Dispersal traits determine passive restoration trajectory of a Nigerian
Andrew D. Barnes
, Hazel M. Chapman
School of Biological Sciences, University of Canterbury, Private Bag 4800, Christchurch, New Zealand
Systemic Conservation Biology, J.F. Blumenbach Institute of Zoology and Anthropology, University of Göttingen, Berliner Strasse 28,
37073 Göttingen, Germany
Received 7 February 2014
Accepted 25 February 2014
Passive restoration methods offer great promise for tropical regions where resources are limited but the
success of such efforts can be variable. Using trait-based theory, we investigated the likely trajectories of
passive restoration efforts in a degraded Nigerian montane forest system recently protected from
burning and cattle grazing. We quantiﬁed the density, species richness, and functional trait dispersion of
dispersed seeds and seedling communities at increasing distances from the forest edge. We then
determined which plant traits are responsible for colonisation by quantifying changes in functional-trait
dispersion and relative frequencies of dispersal-linked traits with increasing distance from the forest. We
found a rapid decrease in density and species richness, and signiﬁcant species turnover in both seeds and
seedlings just beyond the forest edge. This was mirrored by a signiﬁcant decline in functional-trait
dispersion and a shift in the relative frequencies of dispersal-linked traits. These ﬁndings suggest that
the reassembly of plant communities adjacent to remnant forest is dependent on functional traits pre-
sent in these remnant source populations, providing support for the incorporation of trait-based theory
in restoration management.
Ó2014 Elsevier Masson SAS. All rights reserved.
Throughout the tropics, anthropogenic pressures have led to
such severe forest loss and degradation (Asner et al., 2009; Geist
and Lambin, 2002) that there is now a global effort towards their
restoration (Chazdon, 2008; Holl, 2012) and the return of
ecosystem goods and services to local communities (Benayas et al.,
2009; Lamb et al., 2005). ‘Active’restoration strategies, where
intervention techniques are used to re-establish forest, can be
costly and impractical in areas where community involvement is
essential but resources are very limited (Holl and Aide, 2011;
Parrotta et al., 1997). Alternatively, ‘passive’restoration strategies,
which involve only the restriction or total prevention of land-use
practises from degraded land, are more easily employed and
require minimal resources (Holl and Aide, 2011; Morrison and
Although potentially useful, passive restoration can be very slow
and often ineffective depending on landscape and ecological con-
ditions (Duncan and Chapman, 1999; Laing et al., 2011; Myster,
2004). Factors such as insufﬁcient seed rain (Cubiña and Aide,
2001; Duncan and Chapman, 1999; Martínez-Garza and Howe,
2003; Muñiz-Castro et al., 2006), seed bank composition
(Kalesnik et al., 2013), seed and seedling predation (Myster, 2004;
Nepstad et al., 1990), lack of suitable microsites for germination
(Eriksson and Ehrlén, 1992), and competition from grasses
(Chapman and Chapman, 1999; Duncan and Chapman, 1999) can
collectively hinder forest regeneration. Moreover, passive strategies
can lead to undesirable trajectories of ecosystem restoration
because they are highly dependent on the potential for natural seed
dispersal from nearby remnant habitat (Cole et al., 2011; del Castillo
and Ríos, 2008; Martínez-Garza and Howe, 2003). For example
(Kalesnik et al., 2013) showed that after 30 years of abandonment,
commercial forests of exotic willow and poplar in Argentina
remained mixed secondary forest with a high frequency of invasive
The likelihood of forest tree species dispersing into and estab-
lishing within adjacent degraded habitat is highly variable and
depends on a wide array of measurable factors including fruit and
*Corresponding author. Systemic Conservation Biology, J.F. Blumenbach Institute
of Zoology and Anthropology, University of Göttingen, Berliner Strasse 28, 37073
Göttingen, Germany. Tel.: þ49 551 395040.
E-mail address: email@example.com (A.D. Barnes).
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/actoec
1146-609X/Ó2014 Elsevier Masson SAS. All rights reserved.
Acta Oecologica 56 (2014) 32e40
seed traits (Cole, 2009; Dosch et al., 2007; Ingle, 2003; Muller-
Landau et al., 2008; Muñiz-Castro et al., 2006; Teegalapalli et al.,
2010). Seed traits have also been shown to be important for
determining the success of propagules from seed banks, which may
explain further variation in the reassembly of regenerating plant
communities (Pywell et al., 2003). However, despite the wealth of
evidence illustrating the importance of plant traits in mediating
community assembly (Cornwell and Ackerly, 2009; Shipley et al.,
2006), the role that fruit and seed traits play in determining
dispersal of propagules into adjacent regenerating habitats and
their germination potential at the early colonisation stage still re-
quires further investigation (Lebrija-Trejos et al., 2010). This is
especially necessary in African forests where, to our knowledge,
there have been no studies that have identiﬁed the trait de-
terminants of both seed dispersal distances and resulting seedling
establishment with increasing distance from remnant forest sys-
tems into adjacent grassland.
Here we go beyond differentiating between seed size, wind and
animal dispersal (Cubiña and Aide, 2001; Duncan and Chapman,
1999; Muñiz-Castro et al., 2006; Teegalapalli et al., 2010), and in
addition include dispersal traits such as fruit colour and fruit type,
which may affect frugivore choice (Gautier-Hion et al., 1985b) and
even secondary dispersal (Babaasa et al., 2004). To detect the role of
dispersal-linked traits in shaping assembly trajectories of forest
regeneration, we ﬁrst measured the potential for seed dispersal and
seedling establishment from nearby remnant-forest habitat into
grassland recently protected from cattle grazing and burning.
Secondly, we quantiﬁed the dependence of seed rain and seedling
species functional-trait composition on the distance from these
remnant forests. As such, this study aims to provide an indication of
the potential for using dispersal-linked traits in the prediction of
restoration trajectories for future passive-restoration attempts in
2.1. Study site
The study was conducted at Ngel Nyaki Forest Reserve on the
Mambilla plateau in Taraba State, Nigeria. The plateau is part of the
Cameroonian Highlands Ecoregion (Olson et al., 2001). Ngel Nyaki
reserve covers a total area of 4600 ha and includes 750 ha of
continuous forest embedded within a savannah-grassland land-
scape (Beck and Chapman, 2008; Chapman and Chapman, 2001).
The forest is mid-altitude to sub-montane at 1400e1600 m asl
(Chapman and Chapman, 2001), the mean annual rainfall is
approximately 1800 mm (Nigerian Montane Forest Project [NMFP]
rainfall data) and the mean monthly maximum and minimum
temperatures for the wet and dry seasons are 26e13
C, and 23e
C, respectively (Matthesius et al., 2011).
Since the 1950’s, when cattle grazing pressure became severe on
the Mambilla Plateau (Bawden and Tuley, 1966), areas of over-
grazed grassland dominated by Sporobolus pyramidalis and Hyper-
rhenia rufa have been created within Ngel Nyaki forest by the
annual ﬁres lit by Fulani pastoralists to clear forest and stimulate
grass growth in the grasslands around the forest perimeter. Fires
run down grassy spurs leading into the forest and penetrate the
forest edge so that, over time, forest gaps comprising overgrazed
grassland have been created. These grassland areas are predomi-
nantly grassland with a scattering of tall herbs including Dissotis
species (Melastomaceae), Ocimum gratisimum and O. basilicum
(Lamaceae) and Guizotia species (Asteraceae). Small shrubs of
Maesa lanceolata and Psorospermun febrifugum were occasionally
present in all sites. As part of an initiative to promote forest
regeneration, in 2006 we established fences and ﬁre breaks across
the opening of these grassland areas where the pasture penetrates
into the forest to prevent further livestock and ﬁre encroachment.
The grassland sites described in this study had therefore all been
free of grazing and burning for four years. The sites were at least
1000 m apart, located around the perimeter of Ngel Nyaki forest.
2.2. Quantiﬁcation of seed rain and seedling establishment
Sampling was carried out within four grassland areas that
ranged in size from ca 4400 m
to ca 8800 m
throughout Ngel Nyaki forest. Within each of the four areas, ﬁve
sampling transects were established, spaced 10 m apart, at least
10 m from the fence-line and at least 30 m from the forest edge at
the end of the grassland area undergoing restoration (Fig. 1). Each
transect consisted of eight seed traps, spaced equidistantly at 5 m
intervals, from 5 m within the forest edge to 30 m out from the edge
in the adjacent grassland (Fig. 1). The edge (0 m) was deﬁned as the
drip line of the outermost canopy trees at the forest perimeter. Seed
traps consisted of a 0.5 0.5 m piece of mesh netting held 0.3 m
above the ground with a wooden frame in order to prevent the
surrounding vegetation from interfering with seeds falling into the
traps and were protected by chicken wire to prevent the removal of
seeds by seed predators.
All seed traps were visited weekly over a ﬁve year period from
January 2006 to December 2010, upon which all seeds found within
the traps were counted and identiﬁed to the species level. In the
most recent year of sampling (2010), we sampled seedling estab-
lishment across all trap locations whereby all trees and
shrubs 1 m tall and present within 2 m in any direction from a
seed trap (an area of approximately 12.5 m
) were counted and
Fig. 1. Layout of the four sampling-point transects in the regenerating grassland sites.
Values marked on the transects indicate the distance from the forest edge with
negative values used to denote sampling points within the forest and “0”to denote the
A.D. Barnes, H.M. Chapman / Acta Oecologica 56 (2014) 32e40 33
identiﬁed to species. We then calculated the number of seeds and
to obtain a standardised measure of density.
2.3. Measurement of traits and functional dispersion
Traits that may contribute to the dispersal and recruitment
success of a species were obtained from the Nigerian Montane
Forest Project fruit database. Seed size was approximated by seed
diameter in mm and fruit traits included three categorical traits
that characterised the fruit: dispersal mode, colour, and type
(Table A.1). For dispersal mode, we speciﬁed whether the fruit is
ballistic, wind dispersed, gravity dispersed, and small (<10 mm
diameter) or large (>10 mm diameter) endozoochorous. Classiﬁ-
cation of fruit into small and large zoochorous was aimed at dis-
tinguishing between seeds that were dispersed by small and large
animal dispersers (Wheelwright, 1985); speciﬁcally, we assumed
that the large fruit could only be taken by primates and hornbills,
whereas the small zoochorous fruit could be taken by all bird dis-
persers and also primates. Fruit colour was assessed by observation
to give a potential indication of both the attractiveness of fruits to
animal dispersers and the type of animal disperser likely to be
attracted (Gautier-Hion et al., 1985a; Willson and Whelan, 1990).
Fruit type provided a proximate morphological description by
indicating whether the fruit is a capsule, pod, winged, drupe, or
berry (Table A.1).
To determine if there was a community-wide shift in the
dispersion of seed and fruit traits between the forest and adjacent
regenerating grassland, we calculated a distance-based metric of
trait dispersion ‘FDis’(Functional Dispersion)using the “FD”package
(Laliberte and Legendre, 2010) in R 3.0.1 (R Development Core Team,
2013). The FDis metric takes into account multiple traits of species
within a community and measures the distance of each species to
the trait-mean centroid of the whole community. It is a multivariate
adaptation of weighted-mean absolute deviation where the
weighting is given by the relative abundance of species (Laliberte
and Legendre, 2010). Therefore, FDis is a weighted measure of trait
variation among species ina given community. As such, this measure
provided an indication of the potential for trait-based ﬁltering in
seed rain and seedling establishment into the adjacent grassland.
2.4. Statistical analysis
We analysed density, species richness, and functional dispersion
in seed rain and seedling communities as non-linear functions of
distance from the forest edge into the adjacent grassland using
Generalised Additive Mixed Models (Wood, 2011) using the ‘mgcv’
package in R 3.0.1 (R Development Core Team, 2013). Seed-trap
observations were nested within site (n¼4) as random effects in
order to take into account the hierarchical layout of the sampling
design and avoid pseudoreplication (Zuur, 2009). To account for
non-normality and heterogeneity of variance, we modelled re-
sponses of seed and seedling densities and species richness on a
Negative Binomial distribution, which also accounts for over-
dispersion in the data (Zuur, 2009). We optimised smoothing pa-
rameters by selecting models based on the generalised cross-
validation (GCV) criterion, whereby a lower GCV indicates lower
prediction error (Wood, 2011). As the GCV criterion has a tendency
to over-ﬁtting, we applied a penalty on each degree of freedom,
whereby each effective degree of freedom (edf) was forced to be
counted as 1.4 degrees of freedom in the GCV criterion (Kim and Gu,
To test for spatial turnover in species relative abundances for
dispersed seeds and seedling communities, we ﬁrst calculated the
dissimilarity of species composition between sampling points using
a log-base ten Modiﬁed-Gower distance metric (Anderson et al.,
2011) with the ‘vegan’package in R 3.0.1 (R Development Core
Team, 2013). Modiﬁed-Gower dissimilarity considers an order-of-
magnitude change in abundance (e.g., from 0.01 to 0.1) equal to a
change in composition (i.e. from 0 to 1 species), which therefore
accounts for changes in relative abundance of species in addition to
changes in the community composition per se (Anderson et al.,
2006). Compositional dissimilarity between the forest and adja-
cent grassland was visualised using non-metric multidimensional
scaling (NMDS) ordination. We then tested to see if the composi-
tional dissimilarity between samples could be explained by the
distance of the sampling point from the forest using a permuta-
tional distance MANOVA. ‘Distance to the forest edge’was included
as a continuous predictor, with ‘site’(n¼4) speciﬁed as the strata
within which to constrain permutations, thus avoiding pseudor-
eplication resulting from the nested sampling design.
In order to quantify the trait determinants of community as-
sembly with increasing distance from the forest, we adopted a
multivariate approach for categorical variables whereby individuals
were coded by their relative functional trait values as opposed to
their taxonomic classiﬁcation. With these data, we calculated
resemblance matrices derived from the Gower dissimilarity metric
as this measure can deal with categorical variable types and is not
affected by missing values (Laliberte and Legendre, 2010). From
these resemblance matrices, we performed a permutational
MANOVA for each categorical trait (i.e. fruit type, dispersal mode,
and fruit colour), with ‘distance to the forest edge’as a continuous
predictor and ‘site’as a blocking factor, totest the effect of distance
from the forest edge on both seed rain and seedling community
trait-based compositional dissimilarity. For seed diameter, a
continuous variable, we used Generalised Additive Mixed Effects
Models to test for the effect of distance from the forest edge on
community-weighted mean seed diameter with ‘seed-trap obser-
vations’nested within ‘site’as random effects.
3.1. Rapid decline in seed and seedling density across the forest edge
From ﬁve years of seed trapdata we recorded a total of 6332seeds
comprising 31 species of trees and shrubs. We recorded a total of
2010 seedlings from 30 species across all sampling transects
(Table 1). 71% of these seedling species were also present in the seed
trap data. There was a signiﬁcant decline in seed-rain density from
the remnant forest into the grassland (edf ¼5.701; P<0.001), with a
98% average decrease in seeds per m
from the forest edge to 30 m
into the grassland (Fig. 2a). This considerable decline in seed density
with increasing distance into the grassland was mirrored by a 96%
decrease in seedling densities (edf ¼6.177; P<0.001; Fig. 2b). These
observed declines were only evident up to 10 m into the grassland,
reaching asymptote beyond this point. Likewise, our results showed
that the number of species from the seed rain data also declined
strongly with increasing distance from the forest (a 92% decrease;
edf ¼1.4 2 2; P<0.001; Fig. 2c) and the same trend was followed by
the number of seedling species (an 83% decrease; edf ¼1; P<0.001;
Fig. 2d). In contrastto the density responses, species richnessshowed
a more continuous decline with increasing distance from the edge,
without reaching any clear asymptote within the 30 m range
sampled. This was especially evident in seedling species richness,
which showed no departure from a linear relationship (edf ¼1).
3.2. Distance from source populations drives species composition
and a decline in functional dispersion
The NMDS visualisation indicated that the decrease in seed and
colonising plant densities with increasing distance from the forest
A.D. Barnes, H.M. Chapman / Acta Oecologica 56 (2014) 32e4034
was not equally distributed among species. This was suggested by
the gradient of increasing compositional dissimilarity from the
forest to adjacent grassland (Fig. B.1), indicating an overall shift in
species’relative abundances for seed rain and seedling commu-
nities. Furthermore, the permutational multivariate ANOVA
revealed a highly signiﬁcant effect of distance from the forest edge
on the dissimilarity of seed species composition (F¼6.574,
P<0.001) and the same was found for seedling communities
(F¼12.390, P<0.001). Interestingly, we found that distance from
the forest explained considerably more variation in seedling
¼0.127) compared to seed rain composition
With the signiﬁcant decline in seed rain and seedling estab-
lishment away from the forest, we also found a decline in the range
of trait composition. In particular, there was a signiﬁcant decline in
functional trait dispersion (FDis) from the forest to the adjacent
grassland (edf ¼4.182; P<0.001; Fig. 3a). While this was also the
case for seedling communities (edf ¼1; P<0.001; Fig. 3b), the
degree of change in the functional trait dispersion of communities
was greater for seed rain composition than for establishing plant
communities (86% compared to only 75% decline for seed rain and
seedlings, respectively), with the greatest rate of change in FDis
within 10 m of the forest edge for the seed rain.
3.3. Trait-based ﬁltering of propagules determines composition of
colonising plant communities
Community-weighted mean seed diameter did not signiﬁcantly
respond to the distance from forest edge for either seed rain
(edf ¼1; P¼0.749) or seedling communities (edf ¼1.507;
P¼0.331) and appeared to vary idiosyncratically with increasing
distance from the edge (Fig. B.2a and b). However, for all three
categorical traits (i.e. fruit type, dispersal mode, and colour) there
were clear changes in the relative proportion of traits with
increasing distance from the forest edge. In particular, fruit
dispersal mode in the seed rain data signiﬁcantly responded to
distance from the forest (F¼7.420, R
¼0.078, P<0.001) with
wind, small zoochorous and large zoochorous fruited species such
as Combretum molle, Bridelia speciosa, and Landolphia landolphioides
respectively, having the highest dispersal potential beyond w20 m
(Fig. 4c, Table 1). Furthermore, there was a signiﬁcant effect of
distance on seed rain community dissimilarity based on fruit type
¼0.094, P<0.001) and colour (F¼5.306, R
P<0.001), whereby seed from drupes and capsules were most
likely to reach distances up to 30 m, with red and blue fruits
having the highest frequency of dispersal to greater distances
(Fig. 3e and g).
List of all species and associated families found in samples with the total number of individuals recorded from seed traps and seedling plots for distance categories: 5to0m,
5e15 m, and 20e30 m from the forest edge.
Family Species Seeds Seedlings
5 to 0 m 5 to 15 m 20 to 30 m 5 to 0 m 5 to 15 m 20 to 30 m
Mimosaceae Albizia gummifera 4541
Caesalpiniaceae Anthonotha noldeae 48 9 43 23 13
Euphorbiaceae Bridelia speciosa 2474 126 355 153 48 20
Meliaceae Carapa oreophylla 51 1
Cannabaceae Celtis gomphophylla 1
Oleaceae Chionanthus africanus 1
Rutaceae Clausena anisata 100 1 9 241 3
Combretaceae Combretum molle 66 31 4 1 2
Euphorbiaceae Croton macrostachyus 1
Sapindaceae Deinbollia pinnata 10 60
Ebenaceae Diospyros sp.127 595 5
Malvaceae Dombeya ledermannii 53 45 1
Meliaceae Entandrophragma angolense 22 1 96
Mimosaceae Entada abyssinica 25 3 1
Myrtaceae Eugenia gilgii 23 4
Clusiaceae Garcinia smeathmanii 7
Annonaceae Isolona deightonii 95 7
Apocynaceae Landolphia landolphioides 270 22 3
Leeaceae Leea guineensis 137 8 150 16 1
Myrsinaceae Maesa lanceolata 274 4
Mimosaceae Newtonia buchananii 61 35 123 13
Buddlejaceae Nuxia congesta 111 9 4 10 2
Rubiaceae Oxyanthus recemosus 15
Mimosaceae Parkia ﬁlicoidea 1
Sapindaceae Paullinia pinnata 40 13
Araliaceae Polyscias fulva 127 2 10 2
Sapotaceae Pouteria altissima 129
Clusiaceae Psorospermun febrifugum 60533885018
Rubiaceae Psychotria sp.435 28 4
Rubiaceae Psychotria succulenta 413 5 107 48 6 146
Apocynaceae Rauvolﬁa vomitoria 7
Malvaceae Sterculia setigera 16
Olacaceae Strombosia schefﬂeri 16
Clusiaceae Symphonia globulifera 7
Sapotaceae Synsepalum sp.3
Myrtaceae Syzygium guineense
sub sp. guineense
276 8 3 24 8 1
Ulmaceae Trema orientalis 103 96
Moraceae Trilepesium madagascariense 22
Rutaceae Zanthoxylum leprieurii 1
A.D. Barnes, H.M. Chapman / Acta Oecologica 56 (2014) 32e40 35
In contrast to the seed rain data we found that along with small-
zoochorous and wind-dispersed propagules, ballistic propagules
(mainly from the species Anthonotha noldeae and Dombeya leder-
mannii) also made it to larger distances and successfully germi-
nated (Fig. 3d, Table 1) with an overall signiﬁcant effect of distance
from the forest on relative abundances of fruit dispersal modes
¼0.153, P<0.001). We also found highly signiﬁcant
effects of the distance from the forest edge on trait-based compo-
sitional dissimilarity for fruit type (F¼12.702, R
P<0.001) and colour (F¼11.067, R
¼0.115, P<0.001) in seedling
communities. Whilst dispersed seedlings from drupes were still the
most frequent across all distances, we found that seedlings from
species such as A. noldeae and C. molle with pods and winged
propagules were also present at greater distances, albeit in low
Fig. 3. Generalised additive mixed models demonstrating the relationships between distance from the forest edge and functional trait dispersion (FDis) for seed-rain (a) and
seedling communities (b). Smoothed lines are ﬁtted values and shaded area is the 95% conﬁdence interval. R
is the proportion of explained deviance. Negative and positive x-axis
values denote forest and matrix, respectively, with 0 to denote the edge.
Fig. 2. Generalised additive mixed models demonstrating the relationships between distance from the forest edge and density (abundance m
) of seeds (a) and seedlings (b), and
the number of species recorded from the seed-rain (c) and seedling data (d). Smoothed lines are ﬁtted values and shaded area is the 95% conﬁdence interval. R
is the proportion of
explained deviance. Negative and positive x-axis values denote forest and matrix, respectively, with 0 to denote the edge.
A.D. Barnes, H.M. Chapman / Acta Oecologica 56 (2014) 32e4036
frequencies (Fig. 3f, Table 1). However, seedlings from red and blue
fruits remained the most frequent across all distances from the
4.1. What is the likelihood of success for passive restoration
attempts in Afromontane forests?
Our study demonstrates that dispersal-linked traits mediate the
reassembly trajectory of regenerating plant communities under-
going passive restoration following severe human-induced land-
scape degradation. Through a combination of ecological ﬁlters such
as dispersal limitation and factors inhibiting germination, our re-
sults suggest that simply protecting cleared areas of West African
montane forest from burning and cattle grazing will not lead
directly to the rapid reassembly of forest communities, but instead
to communities dominated by tree and shrub species characterised
by small, ﬂeshy, red or blue drupes mainly dispersed by birds and
primates. However, mostly within this functional group we found
much more potential for forest regeneration in seed rain and early
seedling establishment than has been recorded elsewhere in Africa
(Duncan and Chapman, 1999). We found seedlings in the grassland
up to 30 m away from the forest edge of grassland trees and shrubs
such as D. ledermannii and Psorospermun febrifugum, forest-edge
tree species including Eugenia gilgii and Nuxia congesta and the
large, leguminous forest tree species Albizia gummifera and
Of those seed species recovered from the traps, only 71% were
present as seedlings, and of those seedlings found beyond ﬁve m
−5 0 5 10 15 20 25 30
Distance from edge (m)
−5 0 5 10 15 20 25 30
Distance from edge (m)
Fig. 4. Relative proportions of dispersal mode categories for seed rain (a) and seedlings (b), frequencies of fruit type categories for seed rain (c) and seedlings (d), and frequencies of
fruit colours for seed rain (e) and seedling communities (f). Legend abbreviations for dispersal categories are: ballistic (B), gravity (G), large zoochorous (LZ), small zoochorous (SZ),
and wind (W). Fruit types are: berry (B), capsule (C), drupe (D), pod (P), and winged (W). Proportion areas for fruit colour are indicated by their actual colour. X-axis values of 5 and
0 denote 5 m within the forest and the forest edge, respectively.
A.D. Barnes, H.M. Chapman / Acta Oecologica 56 (2014) 32e40 37
from the forest edge, over 90% germinated from small, ﬂeshy
drupes. The 29% of seed species found in traps but never as seed-
lings comprised a wide range of seed types and their apparent lack
of regeneration is presumably explained by a combination of post
dispersal factors such as unsuitable microsites and seed predators
(Chapman and Chapman, 1999), competition from grasses such as
the introduced Sporobilis sp. and Hyperrhenia sp.(Chapman and
Chapman, 2001) and drought during the six month dry season
(NMFP rainfall records). Additionally, while there may be sufﬁcient
seed dispersal and even germination of forest plant species, there
are likely to be other barriers limiting their recruitment, such as the
requirement of larger established trees for colonizing vine species
like Landolphia landolphioides (Table 1).
Seed rain density declined dramatically only just beyond the
forest edge, indicating that the matrix is important for mediating
rates of dispersal across the forestegrassland interface (Barnes
et al., 2014; Holl, 2012). Relatively high numbers of dispersed
seed were still found 5 m out from the forest edge, but beyond this
point seed densities dropped to only 2% of the densities found
under the forest canopy within 5 m of the forest edge. While
densities of animal dispersed seeds appeared to remain relatively
constant beyond 5 m out from the forest edge, even at the
maximum distance measured of 30 m, this was not the case for
wind dispersed seed which had dropped off dramatically ten m into
the grassland. Such a rapid decrease in density of wind dispersed
seed is to be expected and has been recorded in previous studies
from the Neotropics (Cubiña and Aide, 2001; Holl, 1999) and dry
forest in India (Teegalapalli et al., 2010). In regard to animal
dispersed seed our ﬁndings suggests that in these regenerating
grasslands, which are often no wider than 60 m, regardless of the
distance from the forest edge we can expect equal densities of seed
rain from the adjacent forest.
In addition to the decline in seed and seedling densities, we
found that there was signiﬁcant turnover in species composition
with increasing distance from the forest edge, despite the relatively
ﬁne-scale gradient of 5 m incremental changes in distance. For
example, seedlings of pioneer species with small, ﬂeshy fruit such
as Bridelia speciosa and Psychotria succulenta became relatively
more common with distance from the forest edge relative to forest
species such as Pouteria altissima and Deinbollia pinnata, which
have larger, green fruit. While it is possible that recruitment is
strongly limited after seedlings germinate due to factors such as
those described above (Duncan and Chapman, 1999; Holl, 1999;
Nepstad et al., 1990) our ﬁndings suggest that in naturally regen-
erating West African montane systems, dispersal limitation may be
playing an important role in the early stages of community reas-
sembly, even across small spatial scales.
4.2. Can fruit and seed traits predict restoration outcomes?
From these ﬁndings, the question arises as to which factors are
determining the varying levels of dispersal limitation. The answer
may be pivotal to understanding how forest communities are likely
to reassemble during passive restoration efforts. We found that
from the four selected dispersal-linked trait measures used, in
almost all instances they changed in relative frequency with
increasing distance away from the parent populations. Most
obvious was the ﬁnding that seed from fruits that were small-
zoochorous red drupes were dispersed relatively more often and
in greater quantities across most distances and were found at least
up to distances of 30 m more frequently than any other propagules.
These results point to the importance of traits that are linked to
avian dispersal as these would be among the most attractive fruit to
such dispersers (Duncan and Chapman, 1999; Willson and Whelan,
1990). However, birds are known to be extremely habitat selective
and have been found not to contribute usefully to seed rain in forest
regeneration elsewhere in Africa (Duncan and Chapman, 1999).
While investigation of the role of birds in dispersing seeds into
these grasslands is currently underway at Ngel Nyaki, the fact that
such fruits are also dispersed by several primate species including
tantalus monkeys (Chlotocebus tantalus), which regularly enter and
feed in grassland (Agmen et al., 2010), may explain the pattern of
seed rain we observed. Moreover, our ﬁnding that seed diameter
appeared to be of little importance for the dispersal potential of
seeds and germination of seedlings with increasing distance from
the forest edge is also most likely explained by primate dispersal as
the wider gape of primates relative to passerine birds allow them to
swallow larger seed (Jordano et al., 2007). While such ﬁndings are
also in contrast to previous Neotropical studies (Cubiña and Aide,
2001; Holl, 1999) they are similar to those of Teegalapalli et al.
(2010) who investigated the dispersal of large seed from Indian
dry forest into pasture related to grazing patterns of large
A particularly interesting ﬁnding was the relatively high density
of seedlings from ballistic-dispersed seeds across larger distances
from the forest, despite relatively lower numbers within the forest
sampling points. For example the very large, nutritious seeds of
Anthonotha noldeae were rarely found in seed traps beyond 5 m into
the grassland, yet seedlings were found up to 30 m away from the
forest edge. Such a pattern is indicative of dispersal by secondary
seed dispersers such as scatter-hoarding rodents (Nyiramana et al.,
2011). While not included in this study, we have observed a similar
phenomenon in Carapa oreophylla whose seedlings establish many
metres from the forest edge (personal observation). Wind was also
found to be a successful mode of dispersal for trees to colonise
these grasslands as both seed and/or germinating seedlings from
wind/winged classed species, such as Albizia gummifera,Com-
bretum molle and Entandrophragma angolense, were found at low
but relatively consistent numbers across all distances.
4.3. Applications for trait-based theory in habitat restoration
By quantifying patterns in functional trait dispersion and
changes in community trait composition of seeds and seedlings
across the interface of remnant forest and passively regenerating
grassland, this study shows that regenerating plant communities
undergo trait-based ﬁltering, which appears to increase in intensity
with increasing distance from the forest. Despite the strong
ﬁltering of species dispersing into these adjacent grasslands, the
somewhat high density and diversity of germinating seedlings in-
dicates that there is still promise for passive restoration efforts in
Afromontane forest landscapes. It must be noted, however, that this
study has only taken into account a relatively short time-scale of
regeneration time (over just 5 years) and therefore provides insight
into the initial recolonisation process. It is likely that over a much
greater time-scale there may be rare, chance dispersal events that
result in the dispersal of important pioneer species into these
regenerating communities which may in fact catalyse the restora-
tion process (Rodrigues et al., 2004). Furthermore, there is evidence
in the Afrotropics that regeneration can be arrested due to dispersal
limitation and/or the competitive dominance of other non-forest
species (Babaasa et al., 2004; Duncan and Chapman, 1999). Still,
our study provides a characterisation of the early recolonisation
processes that are likely to take place in passive restoration at-
tempts, which is important for identifying the mechanisms that can
potentially lead to undesirable restoration trajectories.
While previous studies have shown that plant traits play an
important role in later phases of community assembly within
passively restored habitats (Muñiz-Castro et al., 2006; Pywell et al.,
2003) we still lack an understanding of how these traits determine
A.D. Barnes, H.M. Chapman / Acta Oecologica 56 (2014) 32e4038
community reassembly at the initial stages of habitat regeneration.
This study highlights the value of quantifying these trait de-
terminants so that they can be utilised to gain a priori knowledge of
the potential for success and likely trajectory of passive restoration
efforts, depending on the functional-trait composition of nearby
parent populations. Therefore, we recommend that future attempts
to passively restore tropical forests should ﬁrst quantify the po-
tential for species to colonise target-restoration areas from nearby
remnant forests by utilising the ‘library’of functional traits found
within these communities.
We thank the Taraba State Forest Service for inviting us to work
in Ngel Nyaki Forest, Usman Abubakar for assistance in the ﬁeld,
and the other Nigerian Montane Forest Project staff for logistical
support at the NMFP ﬁeld station. Kristy Udy, Laura Young, and the
Chapman lab group provided invaluable comments on earlier drafts
of the manuscript. We also thank two anonymous reviewers for
providing comments and suggestions that substantially improved
this paper. The study was funded by the North of England
Zoological Society (Chester zoo), Nexen Inc. and the A. G. Leventis
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://
Agmen, F.L., Chapman, H.M., Bawuro, M., 2010. Seed dispersal by tantalus monkeys
(Chlorocebus tantalus tantalus) in a Nigerian montane forest. Afr. J. Ecol. 48,
112 3e112 8 .
Anderson, M.J., Crist, T.O., Chase, J.M., Vellend, M., Inouye, B.D., Freestone, A.L.,
Sanders, N.J., Cornell, H.V., Comita, L.S., Davies, K.F., Harrison, S.P., Kraft, N.J.B.,
Stegen, J.C., Swenson, N.G., 2011. Navigating the multiple meanings of
versity: a roadmap for the practicing ecologist. Ecol. Lett. 14, 19e28.
Anderson, M.J., Ellingsen, K.E., McArdle, B.H., 2006. Multivariate dispersion as a
measure of beta diversity. Ecol. Lett. 9, 683e693.
Asner, G.P., Rudel, T.K., Aide, T.M., Defries, R., Emerson, R., 2009.
A contemporary assessment of change in humid tropical forests. Conserv.
Biol. 23, 1386e1395.
Babaasa, D., Eilu, G., Kasangaki, A., Bitariho, R., McNeilage, A., 2004. Gap charac-
teristics and regeneration in Bwindi Impenetrable National Park, Uganda. Afr. J.
Ecol. 42, 217e224.
Barnes, A.D., Emberson, R.M., Chapman, H.M., Krell, F.-T., Didham, R.K., 2014. Matrix
habitat restoration alters dung beetle species responses across tropical forest
edges. Biol. Conserv. 170, 28e37.
Bawden, M.G., Tuley, P., 1966. The Land Resources of Southern Sardauna and
Southern Adamawa Provinces, Northern Nigeria (Land Resource Study No. 2).
Directorate of Overseas Surveys, Tolworth, Surrey.
Beck, J., Chapman, H., 2008. A population estimate of the endangered chimpanzee
Pan troglodytes vellerosus in a Nigerian montane forest: implications for con-
servation. Oryx 42, 448e451.
Benayas, J.M.R., Newton, A.C., Diaz, A., Bullock, J.M., 2009. Enhancement of biodi-
versity and ecosystem services by ecological restoration: a meta-analysis. Sci-
ence 325, 1121e11 24.
Chapman, C.A., Chapman, L.J., 1999. Forest restoration in abandoned agricultural
land: a case study from East Africa. Conserv. Biol. 13, 1301e1311.
Chapman, J.D., Chapman, H.M., 2001. The Forests of Taraba and Adamawa States,
Nigeria: an Ecological Account and Plant Species Checklist. University of Can-
terbury, Christchurch NZ.
Chazdon, R.L., 2008. Beyond deforestation: restoring forests and ecosystem services
on degraded lands. Science 320, 1458e1460.
Cole, R.J., 2009. Postdispersal seed fate of tropical montane trees in an agricultural
landscape, Southern Costa Rica. Biotropica 41, 319e327.
Cole, R.J., Holl, K.D., Keene, C.L., Zahawi, R.A., 2011. Direct seeding of late-
successional trees to restore tropical montane forest. For. Ecol. Manag. 261,
Cornwell, W.K., Ackerly, D.D., 2009. Community assembly and shifts in plant trait
distributions across an environmental gradient in coastal California. Ecol.
Monogr. 79, 109e126.
Cubiña, A., Aide, T.M., 2001. The effect of distance from forest edge on seed rain and
soil seed bank in a tropical pasture. Biotropica 33, 260e267.
del Castillo, R.F., Ríos, M.A.P., 2008. Changes in seed rain during secondary suc-
cession in a tropical montane cloud forest region in Oaxaca, Mexico. J. Trop.
Ecol. 24, 433e444.
Dosch, J.J., Peterson, C.J., Haines, B.L., 2007. Seed rain during initial colonization of
abandoned pastures in the premontane wet forest zone of southern Costa Rica.
J. Trop. Ecol. 23, 151e159.
Duncan, R.S., Chapman, C.A., 1999. Seed dispersal and potential forest succession in
abandoned agriculture in tropical Africa. Ecol. Appl. 9, 998e1008.
Eriksson, O., Ehrlén, J., 1992. Seed and microsite limitation of recruitment in plant
populations. Oecologia 91, 360e364.
Gautier-Hion, A., Duplantier, J.-M., Quris, R., Feer, F., Sourd, C., Decoux, J.-P.,
Dubost, G., Emmons, L., Erard, C., Hecketsweiler, P., 1985a. Fruit characters as a
basis of fruit choice and seed dispersal in a tropical forest vertebrate commu-
nity. Oecologia 65, 324e337.
Gautier-Hion, A., Duplantier, J.M., Quris, R., Feer, F., Sourd, C., Decoux, J.P., Dubost, G.,
Emmons, L., Erard, C., Hecketsweiler, P., Moungazi, A., Roussilhon, C.,
Thiollay, J.M.,1985b. Fruit characters as a basis of fruit choice and seed dispersal
in a tropical forest vertebrate community. Oecologia 65, 324e337.
Geist, H.J., Lambin, E.F., 2002. Proximate causes and underlying driving forces of
tropical deforestation. Bioscience 52, 143e150.
Holl, K.D., 1999. Factors limiting tropical rain forest regeneration in abandoned
pasture: seed rain, seed germination, microclimate, and soil. Biotropica 31,
Holl, K.D., 2012. Tropical forest restoration. In: Andel, J.V., Aronson, J. (Eds.),
Restoration Ecology. Blackwell Publishing, Malden, MA, pp. 103e114.
Holl, K.D., Aide, T.M., 2011. When and where to actively restore ecosystems? For.
Ecol. Manag. 261, 1558e1563.
Ingle, N.R., 2003. Seed dispersal by wind, birds, and bats between Philippine
Montane rainforest and successional vegetation. Oecologia 134, 251e261.
Jordano, P., García, C., Godoy, J.A., García-Castaño, J.L., 2007. Differential contribution
of frugivores to complex seed dispersal patterns. Proc. Natl. Acad. Sci. 104,
Kalesnik, F., Sirolli, H., Collantes, M., 2013. Seed bank composition in a
secondary forest in the Lower Delta of the Paraná River (Argentina). Acta Bot.
Bras. 27, 40e49.
Kim, Y.-J., Gu, C., 2004. Smoothing spline Gaussian regression: more scalable
computation via efﬁcient approximation. J. Royal Stat. Soc. Ser. B Stat. Methodol.
Laing, J.M., Shear, T.H., Blazich, F.A., 2011. How management strategies have affected
Atlantic White-cedar forest recovery after massive wind damage in the Great
Dismal swamp. For. Ecol. Manag. 262, 1337e1344.
Laliberte, E., Legendre, P., 2010. A distance-based framework for measuring func-
tional diversity from multiple traits. Ecology 91, 299e305.
Lamb, D., Erskine, P.D., Parrotta, J.A., 2005. Restoration of degraded tropical forest
landscapes. Science 310, 1628e1632.
Lebrija-Trejos, E., Pérez-García, E.A., Meave, J.A., Bongers, F., Poorter, L., 2010.
Functional traits and environmental ﬁltering drive community assembly in a
species-rich tropical system. Ecology 91, 386e398.
Martínez-Garza, C., Howe, H.F., 2003. Restoring tropical diversity: beating the time
tax on species loss. J. Appl. Ecol. 40, 423e429.
Matthesius, A., Chapman, H., Kelly, D., 2011. Testing for Janzen-Connell effects in a
West African montane forest. Biotropica 43, 77e83.
Morrison, E.B., Lindell, C.A., 2011. Active or passive forest restoration?
Assessing restoration alternatives with avian foraging behavior. Restor. Ecol.19,
Muller-Landau, H.C., Wright, S.J., Calderón, O., Condit, R., Hubbell, S.P., 2008.
Interspeciﬁc variation in primary seed dispersal in a tropical forest. J. Ecol. 96,
Muñiz-Castro, M.A., Williams-Linera, G., Benayas, J.M.R., 2006. Distance effect from
cloud forest fragments on plant community structure in abandoned pastures in
Veracruz, Mexico. J. Trop. Ecol. 22, 431e440.
Myster, R., 2004. Regeneration ﬁlters in post-agricultural ﬁelds of Puerto Rico and
Ecuador. Plant Ecol. 172, 199e209.
Nepstad, D., Uhl, C., Serrao, E.A., Anderson, A., 1990. Surmounting Barriers to Forest
Regeneration in Abandoned, Highly Degraded Pastures: a Case Study from
Paragominas, Pará, Brazil. In: Alternatives to Deforestation: Steps towards
Sustainable Use of the Amazon Rain Forest, pp. 215e229.
Nyiramana, A., Mendoza, I., Kaplin, B.A., Forget, P.-M., 2011. Evidence for
seed dispersal by rodents in tropical montane forest in Africa. Biotropica 43,
Olson, D.M., Dinerstein, E., Wikramanayake, E.D., Burgess, N.D., Powell, G.V.N.,
Underwood, E.C., D’Amico, J.A., Itoua, I., Strand, H.E., Morrison, J.C., Loucks, C.J.,
Allnutt, T.F., Ricketts, T.H., Kura, Y., Lamoreux, J.F., Wettengel, W.W., Hedao, P.,
Kassem, K.R., 2001. Terrestrial ecoregions of the world: a new map of life on
Earth. BioScience 51, 933e938.
Parrotta, J.A., Turnbull, J.W., Jones, N.,1997. Catalyzing native forest regeneration on
degraded tropical lands. For. Ecol. Manag. 99, 1e7.
Pywell, R.F., Bullock, J.M., Roy, D.B., Warman, L., Walker, K.J., Rothery, P., 2003. Plant
traits as predictors of performance in ecological restoration. J. Appl. Ecol. 40,
R Development Core Team, 2013. A Language and Environment for Statistical
Computing. R Foundation for Statistical Computing, Vienna, Austria.
Rodrigues, R.R., Martins, S.V., de Barros, L.C., 2004. Tropical rain forest regeneration
in an area degraded by mining in Mato Grosso state, Brazil. For. Ecol. Manag.
A.D. Barnes, H.M. Chapman / Acta Oecologica 56 (2014) 32e40 39
Shipley, B., Vile, D., Garnier, É., 2006. From plant traits to plant communities: a
statistical mechanistic approach to biodiversity. Science 314, 812e814.
Teegalapalli, K., Hiremath, A.J., Jathanna, D., 2010. Patterns of seed rain and seedling
regeneration in abandoned agricultural clearings in a seasonally dry tropical
forest in India. J. Trop. Ecol. 26, 25e33.
Wheelwright, N.T., 1985. Fruit-size, gape width, and the diets of fruit-eating birds.
Willson, M.F., Whelan, C.J., 1990. The evolution of fruit color in ﬂeshy-fruited plants.
Am. Nat., 790e809.
Wood, S.N., 2011. Fast stable restricted maximum likelihood and marginal likeli-
hood estimation of semiparametric generalized linear models. J. Royal Stat. Soc.
Ser. B Stat. Methodol. 73, 3e36.
Zuur, A.F., 2009. Mixed Effects Models and Extensions in Ecology with R. Springer,
A.D. Barnes, H.M. Chapman / Acta Oecologica 56 (2014) 32e4040