Wassenaar van Aarde 2001
Short-term responses of rehabilitating coastal dune
forest ground vegetation to livestock grazing
Theo D.Wassenaar and Rudi J. van Aarde
Conservation Ecology Research Unit, Department of Zoology & Entomology, University of Pretoria, Pretoria, South Africa
¤
rage dependaient apparemment des changements de
Abstract
¤ ¤ ' ¤
vegetation intrinseques speci¢ques des sites, et certains
We investigated the responses of the ground vegetation in indices laissent penser qu’il y a des interactions entre les
a17-year-old coastal dune forest plant community to four “
chutes de pluies et le taux de paturage. Le paturage qui“
levels of experimentally applied livestock grazing (three avait certains e¡ets apparents, mais non signi¢catifs sur
grazing levels and one ungrazed control) from May ¤
la composition speci¢que des plantes, a¡ectait signi¢cati-
1994 to March 1996. The e¡ects of grazing were appar- '
vement la richesse en especes avec le temps, et augmen-
ently subordinate to site-speci¢c intrinsic vegetation ¤
tait signi¢cativement l’etendue de la richesse en especes '
change and there were some indications that rainfall ¤ ¤ ¤ ¤
vegetales et la couverture vegetale ainsi que l’abondance
interacted with grazing level. Grazing had some apparent ' ¤ ¤
relative et le nombre d’especes vegetales de forme erigee. ¤ ¤
but no signi¢cant e¡ects on plant species composition, ¤ ¤
Le couvert vegetal changeait signi¢cativement au cours
signi¢cantly a¡ected plant species richness over time, “ ¤
du temps, quel que soit le paturage. Nos resultats indi-
and signi¢cantly increased the range of species richness ¤
quent deux mecanismes importants, faciles a mesurer, '
and vegetation cover values as well as the relative abun- “
pour la gestion conservatoire des forets de dunes cotieres “ '
dance and numbers of plant species with erect growth ^ l’ interaction du type de perturbation avec la forme de
forms. Vegetation cover changed signi¢cantly over time, la croissance des plantes, et l’accroissement de la variation
independently of grazing. Our results point to two impor- des variables structurelles de la communaute vegetale ¤ ¤ ¤
tant, easily measured mechanisms for the conservation ¤ “
en cas de perturbation. Ces mecanismes, meme s’ ils peu-
management of coastal dune forests ^ the interaction of vent avoir une large application et une valeur predictive, ¤
disturbance type with plant growth form and the ¤ ¤
n’ont pas encore ete convenablement etudies. ¤ ¤
increase of variation in community structural variables
under disturbance. These mechanisms, although they
potentially have wide application and predictive power, Introduction
have not been studied adequately.
Two main factors are relevant to grazing in African
Key words: disturbance, growth form, species richness, coastal dune forests. The ¢rst of these is that large herbi-
variability vores can in£uence the organization of almost every plant
community at many di¡erent scales and on all hierarch-
ical levels (Glenn-Lewin & van der Maarel, 1992; van de
¤
Resume¤
Koppel, Rietkerk & Weissing, 1997). Although coastal
¤ ¤ ¤ ¤ ¤
Nous avons etudie les reponses de la vegetation terrestre dune forests are generally regarded as a resilient habitat
' “ '
d’une plantation forestiere de17 ans sur une dune cotiere type with a fairly predictable development from grassland
' ¤ “
a quatre niveaux experimentaux de paturage par du to mature dune forest (Weisser, 1978; Weisser & Muller,
¤ “ “
betail (trois niveaux de paturage et un controle non 1983), current insights into community development
“ ¤ “
pature) entre mai 1994 et mars 1996. Les e¡ets du patu- stress the existence of multiple possible stable states
(Sutherland, 1990; see also von Maltitz, van Wyk &
Correspondence:Theo D.Wassenaar, Conservation Ecology Research
Everard, 1990).
Unit, Department of Zoology & Entomology, University of Secondly, few vertebrate herbivores inhabit the coastal
Pretoria, Pretoria 0002, South Africa. Tel: þ27 12 4202561; Fax: dune forests of the southeast coast of Africa (Skinner &
þ27 12 4204523; E-mail: tdwassenaar@zoology.up.ac.za Smithers, 1990). Of the 26 mammal species listed by
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339 329
330 Theo D.Wassenaar and Rudi J. van Aarde
Ferreira (1993) for the coastal dune forests north of are uniform throughout the area, with little horizontal
Richards Bay, KwaZulu-Natal, South Africa, only three di¡erentiation (Avis, 1992). Most rain falls from January
herbivores (the bushpig Potamochoerus porcus [Linnaeus to March (peak in February; annual mean 1292 mm).
1758], the red duiker Cephalophus natalensis [Smith Extended droughts are uncommon and approximately
1834] and the bushbuck Tragelaphus scriptus [Pallas 30% of annual precipitation occurs in the winter. Daily
1766]) are large enough to cause signi¢cant disturbance maximum temperatures range from 22.6 8C to 30.0 8C
to vegetation. Both red duiker and bushbuck occur in and minimum temperatures from 10.0 8C to 20.6 8C in
relatively low numbers and are highly selective in their June and January, respectively (Ferreira, 1996).
feeding habits, while the distribution of the bushpig is The regenerating forest consists of an Acacia karroo
typically patchy and they are not common anywhere (Hayne) woodland 9^12 m high, with secondary dune
(Skinner & Smithers, 1990). Livestock grazing, which forest tree species colonizing. Compared to a mature
is therefore likely to be an ‘unnatural’ type of distur- forest there is relatively little vertical strati¢cation. The
bance for coastal dune forests, could thus have far- herb layer (%1m or less) is dominated by a number of sto-
reaching implications, not only for the conservation of loniferous grasses, creepers and decumbent herbs. Van
coastal dune forests, but also for their post-disturbance Aarde et al. (1996) provide a detailed description of the
development. study area and vegetation.
The mining company Richards Bay Minerals (RBM)
has been mining and subsequently rehabilitating a strip
Experimental design and data recording
of coastal sand dunes on the northeast coast of South
Africa since July 1977 (Camp, 1990; van Aarde et al., Grazing was applied to 0.125 ha fenced paddocks,
1996). The mining lease and rehabilitating areas are situ- arranged in four blocks with a Control, Low, Medium
ated in a regional development node with large in£uxes and High grazing level in each (i.e. four replicates per
of people (23% increase from 1980 to 1991) and their grazing level). Blocks were randomly placed within the
associated domestic livestock. This has led to illegitimate study site and grazing levels randomly assigned to pad-
grazing of regenerating forests, and concerns over its docks within each block. In this instance the term ‘graz-
possible e¡ect on the rehabilitation programme. ing’ includes all other forms of disturbance by cattle to
We expected livestock grazing to signi¢cantly a¡ect the vegetation, i.e. trampling, defecating, etc. Five grazing
the restoration of coastal dune forest in the short term cycles were applied over 16 months (Fig.1), the period
through its e¡ect on ground vegetation community between grazing cycles varying from 90 to 150 days. A
structure. The present study therefore examines the grazing cycle consisted of 2 days’grazing in the low graz-
short-term e¡ects of cattle on ground vegetation (i) ing level paddocks, 4 days in the medium level paddocks
species richness and vegetation cover (ii) species compo- and 8 days in the high level paddocks. For each grazing
sition, and (iii) numbers and abundances of species in cycle, eight heifers (200^300 kg) were sorted into four
two plant growth form groups in a17-year-old rehabilitat- pairs (so that mean pair-weight % mean group-weight)
ing coastal dune forest. and each pair randomly assigned to one of the four repli-
cates in a grazing level every day.
Species presence and vegetation cover were recorded at
Materials and methods
six1m2 quadrats per paddock prior to each grazing cycle.
Vegetation cover was recorded with a point-bridge
Study area
adapted from Barbour, Burk & Pitts, (1987). Plant species
The study was conducted in the oldest regenerating were assigned to one of two categories: (i) decumbent
part of an area of mined dunes north of Richards Bay, (all species, excluding lianas, with mostly horizontal
KwaZulu-Natal, South Africa (288430 S and 328120 E), vegetative growth, i.e. creepers and stoloniferous
which has been under ecological rehabilitation since1977 grasses), and (ii) erect (large and small annual and peren-
(Stand 1 in Ferreira & van Aarde, 1996). The area is char- nial forbs, tussock grasses, juvenile trees and shrubs).
acterized by longitudinal sand dunes, lying parallel to Rainfall was measured using standard rain gauges at
the coastline, and rising to an elevation of 40^90 m above two points situated halfway between the four blocks of
sea level. Soils (¢ne to medium-grained aeolian sands) grazing paddocks.
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
Ground vegetation response to livestock grazing 331
Fig 1 Mean (n ¼ 3) daily rainfall in the
rehabilitating area from January 1994 to
August 1996 and schedule of plant surveys
and grazing cycles. The dotted line
represents the mean daily rainfall over a
10-year period, including the period of our
study. S1^S6 represent plant surveys and
G1^G5 grazing applications
The data were used to calculate species density (SD or Kruskal^Wallis test where applicable, for the overall
species/m2), vegetation cover below 10 cm (VC), relative e¡ect of cattle on VC (arcsine transformed) and SD with
abundance of growth form (Erect/Decumbent[Ab]) and multivariate analysis of variance for repeated measures
number of species per growth form (Erect/Decum- (manovar; Potvin, Lechowicz & Tardif, 1990), and for
bent[Sp]), as well as species composition. For species comparisons between grazing levels we used Tukey’s
composition analysis we de¢ned three Bray^Curtis simi- honestly signi¢cant di¡erence test (T-HSD; Sokal & Rohlf,
larityclasses (Ludwig & Reynolds,1988): (i) temporal simi- 1995). We used least squares linear regression (LSLR) to
larity (Tsim): similarity of a paddock to itself over time; test for the e¡ect of cattle on the range inVC and SD values
(ii) close spatial similarity (CSsim): similarity between as well as on all growth form variables and similarity
paddocks in the same block, but from di¡erent grazing changes over time (independent variable: days since start
levels; and (iii) far spatial similarity (FSsim): similarity of ¢rst survey). A signi¢cant di¡erence between the slopes
between paddocks in the same grazing level, but from dif- of control and any of the other grazing levels indicated a
ferent blocks. grazing e¡ect, and a slope signi¢cantly di¡erent from zero
To determine whether the microhabitats at the quad- for any of the grazing levels indicated a time e¡ect.Analy-
rats were similar, we recorded the structure of the canopy sis of similarity (anosim; Clarke, 1993) was used to test
vegetation surrounding each quadrat (name, height, for di¡erences in species composition between grazing
stem diameter and number of stems of closest species levels, paddocks and sequential surveys.
of canopy tree in each of four quarters surrounding
the quadrat), percentage canopy cover at each quadrat
Results
(measured on a colour negative of the underside of
the canopy with a Quantimet 520 image analyser;
Microhabitat
Cambridge Instruments, London), and the light inten-
sity (lux) on the forest £oor (Measuring Instruments There were no signi¢cant di¡erences between grazing
Technologies, Pretoria). levels in canopy tree structure and density (Kruskal^
Wallis: all H-values 3.22, all P-values >0.05), incandes-
cent light (overall mean Æ SE ¼ 21.40 Æ1.43 lux; anova:
Statistical analysis
F ¼1.31, P ¼ 0.20) and percentage canopy cover (overall
We tested for di¡erences in microhabitat between mean ÆSE ¼ 84.9 Æ 0.78%; anova: F ¼1.18, P ¼ 0.36).
paddocks with analysis of variance (anova) and the Only one species of canopy tree was recorded, namely
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
332 Theo D.Wassenaar and Rudi J. van Aarde
A. karroo.We therefore assumed that di¡erences in micro- time^grazing level interaction (linear) and time e¡ect
climatic conditions at each quadrat were small enough (quadratic) were signi¢cant (manovar; overall time Â
for the herb layer vegetation not to have been a¡ected by level: Roy’s GR ¼ 0.34, F ¼ 6.20, P < 0.001, overall time:
it. Furthermore, although precipitationwas not measured Roy’s GR ¼1.94, F ¼ 34.20, P < 0.001; linear (time Â
at all survey points, we assumed that paddocks were close level): F ¼ 4.26, P ¼ 0.01; quadratic (time): F ¼10.48,
enough to each other not to have been in£uenced by spa- P ¼ 0.002). Control di¡ered from Low and High over all
tial variation in rainfall patterns. Rainfall followed a sea- surveys (T-HSD, msd(0.05,20) ¼ 1.87). During survey 2 Con-
sonal pattern during the study period, although daily trol and High di¡ered signi¢cantly (msd(0.05,12) ¼ 1.803),
rainfall was below the 10-year average for a considerable during survey 4 Control di¡ered from both Low and High
period before the third survey (Fig.1). It increased some- (msd(0.05,12) ¼ 2.35), during survey 5 (msd(0.05,12) ¼ 2.37)
what after the second survey and at the time of the fourth and survey 6 (msd(0.05,12) ¼ 1.97) Control di¡ered from
survey increased above the 10-year average. Low, Medium and High.
Increased grazing level apparently increased the
range of SD values at each survey over time; in con-
Plant species density (SD)
trast the range decreased in Control (Fig.3). The slope
Mean SD decreased over time in Control, while it stayed at of the relationship between time and range of values
the same level in Low, Medium and High (Fig. 2). The at each survey was signi¢cant for Control (LSLR;
Fig 2 The number of species per1m2 quadrat (species density or SD ^ see text) over 6 plant surveys in (a) Control (b) Low (c) Medium and (d)
High. Symbols represent different paddocks. Lines were fitted with least squares linear regression on the mean values per survey
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
Ground vegetation response to livestock grazing 333
Roy’s GR ¼ 3.96, F ¼ 8.91, P ¼ 0.003, quadratic (time):
F ¼ 26.81, P < 0.001). The range of values per survey
increased with grazing level (Fig.3), but none of the slopes
di¡ered signi¢cantly from zero (slopes from À0.004 to
0.06, all P > 0.05, r2 from 0.006 to 0.30) or from each other
(F ¼ 0.44; P > 0.05).
Growth form
Decumbent[Ab] values for Control remained higher than
Low, Medium and High throughout and apparently
increased more over time (Fig.5a). Slopes were not signif-
icantly di¡erent (F ¼1.07, P ¼ 0.39). In contrast Erect[Ab]
values for Control remained lower than the other grazing
levels throughout (Fig.5b), the slope for Control di¡ering
signi¢cantly from High (Control: y ¼15.69 À 0.008x;
P ¼ 0.15, r2 ¼0.44; High: y ¼ 23.84 þ 0.009x; P ¼ 0.06,
r2 ¼0.63; F ¼ 3.90, P ¼ 0.03). The slope of the relationship
between time and Erect[Sp] was signi¢cant only for
Low and High (Low: y ¼18.53 þ0.009x; P ¼ 0.04,
r2 ¼0.70; High: y ¼15.29 þ 0.01x; P ¼ 0.04, r2 ¼0.70).
Both Erect and Decumbent species were apparently lost
faster from Control (Fig.5c,d). Di¡erences were more pro-
nounced in Erect plants however (Fig.5d), where the
slope for Control di¡ered signi¢cantly from High
(F ¼ 5.39, P ¼ 0.009).
Similarity/species composition
Fig 3 The relationship between time (days since start of first Species composition of Control paddocks apparently
plant survey) and range in values (difference between changed faster over time (similarity to initial species com-
minimum and maximum for each survey) of (a) SD and position, or Tsim, declined faster; Fig.6a) and stayed more
(b) VC. Least squares linear regression models are described
similar to each other than the other grazing levels (mean
in the text
similarity within grazing level, or FSsim, was higher;
Fig.6b). For Tsim, slopes were signi¢cant in Control
(y ¼ 94.06 À3.57x, P ¼ 0.05; r2 ¼0.79, P ¼ 0.05) and
y ¼ 3.70 À 0.004x, P ¼ 0.03, r2 ¼0.74), but not for Low Low (y ¼ 86.66 À1.51x, P ¼ 0.01; r2 ¼0.93, P ¼ 0.01), but
to High (slopes from 0.001 to 0.006, all P > 0.05; r2 there were no overall signi¢cant di¡erences between
from 0.08 to 0.53). Slopes were signi¢cantly di¡erent the slopes (F ¼1.22, P ¼ 0.35). Slopes for FSsim (Fig.6b)
overall (F ¼ 3.47; P ¼ 0.04), with Control di¡ering from were not signi¢cantly di¡erent from zero or from each
Low. other (range ¼ À1.56 to À0.11, all P-values >0.05; all
r2-values 0.40; F ¼ 0.35, P ¼ 0.79). Similarity within
blocks (CSsim) (Fig.6c) in two of the blocks decreased
Vegetation cover (VC)
signi¢cantly over time (Block 2: y ¼ 66.45 À0.01x, P ¼
Mean VC apparently increased over time in Control, but 0.009, r2 ¼0.18; Block 4: y ¼ 71.16 À 0.02x, P ¼ 0.007,
decreased slightly in High (Fig. 4). There was a signi¢cant r2 ¼0.20), but the slopes for the di¡erent blocks did not
quadratic time e¡ect, but no signi¢cant time^grazing di¡er signi¢cantly (F ¼1.26, P ¼ 0.29). Mean Tsim (over
level interaction or grazing e¡ect (manovar; overall time: all grazing levels) stayed higher than both mean CSsim
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
334 Theo D.Wassenaar and Rudi J. van Aarde
Fig 4 The percentage ground cover (VC ^ see text) (arcsine transformed) over 5 plant surveys in (a) Control (b) Low (c) Medium and (d) High.
Symbols represent different paddocks. Lines were fitted with least squares linear regression on the mean values per survey
and mean FSsim, and CSsim stayed higher than FSsim
Discussion
throughout the study period (Fig.6d).
There were no signi¢cant di¡erences between either
Site and climate
the grazing levels or the di¡erent surveys in species
composition (two-way crossed anosim; R ¼ À0.004, Our study was conducted over a relatively short time-
P ¼ 0.52 and R ¼ À0.09, P ¼ 0.98, respectively). There scale, with an emphasis on the e¡ects of grazing on
were also no signi¢cant di¡erences between the grazing ground vegetation community variables. Results suggest
levels within any of the surveys (one-way anosim; global that grazing had some e¡ects on plant community struc-
R-values range À0.11 to 0.11, all P-values >0.05). Blocks ture, although these were not always signi¢cant and
of paddocks on the other hand di¡ered signi¢cantly over appeared to be subordinate to both intrinsic vegetation
all (global R ! 0.45, P 0.001 in all surveys). In these change (i.e. site e¡ects) and rainfall e¡ects. Site e¡ects
comparisons the pair of blocks situated closer to each can clearly be seen in the signi¢cant di¡erences in species
other were never signi¢cantly di¡erent (R 0.46, composition between the blocks of paddocks at all times,
P > 0.05 in all cases), while the other blocks, situated while at the same time none of the grazing levels di¡ered.
further away from each other, were always di¡erent Furthermore, if grazing had an e¡ect over and above site
(R ! 0.43, P < 0.05 in all cases). and environment, there would have been a change in
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
Ground vegetation response to livestock grazing 335
Fig 5 The relationship between time (days since start of first plant survey) and (a) relative abundance of decumbent plants, (b) relative
abundances of erect plants, (c) number of decumbent species, and (d) number of erect species; in four grazing levels. Least squares linear
regression models are described in the text
the site-dependent theoretical ranking of temporal (Tsim; relatively dry preceding period (Fig.1). Furthermore,
highest) ^ close spatial (CSsim) ^ and far spatial (FSsim) there were some indications of a possible interaction
similarity over time, and/or a di¡erence between control between grazing level and rainfall e¡ect, because the
and treatment in this pattern. This did not happen, nor magnitude of the drop in both SD and VC during survey
did paddocks within the same grazing level and paddocks 3 (Figs. 2 and 4) was apparently dependent on the level
within the same block become more (i.e. increased FSsim) of grazing. It is well known that rainfall (and climate in
and less (i.e. decreased CSsim) similar to each other, general) can have a strong in£uence on plant community
respectively, as expected. Also, although CSsim did variables, overshadowing other abiotic and biotic distur-
decrease in two of the blocks, FSsim, contrary to expecta- bances (e.g. Walker & Knoop, 1987; O’Connor, 1991; Peel,
tion, actually decreased. Grazing could therefore not Grossman & van Rooyen, 1991).
overcome the high Tsim and CSsim, or the low FSsim,
which suggests that the ground vegetation community
Grazing
of these coastal dunes is relatively resistant to grazing dis-
turbance in the short term. In contrast to our study, earlier studies noted decreases
Rainfall e¡ect was not speci¢cally investigated, but in vegetation cover under grazing (e.g. Belsky &
the drop in almost all variables and grazing levels during Blumenthal, 1997, cites several examples in western US
survey 3 followed on and was probably the result of a forests). Presumably our result was due to a combination
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
336 Theo D.Wassenaar and Rudi J. van Aarde
Fig 6 The relationship between time (days since start of first plant survey) and (a) temporal similarity (Tsim ¼ similarity to initial
species composition), (b) far spatial similarity (FSsim ¼ paddocks in the same grazing level but in different blocks), (c) close spatial similarity
(CSsim ¼ paddocks in the same block but in different grazing levels), and (d) a summary of changes in mean similarity over all grazing
levels (for Tsim and FSsim) and over all blocks (for CSsim); in four grazing levels. Least squares linear regression models are described in
the text
of (possibly) too light a grazing pressure and relatively varying duration) in fact report higher species richness
long intervals between successive surveys, providing under grazing (Gibson et al., 1987 [moist grassland],
the fast-growing herbaceous vegetation su⁄cient time O’Connor,1991 [savanna], Pandey & Singh,1992 [season-
to recover. However, grazing had a signi¢cant level- ally dry savanna], Bullock et al., 1994 [grassland], Ash &
dependent e¡ect on SD (species density, an index of rich- McIvor, 1998 [tall grassland]), usually as a result of
ness), which evidently became progressively more intense increases in annual grasses and forbs. A distinction
with time, mostly because of a progressive decrease should however be made between grazing e¡ect and
in mean SD in the ungrazed Control. Grazing itself exclusion e¡ect ^ in those studies carried out in a back-
apparently maintained mean species richness (Smith & ground of free-range grazing (as our own was), species
Rushton, 1994; Wilson, 1994), in our case probably as richness does not increase under grazing, it rather
a result of an increase in the number of erect forb species decreases where grazing has been excluded (e.g. Pandey
as they were released from competition with the domi- & Singh,1992; Smith & Rushton,1994). This suggests that
nant decumbent plants (see also Gibson, Watt & Brown, plant species numbers in the area a¡ected by free-range
1987; Smith & Rushton, 1994). grazing are maintained by grazing. Depending on the
The highest grazing level we applied, although initially vegetation type, grazing may thus be an important tool
estimated to be su⁄cient to emulate high grazing inten- in managing plant diversity.
sity, was probably not high enough to decrease species In contrast to most other studies (e.g. Smith & Rushton,
richness in the short term. Decreased richness/diversity 1994), in our study grazing had relatively little direct
apparently occurs only in long-term overgrazing situa- in£uence on species composition. With the exception of
tions (Barker, Herlocker & Young,1989 [grassland]; McIn- CSsim, which decreased signi¢cantly in two of the blocks
tyre & Lavorel, 1994 [grassland]). Many studies (of (because of either grazing or intrinsic succession), none
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
Ground vegetation response to livestock grazing 337
of the e¡ects were signi¢cant. Nevertheless, there were compared to10% to 35% erect plants; Fig.5). This is prob-
some indications of a grazing e¡ect in that the average ably because the strategy of colonization by horizontal
similarity in a grazing level (FSsim), tended to decline fas- vegetative growth has a competitive advantage in a
ter in grazed than ungrazed paddocks, while at the same dynamic and large-scale patchy resource environment
time Tsim decreased faster in Control than in other graz- (Dushyantha & Hutchings,1997). Coastal dune vegetation
ing levels.This indicates a faster average species turnover may therefore be adapted to disturbance on several di¡er-
across blocks in the absence of grazing. If turnover is a ent evolutionary and ecological time-scales (Mentis &
measure of the rate of succession, then livestock grazing Ellery, 1998). However, this adaptation is likely to be spe-
may ‘retard’ succession (Davidson, 1993). This may be ci¢c to relatively large disturbances, in the size range of
because grazing-tolerant plant species persist longer tree-falls to cyclonic blowouts (although there is some
(Tsim declines relatively slowly, explaining the grazed evidence that disturbances larger than tree-falls are
results), or it may be an artefact of the decrease in number uncommon in African forests; Chapman et al., 1999),
of species in Control (Tsim declines relatively fast, and not to disturbances by large gregarious herbivores.
explaining the ungrazed results), or both. ‘Natural’ disturbances in this type of environment are
A level-dependent e¡ect on the range of both SD andVC typically intense, of short duration and cause a large-
values was the most noticeable and also most interesting scale (relative to plant size) mosaic of open areas (Mentis
grazing e¡ect. For instance, exclusion of grazing appar- & Ellery, 1998), while cattle disturbance is low-grade,
ently decreases any a priori site di¡erences in SD, suggest- cumulative and causes a small-scale patchy resource en-
ing that there is a general limit to species richness in vironment [implicit in McNaughton’s (1984) argument].
the absence of disturbance. Furthermore, the increased Scale-dependent e¡ects of disturbance on plant com-
range of SD values under grazing was site-speci¢c, some munities have often been recorded (Co⁄n & Lauenroth,
sites within the same grazing level remaining relatively 1988; Walker, Langridge & McFarlane, 1997). Further-
una¡ected (Figs. 4 and 5). Giventhat each site has its char- more, a survey of the literature on grazing and its e¡ects
acteristic complement of species (evidenced by relatively onvegetation suggests that there is a strong e¡ect on plant
low FSsim), the resistance of particular species to grazing life attributes, particularly in environments that do not
is probably crucial in determining the response of a site have a long history of grazing by large herbivores (for
to grazing. A site that originally consisted of a few some recent examples see McIntyre et al., 1995; Belsky &
patch-dominant plant species might become more het- Blumenthal, 1997; Fensham, Holman & Cox, 1999). Our
erogeneous in response to grazing and consequently study tends to con¢rm this.
allow a more even spread of species (in e¡ect decreasing
the scale of patchiness). On the other hand a site
Conclusion
that was heterogeneous before the start of grazing may
react in exactly the opposite way ^ a few robust plants Our results point to two important, easily measured
may become dominant at the cost of a number of more mechanisms with wide application in coastal dune forest
disturbance-sensitive plants. Also, a di¡erential response management ^ the interaction of disturbance type with
of the woody and herbaceous plants to a disturbance by ¤
plant growth form (Adamoli et al., 1990; D|¤ az, Acosta &
cattle may lead to increases or decreases in species Cabido, 1992; McIntyre, Lavorel & Tremont, 1995), and
numbers. the increase in variation in community structural vari-
We know of no other studies that have looked at the ables under disturbance (Warwick & Clarke, 1993).
e¡ect of grazing on variation in plant species numbers However, on the east coast of Africa where livestock
and vegetation cover. Indeed, apart from Warwick & numbers are increasing, these mechanisms have not
Clarke (1993), who used increased variabilityas an indica- been studied adequately ^ there is for instance no quanti-
tion of ecosystem stress in a marine environment, we tative information available on the levels and e¡ects of
were unable to ¢nd any references to this phenomenon. anthropogenically-derived disturbances (including cat-
Finally, grazing tended to increase the proportion of tle) in any coastal forested areas in southern Africa. In
erect compared to decumbent plants. The ground vegeta- view of increased development and increasing human
tion in our study area was dominated by decumbent population pressures in the coastal regions, a trend
plants (from 40% to 65% frequency over all grazing levels, that occurs not just in South Africa but also across the
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
338 Theo D.Wassenaar and Rudi J. van Aarde
African continent, this is a factor that warrants urgent Dushyantha, K.W. & Hutchings, M.J. (1997) The effects of spatial
investigation. scale of environmental heterogeneity on the growth of a clonal
plant: an experimental study with Glechoma hederacea. J. Ecol.
85, 17^28.
Acknowledgements Fensham, R.J., Holman, J.E. & Cox, M.J. (1999) Plant species
responses along a grazing disturbance gradient in Australian
Richards Bay Minerals (RBM), the National Research grassland. J.Veg. Sci. 10,77^86.
Foundation and the Department of Trade & Industry Ferreira, S.M. (1993) The Effects of Habitat Rehabilitation at
supported this study. We thank Paul Camp and Andrew Richards Bay on Small Mammal Community Structure. MSc Thesis,
Denton of the Ecology Department at RBM, Bheki University of Pretoria, South Africa.
Mpanza (¢eld assistance) and Caroline Lamb (prepara- Ferreira, S.M. (1996) Determinants of Small Mammal Community
Structure on Rehabilitating Coastal Dune Forests in Northern
tion of manuscript).
KwaZulu-Natal, South Africa. PhD Thesis, University of Pretoria,
South Africa.
Ferreira, S.M. & Van Aarde, R.J. (1996) Changes in community
References
characteristics of small mammals in rehabilitating coastal
Avis, M.A. (1992) Climate, soils and land use potential. In: Environ- dune forests in northern KwaZulu/Natal. Afr. J. Ecol. 34,
mental Impact Assessment. Eastern Shores of Lake St Lucia 113^130.
(Kingsa/Tojan Lease Area). Specialist ReportsVolume 1: Biophysical Gibson, C.W.D.,Watt,T.A. & Brown,V (1987) The use of sheep
.K.
Environment. Coastal and Environmental Services, grazing to recreate species-rich grassland from abandoned
Grahamstown. arable land. Biol. Conserv. 42, 165^183.
Ada moli, J., Sennhauser, E., Acero, J.M. & Rescia, A. (1990) Stress
¤ Glenn-Lewin, D.C. & Van Der Maarel, E. (1992) Patterns and
and disturbance: vegetation dynamics in the dry Chaco region processes of vegetation dynamics. In: Plant Succession.Theory and
of Argentina. J. Biogeogr. 17, 491^500. Prediction (Eds D. C. Glenn-Lewin, R. K. Peet and T. T.Veblen).
Ash, A.J. & McIvor, J.G. (1998) How season of grazing and herbivore Chapman & Hall, London.
selectivity influence monsoon tall-grass communities of Ludwig, J.A. & Reynolds, J.F. (1988) Statistical Ecology. A Primer on
northern Australia. J.Veg. Sci. 9, 123^132. Methods and Computing. John Wiley & Sons, NewYork.
Barbour, M.G., Burk, J.H. & Pitts,W.D. (1987) Terrestrial Plant McIntyre, S. & Lavorel, S. (1994) How environmental and
Ecology, 2nd edn. Benjamin/Cummings, Menlo Park, CA. disturbance factors influence species composition in temperate
Barker, J.R., Herlocker, D.J. & Young, S.A. (1989) Vegetal Australian grasslands. J.Veg. Sci. 5, 373^384.
dynamics along a grazing gradient within the coastal grass- McIntyre, S., Lavorel, S. & Tremont, R.M. (1995) Plant life-
land of central Somalia. Afr. J. Ecol. 27, 283^289. history attributes: their relationship to disturbance responses in
Belsky, A.J. & Blumenthal, D.M. (1997) Effects of livestock grazing herbaceous vegetation. J. Ecol. 83, 31^44.
on stand dynamics and soil in upland forests of the interior west. McNaughton, S.J. (1984) Grazing lawns: animals in herds, plant
Biol. Conserv. 11, 315^327. form and coevolution. Am. Nat. 124, 863^886.
Bullock, J.M., Clear Hill, B., Dale, M.P. & Silvertown, J. (1994) Mentis, M.T. & Ellery,W.N. (1998) Environmental effects of mining
An experimental study of the effects of sheep grazing on coastal dunes: conjectures and refutations. S. Afr. J. Sci. 94,
vegetation change in a species-poor grassland and the role of 215^222.
seedling recruitment into gaps. J. appl. Ecol. 31, 493^507. O’Connor,T. (1991) Influences of rainfall and grazing on the
Camp, P.D. (1990) Rehabilitation after dune mining at Richards Bay compositional change of the herbaceous layer of a sandveld
Minerals. S. Afr. Min.World October 1990, 34^37. savanna. J. Grassl. Soc. S. Afr. 8, 103^109.
Chapman, C.A., Chapman, L.J., Kaufman, L. & Zanne, A.E. (1999) Pandey, C.B. & Singh, J.S. (1992) Influence of rainfall and grazing
Potential causes of arrested succession in Kibale National park, on herbage dynamics in a seasonally dry tropical savanna.
Uganda: growth and mortality of seedlings. Afr. J. Ecol. 37, 81^92. Vegetatio 102, 107^124.
Clarke, K.R. (1993) Non-parametric multivariate analyses of Peel, M.J.S., Grossman, D. & Van Rooyen, N. (1991) Determinants
changes in community structure. Aust. J. Ecol. 18, 117^143. of herbaceous plant species composition on a number of ranches
Coffin, D.P. & Lauenroth,W.K. (1988) The effects of disturbance in the north-western Transvaal. J. Grassl. Soc. S. Afr. 8, 99^102.
size and frequency on a shortgrass plant community. Ecology 69, Potvin, C., Lechowicz, M.J. & Tardif, S. (1990) The statistical
1609^1617. analysis of ecophysiological response curves obtained from
Davidson, D.W. (1993) The effects of herbivory and granivory on experiments involving repeated measures. Ecology 71, 1389^
terrestrial plant succession. Oikos 6, 23^25. 1400.
D|¤ az, S., Acosta, A. & Cabido, M. (1992) Morphological analysis of Skinner, J.D. & Smithers, R.H.N. (1990) The Mammals of the
herbaceous communities under different grazing regimes. J.Veg. Southern African Subregion. University of Pretoria,
Sci. 3, 689^696. South Africa.
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
Ground vegetation response to livestock grazing 339
Smith, R.S. & Rushton, S.P. (1994) The effects of grazing Walker, B.H. & Knoop,W.T. (1987) The response of the herbaceous
management on the vegetation of mesotrophic (meadow) layer in a dystrophic Burkea africana savanna to increased levels
grassland in Northern England. J. appl. Ecol. 31, 13^24. of nitrogen, phosphate and potassium. J. Grassl. Soc. S. Afr. 4,
Sokal, R.R. & Rohlf, F.J. (1995) Biometry. The Principles and 31^34.
Practice of Statistics in Biological Research.W.H. Freeman, Walker, B.H., Langridge, J.L. & McFarlane, F. (1997) Resilience of
NewYork. an Australian savanna grassland to selective and non-selective
Sutherland, J.P. (1990) Perturbations, resistance, and alternative perturbations. Aust. J. Ecol. 22, 125^135.
views of the existence of multiple stable points in nature. Am. Warwick, R.M. & Clarke, K.R. (1993) Increased variability as a
Nat. 136, 270^275. symptom of stress in marine communities. J. Exp. Mar. Biol. Ecol.
Van Aarde, R.J., Ferreira, S.M., Kritzinger, J.J.,Van Dyk, P.J., 172, 215^226.
Vogt, M. & Wassenaar,T.D. (1996) An evaluation of habitat Weisser, P.J. (1978) Changes in the area of grasslands on the dunes
rehabilitation on coastal dune forests in northern KwaZulu- between Richards Bay and the Mfolozi river, 1937^74. Proc.
Natal, South Africa. Rest. Ecol. 4, 334^345. Grassl. Soc. S. Afr. 13, 95^97.
Van De Koppel, J., Rietkerk, M. & Weissing, J. (1997) Catastrophic Weisser, P.J. & Muller, R. (1983) Dune vegetation dynamics from
vegetation shifts and soil degradation in terrestrial grazing 1937 to 1976 in the Mlalazi-Richards Bay area, Natal, South
systems. Tr. Ecol. Evol. 12, 352^356. Africa. Bothalia 14, 661^667.
Von Maltitz, G.P.,Van Wyk, G.F. & Everard, D.A. (1996) Wilson, S.D. (1994) The contribution of grazing to plant diversity in
Successional pathways in disturbed coastal dune forest on the alpine grassland and heath. Aust. J. Ecol. 19, 137^140.
coastal dunes in north-east KwaZulu-Natal, South Africa. S. Afr.
J. Bot. 62, 188^195. (Manuscript accepted 12 December 2000)
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
forest ground vegetation to livestock grazing
Theo D.Wassenaar and Rudi J. van Aarde
Conservation Ecology Research Unit, Department of Zoology & Entomology, University of Pretoria, Pretoria, South Africa
¤
rage dependaient apparemment des changements de
Abstract
¤ ¤ ' ¤
vegetation intrinseques speci¢ques des sites, et certains
We investigated the responses of the ground vegetation in indices laissent penser qu’il y a des interactions entre les
a17-year-old coastal dune forest plant community to four “
chutes de pluies et le taux de paturage. Le paturage qui“
levels of experimentally applied livestock grazing (three avait certains e¡ets apparents, mais non signi¢catifs sur
grazing levels and one ungrazed control) from May ¤
la composition speci¢que des plantes, a¡ectait signi¢cati-
1994 to March 1996. The e¡ects of grazing were appar- '
vement la richesse en especes avec le temps, et augmen-
ently subordinate to site-speci¢c intrinsic vegetation ¤
tait signi¢cativement l’etendue de la richesse en especes '
change and there were some indications that rainfall ¤ ¤ ¤ ¤
vegetales et la couverture vegetale ainsi que l’abondance
interacted with grazing level. Grazing had some apparent ' ¤ ¤
relative et le nombre d’especes vegetales de forme erigee. ¤ ¤
but no signi¢cant e¡ects on plant species composition, ¤ ¤
Le couvert vegetal changeait signi¢cativement au cours
signi¢cantly a¡ected plant species richness over time, “ ¤
du temps, quel que soit le paturage. Nos resultats indi-
and signi¢cantly increased the range of species richness ¤
quent deux mecanismes importants, faciles a mesurer, '
and vegetation cover values as well as the relative abun- “
pour la gestion conservatoire des forets de dunes cotieres “ '
dance and numbers of plant species with erect growth ^ l’ interaction du type de perturbation avec la forme de
forms. Vegetation cover changed signi¢cantly over time, la croissance des plantes, et l’accroissement de la variation
independently of grazing. Our results point to two impor- des variables structurelles de la communaute vegetale ¤ ¤ ¤
tant, easily measured mechanisms for the conservation ¤ “
en cas de perturbation. Ces mecanismes, meme s’ ils peu-
management of coastal dune forests ^ the interaction of vent avoir une large application et une valeur predictive, ¤
disturbance type with plant growth form and the ¤ ¤
n’ont pas encore ete convenablement etudies. ¤ ¤
increase of variation in community structural variables
under disturbance. These mechanisms, although they
potentially have wide application and predictive power, Introduction
have not been studied adequately.
Two main factors are relevant to grazing in African
Key words: disturbance, growth form, species richness, coastal dune forests. The ¢rst of these is that large herbi-
variability vores can in£uence the organization of almost every plant
community at many di¡erent scales and on all hierarch-
ical levels (Glenn-Lewin & van der Maarel, 1992; van de
¤
Resume¤
Koppel, Rietkerk & Weissing, 1997). Although coastal
¤ ¤ ¤ ¤ ¤
Nous avons etudie les reponses de la vegetation terrestre dune forests are generally regarded as a resilient habitat
' “ '
d’une plantation forestiere de17 ans sur une dune cotiere type with a fairly predictable development from grassland
' ¤ “
a quatre niveaux experimentaux de paturage par du to mature dune forest (Weisser, 1978; Weisser & Muller,
¤ “ “
betail (trois niveaux de paturage et un controle non 1983), current insights into community development
“ ¤ “
pature) entre mai 1994 et mars 1996. Les e¡ets du patu- stress the existence of multiple possible stable states
(Sutherland, 1990; see also von Maltitz, van Wyk &
Correspondence:Theo D.Wassenaar, Conservation Ecology Research
Everard, 1990).
Unit, Department of Zoology & Entomology, University of Secondly, few vertebrate herbivores inhabit the coastal
Pretoria, Pretoria 0002, South Africa. Tel: þ27 12 4202561; Fax: dune forests of the southeast coast of Africa (Skinner &
þ27 12 4204523; E-mail: tdwassenaar@zoology.up.ac.za Smithers, 1990). Of the 26 mammal species listed by
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339 329
330 Theo D.Wassenaar and Rudi J. van Aarde
Ferreira (1993) for the coastal dune forests north of are uniform throughout the area, with little horizontal
Richards Bay, KwaZulu-Natal, South Africa, only three di¡erentiation (Avis, 1992). Most rain falls from January
herbivores (the bushpig Potamochoerus porcus [Linnaeus to March (peak in February; annual mean 1292 mm).
1758], the red duiker Cephalophus natalensis [Smith Extended droughts are uncommon and approximately
1834] and the bushbuck Tragelaphus scriptus [Pallas 30% of annual precipitation occurs in the winter. Daily
1766]) are large enough to cause signi¢cant disturbance maximum temperatures range from 22.6 8C to 30.0 8C
to vegetation. Both red duiker and bushbuck occur in and minimum temperatures from 10.0 8C to 20.6 8C in
relatively low numbers and are highly selective in their June and January, respectively (Ferreira, 1996).
feeding habits, while the distribution of the bushpig is The regenerating forest consists of an Acacia karroo
typically patchy and they are not common anywhere (Hayne) woodland 9^12 m high, with secondary dune
(Skinner & Smithers, 1990). Livestock grazing, which forest tree species colonizing. Compared to a mature
is therefore likely to be an ‘unnatural’ type of distur- forest there is relatively little vertical strati¢cation. The
bance for coastal dune forests, could thus have far- herb layer (%1m or less) is dominated by a number of sto-
reaching implications, not only for the conservation of loniferous grasses, creepers and decumbent herbs. Van
coastal dune forests, but also for their post-disturbance Aarde et al. (1996) provide a detailed description of the
development. study area and vegetation.
The mining company Richards Bay Minerals (RBM)
has been mining and subsequently rehabilitating a strip
Experimental design and data recording
of coastal sand dunes on the northeast coast of South
Africa since July 1977 (Camp, 1990; van Aarde et al., Grazing was applied to 0.125 ha fenced paddocks,
1996). The mining lease and rehabilitating areas are situ- arranged in four blocks with a Control, Low, Medium
ated in a regional development node with large in£uxes and High grazing level in each (i.e. four replicates per
of people (23% increase from 1980 to 1991) and their grazing level). Blocks were randomly placed within the
associated domestic livestock. This has led to illegitimate study site and grazing levels randomly assigned to pad-
grazing of regenerating forests, and concerns over its docks within each block. In this instance the term ‘graz-
possible e¡ect on the rehabilitation programme. ing’ includes all other forms of disturbance by cattle to
We expected livestock grazing to signi¢cantly a¡ect the vegetation, i.e. trampling, defecating, etc. Five grazing
the restoration of coastal dune forest in the short term cycles were applied over 16 months (Fig.1), the period
through its e¡ect on ground vegetation community between grazing cycles varying from 90 to 150 days. A
structure. The present study therefore examines the grazing cycle consisted of 2 days’grazing in the low graz-
short-term e¡ects of cattle on ground vegetation (i) ing level paddocks, 4 days in the medium level paddocks
species richness and vegetation cover (ii) species compo- and 8 days in the high level paddocks. For each grazing
sition, and (iii) numbers and abundances of species in cycle, eight heifers (200^300 kg) were sorted into four
two plant growth form groups in a17-year-old rehabilitat- pairs (so that mean pair-weight % mean group-weight)
ing coastal dune forest. and each pair randomly assigned to one of the four repli-
cates in a grazing level every day.
Species presence and vegetation cover were recorded at
Materials and methods
six1m2 quadrats per paddock prior to each grazing cycle.
Vegetation cover was recorded with a point-bridge
Study area
adapted from Barbour, Burk & Pitts, (1987). Plant species
The study was conducted in the oldest regenerating were assigned to one of two categories: (i) decumbent
part of an area of mined dunes north of Richards Bay, (all species, excluding lianas, with mostly horizontal
KwaZulu-Natal, South Africa (288430 S and 328120 E), vegetative growth, i.e. creepers and stoloniferous
which has been under ecological rehabilitation since1977 grasses), and (ii) erect (large and small annual and peren-
(Stand 1 in Ferreira & van Aarde, 1996). The area is char- nial forbs, tussock grasses, juvenile trees and shrubs).
acterized by longitudinal sand dunes, lying parallel to Rainfall was measured using standard rain gauges at
the coastline, and rising to an elevation of 40^90 m above two points situated halfway between the four blocks of
sea level. Soils (¢ne to medium-grained aeolian sands) grazing paddocks.
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
Ground vegetation response to livestock grazing 331
Fig 1 Mean (n ¼ 3) daily rainfall in the
rehabilitating area from January 1994 to
August 1996 and schedule of plant surveys
and grazing cycles. The dotted line
represents the mean daily rainfall over a
10-year period, including the period of our
study. S1^S6 represent plant surveys and
G1^G5 grazing applications
The data were used to calculate species density (SD or Kruskal^Wallis test where applicable, for the overall
species/m2), vegetation cover below 10 cm (VC), relative e¡ect of cattle on VC (arcsine transformed) and SD with
abundance of growth form (Erect/Decumbent[Ab]) and multivariate analysis of variance for repeated measures
number of species per growth form (Erect/Decum- (manovar; Potvin, Lechowicz & Tardif, 1990), and for
bent[Sp]), as well as species composition. For species comparisons between grazing levels we used Tukey’s
composition analysis we de¢ned three Bray^Curtis simi- honestly signi¢cant di¡erence test (T-HSD; Sokal & Rohlf,
larityclasses (Ludwig & Reynolds,1988): (i) temporal simi- 1995). We used least squares linear regression (LSLR) to
larity (Tsim): similarity of a paddock to itself over time; test for the e¡ect of cattle on the range inVC and SD values
(ii) close spatial similarity (CSsim): similarity between as well as on all growth form variables and similarity
paddocks in the same block, but from di¡erent grazing changes over time (independent variable: days since start
levels; and (iii) far spatial similarity (FSsim): similarity of ¢rst survey). A signi¢cant di¡erence between the slopes
between paddocks in the same grazing level, but from dif- of control and any of the other grazing levels indicated a
ferent blocks. grazing e¡ect, and a slope signi¢cantly di¡erent from zero
To determine whether the microhabitats at the quad- for any of the grazing levels indicated a time e¡ect.Analy-
rats were similar, we recorded the structure of the canopy sis of similarity (anosim; Clarke, 1993) was used to test
vegetation surrounding each quadrat (name, height, for di¡erences in species composition between grazing
stem diameter and number of stems of closest species levels, paddocks and sequential surveys.
of canopy tree in each of four quarters surrounding
the quadrat), percentage canopy cover at each quadrat
Results
(measured on a colour negative of the underside of
the canopy with a Quantimet 520 image analyser;
Microhabitat
Cambridge Instruments, London), and the light inten-
sity (lux) on the forest £oor (Measuring Instruments There were no signi¢cant di¡erences between grazing
Technologies, Pretoria). levels in canopy tree structure and density (Kruskal^
Wallis: all H-values 3.22, all P-values >0.05), incandes-
cent light (overall mean Æ SE ¼ 21.40 Æ1.43 lux; anova:
Statistical analysis
F ¼1.31, P ¼ 0.20) and percentage canopy cover (overall
We tested for di¡erences in microhabitat between mean ÆSE ¼ 84.9 Æ 0.78%; anova: F ¼1.18, P ¼ 0.36).
paddocks with analysis of variance (anova) and the Only one species of canopy tree was recorded, namely
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
332 Theo D.Wassenaar and Rudi J. van Aarde
A. karroo.We therefore assumed that di¡erences in micro- time^grazing level interaction (linear) and time e¡ect
climatic conditions at each quadrat were small enough (quadratic) were signi¢cant (manovar; overall time Â
for the herb layer vegetation not to have been a¡ected by level: Roy’s GR ¼ 0.34, F ¼ 6.20, P < 0.001, overall time:
it. Furthermore, although precipitationwas not measured Roy’s GR ¼1.94, F ¼ 34.20, P < 0.001; linear (time Â
at all survey points, we assumed that paddocks were close level): F ¼ 4.26, P ¼ 0.01; quadratic (time): F ¼10.48,
enough to each other not to have been in£uenced by spa- P ¼ 0.002). Control di¡ered from Low and High over all
tial variation in rainfall patterns. Rainfall followed a sea- surveys (T-HSD, msd(0.05,20) ¼ 1.87). During survey 2 Con-
sonal pattern during the study period, although daily trol and High di¡ered signi¢cantly (msd(0.05,12) ¼ 1.803),
rainfall was below the 10-year average for a considerable during survey 4 Control di¡ered from both Low and High
period before the third survey (Fig.1). It increased some- (msd(0.05,12) ¼ 2.35), during survey 5 (msd(0.05,12) ¼ 2.37)
what after the second survey and at the time of the fourth and survey 6 (msd(0.05,12) ¼ 1.97) Control di¡ered from
survey increased above the 10-year average. Low, Medium and High.
Increased grazing level apparently increased the
range of SD values at each survey over time; in con-
Plant species density (SD)
trast the range decreased in Control (Fig.3). The slope
Mean SD decreased over time in Control, while it stayed at of the relationship between time and range of values
the same level in Low, Medium and High (Fig. 2). The at each survey was signi¢cant for Control (LSLR;
Fig 2 The number of species per1m2 quadrat (species density or SD ^ see text) over 6 plant surveys in (a) Control (b) Low (c) Medium and (d)
High. Symbols represent different paddocks. Lines were fitted with least squares linear regression on the mean values per survey
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
Ground vegetation response to livestock grazing 333
Roy’s GR ¼ 3.96, F ¼ 8.91, P ¼ 0.003, quadratic (time):
F ¼ 26.81, P < 0.001). The range of values per survey
increased with grazing level (Fig.3), but none of the slopes
di¡ered signi¢cantly from zero (slopes from À0.004 to
0.06, all P > 0.05, r2 from 0.006 to 0.30) or from each other
(F ¼ 0.44; P > 0.05).
Growth form
Decumbent[Ab] values for Control remained higher than
Low, Medium and High throughout and apparently
increased more over time (Fig.5a). Slopes were not signif-
icantly di¡erent (F ¼1.07, P ¼ 0.39). In contrast Erect[Ab]
values for Control remained lower than the other grazing
levels throughout (Fig.5b), the slope for Control di¡ering
signi¢cantly from High (Control: y ¼15.69 À 0.008x;
P ¼ 0.15, r2 ¼0.44; High: y ¼ 23.84 þ 0.009x; P ¼ 0.06,
r2 ¼0.63; F ¼ 3.90, P ¼ 0.03). The slope of the relationship
between time and Erect[Sp] was signi¢cant only for
Low and High (Low: y ¼18.53 þ0.009x; P ¼ 0.04,
r2 ¼0.70; High: y ¼15.29 þ 0.01x; P ¼ 0.04, r2 ¼0.70).
Both Erect and Decumbent species were apparently lost
faster from Control (Fig.5c,d). Di¡erences were more pro-
nounced in Erect plants however (Fig.5d), where the
slope for Control di¡ered signi¢cantly from High
(F ¼ 5.39, P ¼ 0.009).
Similarity/species composition
Fig 3 The relationship between time (days since start of first Species composition of Control paddocks apparently
plant survey) and range in values (difference between changed faster over time (similarity to initial species com-
minimum and maximum for each survey) of (a) SD and position, or Tsim, declined faster; Fig.6a) and stayed more
(b) VC. Least squares linear regression models are described
similar to each other than the other grazing levels (mean
in the text
similarity within grazing level, or FSsim, was higher;
Fig.6b). For Tsim, slopes were signi¢cant in Control
(y ¼ 94.06 À3.57x, P ¼ 0.05; r2 ¼0.79, P ¼ 0.05) and
y ¼ 3.70 À 0.004x, P ¼ 0.03, r2 ¼0.74), but not for Low Low (y ¼ 86.66 À1.51x, P ¼ 0.01; r2 ¼0.93, P ¼ 0.01), but
to High (slopes from 0.001 to 0.006, all P > 0.05; r2 there were no overall signi¢cant di¡erences between
from 0.08 to 0.53). Slopes were signi¢cantly di¡erent the slopes (F ¼1.22, P ¼ 0.35). Slopes for FSsim (Fig.6b)
overall (F ¼ 3.47; P ¼ 0.04), with Control di¡ering from were not signi¢cantly di¡erent from zero or from each
Low. other (range ¼ À1.56 to À0.11, all P-values >0.05; all
r2-values 0.40; F ¼ 0.35, P ¼ 0.79). Similarity within
blocks (CSsim) (Fig.6c) in two of the blocks decreased
Vegetation cover (VC)
signi¢cantly over time (Block 2: y ¼ 66.45 À0.01x, P ¼
Mean VC apparently increased over time in Control, but 0.009, r2 ¼0.18; Block 4: y ¼ 71.16 À 0.02x, P ¼ 0.007,
decreased slightly in High (Fig. 4). There was a signi¢cant r2 ¼0.20), but the slopes for the di¡erent blocks did not
quadratic time e¡ect, but no signi¢cant time^grazing di¡er signi¢cantly (F ¼1.26, P ¼ 0.29). Mean Tsim (over
level interaction or grazing e¡ect (manovar; overall time: all grazing levels) stayed higher than both mean CSsim
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
334 Theo D.Wassenaar and Rudi J. van Aarde
Fig 4 The percentage ground cover (VC ^ see text) (arcsine transformed) over 5 plant surveys in (a) Control (b) Low (c) Medium and (d) High.
Symbols represent different paddocks. Lines were fitted with least squares linear regression on the mean values per survey
and mean FSsim, and CSsim stayed higher than FSsim
Discussion
throughout the study period (Fig.6d).
There were no signi¢cant di¡erences between either
Site and climate
the grazing levels or the di¡erent surveys in species
composition (two-way crossed anosim; R ¼ À0.004, Our study was conducted over a relatively short time-
P ¼ 0.52 and R ¼ À0.09, P ¼ 0.98, respectively). There scale, with an emphasis on the e¡ects of grazing on
were also no signi¢cant di¡erences between the grazing ground vegetation community variables. Results suggest
levels within any of the surveys (one-way anosim; global that grazing had some e¡ects on plant community struc-
R-values range À0.11 to 0.11, all P-values >0.05). Blocks ture, although these were not always signi¢cant and
of paddocks on the other hand di¡ered signi¢cantly over appeared to be subordinate to both intrinsic vegetation
all (global R ! 0.45, P 0.001 in all surveys). In these change (i.e. site e¡ects) and rainfall e¡ects. Site e¡ects
comparisons the pair of blocks situated closer to each can clearly be seen in the signi¢cant di¡erences in species
other were never signi¢cantly di¡erent (R 0.46, composition between the blocks of paddocks at all times,
P > 0.05 in all cases), while the other blocks, situated while at the same time none of the grazing levels di¡ered.
further away from each other, were always di¡erent Furthermore, if grazing had an e¡ect over and above site
(R ! 0.43, P < 0.05 in all cases). and environment, there would have been a change in
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
Ground vegetation response to livestock grazing 335
Fig 5 The relationship between time (days since start of first plant survey) and (a) relative abundance of decumbent plants, (b) relative
abundances of erect plants, (c) number of decumbent species, and (d) number of erect species; in four grazing levels. Least squares linear
regression models are described in the text
the site-dependent theoretical ranking of temporal (Tsim; relatively dry preceding period (Fig.1). Furthermore,
highest) ^ close spatial (CSsim) ^ and far spatial (FSsim) there were some indications of a possible interaction
similarity over time, and/or a di¡erence between control between grazing level and rainfall e¡ect, because the
and treatment in this pattern. This did not happen, nor magnitude of the drop in both SD and VC during survey
did paddocks within the same grazing level and paddocks 3 (Figs. 2 and 4) was apparently dependent on the level
within the same block become more (i.e. increased FSsim) of grazing. It is well known that rainfall (and climate in
and less (i.e. decreased CSsim) similar to each other, general) can have a strong in£uence on plant community
respectively, as expected. Also, although CSsim did variables, overshadowing other abiotic and biotic distur-
decrease in two of the blocks, FSsim, contrary to expecta- bances (e.g. Walker & Knoop, 1987; O’Connor, 1991; Peel,
tion, actually decreased. Grazing could therefore not Grossman & van Rooyen, 1991).
overcome the high Tsim and CSsim, or the low FSsim,
which suggests that the ground vegetation community
Grazing
of these coastal dunes is relatively resistant to grazing dis-
turbance in the short term. In contrast to our study, earlier studies noted decreases
Rainfall e¡ect was not speci¢cally investigated, but in vegetation cover under grazing (e.g. Belsky &
the drop in almost all variables and grazing levels during Blumenthal, 1997, cites several examples in western US
survey 3 followed on and was probably the result of a forests). Presumably our result was due to a combination
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
336 Theo D.Wassenaar and Rudi J. van Aarde
Fig 6 The relationship between time (days since start of first plant survey) and (a) temporal similarity (Tsim ¼ similarity to initial
species composition), (b) far spatial similarity (FSsim ¼ paddocks in the same grazing level but in different blocks), (c) close spatial similarity
(CSsim ¼ paddocks in the same block but in different grazing levels), and (d) a summary of changes in mean similarity over all grazing
levels (for Tsim and FSsim) and over all blocks (for CSsim); in four grazing levels. Least squares linear regression models are described in
the text
of (possibly) too light a grazing pressure and relatively varying duration) in fact report higher species richness
long intervals between successive surveys, providing under grazing (Gibson et al., 1987 [moist grassland],
the fast-growing herbaceous vegetation su⁄cient time O’Connor,1991 [savanna], Pandey & Singh,1992 [season-
to recover. However, grazing had a signi¢cant level- ally dry savanna], Bullock et al., 1994 [grassland], Ash &
dependent e¡ect on SD (species density, an index of rich- McIvor, 1998 [tall grassland]), usually as a result of
ness), which evidently became progressively more intense increases in annual grasses and forbs. A distinction
with time, mostly because of a progressive decrease should however be made between grazing e¡ect and
in mean SD in the ungrazed Control. Grazing itself exclusion e¡ect ^ in those studies carried out in a back-
apparently maintained mean species richness (Smith & ground of free-range grazing (as our own was), species
Rushton, 1994; Wilson, 1994), in our case probably as richness does not increase under grazing, it rather
a result of an increase in the number of erect forb species decreases where grazing has been excluded (e.g. Pandey
as they were released from competition with the domi- & Singh,1992; Smith & Rushton,1994). This suggests that
nant decumbent plants (see also Gibson, Watt & Brown, plant species numbers in the area a¡ected by free-range
1987; Smith & Rushton, 1994). grazing are maintained by grazing. Depending on the
The highest grazing level we applied, although initially vegetation type, grazing may thus be an important tool
estimated to be su⁄cient to emulate high grazing inten- in managing plant diversity.
sity, was probably not high enough to decrease species In contrast to most other studies (e.g. Smith & Rushton,
richness in the short term. Decreased richness/diversity 1994), in our study grazing had relatively little direct
apparently occurs only in long-term overgrazing situa- in£uence on species composition. With the exception of
tions (Barker, Herlocker & Young,1989 [grassland]; McIn- CSsim, which decreased signi¢cantly in two of the blocks
tyre & Lavorel, 1994 [grassland]). Many studies (of (because of either grazing or intrinsic succession), none
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
Ground vegetation response to livestock grazing 337
of the e¡ects were signi¢cant. Nevertheless, there were compared to10% to 35% erect plants; Fig.5). This is prob-
some indications of a grazing e¡ect in that the average ably because the strategy of colonization by horizontal
similarity in a grazing level (FSsim), tended to decline fas- vegetative growth has a competitive advantage in a
ter in grazed than ungrazed paddocks, while at the same dynamic and large-scale patchy resource environment
time Tsim decreased faster in Control than in other graz- (Dushyantha & Hutchings,1997). Coastal dune vegetation
ing levels.This indicates a faster average species turnover may therefore be adapted to disturbance on several di¡er-
across blocks in the absence of grazing. If turnover is a ent evolutionary and ecological time-scales (Mentis &
measure of the rate of succession, then livestock grazing Ellery, 1998). However, this adaptation is likely to be spe-
may ‘retard’ succession (Davidson, 1993). This may be ci¢c to relatively large disturbances, in the size range of
because grazing-tolerant plant species persist longer tree-falls to cyclonic blowouts (although there is some
(Tsim declines relatively slowly, explaining the grazed evidence that disturbances larger than tree-falls are
results), or it may be an artefact of the decrease in number uncommon in African forests; Chapman et al., 1999),
of species in Control (Tsim declines relatively fast, and not to disturbances by large gregarious herbivores.
explaining the ungrazed results), or both. ‘Natural’ disturbances in this type of environment are
A level-dependent e¡ect on the range of both SD andVC typically intense, of short duration and cause a large-
values was the most noticeable and also most interesting scale (relative to plant size) mosaic of open areas (Mentis
grazing e¡ect. For instance, exclusion of grazing appar- & Ellery, 1998), while cattle disturbance is low-grade,
ently decreases any a priori site di¡erences in SD, suggest- cumulative and causes a small-scale patchy resource en-
ing that there is a general limit to species richness in vironment [implicit in McNaughton’s (1984) argument].
the absence of disturbance. Furthermore, the increased Scale-dependent e¡ects of disturbance on plant com-
range of SD values under grazing was site-speci¢c, some munities have often been recorded (Co⁄n & Lauenroth,
sites within the same grazing level remaining relatively 1988; Walker, Langridge & McFarlane, 1997). Further-
una¡ected (Figs. 4 and 5). Giventhat each site has its char- more, a survey of the literature on grazing and its e¡ects
acteristic complement of species (evidenced by relatively onvegetation suggests that there is a strong e¡ect on plant
low FSsim), the resistance of particular species to grazing life attributes, particularly in environments that do not
is probably crucial in determining the response of a site have a long history of grazing by large herbivores (for
to grazing. A site that originally consisted of a few some recent examples see McIntyre et al., 1995; Belsky &
patch-dominant plant species might become more het- Blumenthal, 1997; Fensham, Holman & Cox, 1999). Our
erogeneous in response to grazing and consequently study tends to con¢rm this.
allow a more even spread of species (in e¡ect decreasing
the scale of patchiness). On the other hand a site
Conclusion
that was heterogeneous before the start of grazing may
react in exactly the opposite way ^ a few robust plants Our results point to two important, easily measured
may become dominant at the cost of a number of more mechanisms with wide application in coastal dune forest
disturbance-sensitive plants. Also, a di¡erential response management ^ the interaction of disturbance type with
of the woody and herbaceous plants to a disturbance by ¤
plant growth form (Adamoli et al., 1990; D|¤ az, Acosta &
cattle may lead to increases or decreases in species Cabido, 1992; McIntyre, Lavorel & Tremont, 1995), and
numbers. the increase in variation in community structural vari-
We know of no other studies that have looked at the ables under disturbance (Warwick & Clarke, 1993).
e¡ect of grazing on variation in plant species numbers However, on the east coast of Africa where livestock
and vegetation cover. Indeed, apart from Warwick & numbers are increasing, these mechanisms have not
Clarke (1993), who used increased variabilityas an indica- been studied adequately ^ there is for instance no quanti-
tion of ecosystem stress in a marine environment, we tative information available on the levels and e¡ects of
were unable to ¢nd any references to this phenomenon. anthropogenically-derived disturbances (including cat-
Finally, grazing tended to increase the proportion of tle) in any coastal forested areas in southern Africa. In
erect compared to decumbent plants. The ground vegeta- view of increased development and increasing human
tion in our study area was dominated by decumbent population pressures in the coastal regions, a trend
plants (from 40% to 65% frequency over all grazing levels, that occurs not just in South Africa but also across the
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
338 Theo D.Wassenaar and Rudi J. van Aarde
African continent, this is a factor that warrants urgent Dushyantha, K.W. & Hutchings, M.J. (1997) The effects of spatial
investigation. scale of environmental heterogeneity on the growth of a clonal
plant: an experimental study with Glechoma hederacea. J. Ecol.
85, 17^28.
Acknowledgements Fensham, R.J., Holman, J.E. & Cox, M.J. (1999) Plant species
responses along a grazing disturbance gradient in Australian
Richards Bay Minerals (RBM), the National Research grassland. J.Veg. Sci. 10,77^86.
Foundation and the Department of Trade & Industry Ferreira, S.M. (1993) The Effects of Habitat Rehabilitation at
supported this study. We thank Paul Camp and Andrew Richards Bay on Small Mammal Community Structure. MSc Thesis,
Denton of the Ecology Department at RBM, Bheki University of Pretoria, South Africa.
Mpanza (¢eld assistance) and Caroline Lamb (prepara- Ferreira, S.M. (1996) Determinants of Small Mammal Community
Structure on Rehabilitating Coastal Dune Forests in Northern
tion of manuscript).
KwaZulu-Natal, South Africa. PhD Thesis, University of Pretoria,
South Africa.
Ferreira, S.M. & Van Aarde, R.J. (1996) Changes in community
References
characteristics of small mammals in rehabilitating coastal
Avis, M.A. (1992) Climate, soils and land use potential. In: Environ- dune forests in northern KwaZulu/Natal. Afr. J. Ecol. 34,
mental Impact Assessment. Eastern Shores of Lake St Lucia 113^130.
(Kingsa/Tojan Lease Area). Specialist ReportsVolume 1: Biophysical Gibson, C.W.D.,Watt,T.A. & Brown,V (1987) The use of sheep
.K.
Environment. Coastal and Environmental Services, grazing to recreate species-rich grassland from abandoned
Grahamstown. arable land. Biol. Conserv. 42, 165^183.
Ada moli, J., Sennhauser, E., Acero, J.M. & Rescia, A. (1990) Stress
¤ Glenn-Lewin, D.C. & Van Der Maarel, E. (1992) Patterns and
and disturbance: vegetation dynamics in the dry Chaco region processes of vegetation dynamics. In: Plant Succession.Theory and
of Argentina. J. Biogeogr. 17, 491^500. Prediction (Eds D. C. Glenn-Lewin, R. K. Peet and T. T.Veblen).
Ash, A.J. & McIvor, J.G. (1998) How season of grazing and herbivore Chapman & Hall, London.
selectivity influence monsoon tall-grass communities of Ludwig, J.A. & Reynolds, J.F. (1988) Statistical Ecology. A Primer on
northern Australia. J.Veg. Sci. 9, 123^132. Methods and Computing. John Wiley & Sons, NewYork.
Barbour, M.G., Burk, J.H. & Pitts,W.D. (1987) Terrestrial Plant McIntyre, S. & Lavorel, S. (1994) How environmental and
Ecology, 2nd edn. Benjamin/Cummings, Menlo Park, CA. disturbance factors influence species composition in temperate
Barker, J.R., Herlocker, D.J. & Young, S.A. (1989) Vegetal Australian grasslands. J.Veg. Sci. 5, 373^384.
dynamics along a grazing gradient within the coastal grass- McIntyre, S., Lavorel, S. & Tremont, R.M. (1995) Plant life-
land of central Somalia. Afr. J. Ecol. 27, 283^289. history attributes: their relationship to disturbance responses in
Belsky, A.J. & Blumenthal, D.M. (1997) Effects of livestock grazing herbaceous vegetation. J. Ecol. 83, 31^44.
on stand dynamics and soil in upland forests of the interior west. McNaughton, S.J. (1984) Grazing lawns: animals in herds, plant
Biol. Conserv. 11, 315^327. form and coevolution. Am. Nat. 124, 863^886.
Bullock, J.M., Clear Hill, B., Dale, M.P. & Silvertown, J. (1994) Mentis, M.T. & Ellery,W.N. (1998) Environmental effects of mining
An experimental study of the effects of sheep grazing on coastal dunes: conjectures and refutations. S. Afr. J. Sci. 94,
vegetation change in a species-poor grassland and the role of 215^222.
seedling recruitment into gaps. J. appl. Ecol. 31, 493^507. O’Connor,T. (1991) Influences of rainfall and grazing on the
Camp, P.D. (1990) Rehabilitation after dune mining at Richards Bay compositional change of the herbaceous layer of a sandveld
Minerals. S. Afr. Min.World October 1990, 34^37. savanna. J. Grassl. Soc. S. Afr. 8, 103^109.
Chapman, C.A., Chapman, L.J., Kaufman, L. & Zanne, A.E. (1999) Pandey, C.B. & Singh, J.S. (1992) Influence of rainfall and grazing
Potential causes of arrested succession in Kibale National park, on herbage dynamics in a seasonally dry tropical savanna.
Uganda: growth and mortality of seedlings. Afr. J. Ecol. 37, 81^92. Vegetatio 102, 107^124.
Clarke, K.R. (1993) Non-parametric multivariate analyses of Peel, M.J.S., Grossman, D. & Van Rooyen, N. (1991) Determinants
changes in community structure. Aust. J. Ecol. 18, 117^143. of herbaceous plant species composition on a number of ranches
Coffin, D.P. & Lauenroth,W.K. (1988) The effects of disturbance in the north-western Transvaal. J. Grassl. Soc. S. Afr. 8, 99^102.
size and frequency on a shortgrass plant community. Ecology 69, Potvin, C., Lechowicz, M.J. & Tardif, S. (1990) The statistical
1609^1617. analysis of ecophysiological response curves obtained from
Davidson, D.W. (1993) The effects of herbivory and granivory on experiments involving repeated measures. Ecology 71, 1389^
terrestrial plant succession. Oikos 6, 23^25. 1400.
D|¤ az, S., Acosta, A. & Cabido, M. (1992) Morphological analysis of Skinner, J.D. & Smithers, R.H.N. (1990) The Mammals of the
herbaceous communities under different grazing regimes. J.Veg. Southern African Subregion. University of Pretoria,
Sci. 3, 689^696. South Africa.
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339
Ground vegetation response to livestock grazing 339
Smith, R.S. & Rushton, S.P. (1994) The effects of grazing Walker, B.H. & Knoop,W.T. (1987) The response of the herbaceous
management on the vegetation of mesotrophic (meadow) layer in a dystrophic Burkea africana savanna to increased levels
grassland in Northern England. J. appl. Ecol. 31, 13^24. of nitrogen, phosphate and potassium. J. Grassl. Soc. S. Afr. 4,
Sokal, R.R. & Rohlf, F.J. (1995) Biometry. The Principles and 31^34.
Practice of Statistics in Biological Research.W.H. Freeman, Walker, B.H., Langridge, J.L. & McFarlane, F. (1997) Resilience of
NewYork. an Australian savanna grassland to selective and non-selective
Sutherland, J.P. (1990) Perturbations, resistance, and alternative perturbations. Aust. J. Ecol. 22, 125^135.
views of the existence of multiple stable points in nature. Am. Warwick, R.M. & Clarke, K.R. (1993) Increased variability as a
Nat. 136, 270^275. symptom of stress in marine communities. J. Exp. Mar. Biol. Ecol.
Van Aarde, R.J., Ferreira, S.M., Kritzinger, J.J.,Van Dyk, P.J., 172, 215^226.
Vogt, M. & Wassenaar,T.D. (1996) An evaluation of habitat Weisser, P.J. (1978) Changes in the area of grasslands on the dunes
rehabilitation on coastal dune forests in northern KwaZulu- between Richards Bay and the Mfolozi river, 1937^74. Proc.
Natal, South Africa. Rest. Ecol. 4, 334^345. Grassl. Soc. S. Afr. 13, 95^97.
Van De Koppel, J., Rietkerk, M. & Weissing, J. (1997) Catastrophic Weisser, P.J. & Muller, R. (1983) Dune vegetation dynamics from
vegetation shifts and soil degradation in terrestrial grazing 1937 to 1976 in the Mlalazi-Richards Bay area, Natal, South
systems. Tr. Ecol. Evol. 12, 352^356. Africa. Bothalia 14, 661^667.
Von Maltitz, G.P.,Van Wyk, G.F. & Everard, D.A. (1996) Wilson, S.D. (1994) The contribution of grazing to plant diversity in
Successional pathways in disturbed coastal dune forest on the alpine grassland and heath. Aust. J. Ecol. 19, 137^140.
coastal dunes in north-east KwaZulu-Natal, South Africa. S. Afr.
J. Bot. 62, 188^195. (Manuscript accepted 12 December 2000)
# 2001 East African Wild Life Society, Afr. J. Ecol., 39, 329^339