Effects of Fire Size and Pattern on Early Succession in Yellowstone National Park

Abstract

The Yellowstone fires of 1988 affected >250 000 ha, creating a mosaic of burn severities across the landscape and providing an ideal opportunity to study effects of fire size and pattern on postfire succession. We asked whether vegetation responses differed between small and large burned patches within the fire-created mosaic in Yellowstone National Park (YNP) and evaluated the influence of spatial patterning on the postfire vegetation. Living vegetation in a small (1 ha), moderate (70-200 ha), and large (500-3600 ha) burned patch at each of three geographic locations was sampled annually from 1990 to 1993, Burn severity and patch size had significant effects on most biotic responses. Severely burned areas had higher cover and density of lodgepole pine seedlings, greater abundance of opportunistic species, and lower richness of vascular plant species than less severely burned areas. Larger burned patches had higher cover of tree seedlings and shrubs, greater densities of lodgepole pine seedlings and op

Full text

411
Ecological Monographs,
67(4), 1997, pp. 411–433
q
1997 by the Ecological Society of America
EFFECTS OF FIRE SIZE AND PATTERN ON EARLY SUCCESSION IN
YELLOWSTONE NATIONAL PARK
M
ONICA
G. T
URNER
,
1
W
ILLIAM
H. R
OMME
,
2
R
OBERT
H. G
ARDNER
,
3
AND
W
ILLIAM
W. H
ARGROVE
4
1
Department of Zoology, Birge Hall, University of Wisconsin, Madison, Wisconsin 53706 USA
2
Biology Department, Fort Lewis College, Durango, Colorado 81301 USA
3
Appalachian Environmental Laboratory, Gunter Hall, University of Maryland, Frostburg, Maryland 21532 USA
4
Environmental Sciences Division, Oak Ridge National Laboratory, P.O. Box 2008,
Oak Ridge, Tennessee 37831-6036 USA
Abstract.
The Yellowstone fires of 1988 affected
.
250000 ha, creating a mosaic of
burn severities across the landscape and providing an ideal opportunity to study effects of
fire size and pattern on postfire succession. We asked whether vegetation responses differed
between small and large burned patches within the fire-created mosaic in Yellowstone
National Park (YNP) and evaluated the influence of spatial patterning on the postfire veg-
etation. Living vegetation in a small (1 ha), moderate (70–200 ha), and large (500–3600
ha) burned patch at each of three geographic locations was sampled annually from 1990
to 1993. Burn severity and patch size had significant effects on most biotic responses.
Severely burned areas had higher cover and density of lodgepole pine seedlings, greater
abundance of opportunistic species, and lower richness of vascular plant species than less
severely burned areas. Larger burned patches had higher cover of tree seedlings andshrubs,
greater densities of lodgepole pine seedlings and opportunistic species, and lower species
richness than smaller patches. Herbaceous species present before the fires responded in-
dividually to burn severity and patch size; some were more abundant in large patches or
severely burned areas, while others were more abundant in small patches or lightly burned
areas. To date, dispersal into the burned areas from the surrounding unburned forest has
not been an important mechanism for reestablishment of forest species. Most plant cover
in burned areas consisted of resprouting survivors during the first 3 yr after the fires. A
pulse of seedling establishment in 1991 suggested that local dispersal from these survivors
was a dominant mechanism for reestablishment of forest herbs. Succession across much of
YNP appeared to be moving toward plant communities similar to those that burned in 1988,
primarily because extensive biotic residuals persisted even within very large burned areas.
However, forest reestablishment remained questionable in areas of old (
.
400 yr) forests
with low prefire serotiny. Despite significant effects of burn severity and patch size, the
most important explanatory variable for most biotic responses was geographic location,
particularly as related to broad-scale patterns of serotiny in
Pinus contorta.
We conclude
that the effects of fire size and pattern were important and some may be persistent, but that
these landscape-scale effects occurred within an overriding context of broader scale gra-
dients.
Key words: disturbance; fire ecology; landscape ecology; patch size;
Pinus contorta;
secondary
succession; spatial heterogeneity; spatial pattern; Wyoming; Yellowstone National Park.
I
NTRODUCTION
A key challenge in ecological research involves de-
termining the influence of spatial patterns on ecological
processes (Levin 1992, Kareiva 1994). Spatial pattern
has demonstrable effects on habitat use and foraging
dynamics (Pearson 1993, Turner et al. 1994
a
), popu-
lations (Kareiva 1990, Hanski 1993), nutrient move-
ments (Peterjohn and Correll 1984), and the spread of
disturbance (Turner 1987), but few studies have eval-
uated the relative importance of spatial pattern com-
pared to other controlling variables. Because broad-
scale disturbances are heterogeneous in their effects
Manuscript received 13 November 1995; revised 10 Oc-
tober 1996; accepted 4 November 1996; final version received
9 December 1996.
across the landscape, such events provide ideal op-
portunities for investigating the importance of spatial
pattern on succession. Investigations into mechanisms
of plant succession following fire and other distur-
bances often have emphasized the importance of au-
tecology and life history attributes of individual plants
and species in determining vegetation dynamics (e.g.,
Connell and Slatyer 1977, Noble and Slatyer 1980, Peet
and Christensen 1980, Pickett et al. 1987, Halpern
1988, 1989, Peterson and Pickett 1995). These studies
also demonstrated that species responses may vary with
the kind, severity, and spatial and temporal context of
disturbance (also see Pickett 1976, Finegan 1984).
Patch size, heterogeneity, and distance from undis-
turbed sites may differentially influence species having
particular combinations of life history attributes (Den-

412
MONICA G. TURNER ET AL.
Ecological Monographs
Vol. 67, No. 4
slow 1980
a
,
b
, Hartshorn 1980, Miller 1982, Malanson
1984, Green 1989) and may result in multiple pathways
for succession (Fastie 1995). In addition, disturbance
effects on natural communities are influenced by en-
vironmental conditions controlled by landscape posi-
tion (Foster 1988
a
,
b
, Callaway and Davis 1993, Boose
et al. 1994), although this depends on the disturbance
(Frelich and Lorimer 1991). Including spatial consid-
erations in explaining post-disturbance succession may
extend the power of approaches that incorporate life
history factors and environmental variation.
We studied the effects of fire size and pattern on
vegetation dynamics following the 1988 fires in Yel-
lowstone National Park (YNP), Wyoming, USA. Fires
in 1988 affected
.
25
3
10
4
ha in YNP and surrounding
lands as a consequence of unusually prolonged drought
and high winds (Renkin and Despain 1992, Bessie and
Johnson 1995). Such large fires are a major but infre-
quent disturbance in this landscape, occurring at 100-
to 300-yr intervals (Romme 1982, Romme and Despain
1989). As a result of variations in wind, topography,
vegetation, and time of burning (Rowe and Scotter
1973, Wright and Heinselman 1973, Van Wagner
1983), the 1988 fires produced a strikingly heteroge-
neous mosaic of burn severities (effects of fire on the
ecosystem) and islands of unburned vegetation across
the landscape (Christensen et al. 1989, Turner et al.
1994
b
). The spatial extent and heterogeneity of the
1988 fires provided an ideal opportunity to study ef-
fects of fire size and pattern on postfire succession. In
this paper, we ask whether vegetation responses differ
between small and large burned patches within the fire-
created mosaic in YNP and evaluate the importance of
spatial patterning on the postfire vegetation.
Four classes of burn severity that were easily dis-
criminated in the field through 1991 were used to char-
acterize heterogeneity of the 1988 fires: unburned (no
sign of fire effects), light-surface burns (canopy trees
generally survived and retained green needles, stems
often were scorched, and soil organic layer remained
largely intact); and severe-surface burns and crown
fires (extensive tree mortality, soil organic layer com-
pletely consumed). Canopy needles were consumed in
crown fires but not in severe-surface fires, where a litter
layer of dead needles developed rapidly. Spatial anal-
ysis of a map of burn severity derived from a 1989
Landsat Thematic Mapper image indicated a highly
patchy burn mosaic. Approximately 75% of the area
in crown fire was within 200 m of unburned or lightly
burned areas that are potential sources of plant pro-
pagules (Turner et al. 1994
b
). Even large patches of
crown fire contained areas of light- and severe-surface
burn.
Natural disturbances are unplanned, and it is usually
not possible to study their effects with a well-balanced
design. However, events such as the Yellowstone fires
provide much-needed empirical data about disturbance
effects over very large regions (Glenn-Lewin and van
der Maarel 1992) and therefore alternative study de-
signs for assessing effects of such events are necessary.
Wiens and Parker (1995) have identified two main
problems in studying unplanned events: (1) affected
sites are not randomly located and, because they must
be defined post facto, ‘‘reference’’ areas are not true
controls; and (2) pseudoreplication is a problem be-
cause the only true level of independent replication is
the disturbance event itself. In practice, effective strat-
egies can be used to deal with non-independence among
samples, and post facto study designs that document
both initial effects and subsequent recovery (e.g., the
interaction between impact level and time), which is
our approach in YNP, may actually be more useful than
before-after comparisons (Wiens and Parker 1995).
Level
3
time designs require methods consistent be-
tween sampling periods, but the potential severity of
errors due to pseudoreplication can be minimized by
use of repeated-measures analysis, which we have em-
ployed (Wiens and Parker 1995). Although we had
some information on prefire forest stand structure
across the landscape (see
Methods: Study area
), site-
specific data on the herbaceous community would have
been immensely valuable in interpreting postfire com-
munity response (e.g., see Dayton et al. 1992). We
submit that, interpreted appropriately, comparison of
postfire vegetation dynamics among burned patches is
a valid path of inquiry for studying effects of distur-
bance size and pattern.
H
YPOTHESES
We present our initial hypotheses concerning indi-
vidual species and community responses to fire sever-
ity, fire size, and distance to unburned forest by group-
ing similar response variables together (Table 1). We
then synthesize the full set of response variables in the
Discussion to address how burn size and pattern influ-
ence postfire successional dynamics.
We distinguished two categories of plants: (1) forest
species, which were important in prefire communities;
and (2) opportunistic species, which were absent or
only incidental before the fire. Forest species were fur-
ther classified by their mode of reproduction as: (1)
vegetative, in which plants sprouted from surviving
belowground structures, and (2) sexual, in which seed-
ling establishment occurred. The major seed sources
immediately following fire are the canopy and dispersal
from unburned areas, as the number of viable seeds in
boreal and high elevation forest soils tends to be ex-
tremely low (Johnson 1975, Whipple 1978, Archibold
1989).
Conifer trees in YNP reproduce only by seed. Lodge-
pole pine (
Pinus contorta
var.
latifolia
) has serotinous
cones, regenerates well following crown fire, and re-
quires exposed mineral soil for seed germination and
seedling establishment; we therefore expected seedling
density to be greatest in areas severely burned by crown
fires (Table 1). We also expected that seed might dis-

November 1997 413
FIRE SIZE AND PATTERN
T
ABLE
1. Summary of hypotheses about the effects of fire size and pattern on biotic response variables that were tested in
this study of early postfire succession in Yellowstone National Park.
Response variable
Independent variable
Burn severity Patch size Distance to
unburned forest
Biotic cover Decreasing with increasing
burn severity Greater in smaller patches No effect
Lodgepole pine seedling
density and cover Increasing with increasing
burn severity Greater in smaller patches Decreasing with increasing
distance
Sprouts of forest herbs
and shrubs Decreasing with increasing
burn severity No effect No effect
Seedlings of forest herbs
and shrubs Greatest in severe-surface
burns, lower in light-surface
burns and crown fires
Greater in smaller patches Decreasing with increasing
distance
Opportunistic species Increasing with increasing
burn severity Greater in larger patches No effect
Species richness Greatest in severe-surface
burns, lower in light-surface
burns, and lowest in crown
fires
No effect Decreasing with increasing
distance
perse into burned areas from open-cone trees in un-
burned forest and thus hypothesized that seedling den-
sity would be negatively related to the distance from
unburned forest.
Most herbs and shrubs can reproduce both vegeta-
tively and sexually (Lyon and Stickney 1976, Stickney
1986), so we distinguished between sprouts and seed-
lings in our analyses. Forest species used to test for
fire effects on sprouts (Table 1) included three forbs,
one shrub, and two sedges.
Epilobium angustifolium
(fireweed), and
Arnica cordifolia
(heartleaf arnica) are
perennial forbs with light, wind-dispersed seeds. Al-
though
Epilobium angustifolium
is recognized as a fire-
adapted species, it is common in unburned forests in
YNP as an inconspicuous herb that blooms infrequent-
ly.
Lupinus argenteus
(lupine) is a perennial forb with
relatively large, heavy seeds.
Vaccinium scoparium
(grouse whortleberry) is the dominant shrub in the un-
burned forest.
Carex geyeri
(elk sedge) and
Carex ros-
sii
(Ross’s sedge) are common graminoids in unburned
lodgepole pine forest (Despain 1990).
Forest species used to test for fire effects on seedlings
(Table 1) included
Lupinus argenteus, Epilobium an-
gustifolium,
and
Carex
spp. (
C. geyeri
and
C. rossii
).
Seedlings of
Arnica cordifolia
and
Vaccinium scopar-
ium
were never observed during the study, and
A. cor-
difolia
seeds were not successfully germinated even in
a greenhouse setting (Romme et al. 1995). Seedlings
of
C. geyeri
and
C. rossii
could not be differentiated.
Hypotheses about opportunistic species (Table 1)
were tested using four species.
Cirsium arvense
(Can-
ada thistle) is an exotic species that frequently occurred
along horse trails, foot trails, and roadways. The thistle
invaded several areas burned from 1972 to 1988 and
reproduces vegetatively once established. However, it
was not present throughout most of the burned area
before the fires of 1988, so we expected it to invade
burned patches through seed dispersal.
Gayophytum
diffusum
(ground smoke) and
Collinsia parviflora
(blue-eyed Mary) are native annuals that may be pres-
ent in unburned forests but are never abundant or con-
spicuous, and we hypothesized that these species might
respond rapidly to the newly created open habitat.
Lac-
tuca serriola
(prickly lettuce) is an exotic biennial that,
like the annuals, was not conspicuous in unburned for-
est.
S
TUDY
A
REA
Yellowstone National Park encompasses 9000 km
2
on a high forested plateau in northwest Wyoming. Ap-
proximately 80% of the park is dominated by lodgepole
pine forest, although subalpine fir (
Abies lasiocarpa
(Hook.) Nutt.), Engelmann spruce (
Picea engelmannii
Parry), and whitebark pine (
Pinus albicaulis
Engelm.)
may be locally abundant (Despain 1990). Our study
area was the subalpine forested plateau that covers most
of Yellowstone (Fig. 1) and encompasses dry, infertile
rhyolite substrates as well as more mesic and fertile
andesite and lake-bottom substrates. The climate is
generally cool and dry with mean January temperature
of
2
11.4
8
C, mean July temperature of 10.8
8
C, and
mean annual precipitation of 56.25 cm (Dirks and Mart-
ner 1982). The summer of 1988 was the driest on record
since 1886, with precipitation in June, July, and August
at 20%, 79%, and 10%, respectively, of average (Na-
tional Park Service, YNP).
M
ETHODS
Field sampling
In order to establish three replicates of three patch
sizes that differed in size by an order of magnitude,
we selected a small (1 ha), moderate (70–200 ha), and
large (500–3600 ha) patch of crown fire at each of three
geographic locations (Table 2) across the subalpine pla-
teau (Fig. 1). Regions of lesser burn severity were con-

414
MONICA G. TURNER ET AL.
Ecological Monographs
Vol. 67, No. 4
F
IG
. 1. Map of Yellowstone National Park showing the
locations of the three study locations (Cougar Creek, Fern
Cascades, and Yellowstone Lake). A small (1–2 ha), moderate
(80–200 ha), and large (480–3698 ha) patch of crown fire
was studied at each location. The hatched area depicts the
Yellowstone Plateau, the gray shading illustratesmajor lakes,
solid black areas show the nine burned patches used in this
study, and irregular solid lines indicate Park roads.
tained within and around each patch. Selection was
subjectively based on patch size and accessibility based
on digital satellite imagery. Availability of accessible
large crown-fire patches was limited, and the large
patch at Yellowstone Lake was substantially larger than
at the other two locations (Table 2).
In July 1990, four permanent transects were estab-
lished in each of the nine patches, extending from the
center to the edge of the patch along subcardinal di-
rections (NE, NW, SW, and SE). Transects varied in
length (Table 2) depending on patch size and shape.
The edge of the patch was defined by unburned forest,
a light-surface burn, a topographic barrier, or an un-
forested area. Sampling on each transect began 20 m
from the center point of the patch and continued at
fixed intervals (100 m in moderate patches and the large
patches at Fern Cascades and Cougar Creek, 200 m in
the large patch at Yellowstone Lake, and 20 m in the
small patches) as long as the transect continued through
areas affected by crown fire. At transitions between
burn severity classes (e.g., between crown fire and a
severe-surface burn), sampling points were located on
the edge and at three 20-m intervals on either side of
the edge. Subsequent sampling points were again lo-
cated at the fixed intervals until the next edge was
encountered. Sampling points were marked in the field
with wooden stakes, flagging, and rock cairns and were
sampled during July and August of 1990, 1991, 1992,
and 1993.
A 50-m
2
circular plot was centered on each sampling
point, and slope, aspect, and burn severity were re-
corded in 1990. The proportion of prefire serotinous
lodgepole pine trees was recorded within a 50-m radius
of the sampling point in 1992 following methods in
Tinker et al. (1994). For vegetation measurements, an
8-m line was centered on the sampling point and ex-
tended perpendicular to the main axis of the transect.
Percent cover data were recorded within eight 0.25-m
2
plots spaced at 1-m intervals along this line. At each
point in a 25-point 0.5
3
0.5 m point-intercept frame
(cf. Floyd and Anderson 1982, 1987), the underlying
plant species or cover type (exposed mineral soil, un-
burned litter, charred litter, pebble, cobble, or boulder)
was recorded. Percent cover was determined by aggre-
gating the data from the eight 0.25-m
2
plots. Species
richness was measured by recording all species within
an area extending 1 m along the 8-m line. Nomenclature
follows Dorn (1992).
Individuals of
Epilobium angustifolium, Lupinus ar-
genteus, Arnica cordifolia, Carex geyeri, Carex rossii,
Pinus contorta,
and
Vaccinium scoparium
were cen-
sused within each of the eight 0.25-m
2
plots with pe-
rennials classified as seedlings of the year or sprouts
based on morphological characters (M. G. Turner et al.,
unpublished data
).
Pinus contorta
seedlings were
counted by age (in years) to estimate recruitment
through time.
Opportunistic species (
Cirsium arvense, Collinsia
parviflora, Gayophytum diffusum,
and
Lactuca serrio-
la
), which frequently were sparse at individual sam-
pling points, were sampled along 1 m wide belt tran-
sects between sampling points. Belt transects were es-
tablished to record the total number of individuals ob-
served
#
0.5 m from the transect line.
The similarity in species richness between years at
each sampling point was computed by
S
5
2
C
/(
A
1
B
1
2
C
)
where
S
5
the similarity in species richness between
two samples (here, between years),
A
5
the number of
species unique to sample 1,
B
5
the number of species
unique to sample 2, and
C
5
the number of species
that shared in both samples (Pielou 1974:311).
Statistical analyses were performed by using SAS
(SAS Institute 1992). An arcsine transform was applied
to percent cover data and the similarity index values
prior to analysis to eliminate bias in the variance and
mean (Zar 1984). A square-root transform was applied
to counts of sprouts and seedlings of forest species,
including
P. contorta,
and the belt transect data prior
to analysis. Data were subject to repeated-measures
analysis of variance (ANOVA) because measurements
were obtained annually from the same sampling points.
Independent variables included geographic location,
patch size, burn severity, slope, aspect, distance to the

November 1997 415
FIRE SIZE AND PATTERN
T
ABLE
2. General description of the three geographic locations where small, moderate, and large patches of crown fire
were sampled in Yellowstone National Park (YNP) from 1990 to 1993.
Attribute Cougar Creek Fern Cascades Yellowstone Lake
Location in YNP Westcentral YNP near
lower forest ecotone Southwestern YNP near
Old Faithful, extensive
forest
Southeastern YNP, forest
interspersed with occa-
sional meadows
Patch sizes (ha) and no. sampling
points
Large
Moderate
Small
500 (84 points)
91 (46 points)
1 (34 points)
480 (103 points)
200 (82 points)
1 (37 points)
3698 (59 points)
74 (67 points)
1 (40 points)
Elevation range (m) 2150–2300 2270–2500 2400–2700
Geologic substrate (Keefer 1972) Rhyolite and tuff (moder-
ately infertile Quater-
nary volcanics)
Rhyolite (infertile Quater-
nary volcanics) Lake sediments and ande-
site (moderately fertile
Eocene volcanics)
General vegetation of unburned
areas
Pinus contorta
dominated
early and late seral
stages; occasional
Pseudotsuga menziesii
;
Ceanothus velutinus,
Carex rossii,
and
Cala-
magrostis rubescens
dominated ground layer
Pinus contorta
dominated
early and late seral
stages; occasional
Abies
lasiocarpa
;
Vaccinium
scoparium, Carex gey-
eri,
and
Lupinus argen-
teus
dominated ground
layer
Pinus contorta
dominated
early seral stages;
Abies
lasiocarpa
and
Picea en-
gelmannii
dominated late
stages;
Vaccinium sco-
parium, Carex geyeri,
and
Arnica cordifolia
dominated ground layer
Relative density (%) of lodgepole
pine in prefire stands Mean
5
91
SD
5
22
Range
5
12–100
Mean
5
100
SD
5
0Mean
5
63
SD
5
35
Range
5
0–100
Predominant prefire stand age (from
Romme and Despain 1989 and
unpublished data
)
ø
130-yr-old even-aged
stands that originated
after fires in the 1860s
ø
290-yr-old even-aged
stands that originated
after fires in early 1700s
ø
250-yr-old even-aged
stands that originated af-
ter fires in mid 1700s in
small and moderate
patches;
.
400-yr-old un-
even-aged stands in large
patch
Prefire tree density (all species;
stems/ha) Mean
5
610
SD
5
424
Range
5
32–1846
Mean
5
470
SD
5
212
Range
5
96–891
Mean
5
470
SD
5
182
Range
5
64–796
Mean prefire percent serotinous
trees (Tinker et al. 1994) 65.0 5.4 1.9
nearest unburned or lightly burned forest (i.e., live for-
est), distance to the nearest severe-surface burn, and
appropriate interaction terms. Distance to nearest live
forest was included because of the potential for seeds
to disperse into the areas of stand-replacing burn. Dis-
tance to the nearest severe-surface burn was included
because our initial field observations in 1989 and 1990
suggested that these areas could be sources for lodge-
pole pine seeds, as cones were heated but not con-
sumed. The repeated-measures ANOVAs were done by
using the repeated-measures procedure in SAS to test
for between-sampling-point effects and by running an
ANOVA in which all effects and interactions werefully
nested in time to determine the explanatory power (
r
2
)
of the overall model. Means were separated in the fully
nested model by using Tukey’s studentized range test
(
P
,
0.05). To examine the expansion of
Epilobium
angustifolium
and
Lupinus argenteus
across the land-
scape, the proportion of sampling points with sprouts
and seedlings of these species was examined through
time.
R
ESULTS
Our sampling design involved three replicates of the
three patch sizes across the subalpine plateau, but anal-
yses of the data revealed a surprisingly strong effect
of geographic location in explaining variability in post-
fire vegetation. Thus, our results report not only the
effects of the fire-related variables but also the influ-
ence of geographic location and its interaction with
other main effects.
Biotic cover
Total biotic cover increased through time and varied
in a complex way with location, patch size, burn se-
verity, and interactions among these variables (Table
3). Most variability in total biotic cover was explained
by geographic location, with Cougar Creek and Yel-
lowstone Lake having much greater cover than Fern
Cascades (Fig. 2). Patch size was next in importance
(Table 3), with small patches having greater cover than
moderate or large patches (Fig. 3). Biotic cover also
decreased as burn severity increased (Fig. 4). By 1993,
however, severe-surface burns and light-surface burns
had the same percent cover (Fig. 4), although the values
varied with location (60–65% at Cougar Creek and
Yellowstone Lake and
;
40% at Fern Cascades). Total
cover increased two to four times in crown fires and
severe-surface burns between 1990 and 1993 but did

416
MONICA G. TURNER ET AL.
Ecological Monographs
Vol. 67, No. 4
T
ABLE
3. Summary of tests for between-sample-points effects on postfire biotic cover from repeated-measures ANOVA.
Entries indicate the
F
and
P
values for significant effects, and trends for main effects. The two most important effectsper
column (as indicated by the two greatest
F
values) are highlighted in boldface type.
Effect Forbs Graminoids Trees Shrubs Total cover
Location† (df
5
2)
F
5
112.71
P
5
0.0001
L
.
F, C
F
5
43.69
P
5
0.0001
C
.
F, L
F
5
130.28
P
5
0.0001
C
.
F
.
L
F
5
9.02
P
5
0.0001
C
.
F, L
F
5
85.81
P
5
0.0001
C, L
.
F
Patch size‡ (df
5
2)
F
5
28.68
P
5
0.0001
S
.
M
.
L
F
5
17.76
P
5
0.0001
S
.
M, L
NS
F
5
8.32
P
5
0.0003
L
.
M, S
F
5
35.86
P
5
0.0001
S
.
M, L
Burn severity
\
(df
5
3)
NS
F
5
16.18
P
5
0.0001
2, 1
.
3
F
5
6.75
P
5
0.0002
3, 2
.
1
F
5
18.97
P
5
0.0001
1
.
2, 3
F
5
24.17
P
5
0.0001
2, 1
.
3
Slope (df
5
1)
NS NS NS NS NS
Aspect (df
5
2)
NS NS NS NS NS
Distance to unburned forest (df
5
1)
NS
F
5
3.88
P
5
0.0495
Decreasing with
increasing dis-
tance
NS
F
5
7.39
P
5
0.0068
Decreasing with
increasing dis-
tance
NS
Distance to severe-surface burn (df
5
1)
NS NS NS NS NS
Location
3
patch size (df
5
4)
F
5
19.09
P
5
0.0001
F
5
3.88
P
5
0.0041
F
5
5.35
P
5
0.0003
F
5
5.22
P
5
0.0004
F
5
10.85
P
5
0.0001
Location
3
burn severity (df
5
4)
F
5
2.53
P
5
0.0398
F
5
6.22
P
5
0.0001
F
5
7.24
P
5
0.0001
F
5
10.04
P
5
0.0001
F
5
2.92
P
5
0.0210
Patch size
3
burn severity (df
5
4)
F
5
3.94
F
5
9.21
NS
F
5
9.01
F
5
5.80
P
5
0.0037
P
5
0.0001
P
5
0.0001
P
5
0.0001
Location
3
patch size
3
burn severity
(df
5
6)
NS NS NS NS NS
Notes:
Error df
5
497.
NS
indicates that an effect was not significant. Changes through time were significant for all
response variables except shrub cover.
† Abbreviations for locations are: C
5
Cougar Creek, F
5
Fern Cascades, and L
5
Yellowstone Lake.
‡ Abbreviations for patch sizes are: L
5
large, M
5
moderate, and S
5
small.
\
Abbreviations for burn severities are: 1
5
light-surface burn, 2
5
severe-surface burn, and 3
5
crown fire.
not change significantly in light-surface burns. None-
theless, average biotic cover in crown-fire areas in 1993
was less than that of light-surface burns in 1990.
Percent cover of forbs was largely explained by geo-
graphic location and patch size (Table 3). Forb cover
was greatest at Yellowstone Lake (Fig. 2). Patch size
was next in importance, with forb cover greatest in
small patches and lowest in large patches (Fig. 3). Forb
cover at Fern Cascades and YellowstoneLake increased
through time, but forb cover did not change signifi-
cantly at Cougar Creek through time. Burn severity had
no effect.
Variation in graminoid cover was also explained pri-
marily by location differences and patch size (Table 3).
Graminoid cover was greatest at Cougar Creek and
lower at the other locations (Fig. 2). Among patch sizes,
graminoid cover was greatest in small patches (Fig. 3).
Among burn severity classes, graminoid cover was sig-
nificantly greater in the surface burns than in crown
fires. In areas of light-surface burn, graminoid cover
was similar among locations, ranging from
;
15 to
20%. Graminoid cover also was negatively correlated
with distance from unburned forest (Spearman
r
S
5
2
0.31,
P
5
0.0001).
Most variability in percent cover of tree seedlings
(primarily lodgepole pine) was due to differences
among study locations (Table 3), with burn severity
also contributing. Tree cover was greatest at Cougar
Creek (Fig. 2), reaching
;
10% by 1993. Fern Cascades
had
,
1% cover of tree seedlings in 1993, and Yellow-
stone Lake had
,
0.1% tree seedling cover. Tree seed-
ling cover was greatest in crown-fire and severe-surface
burns (Fig. 4). Patch size did not have a significant
effect on tree seedling cover, although there was an
interaction of patch size and location. At Cougar Creek,
where tree seedling cover was highest among locations,
cover was substantially greater in the large patch com-
pared to the moderate and small patches. At both Fern
Cascades and Yellowstone Lake, however, treeseedling
cover was greater in the smaller patches-but it was still
extremely low.
Shrub cover was uniformly low in the burned areas
and varied primarily with burn severity and the inter-
action between location and burn severity (Table 3).
Shrub cover did not change during the period of study
and was greatest in areas of light-surface burn and low-
est in crown fires (Fig. 4). Patch size was next in im-
portance, with large patches having more shrub cover
than moderate or small patches. Among locations,
shrub cover was greater at Cougar Creek than at Fern

November 1997 417
FIRE SIZE AND PATTERN
F
IG
. 2. Percent cover by year at each of the three study locations (Cougar Creek, Fern Cascades, and Yellowstone Lake)
in Yellowstone National Park. Data show means
6
2
SE
.
Cascades or Yellowstone Lake. Shrub cover was also
negatively correlated with distance to the unburned for-
est edge (Spearman
r
S
52
0.15,
P
5
0.0001).
Lodgepole pine reestablishment
Density of postfire
P. contorta
seedlings did not
change between 1990 and 1993, but density of first-
year seedlings declined (Fig. 5). Among locations,
Cougar Creek had the greatest mean seedling density
(11.1 seedlings/m
2
), Fern Cascades was lower by two
orders of magnitude (0.23 seedlings/m
2
), and the Yel-
lowstone Lake location had the lowest density (0.06
seedlings/m
2
). The number of prefire serotinous trees
within 50 m of the sampling point was the most im-
portant variable explaining lodgepole pine seedling
density (Table 4), with a positive correlation observed
(Table 5). Correlation analysis also revealed interesting
variation among locations. Seedling density was pos-
itively correlated with number of prefire serotinous in-
dividuals at Cougar Creek (Table 5), which had the
highest mean serotiny level (65%) among the geo-
graphic locations, but the correlation was not signifi-
cant at Fern Cascades or Yellowstone Lake (Table 5),
where mean serotiny and numbers of seedlings were
much lower.
Mean seedling density was greatest in severe-surface
burns (14 seedlings/m
2
) and least in the crown fires and
light-surface burns (5 seedlings/m
2
). During the four
years of the study, seedlings were present at 51–59%
of sampling points in light-surface burns, 61–73% in
severe-surface burns, and only 34–36% of sampling
points in crown fires. Lodgepole pine seedlings always
occurred more often than expected in the surface burns,
and less often than expected in crown-fire burns (
P
,
0.05 for chi-square analyses for each year). Seedling
density was also negatively correlated with distance to
the nearest severe-surface burn (Table 5), indicating a
potentially important edge effect between burn severity
classes.
Patch size was significant but less important (Table
4), and its effect varied among locations. Overall mean
seedling density declined with increasing patch size
(means of 8.1, 6.8, and 3.9 seedlings/m
2
for large, mod-
erate, and small patches, respectively). At Cougar
Creek, seedling density was lower in the small patch
than in the moderate or large patch. At Fern Cascades
and Yellowstone Lake, however, lodgepole pine seed-
ling density was greater in the small patches than in
the large or moderate patches.
Mean density of lodgepole seedlings by annual age
class each year indicated that seedling recruitment oc-
curred primarily in 1989 and 1990. Density of first-

418
MONICA G. TURNER ET AL.
Ecological Monographs
Vol. 67, No. 4
F
IG
. 3. Percent cover by year for small, moderate, and large burned patches for forb, graminoid, and total biotic cover
after the extensive 1988 fires. Data show means
6
2
SE
.
year seedlings declined annually from 1.4 seedlings/
m
2
in 1990 to 0.01 seedlings/m
2
in 1993 (Fig. 5). As
with total seedling density, most variation in first-year
seedling density was explained by the number of prefire
serotinous individuals, but change through time was
significant (Table 4). Examination of the percentage of
sampling points containing lodgepole seedlings each
year also supported a peak of establishment during the
first and second years after the fire. The percentage of
sampling points that contained lodgepole pine seed-
lings was 43% in 1990, 48% in 1991, 46% in 1992,
and 45% in 1993, with the same sites generally having
seedlings present or absent each year (i.e., new estab-
lishment was negligible after 1990).
Reestablishment of forest species
Density of sprouts of forest species.
—
1.
Epilobium angustifolium.
—Sprout density across
all sampling points increased annually from an average
of 4.8 sprouts/m
2
in 1990 to 23.4 sprouts/m
2
in 1993.
The ANOVA model fully nested in time for
E. angus-
tifolium
explained 57% of the variance in sprout density
(Table 6). Most of the variability in
E. angustifolium
sprout density was explained by differences in location,
patch size, and their interactions (Table6). Sprout den-
sity was greatest at the Yellowstone Lake location (Fig.
6). Among patch sizes, sprout density was generally
greater in the small and moderate patches compared to
the large patches (Fig. 7). Sprout density also varied
with burn severity, with densities being greater in the
severe-surface burns and crown fires (18.6 and 15.0
sprouts/m
2
, respectively) than in light-surface burns
(12.2 sprouts/m
2
).
2.
Lupinus argenteus.
—The ANOVA model for
L.
argenteus
explained only 15% of the variance in sprout
density, and most of the variability was due to location
and patch size (Table 6). There was no significant dif-
ference in sprout density through time. Among loca-
tions,
L. argenteus
was most abundant at Yellowstone
Lake and Fern Cascades and least abundant at Cougar
Creek (Fig. 6). No consistent effect of patch size was
observed (Fig. 7). Although burn severity did not ex-
plain density of
L. argenteus
sprouts, it did relate to
its presence at a sampling point. Chi-square analysis
for each year indicated that sprouts of
L. argenteus
were always present more than expected in light-sur-
face burns, as expected in severe-surface burns, and
less than expected in crown-fire burns (all
x
2
tests sig-
nificant at
P
,
0.05).
3.
Arnica cordifolia.
—Density of
A. cordifolia
sprouts increased through time and was explained pri-
marily by differences in location and patch size (Table

November 1997 419
FIRE SIZE AND PATTERN
F
IG
. 4. Percent cover by year for the three burn severity classes. Data show means
6
2
SE
.
F
IG
. 5. Mean density of
Pinus contorta
(PICO) seedlings
across all sampling points by year. ‘‘All PICOs’’ refers toall
seedlings that germinated following the 1988 fires. The first-
year PICOs are those seedlings that germinated during the
indicated year. Data show means
6
2
SE
.
6). Among locations,
A. cordifolia
sprouts were much
more abundant at the Yellowstone Lake location than
either of the other locations (Fig. 6). Sprouts were also
much more abundant in small patches compared to
large or moderate patches (Fig. 7).
4.
Vaccinium scoparium.
—Sprout density of
V. sco-
parium
was explained primarily by burn severity and
location (Table 6). Sprout density clearlydeclined with
increasing burn severity, varying over an order of mag-
nitude (Fig. 8). Among locations, density was highest
at the Yellowstone Lake location and lowest at Cougar
Creek (Fig. 6). Among patch sizes, density was greatest
in small patches and least in the large patches (Fig. 7).
5.
Carex rossii.
—Sprout density of
C. rossii
did not
change significantly through time. Variation in the den-
sity of
C. rossii
sprouts was explained primarily by
location, patch size, and distance from the nearest un-
burned forest and severe-surface burn (Table 6). Cou-
gar Creek had the greatest sprout density, Fern Cas-
cades was intermediate and Yellowstone Lake the low-
est (Fig. 6).
C. rossii
sprouts were most abundant in
large patches and lower in the moderate and small
patches (Fig. 7). Sprout density was greater in the more
severe burns and increased with distance from the near-
est unburned forest and severe-surface burn.
6.
Carex geyeri.
—Density of sprouts of
C. geyeri
varied primarily in response to location and patch size
with burn severity also an important main effect (Table
6). In contrast to
C. rossii, C. geyeri
was most abundant
at the Yellowstone Lake and Fern Cascades locations
and least abundant at Cougar Creek (Fig. 6).
C. geyeri
sprouts were more abundant in small patches than in
moderate or large patches (Fig. 7), and more abundant
in light-surface burns (5.5 sprouts/m
2
) than in severe-
surface burns (3.1 sprouts/m
2
) or crown fires (0.54

420
MONICA G. TURNER ET AL.
Ecological Monographs
Vol. 67, No. 4
T
ABLE
4. Summary of significant between-sample-points effects on postfire density of lodgepole pine (
Pinus contorta,
PICO) seedlings from repeated-measures ANOVA. Trends are indicated for main effects.
Effect
All postfire PICO seedlings
FP
Trend
First-year PICO seedlings
FP
Trend
Location† (df
5
2)
Patch size‡ (df
5
2)
Burn severity
\
(df
5
3)
Distance to severe-surface burn (df
5
1)
Number of serotinous individuals
surrounding the sampling point
(df
5
1)
15.80
3.70
12.42
5.07
69.12
0.0001
0.0255
0.0001
0.0249
0.0001
C
.
F
.
L
L
.
M
.
S
2
.
3, 1
negative
positive
12.53
12.98
45.40
NS
0.0001
0.0001
NS
0.0001
M
.
L, S
2, 1
.
3
Location
3
patch size (df
5
4)
Location
3
burn severity (df
5
4)
Patch size
3
burn severity (df
5
4)
10.22
16.21 0.0002
0.0001
NS
19.77
7.63
5.27
0.0001
0.0001
0.0004
Location
3
patch size
3
burn severity (df
5
6) 2.96 0.0078 5.51 0.0001
Notes:
Error df
5
388. For all postfire seedlings, overall model
r
2
5
0.64, and change through time was not significant;
for first-year seedlings, overall model
r
2
5
0.51, and change through time was significant.
† Abbreviations for locations are: C
5
Cougar Creek, F
5
Fern Cascades, and L
5
Yellowstone Lake.
‡ Abbreviations for patch sizes are: L
5
large, M
5
moderate, and S
5
small.
\
Abbreviations for burn severities are: 1
5
light-surface burn, 2
5
severe-surface burn, and 3
5
crown fire.
T
ABLE
5. Spearman rank-order correlation coefficients (
r
) between square root of mean seed-
ling density of
Pinus contorta
during 1990–1993 and (a) the numbers of prefire serotinous
individuals within a 50-m radius of the sampling point and (b) distance to the nearest surface
burn (m) at each sampling point. Significant correlations are highlighted in boldface type.
Variable Statistic All locations Cougar Fern Lake
a) Number serotinous
individuals
r
S
P
n
0.56
0.0001
454
0.45
0.0001
156
0.05
0.4890
208
0.10
0.3203
90
b) Distance to severe-
surface burn
r
S
P
n
2
0.23
0.0001
594
2
0.23
0.0001
169
2
0.50
0.0001
230
2
0.39
0.0001
195
sprouts/m
2
). Sprouts also were most abundant on south-
erly aspects.
Density of seedlings of forest species.
—
1.
Epilobium angustifolium.
—Variation in seedling
density of
E. angustifolium
was explained primarily by
location and the interaction of location and patch size
(Table 7). Seedlings were most abundant at the Yel-
lowstone Lake location (6.8 seedlings/m
2
) and least
abundant at Fern Cascades and Cougar Creek (
;
1.4
seedlings/m
2
). Among patches, seedling density was
greater in the small and moderate patches compared to
the large patches. Seedling density did not vary with
burn severity, and the presence of sprouts of
E. an-
gustifolium
during 1990 or 1991 did not explain vari-
ation in seedling density (Table 7).
2.
Lupinus argenteus.
—Variation in
L. argenteus
seedling density was explained almost exclusively by
the presence of sprouts at the sampling point in 1990
or 1991 (Table 7). Seedling densities were muchhigher
when sprouts were present than when they were absent
(Fig. 9). Other effects that were significant, but not
nearly as important, included location (with Fern Cas-
cades having more seedlings than the other two loca-
tions) and patch size (greater density in large and mod-
erate patches than in small).
3.
Carex spp.
—Only 6% of the variance in density
of
Carex
seedlings was explained in the statistical mod-
el (
r
2
5
0.06, df
5
44/2240,
MSE
5
0.42,
F
5
3.29,
P
5
0.0001). Most of the variation was explained by the
presence of
Carex
sprouts at the sampling point in 1990
and 1991. Differences among locations, years, and the
interaction between location and year were the only
other significant variables. Burn severity and patch size
had no effect.
Presence/absence of
Epilobium
and
Lupinus.—In
addition to analyzing the density of sprouts and seed-
lings, we also examined simply the presence of sprouts
and seedlings of
E. angustifolium
and
L. argenteus
through time and as a function of burn severity. Al-
though the proportion of sampling points at which each
species was present increased during the sampling pe-
riod,
E. angustifolium
was much more ubiquitous. By
1993,
E. angustifolium
sprouts were present at nearly
90% of the sampling points in crown fires but
L. ar-
genteus
sprouts were present at only
;
20% (Fig. 10).
However, the proportion of sampling points with seed-
lings present peaked in 1991 for both species and was
quite low by 1993 (Fig. 10). Indeed, we observed very
few seedlings of any species in 1993.
Opportunistic species
Between 19 and 46% of the variance in density of
opportunistic species was explained by main effects of

November 1997 421
FIRE SIZE AND PATTERN
T
ABLE
6. Summary of tests for between-sample-points effects on postfire density of sprouts of forest species from repeated-
measures ANOVA. Entries indicate the
F
and
P
values for significant effects, and trends for main effects. The two most
important effects per column are shown in boldface type.
Effect
Epilobium
angustifolium Lupinus
argenteus Arnica
cordifolia Vaccinium
scoparium Carex rossii Carex geyeri
Overall
r
2
for ANOVA ful-
ly nested in time 0.57 0.15 0.32 0.23 0.23 0.44
Location† (df
5
2)
F
5
52.42
P
5
0.0001
L
.
F, C
F
5
7.85
P
5
0.0004
L, F
.
C
F
5
57.27
P
5
0.0001
L
.
C
.
F
F
5
16.99
P
5
0.0001
F, L
.
C
F
5
12.32
P
5
0.0001
C
.
F
.
L
F
5
99.41
P
5
0.0001
L, F
.
C
Patch size‡ (df
5
2)
F
5
3.50
P
5
0.0311
S, M
.
L
F
5
5.15
P
5
0.0061
L, S
.
M
F
5
9.91
P
5
0.0001
S
.
L
.
M
F
5
8.52
P
5
0.0002
S
.
M
.
L
F
5
9.44
P
5
0.0001
L
.
M, S
F
5
80.18
P
5
0.0001
S
.
M, L
Burn severity
\
(df
5
2)
F
5
10.02
P
5
0.0001
3, 2
.
1
NS
F
5
4.73
P
5
0.0092
1, 2
.
3
F
5
18.60
P
5
0.0001
1
.
2
.
3
F
5
3.23
P
5
0.0402
2, 3
.
1
F
5
52.66
P
5
0.0001
1
.
2
.
3
Aspect (df
5
2)
NS NS NS NS NS
F
5
4.06
P
5
0.0178
Greatest on
southerly
Distance to unburned forest
(df
5
1)
NS NS NS NS
F
5
9.04
P
5
0.0028
Increases with
distance
NS
Distance to severe-surface
burn (df
5
1)
NS NS NS NS
F
5
10.35
P
5
0.0014
Increases with
distance
NS
Location
3
patch size
(df
5
4)
F
5
21.19
P
5
0.0001
NS NS
F
5
5.16
P
5
0.004
F
5
4.61
P
5
0.0012
F
5
24.34
P
5
0.0001
Location
3
burn severity
(df
5
4)
NS NS NS
F
5
5.85
P
5
0.0001
F
5
2.89
P
5
0.0218
F
5
17.10
P
5
0.0001
Patch size
3
burn severity
(df
5
4)
F
5
6.72
P
5
0.0001
F
5
4.67
P
5
0.0010
NS
F
5
10.06
P
5
0.0001
F
5
3.98
P
5
0.0034
F
5
16.07
P
5
0.0001
Location
3
patch size
3
burn severity (df
5
7)
F
5
2.49
P
5
0.0161
F
5
2.64
P
5
0.0108
F
5
3.42
P
5
0.0014
NS
F
5
2.51
P
5
0.0153
F
5
2.20
P
5
0.330
Notes:
Error df
5
499.
NS
indicates that an effect was not significant. Changes through time were significant for allresponse
variables except
Lupinus argenteus
and
Carex rossii.
† Abbreviations for locations are: C
5
Cougar Creek, F
5
Fern Cascades, and L
5
Yellowstone Lake.
‡ Abbreviations for patch sizes are: L
5
large, M
5
moderate, and S
5
small.
\
Abbreviations for burn severities are: 1
5
light-surface burn, 2
5
severe-surface burn, and 3
5
crown fire.
location, patch size, burn severity, and year, as well as
interaction effects (Table 8). Three of the four oppor-
tunistic species increased in abundance through time
(Table 8), some reaching densities approaching 50
3
10
3
stems/ha. Density of three of the four opportunistic
species varied among locations. The two native an-
nuals,
Gayophytum diffusum
and
Collinsia parviflora,
were most abundant at the Cougar Creek location, less
at Fern Cascades, and least abundant at the Yellowstone
Lake location. Of the exotic perennials,
Cirsium ar-
vense
was most abundant at the Yellowstone Lake lo-
cation, reaching mean densities of
;
1100 stems/ha,
whereas
Lactuca serriola
did not differ in density
among locations (Table 8).
Patch size influenced the density of the native an-
nuals but not the exotic perennials (Table 8).
G. dif-
fusum
and
C. parviflora
were more abundant in the
large and moderate patches compared to the small
patches (Fig. 11a, b). Burn severity was important for
three of the four opportunistic species, and the direction
of the effect was the same: density was greatest with
the more severe crown-fire burns (Table 8).
C. arvense
was increasing through time in all burn severities, al-
though density was lowest in the light-surface burns
and greater in the stand-replacing burns (Fig. 11c).
Density of
L. serriola
was negligible in the light-sur-
face burns and peaked in the stand-replacing burns in
1991 (Fig. 11d).
G. diffusum
was most abundant in
crown-fire burns, and
C. parviflora
showed no response
to burn severity.
For the two native annuals, there was a significant
interaction between burn severity and location.
G. dif-
fusum
was always more abundant in the more severe
burns, but was decreasing through time at Cougar
Creek and Fern Cascades between 1991–1993 and in-
creasing at Yellowstone Lake.
C. parviflora
was most
abundant in the crown-fire and severe-surface burns at
Cougar Creek and Fern Cascades, but at Yellowstone

422
MONICA G. TURNER ET AL.
Ecological Monographs
Vol. 67, No. 4
F
IG
. 6. Annual density of sprouts of five forest species by study location. Data show means
6
2
SE
.
Lake was notably more abundant in severe-surface
burns than in crown fires. In addition, neither species
ever occurred in light-surface burns at Yellowstone
Lake.
Species richness
Species richness from 1991 through 1993 varied pri-
marily by year, location, and patchsize (Table 9). Over-
all richness increased through time, averaging 7.3 spe-
cies per 8-m
2
plot across all sampling points in 1991
and 11.8 species per plot in 1993. Among the three
study locations, species richness was lowest at Fern
Cascades and similar at Cougar Creek and Yellowstone
Lake (Fig. 12a). Among patch sizes, small patches con-
tained the greatest number of species, followed by large
and then moderate patches (Fig. 12c). Species richness
was also influenced by burn severity (Table 9), with
greater richness observed in the less severe burns (Fig.
12b). Burn severity also contributed to significant in-
teraction effects with location and patch size, however
(Table 9). At Cougar Creek, species richness was sim-
ilar across all burn severities (10.4–11.4 species). At
the Yellowstone Lake location, however, species rich-
ness decreased with increasing burn severity (average
of 15.9, 12.6, and 10.1 species for the light-surface,
severe-surface, and crown fires, respectively). At Fern
Cascades, species richness was similar in the surface
burns (9.4 and 8.3 species for light- and severe-surface
burns) and lower in areas of crown fire (6.3 species).
The interaction effect between burn severity and patch
size reflected a decline in species richness with in-
creasing burn severity in the moderate patches (11.9,
8.8, and 6.4 species for the light-surface, severe-sur-
face, and crown fires, respectively). Species richness
was not influenced by slope, aspect, distance to the
nearest unburned forest, or distance to the nearest se-
vere-surface burn (Table 9).
We observed a high degree of similarity between
years at each sample point (i.e., similarity between
1991 and 1992 and between 1992 and 1993), averaging
between 0.837 and 0.915 (Table 10). This high simi-
larity suggests that most species present at a point per-
sisted between years. Some of the variance in between-
year similarity was explained by differences in location
(ANOVA for similarity index results for 1991 vs. 1992:
r
2
5
0.15, df
5
25/515,
MSE
5
0.122,
F
5
3.64,
P
5
0.0001; ANOVA for similarity index results for 1992
vs. 1993:
r
2
5
0.20, df
5
25/551,
MSE
5
0.157,
F
5

November 1997 423
FIRE SIZE AND PATTERN
F
IG
. 7. Annual density of sprouts of five forest species by patch size. Data show means
6
2
SE
.
F
IG
. 8. Mean annual density of
Vaccinium scoparium
sprouts by burn severity. Data show means
6
2
SE
.
5.61,
P
5
0.0001). The Cougar Creek and Fern Cas-
cades locations had higher between-year similarities
than did the Yellowstone Lake location (Table 10), in-
dicating more community-level change at Yellowstone
Lake. The Cougar Creek location also had the narrow-
est range of similarity values, with a minimum value
of 0.61 compared to 0.40 for the other two locations
(Table 10). Between-year similarity was not influenced
by burn severity or by patch size, suggesting that the
rates of change in species composition were affected
primarily by broad-scale environmental gradients cap-
tured by differences among locations.
D
ISCUSSION
We organize our discussion by first addressing the
results for particular biotic response variables, focusing
especially on the differential traits among species that
may explain observed responses. We then synthesize
our findings to address the general question of how fire
size and pattern influence early secondary succession.
Trends in biotic response variables
Biotic cover.
—Biotic cover generally responded as
we hypothesized (Table 1), but more complex patterns

424
MONICA G. TURNER ET AL.
Ecological Monographs
Vol. 67, No. 4
T
ABLE
7. Summary of tests for between-sample-points effects on postfire density of first-year
seedlings of forest species from repeated-measures ANOVA. Entries indicate the
F
and
P
values for significant effects, and trends for main effects. The two most important effects
are highlighted in boldface type.
Effect
Epilobium
angustifolium Lupinus argenteus
Overall
r
2
for ANOVA fully nested
in time 0.41 0.24
Location† (df
5
2)
F
5
46.58
P
5
0.0001
L
.
F, C
F
5
3.99
P
5
0.0190
F
.
C, L
Patch size‡ (df
5
2)
F
5
4.08
P
5
0.0174
S
.
M
.
L
F
5
4.03
P
5
0.0183
L, M
.
S
Burn severity
\
(df
5
2)
NS NS
Presence of sprouts in 1990 or 1991 of
this species (df
5
1)
NS
F
5
26.33
P
5
0.0001
Greater when sprouts
present
Slope (df
5
1)
NS NS
Aspect (df
5
2)
NS NS
Distance to unburned forest (df
5
1)
NS NS
Distance to severe-surface burn (df
5
1)
NS NS
Location
3
patch size (df
5
4)
F
5
5.56
P
5
0.0002
NS
Location
3
burn severity (df
5
4)
F
5
4.24
P
5
0.0022
NS
Patch size
3
burn severity (df
5
4)
F
5
3.90
P
5
0.0040
NS
Location
3
patch size
3
burn severity
(df
5
7)
F
5
2.17
P
5
0.0359
NS
Notes:
Error df
5
498.
NS
indicates that an effect was not significant. Changes through time
were significant for both species.
† Abbreviations for locations are: C
5
Cougar Creek, F
5
Fern Cascades, and L
5
Yellow-
stone Lake.
‡ Abbreviations for patch sizes are: L
5
large, M
5
moderate, and S
5
small.
\
Abbreviations for burn severities are: 1
5
light-surface burn, 2
5
severe-surface burn, and
3
5
crown fire.
F
IG
. 9. Seedling density of
Lupinus argenteus
on sam-
pling points that did or did not contain sprouts of
L. argenteus
during 1990 or 1991. Data show means
6
2
SE
.
emerged. Local differences in plant community com-
position were important in interpreting cover patterns.
For example, the Yellowstone Lake and Cougar Creek
locations had comparable total biotic cover, but for
different reasons. Forbs contributed substantially to to-
tal cover at Yellowstone Lake, but tree seedling cover
was extremely low. In contrast, tree seedling cover
dominated at Cougar Creek, but forb cover was rela-
tively low.
The effect of patch size on biotic cover was as hy-
pothesized for total, forb, graminoid, and shrub cover
(Table 1), with more cover observed in small than in
large patches. Surprisingly, tree seedling cover exhib-
ited the opposite trend. We suspect that increased light
availability in large patches plus potentially less com-
petition from herbaceous flora may help explain this
result. The 1-ha patches were more shaded by the sur-
rounding forest and had more herbaceous cover, which
together may limit tree seedling growth. Although we
did not measure the size of lodgepole seedlings through

November 1997 425
FIRE SIZE AND PATTERN
F
IG
. 10. Proportion of sampling points for crown-fire
burn severity on which postfire seedlings and sprouts and
seedlings of
Epilobium angustifolium
(EPAN) and
Lupinus
argenteus
(LUAR) were present in each year following ex-
tensive forest fire in Yellowstone National Park in 1988.
T
ABLE
8. Summary of analysis of variance results for density of opportunistic species and
trends in between-subject effects.
Native annuals
Gayophytum
diffusum Collinsia
parviflora
Exotic perennials
Cirsium
arvense Lactuca
serriola
a) Full model ANOVA nested in time
r
2
df
MSE
F
P
0.46
130/1963
223.46
12.73
0.0001
0.39
130/1963
95.23
9.74
0.0001
0.30
130/1963
21.85
6.61
0.0001
0.19
130/1963
4.70
3.45
0.0001
b) Between-subject effects from repeated-measures ANOVA
Year
Location†
Patch‡
Burn severity
\
Slope
Dist. to light-surface
Dist. to severe-surface
Aspect
Location
3
patch
Location
3
burn
Patch
3
burn
Location
3
patch
3
burn
increasing
C
.
F
.
L
L
.
M, S
3
.
2, 1
NS
NS
NS
NS
NS
***
NS
NS
increasing
C
.
F, L
L
.
M, S
NS
NS
positive
NS
NS
***
***
NS
NS
increasing
L
.
F, C
NS
3
.
2
.
1
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
3
.
2
.
1
NS
NS
positive
NS
NS
NS
NS
NS
***
P
,
0.0001;
NS
5
not significant.
† Abbreviations for locations are: C
5
Cougar Creek, F
5
Fern Cascades, and L
5
Yellow-
stone Lake.
‡ Abbreviations for patch sizes are: L
5
large, M
5
moderate, and S
5
small.
\
Abbreviations for burn severities are: 1
5
light-surface burn, 2
5
severe-surface burn, and
3
5
crown fire.
time, individual plants may grow more rapidly in open
areas (J. E. Anderson,
personal communication
).
The strong effect of burn severity on shrub cover
and lack of any significant changes through time prob-
ably reflect the slow-growing habit of
Vaccinium sco-
parium,
the dominant shrub. We never observed a
Vac-
cinium scoparium
seedling in either burned or unbur-
ned forests in Yellowstone, although we observed ber-
ries with viable seeds on plants in unburned forests
(Romme et al. 1995); this response is not unusual for
Vaccinium
spp. or ericaceous shrubs in general (Ma-
tlack et al. 1993, Motzkin et al. 1996). All postfire
shrub regeneration to date appeared to be vegetative.
Hence, shrubs grew only in areas in which they oc-
curred prior to the fire, and cover was strongly influ-
enced by local fire severity. Rhizomes of mature
Vac-
cinium scoparium
are in the litter and upper soil layers,
where they are especially vulnerable to fire damage.
Lodgepole pine reestablishment.
—In contrast to our
expectation (Table 1), lodgepole pine seedling density
was greatest in areas of severe-surface burn rather than
in crown fires, although tree seedling cover did not
differ between these two severities of stand-replacing
burn. Lodgepole pine seedlings germinate best in full
sunlight and exposed mineral soil, and the more open-
grown seedlings in the crown fires may have grown
more rapidly than the more crowded seedlings in the
severe-surface burns. However, spatial variation in
seedling density as a function of burn severity in stand-
replacing fires has only recently been reported (An-
derson and Romme 1991). Fire intensities and spread
rates in severe-surface burns may have been optimal
for the opening of serotinous cones and release of seed,
but fire conditions in severe crown fires may have re-
sulted in cone ignition or substantially reduced seed

426
MONICA G. TURNER ET AL.
Ecological Monographs
Vol. 67, No. 4
F
IG
. 11. Effects of patch size and burn severity on density of opportunistic species: two native annuals,
Gayophytum
diffusum
and
Collinsia parviflora,
influenced by patch size; and two exotic species,
Cirsium arvense
and
Lactuca serriola,
influenced by burn severity. Data show means
6
2
SE
.
T
ABLE
9. Summary of repeated-measures analysis of variance for species richness in burned
forests of the Yellowstone Plateau as measured from 1991 to 1993.
Effect df
MSE
FP
Location
Patch size
Burn severity
Slope
Distance to unburned forest
Distance to severe-surface burn
Aspect
Location
3
patch size
Location
3
burn severity
Patch size
3
burn severity
Location
3
patch size
3
burn severity
2
2
3
1
1
2
2
4
4
4
7
1640.2
1090.2
186.6
1.9
0.6
44.3
20.1
73.9
216.9
146.4
76.1
72.14
47.95
8.21
0.09
0.03
1.95
0.88
3.25
9.54
6.44
3.35
0.0001
0.0001
0.0001
0.7672
0.8688
0.1632
0.4139
0.0119
0.0001
0.0001
0.0017
Error 544 22.7
Note:
Overall
r
2
5
0.60 for ANOVA fully nested in time.
viability (Johnson and Gutsell 1993). Rather than un-
burned forest, it was these severe-surface burns that
served as the major seed source for stand regeneration
in crown fires, as illustrated by the negative correlation
of seedling density and distance to severe-surface burn.
Some seeds also survived in most crown fires, es-
pecially where serotiny was high, as pine seedlings
were found in areas even
.
200 m from the nearest
severe-surface burn.
Lodgepole pine seedling density did not change be-
tween 1990 and 1993. Within-stand variability in ages
for lodgepole pine stands that regenerate following
stand-replacing fires is generally low (Horton 1953,
Tande 1979, Muir 1993). In areas of high serotiny,
lodgepole pine recruitment may continue through the
first decade following fire, as
;
10–15 yr of cones may
be available to supply seeds (Johnson and Fryer 1989).
The duration of time over which present spatial vari-
ation in lodgepole pine seedling density will persist
across the YNP landscape is not known. The density

November 1997 427
FIRE SIZE AND PATTERN
F
IG
. 12. Annual species richness in burned forests of the
Yellowstone Plateau: (a) by location, (b) by burn severity,
and (c) by patch size. Data show means
6
2
SE
.
T
ABLE
10. Values of similarity index comparing species
richness at individual sampling points between years for
the three study locations.
Statistic Cougar
Creek Fern
Cascades Yellowstone
Lake
a) Similarity index for 1991 vs. 1992
Mean
N
SD
Minimum
Maximum
0.884
156
0.088
0.615
1.000
0.867
214
0.116
0.400
1.000
0.810
188
0.121
0.400
1.000
b) Similarity index for 1992 vs. 1993
Mean
N
SD
Minimum
Maximum
0.915
168
0.080
0.667
1.000
0.903
226
0.087
0.500
1.000
0.837
200
0.111
0.400
1.000
of older lodgepole stands varied substantially across
the YNP landscape, and these variations in tree density
within stands of similar age may have resulted, at least
in part, from the spatial heterogeneity of burn severities
in past fires.
Lodgepole seedling density was positively related to
the number of prefire serotinous trees surrounding the
sampling point (see also Tinker et al. 1994). Any par-
ticular stand of lodgepole pine may contain a mixture
of serotinous trees and trees having cones that open at
maturity. The percentage of serotinous trees is extreme-
ly variable in Greater Yellowstone, ranging from 0 to
72% and being greatest at low-to-mid elevation (1900–
2300 m) sites (Lotan 1975, Ellis et al. 1994). The three
locations we studied varied from a mean of 2% serot-
inous trees at Yellowstone Lake to 65% at Cougar
Creek (Table 2). Tinker et al. (1994) found scale-de-
pendent variation in the occurrence of serotinous cones
in YNP, with low variability in percent serotiny ob-
served across short distances (
,
1 km) and large dis-
tances (
.
10 km) and high variability over intermediate
distances (1–10 km). The strong differences in lodge-
pole seedling density among our three geographic lo-
cations reflected broad-scale patterns of variability in
serotiny levels.
Reestablishment of forest species.
—Forest species
responded individualistically to the 1988 fires. Some
species (e.g.,
Lupinus argenteus, Vaccinium scopar-
ium,
and
Carex geyeri
) demonstrated a negative rela-
tionship between sprout density and fire severity. Other
species, (e.g.,
Epilobium angustifolium
and
Arnica cor-
difolia
), achieved greater sprout densities in more se-
vere burns. Differences in depth distribution of rhi-
zomes in the soils may be most important for survival
and subsequent resprouting of individuals and species
(Granstrom and Schimmel 1993).
Epilobium angustifolium
plants readily survive fire
and produce great quantities of easily dispersed seeds
that germinate quickly in open sites. Dormant buds at
depths of 2–8 cm sprout during the first or second sub-
sequent growing season. Seed dispersal may extend
over hundreds of kilometers (Archibold 1980, Solbreck
and Andersson 1986), and seeds germinate promptly
in suitable growing conditions (Granstrom 1987, Rom-
me et al. 1995).
E. angustifolium
flowered profusely in
YNP in 1990 and appeared to reach its peak in 1991
when it formed thick patches of waist-high, flowering
stems in many areas. Subsequently, its stature de-
creased, flowering declined, and density of seedlings
was reduced.
Lupinus argenteus
appears relatively poorly adapted
to fire, having heavy seeds with limited dispersal ca-
pabilities (Wood and del Moral 1988) that require scar-

428
MONICA G. TURNER ET AL.
Ecological Monographs
Vol. 67, No. 4
ification to ensure rapid germination (Oberbauer and
Miller 1982, Romme et al. 1995). Following the fires,
L. argenteus
plants sprouted in many areas of YNP and
were subsequently surrounded by seedlings. On sites
where adults were killed or were absent before the fires,
however,
Lupinus
was rare or absent by the end of our
study.
Local dispersal from reproductive plants that sur-
vived the fire appeared to be the dominant mechanism
for seedling establishment of forest herbs. Although
long-distance dispersal is not typically required for ini-
tial colonization in small disturbed patches (e.g.,
,
0.5
ha; Holt et al. 1995), we were surprised by the relative
unimportance of long-distance dispersal for forest spe-
cies in the larger patches. However, seedling densities
varied considerably among the three geographic lo-
cations, suggesting that seed sources may vary sub-
stantially across the Yellowstone Plateau—likely due
to prefire species distributions. When combined with
our anecdotal observations of the spatial distribution
of other forest species (e.g.,
Hieracium albiflorum, Cal-
amagrostis canadensis, Calamagrostis rubescens
),
postfire vegetation patterns suggested that seed dis-
persal into the burned areas from the surrounding un-
burned forest was not an important mechanism for the
reestablishment of forest species.
Opportunistic species.
—The two native annuals
showed a response to patch size, being more abundant
in the large than in the small patches, but the exotic
perennials did not. The exotics, however, showed a
response to burn severity; both were more abundant in
the more severely burned areas, and Canada thistle was
still increasing in density when our study ended in
1993. We hypothesized no relationship between dis-
tance to nearest unburned or light-surface burned area
and density of opportunistic species, and this hypoth-
esis was rejected for only one species,
Collinsia par-
viflora,
which increased with distance. This species
may have been better able to establish in the centers
of the larger patches where competition from other her-
baceous species was reduced.
The annual species (
Gayophytum diffusum
and
Col-
linsia parviflora
) may go through a relatively quick
‘‘boom and bust’’ cycle. Our field observations suggest
their abundance may have begun to decline in 1993
when the plants were much smaller and less robust than
at their peak in 1991 (M. G. Turner and W. H. Romme,
personal observation
). The exotic Canada thistle, how-
ever, showed no sign of decline in 1993 and may have
been able to expand its range following the fires.
Species richness.
—The number of species per 8-m
2
plot increased significantly during the first 5 yr of post-
fire succession at all locations. However, there were
differences in richness among locations, patch sizes,
and burn severity classes. The Yellowstone Lake lo-
cation had the lowest mean index of similarity from
year to year, indicating more rapid addition of new
species (or turnover of previous species) in the plant
communities in this area. The highest mean similarity
index was seen at Cougar Creek, indicating a slow rate
of change in plant community composition. The Cougar
location also had the greatest lodgepole pine seedling
densities, suggesting that dense tree seedlings may
have inhibited development of the herbaceous com-
munity. However, more experimental studies clearly
are needed to establish the mechanisms controlling rate
of compositional change.
Richness on a per unit area basis decreased as patch
size increased. We originally hypothesized that this
would be the case because of reduced seed dispersal
from the unburned forest into larger openings. How-
ever, in light of our finding that distance from unburned
edge was rarely an important variable in predicting
densities of sprouts and seedlings, we suggest that the
lower species richness observed in larger burned patch-
es may have been a result of relatively more severe
abiotic conditions (higher light intensity, evapotrans-
piration, and diurnal temperature variation) in these
environments.
Richness was generally lowest in crown-fire areas,
as we hypothesized, but was inconsistent with respect
to the other burn classes. At no location did we see
highest richness in severe-surface burns, as originally
hypothesized. At Cougar Creek and Fern Cascades,
there was no significant difference in richness between
severe-surface and light-surface burns, whereas at Yel-
lowstone Lake richness was greater in the less severely
burned areas. It is possible that the Yellowstone Lake
location showed clearer patterns of response to burn
severity in part because richness was higher and com-
position was changing more rapidly there than at either
of the other two locations.
Significance of the fire-created pattern for
postfire succession
Burn severity and patch size had significant effects
on nearly every biotic response variable we measured,
suggesting an important influence of the fire-created
mosaic on postfire succession. In general, the effects
of burn severity in Yellowstone (Table 11) conformed
to effects observed elsewhere in the Rocky Mountains
(e.g., Habeck and Mutch 1973, Lyon and Stickney
1976, Viereck 1983, Ryan and Noste 1985). The most
enduring legacy of the mosaic of burn severities may
be spatial variability in lodgepole pine density across
the landscape. Areas of severe-surface burn may de-
velop persistent high-density stands of lodgepole pine
that grade into areas of lower density. In the shorter
term, areas of crown fire provided the best sites for
opportunistic species to colonize, although we do not
yet know how long they will persist. Our results sug-
gested that the natural variability in fire severity across
the landscape is an important source of heterogeneity
for the plant community (Table 11).
Patch size influenced early postfire succession in
YNP (Table 12), even when no effect was hypothesized

November 1997 429
FIRE SIZE AND PATTERN
T
ABLE
11. Synthesis of the effects of fire severity on biotic response variables during early postfire succession in Yellowstone
National Park, Wyoming.
Attribute Crown fire Low intensity surface fire
Biotic cover Higher cover of tree seedlings (es-
pecially
Pinus contorta
); lower cover
of herbs
Higher cover of shrubs and herbs
(forbs and graminoids); lower cover
of tree seedlings
Density of
Pinus contorta
seedlings Lower density of seedlings, although
the seedlings were generally larger Higher density of seedlings, but seed-
lings generally smaller
Density of forest understory species Higher sprout densities of some forest
species (e.g.,
Epilobium angustifol-
ia
); lower densities of many others
Higher sprout densities in many forest
species (e.g.,
Vaccinium scoparium,
Lupinus argenteus, Carex geyerii
)
Opportunistic species Greater densities of opportunists (e.g.,
Gayophytum diffusum, Cirsium ar-
vense, Lactuca serriola
)
Lower densities, with the two native
annuals (
G. diffusum
and
C. parviflo-
ra
) not occurring at all in light-sur-
face burns
Plant species richness Lower richness of vascular plants Higher richness of vascular plants
T
ABLE
12. Synthesis of the effects of patch size on biotic response variables during early postfiresuccession in Yellowstone
National Park, Wyoming.
Attribute Large patches Small patches
Biotic cover Higher cover of woody plants (tree
seedlings and shrubs); lower cover of
herbs
Higher cover of herbaceous plants
(forbs and graminoids); lower cover
of trees and shrubs
Density of
Pinus contorta
seedlings Greater density Lower density
Density of forest understory species Higher sprout densities of some forest
species (e.g.,
Carex rossii
); lower
densities of many others
Higher sprout densities of many forest
species (e.g.,
Vaccinium scoparium,
Epilobium angustifolium, Arnica cor-
difolia,
and
Carex geyerii
)
Opportunistic species Higher densities of the two native an-
nuals, no effect on the two exotic
species
Lower density of the native annuals,
some delay in colonizing small
patches
Plant species richness Lower richness of vascular plant spe-
cies Higher richness of vascular plant spe-
cies
(e.g., for sprout density and species richness). Large
patches had greater densities of lodgepole pine seed-
lings, higher densities of two native annuals considered
opportunistic species, lower species richness, and low-
er cover of herbs. The effect of patch size may be due,
in part, to differential fire intensities that generated
large and small patches. Previous spatial analyses in-
dicated that the proportion of burned area in crown fire
increased with increasing area burned during a day
(Turner et al. 1994
b
). Thus, the crown fires that gen-
erated our large patches may have burned with greater
intensity than the spotty fires that created the small
patches of crown fire. Greater fire intensities would lead
to reduced survival of the prefire plant community and
offer more potential area for colonization by oppor-
tunists. Compared with large patches, small patches are
likely to provide cooler, moister growing conditions
with more shade and wind protection, all of which may
enhance plant survival and growth. Snow ablation (i.e.,
loss by melting or evaporation) in forest openings in
southwest Alberta, Canada, also increases with the size
of the opening, indicating warmer temperatures and
more rapid drying in larger patches (Berry and Roth-
well 1992). Thus, patch size effects may also be due,
in part, to microclimate conditions.
Despite the statistically and biologically significant
effects of fire severity and patch size, geographic lo-
cation was often the most important independent vari-
able in the statistical analyses, reflecting substantial
differences in the plant community among the three
study locations (Table 13). Thus, the best predictor of
postfire vegetation in YNP may be the prefire vege-
tation as influenced by stand history and abiotic gra-
dients. Although we were aware of the ranges of ele-
vation, substrate, and forest ages across the three lo-
cations at the outset of this study (Table 2), the wide
variability in the mean proportion of serotinous lodge-
pole pines across the landscape was a surprise. Our
sites spanned this range of variability, and mean ser-
otiny was strongly related to location. Weconclude that
the effects of fire size and heterogeneity were important
and that at least some will be long lasting, but these
landscape-scale effects of fire occurred within an over-
riding context of gradients, especially in serotiny, on
a broader scale.
Succession across much of YNP appeared in 1993

430
MONICA G. TURNER ET AL.
Ecological Monographs
Vol. 67, No. 4
T
ABLE
13. Synthesis of differences among the three study locations (see Table 2) in earlypostfire succession in Yellowstone
National Park, Wyoming.
Attribute Cougar Creek Fern Cascades Yellowstone Lake
Biotic cover Highest cover of tree seed-
lings (especially
Pinus
contorta
), shrubs (es-
pecially
Ceanothus velu-
tinus
), and graminoids
(especially
Carex rossii
and
Calamagrostis rubes-
cens
)
Lowest total biotic cover Highest cover of forbs (es-
pecially
Arnica cordifolia
and
Epilobium angustifol-
ium
)
Density of
Pinus contorta
seedlings Highest Intermediate Very low
Density of forest understory
species Higher densities of sprouts
of some forest species
(e.g.,
Carex rossii
); lower
densities of others
Higher densities of sprouts
of many forest species
(e.g.,
Vaccinium scopar-
ium, Lupinus argenteus,
and
Carex geyerii
)
Higher densities of sprouts
of many forest species
(e.g.,
Vaccinium scopar-
ium, Epilobium angusti-
folium, Lupinus argen-
teus,
and
Arnica cordifol-
ia
)
Opportunistic species Higher densities of native
annuals
Gayophytum dif-
fusum
and
Collinsia par-
viflora
Moderate densities of na-
tive annuals
Gayophytum
diffusum
and
Collinsia
parviflora
Lowest density of native
annuals
Gayophytum dif-
fusum
and
Collinsia par-
viflora
; highest density of
exotic
Cirsium arvense
Plant species richness High Low High
to be moving toward plant communities very similar
to those that burned in 1988. For example, dense ‘‘dog-
hair’’ thickets of lodgepole pine occurred at Cougar
Creek in 1988, and extensive patches of lodgepole pine
seedlings
.
50 stems/m
2
were present in 1993. Prefire
stands were more open at Fern Cascades, and pine seed-
lings in 1993 were substantially less dense (
,
1 stem/
m
2
) than at Cougar Creek. Moreover, except for some
extremely fire-sensitive, late-successional species like
Linnaea borealis
(Eriksson 1992), all plant species
present in unburned forests appeared to occur in nearby
burned areas (although densities differed).
Forest reestablishment appeared questionable, how-
ever, at our Yellowstone Lake location. In 1993, 5 yr
postfire, we found
,
10 tree seedlings/ha (pine, spruce,
and fir combined). Seed viability is
,
5 yr in lodgepole
pine, 3 yr in Engelmann spruce, and only 1 yr in sub-
alpine fir (Archibold 1989, Johnson and Fryer 1989),
suggesting that the opportunity for immediate postfire
tree seedling establishment had been missed. Different
successional trajectories may be initiated within similar
abiotic environments because of local variation in dis-
turbance intensity or availability of plant propagules
(Glenn-Lewin and van der Maarel 1992, Fastie 1995,
Baker and Walford 1995); the paucity of tree seedlings
at the Yellowstone Lake location may be due to effects
of very severe fire on propagule availability. Refor-
estation of this area where the local sources for tree
seeds were apparently destroyed by the fire will depend
on seed dispersal from sources outside the burned area,
but much of the 3700-ha large patch is well beyond
the effective dispersal distance of conifer seeds (Ar-
chibold 1989, Johnson and Fryer 1989). Furthermore,
conifer seeds that gradually disperse into the area may
be unable to establish because of competition from the
well-developed herbaceous community (Lotan and Per-
ry 1983, Coates et al. 1991, Lieffers et al. 1993). Thus,
although only from the earliest stages of postfire suc-
cession in YNP, our data suggest that pathways of suc-
cession potentially leading to nonforest communities
were initiated following the 1988 fires.
Concluding remarks
The answer to our question of whether vegetation
responses differed between small and large burned
patches is not a simple yes or no. Based on our data
for early postfire succession, most of the burned forests
in YNP appeared to be reestablishing community com-
positions similar to the prefire vegetation, primarily
because extensive biotic residuals persisted even within
very large burned areas. Thus, large fires presently are
not a threat to plant communities in YNP. Nonetheless,
fire size, fire severity, and the spatial heterogeneity of
burn severities across the landscape significantly influ-
enced postfire succession. The legacy of variable lodge-
pole pine density across the landscape and potential
conversion from forest to nonforest vegetation in some
areas may be detectable for decades to centuries. Fur-
ther study of these longer term dynamics is clearly
needed. Predictability of successional trajectories has
received relatively little study (Peet 1992, Wood and
del Moral 1993), and it remains challenging to identify
the factors controlling vegetation dynamics at multiple
spatial scales.
Large fires will likely occur again in YNP, and many
other areas in western and northern North America are

November 1997 431
FIRE SIZE AND PATTERN
covered by vast expanses of coniferous forest subject
to infrequent, large, severe fires (Heinselman 1973,
Hemstrom and Franklin 1982, Turner and Romme
1994). Understanding the effects of large fires assumes
added significance when potential implications of glob-
al climate change are considered. Variation in area
burned in the Yellowstone region is related to the in-
creased aridity observed there since 1895 (Balling et
al. 1992
a, b
). Fire frequency may increase in YNP if
the future climate becomes warmer and drier (Romme
and Turner 1991, Gardner et al. 1996). Since the late
1970s, there has been a marked increase in the annual
extent of wildfires in Canada and the western United
States (Van Wagner 1988, Flannigan and Van Wagner
1991 Auclair and Carter 1993), although whether this
reflects a global warming trend cannot yet be discerned.
Continued research on succession following the 1988
Yellowstone fires should enhance our ability to project
future landscape condition under altered fire regimes
and provide valuable insight into the patterns and pro-
cesses expected in other landscapes affected by large
infrequent disturbances.
A
CKNOWLEDGMENTS
We are extremely grateful to the following people who
assisted with the field work: Laura Bohland, Teresa Cary,
Yvonne Corcoran, Christine Eckhardt, Jeremy Gardner, Na-
than Gardner, Sandi Gardner, Dorothy Goigel, Priscilla Gol-
ley, John Goodlaxson, Katherine Gould, Gregory Haines, Ter-
ri Hargrove, Leonard Hordijk, Sim Huang, Thomas Knight,
Susan Lindahl, Stephen Nold, Cyndi Persichetty O’Hara, Mi-
chael O’Hara, Anita Parker, Mitzi Pearson, Scott Pearson,
Howard Romme, Janelle Stith, Michael Turner, Dan Tinker,
Tim Twining, Joe Van Landschoot, and Michi Vojta. The
University of Wyoming-National Park Service Research Cen-
ter provided excellent logistical support, and we especially
thank Mark Boyce, Glen Plumb and Hank Harlow for their
assistance. The staff of Yellowstone National Park shared
their complete GIS data bases and provided much-appreciated
logistical assistance. We are especially grateful toSouth Dis-
trict Ranger Gerry Mernon for assistance with our trips to
the South Arm of Yellowstone Lake. John Beauchamp from
ORNL provided guidance on the statistical analyses, and Sid-
ey Timmons helped with the GIS work required for study site
selection. Our ideas benefited from discussions with Park
Biologist Don Despain. The manuscript was improved by
thoughtful suggestions from Virginia H. Dale and Scott M.
Pearson and insightful reviews from David R. Foster and three
anonymous reviewers. This research was supported by fund-
ing from the National Science Foundation (BSR-9016281 and
BSR-9018381), National Geographic Society (Grant No.
4284-90), and the Ecological Research Division, Office of
Health and Environmental Research, U.S. Department of En-
ergy, under contract no. DE-AC05-84OR21400 with Martin
Marietta Energy Systems.
L
ITERATURE
C
ITED
Anderson, J. E., and W. H. Romme. 1991. Initial floristics
in lodgepole pine (
Pinus contorta
) forests following the
1988 Yellowstone fires. International Journal of Wildland
Fire 1:119–124.
Archibold, O. W. 1980. Seed input into a postfire forest site
in northern Saskatchewan. Canadian Journal of Forest Re-
search 10:129–134.
. 1989. Seed banks and vegetation processes in co-
niferous forests. Pages 107–122
in
M. A. Leck, V. T. Parker,
and R. L. Simpson, editors. Ecology of soil seed banks.
Academic Press, San Diego, California, USA.
Auclair, A. N. D., and T. B. Carter. 1993. Forest wildfires
as a recent source of CO
2
at northern latitudes. Canadian
Journal of Forest Research 23:1528–1536.
Baker, W. L., and G. M. Walford. 1995. Multiple stable states
and models of riparian vegetation succession on the Animas
River, Colorado. Annals of the Association of American
Geographers 85:320–338.
Balling, R. C., Jr., G. C. Meyer, and S. G. Wells. 1992
a.
Relation of surface climate and burned area in Yellowstone
National Park. Agricultural and Forest Meteorology 60:
285–293.
Balling, R. C., Jr., G. C. Meyer, and S. G. Wells. 1992
b.
Climate change in Yellowstone National Park: is the
drought-related risk of wildfires increasing? Climatic
Change 22:35–45.
Berry, G. J., and R. L. Rothwell. 1992. Snow ablation in
small forest openings in southwest Alberta. Canadian Jour-
nal of Forest Research 22:1326–1331.
Bessie, W. C., and E. A. Johnson. 1995. The relative im-
portance of fuels and weather on fire behavior in subalpine
forests. Ecology 76:747–762.
Boose, E. R., D. R. Foster, and M. Fluet. 1994. Hurricane
impacts to tropical and temperate forest landscapes. Eco-
logical Monographs 64:369–400.
Callaway, R. M, and F. W. Davis. 1993. Vegetation dynamics,
fire, and the physical environment in coastal central Cali-
fornia. Ecology 74:1567–1578.
Christensen, N. L., J. K. Agee, P. F. Brussard, J. Hughes, D.
H. Knight, G. W. Minshall, J. M. Peek, S. J. Pyne, F. J.
Swanson, J. W. Thomas, S. Wells, S. E. Williams, and H.
A. Wright. 1989. Interpreting the Yellowstone fires of
1988. BioScience 39:678–685.
Coates, K. D., W. H. Emmingham, and S. R. Radosevich.
1991. Conifer-seedling success and microclimate at dif-
ferent levels of herb and shrub cover in a
Rhododendron-
Vaccinium-Menziesia
community of south central British
Columbia. Canadian Journal of Forest Research 21:858–
866.
Connell, J. H., and R. O. Slatyer. 1977. Mechanisms of suc-
cession in natural communities and their role in community
stability and organization. American Naturalist 111:1119–
1144.
Dayton, P. K., M. J. Tegner, P. E. Parnell, and P. B. Edwards.
1992. Temporal and spatial patterns of disturbance and
recovery in a kelp forest community. Ecological Mono-
graphs 62:421–445.
Denslow, J. D. 1980
a.
Gap partitioning among tropical trees.
Biotropica 12(Supplement):47–55.
. 1980
b.
Patterns of plant species diversity during
succession under different disturbance regimes. Oecologia
46:18–21.
Despain, D. G. 1990. Yellowstone vegetation: consequences
of environment and history. Roberts Rinehart, Boulder,
Colorado, USA.
Dirks, R. A., and B. E. Martner. 1982. The climate of Yel-
lowstone and Grand Teton National Parks. Occasional Pa-
per No. 6, U.S. National Park Service, Washington, D. C.,
USA.
Dorn, R. D. 1992. Vascular plants of Wyoming. Second edi-
tion. Mountain West Publishing, Cheyenne, Wyoming,
USA.
Ellis, M., C. D. von Dohlen, J. E. Anderson, and W. H. Rom-
me. 1994. Some important factors affecting density of
lodgepole pine seedlings following the 1988 Yellowstone
fires. Pages 139–150
in
D. Despain, editor.Plants and their
environments: proceedings of the first biennial scientific
conference on the Greater Yellowstone Ecosystem. Tech-
nical Report NPS/NRYE/NRTR-93/XX. U.S. Department

432
MONICA G. TURNER ET AL.
Ecological Monographs
Vol. 67, No. 4
of the Interior, National Park Service, Natural Resources
Publication Office, Denver, Colorado, USA.
Eriksson, O. 1992. Population structure and dynamics of the
clonal dwarf-shrub
Linnaea borealis.
Journal of Vegetation
Science 3:61–68.
Fastie, C. L. 1995. Causes and ecosystem consequences of
multiple pathways of primary succession at Glacier Bay,
Alaska. Ecology 76:1899–1916.
Finegan, B. 1984. Forest succession. Nature 312:109–114.
Flannigan, M. D., and C. E. Van Wagner. 1991. Climate
change and wildfire in Canada. Canadian Journal of Forest
Research 21:66–72.
Floyd, D. A., and J. E. Anderson. 1982. A new point inter-
ception frame for estimating cover of vegetation. Vegetatio
50:185–186.
Floyd, D. A., and J. E. Anderson. 1987. A comparison of
three methods for estimating plant cover. Journal of Ecol-
ogy 75:221–228.
Foster, D. R. 1988
a.
Disturbance history, community orga-
nization and vegetation dynamics of the old-growth Pisgah
Forest, southwestern New Hampshire, USA. Journal of
Ecology 76:105–34.
. 1988
b.
Species and stand response to catastrophic
wind in central New England, USA. Journal of Ecology
76:135–51.
Frelich, L. E., and C. G. Lorimer. 1991. Natural disturbance
regimes in hemlock-hardwood forests of the Upper Great
Lakes region. Ecological Monographs 61:145–164.
Gardner, R. H., W. W. Hargrove, M. G. Turner, and W. H.
Romme. 1996. Global change, disturbances and landscape
dynamics. Pages 149–172
in
B. Walker and W. Steffen,
editors. Global change and terrestrial ecosystems. Cam-
bridge University Press, Cambridge, UK.
Glenn-Lewin, D. C., and E. van der Maarel. 1992. Patterns
and processes of vegetation dynamics. Pages 11–59
in
D.
C. Glenn-Lewin, R. K. Peet, and T. T. Veblen. Plant suc-
cession—theory and prediction. Chapman and Hall, New
York, New York, USA.
Granstrom, A. 1987. Seed viability of fourteen species dur-
ing five years of storage in a forest soil. Journal of Ecology
75:321–331.
Granstrom, A., and J. Schimmel. 1993. Heat effects on seeds
and rhizomes of a selection of boreal forest plants and
potential reaction to fire. Oecologia 94:307–313.
Green, D. G. 1989. Simulated effects of fire, dispersal and
spatial pattern on competition within forest mosaics. Ve-
getatio 82:139–153.
Habeck, J. R., and R. W. Mutch. 1973. Fire-dependent forests
in the northern Rockies. Quaternary Research 3:408–424.
Halpern, C. B. 1988. Early successional pathways and the
resistance and resilience of forest communities. Ecology
69:1703–1715.
. 1989. Early successional patterns of forest species:
interactions of life history traits and disturbance. Ecology
70:704–720.
Hanski, I. 1993. Coexistence of competitors in patchy en-
vironment. Ecology 64:493–500.
Hartshorn, G. S. 1980. Neotropical forest dynamics. Biotro-
pica 12(Supplement):23–30.
Heinselman, M. L. 1973. Fire in the virgin forests of the
Boundary Waters Canoe Area, Minnesota. Quaternary Re-
search 3:329–382.
Hemstrom, M. A., and J. F. Franklin. 1982. Fire and other
disturbances of the forests in Mount Rainier National Park.
Quaternary Research 18:32–51.
Holt, R. D., G. R. Robinson, and M. S. Gaines. 1995. Veg-
etation dynamics in an experimentally fragmented land-
scape. Ecology 76:1610–1624.
Horton, K. W. 1953. Causes of variation in the stocking of
lodgepole pine regeneration following fire. Canada De-
partment of Northern Affairs and Natural Resources, For-
estry Research Division. Silviculture Leaflet Number 95.
Johnson, E. A. 1975. Buried seed populations in the subarctic
forest east of Great Slave Lake, Northwest Territories. Ca-
nadian Journal of Botany 53:2933–2941.
Johnson, E. A., and G. I. Fryer. 1989. Population dynamics
in lodgepole pine-Engelmann spruce forests. Ecology 70:
1335–1345.
Johnson, E. A., and S. L. Gutsell. 1993. Heat budget and
fire behavior associated with the opening of serotinous
cones in two
Pinus
species. Journal of Vegetation Science
4:745–750.
Kareiva, P. 1990. Population dynamics in spatially complex
environments: theory and data. Philosophical Transactions
of the Royal Society of London B330:175–190.
. 1994. Space: the final frontier for ecological theory.
Ecology 75:1.
Levin, S. A. 1992. The problem of pattern and scale in ecol-
ogy. Ecology 73:1943–1983.
Lieffers, V. J., S. E. MacDonald, and E. H. Hogg. 1993.
Ecology of and control strategies for
Calamagrostis can-
adensis
in boreal forest sites. Canadian Journal of Forest
Research 23:2070–2077.
Lotan, J. E. 1975. The role of cone serotiny in lodgepole
pine forests. Pages 471–495
in
D. M. Baumgartner,editor.
Management of lodgepole pine ecosystems, symposium
proceedings. Cooperative Extension Service, Washington
State University, Pullman, Washington, USA.
Lotan, J. E., and D. A. Perry. 1983. Ecology and regeneration
of lodgepole pine. Agricultural Handbook 606, Washing-
ton, D.C., USA.
Lyon, L. J., and P. F. Stickney. 1976. Early vegetal succession
following large northern Rocky Mountain wildfires. Pages
355–375
in
Proceedings of the Montana Tall Timbers Fire
Ecology Conference and Fire and Land Management Sym-
posium, Number 14. Tall Timbers Research Station, Tal-
lahassee, Florida, USA.
Malanson, G. P. 1984. Intensity as a third factor of distur-
bance regime and its effect on species diversity. Oikos 43:
411–413.
Matlack, G. R., D. J. Gibson, and the late R. E. Good. 1993.
Clonal propagation, local disturbance, and the structure of
vegetation: ericaceous shrubs in the Pine Barrens of New
Jersey. Biological Conservation 63:1–8.
Miller, T. E. 1982. Community diversity and interactions
between the size and frequency of disturbance. American
Naturalist 120:533–536.
Motzkin, G., D. Foster, A. Allen, J. Harrod, and R. Boone.
1996. Controlling site to evaluate history: vegetation pat-
terns of a New England sand plain. Ecological Monographs
66:345–365.
Muir, P. S. 1993. Disturbance effects on structure and tree
species composition of
Pinus contorta
forests in western
Montana. Canadian Journal of Forest Research 23:1617–
1625.
Noble, I. R., and R. O. Slatyer. 1980. The use of vital at-
tributes to predict successional changes in plant commu-
nities subject to recurrent disturbances. Vegetatio 43:5–21.
Oberbauer, S., and P. C. Miller. 1982. Effect of water po-
tential on seed germination. Holarctic Ecology 5:218–220.
Pearson, S. M. 1993. The spatial extent and relative influence
of landscape-level factors on wintering bird populations.
Landscape Ecology 8:3–18.
Peet, R. K. 1992. Community structure and ecosystem func-
tion. Pages 103–151
in
D. C. Glenn-Lewin, R. K. Peet, and
T. T. Veblen. Plant succession—theory and prediction.
Chapman and Hall, New York, New York, USA.
Peet, R. K., and N. L. Christensen. 1980. Succession: a pop-
ulation process. Vegetatio 43:131–140.
Peterjohn, W. T., and D. L. Correll. 1984. Nutrient dynamics

November 1997 433
FIRE SIZE AND PATTERN
in an agricultural watershed: observations on the role of a
riparian forest. Ecology 65:1466–75.
Peterson, C. J., and S. T. A. Pickett. 1995. Forest reorga-
nization: a case study in an old-growth forest catastrophic
blowdown. Ecology 76:763–774.
Pickett, S. T. A. 1976. Succession: an evolutionary inter-
pretation. American Naturalist 110:107–119.
Pickett, S. T. A., S. L. Collins, and J. J. Armesto. 1987.
Models, mechanisms, and pathways of succession. Botan-
ical Review 53:335–371.
Pielou, E. C. 1974. Population and community ecology,prin-
ciples and methods. Gordon and Breach Science, New
York, New York, USA.
Renkin, R. A., and D. G. Despain. 1992. Fuel moisture, forest
type and lightning-caused fire in Yellowstone National
Park. Canadian Journal of Forest Research 22:37–45.
Romme, W. H. 1982. Fire and landscape diversity in sub-
alpine forests of Yellowstone National Park. Ecological
Monographs 52:199–221.
Romme, W. H., and D. G. Despain. 1989. Historical per-
spective on the Yellowstone fires of 1988. BioScience 39:
695–699.
Romme, W. H., and M. G. Turner. 1991. Implications of
global climate change for biodiversity in the Greater Yel-
lowstone Ecosystem. Conservation Biology 5:373–386.
Romme, W. H., L. Bohland, C. Persichetty, and T. Caruso.
1995. Germination ecology of some common forest herbs
in Yellowstone National Park. Arctic and Alpine Research
27:407–412.
Rowe, J. S., and G. W. Scotter. 1973. Fire in the boreal forest.
Quaternary Research 3:444–464.
Ryan, K. C., and N. V. Noste. 1985. Evaluating prescribed
fires. Pages 230–238
in
Proceedings of the Symposium and
Workshop on Wilderness Fire. United States Forest Service
General Technical Report INT-182.
SAS Institute. 1992. SAS user’s guide: statistics. SAS In-
stitute, Cary, North Carolina, USA.
Solbreck, C., and D. Andersson. 1986. Vertical distribution
of fireweed,
Epilobium angustifolium,
seeds in the air. Ca-
nadian Journal of Botany 65:2177–2178.
Stickney, P. F. 1986. First decade plant succession following
the Sundance fire, northern Idaho. U.S. Department of Ag-
riculture Forest Service General Technical Report INT-
197.
Tande, G. F. 1979. Fire history and vegetation pattern of
coniferous forests in Jasper National Park, Alberta. Ca-
nadian Journal of Botany 57:1912–1931.
Tinker, D. B., W. H. Romme, W. W. Hargrove, R. H. Gardner,
and M. G. Turner. 1994. Landscape-scale heterogeneity in
lodgepole pine serotiny. Canadian Journal of Forest Re-
search 24:897–903.
Turner, M. G., editor. 1987. Landscape heterogeneity and
disturbance. Springer-Verlag, New York, New York, USA.
Turner, M. G., and W. H. Romme. 1994. Landscape dynamics
in crown fire ecosystems. Landscape Ecology 9:59–77.
Turner, M. G., Y. Wu, W. H. Romme, L. L. Wallace, and A.
Brenkert. 1994
a.
Simulating winter interactions between
ungulates, vegetation and fire in northern YellowstonePark.
Ecological Applications 4:472–496.
Turner, M. G., W. H. Hargrove, R. H. Gardner, and W. H.
Romme. 1994
b.
Effects of fire on landscape heterogeneity
in Yellowstone National Park, Wyoming. Journal of Veg-
etation Science 5:731–742.
Van Wagner, C. E. 1983. Fire behavior in northern conifer
forests and shrublands. Pages 65–80
in
R. W. Wein and D.
A. MacLean, editors. The role of fire in northern circum-
polar ecosystems. John Wiley and Sons, New York, New
York, USA.
. 1988. The historical pattern of annual burned area
in Canada. Forestry Chronicle 64:182–185.
Viereck, L. A. 1983. The effects of fire in black spruce
ecosystems of Alaska and Northern Canada. Pages 201–
220
in
R. W. Wein and D. A. MacLean, editors. The role
of fire in northern circumpolar ecosystems. John Wiley and
Sons, New York, New York, USA.
Whipple, S. A. 1978. The relationship of buried, germinating
seeds to vegetation in an old-growth Colorado subalpine
forest. Canadian Journal of Botany 56:1505–1509.
Wiens, J. A., and K. R. Parker. 1995. Analyzing the effects
of accidental environmental impacts: approaches and as-
sumptions. Ecological Applications 5:1069–1983.
Wood, D. M., and R. del Moral. 1988. Colonizing plants on
the pumice plains, Mount St. Helens, Washington. Amer-
ican Journal of Botany 75:1228–1237.
Wood, D. M., and R. Del Moral. 1993. Early primary suc-
cession on the volcano Mount St. Helens. Journal of Veg-
etation Science 4:223–234.
Wright, H. E., Jr., and M. L. Heinselman. 1973. Introduction
to symposium on the ecological role of fire in natural co-
niferous forests of western and northern America. Quater-
nary Research 3:319–328.
Zar, J. H. 1984. Biostatistical analysis. Prentice Hall, Upper
Saddle River, New Jersey, USA.