Forest Ecology and Management 256 (2008) 106–113

Contents lists available at ScienceDirect

Forest Ecology and Management

journal homepage: www.elsevier.com/locate/foreco

Establishment patterns of woody species in low intensity-grazed pasturesafter the cessation of intensive agricultural use

Jan Van Uytvanck a,b,*, Kris Decleer a, Maurice Hoffmann a,b

a Research Institute for Nature and Forest, Kliniekstraat 25, B-1070 Brussels, Belgiumb Terrestrial Ecology Unit, Department of Biology, Ghent University, K.L. Ledeganckstraat 35, B-9000 Ghent, Belgium

A R T I C L E I N F O

Article history:

Received 1 February 2008

Received in revised form 31 March 2008

Accepted 1 April 2008

Keywords:

Fraxinus excelsior

Facilitation

Grazing

Herbivory

Tree establishment

Woodland expansion and development

A B S T R A C T

Low intensity grazing with large herbivores is increasingly used to steer the development of mosaic

landscapes (including wood-pastures) and woodland expansion on former agricultural land. We studied

the establishment and early growth pattern of woody species in grasslands (formerly used for high

intensity summer cattle grazing) in which a low intensity-grazing regime was set up between 4 and 11

years ago. We tested whether the developing vegetation patches, consisting of tall herbs and scrub,

influenced this pattern. Further, we investigated if increased susceptibility to trampling disturbance by

large herbivores on wet soils in floodplains influenced sapling establishment.

Formerly intensively used grasslands were studied at six sites, four in floodplains and two in

interfluvial zones on nutrient rich, heavy soils. In total, we sampled 141 plots with grassland, tall herb or

scrub vegetation and recorded individuals of established tree and shrub species, their height, diameter,

browsing damage and distance to potential seed sources. In the floodplains, we recorded soil disturbance

by large herbivores (hoofprints) in wet and moist soils.

Sapling frequencies, with Fraxinus excelsior as the most common species (>50% of all individuals),

were two to three times higher in tall herb and scrub patches compared to grassland. Only the spiny

Crataegus monogyna established in higher frequencies in grassland plots. In floodplains, we found two to

three times more established woody saplings in wet tall herb plots, compared to moist and wet grassland

plots and moist tall herb plots.

Browsing reduced sapling height in grassland and tall herb patches. Only in scrub patches did average

height (�2.5 m) reach above the browse line. We found 30–70% less browsing damage to saplings in scrub

compared to damage in grassland and tall herb plots. In particular, damage to the apical shoot was strongly

reduced.

Our findings show that tree establishment occurs in spatial association with non-grassland patches

that offer protection against browsing. Fast growing spiny shrubs like Rubus sp. offer protection for

palatable saplings, allowing establishment and growing out beyond the browse line. Alternatively,

increased establishment occurs in wet tall herb patches, where soil disturbance by moderate trampling

favours germination and, subsequently, saplings are protected at least temporarily against browsing by

unpalatable tall herbs.

� 2008 Elsevier B.V. All rights reserved.

1. Introduction

Low intensity grazing with large herbivores is increasingly usedin conservation management (WallisDeVries et al., 1998; Bakkerand Londo, 1998; Bakker et al., 2006). The principal idea is thatlarge herbivores, whether wild or livestock species, can play a key

* Corresponding author at: Research Institute for Nature and Forest, Kliniekstraat

25, B-1070 Brussels, Belgium. Tel.: +32 2 558 18 64; fax: +32 2 558 18 05.

E-mail addresses: jan.vanuytvanck@inbo.be, jan_ingrid@telenet.be

(J. Van Uytvanck).

0378-1127/$ – see front matter � 2008 Elsevier B.V. All rights reserved.

doi:10.1016/j.foreco.2008.04.008

role in restoring the functioning of ecological processes. Nativewoodland regeneration on formerly intensively used grassland isan important item in this view. In the British uplands for example,there is interest in the concept of ‘re-wilding’ and large-scalenative woodland restoration in conjunction with extensification ofagriculture (Worrell et al., 2002; Kirby, 2003; Hodder et al., 2005).In this context, Holl and Smith (2007) argue that targeted grazingmanagement is needed to restore the ecological functioning ofsuch landscapes, with a range of representative habitats (includingdifferent woodland types), communities and species at differentscales. In Flanders (N-Belgium) low intensity grazing by semi-domesticated animals is currently explored as an alternative for

J. Van Uytvanck et al. / Forest Ecology and Management 256 (2008) 106–113 107

active afforestation or spontaneous regeneration and developmentof wood-pastures.

It is generally known that the regeneration of woody species isstrongly affected by large herbivores, but local site differences,including spatial effects, palatability and competitive capacities ofground vegetation, grazing intensity, herbivore and tree species,will influence the interactions between herbivores and woodyspecies (Kramer et al., 2006). Large herbivores like deer are able toregulate the balance between woody and herbaceous vegetationwhere agricultural land has been abandoned (Pepin et al., 2006),but also free ranging domestic large herbivores may maintain ordevelop fine-meshed (often mid-successional) mosaic land-scapes, including grassland, scrub, scattered trees and woodlandpatches (Bokdam and Gleichman, 2000; Vera, 2000; Hodder et al.,2005). However, livestock herbivory has a complex relationshipwith tree regeneration (Sankey et al., 2006). On one hand, largeherbivores have been found to retard or prevent tree regenerationin wooded landscapes, dehesas and wood-pastures (Putman,1986; Moser et al., 2006; Linares, 2007; Vandenberghe et al.,2007). On the other hand, large herbivores may enhance woodyspecies establishment, e.g. in grassland vegetation in temperateregions (Kirby et al., 1994; Olff et al., 1999; Vera, 2000), but also inlogged forests with a palatable understory (e.g. bamboo, Darabantet al., 2007). Generally, a dense and closed grass canopy reducesfavourable sites for germination and limits growth and survivalrate of established saplings (Davis et al., 1999). The lack of suitablegermination sites in grassland may be a limiting factor forseedling recruitment (Mazia et al., 2001; Setterfield, 2002).Grazing, trampling and uprooting by large herbivores can opendense grass swards, increase favourable microsite availability,temporarily decrease competitive ability of grasses and favour thegermination and establishment potential of woody species(Posada et al., 2000; de Villalobos et al., 2005; Vandenbergheet al., 2006).

Woody species establishment in open and grazed landscapesoccurs in a non-random spatial pattern: saplings are often found inclose association with plants or vegetation patches that facilitatetheir establishment (Smit et al., 2005) and provide a certain level ofprotection against herbivory. The latter seems to be crucial forsuccessful establishment of woody species in grassland ecosys-

Fig. 1. Location of the study sites in Europe (left) and in the Western and Central Hills E

BM = Boelaremeersen; RB = Rietbeemd; 6 = Valier.

tems. Saplings may be protected against browsing by unpalatable,thorny or spiny species (often shrubs), which provide safe sites,although higher seed removal by predators under these shrubsmay limit establishment (Smit et al., 2006). Moreover, browsingdamage by small rodents, finding cover against predators indense ground vegetation, may increase (Ostfeld and Canham,1993). Facilitating shrubs usually have the ability to establishin open sites, produce attractive fruits, discourage growth ofcompetitive grasses, ameliorate soil conditions, sprout readily afterbrowsing damage and compete relatively little for water andnutrients (Vieira et al., 1994). The process of facilitation in grazedecosystems is known as associational resistance (Huntly, 1991;Olff et al., 1999). It is determined by a large spectrum of variables,such as herbivore selectivity, herbivore density, palatability offacilitating plants, plant strategies and light requirements of thefacilitated trees and their surrounding, facilitating plants (Call-away, 1992; Rousset and Lepart, 2000). Evidence is growing that itis a key process in the dynamics of grazed ecosystems that includewoodland (Olff et al., 1999; Bokdam, 2001; Bakker et al., 2004). Olffet al. (1999) presented a descriptive model of long-term cyclicsuccession, in which low intensity grazing by free rangingherbivores, associational resistance and competition play a deter-mining role in succession.

In this paper, we focus on the early stages of tree establishmentin grasslands, that were formerly used for high intensity summergrazing, in which the grazing regime was recently altered to a lowintensity-grazing regime. We studied the tree establishmentpattern and the developing spatial associations in six naturereserves on former intensively used agricultural grasslands. Weinvestigated whether the frequencies of the five most commonwoody species colonising these grasslands were influenced bypresence and distance to potential seed sources. Negativecorrelations were expected. Further, we explored tree establish-ment on soils with different sensitivity to large herbivoretrampling. We hypothesized that on wet soils, which are moresusceptible to trampling disturbance, the opening of the swardby this trampling offers better establishment conditions. Wehypothesized that, ultimately, the protective capacities againstbrowsing of the present structural vegetation types will determinethe establishment pattern of newly colonising woody species.

coregion in Belgium (right): BE = Bos t’ Ename; BU = Burreken; MB = Moenebroek;

J. Van Uytvanck et al. / Forest Ecology and Management 256 (2008) 106–113108

2. Methods

2.1. Study sites

A survey study was conducted in six nature reserves situated inthe Western and Central Hills Ecoregion in the southern part ofFlanders (N-Belgium, Fig. 1, Table 1). The climate is temperate(annual precipitation: �700 mm). The rather small (10–100 ha)nature reserves are located in a hilly landscape, composed of theupper courses, floodplains and interfluvia of two medium-sizedlowland rivers (Schelde and Dender). At all sites, a mosaic of grasslandand woodland is present. Four sites are situated in floodplains, twosites are situated on the slopes of interfluvial zones. Due to localembanking and damming of the rivers, the floodplain nature reservesare only flooded after heavy rainfall during winter. Flooding periodsonly last a few days, with a maximum length of 2 weeks. Soil typesrange from moist sandy loam and loam to wet clay (Table 1).

Before the start of low intensity grazing, all sites consisted ofhomogeneously structured, short-grazed pasture, without anywoody species, tall herb or scrub patches. The moist to wetgrasslands are moderate nutrient rich to nutrient rich types:Alopecurion pratensis, Cynosurion cristati, Lolio-Potentillion and, dueto manuring, impoverished communities that are dominated bythe palatable species Poa trivialis or Lolium perenne. Tall herbpatches are numerous and are mainly dominated by unpalatablespecies such as Juncus effusus, Carex acutiformis and Epilobium

hirsutum. Moreover, these species and the also abundant Phalaris

arundinacea have rhizomes allowing them to spread fast ingrasslands where grazing pressure has been decreased. Tall herbpatches occur on moist and wet soils, but are more numerous onwet soils. At all sites, sparse natural regeneration of woody speciesis present. Saplings of Fraxinus excelsior, Quercus robur, Salix alba

and Alnus glutinosa are most frequent. Scrub patches mainlyconsist of fast growing bramble thickets (Rubus sp.). Individualshrubs of spiny and thorny species include Prunus spinosa,Crataegus monogyna and Rosa canina. Forest patches are mainlyof the Stellario-Carpinetum (slopes) or the Fraxino-Ulmetum

(floodplains) type.Two grazing systems are applied, both concerning free ranging

animals in unpartitioned pastures: year-round, where animals are

Table 1Study site characteristics

Study site BE BU BM

Coordinates 508860N 508800N 508738650E 38700N 3888

Type IS IS Fl

Dominant soil type Sandy loam Loam Clay

Grazed area (ha) 63.5 2.1 8.9

Grazed vegetation

types

G F A G F A G F L

Sampling area (ha) 2.0 1.5 4.0

Grazing type Y S S

Large herbivores C + H C C

AU ha�1 year�1 0.4 0.8 0.4

Start grazing

management

1996 2000 2001

Number of plots and proportion (%) of sampling area (between brackets)

Grassland 6 (32) 9 (80) 27 (5

Tall herb 6 (39) 0 (4) 24 (3

Scrub 8 (25) 4 (13) 3 (7)

Other 0 (4) 0 (3) 0 (2)

BE = Bos t’ Ename; BU = Burreken; BM = Boelaremeersen; RB = Rietbeemd; MB = Moen

vegetation types: G = grassland; F = forest; A = former Arable land; L = logged forest;

herbivores: C = cattle; H = horses. AU = animal units (1 AU = 450–500 kg).

in the pasture during the whole year and seasonal, where grazing isrestricted to spring and summer (from the beginning of May tillthe end of October). Low intensity grazing, as a conservationmanagement measure on the formerly intensively used grasslands,started between 4 and 11 years before data collecting in this study.Average grazing period after cessation of agricultural use is 6.3, 6.9and 7.5 years for grassland, tall herb and scrub plots, respectively.Actual grazing blocks may also include forest, logged forest andformer arable land (Table 1). Grazing pressure is rather high (�0.4animal units (=AU) ha�1 year�1, i.e. � � 200 kg herbivore biomassha�1 year�1, cf. WallisDeVries, 1998) and fairly constant on all sites,allowing the development of a varied vegetation structure. Cattle areused on all sites, except in Valier, where horses are grazing. In Bos t’Ename and Rietbeemd, respectively, two horses and two ponies areadded to the cattle herd (Table 1). No other large vertebrate grazers ordeer are present on the sites. Rabbits are present at low densities atBos t’ Ename and Burreken and almost absent on the others sites.

2.2. Data collection

Fieldwork was carried out between May and August 2005. Ateach study site, we selected one sampling area (0.5–4 ha, Table 1)where a mosaic of grassland, tall herb and scrub patches appeared.In the sampling area, we established two to five linear transects(evenly spread and ranging between 100 and 300 m length) andlaid out plots with a radius of 10 m. Transects, and the centres ofplots along transects, were 20 m apart. Plots were assigned to oneof three structural vegetation types (grassland, tall herb or scrub)on the basis of a visually estimated percentage cover of plantspecies and vegetation height. We made one vegetation survey(decimal scale Londo, 1984) per plot. Vegetation height wasmeasured at four fixed points in each plot. We used a verticallyheld marked rod. Vegetation height was the maximum height ofthe foliage at the point touched by the rod. For grassland(height < 50 cm), tall herb (height > 50 cm) and scrub (>50 cm),average vegetation height was 13.6 (�8.4), 72.3 (�18.7) and 86.2(�27.8) cm, respectively.

Plots, in which one of these three structural vegetation typesdid not exceed a cover of 80% were excluded. We sampled 141plots: 64 in grassland, 49 in tall herb and 28 in scrub patches.

RB MB VA

80N 508750N 508840N 508880N0E 38880E 38840N 48100E

Fl Fl Fl

Clay Loam Clay

22.5 5.8 1.5

G L G A G

2.5 1.5 1.5

Y S S

C + H C H

0.4 0.5 0.6

1997 1994 1997

3) 5 (43) 7 (37) 10 (67)

8) 7 (24) 12 (48) 0 (4)

6 (28) 1 (13) 6 (27)

0 (5) 0 (3) 0 (2)

ebroek; VA = Valier. Study site type: IS = interfluvial slope; Fl = floodplain. Grazed

grazing type: S = seasonal grazing (May–October); Y = year round grazing. Large

Table 2Correlation between established woody species frequencies and distance to the

nearest potential seed sources (R = Spearman rank order correlation coefficient)

Species R p Analysed

range (m)

No. of

sites

No. of

plots

Fraxinus excelsior �0.78 <0.01 0–200 6 75

Quercus robur �0.90 <0.01 0–160 5 30

Alnus glutinosa �0.62 0.561 0–180 4 22

Salix alba 0.19 0.655 0–460 4 28

Crataegus monogyna �0.95 <0.01 0–200 4 25

J. Van Uytvanck et al. / Forest Ecology and Management 256 (2008) 106–113 109

Proportion of these vegetation structures in the sampling area wasmeasured using a geographical information system (Arcview2) andaerial photographs (AGIV 2006 Eastern Flanders; Table 1). In eachplot, we recorded sapling number, height and diameter (30 cmabove ground level). Only individuals taller than 40 cm or, ifsmaller, older than 1 year (in case of heavily browsed individuals),were recorded (to assess survival and establishment, rather thangermination and emergence). Presence or absence of browsingdamage in general and browsing damage to apical shoots inparticular were recorded. Whether the stem was branched or notat ground level, was recorded as a measure for grazing damage inan early stage. In floodplain plots, we recorded presence or absenceof hoofprints, causing visible soil disturbance. Soil moisture regimeof all plots was estimated based on the vegetation survey by usingweighted averages of Ellenberg’s indicator values (Ellenberg et al.,1992). We divided plots in a moist and a wet category (weightedaverages of Ellenberg’s indicator values between 5.00–6.60 and7.00–8.99, respectively). The position of potential seed sources(i.e. seed producing trees) within a range of 30 m around thegrazed area (Table 1) was recorded for all woody species present inthe sampling area, using a Global Positioning System (GarminGPSMAP76).

2.3. Data analysis

We used generalised linear mixed models (SAS 9.1, proc glimmix)to model established sapling frequencies, presence of browsingdamage and stem branching. We performed a log transformation forPoisson distributed data (frequencies) and a logit transformation forbinomial distributed data (browsing damage and stem branching).We used general linear mixed models to analyse sapling height andstem diameter (SAS 9.1, proc mixed). Both proc glimmix and procmixed cope with unbalanced data.

In order to determine the general tree regeneration pattern, afirst analysis was made with all species pooled together.Subsequently, we analysed the regeneration pattern of speciespresent in at least 20 plots. This was the case for the five mostfrequent species. Analyses on sapling height and diameter andbrowsing damage were performed for F. excelsior (more than 50% ofall individuals) only. In these analyses, we additionally tested theeffect of grazing period after cessation. Impact of soil moisture(moist or wet) was analysed for the floodplains only (no wet plotswere available on the interfluvial slope sites), excluding scrub plots(only present in moist plots). For this analysis, frequencies of allspecies were pooled together.

In the analyses, we entered structural vegetation type, species,grazing period after cessation of agricultural use and moistcategory as fixed factors. Site and plot were modelled as randomfactors. The number of degrees of freedom was estimated using the

Table 3Mean (�S.E.) values for frequency, height, diameter and three browsing damage parameter

results of the GLM-analyses (effect of structural vegetation type)

Response Mean (�S.E.)

Grassland Tall Herb

All species

Sapling frequency per plot 2.23 (�0.54)a 5.94 (�1.74)b

Fraxinus excelsior

Sapling frequency per plot 0.91 (�0.21)a 2.49 (�0.81)b

Sapling height (m) 1.72 (�0.27)a 1.93 (�0.26)a

Sapling diameter (cm) 2.10 (�0.28)a 1.99 (�0.27)a

Browsing damage (general) (%) 84.9 (�4.6)a 79.5 (�4.9)a

Apical shoot damage (%) 56.5 (�11.1)a 52.8 (�10.9)a

Stem branching (%) 50.1 (�12.0)a 50.1 (�11.5)a

Different letters indicate significantly different mean values (post hoc Tukey’s tests).

method of Kenward and Roger (1997). Initial models included allvariables and interactions. They were simplified using a backwardprocedure, using AIC values for goodness of model fit (Akaike,1987), to eliminate non-significant interactions. Site was alwaysretained in the model, because of its inherent role in the survey’sdesign (including both floodplains and slopes). Tukey’s tests wereperformed to make multiple comparisons in all mixed modelanalyses.

For the five most frequent species, the distance to the nearestpotential seed source was calculated from the centre of each plot inwhich the species was present. For this, we used Arcview3.2. Thecorrelation between frequency per plot and distance to the nearestseed source was tested using the Spearman Rank Order Correlation.

3. Results

3.1. Sapling frequencies and height

We recorded a total number of 595 established saplingsbelonging to 15 different woody species. F. excelsior was the mostfrequent species at all sites (305 individuals). F. excelsior, Q. robur,A. glutinosa, S. alba and C. monogyna were present in more than 20plots. In 38 plots, no woody species were recorded (28 grassland, 6tall herb and 4 scrub plots). Distance to the nearest potential seedsource was negatively correlated with sapling frequency per plotfor F. excelsior, Q. robur and C. monogyna. No significant correlationwas found for A. glutinosa and S. alba (Table 2). A series of species, ofwhich numerous seed trees were present in the examined zonesurrounding the sampling areas, did not establish (e.g. Sorbus

aucuparia, Castanea sativa and Salix cinerea) or established in verylow numbers (e.g. Carpinus betulus and Corylus avellana).

The effect of structural vegetation type on the frequency ofestablished woody species (all species together) was significant(Table 3): frequencies in scrub and tall herb were two to threetimes higher than in grassland (Fig. 2). We found no significantdifferences in total sapling frequencies between tall herb andscrub. For the five most numerous species combined, structuralvegetation type (F = 7.13; p < 0.001), species (F = 6.47; p < 0.001)

s of established saplings (plots = circles of 314 m2) per structural vegetation type and

GLM-analysis

Scrub Num d.f. Den d.f. F value Pr > F

4.76 (�1.40)b 2 135 47.90 <0.001

3.32 (�1.13)b 2 71 53.2 <0.001

2.42 (�0.27)b 2 301 10.52 <0.001

2.60 (�0.27)b 2 301 10.20 <0.001

47.7 (�7.5)b 2 301 21.64 <0.001

10.9 (�4.8)b 2 301 24.35 <0.001

31.3 (�10.1)b 2 301 5.28 0.005

Fig. 2. Densities of established woody species saplings in three structural vegetation

types.

Fig. 4. Browsing damage pattern of established F. excelsior saplings in three structural

vegetation types. Different letters indicate significant differences between vegetation

structures for a particular browsing damage type (post hoc Tukey’s tests).

J. Van Uytvanck et al. / Forest Ecology and Management 256 (2008) 106–113110

and the interaction between structural vegetation type and species(F = 3.48; p < 0.001) had significant effects on frequency: higherfrequencies for F. excelsior, Q. robur, A. glutinosa and S. alba in tallherb and scrub; only C. monogyna had higher frequencies ingrassland (Fig. 2). Sapling height and stem diameter of F. excelsior

were significantly higher in scrub patches, but differences in meanvalues were rather small (Table 3). Grazing period (in the range of4–11 years) after agricultural cessation had no effect on saplingdiameter and height ( p = 0.13 and p = 0.58, respectively), but therewas a significant interaction with vegetation type for diameter( p < 0.001) and a trend for height ( p = 0.07). The positive effect ofincreasing grazing period was most expressed in tall herb plots forboth diameter and height.

When investigating the effect of moisture conditions (only inthe floodplain study sites), overall sapling frequency depended onstructural vegetation type (F = 12.63; p = 0.02), moist category(F = 5.28; p < 0.01), species (F = 4.43; p = 0.05) and the interactionbetween structural vegetation type and species (F = 1.74; p = 0.03)and showed a clear pattern: higher frequencies in wet plots and intall herb (Fig. 3). A significant effect of species and the interactionbetween species and structural vegetation type was due to thedeviant establishment pattern of C. monogyna, which establishedsignificantly more in grassland than in tall herb vegetation. Nosignificant interactions between structural vegetation type andmoist ( p = 0.76) and between moist and species ( p = 0.15) were

Fig. 3. Densities of established woody species in moist and wet vegetation patches

in floodplains. Different letters indicate significantly different frequencies (post hoc

Tukey’s tests).

found. We found hoofprints in all wet plots and in 45% and 32% ofmoist grassland and moist tall herb plots, respectively.

3.2. Browsing damage

Overall, high numbers of established F. excelsior saplings werebrowsed (48–85%). In scrub patches, a significantly lowerpercentage of saplings showed general browsing damage, apicalshoot damage and reduced stem branching (Table 3; Fig. 4).Differences between grassland and tall herb patches were notsignificant.

4. Discussion

Palatable saplings were found in close association with scrub,tall herb vegetation and wet patches. Compared to dry grasslandpatches, scrub and tall herb patches must have provided protectionagainst browsing and wetter patches must have offered betterestablishment conditions.

4.1. Woody species assemblages

Sapling communities are good indicators of the dynamic trendsof woodlands, since they represent the result of recent dynamics ofa given site and indicate how forest expansion will evolve (Duarteet al., 2006). We expected in our study sites a colonisation ofgrazing tolerant species-like spiny shrubs (Archer et al., 1988),unpalatable saplings or a combination of protective scrub andpalatable saplings (Bakker et al., 2004). We mainly found palatablesaplings, established in tall herb and scrub patches.

In general, 4–11 years after the cessation of intensive grasslandsummer grazing, establishment of woody species was sparse, butnevertheless present in all sites. It is known that on nutrient richand productive sites, it takes 30–45 years before woody species onformer open vegetation types become dominant (Smit and Olff,1998). Closed grassland swards inhibit regeneration of woodyspecies (Harmer et al., 2001) and retard succession. Woody pioneerand wind dispersed species like Betula pendula and Salix caprea,which often dominate the early stages of abandoned arable land(Finegan, 1984) on heavy soils, occurred in negligible numbers. Incontrast, mainly animal and water dispersed species of mid-successional or mature forest stages were numerous. The mostfrequently found F. excelsior is, to a certain extent, also a winddispersed species, but it has rather heavy seeds. Walker et al.(2000) found contrasting establishment patterns for F. excelsior andQ. robur, the latter being widely dispersed by animals over a newlydeveloping woodland, the former being dispersed at short distancefrom seed trees. In our study sites, increasing distance to seed treeswas correlated with decreasing frequencies of both F. excelsior and

J. Van Uytvanck et al. / Forest Ecology and Management 256 (2008) 106–113 111

Q. robur, and also of C. monogyna. However, sapling numbersdecreased gradually with increasing distance to potential seedsource and no clear edge effects (i.e. saplings only found in thedirect vicinity of the seed trees in adjacent forest; Walker et al.,2000) were found. In floodplain sites, seeds of A. glutinosa probablywere mainly spread through inundations. In this study, we wereonly able to examine rather small areas, with a maximum distanceof 460 m between plot and potential seed source and 200 mbetween plot and forest edge. In these small areas, the presence ofseed sources, wind, inundations, plant–herbivore interactions andeven distance to seed sources seemed to be important factorsinfluencing establishment patterns of woody species.

4.2. Herbivory and vegetation structure

In early successional stages in a low productive environment,woody species establishment is limited by abiotic stress, ratherthan by herbivory or competition (Menge and Sutherland, 1987). Inthe productive former agricultural landscapes that we studied,top-down control by large herbivores, as predicted by theexploitation ecosystem theory (Oksanen et al., 1981), failed. vande Koppel et al. (1996) showed that in productive environments,high plant standing crop decreased foraging efficiency ofherbivores (hares, rabbits, geese), leading to the development ofdense vegetations. A comparable mechanism of decreased foragingefficiency may have occurred in tall herb patches in our study sites.However, the development of these patches was primarily due tothe reduction of grazing pressure after the cessation of agriculturaluse. Consequently, grazing activities structured the vegetation,leading to the development of short grazed patches, dense tall herbpatches with mainly unpalatable and spiny species, and scrub. Avarying herbivore pressure across such a mosaic with differentsuccessional stages, will affect the establishment of woody species.Our findings of increased establishment of palatable saplings in tallherb and scrub (mainly bramble, Rubus sp.) patches correspondwith the findings of Kuiters and Slim (2003) on woody establish-ment on former arable land grazed by horses. In a previous study(Van Uytvanck et al., 2008) high percentages of seedlings of Q.

robur and F. excelsior survived in tall herb patches of Juncus or Carex

species and these tree species were also able to grow out beyondthe browse line in bramble patches. These experimental findingsaccord to the observational results of this study, but thismechanism was now proven to work for other palatable speciesas well. However, one of the five common species, C. monogyna

(palatable but spiny) established mainly in open grassland. Thisconfirms findings of Benjamin et al. (2005), who stated thatselective browsing may even favour regeneration of spiny shrubsin intensively grazed pastures. Large herbivore grazing or mowingare able to reduce competition between saplings or less-preferredwoody saplings like C. monogyna and the surrounding vegetation(Archer, 1995; Lof et al., 2004; Vandenberghe et al., 2008).

Damage by herbivores to established saplings depended onthe different structural vegetation types. Fewer seedlings werebrowsed in scrub patches. This was most clearly expressed indamage to the apical shoot of the saplings. The mechanisms heremay be related to associational resistance (for scrub), to associa-tional palatability (for grassland, Olff et al., 1999) and to neighbourcontrast susceptibility (for tall herb, Bergvall and Leimar, 2005),meaning that scrub patches offer direct protection to saplingsthrough impenetrable and spiny branches, that in grasslandsaplings are eaten together with palatable grasses and that in tallherb (mainly unpalatable and hence less visited) palatable saplings,growing out above the canopy, may attract herbivores like cattle. Insuch a case, selective browsing by cattle in an unpalatable patchmay explain the damage pattern of palatable saplings and their

apparent inability to grow out beyond the browse line. Kuiters andSlim (2003) also found that established saplings in soft rush andheather patches could not escape from browsing.

However, we believe that direct regeneration in tall herbpatches, without a scrub stage, may be an operative short cut in thecyclic vegetation succession in low intensity grazed ecosystems asproposed by Olff et al. (1999). First, a few individual saplings in ourstudy sites had already grown out beyond the browse line. Ourdata cover only 4–11 years. So, it may be only a question of timebefore forest patches develop. Second, a temporal decrease orabsence of grazing pressure (managed or natural) may lead to asudden and fast growth of already established saplings in tall herbpatches (Van Uytvanck et al., 2008). Kuiters and Slim (2002) arguethat browse-sensitive woody species such as Quercus species needa window of opportunity to regenerate. As a consequence, weexpect that species that are less browse-sensitive (e.g. F. excelsior

and A. glutinosa have better regeneration capacities after browsingthan Quercus sp., moreover, A. glutinosa is less palatable) will playan important role in woodland development in grazed floodplains,in particular via regeneration in wet tall herb patches, which seemto provide favourable microsites for germination.

4.3. Trampling damage and microsites for germination

Gaps or openings in swards, created by trampling or uprooting,may be considered as microsites, allowing partial release fromcompetition and provide opportunities for establishment of woodyplants (Jurena and Archer, 2003; Vandenberghe et al., 2006).Mercier et al. (2001) and Staaland et al. (1998) showed thatimperfect soil drainage on wet soils slowed down the colonisationof woody species in ungrazed, abandoned agricultural land. In therather intensively grazed sites in this study however, the oppositehappened. Wet soils are more susceptible to soil damage bytrampling and the increase in suitable germination sites could havefavoured establishment. This establishment may logically happenwhen grazing disturbance is moderate, e.g. in poorly used orunpalatable vegetation types, like tall herb patches (whichdeveloped more numerous on wet than on dry soils). Increaseddesiccation of seedlings in gaps, a major cause of seedling mortalityin mountain wood-pastures (Vandenberghe et al., 2006) is unlikelyto happen in trampled zones in wet tall herb patches in floodplains.Nevertheless, moist conditions in disturbed soils are often a criticalconstraint for tree germination and establishment success maydiffer over time (Nilsson et al., 2002).

Recruitment of woody species in livestock grazed systems isoften episodic and disturbance-driven (Clarke, 2002). Distur-bances in floodplains like local erosion and sedimentation causedby inundation (Van Splunder et al., 1995), dung deposition (deVillalobos et al., 2005) and local enrichment may induce naturalrecruitment of woody species via tall herb and scrub patches,developing in grassland lawns (Olff et al., 1999; Vera, 2000). Ourdata suggest that woodland development in floodplains addi-tionally may start via microsite creation caused by largeherbivore trampling in wet tall herb patches, which subsequentlyoffer (at least temporal) protection to saplings via associationalresistance. We started germination experiments with simulateddisturbance regimes in floodplain grasslands to further explorethis mechanism.

4.4. Conclusions and implications for management

The development of new woodlands in low intensity-grazedgrasslands is strongly determined by vegetation structure. Thepattern of spiny scrub and unpalatable tall herb patchesdeveloping in such grasslands reflects future woodland patches,

J. Van Uytvanck et al. / Forest Ecology and Management 256 (2008) 106–113112

although a temporal decrease in grazing pressure or a temporalabsence of large herbivores may be needed for successfuloutgrowth of established woody species in tall herb. Moderatesoil disturbance, caused by trampling, probably is important tocreate sufficient microsites for germination in tall herb patches.However, local disturbances like flooding, availability of seeds,distance to seed trees and differences in soil moisture influence theregeneration pattern and these factors may interact with largeherbivore grazing. Longer time series of woody species develop-ment after the cessation of agricultural use are needed to confirmour findings.

Although there’s evidence that large herbivores increasebiodiversity in productive ecosystems (Proulx and Mazumder,1998; Bakker et al., 2006), the use of large herbivores for woodlandexpansion at productive sites on former agricultural land is arelatively unknown tool in conservation management. Moreexperience and a process-based approach are needed to fine-tunegrazing management strategies and to quantify multiple interac-tion processes of inhibition and facilitation (Sankey et al., 2006). Inthis study on nutrient-rich soils, we found that a grazing pressurebelow 0.5–0.6 AU ha�1 year�1 allows rather fast tree establish-ment when protective vegetation structures are available. Whenspiny scrub develops in the grasslands, an even higher grazingpressure will not prevent woodland regeneration. However, in theshort term (20–30 years), woodland expansion steered by grazingwill only realise open woodland types, but as goal in itself or as apathway to more closed forest types, these open woodlands mayrefer to more natural (Rotherham, 2007) or attractive park-likelandscapes with a high conservation value (Piek, 1998).

Acknowledgements

We thank Johnny Cornelis, Godfried Merlevede, Carlos D’Hae-seleer, Guido Tack and Dirk Van den Berghe for granting uspermission to conduct fieldwork in the study sites. We thankCharlotte Vandenberghe and Dirk Maes for their useful commentson the manuscript.

References

Akaike, H., 1987. Factor analysis and AIC. Psychometrika 52, 317–332.Archer, S., 1995. Harry Stobbs Memorial Lecture, 1993.Herbivore mediation of

grass–woody plant interactions. Trop. Grasslands 29, 218–235.Archer, S., Scifres, C., Bassham, C.R., Maggio, R., 1988. Autogenic succession in a sub-

tropical savanna—conversion of grassland to thorn woodland. Ecol. Monogr. 58,111–127.

Bakker, E.S., Olff, H., Vandenberghe, C., De Maeyer, K., Smit, R., Gleichman, J.M., Vera,F.W.M., 2004. Ecological anachronisms in the recruitment of temperate light-demanding tree species in wooded pastures. J. Appl. Ecol. 41, 571–582.

Bakker, E.S., Ritchie, M.E., Olff, H., Milchunas, D.G., Knops, J.M.H., 2006. Herbivoreimpact on grassland plant diversity depends on habitat productivity andherbivore size. Ecol. Lett. 9, 780–788.

Bakker, J.P., Londo, G., 1998. Grazing for conservation management in historicalperspective. In: Wallis DeVries, M., Bakker, J.P., van Wieren, S.E. (Eds.), Grazingand Conservation Management. Kluwer Academic Publishers, Dordrecht, pp.23–54.

Benjamin, K., Domon, G., Bouchard, A., 2005. Vegetation composition and succes-sion of abandoned farmland: effects of ecological, historical and spatial factors.Landscape Ecol. 20, 627–647.

Bergvall, U.A., Leimar, O., 2005. Plant secondary compounds and the frequency offood types affect food choice by mammalian herbivores. Ecology 86, 2450–2460.

Bokdam, J., 2001. Effects of browsing and grazing on cyclic succession in nutrient-limited ecosystems. J. Veg. Sci. 12, 875–886.

Bokdam, J., Gleichman, J.M., 2000. Effects of grazing by free-ranging cattle onvegetation dynamics in a continental north-west European heathland. J. Appl.Ecol. 37, 415–431.

Callaway, R.M., 1992. Effect of Shrubs on Recruitment of Quercus douglasii andQuercus lobata in California. Ecology 73, 2118–2128.

Clarke, P.J., 2002. Experiments on tree and shrub establishment in temperate grassywoodlands: seedling survival. Aust. Ecol. 27, 606–615.

Darabant, A., Rai, P.B., Tenzin, K., Roder, W., Gratzer, G., 2007. Cattle grazingfacilitates tree regeneration in a conifer forest with palatable bamboo unders-tory. Forest Ecol. Manage. 252, 73–83.

Davis, M.A., Wrage, K.J., Reich, P.B., Tjoelker, M.G., Schaeffer, T., Muermann, C., 1999.Survival, growth, and photosynthesis of tree seedlings competing with herbac-eous vegetation along a water-light-nitrogen gradient. Plant Ecol. 145, 341–350.

de Villalobos, A.E., Pelaez, D.V., Elia, O.R., 2005. Factors related to establishment ofProsopis caldenia Burk. Seedlings in central rangelands of Argentina. Acta Oecol.27, 99–106.

Duarte, L.D.S., Machado, R.E., Hartz, S.M., Pillar, V.D., 2006. What saplings cantell us about forest expansion over natural grasslands. J. Veg. Sci. 17, 799–808.

Ellenberg, H., Weber, H.E., Dull, R., Wirth, V., Werner, W., Paulißen, D., 1992.Zeigerwerte von Pflanzen in Mitteleuropa. Scr. Geobot. 18.

Finegan, B., 1984. Forest succession. Nature 312, 109–114.Harmer, R., Peterken, G., Kerr, G., Poulton, P., 2001. Vegetation changes during 100

years of development of two secondary woodlands on abandoned arable land.Biol. Conserv. 101, 291–304.

Hodder, K.H., Bullock, J.M., Buckland, B.C., Kirby, K.J., 2005. Large herbivores in thewildwood and in modern naturalistic grazing systems. English Nature ResearchReport No. 648, Peterborough.

Holl, K., Smith, M., 2007. Scottish upland forests: history lessons for the future.Forest Ecol. Manage. 249, 45–53.

Huntly, N.J., 1991. Herbivores and the dynamics of communities and ecosystems.Ann. Rev. Ecol. Syst. 22, 477–503.

Jurena, P.N., Archer, S., 2003. Woody plant establishment and spatial heterogeneityin grasslands. Ecology 84, 907–919.

Kenward, M.G., Roger, J.H., 1997. Small sample inference for fixed effects fromrestricted maximum likelihood. Biometrics 53, 977–983.

Kirby, K.J., 2003. What might a British forest-landscape driven by large herbivoreslook like. English Nature Research Report No. 530, Peterborough.

Kirby, K.J., Mitchell, F.J., Hester, A.J., 1994. A role for large herbivores (deer anddomestic stock) in nature conservation management in British semi-naturalwoods. Arboric. J 18, 381–399.

Kramer, K., Bruinderink, G.W.T.A., Prins, H.H.T., 2006. Spatial interactionsbetween ungulate herbivory and forest management. Forest Ecol. Manage.226, 238–247.

Kuiters, A.T., Slim, P.A., 2002. Regeneration of mixed deciduous forest in a Dutchforest-heathland, following a reduction of ungulate densities. Biol. Conserv.105, 65–74.

Kuiters, A.T., Slim, P.A., 2003. Tree colonisation of abandoned arable land after 27years of horse-grazing: the role of bramble as a facilitator of oak woodregeneration. Forest Ecol. Manage. 181, 239–251.

Linares, A.M., 2007. Forest planning and traditional knowledge in collective wood-lands of Spain: the dehesa system. Forest Ecol. Manage. 249, 71–79.

Lof, M., Thomsen, A., Madsen, P., 2004. Sowing and transplanting of broadleaves(Fagus sylvatica L., Quercus robur L., Prunus avium L. and Crataegus monogynaJacq.) for afforestation of farmland. Forest Ecol. Manage. 188, 113–123.

Londo, G., 1984. The decimal scale for releves of permanent quadrats. In: Knapp, R.(Ed.), Handbook of Vegetation Science 4. Sampling Methods and Taxon Analysisin Vegetation Science. Dr. W. Junk Publishers, The Hague, pp. 45–50.

Mazia, C.N., Chaneton, E.J., Ghersa, C.M., Leon, R.J.C., 2001. Limits to tree speciesinvasion in pampean grassland and forest plant communities. Oecologia 128,594–602.

Menge, B.A., Sutherland, J.P., 1987. Community regulation—variation in distur-bance, competition, and predation in relation to environmental stress andrecruitment. Am. Nat. 130, 730–757.

Mercier, C., Brison, J., Bouchard, A., 2001. Demographic analysis of tree colonizationin a 20-year-old right-of-way. Environ. Manage. 28, 777–787.

Moser, B., Schutz, M., Hindenlang, K.E., 2006. Importance of alternative foodresources for browsing by roe deer on deciduous trees: the role of foodavailability and species quality. Forest Ecol. Manage. 226, 248–255.

Nilsson, U., Gemmel, P., Johansson, U., Karlsson, M., Welander, T., 2002. Naturalregeneration of Norway spruce, Scots pine and birch under Norway spruceshelterwoods of varying densities on a mesic-dry site in southern Sweden.Forest Ecol. Manage. 161, 133–145.

Oksanen, L., Fretwell, S.D., Arruda, J., Niemela, P., 1981. Exploitation ecosystems ingradients of primary productivity. Am. Nat. 118, 240–261.

Olff, H., Vera, F.W.M., Bokdam, J., Bakker, E.S., Gleichman, J.M., De Maeyer, K., Smit,R., 1999. Shifting mosaics in grazed woodlands driven by the alternation ofplant facilitation and competition. Plant Biol. 1, 127–137.

Ostfeld, R.S., Canham, C.D., 1993. Effects of meadow vole population-density on treeseedling survival in old fields. Ecology 74, 1792–1801.

Pepin, D., Renaud, P.C., Boscardin, Y., Goulard, A., Mallet, C., Anglard, F., Ballon, P.,2006. Relative impact of browsing by red deer on mixed coniferous and broad-leaved seedlings—an enclosure-based experiment. Forest Ecol. Manage. 222,302–313.

Piek, H., 1998. The practical use of grazing in nature reserves in the Netherlands. In:Wallis DeVries, M.,Bakker, J.P., van Wieren, S.E. (Eds.),Grazing and ConservationManagement. Kluwer Academic Publishers, Dordrecht, pp. 253–273.

Posada, J.M., Aide, T.M., Cavelier, J., 2000. Cattle and weedy shrubs as restorationtools of tropical montane rainforest. Restor. Ecol. 8, 370–379.

Proulx, M., Mazumder, A., 1998. Reversal of grazing impact on plant species richnessin nutrient-poor vs. nutrient-rich ecosystems. Ecology 79, 2581–2592.

J. Van Uytvanck et al. / Forest Ecology and Management 256 (2008) 106–113 113

Putman, R.J., 1986. Grazing in temperate ecosystems. Large herbivores and theecology of the New Forest. Croom Helm, Beckenham.

Rotherham, I.D., 2007. The implications of perceptions and cultural knowledge lossfor the management of wooded landscapes: a UK case-study. Forest Ecol.Manage. 249, 100–115.

Rousset, O., Lepart, J., 2000. Positive and negative interactions at different life stagesof a colonizing species (Quercus humilis). J. Ecol. 88, 401–412.

Sankey, T.T., Montagne, C., Graumlich, L., Lawrence, R., Nielsen, J., 2006. Twentiethcentury forest-grassland ecotone shift in Montana under differing livestockgrazing pressure. Forest Ecol. Manage. 234, 282–292.

Setterfield, S.A., 2002. Seedling establishment in an Australian tropical savanna:effects of seed supply, soil disturbance and fire. J. Appl. Ecol. 39, 949–959.

Smit, C., Beguin, D., Buttler, A., Muller-Scharer, H., 2005. Safe sites for tree regen-eration in wooded pastures: a case of associational resistance? J. Veg. Sci. 16,209–214.

Smit, C., Gusberti, M., Muller-Scharer, H., 2006. Safe for saplings; safe for seeds?Forest Ecol. Manage. 237, 471–477.

Smit, R., Olff, H., 1998. Woody species colonisation in relation to habitat produc-tivity. Plant Ecol. 139, 203–209.

Staaland, H., Holand, O., Nelleman, C., Smith, M., 1998. Time scale for forestregrowth: abandoned grazing and agricultural areas in southern Norway.Ambio 27, 456–460.

van de Koppel, J., Huisman, J., van der Wal, R., Olff, H., 1996. Patterns of herbivoryalong a productivity gradient: an emperical and theoretical investigation.Ecology 77, 736–745.

Van Splunder, I., Coops, H., Voesenek, L.A.C., Blom, C.W.P., 1995. Establishmentof alluvial forest species in floodplains: the role of dispersal timing,germination characteristics and water level fluctuations. Acta Bot. Neerl.44, 269–278.

Van Uytvanck, J., Maes, D., Vandenhaute, D., Hoffmann, M., 2008. Restoration ofwoodpasture on former agricultural land: the importance of safe sites and timegaps before grazing for tree seedlings. Biol. Conserv. 141, 78–88.

Vandenberghe, C., Frelechoux, F., Gadallah, F., Buttler, A., 2006. Competitive effectsof herbaceous vegetation on tree seedling emergence, growth and survival:Does gap size matter? J. Veg. Sci. 17, 481–488.

Vandenberghe, C., Frelechoux, F., Moravie, M.A., Gadallah, F., Buttler, A., 2007. Short-term effects of cattle browsing on tree sapling growth in mountain woodedpastures. Plant Ecol. 188, 253–264.

Vandenberghe, C., Frelechoux, F., Buttler, A., 2008. The influence of competitionfrom herbaceous vegetation and shade on simulated browsing tolerance ofconiferous and deciduous saplings. Oikos 117, 415–423.

Vera, F.W.M., 2000. Grazing ecology and Forest History. CABI Publishing, Oxon, UK.Vieira, I.C.G., Uhl, C., Nepstad, D., 1994. The role of the shrub Cordia multispicata

Cham as a succession facilitator in an abandoned pasture, paragominas. Ama-zonia Vegetatio 115, 91–99.

Walker, K.J., Sparks, T.H., Swetnam, R.D., Boatman, N.D., Clay, D.V., Goodman, A.,Marrs, R.H., Marshall, E.J.P., Newman, J.R., Putwain, P.D., Pywell, R.F., 2000. Thecolonisation of tree and shrub species within a self-sown woodland: the MonksWood Wilderness. Aspects Appl. Biol. 58, 337–344.

WallisDeVries, M.F., Bakker, J.P., van Wieren, S.E., 1998. Grazing and ConservationManagement. Kluwer Academic Press, Dordrecht.

WallisDeVries, M.F., 1998. Habitat quality and the performance of large herbivores.In: WallisDeVries, M.F., Bakker, J.P., van Wieren, S.E. (Eds.), Grazing and Con-servation Management. Kluwer Academic Publishers, Dordrecht, pp. 275–320.

Worrell, A.M., Peterken, G.F., Scott, A., Pryor, S., Taylor, K., Knightbridge, R., Brown,N., 2002. New wildwoods: developing the role of large-scale new native wood-lands in the uplands. Land Use Policy Group, Joint Nature Conservation Com-mittee, Peterborough.