soil disturbance and tree
distribution in New Brunswick, Canada. Harvard
Forest Paper 15.
Macdonald, S.E. 2007 Effects of partial post-fire salvage harvesting on vegetation communities in the
boreal mixedwood forest region of northeastern Alberta, Canada. For. Ecol. Manage. 239, 21–31.
McCune, B. and Mefford, M.J. 1999 Multivariate
analysis of ecological data. MjM Software, Gleneden
Beach, OR.
375
McIver, J.D. and Starr, L. 2001 A literature review on
the environmental effects of postfire logging. West. J.
Appl. For. 16, 159–168.
Nakashizuka, T. 1989 Role of uprooting in composition and dynamics of an old-growth forest in Japan.
Ecology. 70, 1273–1278.
National Oceanic and Atmospheric Administration
(NOAA) 1999 Storm data. 41 (5), National Climatic
Data Center, Asheville, NC.
Newton, M., Fitzgerald, S., Rose, R.R., Adams, P.W.,
Tesch, S.D. and Sessions, J. 2006 Comment on “Postwildfire logging hinders regeneration and increases
fire risk”. Science. 313, 615a.
Oliver, C.D. and Larson, B.C. 1996 Forest Stand Dynamics. Wiley, New York, NY.
Paine, R.T., Tegner, M.J. and Johnson, E.A. 1998
Compounded perturbations yield ecological surprises. Ecosystems. 1, 535–545.
Palmer, M.W. 1993 Putting things in even better
order—the advantages of canonical correspondence
analysis. Ecology. 74, 2215–2230.
Peterson, C.J. and Campbell, J.E. 1993 Microsite differences and temporal change in plant communities
of treefall pits and mounds in an old-growth forest.
Bull. Torrey Bot. Club. 120, 451–460.
Peterson, C.J. and Carson, W.P. 1996 Generalizing forest regeneration models: the dependence of propagule availability on disturbance history and stand
size. Can. J. For. Res. 26, 45–52.
Peterson, C.J., Carson, W.P., Mccarthy, B.C. and
Pickett, S.T.A. 1990 Microsite variation and soil
dynamics within newly created treefall pits and
mounds. Oikos. 58, 39–46.
Peterson, C.J. and Leach, A.D. 2008 Limited salvage
logging effects on forest regeneration after moderateseverity windthrow. Ecol. Appl. 18, .
Peterson, C.J. and Pickett, S.T.A. 1990 Microsite
and elevational Influences on early forest regeneration after catastrophic windthrow. J. Veg. Sci. 1,
657–662.
Peterson, C.J. and Pickett, S.T.A. 1995 Forest reorganization—a case-study in an old-growth forest catastrophic blowdown. Ecology. 76, 763–774.
Peterson, C.J. and Pickett, S.T.A. 2000 Patch type
influences on regeneration in a western Pennsylvania, USA, catastrophic windthrow. Oikos. 90,
489–500.
Putz, F.E. 1983 Treefall pits and mounds, buried seeds,
and the importance of soil disturbance to pioneer
trees on Barro Colorado Island, Panama. Ecology.
64, 1069–1074.
376
FORESTRY
Roberts, M.R. 2004 Response of the herbaceous layer
to natural disturbance in North American forests.
Can. J. Bot. 82, 1273–1283.
Robison, S.A. and McCarthy, B.C. 1999 Potential
factors affecting the estimation of light availability
using hemispherical photography in oak forest understories. J. Torrey Bot. Soc. 126, 344–349.
Rumbaitis Del Rio, C.M. 2006 Changes in understory
composition following catastrophic windthrow and
salvage logging in a subalpine forest ecosystem. Can.
J. For. Res. 36, 2943–2954.
SAS Institute 1995 SAS/STAT User’s Guide. Cary, NC.
Release 6.03 Edition.
Smalley, G.W. 1991 Classification and Evaluation of
Forest Sites on the Natchez Trace State Forest, State
Resort Park, and Wildlife Management Area in
West Tennessee. General Technical Report SO-85.
U.S.D.A. Forest Service, New Orleans, LA.
Sokal, R.R. and Rohlf, F.J. 1995 Biometry. 3rd edition.
W.H. Freeman and Company, New York, NY.
USDA and NRCS 2002 The PLANTS Database, Version 3.5 [online]. National Plant Data Center, Baton
Rouge, LA. http://plants.usda.gov(accessed on 21
April, 2003).
Webb, S.L. and Scanga, S.E. 2001 Windstorm disturbance without patch dynamics: twelve years
of change in a Minnesota forest. Ecology. 82,
893–898.
Received 25 August 2007