Intermediate wind disturbance in an old-growth beech-fir forest in southeastern Slovenia

Abstract

We studied the immediate effects of two successive storms in 1983 and 2004 in an old-growth Fagus sylvatica L. – Abies alba Mill. forest in the Dinaric Alps, southeastern Slovenia. In the 1983 and 2004 storms the density and basal area of wind-killed trees were 27.4·ha–1 and 5.98 m2·ha–1 and 11.2·ha–1and 4.02 m2·ha–1, respectively. In both storm events, mid-sized to large stems were more prone to wind mortality than small stems, and A. alba was more susceptible than F. sylvatica. In the 2004 storm, 70% of wind-killed trees were uprooted and 30% were snapped. Tree size (diameter at breast height) was more important than species in determining damage type, smaller stems being more prone to uprooting and larger stems to snapping. The combined mortality due to the two storms was similar to a decade of baseline mortality, and wind-killed trees were larger than trees that died from background mortality processes. In both storms, wind-killed trees had a clumped spatial distribution, which resulted in the formation of many multiple-tree-fall gaps up to 1500 m2 in size. We suggest that intermediate wind disturbance occurs at time intervals similar to or less than the life expectancy of most canopy tree species in central Europe, and may play a more important role in forest dynamics than previously thought.

Full text

Intermediate wind disturbance in an old-growth
beech–fir forest in southeastern Slovenia
Thomas A. Nagel and Jurij Diaci
Abstract: We studied the immediate effects of two successive storms in 1983 and 2004 in an old-growth Fagus
sylvatica L.–Abies alba Mill. forest in the Dinaric Alps, southeastern Slovenia. In the 1983 and 2004 storms the den-
sity and basal area of wind-killed trees were 27.4·ha–1 and 5.98 m2·ha–1 and 11.2·ha–1and 4.02 m2·ha–1, respectively. In
both storm events, mid-sized to large stems were more prone to wind mortality than small stems, and A. alba was
more susceptible than F. sylvatica. In the 2004 storm, 70% of wind-killed trees were uprooted and 30% were snapped.
Tree size (diameter at breast height) was more important than species in determining damage type, smaller stems being
more prone to uprooting and larger stems to snapping. The combined mortality due to the two storms was similar to a
decade of baseline mortality, and wind-killed trees were larger than trees that died from background mortality pro-
cesses. In both storms, wind-killed trees had a clumped spatial distribution, which resulted in the formation of many
multiple-tree-fall gaps up to 1500 m2in size. We suggest that intermediate wind disturbance occurs at time intervals
similar to or less than the life expectancy of most canopy tree species in central Europe, and may play a more impor-
tant role in forest dynamics than previously thought.
Résumé : Les auteurs ont étudié l’effet immédiat de deux tempêtes consécutives survenues en 1983 et 2004 dans une
forêt ancienne de Fagus sylvatica L. – Abies alba Mill. dans les Alpes Dinariques, au sud-est de la Slovénie. Lors des
tempêtes de 1983 et 2004, la densité et la surface terrière des arbres tués par la tempête atteignaient respectivement de
27,4·ha–1,5,98m
2·ha–1, 11,2·ha–1 et 4,02 m2·ha–1. Dans les deux cas, les arbres de dimension moyenne à supérieure
étaient plus vulnérables à la mortalité par le vent que les plus petits et A. Alba était plus vulnérable que F. Sylvatica.
Lors de la tempête de 2004, 70 % des arbres tués par le vent ont été déracinés et 30 % ont été cassés. La taille de
l’arbre (diamètre à hauteur de poitrine) avait plus d’influence que l’espèce sur le type de dommage, les petites tiges
étant plus susceptibles au déracinement et les plus grosses, à la rupture. La mortalité combinée des deux tempêtes se
comparait en importance à la mortalité décennale de base et les arbres tués par le vent étaient plus gros que ceux qui
étaient morts suite aux processus associés à la mortalité de base. Dans les deux cas, les arbres tués par le vent étaient
regroupés, ce qui a occasionné la formation de nombreuses trouées résultant de la chute de plusieurs arbres et pouvant
atteindre jusqu’à 1500 m2. Les auteurs croient que les perturbations intermédiaires causées par le vent surviennent à
des intervalles de temps semblables ou inférieurs à la longévité de la plupart des espèces d’arbres de l’Europe Centrale
et pourraient jouer un rôle dans la dynamique forestière plus important qu’on l’avait crû auparavant.
[Traduit par la Rédaction] Nagel and Diaci 638
Introduction
Wind disturbance plays a major role in the dynamics of
forest ecosystems throughout the world (Lawton and Putz 1988;
Foster and Boose 1992; Batista et al. 1998; Ida 2000; Veblen
et al. 2001). Wind creates a continuous gradient of distur-
bance severity ranging from damage to scattered individuals
to catastrophic events that blow down entire stands (Everham
and Brokaw 1996). Most studies involving wind damage,
however, have centered on the extreme ends of the spectrum,
focusing on small canopy gaps or large areas disturbed by
catastrophic winds, such as tornados and hurricanes. Less at-
tention has been focused on intermediate wind-disturbance
events that damage moderate portions of a stand, creating a fine-
grained mosaic of canopy damage in a matrix of undisturbed
forest. Disturbances of this type may create a range of can-
opy gap sizes, including “messy” mid-sized openings without
discrete edges and standing trees remaining in gap interiors
(Webb 1989; Greenberg and McNab 1998; Woods 2004).
Wind is the dominant cause of canopy disturbance in tem-
perate forest ecosystems throughout Europe. With the excep-
tion of the extraordinary storms in 1990 and 1999 that caused
unprecedented damage to forests throughout Central Europe
(Schönenberger et al. 2003; Angst et al. 2004), winds that
cause large-scale catastrophic blowdowns are infrequent in
Europe, while strong winds associated with localized thun-
derstorms are more common (Schelhaas et al. 2003). These
events may play an important role in intermediate canopy
disturbance. However, there have been few attempts to describe
the direct effects of intermediate wind disturbance on the
structure of forests, especially old-growth forests, in Europe.
In old-growth temperate forests of Central Europe, gap
formation due to the death of individual trees is often noted
as being the dominant process driving forest dynamics (Prusa
1985; Leibundgut 1987; Korpel 1995). However, intermedi-
ate wind disturbance may have more important and distinc-
Can. J. For. Res. 36: 629–638 (2006) doi:10.1139/X05-263 © 2006 NRC Canada
629
Received 1 April 2005. Accepted 9 November 2005
Published on the NRC Research Press Web site at
http://cjfr.nrc.ca on 14 March 2005.
T.A. Nagel1and J. Diaci. Department of Forestry and
Renewable Forest Resources, Biotechnical Faculty, University
of Ljubljana, Vecna Pot 83, 1000 Ljubljana, Slovenia.
1Corresponding author (e-mail: tom.nagel@bf.uni-lj.si).

tive effects on stand structure and dynamics than the more
frequent small-scale gap-disturbance processes resulting from
the death of individual canopy trees. For example, in studies
that examined the direct effects of intermediate wind distur-
bance on forest structure, larger trees were found to be more
prone to wind mortality (Webb 1989; Greenberg and McNab
1998; Canham et al. 2001; Woods 2004), while smaller,
subcanopy stems were more susceptible to background mor-
tality processes (Wolf et al. 2004; Woods 2004). Intermediate
windstorms have also been reported to cause species-specific
mortality patterns, such as greater damage among conifers
than among broad-leafed trees (Mayer and Neumann 1981;
Webb 1989; Woods 2004). In addition, the mode of mortal-
ity (i.e., uprooting or snapping) has been found to vary with
tree species and size (Greenberg and McNab 1998). Finally,
spatial patterns of canopy gap formation may differ substan-
tially between wind-related and baseline mortality processes.
For instance, several studies have shown clumped distribu-
tions of wind-killed trees and random or hyperdispered patterns
of baseline mortality (Wolf et al. 2004; Woods 2004).
Two storms with unusually intense winds in the summers
of 1983 and 2004 caused intermediate canopy damage in an
old-growth forest stand dominated by European beech, Fagus
sylvatica L., and silver fir, Abies alba Mill., in Slovenia. In
this study, we describe the immediate effects of the two dis-
turbance events on the stand. The specific objectives were
(i) to quantify the trees killed in the storms; (ii) to determine
whether species or tree size is related to susceptibility to
wind-caused mortality; (iii) to identify the type of damage
(i.e., snapping or uprooting); (iv) to compare the characteris-
tics of wind-killed trees with those of trees dying as a result
of baseline mortality processes; and (v) to describe the spa-
tial pattern of wind-killed trees.
Materials and methods
Study area
We conducted our study in the Pecka Forest Reserve, a
60 ha old-growth F. sylvatica –A. alba forest remnant located
on a high karst plateau (900 m) in the Dinaric Alps, south-
eastern Slovenia (45°45′N, 14°59′E) (Fig. 1). The climate is
a combination of continental and Mediterranean, with mean
annual precipitation ca. 1400 mm. The mean annual temper-
ature is 8.3 °C and monthly means range from –4 °C in Jan-
uary to 20.3 °C in July. The microtopography of the site is
very diverse, with abundant sinkholes typical of karst geol-
ogy. Calcareous brown soils on the site are derived from the
© 2006 NRC Canada
630 Can. J. For. Res. Vol. 36, 2006
Fig. 1. Map of the Pecka Forest Reserve, showing windthrown trees resulting from the 1983 and 2004 storm events.

limestone parent material, and soil depth varies between 30
and 70 cm depending on microtopographic position.
The forest is dominated by F. sylvatica (81%) and A. alba
(19%), but other species, including Norway spruce (Picea
abies (L.) Karst.), sycamore maple (Acer pseudoplatanus L.),
and Wych elm (Ulmus glabra Huds.), occur infrequently.
Long-term inventory studies in the reserve indicate that
A. alba declined substantially over the second half of the
20th century (Turk et al. 1985; Rozenbergar 2000), a trend
reported throughout F. sylvatica – A. alba forests in Slovenia
and Central Europe (Bigler et al. 2004). The reasons for the
decline are unknown, but several anthropogenic factors, such
as climate change, pollution, and soil acidification, may be
partly responsible. The reserve has been anthropogenically
influenced in other ways as well. Long-term changes in deer
populations regulated by hunting have strongly influenced
A. alba demography through browsing pressure on regenera-
tion (Debeljak 1997; Diaci and Boncina 2003). Also, a few
scattered cut A. alba stumps in parts of the reserve indicate
that some selective cutting occurred in the past. Although
people have influenced their structure and composition, old-
growth F. sylvatica – A. alba forests in the Dinaric Alps of
Slovenia, Croatia, and Bosnia and Herzegovina are consid-
ered to be some of the largest and best preserved in Europe
(Mlinsek 1969; Leibundgut 1987; Diaci 1999).
Field methods
Within several weeks of both storms, all storm-killed trees
≥10 cm diameter at breast height (DBH) were located and
mapped to the nearest 1m. Trees damaged in the storms were
easily distinguished from older uprooted or snapped trees based
on the freshness of the root masses and leaves. Only trees that
were fully uprooted or snapped below the crown were re-
corded. Trees that sustained less damage, such as partial up-
rooting or damage to large portions of the crown, were not
included in the analysis. Thus, our analysis only takes into ac-
count trees that were killed in the storms, thereby underesti-
mating the total damage due to the storms, as many damaged
trees likely die within a few years (Everham and Brokaw 1996).
The locations of trees that were uprooted and thrown from
their original position were recorded in the middle of the
root pit, while the locations of snapped trees were recorded
at the stump position. For each tree, we recorded the species,
trunk diameter at 1.3 m (DBH), and direction of fall, and
noted the type of damage as either snapped or uprooted
(damage type was only recorded for the 2004 storm). For the
purposes of discussion, we also conservatively described the
prestorm status of trees killed in the 2004 storm, based on
the presence of bark fungi or heart rot. No attempt was made
to distinguish between mortality due to damage resulting
from the direct effects of wind and indirect mortality due to
falling trees, as identification of indirectly damaged trees
was often ambiguous.
Background sampling
We used data from two diameter inventories conducted in
the reserve in 1980 and 2004 to provide a basis for compar-
ing windthrown trees from each storm with overall tree pop-
ulations. Although the 1980 inventory was done 3 years
before the first storm, it comprises the available data nearest
to the 1983 event, and should provide a close approximation
to stand conditions in 1983. No major wind or other distur-
bance events were recorded in the reserve between 1980 and
1983. In the 1980 inventory, trees were sampled in 65 half-
hectare plots, where all trees ≥10 cm DBH were identified
and measured. For the 2004 inventory, all trees ≥10 cm
DBH in the entire reserve were identified and measured.
Baseline mortality data were gathered from four permanent
research plots in the reserve (2.9 ha in total); mortality and
characteristics of dead trees were scored between 1981 and
1998. Only trees ≥10 cm DBH that died standing were in-
cluded in the analysis. None of the permanent plots were af-
fected by the 1983 storm event.
Data analyses
For calculating basic density and basal area, the locations
of all wind-killed trees were transferred to a topographic
map and the approximate area affected by each storm was
calculated using a 0.25 ha grid. Differences between the
mean diameter of all trees and that of trees that died from ei-
ther wind or baseline mortality were compared using ttests.
We used Kolmogorov–Smirnov tests to compare the size dis-
tributions of windthrown trees with the overall tree popula-
tions and with baseline mortality for all species pooled and
within species. Gtests (Sokal and Rohlf 2003) were used to
examine whether trees were windthrown in proportion to
their abundance with respect to tree size for all trees pooled
and within species. Similarly, Gtests were used to test for
significant departures from expected ratios of snapping and
uprooting within species and within size classes. To simulta-
neously examine the influence of species and size (DBH) on
the risk of windthrow mortality, stepwise logistic regression
analyses were used (Grizzle et al. 1969).
Ripley’s Kunivariate analyses (Ripley 1981) were used to
describe the spatial patterns of the windthrown trees resulting
from both storms. Ripley’s Kfunction determines whether
the empirical distribution of individuals, based on all point-
to-point distances, differs significantly from the Poisson dis-
tribution (Ripley 1981; Diggle 1983). Ripley’s K(d) function
was computed using edge-correction as described by Diggle
(1983) and then transformed to the function L(d), which is a
square-root transformation of K(d) that has an expected value
of zero over all distances if individuals are randomly distrib-
uted. Significant deviations of the empirical tree distribution
from a random distribution were tested with 95% confidence
limits computed from 100 randomizations. The resulting pat-
tern is either random, clumped, or regularly dispersed at any
distance up to half the length of the shortest plot dimension.
We examined the spatial patterns of wind-killed trees in the
reserve at three spatial scales, including the entire area of
wind damage (600 m × 900 m), a mid-sized area with a
patch of wind-killed trees (400 m × 400 m), and small plots
within the patch (100 m × 200 m). Because the analysis re-
quires rectangular plots, we fit a large plot within the forest
reserve border for the analysis of the entire windthrow for
each storm, which excluded six windthrown trees near the
edge of the reserve for the 1983 windthrow.
Results
In both storms the prevailing wind was from the north-
west; this is clearly seen in the uniformity of the direction in
© 2006 NRC Canada
Nagel and Diaci 631

which trees fell (Fig. 1). The 1983 storm affected approxi-
mately 12 ha and killed a total of 322 trees, while the 2004
storm damaged roughly 6 ha and killed 70 trees (Table 1).
The density of wind-killed trees in the 1983 and 2004 storms
was 27.4 and 11.2 trees·ha–1, respectively. Both storms dra-
matically reduced the total basal area within the disturbed
parts of the reserve. In the 1983 storm the total basal area of
all storm-killed trees (>10 cm DBH) was 5.98 m2·ha–1,or
about 13% of the total 1980 basal area, of which 3.04 m2·ha–1
was A. alba and 2.94 m2·ha–1 was F. sylvatica. In the 2004
storm the total basal area of storm-killed trees was 4.02 m2·ha–1,
or about 11% of the total 2004 basal area, of which 1.97 m2·ha–1
was A. alba and 2.05 m2·ha–1 was F. sylvatica.
The size structure of the trees killed in both storms differed
from that of the overall tree populations (Fig. 2). In 1983 the
size distribution of the wind-killed trees was unimodal and
differed significantly from the overall tree population in 1980
(Kolmogorov–Smirnov test, p< 0.001). The mean DBH of
wind-killed trees was 49.3 cm, significantly larger than the
39.4 cm DBH of trees in the 1980 population (ttest, t=
6.10, p< 0.001). Similarly, the size distribution of storm-
killed trees in both the F. sylvatica and the A. alba population
differed significantly from that of the overall populations of
each species (Kolmogorov–Smirnov tests, p< 0.01 for both
species), both species showing more storm-related mortality
in larger size classes. The size distribution of storm-killed
trees in 2004 showed similar patterns. Again, over all spe-
cies, larger trees were more prone to wind mortality than
small stems, and the size distribution of the wind-killed trees
differed significantly from that of the overall tree population
in 2004 (Kolmogorov–Smirnov test, p< 0.001). The size
distributions of F. sylvatica and A. alba also differed signifi-
cantly from the overall distributions (Kolmogorov–Smirnov
tests, p< 0.001 for both species), but the mean DBH of
storm-killed F. sylvatica trees (60.1 cm) was larger in the
2004 storm than in the 1983 storm (43.7 cm). Overall, the
mean DBH of wind-killed trees (63.3 cm) was significantly
larger than that of trees in the overall population (40.0 cm) (t
test, t= 7.58, p< 0.001).
In both storms, Gtests show there was a significant depar-
ture from randomness for all wind-killed trees of both spe-
cies with respect to tree size (p< 0.001). For F. sylvatica
killed in the 1983 storm, small size classes (<30 cm DBH)
of wind-killed trees were underrepresented (21%) compared
with small trees in the overall population (28%), while inter-
mediate size classes (30–60 cm DBH) of wind-killed trees
were overrepresented (66%) compared with trees in the over-
all population (59%). Large size classes (>60 cm DBH) were
killed in proportion to their number in the overall popula-
tion. Wind-killed F. sylvatica showed different patterns in
the 2004 storm. Small size classes were underrepresented
(3%) in the wind-killed population compared with the over-
all population (33%), while the percentage of mid-sized trees
was similar (49%), and large size classes of wind-killed
trees were overrepresented (50% of wind-killed trees, but
17% of trees in the overall population). For A. alba, the pat-
terns were similar for both storms, with small size classes
(>30 cm DBH) underrepresented in the wind-killed popula-
tion in 1983 (11%) and 2004 (14%) compared with their
percentage in the overall population (43% and 56%, respec-
tively). Both the intermediate (30–60 cm DBH) and large
(>60 cm DBH) size classes were overrepresented in the
wind-killed population in 1983 (42% and 47%, respectively)
and 2004 (32% and 54%, respectively) compared with their
percentage in the overall population in 1980 (23% and 34%,
respectively) and 2004 (22% and 22%, respectively).
Wind-killed trees were also compared with trees in the
overall population using logistic regression to examine the
influence of both size and species on susceptibility to wind
mortality. In the 1983 storm, size was the only factor that
was significant in the model (G= 30.3, p< 0.001), with
higher wind-caused mortality of larger trees. However, both
size and species had significant predictive value in the
model for the 2004 storm, but size (G= 33.8, p< 0.001)
had a stronger effect than species (G= 8.9, p< 0.01). Again,
larger trees sustained higher mortality than smaller trees, and
mortality was higher in A. alba than in F. sylvatica.
Of the 70 trees that were killed in the 2004 storm, 49
were uprooted and 21 snapped (Table 2). The mean DBH of
snapped trees (74.7 cm) was significantly larger than the
mean DBH of uprooted trees (58.5 cm) (ttest, t= 2.03, p<
0.05). Furthermore, Gtests show a significant departure from
randomness regarding damage type (snap or uproot) with re-
spect to tree size (p< 0.05). Trees <50 cm DBH were uprooted
more often than they snapped, trees between 50 and 70 cm
DBH snapped and were uprooted in similar relative propor-
© 2006 NRC Canada
632 Can. J. For. Res. Vol. 36, 2006
Overall populationaWind-killed populationb
Storm Species
Density
(no.·ha–1)c
Basal area
(m2·ha–1)c
Total no.
of treesc
Density
(no.·ha–1)
Percentage of
overall density
Basal area
(m2·ha–1)c
Percentage of
overall basal area
1983 Fagus sylvatica 191.7 (67) 25.3 (56) 213 (66) 18.1 9.5 2.94 (49) 11.6
Abies alba 95.2 (33) 19.9 (44) 109 (34) 9.3 9.8 3.04 (51) 15.3
Total 286.9 (100) 45.2 (100) 322 (100) 27.4 9.6 5.98 (100) 13.2
2004 Fagus sylvatica 189.5 (81) 30.4 (81) 42 (60) 6.7 3.5 2.05 (51) 6.7
Abies alba 44.1 (19) 7.3 (19) 28 (40) 4.5 10.2 1.97 (49) 26.9
Total 233.6 (100) 37.7 (100) 70 (100) 11.2 4.8 4.02 (100) 10.7
aCalculations of density and basal area are based on inventory data collected in the reserve in 1980 and 2004.
bCalculations of density and basal area for the populations of wind-killed trees are based on the area of land affected by each storm (11.75 ha in 1983
and 6.25 ha in 2004).
cValues in parentheses are percentages.
Table 1. Densities and basal areas of trees (≥10 cm DBH) in the overall and wind-killed populations for two storm events in the
Pecka Forest Reserve.

tions, and trees >70 cm DBH snapped more often than they
were uprooted. However, the relationship between damage
type (snapping and uprooting) and tree species was weak.
The proportion of snapped and uprooted A. alba and
F. sylvatica did not differ significantly from the expected ra-
tio based on the total proportion of snapped and uprooted
trees (Gtest, p= 0.064).
Baseline mortality of trees (≥10 cm DBH) recorded in
permanent plots between 1981 and 1998 amounted to a density
of 4.5 ha–1·year–1, or a basal area of 0.9 m2·ha–1·year–1, much
less than the mortality resulting from each storm event. The
size distributions of A. alba and F. sylvatica that sustained
baseline and wind-caused mortality were also significantly
different (Kolmogorov–Smirnov tests, p< 0.001 for both
species) (Fig. 3). Fagus sylvatica that died from baseline
mortality processes had an approximately inverse J-shaped
size distribution, while the distribution of A. alba baseline
mortality was approximately bimodal, most mortality occur-
ring in trees <20 cm and >50 cm DBH. The mean DBH of
wind-killed F. sylvatica from both storms was 46.4 cm, sig-
nificantly larger than the 30.3 cm DBH for baseline mortal-
ity (ttest, t= 8.42, p< 0.001). Likewise, wind-killed A. alba
averaged 61.7 cm DBH, significantly larger than the 53.2 cm
DBH for baseline mortality (ttest, t= 2.3, p< 0.05).
The spatial pattern of wind-killed trees in both storms
showed clumping at the three scales tested (Fig. 4). At both
© 2006 NRC Canada
Nagel and Diaci 633
Uprooted trees Snapped trees
No. DBH (cm) No. DBH (cm)
Total no.
of trees
Fagus sylvatica 26 54.4±14.5 16 69.4±16.3 42
Abies alba 23 63.1±27.0 5 91.4±41.8 28
Total 49 58.5±21.5 21 74.7±25.3 70
Table 2. Total numbers and DBH (mean ± SD) of uprooted and
snapped trees resulting from the 2004 storm in the Pecka Forest
Reserve.
1983 windthrow,
all trees
0
1
2
3
4
5
6
7
15 25 35 45 55 65 75 85 95 105115125135
2004 windthrow,
all trees
0
1
2
3
15 25 35 45 55 65 75 85 95 105115 125135
1980 population,
all trees
0
10
20
30
40
50
60
70
15 25 35 45 55 65 75 85 95 105115125135
2004 population,
all trees
0
10
20
30
40
50
60
70
15 25 35 45 55 65 75 85 95 105115125135
1983 windthrow
0
1
2
3
4
5
6
15 25 35 45 55 65 75 85 95 105115125135
Fagus sylvatica
Abies alba
2004 windthrow
0
1
2
15 25 35 45 55 65 75 85 95 105115 125135
Fagus sylvatica
Abies alba
1980 population
0
5
10
15
20
25
30
35
40
45
15 25 35 45 55 65 75 85 95 105115125135
DBH class midpoint (cm)
trees·ha
–1
Fagus sylvatica
Abies alba
2004 population
0
5
10
15
20
25
30
35
40
45
15 25 35 45 55 65 75 85 95 105 115 125135
DBH class midpoint (cm)
Fagus sylvatica
Abies alba
trees·ha
–1
trees·ha
–1
trees·ha
–1
Fig. 2. Size-class distribution of all trees, by species, for the overall and wind-killed populations for both storm events in the Pecka
Forest Reserve.

the large (600 m × 900 m) and intermediate scales (400 m ×
400 m), mortality was significantly clumped at all spatial
scales tested (up to half the length of the shortest plot di-
mension). The results of Ripley’s Kanalyses at the smallest
scale (100 m × 200 m), however, showed a finer scale pat-
tern. For the three 2 ha plots tested in the 1983 storm, wind-
killed trees were significantly clumped at scales of about 2–
20 m and randomly dispersed at all other scales. In the 2004
storm, wind-caused mortality was significantly clumped be-
tween 13 and 22 m, but only one 2 ha plot had a density of
wind-killed trees (>20) sufficient for Ripley’s analysis.
Discussion
Strong winds associated with the two summer storms in
1983 and 2004 caused widespread damage across the Pecka
Forest Reserve and created many “messy”, multiple-tree-fall
gaps with scattered wind-firm trees in the gap interiors. The
result was a diverse mosaic of wind damage within an undis-
turbed forest, with discrete canopy openings that were were
often interconnected and difficult to measure (personal ob-
servation). The storms reduced total tree density and basal
area by 5%–10% and 11%–13%, respectively, while the loss
of basal area for individual species, such as A. alba in the
2004 storm, was as high as 27%. These damage levels are
comparable to the ranges reported in other studies of inter-
mediate wind disturbance in temperate forests (Webb 1989;
Woods 2004).
Many studies have reported that tree size is an important
factor that influences susceptibility to wind damage (Everham
and Brokaw 1996). In this study, larger trees had a higher
risk of wind damage. However, damage patterns from the
two storms were different, with mid-sized trees (30–60 cm
DBH) showing the most damage in the 1983 storm and large
trees (>60 cm DBH) in the 2004 storm. Everham and Brokaw
(1996) proposed a relationship between stem size and wind
damage, where the most damage occurs in intermediate size
classes, as small trees are protected from wind and the larg-
est trees are preconditioned and able to withstand high
winds. The unimodal distribution of damage to intermediate
size classes in the 1983 storm follows this pattern, while the
higher mortality in large size classes in the 2004 storm may
have several explanations. First, since both storms occurred
in similar parts of the stand, one possibility is that the re-
moval of mid-sized trees after the first storm formed gaps
that left larger trees more vulnerable in the second storm.
We observed that in the second storm, many of the wind-
killed trees were located at or near the edge of gaps formed
in the 1983 event. This suggests that gap-expansion pro-
cesses may have played an important role in the patterns of
mortality caused by the second storm event (Foster and Reiners
1986; Worrall et al. 2005). As a consequence, open parts of
the stand damaged in these two previous events are more
vulnerable to further wind disturbance and may continue to
break down. Second, it is also likely that many of the larger
A. alba killed by wind in the second storm were vulnerable
because of the severe decline of the A. alba population in
the Pecka Forest Reserve (Turk et al. 1985; Rozenbergar
2000). Bigler et al. (2004) found that dead and declining
A. alba in Slovenia had very low growth levels compared
with living trees, and declining trees often had an unhealthy
crown and nonproductive cambium. Many of the large
A. alba killed in the 2004 storm showed signs of decline: ei-
ther unhealthy crowns, heart rot, or very suppressed growth
in the previous several decades (personal observations).
The greater damage to A. alba in the 2004 storm is in
agreement with the trend for conifers to be more vulnerable
than hardwoods because of either weaker wood (Webb 1989;
Foster and Boose 1992) or taller trees (Mayer and Neumann
1981). Abies alba sustained 40% of the total mortality in the
2004 storm, but made up only 19% of the overall tree popu-
lation, while the proportion of wind-killed A. alba and
F. sylvatica did not differ from the overall population in the
1983 storm. It is possible that the decline of the A. alba pop-
ulation may have also contributed to the 2004 storm mortal-
ity, but this does not explain why the same pattern was not
found in the 1983 event. If large A. alba are more vulnerable
than F. sylvatica, as this study suggests, the decline of the
A. alba population may be enhanced by wind-driven mortality
processes, accelerating the change to a F. sylvatica-dominated
forest on this site.
One of the most robust trends found in this study was that
uprooting was more frequent than snapping among the wind-
thrown trees in the 2004 storm, which is consistent with the
results of other intermediate windthrow studies (Pontailler et
al. 1997; Greenberg and McNab 1998; Ida 2000). Shallow
soils on the study site probably contributed to the large num-
ber of uprooted trees. Rooting depth (Mueller and Cline
1959), tree size and species (Peterson and Pickett 1991;
Greenberg and McNab 1998), and wood strength (Putz et al.
1983) have been noted to influence the mode of windthrow.
In this study, we found a significant relationship between
© 2006 NRC Canada
634 Can. J. For. Res. Vol. 36, 2006
windthrow mortality: 1983 and 2004 storms
0
1
2
3
4
5
6
7
15
25
35
45
55
65
75
85
95
105
115
125
135
Fagus sylvatica
Abies alba
baseline mortality: 1981-1998
0
2
4
6
8
10
12
14
16
15
25
35
45
55
65
75
85
95
105
115
125
135
DBH class midpoint (cm)
trees·ha
–1
trees·ha
–1
Fagus sylvatica
Abies alba
Fig. 3. Size-class distribution of total storm-caused mortality and
baseline mortality recorded in permanent plots between 1981 and
1998 in the Pecka Forest Reserve.

© 2006 NRC Canada
Nagel and Diaci 635
1983, all trees (n=316)
-10
0
10
20
30
40
50
60
0 60 120 180 240 300
L(d)
L(t)
Lower CI
Upper CI
1983, intermediate (n=188)
-100
-50
0
50
100
150
200
250
0 30 60 90 120 150 180
L(d)
1983, small (n=55)
-40
-30
-20
-10
0
10
20
30
40
50
0 1020304050
L(d)
1983, small (n=43)
-6
-4
-2
0
2
4
6
8
0 1020304050
L(d)
1983, small (n=42)
-6
-4
-2
0
2
4
6
8
0 1020304050
distance (m)
L(d)
2004, all trees (n=68)
-40
-20
0
20
40
60
80
100
0 60 120 180 240 300
2004, intermediate (n=52)
-150
-100
-50
0
50
100
150
200
250
300
350
400
0 30 60 90 120 150 180
2004, small (n=22)
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 1020304050
distance (m)
Fig. 4. Results of Ripley’s Kanalyses of wind-killed trees in the Pecka Forest Reserve. Solid lines show L(d) values calculated for all
distances analyzed and broken lines indicate 95% confidence limits (CI) for random distribution. The analyses were performed at three
scales for each storm event, including a large plot with all wind-killed trees, a mid-sized plot with a patch of wind-killed trees, and
small plots wihin the patch. In the 2004 storm, only one small plot had a density of wind-killed trees (>20) sufficient for Ripley’s K
analysis. Values above the confidence limit indicate clustering, whereas values below the confidence limit indicate hyperdispersion.

tree size and mode of windthrow, but no clear pattern with
regard to species. Overall, small to mid-sized trees were
more likely to uproot, while large trees were more likely to
snap, which is similar to trends documented in several other
studies carried out in temperate forests (Webb 1988; Peterson
and Pickett 1991; Pontailler et al. 1997; Ida 2000). One fac-
tor that may lead to greater snapping in larger trees is lower
tree vitality. Twenty-four percent of all snapped trees had
rotten centers, while all the uprooted trees appeared healthy.
It is likely that some of the uprooted trees also had rotten
centers that were not noticeable from the root plate or bark.
Nevertheless, these observations suggest that factors that
weaken the structural properties of a tree contribute to wind-
throw susceptibility (Webb 1989; Canham et al. 2001).
The combined mortality due to the two storm events was
about 9 times the annual baseline mortality recorded in the
permanent plots between 1981 and 1998. Baseline mortality
of both A. alba and F. sylvatica occurred mostly in the
smallest size classes, between 10 and 30 cm DBH, while
storm-caused mortality for both species occurred in the in-
termediate to large size classes. Many large A. alba in the
permanent plots died standing, but this was probably related
to the declining population in the reserve. Overall, the pat-
terns suggest that smaller, subcanopy trees are protected from
wind-caused mortality, while many of the smaller trees that
died as a result of baseline mortality processes died standing,
which suggests that they died from endogenous processes,
such as competition for light or other resources. Woods
(2004) observed similar patterns after an intermediate wind
disturbance in a Michigan mixed temperate forest, where a
single storm caused mortality equal to between one and
three decades of baseline mortality. Similarly, in mixed de-
ciduous forests in Denmark, Wolf et al. (2004) found that
40% of the mortality that occurred during a 10-year observa-
tion period resulted from a single storm that killed mostly
mid-sized and large trees, whereas trees in small size classes
that were killed were mostly standing dead.
One of the more interesting findings is that wind-killed
trees showed different patterns of clumping at the three scales
tested. The larger scale clusters of storm mortality may have
resulted from small-scale heterogeneity in wind gusts and
turbulence during the storms, which interacted with the to-
pography of the site to produce local patches of dense mor-
tality. In both storms, a large number of stems were damaged
on the highest part of the reserve, where trees were more ex-
posed to wind. The finer scale clumping found in the small
plots reflects the large number of multiple-tree-fall gaps,
many of which formed as a result of indirect damage due to
one tree falling on another. The resulting mortality patterns
created a broad range of gap sizes (100–1500 m2) (T.A.
Nagel and J. Diaci, unpublished data). In most of the larger
openings, however, canopy trees were left in the interior, un-
like the discrete canopy openings formed by the death of
single trees. Such small-scale, clumped patterns of wind-
killed trees have been documented in numerous other studies
in temperate forests (Ida 2000; McDonald et al. 2003; Wolf
et al. 2004; Woods 2004).
While wind-killed trees generally show a clumped spatial
distribution, the pattern of baseline mortality is often random
(Wolf et al. 2004) or hyperdispersed (Woods 2004). In a 1 ha
permanent plot in the Pecka Forest Reserve, the spatial pattern
of standing dead trees was random (T.A. Nagel, unpublished
data). Furthermore, because most of the baseline mortality
comprised single standing-dead trees in the permanent plots,
the resulting canopy gaps were small. For example, in the
Strmec Forest Reserve, a smaller old-growth F. sylvatica –
A. alba remnant near our study site, 64% of gaps were be-
tween 25 and 50 m2and most gaps were <100 m2(Konecnik
and Zaplotnik 2001).
These small gaps formed as a result of background mor-
tality processes occur rather frequently compared with the
storm events documented here. On the other hand, cata-
strophic wind disturbances occur very rarely, so intermediate
wind damage from storms may have a major influence on
forest dynamics, especially if these events occur at intervals
less than tree longevity (about 200–300 years at our site).
The interval between the two storms at our site was only
21 years. In the past decade, weather stations in the region
near the study site have recorded approximately seven storms
with wind speeds above 20 m·s–1 each year, including storms
with gusts greater than 40 m·s–1 (T.A. Nagel, unpublished
data). In the storm on 23 July 2004, the maximum wind
speed recorded at the closest weather station (Novo Mesto)
was 24.4 m·s–1, but the study location is approximately 700 m
higher and 14 km from the weather station, so it is impossible
to speculate about wind speeds at the study site. In temper-
ate forests in North America, storms that caused intermedi-
ate forest damage produced wind speeds between 25 and
35 m·s–1 (Webb 1989; Greenberg and McNab 1998; Canham
et al. 2001). This suggests that storms with winds strong
enough to cause intermediate damage occur rather frequently
in the region around the study site.
Surprisingly few studies have documented the immediate
effects of intermediate wind disturbance in European old-
growth forests, probably because of the scarcity of old-growth
conditions. In many of these forests, especially those domi-
nated by F. sylvatica – A. alba in Central Europe, the distur-
bance regime that drives forest dynamics is typically
characterized by small-scale gaps formed by the death of in-
dividual trees (Mlinsek 1967; Leibundgut 1987). However,
recent evidence suggests that infrequent intermediate distur-
bances occur in these forests, and play an important role in
forest dynamics (Splechtna and Gratzer 2005). In this study,
we show that wind disturbance that is intermediate in sever-
ity interacts with tree species and size to create unique mor-
tality patterns that differ both qualitatively and quantitatively
from catastrophic wind disturbance and background mortal-
ity processes. Moreover, the creation of many mid-sized gaps
in the canopy may generate long-term structural features,
such as small patches of even-aged trees. The larger canopy
openings created from multiple-tree-fall gaps may also play
an important role in maintaining shade-intolerant species such
as A. pseudoplatanus (Marinsek and Diaci 2004). Forest
management that tries to mimic natural forest dynamics and
regeneration would benefit from a more thorough under-
standing of the immediate and long-term effects of interme-
diate wind disturbance.
Acknowledgements
This study was partly supported by a Fulbright grant. We
thank M. Svoboda for assistance in the field, as well as all
© 2006 NRC Canada
636 Can. J. For. Res. Vol. 36, 2006

the researchers who collected inventory data in the Pecka
Forest Reserve. We are especially grateful to D. Mlinsek for
starting long-term research and establishing permanent plots
in old-growth forests of Slovenia. For help with inventory
and permanent-plot data we thank D. Rozenbergar. For help
with statistics we are grateful to K. Jerina. Two anonymous
reviewers provided comments that were helpful in revising
the manuscript.
References
Angst, C., Bürgi, A., Duelli, P., Egli, S., Heiniger, U., Hindenlang, K.
et al. 2004. Waldentwicklung nach Windwurf in tieferen Lagen
der Schweiz 2003–2004: Schlussbericht eines Projektes im Rahmen
des Programms « Lothar Evaluations- und Grundlagenprojekte ».
Eidgenössische Forschungsanstalt für Wald, Schnee und
Landschaft, Birmensdorf, Switzerland.
Batista, W.B., Platt, W.J., and Macchiavelli, R.E. 1998. Demogra-
phy of a shade-tolerant tree (Fagus grandifolia) in a hurricane-
disturbed forest. Ecology, 79: 38–53.
Bigler, C., Gricar, J., Bugmann, H., and Cufar, K. 2004. Growth
patterns as indicators of impending tree death in silver fir. For.
Ecol. Manage. 199: 183–190.
Canham, C.D., Papaik, M.J., and Latty, E.F. 2001. Interspecific
variation in susceptibility to windthrow as a function of tree size
and storm severity for northern temperate tree species. Can. J.
For. Res. 31: 1–10.
Debeljak, M. 1997. Jelka (Abies alba Mill.) v pomladku pragozda
Pecka v zadnjih tridesetih letih [Silver fir in the new growth of
the Pecka virgin forest in the last thirty years.] Zb. Gozd. Lesar.
53: 29–48.
Diaci, J. (Editor). 1999. Virgin forests and forest reserves in Cen-
tral and East European countries. Department of Forestry and
Renewable Forest Resources, Biotechical Faculty, University of
Ljubljana, Ljubljana, Slovenia.
Diaci, J., and Boncina, A. 2003. Aktuelle Forschungsergebnisse
zur natürlichen Verjüngung in Bergwäldern Sloweniens. FBVA
Ber. 130: 57–65.
Diggle, P.J. 1983. Statistical analysis of spatial point patterns. Aca-
demic Press, London.
Everham, E.M., and Brokaw, N.V.L. 1996. Forest damage and re-
covery from catastrophic wind. Bot. Rev. 62: 113–185.
Foster, D.R., and Boose, E.R. 1992. Patterns of forest damage re-
sulting from catastrophic wind in central New England, USA. J.
Ecol. 80: 79–98.
Foster, J.R., and Reiners, W.A. 1986. Size distribution and expan-
sion of canopy gaps in a northern Appalachian spruce–fir forest.
Vegetatio, 68: 109–114.
Greenberg, C.H., and McNab, W.H. 1998. Forest disturbance in
hurricane-related downbursts in the Appalachian Mountains of
North Carolina. For. Ecol. Manage. 104: 179–191.
Grizzle, J.E., Starmer, C.F., and Koch, G.G. 1969. Analysis of cat-
egorical data by linear models. Biometrics, 25: 489–504.
Ida, H. 2000. Treefall gap disturbance in an old-growth beech for-
est in southwestern Japan by a catastrophic typhoon. J. Veg. Sci.
11: 825–832.
Konecnik, K., and Zaplotnik, V. 2001. The virgin forest reserve
strmec: research of natural forest structure and comparison of
selected methods. University of Ljubljana, Ljubljana, Slovenia.
Korpel, S. 1995. Die Urwälder der Westkarpaten. Gustav Fischer
Verlag, Stuttgart and Jena, Germany.
Lawton, R.O., and Putz, F.E. 1988. Natural disturbance and gap-
phase regeneration in a wind- exposed tropical cloud forest.
Ecology, 69: 764–777.
Leibundgut, H. 1987. Europaische Urwälder der Bergstufe. Verlag
Paul Haupt, Bern, Switzerland.
Marinsek, A., and Diaci, J. 2004. Razvoj inicialne faze na
vetrolomni povrsini v pragozdnem ostanku Ravna gora. [Devel-
opment of the initial phase after windthrow in the virgin forest
remnant Ravna gora.] Zb. Gozd. Lesar. 73: 31–50.
Mayer, H., and Neumann, M. 1981. Struktureller und entwicklung-
sdynamischer Vergleich der Fichten-Tannen-Buchen-Urwälder
Rothwald-Niederösterreich und Corkova Uvala-Kroatien.
Forstwiss. Centralbl. (Hamb.), 100: 111–132.
McDonald, R.I., Peet, R.K., and Urban, D.L. 2003. Spatial pattern
of Quercus regeneration limitation and Acer rubrum invasion in
a Piedmont forest. J. Veg. Sci. 14: 441–450.
Mlinsek, D. 1967. Rast in sposobnost reagiranja pragozdne bukve.
Zb. Bioteh. Fak. Univ. Ljubl. 15: 63–79.
Mlinsek, D. 1969. Waldschadenuntersuchungen am Stammkern
von erwachsenen Tannen im dinarischen Tannen-Buchen-Wald.
Forstwiss. Centralbl. (Hamb.), 88: 191 – 199.
Mueller, O.P., and Cline, M.G. 1959. Effects of mechanical soil
barriers and soil wetness on rooting of trees and soil mixing by
blow-down in central New York. Soil Sci. 88: 107–111.
Peterson, C.J., and Pickett, S.T.A. 1991. Treefall and resprouting
following catastrophic windthrow in an old-growth hemlock
hardwoods forest. For. Ecol. Manage. 42: 205–217.
Pontailler, J.Y., Faille, A., and Lemee, G. 1997. Storms drive
successional dynamics in natural forests: a case study in
Fontainebleau Forest (France). For. Ecol. Manage. 98: 1–15.
Prusa, E. 1985. Die böhmischen und mährischen Urwälder — ihre
Struktur und Ökologie. Verlag der Tschechoslowakischen
Akademie der Wissenschaften, Prague, Czechoslovakia.
Putz, F.E., Coley, P.D., Lu, K., Montalvo, A., and Aiello, A. 1983.
Uprooting and snapping of trees — structural determinants and
ecological consequences. Can. J. For. Res. 13: 1011–1020.
Ripley, B.D. 1981. Spatial statistics. John Wiley & Sons, New
York.
Rozenbergar, D. 2000. Razvojne znacilnosti sestojev v pragozdnih
ostankih Pecka in Rajhenavski Rog. [Stand dynamics of the vir-
gin forest remnants of Pecka and Rajhenavski Rog.] Gozd. Vestn.
58: 53–55.
Schelhaas, M.J., Nabuurs, G.J., and Schuck, A. 2003. Natural dis-
turbances in the European forests in the 19th and 20th centuries.
Global Change Biol. 9: 1620–1633.
Schönenberger, W., Angst, C., Brundl, M., Dobbertin, M., Duelli,
P., Egli, S. et al. 2003. Vivians Erbe-Waldentwicklung nach
Windwurf im Gebirge. Merkbl. Prax. 36: 1–12.
Sokal, R.R., and Rohlf, F.J. 2003. Biometry. 3rd ed. W. H. Free-
man and Co., New York.
Splechtna, B.E., and Gratzer, G. 2005. Natural disturbances in
Central European forests: approaches and preliminary results
from Rothwald, Austria. For. Snow Landsc. Res. 79: 57–67.
Turk, V., Kastelic, A., and Hartman, T. 1985. Gozdni rezervati
Slovenije: Pragozd Pecka. Biotechnical Faculty, University of
Ljubljana, Ljubljana, Slovenia.
Veblen, T.T., Kulakowski, D., Eisenhart, K.S., and Baker, W.L.
2001. Subalpine forest damage from a severe windstorm in north-
ern Colorado. Can. J. For. Res. 31: 2089–2097.
Webb, S.L. 1988. Windstorm damage and microsite colonization in
two Minnesota forests. Can. J. For. Res. 18: 1186–1195.
Webb, S.L. 1989. Contrasting windstorm consequences in 2 for-
ests, Itasca State Park, Minnesota. Ecology, 70: 1167–1180.
Wolf, A., Moller, P.F., Bradshaw, R.H.W., and Bigler, J. 2004.
Storm damage and long-term mortality in a semi-natural, tem-
perate deciduous forest. For. Ecol. Manage. 188: 197–210.
© 2006 NRC Canada
Nagel and Diaci 637

Woods, K.D. 2004. Intermediate disturbance in a late-successional
hemlock – northern hardwood forest. J. Ecol. 92: 464–476.
Worrall, J.J., Lee, T.D., and Harrington, T.C. 2005. Forest dynam-
ics and agents that initiate and expand canopy gaps in Picea–
Abies forests of Crawford Notch, New Hampshire, USA. J. Ecol.
93: 178–190.
© 2006 NRC Canada
638 Can. J. For. Res. Vol. 36, 2006