Ecology Letters, (2009) 12: 220–228
doi: 10.1111/j.1461-0248.2008.01278.x
LETTER
Size of sampling unit strongly influences detection
of seedling limitation in a wet tropical forest
Richard K. Kobe1* and Corine
F. Vriesendorp1,2
1
Michigan State University
Department of Forestry and
Graduate Program in Ecology,
Evolutionary Biology and
Behavior Natural Resources
Building, East Lansing, MI
48824-1222, USA
2
Environment, Culture and
Conservation, Field Museum of
Natural History, Chicago, IL
60605, USA
*Correspondence: Email:
Abstract
Seedling limitation could structure communities, but often is evaluated with sampling
units that are orders of magnitude smaller than mature plants. We censused seedlings for
5.5 years in five 1 · 200-m transects in a wet Neotropical forest. For 106 common
species (‡ 10 seedlings in a transect), we calculated prevalence (occurrence of ‡ 1 newly
emerged seedlings per sampling unit) at 1 m2 and at 1 m · mature crown diameter units
by aggregating adjacent quadrats. For most species, prevalence was 2–25% at 1 m2, but
20–92% at mature crown scales. Increased prevalence arose from broadly distributed
seedlings within transects, with unoccupied segments generally shorter than crown
diameters. At the landscape scale, 69% of 301 species were locally rare (< 10 seedlings)
and only 16% were represented in all transects (maximally separated by 2.4 km).
Nonetheless, for more common species, much lower estimates of seedling limitation at
mature crown scales suggest weaker influence of seedling limitation on community
dynamics than previously assumed.
[email protected]
Keywords
Competitive exclusion, dispersal limitation, establishment, forest dynamics, recruitment,
recruitment limitation, seedling limitation, seedlings, spatial scale, species coexistence,
species diversity, tropical forest.
Ecology Letters (2009) 12: 220–228
INTRODUCTION
Recruitment limitation (RL) – the failure of a species to have
juveniles at an available site – could contribute to the
maintenance of plant species diversity by allowing inferior
competitors to win local sites by forfeit, effectively slowing
community dynamics (Hurtt & Pacala 1995) and, by
extension, delaying competitive exclusion. Several studies
have suggested that RL structures tropical (Hubbell et al.
1999; Dalling et al. 2002; Muller-Landau et al. 2002; Denslow et al. 2006; Comita et al. 2007; Norden et al. 2007) and
temperate (Ribbens et al. 1994; Clark et al. 1998) forests and
grasslands (Tilman 1997; Zobel et al. 2000; Seabloom et al.
2003). Seedling RL could arise from any of its components,
including adult fecundity or density, distance to seed sources
and dispersal, germination and initial seedling establishment
(Clark et al. 1998).
Empirical support for RL and its components are derived
from three types of studies. First, seed traps and seedling
plots have been used to quantify the percentage of sampling
units in which a species occurs. For example, only seven of
2008 Blackwell Publishing Ltd/CNRS
260 species dispersed ‡ 1 seed into > 75% of 200 seed traps
(of 0.5 m2) over a 10-year period on Barro Colorado Island
(BCI), Panama and 50% of species were present in six or
fewer traps; the most commonly encountered species
occurred in only 14.9% of 1-m2 seedling quadrats (Hubbell
et al. 1999). Pulses in seed arrival can strongly influence the
temporal variation in seedling community structure (Norden
et al. 2007). Seed-addition experiments also have provided
support for seed limitation; about half of the experiments
reviewed by Turnbull et al. (2000) found significant increases
in seedling population density in response to seed additions,
but as pointed out by Clark et al. (2007) only a small fraction
of added seeds actually recruit to the seedling stage,
suggesting the importance of establishment constraints.
Finally, dispersal kernels reconstructed through inverse
modelling (based on seed and seedling densities and
distances to potential parents) suggest dispersal distances
of < 30 m (Ribbens et al. 1994; Clark et al. 1999b), but
longer dispersal distances are supported by genetic reconstructions (Jones & Muller-Landau 2008) and direct
observations of dispersal by animals (Stevenson 2000;
Letter
Spatial scale of seedling limitation 221
Wehncke et al. 2003; Russo et al. 2006). Discordance in
seedling and mature tree species presence in nested plots
also supports active dispersal rather than passive dropping
of seeds (Webb & Peart 2001).
A rigorous test of RL should specify the limiting stage
(typically seed or seedling) and the subsequent stage to
which recruitment is limited (typically the adult plant)
(Muller-Landau et al. 2002). Moreover, the appropriate
spatial scale at which to measure RL depends on the size
of the organism at the life-cycle stage of interest (MullerLandau et al. 2002). In grassland ecosystems, 1-m2 sampling
plots would encompass the size of a mature plant (e.g.
Tilman 1997). In forests, however, if inference extends to
mature tree recruitment, then the appropriate spatial scale at
which to measure seed or seedling limitations is the crown
area or diameter occupied by a mature tree (Muller-Landau
et al. 2002). For example, Ribbens et al. (1994) used fieldcalibrated models of seedling dispersion to populate
5 m · 5 m cells, the space occupied by most mature
canopy trees in their study site, and then calculated the
percentage of unoccupied cells.
In studies that characterize RL as the percentage of
sampling units in which a speciesÕ seed and ⁄ or seedling
occurs, sampling units are typically small and are not
amenable to aggregation to coarser spatial scales (Fig. 1a).
For instance, the long-term BCI seed trap study has
contributed a wealth of data on seed production and
implications for seedling recruitment (Hubbell et al. 1999;
Harms et al. 2000; Wright et al. 2005), but the traps are small
(0.5 m2) and separation by c. 19 m along trails precludes
straightforward aggregation (Wright et al. 2005). Clearly
these studies were aware of the importance of spatial scale in
assessing seedling limitation (Muller-Landau et al. 2002), but
the experimental set-up likely rules out the explicit inclusion
of spatial scale in analyses of seed and seedling limitation.
In this paper, we compare seedling limitation at the
typical 1-m2 quadrat level and at spatial scales commensurate with the crown diameters of mature plants (hereforth
called ÔorganismÕ scale). Spatial relationships between
potential parents and seedling offspring will be reported in
a separate paper (R.K. Kobe and C.F. Vriesendorp,
unpublished data). To test the potential influence of seedling
limitation on mature plant composition, we examined
seedling limitation at a spatial scale commensurate with
adult size because the presence of only one seedling within
the area occupied by an adult is needed to replace that adult.
Seedling prevalence is related to fundamental seedling
limitation (FSL) as FSL = 1 ) (sampling units occupied
by seedlings ⁄ total number of sampling units) (Nathan &
Muller-Landau 2000; Muller-Landau et al. 2002), where the
parenthetical term is the seedling prevalence. We used FSL
and seedling prevalence in order to conceptually separate
seedling arrival, which is largely stochastic, from growth and
survival, which are more strongly influenced by resource
(a)
Figure 1 The sampling units in seedling
limitation studies are often much smaller
than the mature tree that a seedling might
someday replace. (a) Transects used for
seedling censuses consisted of 200 contiguous 1 m2 quadrats. The typical sampling unit
used in studies of recruitment limitation is
£ 1 m2. However, the spatial scale of mature
plants is much larger as illustrated for a 20-m
crown diameter canopy tree. This study
evaluated seedling prevalence at both 1 m2
and mature plant size spatial scales. (b) The
spatial distribution of occupied 1-m2 quadrats will influence the calculation of prevalence at coarser spatial scales. In both
transects, 5 of 40 1-m2 quadrats are occupied
(prevalence1m = 0.125), but because occupied 1-m2 quadrats are clumped in the upper
but broadly distributed in the lower transect,
prevalence8m = 0.25 in the upper and 1.0 in
the lower transect.
. . . to 200 meters
Typical sampling unit
(1 meter)
Linear distance occupied by crown diameter
(e.g., 20 meters for canopy tree)
(b)
occ.
vac.
vac.
vac.
occ.
occ.
occ.
occ.
= 1m unit occupied
occ. = 8m unit occupied
vac. = 8m unit vacant
2008 Blackwell Publishing Ltd/CNRS
222 R. K. Kobe and C. F. Vriesendorp
availability. At 6-week intervals from March 2000–October
2005, we recorded all newly germinating seedlings of woody
species in five belt transects of 200 contiguous 1-m2
quadrats (tagging > 19 k seedlings over 5.5 years). We
assessed the effects of spatial scale on FSL by aggregating
adjacent quadrats into sampling units of larger sizes that
were commensurate with the crown diameter of mature
plants (Fig. 1a). Although seedling prevalence was expected
to increase with sampling unit size, it is important to assess
seedling RL at the spatial scale of the organism that the
seedling might someday replace in order to more rigorously
evaluate the role of seedling limitation in plant community
dynamics.
METHODS
Field site and sampling
We conducted this study in tropical wet forest at La Selva
Biological Station in the Atlantic lowlands of Costa
Rica (1026¢ N, 8400¢ W). La Selva is a 1510-ha forest
bordered by the Puerto Viejo and Sarapiqui Rivers and
contiguous with Braulio Carillo National Park. Mean annual
rainfall is 3859 mm, with a mean monthly minimum of
100 mm (http://www.ots.duke.edu/en/laselva/metereological.
shtml). A broad gradient in soil fertility runs from relatively
rich entisols and inceptisols of alluvial origin to extensive
areas of low fertility ultisols developed on old lava flows. See
McDade et al. (1994) for more detail.
To monitor early seedling establishment, we set up five
belt transects, each composed of 200 contiguous 1-m2
quadrats (Vriesendorp 2002). Within 20 m of each seedling
transect, we mapped and identified all woody plants > 5 cm
diameter at 1.3 m height; this size cut-off encompassed
mature individuals of canopy, subcanopy and treelet species
but excluded lianas and understory plants. Transect locations were chosen randomly based on La SelvaÕs
50 · 100 m grid post system. To represent landscape-level
occurrence of major soils, we stratified transect selection
across three sites of residual, one site of recent alluvial and
one site of older alluvial soils. Grid posts that could serve as
a transect starting point were selected to accommodate 200m length and minimize contact with trails. For each of the
five grid posts, we randomly selected an azimuth angle to
determine transect orientation.
We conducted a census every 6 weeks in each transect,
marking all newly emerging woody dicot and palm seedlings
with numbered tags and identifying each to species. Tropical
seedling taxonomy is difficult, especially when individuals
possess solely cotyledons. For each encountered species ⁄ morphospecies, we took digital photographs as
voucher specimens, which were archived in a virtual
herbarium. Seedling identities were determined by collecting
2008 Blackwell Publishing Ltd/CNRS
Letter
and germinating seeds and comparing germinated individuals to the transect seedlings, identifying seeds attached to
seedlings in the field, observing seedlings under adults in the
field or, in the case of species overlap with Panama,
comparing voucher photographs with the illustrated
seedling guide for BCI (Garwood 2009). Through comparisons with voucher images in the virtual herbarium, most
individuals were positively identified to species and all
seedlings were identified to morphospecies.
For analyses on the effects of spatial scale, we included all
species–transect combinations with ‡ 10 seedlings over
c. 5.5 years (14 March 2000 through 6 October 2005). The
resulting 19 360 seedlings encompassed 106 species, representing canopy and subcanopy trees, treelets, understory
plants and lianas (Appendix S1). We excluded all morphospecies (n = 26) that had not been identified to genus to
minimize taxonomic errors. Exclusion of morphospecies
(sample sizes ranging from 1 to 110 seedlings per transect) is
unlikely to bias results because abundance–rank relationships mirrored that of known species. Presumed species that
had been identified to genera, but not species, were
included. If these presumed species were not distinct
species but variants of other species in the data set, then
we would be underestimating prevalence.
Data analysis
We calculated prevalence – the proportion of sampling units
in which a given species had at least one recently germinated
seedling – at two scales: the typical 1-m2 quadrat level
(prevalence1m) and aggregations of adjacent 1-m2 quadrats
to approximate the crown diameter of a mature individual of
that species (prevalenceorganism) based on its classification to
one of five lifeforms. Canopy trees (including emergents)
and lianas (which occupy the equivalent of one canopy tree
crown) were assigned a crown diameter of 20 m, subcanopy
trees 15 m, treelets 8 m and understory trees 2 m (based on
Clark et al. 2005 and field observations). For example, the
mean crown diameter of all La Selva canopy emergents is
24 m; Ceiba petandra, among the largest, has a mean crown
diameter of 32 m (Clark et al. 2005). We analysed prevalence
for species that had ‡ 10 seedlings in a transect. If a species
were represented by ‡ 10 seedlings in more than one
transect, we report the mean prevalence across sites.
To place bounds on the possible range of prevalenceorganism,
we calculated theoretical minima and maxima based on
occupancy of 1-m2 quadrats (Fig. 1b). Let z be the organism
scale (i.e. crown diameter class in metres). Then prevalenceorganism at scale z
= [(number of occupied units at scale z) ⁄ (total number of
1-m2 units ⁄ z)].
= [(number of occupied units at scale z) ⁄ (total number of
units at scale z)].
Letter
(a)
Prevalenceorganism (prop. of organism–scale sampling units with ≥1 seedling)
Minimum prevalenceorganism will occur when all occupied
1-m2 units are contiguously clustered such that: (number of
occupied units at scale z) = [(number of occupied 1 m2
units) ⁄ z], rounded up to the nearest integer (because a
sampling unit cannot be partially occupied). Maximum
prevalence will occur when each occupied 1-m2 unit
contributes to occupancy of a sampling unit at the coarser
spatial scale z (i.e. number of occupied 1-m2 sampling
units = number of occupied sampling units at scale z)
(similar to Fig. 1b).
To understand how prevalence1m is scaled to prevalenceorganism, we also calculated the number of contiguous
1-m2 sampling units that were vacant. Aggregating adjacent
1-m2 quadrats will lead to the greatest increases in seedling
prevalenceorganism when the vacant transect segment is just
less than the crown diameter of the focal species. Under this
scenario, space occupied by each mature plant would have
the potential to be replaced by a seedling of the focal
species.
We also tested whether seed volume (measured with
digital calipers) and adult abundance explained variation in
prevalence at both the 1 m and organism scales, using leastsquares regression in SYSTAT 12 (Systat Software, Chicago,
IL, USA). We had seed volume for 45 species (five of which
were measured at LaSelva), compiled by an NCEAS
Working Group (see Wright et al. 2007 and Acknowledgements). Relationships between prevalence vs. adult abundance and seed size were assessed using all species
combined and separately by lifeform.
Spatial scale of seedling limitation 223
(b)
Canopy trees
1.00
1.00
0.75
0.75
0.50
0.50
0.25
0.25
0.00
0.0
(c)
0.1
0.2 0.3
0.4
0.00
0.0
0.5
(d)
Treelet
1.00
1.00
0.75
0.75
0.50
0.50
0.25
0.25
0.00
0.0
0.1
(e)
0.2
0.3
0.00
0.0
Subcanopy trees
0.2
0.4
0.6
0.8
Understory
0.1
0.2
Lianas
1.00
0.75
0.50
0.25
0.00
0.0
0.1
0.2
0.3
Prevalence1m (prop. of 1m2 sampling units with ≥1 seedling)
RESULTS
Estimates of seedling limitation were much lower when
considered at a spatial scale commensurate with mature
plant size for the 106 more common species (‡ 10 seedlings
in ‡ 1 transect; Appendix S1). Most canopy and subcanopy
tree species had seedlings in 2–25% of 1-m2 sampling units;
when combining adjacent units to approximate crown
diameter (i.e. the linear distance taken up by mature trees),
most species were represented in 23–92% of mature
organism-scale sampling units (Fig. 2a,b). The species in
the treelet and liana lifeform groups had similar magnitudes
of response, but tended to have a lower and more restricted
range of prevalence1m (Fig. 2c,e), with most species having
seedlings in 2–20% of 1-m2 sampling units and 10–80% of
organism-scale units. Understory species had the smallest
change in estimated seedling limitation (Fig. 2d), likely
because understory species occupy the shortest linear
distance (2 m).
Most species were closer to the theoretical maximum
than minimum prevalenceorganism at a given prevalence1m
(Fig. 2), suggesting that seedlings of most species occurred
throughout a given transect. Indeed, there were relatively
Figure 2 Species-specific seedling prevalence at organism vs. 1-m
scales by lifeform category (a–e). The lines represent minimum and
maximum theoretical prevalence at the organism scale given
seedling prevalence at the 1 m scale. See text for detail.
few cases where vacant stretches along the transects
exceeded the crown diameter for canopy and subcanopy
trees (Fig. 3a). For canopy trees, only 7% of the vacant
stretches of transect were ‡ 20 m but for lianas this was
13%; this difference is consistent with canopy trees being
closer to the theoretical maximum prevalenceorganism than
lianas. For subcanopy trees, 11% of the vacant stretches of
transect were > 15 m (=crown diameter for subcanopy
lifeform). Vacant lengths of transect were more pronounced
for treelets and understory plants: 27% of vacant regions
were ‡ 8 m, the crown diameter of treelets, and 57% of
vacant regions were ‡ 2 m, the crown diameter of
understory plants (Fig. 3b).
Prevalence at both the 1-m and mature organism scales
was positively but weakly related to mature tree abundance
within 20 m for canopy, subcanopy and treelet species
combined (linear regression, n = 71, P = 0.006 and 0.044;
2008 Blackwell Publishing Ltd/CNRS
224 R. K. Kobe and C. F. Vriesendorp
Letter
0.7
0.6
0.5
Canopy
0.4
Subcanopy
Liana
0.3
Proportion of cases
0.2
0.1
0
0–1
2–7
8–15
16–19
20–39
≥ 40
0.6
0.5
0.4
Treelets
Understory
0.3
lifeforms. However, Pentaclethra macroloba, the most common
tree species at La Selva, strongly influenced significant
relationships. Excluding P. macroloba, prevalences at 1-m and
organism scales were not related to adult abundance for
canopy trees nor the combined lifeforms (lowest P = 0.12
and highest R2 = 0.02, which was for all lifeforms combined at 1-m scale prevalence). Lianas and understory
species were not included in analyses because they were not
mapped. Excluding the large-seeded species Pentaclethra
macroloba, seed size was not significantly related to prevalence at the 1-m or organism scales for all lifeforms
combined and lifeform-specific analyses, regardless of
whether adult abundance was included in the regression
model (lowest P = 0.181 and highest R2 = 0.06, for 12
subcanopy tree species at 1 m prevalence).
Relatively few species occur in all transects, supporting
seedling limitation at a landscape scale (maximum distance
between transects = 2.4 km). Considering all 301 species
and 20 804 seedlings, only 16% of species were represented
by at least one individual in all five transects. Similarly,
considering the 106 more common species (n ‡ 10 in at
least one transect) that were represented by 19 360
seedlings, only 8.5% of species were represented by ‡ 10
individuals in all five transects (Fig. 4). Conversely, 41% and
45% of species were present in only one transect,
represented by ‡ 1 or ‡ 10 seedlings respectively. Most
species were relatively rare (Fig. 5), with 28% of species (out
of 301 species total in the data set) represented by a single
seedling in a given transect. An equivalent percentage (31%)
0.5
0.2
0.45
0.4
0
0–1
2-7
8-15
≥ 16
Distance between occupied
1 m2 quadrats (m)
Figure 3 Histogram of the lengths of transect portions that were
unoccupied for a given species, shown by lifeform category. A
single case was calculated as the distance between seedlingoccupied 1-m2 quadrats for a given species. All cases of
unoccupied transect lengths across all species within a lifeform
were combined to develop the histogram.
R2 = 0.099 and 0.044 respectively for 1 m and mature
organism scales). In the within-lifeform analyses, prevalence
at 1 m was related to adult abundance for canopy species
(P < 0.001; R2 = 0.64) but not subcanopy or treelet
2008 Blackwell Publishing Ltd/CNRS
Proportion of species
0.1
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
1
2
3
4
5
Number of transects in which
species was present
Figure 4 Histogram of species presence across transects. Solid fill
indicates that ‡ 10 seedlings were present in a transect (among the
106 species that had 10 seedlings in at least one transect) and the
grey bars indicate the presence of ‡ 1 seedling in a transect (among
the 301 total species encountered).
Letter
Spatial scale of seedling limitation 225
300
0.4
Count
0.2
100
Proportion per bar
0.3
200
0.1
0
1
2
3
4
5
6
7
8
9
10 ≥ 11
0.0
Number of seedlings per species
per transect
Figure 5 Histogram of the number of seedlings present per species
per transect over the 5.5-year monitoring period. ÔCountÕ is the
number of species–transect combinations with the specified
seedling number. Species–transect combinations with < 1 seedling
are considered in Fig. 4 and are excluded here.
of species were represented by at least 10 individuals in at
least one transect, which were the subset of species used in
testing effects of sampling unit size on detection of seedling
limitation.
DISCUSSION
Understanding ecological processes at their relevant spatial
scales is a central problem in ecology (Levin 1992). To
rigorously test whether seedling limitation influences mature
plant composition, seedling limitation should be assessed at
a spatial scale commensurate with the size of the mature
plants that a seedling might someday replace (Muller-Landau
et al. 2002). In grasslands and other non-forest ecosystems,
relatively small sampling units encompass the space occupied by a mature plant (Tilman 1997; Zobel et al. 2000;
Seabloom et al. 2003). However, there has been strong
emphasis on RL as an important influence in both
temperate (Ribbens et al. 1994; Clark et al. 1998) and
especially tropical forests (Hubbell et al. 1999; Dalling et al.
2002; Muller-Landau et al. 2002; Denslow et al. 2006;
Comita et al. 2007; Norden et al. 2007), where mature plants
are much larger. For a wet tropical forest, we assessed how
FSL – the failure for a sampling unit to have at least one
seedling of a species – was influenced by spatial scale. For
more common species (‡ 10 new seedlings in ‡ 1 transect
over 5.5 years), estimates of seedling limitation were greatly
reduced when the spatial scale over which we evaluated
prevalence was commensurate with the spatial scale of
mature plants. For more common species, these results
suggest that seedling limitation would have negligible
influence on slowing community dynamics and competitive
exclusion (Hurtt & Pacala 1995).
Fundamental seedling limitation (=1 ) seedling prevalence), used here and in several other studies (e.g. Hurtt &
Pacala 1995; Hubbell et al. 1999), is a strict definition of RL.
We used this strict definition because ultimately only one
seedling is necessary to replace a given mature tree, as long
as that seedling survives and grows into canopy status.
Certainly a greater number of seedlings would enhance a
speciesÕ probability of mature tree recruitment by having
more opportunities to overcome mortality. But we argue
that after one seedling has arrived, other processes such as
establishment (Clark et al. 2007) and growth and survival
(Kobe 1999) become more critical recruitment bottlenecks
to mature stages. We acknowledge that seedling arrival as
examined here could have been broken down into more
distinct processes. However, we will be examining dispersal
(inferred from distances between potential parents and
seedlings), seedling growth and mortality in other papers. It
is especially important to distinguish among early life-history
processes because seed arrival may involve a greater element
of chance in contrast with seedling establishment (Clark
et al. 2007) and growth and mortality responses to resources
(Kobe 1999, 2006).
Our analyses of spatial scale effects focused on more
common species, but in other studies even common species
are interpreted as being recruitment limited because they
were analysed at a fine spatial scale. For example, on BCI,
the most commonly encountered species of seedling (of an
unidentified species) occurred in only 14.9% of 1-m2
seedling quadrats (Hubbell et al. 1999). Based on our results
for canopy trees, species that occupy 10–18% of 1-m2
seedling quadrats would be represented by at least one
seedling in 65–85% of 20-m length sampling units. It is
important to note that these estimates are conservative and
that true seedling limitation may be even less severe because
we based species occupancy upon a 1 · 20-m strip (to
approximate crown diameter) rather than total area occupied by a tree crown. For a 20-m diameter crown, this
studyÕs 1 · 20-m swath sampled only 6.4% of the crown
area (=area of swath ⁄ area of crown = 20 m2 ⁄ 314.2 m2).
We did not scale the 1-m2 quadrats to crown area to avoid
the assumption of isotropic seed dispersal and early seedling
establishment.
Both the increase in prevalence when scaled to mature
plant size and the approach of most species to the
maximum theoretical prevalenceorganism arose from broad
distributions of seedlings within transects. Indeed, few
vacant stretches of transect were longer than the crown
diameter of the focal species. Thus, as adjacent quadrats
were aggregated – because a given species was represented
in different portions of the transect – prevalence increased.
2008 Blackwell Publishing Ltd/CNRS
226 R. K. Kobe and C. F. Vriesendorp
If quadrat occupancy were strongly clumped, then aggregating adjacent quadrats to approximate crown diameters of
mature plants would not result in an increase in prevalence
at the organism scale. Species occupancy of quadrats could
be broad and non-clumped even as seedlings are highly
clumped due to exponential decays in seed ⁄ seedling
numbers with distance from source (Clark et al. 1999b).
The paucity of unoccupied stretches of transect is consistent
with the dominance of animal dispersal for most woody
species at La Selva (http://sloth.ots.ac.cr/local/florula3/
docs/lista_arboles_sindromes_OVR05.pdf; Chazdon et al.
2003) and other tropical forests (Brewer & Rejmanek 1999;
Webb & Peart 2001). Similarly, weak relationships between
prevalence and local adult abundance suggest that propagules are readily dispersing.
Our results suggest less-severe seedling limitation than
assumed by an important theoretical paper (Hurtt & Pacala
1995), which has implications for the degree to which
seedling limitation can slow community dynamics and delay
competitive exclusion. In their model, the probability that
no juveniles of species i are present at a local site can be
expressed as:
Fi
Lðxi Þ ¼ exp
;
ð1Þ
Q
Letter
where Fi is per plant fecundity for species i and Q is the
number of species. Thus, Fi ⁄ Q is a measure of seedling
limitation. We can solve for Fi ⁄ Q by setting eqn 1 equal to
this studyÕs empirically observed seedling limitation at both
spatial scales of analysis. At the 1-m scale, FSL
(=1 ) prevalence) for common canopy and subcanopy
species generally ranged from 0.8 to 0.95, resulting in
F ⁄ Q = 0.22–0.051. At these F ⁄ Q values, a species winning a
local site is dominated by forfeit rather than competitive
dominance, resulting in slower community dynamics (see
Figs 2 and 4 in Hurtt & Pacala 1995). However, at the
mature organism scale, seedling limitation for common
canopy and subcanopy species generally ranged from 0.1 to
0.4 with F ⁄ Q = 2.3–0.92, which would correspond with
competitive dominance strongly governing the winning of
sites to equal roles of competitive dominance and winning
by forfeit. Thus for the more common species, seedling
limitation would not substantially slow population dynamics
and delay their exclusion of less competitive species.
fecund species at La Selva. The rarer species that we did not
analyse are likely seedling limited even at spatial scales
commensurate with the area occupied by a mature crown. In
addition, few species were represented in all five transects,
which were separated by a maximum distance of 2.4 km.
Thus at a landscape-level (i.e. variation among sites),
seedling limitation could be operating to enhance beta
diversity. Sparse representation of seedling species across
the five sites could arise from mature tree associations with
soil characteristics (Clark et al. 1999a; John et al. 2007), but
we do not have the data to test this idea rigorously.
On the other hand, seedling prevalence could be
underestimated because the 5.5-year study length is a
fraction of the lifespan of a typical canopy tree (MullerLandau et al. 2002). Source limitations might have been
alleviated if the study had encompassed mast events that
occur less frequently than the length of monitoring. In
addition, we could be underestimating prevalence for both
rare and more common species if seedlings were appearing
and dying within a single 6-week census interval and thus
went undetected; density-dependent mortality mediated by
soil fungal pathogens can occur in seeds or seedlings within
a 2- to 3-week time frame (McCarthy-Neumann & Kobe
2008).
The presence of newly germinated seedlings examined
here is the final result of several processes (seed production,
dispersal, escape from predation, germination and establishment), which limits mechanistic interpretation. Nevertheless, using newly germinated seedlings had several
advantages in comparison to enumerating seeds in traps.
Seed traps likely under-sample seeds that have been
secondarily dispersed by animals and thus seed traps could
exaggerate dispersal limitation. For example, species of
seedlings found in plots often were not present in adjacent
seed traps (Harms et al. 2000), which suggests high levels of
fine-scale spatial heterogeneity, persistent seed banks and ⁄ or
seed trap bias. Seedlings in quadrats are unlikely to be biased
by dispersal mode. Seedling transects also facilitated
sampling larger areas than more labour-intensive seed traps.
Finally, using seedlings enabled a continuous belt transect
where quadrats could be aggregated at different spatial
scales; it would have been very difficult to construct a
continuous array of adjacent seed traps without greatly
disturbing study sites.
Caveats
CONCLUSIONS
There are c. 950–1000 woody species at La Selva (http://
sura.ots.ac.cr/local/florula3/index.htm) and our seedling
database included 301 species identified to genus, but only
106 species met the criterion of ‡ 10 seedlings in at least
one transect (see Appendix S1) to be included in our
analysis. Thus our results apply to the more common or
Numerous studies in tropical forests have assessed seed or
seedling limitation using sampling units that are generally
£ 1 m2 (e.g. Hubbell et al. 1999; Dalling et al. 2002) and
have concluded that even common species are severely
seedling limited. In the present study, seedling limitation
may be operating for rare species and even for more
2008 Blackwell Publishing Ltd/CNRS
Letter
common species at the landscape level. Losing sites by
forfeit because of seedling limitation could slow the
population dynamics of these rarer species and could delay
their exclusion of potentially less competitive species (Hurtt
& Pacala 1995); however, it is not known if the rare species
here would be competitive dominants. Nevertheless, our
results support that 1-m2 quadrats substantially underestimate prevalence for more common species; estimates of
seedling limitation were much lower when considered at
spatial scales commensurate with the size of mature woody
plants. It is unlikely that competitive exclusion by these
common species is delayed through seedling limitation,
suggesting that other processes such as differences among
species in resource- and density-dependent performance
and plant–soil feedbacks (e.g. McCarthy-Neumann & Kobe
2008) may more strongly regulate their population
dynamics.
ACKNOWLEDGEMENTS
We acknowledge financial support from NSF (DEB
0075472, 0640904, 0743609) and the MSU Intramural
Research Grant Program. We thank Ademar Hurtado,
Martin Cascante and Yehudi Hernandez for hard work in
the field; Orlando Vargas for help with seedling taxonomy
and OTS for logistical support. Seed size data were
compiled by participants in the ÔLife-history variation and
community structure in Neotropical rainforest communities:
Ecological and phylogenetic influencesÕ Working Group
supported by the National Center for Ecological Analysis
and Syntheses, a Center funded by NSF (Grant DEB-9421535), UCSB and the State of California. This paper was
improved through the constructive comments of Sarah
McCarthy-Neumann, Tom Baribault, Ellen Holste, David
MacFarlane, George Hurtt, Joe Wright, Marcel Rejmanek
and two anonymous referees.
REFERENCES
Brewer, S.W. & Rejmanek, M. (1999). Small rodents as significant
dispersers of tree seeds in a Neotropical forest. J. Veg. Sci., 10,
165–174.
Chazdon, R.L., Careaga, S., Webb, C. & Vargas, O. (2003).
Community and phylogenetic structure of reproductive traits
of woody species in wet tropical forests. Ecol. Monogr., 73,
331–348.
Clark, J.S., Macklin, E. & Wood, L. (1998). Stages and spatial scales
of recruitment limitation in southern Appalachian forests. Ecol.
Monogr., 68, 213–235.
Clark, D.B., Palmer, M.W. & Clark, D.A. (1999a). Edaphic factors
and the landscape-scale distributions of tropical rain forest trees.
Ecology, 80, 2662–2675.
Clark, J.S., Silman, M., Kern, R., Macklin, E. & HilleRisLambers, J.
(1999b). Seed dispersal near and far: patterns across temperate
and tropical forests. Ecology, 80, 1475–1494.
Spatial scale of seedling limitation 227
Clark, M.L., Roberts, D.A. & Clark, D.B. (2005). Hyperspectral
discrimination of tropical rain forest tree species at leaf to crown
scales. Rem. Sens. Environ., 96, 375–398.
Clark, C.J., Poulsen, J.R., Levey, D.J. & Osenberg, C.W. (2007).
Are plant populations seed limited? A critique and meta-analysis
of seed addition experiments Am. Nat., 170, 128–142.
Comita, L.S., Aguilar, S., Perez, R., Lao, S. & Hubbell, S.P. (2007).
Patterns of woody plant species abundance and diversity in the
seedling layer of a tropical forest. J. Veg. Sci., 18, 163–174.
Dalling, J.W., Muller-Landau, H.C., Wright, S.J. & Hubbell, S.P.
(2002). Role of dispersal in the recruitment limitation of neotropical pioneer species. J. Ecol., 90, 714–727.
Denslow, J.S., Uowolo, A.L. & Hughes, R.F. (2006). imitations to
seedling establishment in a mesic Hawaiian forsest. Oecologia,
148, 118–128.
Garwood, N.C. (with M. Tebbs, illustrator) (2009). Seedlings of Barro
Colorado Island and the Neotropics. Cornell University Press, Ithaca,
NY.
Harms, K.E., Wright, S.J., Calderon, O., Hernandez, A. & Herre,
E.A. (2000). Pervasive density-dependent recruitment enhances
seedling diversity in a tropical forest. Nature, 404, 493–495.
Hubbell, S.P., Foster, R.B., O’Brien, S.T., Harms, K.E., Condit, R.,
Wechsler, B. et al. (1999). Light-gap disturbances, recruitment
limitation, and tree diversity in a neotropical forest. Science, 283,
554–557.
Hurtt, G.C. & Pacala, S.W. (1995). The consequences of recruitment limitation – reconciling chance, history and competitive
differences between plants. J. Theor. Biol., 176, 1–12.
John, R., Dalling, J.W., Harms, K.E., Yavitt, J.B., Stallard, R.F.,
Mirabello, M. et al. (2007). Soil nutrients influence spatial distributions of tropical tree species. PNAS, 104, 864–869.
Jones, F.A. & Muller-Landau, H.C. (2008). Measuring long-distance
seed dispersal in complex natural environments: an evaluation and
integration of classical and genetic methods. J. Ecol., 96, 642–652.
Kobe, R.K. (1999). Light gradient partitioning among tropical tree
species through differential seedling mortality and growth.
Ecology, 80, 187–201.
Kobe, R.K. (2006). Sapling growth as a function of light and
landscape-level variation in soil water and foliar nitrogen in
northern Michigan. Oecologia, 147, 119–133.
Levin, S.A. (1992). The problem of pattern and scale in ecology.
Ecology, 73, 1943–1967.
McCarthy-Neumann, S. & Kobe, R.K. (2008). Tolerance of
neighborhood soil pathogens co-varies with shade tolerance
across species of tropical tree seedlings. Ecology, 89, 1883–1892.
McDade, L.A., Bawa, K.S., Hespenheide, H.A. & Hartshorn, G.S.
(1994). La Selva: Ecology and Natural History of a Neotropical Rain
Forest. University of Chicago Press, Chicago.
Muller-Landau, H., Wright, S.J., Calderón, O., Hubbell, S.P. & Foster,
R.B. (2002). Assessing recruitment limitation: concepts, methods
and examples for tropical forest trees. In: Seed Dispersal and Frugivory:
Ecology, Evolution and Conservation (eds Levey, J., Silva, W.R. & Galetti, M.). CAB International, Oxfordshire, UK, pp. 33–53.
Nathan, R. & Muller-Landau, H.C. (2000). Spatial patterns of seed
dispersal, their determinants and consequences for recruitment.
TREE, 15, 278–285.
Norden, N., Chave, J., Caubere, A., Chatelet, P., Ferroni, N.,
Forget, P.M. et al. (2007). Is temporal variation of seedling
communities determined by environment or by seed arrival? A
test in a neotropical forest J. Ecol., 95, 507–516.
2008 Blackwell Publishing Ltd/CNRS
228 R. K. Kobe and C. F. Vriesendorp
Ribbens, E., Silander, J.A. & Pacala, S.W. (1994). Seedling
recruitment in forests – calibrating models to predict patterns of
tree seedling dispersion. Ecology, 75, 1794–1806.
Russo, S.E., Portnoy, S. & Augspurger, C.K. (2006). Incorporating
animal behavior into seed dispersal models: implications for seed
shadows. Ecology, 87, 3160–3174.
Seabloom, E.W., Borer, E.T., Boucher, V.L., Burton, R.S., Cottingham, K.L., Goldwasser, L. et al. (2003). ompetition, seed
limitation, disturbance, and reestablishment of California native
annual forbs. Ecol. Appl., 13, 575–592.
Stevenson, P.R. (2000). Seed dispersal by woolly monkeys (Lagothrix lagothricha) at Tinigua National Park, Colombia: dispersal
distance, germination rates, and dispersal quantity. Am. J. Primatol., 50, 275–289.
Tilman, D. (1997). ommunity invasibility, recruitment limitation,
and grassland biodiversity. Ecology, 78, 81–92.
Turnbull, L.A., Crawley, M.J. & Rees, M. (2000). Are plant populations seed-limited? A review of seed sowing experiments Oikos,
88, 225–238.
Vriesendorp, C.F. (2002). Maintenance of diversity in a neotropical seedling
community. PhD Dissertation, Michigan State University, East
Lansing, MI, USA.
Webb, C.O. & Peart, D.R. (2001). High seed dispersal rates in
faunally intact tropical rain forest: theoretical and conservation
implications. Ecol. Lett., 4, 491–499.
Wehncke, E.V., Hubbell, S.P., Foster, R.B. & Dalling, J.W. (2003).
Seed dispersal patterns produced by white-faced monkeys:
implications for the dispersal limitation of neotropical tree
species. J. Ecol., 91, 677–685.
Wright, S.J., Muller-Landau, H.C., Calderon, O. & Hernandez, A.
(2005). Annual and spatial variation in seedfall and seedling
recruitment in a neotropical forest. Ecology, 86, 848–860.
2008 Blackwell Publishing Ltd/CNRS
Letter
Wright, I.J., Ackerly, D.D., Bongers, F., Harms, K.E., IbarraManriquez, G., Martinez-Ramos, M. et al. (2007). Relationships
among ecologically-important dimensions of plant trait
variation in seven Neotropical forests. Ann. Bot., 99, 1003–
1015.
Zobel, M., Otsus, M., Liira, J., Moora, M. & Mols, T. (2000). Is
small-scale species richness limited by seed availability or microsite availability? Ecology, 81, 3274–3282.
SUPPORTING INFORMATION
Additional Supporting Information may be found in the
online version of this article:
Appendix S1 Species used in the study, with information on
family, lifeform, occurrence in transects and sample sizes in
each transect.
Please note: Wiley-Blackwell are not responsible for the
content or functionality of any supporting materials
supplied by the authors. Any queries (other than missing
material) should be directed to the corresponding author for
the article.
Editor, Marcel Rejmanek
Manuscript received 3 July 2008
First decision made 11 August 2008
Second decision made 24 October 2008
Manuscript accepted 19 November 2008