Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Forest Ecology and Management 161 (2002) 109±122

Permeability of ancient forest edges for weedy


plant species invasion
Olivier Honnay*, Kris Verheyen, Martin Hermy
Laboratory for Forest, Nature and Landscape Research, University of Leuven, Belgium,
Vital Decosterstraat 102, B-3000 Leuven, Belgium
Received 24 May 1999; received in revised form 13 December 2000; accepted 15 January 2001

Abstract

Within highly fragmented forest landscapes, the ¯ux of matter, energy and species from the landscape matrix into the forest
fragment interior is an important ecological phenomenon. We studied the invasion of ancient forest edges by weedy plant
species, which are normally con®ned to the agricultural landscape matrix. We related this to edge orientation and to the
occurrence of natural and anthropogenical environmental edge gradients like microclimate and soil chemistry. The plant
community composition of the forest edge zone differed signi®cantly from the community composition of the forest core area,
and is characterized by a relatively high number of competitive and light demanding species. The maximal width of this edge
zone is more extensive at south facing (maximum ca. 20±23 m) than at north facing edges (ca. 0±3 m). However, most of
the weedy plant species were unable to penetrate the forest and were con®ned to the boundary between forest and arable land.
We found no negative edge effects on the more stress tolerant `true' forest species going further than 0±3 m (only at south
facing edges). The invasion pattern generally ®tted the edge gradient for photosynthetically active radiation (PAR). The
negative effects of fertilizer misplacement from adjacent arable land on forest soil chemistry (and community composition) are
also restricted to the boundary of the forest edge zone. Based on the derived penetration distances of the edge effects and on
the shape index of a forest patch it is possible to derive the minimal required forest patch area for forest plant species
conservation. # 2002 Elsevier Science B.V. All rights reserved.

Keywords: Forest invasion; Edge effects; Fertilizer drift; Forest microclimate

1. Introduction at the landscape scale is usually quanti®ed by edge


density (McGarigal and Marks, 1995; Hargis et al.,
Forest fragmentation has led to increasing spatial 1998). Within high edge density landscapes, the ¯ux
isolation and decreasing patch size of forest habitat of matter, energy and species from the landscape
with subsequent negative implications for the survival matrix into the forest fragment interior and vice versa
of the inhabiting natural populations (Harris, 1984). is an important ecological phenomenon (Wiens, 1992;
An important aspect of the forest fragmentation Ryszkowski, 1992).
process is the relative increase of edge habitat, which The main concern in a plant conservation context is
how the ¯uxes directed into the forest fragment may
*
Corresponding author. Tel.: ‡32-16-32-97-73;
in¯uence forest plant dynamics such as regeneration
fax: ‡32-16-32-97-60. and interspecies competition (Murcia, 1995). The
E-mail address: olivier.honnay@agr.kuleuven.ac.be (O. Honnay). impact on forest plant dynamics and diversity will

0378-1127/02/$ ± see front matter # 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 3 7 8 - 1 1 2 7 ( 0 1 ) 0 0 4 9 0 - X
110 O. Honnay et al. / Forest Ecology and Management 161 (2002) 109±122

depend on the penetration distance of the different Belgium, De Schrijver et al. (1998) report a more than
¯uxes, i.e. on the width of the edge zone. Edge width two-fold increase of NH4‡ deposition at the edge of a
can be de®ned as the part of the forest fragment where pine forest. Second a direct input of nutrients (mainly N,
environmental conditions differ signi®cantly from P and K compounds) by drifting fertilizer application,
those in the interior of the fragment, and, hence where runoff and wind erosion from adjacent agricultural land
species composition and abundance also differ (Forman can be expected (Carlson et al., 1990; Kleijn and
and Moore, 1992). Edge width is the result of the Snoeijing, 1997).
penetration distance of various environmental vari- It is known that arti®cial nutrient addition can have
ables. A lot of the available studies, however, did not major consequences for plant species diversity,
directly measure these environmental edge gradients originating from changes in competitive relations
and purely focused on the resulting vegetation patterns between species (Bobbink et al., 1998; Lee, 1998;
(e.g. Ranney et al., 1981; Palik and Murphy, 1990; Turkington et al., 1998). In a forest environment this
Laurance, 1991; Chen et al., 1992; Fraver, 1994; means the competitive replacement of stress tolerant
Kupfer, 1996; Medley, 1997; Buckley et al., 1997; forest plants species by more competitive species
Esseen and Renhorn, 1998; Laurance et al., 1998; (sensu Grime et al., 1988). For forest edge plants the
Meiners and Pickett, 1999). The effects of naturally effects of atmospheric nitrogen deposition on wind
occurring environmental edge gradients like soil exposed ridges in southern Germany (Wilmanns et al.,
temperature, air temperature, photosynthetic active 1986) and the effect of direct fertilizer misplacement
radiation (PAR) and litter moisture on vegetation in Australia have been studied (Hester and Hobbs,
patterns and, more speci®c, on alien plant species 1992). These studies report an important increase of
invasion of the edge zone have also received attention the cover of competitive species in the forest edge
(e.g. Williams-Linera, 1990; Brothers and Spingarn, zone, and an increasing number of exotic species
1992; Matlack, 1993, 1994; Chen et al., 1995; Jose suppressing the original forest ¯ora.
et al., 1996; Williams-Linera et al., 1998). Depending In this study we will investigate the effect of the
on edge orientation, age and physiognomy, the authors interaction between natural and arti®cial environ-
report a transient zone up to maximal 50 m between mental edge gradients on edge community composi-
the landscape matrix and the typical forest core tion and especially on forest edge invasibility by
microclimate. Brothers and Spingarn (1992) argued weedy species which are usually con®ned to the
the typical forest-core microclimate conditions to be a agricultural landscape matrix. We adopt a very broad
relatively ef®cient protection against invasion of alien, de®nition of invasion involving the expansion of a
mostly weedy, species from the landscape matrix. species into a habitat where it is naturally not occurring
Almost all mentioned studies considered north or or only at very low importance values (see, in contrast,
south American forest edges. In the agricultural the more restrictive de®nition in Thompson et al.
landscapes of western Europe another important kind (1995)). We addressed the following questions:
of environmental variation across edges has been
1. Is the plant community composition of the forest
underexplored. In addition to microclimate variation,
edge different from the forest core?
anthropogenic soil nutrient gradients may be expected
2. What is the effect of edge orientation on abiotic
across forest edges as a result of intensive agricultural
edge gradients and on the penetration of weedy
activity in the surrounding landscape matrix. Two
plant species into the forest core?
different physical processes can generate such gra-
3. Do increased nutrient levels in the edge zone
dients. First, forest edges are exposed to relatively
facilitate invasion or does forest microclimate
high atmospheric ion deposition rates because they
inhibits the invasion of non-forest species?
disturb the vertical wind pro®le and cause air turbulence
(Draaijers et al., 1988). At windward spruce forest edge Our study area, located within the central Belgian
zones in the USA, Wheathers et al. (1995) noted on an loam belt consists of large and ancient deciduous
average three times more ion deposition (H‡, NH4‡, forest fragments within a matrix of intensively used
NO3 , SO42 ) than in the interior of the forest. In a zone arable land. The landscape is representative for many
of industrial livestock breeding in the northern part of western European agricultural landscapes.
O. Honnay et al. / Forest Ecology and Management 161 (2002) 109±122 111

2. Materials and methods Table 1


Characteristics of the ®ve surveyed transects
2.1. Study area and data collection Transect Location Orientation Number of 3 m
 3 m quadrats
Our study area is situated in the central Belgian surveyed
loam belt, 25 km to the east of Brussels. It consists of A Bertem forest N 59
three large forest patches that are mainly deciduous B Grevens forest SSW 66
(Meerdaal forest (ca. 1200 ha), Bertem forest (ca. C Meerdaal forest SW 60
D Meerdaal forest NNE 70
200 ha) and Grevens forest (ca. 100 ha)). The forests
E Grevens forest SE 70
are classi®ed as Milio-Fagetum communities (sensu
Noirfalize, 1984). Within this study area we selected
®ve edges, at least 100 m long, and adjacent to
intensively used arable land (rotation of wheat and 20, 30, 40 and 50 m from the forest edge, and 10
sugar beets). All edges are at least 220 years old (i.e. repetitions at each distance, resulting in seventy
present on the oldest available topographical map), are 3 m  3 m quadrats per edge and an overall total of
very sharply delineated, and have a relatively homo- 350 quadrats (Fig. 1). Matlack (1990, 1994) used a
geneous and closed structure consisting of adventi- similar grid pattern. Ultimately 25 quadrats had to be
tious limbs of the boundary canopy trees and a shrub omitted because they were too disturbed (e.g. in the
belt under the canopy. Two of them are north facing vicinity of a ride). The location of the 0 m quadrat at
(N and NNE) and three are south facing (SE, SSE and the forest margin was determined by a visual analysis
SW) (Table 1). Study sites are covered with mature of the Ah soil horizon, indicating the boundary of
mixed deciduous forest (mainly Acer pseudoplatanus, (historical) plow activity.
Quercus robur, Fagus sylvatica and Prunus avium). In May 1998 the vegetation of each quadrat was
The same species are occurring in the shrub (0±2 m), surveyed using a decimal cover scale (Londo, 1974).
the coppice (5±10 m) and in the canopy layer (20±25 m). In September 1998 an additional survey was per-
Soils of the study sites are luvisols. formed to assess seasonal variation. Species not
Within each study site we established a regular present in the May survey were added to the species
53 m  30 m grid (i.e. a transect). Each transect versus quadrats data matrix. Only the herbaceous and
consisted of seven 3 m  3 m quadrats at 0, 5, 10, the shrub layer were surveyed. Coppice and canopy

Fig. 1. Each sampling grid (i.e. transect) consists of seven 3 m  3 m quadrats at 0, 5, 10, 20, 30, 40 and 50 m from the forest edge, together
with nine replicates at each distance, resulting in seventy 3 m  3 m plots per transect.
112 O. Honnay et al. / Forest Ecology and Management 161 (2002) 109±122

layers were omitted because they usually have been Hurlbert, 1984). This means that many transects
planted. For each quadrat we derived the total plant (pseudoreplicates) were surveyed and analyzed, but
species richness and the number of species belonging that they where all lying in the same forest edge
to two functional ecological species groups according (Murcia, 1995). For a landscape scale study, proper
to Grime's CSR model (Grime et al., 1988; Hodgson replicates should be located in different edges. To
et al., 1995); we derived the number of competitive avoid pseudoreplication, we averaged cover values,
species (i.e. competitives (C), competitive ruderals CIVs, number of competitives/stress tolerators and
(CR), C/CR, C/stress tolerant competitives (SC), CR/ microclimate measurements per distance in each of
CSR) and stress tolerant species (i.e. stress tolerators the ®ve studied transects. This resulted in ®ve repli-
(S), S/CSR) per quadrat together with their sum cover. cates per distance category, which are all based on the
We also calculated the characteristic indicator value average of 10 (vegetation data) or 50 (microclimate
for nitrogen (CIV-N), light (CIV-L) and reaction ®gure data) measurements, and hence are considered to be
(CIV-R) for each quadrat (Persson, 1981) based on the very reliable estimates.
Ellenberg indicator ®gures (Ellenberg et al., 1992). Only in a ®rst data exploratory phase we used the
Microclimate measurements (one transect each day) basic quadrat level data (n ˆ 325) in a correspondence
were made between 11 and 18 September 1998, a analysis (CA) to identify general patterns of plant
period with stable, clear weather. We only measured community variation. Vegetation data (% cover) were
between 12.00 and 14.00 h. In each quadrat we made square root transformed to lower the effects of extreme
®ve measurements which were then averaged to one values. CA-quadrat scores were then correlated with
value per quadrat. We also averaged ®ve measure- transect orientation and quadrat distance to the forest
ments in the open ®eld, at 50 m from the forest edge, to edge. We used a Mann±Whitney test and a Spearman
one value. We measured soil temperature (8C), photo- rank correlation, respectively (Siegel and Castellan,
synthetically active radiation (PAR, 400±700 nm, 1988).
W/m2), air temperature (8C) and relative air humidity We then focused on the penetration distance of the
(%), the latter three variables, all at 30 cm above ¯ux of species, into the forest core. We applied a
the ground level. Microclimate measurements were generalized linear modeling (GLM) and analysis of
expressed as fractions of the open ®eld values. For the variance (ANOVA) procedure. Dependent variables
PAR we used a Solar Hog sensor (SDL 5000 series were species cover, CIVs and sum cover and number
solar cell PSU). All data were logged on line with the of competitives and stress tolerators. Independent
Skye Datahog 2 (Skye Instruments Ltd., Powys, UK) model variables were distance to the edge (covariate),
connected to a laptop computer. edge orientation (N or S, ®xed factor) and the inter-
In four of the ®ve transects (not in transect D), soil action term between distance and orientation. To
was sampled to 5 cm depth (after litter removal) in quantify penetration distances for dependent variables
each quadrat (December 1998). For each distance, exhibiting a signi®cant distance effect, we performed
samples were bulked together resulting in one bulk Kruskal±Wallis pairwise comparisons between dis-
sample per distance (i.e. seven per transect). The bulk tances to specify the penetration distance (Siegel
samples were analyzed according to the standard and Castellan, 1988). For dependent variables with
methods of the Belgian Soil Service for available a signi®cant GLM interaction term, penetration
cations (Ca2‡ and K‡) and available phosphate (both distances were calculated for N and S exposed
Egner±Riehm method), carbon content (Walkley and transects separately. Note that for north facing edges
Black method), pH(KCl), available nitrogen (NO3 and our observations are based on only two replicates per
NH4‡) and total N-content (Kjeldahl±Lauro method), distance (however, both based on 10 measurements).
see Hendrickx (1992) for details. Species only occurring in the 0 m edge quadrats (i.e.
unable to penetrate the forest edge zone any further)
2.2. Analyses were also omitted and were separately listed. The
same procedure was followed for the determination
Many studies dealing with the characterization of of the edge width for the microclimate and the soil
edge effects suffered from pseudoreplication (sensu variables. Microclimate variables were standardized
O. Honnay et al. / Forest Ecology and Management 161 (2002) 109±122 113

by dividing them by the open ®eld value and that the residuals were not spatially autocorrelated.
multiplying them by 100. Hence, inferences from our model are valid. VAR-
Besides spatial autocorrelation parallel to the forest IOWIN (Pannatier, 1996) was used to construct the
edge (resulting in pseudoreplication), that can be scatter plots.
avoided by a proper sampling design, autocorrelation For data analysis we used SPSS 9.0 (SPSS Inc.,
perpendicular to the forest edge is an intrinsic problem 1999) and CANOCO for Windows 4.0 (Ter Braak and
of gradient studies. However, from scatter plots of Smilnauer, 1998). Plant species nomenclature follows
model residuals of neighboring quadrats it appeared De Langhe et al. (1988).

Fig. 2. Quadrat scores of the CA (species cover values) on the ®rst two CA axes (pooled dataset, n ˆ 325). Transect identity and quadrat
distance to the forest edge are indicated, respectively.
114 O. Honnay et al. / Forest Ecology and Management 161 (2002) 109±122

3. Results Table 2
Correlation between CA quadrat scores and quadrat characteristics
(n ˆ 325)
The variation in community composition in the
surveyed quadrats can be attributed to the transect Quadrat distance Orientation (north or
orientation (CA axis I; eigenvalue 0.61) and to to forest edgea south facing transect)b
distance to the forest edge (CA axis II; eigenvalue CA x 0.07 11.32***
0.41) (Fig. 2, Table 2). CA y 0.69*** 1.96*
The ANOVA results (Table 3) con®rm the impor- a
Spearman correlation coef®cient.
tance of distance to edge and edge orientation in b
Mann±Whitney z-value.
determining community composition. More speci- *
0:01  P < 0:05.
***
®cally, the signi®cance of the interaction term P < 0:001.
`distance  orientation' for CIV-L, species number,
number and cover of competitives, and the cover of the 0±3 m quadrats (Table 4, Fig. 3). At south facing edges
invading species Galium aparine and Urtica dioica, U. dioica exhibits an increased cover up to 5±8 m, and
indicates that community composition and penetration G. aparine up to 10±13 m into the forest. There is an
distance strongly depends on transect orientation. increased cover of competitive species up to 20±23 m.
At northern edges signi®cantly increased values of Independent from edge orientation, Rubus fruticosus
mentioned response variables are restricted to the coll. has an increased cover value up to 10±13 m,

Table 3
F-values and signi®cance of the GLM ANOVA modela

Intercept Distance Orientation Distance  orientation Model

Species with decreasing cover values


Galium aparine 24.08*** 9.08** 24.08*** 9.08** 13.33***
Glechoma hederacea 5.73* 4.29* 5.74* 3.09 3.55*
Millium effusum 19.73*** 2.56 10.73** 0.88 6.85**
Poa trivialis 16.87** 9.08** 0.19 0.05 3.39*
Prunus serotina 5.15* 0.00 4.35* 0.13 3.76*
Ribes rubrum 5.41* 2.83 5.41* 2.83 3.31*
Rubus fruticosus coll. 40.63*** 11.18*** 1.47 0.22 4.85**
Sambucus nigra (seedling) 12.21** 4.08* 3.18 1.32 2.93*
Ulmus minor 4.75* 2.09 4.75* 2.09 2.74*
Urtica dioica 25.67*** 8.10** 17.56*** 4.64* 10.73***
Species with increasing cover values
Anemone nemorosa 10.5* 7.00* 3.6 4.5 18.1**
Convallaria majalis 0.52 7.70** 0.01 4.40* 5.50**
Community composition indicators
CIV-L 7.15* 9.76** 0.12 3.41* 2.71*
CIV-N 77.52 2.08 10.4* 0.13 2.39*
CIV-R 135.1 0.21 2.11 1.73 2.36*
Species number 230.0 4.39* 8.85** 3.68* 8.68*
Number of stress tolerators 78.94*** 0.13 10.00** 1.48 2.79*
Number of competitives 80.99*** 4.94* 3.23 4.20* 3.34*
Cover of stress tolerators 109.19*** 4.56** 45.71*** 1.91 7.12***
Cover of competitives 16.52*** 8.07** 8.01** 3.99* 6.36**
a
Response ˆ intercept ‡ distance ‡ orientation ‡ distance  orientation. Species cover values were square root transformed. Only
species with a signi®cant model F-value, and not occurring exclusively in the 0 m quadrats are presented.
*
0:01  P < 0:05.
**
0:001  P < 0:01.
***
P < 0:001.
O. Honnay et al. / Forest Ecology and Management 161 (2002) 109±122 115

Table 4
Differences (P < 0:05) in mean cover, mean indicator value and mean species number between different distances to the forest edge according
to pairwise comparisons (indicated with letters)a

0±3 m 5±8 m 10±13 m 20±23 m 30±33 m 40±43 m 50±53 m

Species with decreasing cover values


Galium aparine (south) 0.86 a 0.82 a 0.31 ab 0.11 b 0.18 b 0.03 b 0b
Glechoma hederacea 4.41 a 2.17 a 0b 0b 0b 0b 0b
Hedera helix 1.44 a 0.61 ab 0.13 b 0b 0.05 b 0.03 b 0b
Moehringia trinervis 0.20 0.17 0.03 0 0 0 0
Poa trivialis 2.90 a 0.22 b 0b 0.07 b 0b 0b 0b
Rubus fruticosus coll. 13.00 a 5.82 b 2.14 b 0.64 c 0.44 c 1.07 c 0.94 c
Sambucus nigra (seedling) 0.24 0.39 0.13 0.18 0 0 0
Urtica dioica (south) 40.31 a 8.21 b 2.03 c 2.93 c 2.44 c 1.21 c 0.13 c
Species with increasing cover
Anemone nemorosa 1.28 16.43 48.24 37.60 33.48 36.87 34.16
Convallaria majalis (south) 0.0 1.47 6.10 0.40 0.20 4.40 3.87
Convallaria majalis (north) 0.20 0.80 4.05 11.75 4.85 22.25 10.45
Community composition indicators
CIV-L (south) 7.10 a 6.15 ba 5.85 b 5.15 c 5.12 c 4.40 c 4.24 c
CIV-L (north) 6.97 a 5.18 bc 5.35 bcd 5.42 bcd 4.96 b 5.1 bcd 4.84 d
Species number (south) 10.79 a 10.35 a 7.67 b 6.68 c 5.67 c 6.17 bc 6.38 c
Species number (north) 6.89 a 4.71 ab 5.9 ab 4.23 b 5.59 ab 4.83 ab 6.10 ab
Number of competitives (south) 9.09 a 5.69 b 2.10 b 1.11 c 1.0 c 1.03 c 0.93 c
Number of competitives (north) 2.42 a 0.85 b 1.05 b 1.18 b 1.88 b 1.28 b 1.10 b
Cover stress tolerators 1.82 a 20.78 b 55.50 c 43.74 bc 29.21 bc 55.59 c 48.79 bc
Cover competitives (south) 60.13 a 14.33 b 4.77 b 4.43 b 3.00 c 2.48 c 0.87 c
Cover competitives (north) 10.65 a 0.55 b 0.55 b 0.25 b 0.25 b 0.70 b 0.20 b
a
Only variables with a signi®cant distance effect in the full model are presented (see Table 3). Variables with a signi®cant interaction term
in the full model are evaluated for north and south facing edges separately when enough data available. Edge width, if derivable, is indicated in
bold.

Glechoma hederacea up to 5±8 m. Two species 4. Discussion


(Anemone nemorosa and Convallaria majalis) exhibit
a positive distance effect, their cover increases core 4.1. Biotic edge gradients
inwards. Table 5 lists the species which are exclusively
occurring in the 0±3 m quadrats, i.e. species which are The fringe of forest patches consists of a zone
unable to penetrate the forest edge zone further than where the plant community composition differs from
0±3 m. The boundary of south facing edges is more the community composition of the forest core area.
species rich than the one of north facing edges. The studied edge communities are characterized by a
Also air and soil temperature vary with distance, relatively high abundance of competitive and light
dependent on edge orientation (Table 6). There is no demanding species, especially at south facing edges.
temperature gradient at north facing edges while there Based on the sum cover of all competitive species, we
is an edge effect up to 20±23 m for air and 0±3 m for suggest an edge zone of at least 20±23 m at south
soil temperature at south facing edges (Table 7, Fig. 4). facing, and of 0±3 m at north facing edges. It is known
Edge width for PAR is 0±3 m (independent from edge that the vegetation pattern with distance from the edge
orientation) while there is no distance effect for varies with time since edge formation and also with
relative air humidity. Soil chemical characteristics edge physiognomy (e.g. Matlack, 1993; Kupfer, 1996;
vary with distance independent from edge orientation. Esseen and Renhorn, 1998). In studies concerned with
For pH, Ca and K content, The Kruskal±Wallis late successional edges of temperate regions, like the
pairwise comparisons reveal a small edge zone. ones examined in this study, the edge widths reported
116 O. Honnay et al. / Forest Ecology and Management 161 (2002) 109±122

Fig. 3. Some typical examples of the biotic variation across the forest edge. The variation of the cover value of Rubus fruticosus coll. and
Urtica dioica (%) as a function of the distance to the forest edge is shown together with the sum cover (%) of all competitive plant species.
Bars represent 1 S.E.

by Palik and Murphy (1990) and Matlack (1994) in the Although changes in community composition can
eastern part of the USA, are comparable with the ones be detected up to 20±23 m in the forest (at south facing
we found. Fraver (1994), studying mixed hardwood edges), far most of the invading weedy species are
forests in North Carolina (USA), mentions much con®ned to the 0±3 m quadrats and only a very
larger edge widths varying between 20 m (N) and restricted number of species is able to penetrate the
60 m (S) in at least 50 years old edges. However, forest edge zone deeper than the established edge
it is very dif®cult to assess and compare the edge boundary quadrats. These few deeper penetrating
physiognomy and the corresponding degree of edge species exhibit sharply decreasing cover values when
closure in the different studies. Because edge physiog- expanding into the forest core, results which are
nomy affects the amount of light that reaches the corroborated by the ®ndings of Fraver (1994), Matlack
forest ¯oor and consequently also the intensity of (1993) and Brothers and Spingarn (1992). Particularly
most other environmental edge gradients, it is a major the latter authors explicitly studied forest edge
source of variation. invasion by alien (mostly weedy) species in Indiana
O. Honnay et al. / Forest Ecology and Management 161 (2002) 109±122 117

Table 5 species, although we found a negative effect on their


Species occurring exclusively at 0 m quadrats at south and north
sum cover value in the 0±3 m plots, both at south and
facing edges
north facing edges. This means that the edge effects on
Species occurring at south facing edges forest interior plant communities are relatively minor
Agrostis capillaris in our study area. Fraver (1994) also found no
Arctium spp.
Artemisia vulgaris
evidence of suppression of forest `interior' species,
Cirsium arvense even at the edge boundary. The question now remains
Deschampsia cespitosa which environmental factors prevent the invaders from
Festuca gigantea penetrating further into the forest edge and what is the
Holcus lanatus role of the soil chemistry gradients across the edge.
Heracleum sphondylium
Lamium album
Poa annua 4.2. Abiotic edge gradients
Poa nemoralis
Prunus avium (seedling) Although microclimate measurements were only
Ranunculus repens recorded during 1 week and one time period, and give
Rumex obtusifolius
Solanum dulcamara
no de®nitive representation of the micro climate at the
Stachys sylvatica site, our measurements are comparable with other
Stellaria media studies. Including diurnal microclimate variation,
Symphytum officinale Chen et al. (1995) already concluded that at temperate
Trifolium repens latitudes edge effect are strongest at SW facing edges
Taraxacum officinale
Vicia cracca
and weakest at NE facing edges. Our study con®rms
this for soil and air temperature in deciduous European
Species occurring at north facing edges
forests. At south facing, completely closed edges in
Agrostis capillaris
Holcus lanatus Pennsylvania and Delaware (USA), Matlack (1993)
Luzula multiflora could not detect microclimate gradients. At very
Poa annua recently established (<5 years), south facing edges,
Poa trivialis edge width ranged from 13 m (PAR) over 14 m (RH)
Stellaria media
up to 24 m for air temperature. Cadenasso et al.
Taraxacum officinale
Teucrium scorodonia (1997), studying NE and NW facing edges in New
Urtica dioica York (USA) report edge widths of 6.5±16.3 m for noon
temperature, of 8.4±17 m for soil temperature and of
11.8±25.1 m for relative air humidity. Although more
(USA) and report that most of the invaders at forest precise values were obtained in these studies, due to
margins have moved no more than two meter into the the used sampling design and the analysis methods,
forest interior. West and south facing edges appeared these ®gures are higher compared with our own
to be more rich in alien species. Meiners and Pickett results. This suggests, that ancient forest edges as the
(1999) report peak abundances of exotic species in ones studied have a relatively closed structure and are
New Jersey (USA) at 20 m of the forest edge. The relatively impermeable. This is due to the well
authors, however, studied relatively young, early developed shrub and coppice layer and to the
successional forests on former arable land. It is known sprouting of adventitious limbs of the boundary trees.
that soil disturbance, even historical, may increase the The reported pH values of the forest interior soil (ca.
likelihood of the invasion of a community (Hobbs and 3.5) correspond with the typical pH of the Ah and E
Huenneke, 1992). soil horizon under ancient forest developed on loess
The restricted invasive capacity of the weedy (Bossuyt et al., 1999). The increased pH values at the 0
species apparently prevents them to negatively affect and the 5 m quadrats suggest the input of fertilizer
the interior forest communities with their typical stress compounds from neighboring arable land (where the
tolerant forest plant species; we found no negative pH increases up to 6.5). It is known that liming and
effect of distance on the number of stress-tolerating fertilizing prevents soils under agricultural use from a
118 O. Honnay et al. / Forest Ecology and Management 161 (2002) 109±122

Table 6
F-values and signi®cance of the GLM ANOVA modela

Intercept Distance Orientation Distance  orientation Model


*** ** * **
Tair (8C) 5179.64 7.74 5.09 8.65 6.82***
Tsoil (8C) 761.91*** 5.16* 1.69 3.90* 3.83*
Relative humidity (%) 152.85*** 4.19 0.15 2.09 3.88*
PAR (W/m2) 84.16*** 8.00** 2.36 0.00 4.75**
pH(KCl) 463.14*** 10.08** 1.14 0.12 4.86*
Ca 113.87*** 24.52*** 1.20 0.69 8.83***
K 6.59*** 13.75** 0.42 0.41 6.59**
Mg 301.41*** 12.03** 5.08* 0.11 10.09***
P 29.87*** 6.48* 1.24 1.96 5.21*
Ntot 66.23*** 1.88 0.21 0.25 1.32
NO3 38.10*** 0.17 0.73 0.04 0.57
NH4‡ 62.33*** 4.23* 11.93** 4.02 4.84*
C 43.49*** 5.44 0.26 0.63 1.86
a
Response ˆ intercept ‡ distance ‡ orientation ‡ distance  orientation. Soil chemical characteristics in mg/100 g soil.
*
0:01  P < 0:05.
**
0:001  P < 0:01.
***
P < 0:001.

pH drop and soil degradation (pedological leaching) re¯ected in an increased plant available P or increased
which typically follows afforestation (HuÈttl and NO3 /NH4‡ levels. The N or P input from neighbor-
Schaaf, 1995; Markewitz et al., 1998; Bossuyt et al., ing arable land is clearly not important enough to
1999). Also Wilmanns et al. (1986) and Hester and cause persistent soil chemistry changes and to mask
Hobbs (1992) noticed increased soil pH values at the naturally occurring litter composition gradients in
forest edges which were exposed to drift of agro- forest soils, which re¯ect the canopy composition. We
chemicals. The fertilizer misplacement hypothesis is cannot explain this, but probably this inconsistency is
further supported by the higher K content of the related to the application form of the fertilizer. In the
topsoil of the 0 m quadrats but is on the other hand not study area, Ca and K are applied as powders and hence

Table 7
Differences in mean value of the environmental variables between different distances to the forest edge according to Kruskal±Wallis pairwise
comparisonsa

0±3 m 5±8 m 10±13 m 20±23 m 30±33 m 40±43 m 50±53 m

PAR 0.032 a 0.015 b 0.012 b 0.016 b 0.011 b 0.016 b 0.013 b


T (north) 94.38 94.39 94.48 94.38 94.36 94.34 94.39
T (south) 96.73 a 94.81 b 94.24 b 93.74 bc 94.02 c 93.79 c 93.35 c
Tsoil (north) 64.09 63.73 63.60 63.80 63.64 63.55 63.51
Tsoil (south) 68.61 a 64.42 ab 62.75 b 62.23 b 63.34 b 62.97 b 62.65 b
pH 4.62 a 4.10 a 3.58 ab 3.50 ab 3.45 ab 3.30 b 3.23 b
Ca 150.21 a 116.75 ab 89.75 b 70.00 b 56.75 b 53.00 b 43.33 b
K 17.00 a 14.25 ab 9.50 b 7.75 b 7.00 b 7.33 b 6.00 b
Mg 22.00 19.00 15.75 13.25 14.25 15.00 13.00
P 15.50 15.75 14.75 19.25 17.75 20.33 25.33
NH4‡ 2.76 3.09 5.68 3.36 2.84 3.55 1.88
a
Differences (P < 0:05) are indicated with letters. Only variables with a signi®cant distance effect in the full model are presented
(see Table 6). Variables with a signi®cant interaction term in the full model are evaluated for north and south facing edges separately.
The microclimate variables were standardized with respect to the open ®eld value (ˆ100).
O. Honnay et al. / Forest Ecology and Management 161 (2002) 109±122 119

Fig. 4. Soil chemistry gradients (potassium and pH) and PAR variation across the forest edge. Bars represent 1 S.E.

are able to penetrate the forest edge boundary 4.3. Linking biotic and abiotic gradients
(attached to soil particulates or by runoff of surface
water). P and N compounds to the contrary are applied We can conclude that most weedy species are
as granulates which ®nd more dif®culties to penetrate unable to penetrate the forest edge any further than the
the forest edge. boundary quadrats. This pattern very well ®ts the PAR
Finally, soil acidi®cation of the edge boundary gradient. Increased PAR values are only noticed at the
quadrats (which are expected as a result of the 0 m quadrats. Other authors have already suggested
oxidation of excessive NH4‡ deposition) could not be that the penetration distance of weedy species is
noticed. In the Belgian loess belt these deposition rates mainly restricted because of the limited penetration of
are rather low and, moreover, the NH4‡ atmospheric light (Brothers and Spingarn, 1992). This seems to be
deposition gradient across the forest edge may be too con®rmed for most of the encountered invasive
¯at to be detected within a 50 m transect. De Schrijver species, which are unable to penetrate the forest core
et al. (1998) have reported a gradient over more than deeper than the 0 m quadrats. In contrast with mid-
180 m in the sandy northern part of Belgium. successional species, the early successional species of
120 O. Honnay et al. / Forest Ecology and Management 161 (2002) 109±122

the landscape matrix have no adaptations for shade analysis of the soil samples and to Bea Bossuyt and
tolerance (e.g. strong apical dominance). Hence, they two anonymous reviewers for reading and discussing
have no chance to penetrate and to reproduce under the manuscript.
canopy cover (Henry and Aarssen, 1997). The two
furthest penetrating species are R. fruticosus coll. and
U. dioica, both species which can grow under References
relatively low PAR levels (`stress tolerating compe-
titives' sensu Grime). This pattern is also in accor- Bobbink, R., Hornung, M., Roelofs, J.G.M., 1998. The effects of
dance with the patterns of increased nutrient levels air-borne nitrogen pollutants on species diversity in natural and
seminatural European vegetation. J. Ecol. 86, 717±738.
across the edge. The increased agro-nutrient input (Ca,
Bossuyt, B., Deckers, J., Hermy, M., 1999. A ®eld methodology for
K) is restricted to the 0 m quadrats and the 5 m assessing man-made disturbance in forest soils developed in
quadrats (for pH) and works probably synergistically loess. Soil Use Manage. 15, 14±20.
with the increased PAR values of the edge. Generally Brothers, S.T., Spingarn, A., 1992. Forest fragmentation and alien
we can conclude, however, that the studied ancient plant invasion of central Indiana old-growth forests. Conserv.
Biol. 6, 91±100.
forest edges are relatively impermeable for weedy
Buckley, G.P., Howell, R., Watt, T.A., Ferris-Kaan, R., Anderson,
species invasion. M.A., 1997. Vegetation succession following ridge edge manage-
ment in lowland plantations and woods. 1. The in¯uence of site
4.4. Implications for conservation factors and management practices. Biol. Conserv. 82, 289±304.
Cadenasso, M.L., Traynor, M.M., Pickett, S.T.A., 1997. Functional
location of forest edges: gradients of multiple physical factors.
Harrison and Bruna (1999) have recently suggested
Can. J. For. Res. 27, 774±782.
that habitat fragmentation may be more a problem of Carlson, C.G., Dean, R., Lemme, G., 1990. Prescription planning:
edge effects and less of spatial habitat con®guration an approach to nonpoint pollution problems. J. Water Soil
(i.e. patch isolation and patch area). For forest plant Conserv. 34, 239±241.
species in forest fragments this has been con®rmed Chen, J., Franklin, J.F., Spies, T.A., 1992. Vegetation responses to
edge environments in old growth Douglas-Fir forests. Ecol.
by Honnay et al. (1999). Therefore, the importance
Appl. 2, 387±396.
of the determination of the forest core area is a very Chen, J., Franklin, J.F., Spies, T.A., 1995. Growing-season
important conservation issue. The area of a forest microclimatic gradients from clear-cut edges into old-growth
patch that is unaffected by edge effects not only Douglas-Fir forests. Ecol. Appl. 5, 74±86.
depends on the penetration distance of the edge De Langhe, J.E., Delvosalle, L., Duvigneaud, J., Van den Berghen,
C., 1988. Flora van BelgieÈ, het Groothertogdom Luxemburg,
effect, and on the patch area but also on the shape of
Noord Frankrijk en de aangrenzende gebieden. Patrimonium
the patch (Laurance and Yensen, 1991). The shape of a van de Nationale Plantentuin van BelgieÈ.
habitat patch can be described by the Patton shape De Schrijver, A., Nachtergale, L., Roskams, P., De Keersmaeker,
index (PSI) which is de®ned by P/(200(pA))0.5 with P L., Mussche, S., Lust, N., 1998. Soil acidi®cation along an
the perimeter (m) and A the area (ha) of the forest ammonium deposition gradient in a Corsican pine stand in
northern Belgium. Environ. Pollut. 102S1, 427±431.
patch (Patton, 1975). Still according to Laurance and
Draaijers, G.P.J., Ivens, P.M.F., Bleuten, W., 1988. Atmospheric
Yensen, the core area can be calculated as follows: deposition in forest edges measured by canopy throughfall.
total area …TA† affected area (AA) with AA ˆ Water Air Soil Pollut. 42, 129±136.
3:55d  PSI  …TA/10 000)0.5, where d (m) is the Ellenberg, H., Weber, H.E., DuÈll, R., Wirth, V., Werner, W.,
penetration distance of the edge effects. Based on the Paulissen, D., 1992. Zeigerwertenvon P¯anzen in Mitteleuropa.
Scripta Geobotanica 181±258
earlier derived penetration distances, it is possible to
Esseen, P.A., Renhorn, K.E., 1998. Edge effects on an epiphytic
assess the extent of edge effects for any given forest lichen in fragmented forests. Conserv. Biol. 12, 1307±1317.
fragment. Forman, R.T.T., Moore, P.N., 1992. Theoretical foundations for
understanding boundaries in landscape mosaics. In: Hansen,
A.J.H., Di Castri, F. (Eds.), Landscape BoundariesÐConse-
quences for Biotic Diversity and Ecological Flows. Springer,
Acknowledgements
Berlin, pp. 236±258.
Fraver, S., 1994. Vegetation responses along edge-to-interior
Thanks to Peter Cokelaere for help with data gradients in the mixed hardwood forests of the Roanoke river
preparation and data collection and the chemical basin, North Carolina. Conserv. Biol. 8, 822±832.
O. Honnay et al. / Forest Ecology and Management 161 (2002) 109±122 121

Grime, J.P., Hodgson, J.G., Hunt, R., 1988. Comparative Plant Londo, G., 1974. De decimale schaal voor de vegetatiekundige
Ecology: A Functional Approach to Common British Species. opnamen van permanente kwadraten. Gorteria 7, 106±107.
Unwin±Hyman, UK. Markewitz, D., Richter, D.D., Allen, H.L., Urrego, J.B., 1998.
Hargis, C.D., Bissonette, J.A., David, J.L., 1998. The behavior of Three decades of observed soil acidi®cation in the Calhoun
landscape metrics commonly used in the study of habitat experimental forest: has acid rain made a difference? Soil Sci.
fragmentation. Landsc. Ecol. 13, 167±186. Soc. Am. J. 62, 1428±1439.
Harris, L.D., 1984. The Fragmented Forest: Island Biogeography Matlack, G.R., 1993. Microenvironment variation within and
Theory and the Preservation of Biotic Diversity. University of among forest edge sites in the eastern United States. Biol.
Chicago Press, Chicago, USA. Conserv. 66, 185±194.
Harrison, S., Bruna, E., 1999. Habitat fragmentation and large- Matlack, G.R., 1994. Vegetation dynamics in the forest edgeÐ
scale conservation: what do we know for sure? Ecography 22, trends in space and successional time. J. Ecol. 82, 113±123.
225±232. McGarigal, K., Marks, B., 1995. FRAGSTATS: spatial pattern
Hendrickx, H., 1992. De chemische bodemvruchtbaarheid van het analysis program for quantifying landscape structure. USDA
Vlaamse akker- en weilandareaal. Bodemkundige Dienst van Forest Service.
BelgieÈ. Medley, K.E., 1997. Distribution of the non-native shrub Lonicera
Henry, H.A.L., Aarssen, L.W., 1997. On the relationship between maackii in Kramer woods, Ohio. Phys. Geogr. 18, 18±36.
shade tolerance and shade avoidance strategies in woodland Meiners, S.J., Pickett, S.T.A., 1999. Changes in community and
plants. Oikos 80, 575±581. population responses across a forest±®eld gradient. Ecography
Hester, A.J., Hobbs, R.J., 1992. In¯uence of ®re and soil nutrients 22, 261±267.
on native and non-native annuals at remnant vegetation edges in Murcia, C., 1995. Edge effects in fragmented forests: implications
the western Australian wheatbelt. J. Veg. Sci. 14, 101±108. for conservation. TREE 10, 58±62.
Hobbs, R.J., Huenneke, L.F., 1992. Disturbance, diversity and Noirfalize, A., 1984. ForeÃts et stations forestieres en Belgique.
invasion: implications for conservation. Conserv. Biol. 6, 324± Presse agronomique, Gembloux, Belgique.
337. Palik, B.J., Murphy, P.G., 1990. Disturbance versus edge effects in
Hodgson, J.G., Grime, J.P., Hunt, R., Thompson, K., 1995. The sugar maple/beech forest fragments. For. Ecol. Manage. 32,
Electronic Comparative Plant Ecology. Chapman & Hall, New 187±202.
York. Pannatier, Y., 1996. VARIOWIN: Software for Spatial Data
Honnay, O., Hermy, M., Coppin, P., 1999. Effects of patch area, Analysis in 2D. Springer, New York.
habitat diversity and age of patch on forest plant species Patton, D.R., 1975. A diversity index for quantifying habitat edge.
richness in Belgium and implications for conservation and Wildl. Soc. Bull. 3, 171±173.
reforestation. Biol. Conserv. 87, 73±84. Persson, S., 1981. Ecological indicator values as an aid in the
Hurlbert, S.H., 1984. Pseudoreplication and the design of interpretation of ordination diagrams. J. Ecol. 69, 71±84.
ecological experiments. Ecol. Monogr. 54, 187±211. Ranney, J.W., Bruner, M.C., Levenson, J.B., 1981. The importance
HuÈttl, R.F., Schaaf, W., 1995. Nutrient supply of forest soils in of edge in the structure and dynamics of forest islands. In:
relation to management and site history. Plant and Soil 168169, Burgess, R.L., Sharpe, D.M. (Eds.), Forest Island dynamics in
31±41. man dominated landscapes. Springer, Berlin, pp. 65±95.
Jose, S., Gillespie, A.R., George, S.J., Kumar, B.M., 1996. Ryszkowski, L., 1992. Energy and material ¯ows across boundaries
Vegetation responses along edge-to-interior gradients in a high in agricultural landscapes. In: Hansen, A.J.H., Di Castri, F.
altitude tropical forest in peninsular India. For. Ecol. Manage. (Eds.), Landscape BoundariesÐConsequences for Biotic Di-
87, 51±62. versity and Ecological Flows. Springer, Berlin, pp. 270±284.
Kleijn, D., Snoeijing, G.I.J., 1997. Field boundary vegetation and Siegel, S., Castellan, N.J., 1988. Non-parametric Statistics for the
the effects of agrochemical drift: botanical change caused by low Behavioral Sciences. McGraw-Hill, New York.
levels of herbicide and fertilizer. J. Appl. Ecol. 34, 1413±1425. Ter Braak, C.J.F., Smilnauer, P., 1998. Canoco for Windows,
Kupfer, J., 1996. Patterns and determinants of edge vegetation of Version 4.0. CPRO-DLO, Wageningen.
midwestern forest preserve. Phys. Geogr. 17, 62±76. Thompson, K., Hodgson, J.G., Rich, T.C.G., 1995. Native and alien
Laurance, W.F., 1991. Edge effects in tropical forest fragments: invasive plants: more of the same? Ecography 18, 390±402.
application of a model for the design of nature reserves. Biol. Turkington, R., John, E., Krebs, C.J., Dale, M.R.T., Nams, V.O.,
Conserv. 57, 205±219. Boonstra, R., Boutin, S., Martin, K., Sinclair, A.R.E., Smith,
Laurance, W.F., Yensen, E., 1991. Predicting the impacts of edge J.N.M., 1998. The effects of NPK fertilization for nine years on
effects in fragmented habitats. Biol. Conserv. 55, 77±92. boreal forest vegetation in northwestern Canada. J. Veg. Sci. 9,
Laurance, W.F., Ferreira, L.V., Rankin-De Merona, J.M., Laurance, 333±346.
S.G., Hutchings, R.W., Lovejoy, T.E., 1998. Effects of forest Wheathers, K.C., Lovett, G.M., Likens, G.E., 1995. Cloud
fragmentation on recruitment patterns in Amazonian tree deposition to a spruce forest edge. Atmos. Environ. 29, 665±672.
communities. Conserv. Biol. 12, 460±464. Wiens, J.A., 1992. Ecological ¯ows across landscape boundaries: a
Lee, J.A., 1998. Unintentional experiments with terrestrial conceptual model. In: Hansen, A.J.H., Di Castr, F. (Eds.),
ecosystems: ecological effects of sulfur and nitrogen pollutants. Landscape BoundariesÐConsequences for Biotic Diversity and
J. Ecol. 86, 1±12. Ecological Flows. Springer, Berlin, pp. 217±235.
122 O. Honnay et al. / Forest Ecology and Management 161 (2002) 109±122

Williams-Linera, G., 1990. Vegetation structure and environmental Wilmanns, O., Bogenrieder, A., MuÈller, W.H., 1986. Der nachweis
conditions of forest edges in Panama. J. Ecol. 78, 356±373. spontaner, teils autogener, teils immisionsbedinter Anderungen
Williams-Linera, G., Dominguez-Gastelu, V., Garcia-Zurita, M.E., von Eichen-HainbuchenwaÈldernÐene Fallstudie im Kaiserstuhl/
1998. Microenvironment and ¯oristics of different edges in a Baden. Natur und Landschaft 61 Jg. Heft 11, 415±422.
fragmented tropical rainforest. Conserv. Biol. 12, 1091±1102.

You might also like