Content uploaded by Rommel Montúfar
Author content
All content in this area was uploaded by Rommel Montúfar on Apr 20, 2018
Content may be subject to copyright.
How anthropogenic disturbances affect the resilience of a keystone palm tree
in the threatened Andean cloud forest?
Fabien Anthelme
a,b,
⇑
, Juan Lincango
b
, Charlotte Gully
a,b
, Nina Duarte
b
, Rommel Montúfar
b
a
Institut de Recherche pour le Développement (IRD), UMR DIAPC, 911 Avenue Agropolis, BP 64501, 34394 Montpellier Cedex 5, France
b
Pontificia Universidad Católica del Ecuador, Av 12 de Octubre y Roca, Quito, Ecuador
article info
Article history:
Received 24 September 2010
Received in revised form 16 December 2010
Accepted 26 December 2010
Available online 22 January 2011
Keywords:
Ceroxylon echinulatum
Cloud forest
Deforestation
Ecuador
Facilitation
Selective logging
abstract
To conserve tropical forests, it is crucial to characterise the disturbance threshold beyond which popula-
tions of tropical trees are no longer resilient. This approach is still not widely employed, especially with
respect to the effects of moderate disturbances. Compensation effects, such as positive interactions
among plants, are addressed even more rarely. We attempt to identify the extents to which the distribu-
tion of the keystone palm tree Ceroxylon echinulatum is regulated by various regimes of deforestation in a
threatened tropical montane cloud forest in the North-West Andes of Ecuador. The demographic struc-
ture of this palm tree was examined in three habitats: old-growth forest, forest disturbed by selective log-
ging, and deforested pasture. Patterns were related to stand structure, microclimate, and soil
composition. Seedling desiccation owing to severe aboveground water stress led to the absence of juve-
nile palms in pastures, and thus was predictive of a near extinction of the species in this habitat. How-
ever, shade provided by dominant bunchgrass in pastures considerably reduced above- and
belowground water stress by diminishing light intensity. Selective logging resulted in a higher density
of individuals in disturbed forests than in old-growth forests, but was associated with a spoiled spatial
structure. Therefore, the protection of residual old-growth forests is a prerequisite for the conservation
of C. echinulatum, although secondary forests might act as provisional refuges that promote its resilience.
The reduction of water stress by nurse grasses in pastures represents a promising approach to promote
the resilience of tropical tree species and their associated communities after deforestation.
Ó2011 Elsevier Ltd. All rights reserved.
1. Introduction
The pervasiveness of deforestation was recently evidenced on a
global scale (Hansen et al., 2010). Humid Tropics present the high-
est portion of forests on Earth, and at the same time they lost as
much as 2.4% of their total cover in the 2000–2005 period (Hansen
et al., 2010). It is largely accepted that intense disturbances such as
deforestation and habitat fragmentation affect the performance
and abundance of plant species deeply in this biome (Debinski
and Holt, 2000), and thereby influence biodiversity as well as eco-
system functions and services (Hering, 2003; Cayuela et al., 2006).
However, it remains uncertain whether tropical forest species
might benefit from intermediate disturbances (e.g., selective log-
ging), although this is a central hypothesis with respect to explain-
ing the effects of such activities on the biodiversity in species-rich
ecosystems (Bongers et al., 2009).
Tropical montane cloud forests (TMCF) in the Andes are among
the ecosystems with the highest biodiversity globally (Brummit
and Lughada, 2003). The effects of deforestation on these ecosys-
tems are thought to be highly negative (Goerner et al., 2007) but
they remain poorly studied (however, see Svenning, 1998;
Svenning et al., 2009) and are in great need of basic research as
well as specific investigations from the perspectives of conserva-
tion and restoration (Wright, 2005). Especially, deforestation gen-
erates strong micro-climatic shifts at the plant scale (e.g., Holl,
1999). Resulting increasing abiotic stress may create conditions
that exceed the physiological limits of juvenile trees. Accordingly,
facilitative interactions with nurse plants may become a central
ecological process in the survival of forest trees in deforested
pastures (stress gradient hypothesis, Brooker et al., 2008). This
hypothesis needs to be tested in TMCF.
Neotropical palms have been shown highly sensitive to human-
induced disturbances, although each type of disturbance (e.g. hab-
itat fragmentation, fire, hunting, non lethal harvest) may affect
each palm population in a singular way (Montúfar et al., in press).
We focus here on the Andean wax palm Ceroxylon echinulatum
Galeano (hereafter Ceroxylon), syn. Ceroxylon alpinum Bonpl. ex
DC. subsp. ecuadorense Galeano (Sanín and Galeano, in press).
0006-3207/$ - see front matter Ó2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.biocon.2010.12.025
⇑
Corresponding author at: Institut de Recherche pour le Développement (IRD),
UMR DIAPC, 911 Avenue Agropolis, BP 64501, 34394 Montpellier Cedex 5, France.
Tel.: +33 467416478.
E-mail address: fabien.anthelme@ird.fr (F. Anthelme).
Biological Conservation 144 (2011) 1059–1067
Contents lists available at ScienceDirect
Biological Conservation
journal homepage: www.elsevier.com/locate/biocon
The restricted, fragmented distribution of Ceroxylon in the TMCF of
Ecuador and Northern Peru makes populations of this species par-
ticularly sensitive to anthropogenic alterations of their natural
habitat (Trenel et al., 2008).
Through a demographic study of Ceroxylon combined with local
biotic and abiotic measurements, we specifically tested (1) the ex-
tent to which populations can endure various types of disturbance
related to deforestation and (2) how plant–plant interactions inter-
fere with the dynamics and structure of populations. We hypothe-
sised that this latter process might have a significant role in
explaining the distribution of tropical trees in a highly productive
system that is subject to consumer pressure (see Callaway, 2007).
In light of the obtained results, we discuss possible approaches to
manage the conservation of Ceroxylon in TMCF.
2. Materials and methods
2.1. Study area and study sites
This study was undertaken in the mountainous area of the
Ecuadorian Chocó, on the western slopes of the Andes (North-West
Pichincha district). Two sites that exhibit ecological conditions that
are optimal for Ceroxylon were selected: the Inti Llacta Reserve
(00°02
0
N, 78°43
0
W; mean altitude 1861 m a.s.l.) and the Rio Bravo
Reserve within the Mindo–Nambillo Protected Forest (00°04
0
S,
78°44
0
S; 1548 m a.s.l.). Both sites contain TMCF and receive
approximately 3200 mm of precipitation each year, as reported
for an adjacent area (Svenning, 1998). Historically, the study area
has been influenced by an indigenous population (Yumbo) for
hundreds of years (Valarezo, 2001).
In modern times, Mindo–Nambillo has been affected very little
by human-related disturbances owing to its remoteness; therefore,
it constitutes an old-growth forest with the presence of the long-
lived trees Licaria limbosa,Ocotea cernua and the palms Aiphanes
erinacea and Geonoma undata, representative of undisturbed TMCF
(Svenning, 1998). In contrast, Inti-Llacta has recently experienced
deforestation and selective logging, as indicated by a loss of 41%
of forested areas from 1966 to 1990 in an adjacent zone (Guevara
et al., 2001). Pastures were established during the same period, and
were soon dominated by the bunchgrass Setaria sphacelata. This
species can reach more than 2 m in height and is almost mono-
dominant in the current pastures. Forests are dominated by a
patchwork of early- and late-successional trees, which include
Cecropia pentandra,Cinchona pubescens, and several arboreal ferns
(Cyathea spp.). The presence of the palm Chamadorea linearis is
indicative of a moderately disturbed environment (Svenning,
1998). In the pastures, residual trees are dominated by Ceroxylon
and Alnus acuminata.
2.2. Target species
Ceroxylon echinulatum is an endemic, long-lived palm tree that
reaches an average stem height of 20 m (Paredes-Ruiz, 1995) and
displays keystone properties owing to a local high density and in-
tense interactions with seed dispersers (Mejía Londoño, 1999).
From a socio-economic point of view, the important harvest of
its leaves and, to a lesser extent, its stems (Borchsenius and Mor-
aes, 2006) might be detrimental to its long-term conservation.
Although the current distribution of Ceroxylon in Ecuador might
be underestimated, some populations in the western Andes of
Ecuador are thought to be threatened by extinction through defor-
estation (Svenning and Balslev, 1998). However, during its long
establishment phase – or rosette phase – Ceroxylon develops a
short subterranean stem and a rosette of leaves very difficult to
eradicate when the area is cleared out, which may increase its
resilience in the face of deforestation (R. Bernal, pers. com.).
2.3. Sampling design
The population structure of Ceroxylon was observed through a
snapshot study along a gradient of increasing intensity of human
activities in old-growth forests (Mindo), disturbed forests where
selective logging of large tree species is taking place (Inti-Llacta),
and deforested pastures dominated by S. sphacelata with a few
residual trees (Inti-Llacta), following a protocol developed within
the international project PALMS (see Anthelme et al. (2010) for de-
tails on the method). Although old-growth forest was not at the
same site as the other habitats and its altitude was slightly lower,
all habitats displayed comparable volcanic bedrock and climate. A
total of 91 square plots of 400 m
2
in size were set randomly, sepa-
rated by at least 40 m, at 1412 to 1985 m a.s.l. A total of 20 plots
were established in old-growth forests, 46 in disturbed forests,
and 25 in pastures, in accordance with the structural heterogeneity
of each habitat.
Plot randomisation consisted of identifying the nearest
stemmed individual of Ceroxylon from a random point within a
habitat, and using it as one of the four corners of the plot.
2.4. Biotic observations
Demographic observations were carried out in June and July
2008 by counting the number of individuals within each plot, clas-
sifying them into one of five life stages (Table 1) and separating
males and females among adults.
Seedling stage was observed using a specially designed method
because of the particularly high density of individuals in some
areas. Three types of zone were identified within each plot: high
density (>300 individuals/m
2
), medium density (between 30 and
300 individuals/m
2
), and low density (less than 30 individuals/
m
2
). Within the high-density zone, individuals were counted in
1m
2
subplots to estimate the total number of individuals. In the
two other zones, seedlings were counted directly.
The stand structure of each plot was estimated by measuring
the basal area of all trees P10 cm in diameter at breast height.
2.5. Abiotic measurements
The locations of the plots were determined using a Garmin GPS
60 (UTM coordinates, WGS 84). Slope inclination was calculated
with a clinometer (Sunnto Tandem) from three 20-m-long mea-
surements along the direction of the slope.
During a relatively dry period (March 2009), we measured var-
ious micro-climatic parameters in disturbed forests and pastures.
Pastures were divided into two sub-habitats on the basis of the
presence or absence of neighbouring S. sphacelata to provide an
indication of potential plant–plant interactions between palms
and grasses. We concentrated the acquisition of data at the Inti-
Llacta site (disturbed forests and pastures only) to ensure more
comparable data between forests and pastures instead of observing
differences between old-growth forests and disturbed forests.
Vapour pressure deficit (VPD, kPa) was monitored with
HOBO-Pro RH/Temp data loggers as an indicator of atmospheric
water stress for plants, with 4 (5) spatial repetitions at each habitat
assigned randomly in the initial plot design. Relative humidity in
the soil at a depth of 5 cm was determined using U23-01 data log-
gers (three spatial repetitions at each habitat). Light intensity was
measured using UA-002 data loggers (eight spatial repetitions at
each habitat). All data loggers (Onset Computer Corporation,
Pocasset, MA) generated data every 15 min during a 12-day period,
from which daily maximal and/or minimal values were extracted
1060 F. Anthelme et al. / Biological Conservation 144 (2011) 1059–1067
to detect physical stresses for plants in the habitats. Data loggers in
pastures with Setaria were placed directly at the base of Setaria
stems.
Soil was collected in disturbed forests, pastures (February
2009), and old-growth forests (June 2009) from a sub-sample of
10 plots that were selected randomly within each habitat after a
relatively dry climatic period. Soil samples (at a depth of
0–20 cm) were collected at the four corners and the centre of each
plot. The concentrations of available macro- and micronutrients,
organic matter (%), and pH were measured following standardized
methods described by OSU (2009).
2.6. Data analysis
Data were analysed using Kruskal–Wallis and Mann–Whitney
non-parametric tests, except when indicated otherwise in the text.
On a landscape scale, the density of Ceroxylon individuals with
stems was calculated using the basic distance to the closest indi-
vidual estimator (BDCI; Cottam and Curtis, 1956):
BDCI ðindividuals=haÞ¼1=ð4½
R
Ri=N
2
Þ10;000;
where Ris the distance between a random point and the closest
stemmed Ceroxylon, and Nis the number of random points. Density
at plot level was later divided by density extracted from BDCI,
which provided an aggregation index for stemmed individuals.
The combined effects of biotic and abiotic parameters on the
demography of Ceroxylon at each plot were tested through step-
wise regressions (alpha to enter = 0.10; alpha to exit = 0.15). Spa-
tial repetitions of micro-climatic data (from three to eight) were
assembled to yield 12 mean values (corresponding to 12 days),
which we correlated with demographic data. We considered that
the potential temporal autocorrelation from day to day was insig-
nificant considering the robustness of the data obtained (mean val-
ues from several spatial locations) and the high day-to-day climatic
variability in this region. Statistical analyses were conducted with
MINITAB and the ADE-4 package available in R.
3. Results
3.1. Habitat structure
Slope inclination varied among habitats (p< 0.001). Slopes were
steeper in old-growth forests (25.01°± 1.95), and to a lesser extent
in disturbed forests (19.25°± 1.16), than in pastures, which con-
tained gentle slopes (14.24°± 1.68).
Overall, basal area (BA) varied among habitats (p< 0.001, Fig. 1),
but was not significantly different between old-growth forests and
disturbed forests (618.5 m
2
/ha and 641.6 m
2
/ha, respectively). BA
was 10 times lower in pastures (64.5 m
2
). BA assignable to
Ceroxylon (p< 0.001) reached 12.4% of total BA in disturbed forests
and 7.1% of that in old-growth forests. It clearly dominated tree
cover in (40.0%).
The BA distribution of diameter classes in disturbed forests was
typical of regenerating tropical forests (‘‘inverted J shape’’;
La Torre-Cuadros et al., 2007). Old-growth forests displayed less
slender trees (20–30 cm in diameter) than disturbed forests and
a larger number of old trees, as expected in mature forests.
3.2. Demography of Ceroxylon among habitats
A total of 129,069 individuals were included in the analysis, 98%
of which were seedlings. The rarity of juveniles 1 (J1) and the ab-
sence of juveniles 2 (J2) and sub-adults (J3) in pastures denoted
a highly abnormal demographic structure. In contrast, seedlings
were particularly frequent, and adults seemed to be present at
the same density as in disturbed forests at the plot level
(Table 1). Seedlings in old-growth forests were very scarce, espe-
cially when compared with the number in disturbed forests –
14.04 and 2159.72, respectively – but the estimated rate of survival
from seedlings to J1 was far higher in old-growth forests (21.9%)
than in the other two habitats (up to 2%). All life stages were rep-
resented in forested habitats. Almost all seedlings found in pas-
Table 1
Mean density of Ceroxylon individuals among life stages and habitats (±S.E.). Differences among treatments are compared using the Kruskal–Wallis test. Common letters indicate
no difference among pairs of treatments within each life stage (p< 0.05, Mann–Whitney test, hypothesis: not equal, adjusted for ties). Sign.: significance; ***: p< 0.001;
**: p< 0.01; *: p< 0.05.
Life stages Description Old-growth forest Disturbed forest Pasture Sign.
Seedlings Leaves not divided 14.04 ± 3.11
a
2159.72 ± 637
b
1326.46 ± 871
a
***
– High density 0
a
1979.28 ± 623
b
1292.11 ± 859
b
***
– Medium density 0
a
101.97 ± 18.2
b
30.05 ± 14.8
c
***
– Low density 14.04 ± 3.11
a
84.11 ± 9.83
b
5.80 ± 3.61
c
***
J1 Leaves divided, <2 m 3.08 ± 0.83
a
48.09 ± 7.94
b
2.35 ± 1.97
c
***
J2 >2 m, absence of stem 0.96 ± 0.34
a
7.89 ± 2.45
b
0
c
***
Sub-adults Stem easily identifiable 0.32 ± 0.13
a
0.50 ± 0.09
a
0
b
**
Adults Reproductive material 1.00 ± 0.10
a
1.61 ± 0.22
ab
1.65 ± 0.25
b
–
Fig. 1. Mean basal area (BA) among classes of tree diameter at each habitat (A), and
total BA with relative BA assignable to Ceroxylon (B). Common letters indicate no
difference among pairwise treatments for each diameter and each habitat
(Mann–Whitney test, p< 0.05). Error bars represent 95% confidence intervals. OF:
old-growth forest; DF: disturbed forest; P: pasture.
F. Anthelme et al. / Biological Conservation 144 (2011) 1059–1067 1061
tures were concentrated in small high-density areas, very close to
female trees (Table 1). In disturbed forests, seedlings were also
concentrated in high-density areas but a few seedlings were ob-
served in low- and medium-density areas as well. The few seed-
lings detected in old-growth forests were found systematically at
a low density.
3.3. Density and aggregation of stemmed individuals
Taking BDCI into account, the density of adults reached 15.6
individuals/ha in disturbed forests but only 5.2 in pastures. In
old-growth forests, the Rdistance exceeded 30 m almost systemat-
ically; therefore, the density of Ceroxylon could not be calculated,
but it was certainly less than that in pastures.
The aggregation index, which was extracted from the ratio of lo-
cal density/BDCI of stemmed individuals, was 2.95 in pastures and
1.85 in disturbed forests. Both habitats were significantly aggre-
gated (one-sample T-test that tested the hypothesis that local den-
sity was greater than BDCI, p< 0.05).
3.4. Interdependence between life stages and basal area
Stepwise regression models provided evidence of key effects of
the number of female palms and BA on the distribution of Ceroxy-
lon (Table 2). Seedlings in low-density areas were less dependent
on female density than seedlings in higher density areas and
showed a higher degree of correlation with the presence of older
life stages than the total number of seedlings. BA affected the
abundance of several life stages positively in all habitats. In dis-
turbed forests, this was due to the high density of young, regener-
ating trees (effects of BA 20–30 cm diameter on J1 and J2; Fig. 2),
whereas the effects of old, large trees (>60 cm diameter) were neu-
tral or slightly negative. Positive effects of BA in old-growth forests
(J1) and pastures (seedlings at low density) were not correlated
with the density of young or old trees (Fig. 2). In pastures, seedling
distribution was accurately predicted by tree canopy plus the pres-
ence of females (Table 2). Finally, the distribution of low-density
seedlings in disturbed forests was not predictable, which was
probably indicative of an efficient dispersal of these few
individuals.
3.5. Micro-climatic variations
Maximal light intensity (MLI) decreased drastically from pas-
tures (113,295 lux) to forests (13,097 lux, p< 0.001, Fig. 3). Inter-
estingly, the shading effect provided by Setaria reduced MLI by
more than one-half in pastures (49,787 lux).
Aboveground water stress (maximal VPD) was significantly var-
iable among habitats (p< 0.001). It peaked at 1.01 kPa in pastures,
with the absolute highest value exceeding 2 kPa on day 9. With re-
spect to MLI, Setaria in pastures reduced VPD by one-half, whereas
canopy virtually eliminated VPD in forest. MLI affected maximal
VPD strongly in pastures without Setaria (Appendix A:R
2
= 0.82,
p< 0.001) and in pastures with Setaria (R
2
= 0.87, p< 0.001), but
not in disturbed forests (R
2
= 0.08, p= 0.35).
Minimal values of relative humidity in the soil atmosphere
were especially low in pastures (35.3%) whereas protection by Se-
taria helped to maintain a similar level of minimal humidity to that
in disturbed forests (76.6% and 73.6%, respectively, Fig. 3). This var-
iable was also affected by MLI in pastures (Appendix A:R
2
= 0.38,
p< 0.05), but not in the two other habitats.
3.6. Soil composition
The most notable variations in the availability of soil nutrients
among habitats concerned NH
4
, Fe, and phosphorus (Table 3).
The concentrations of NH
4
and Fe increased significantly along
the gradient of disturbance, and, for Fe, reached a level 10 times
higher than the initial content in pastures. Levels of phosphorus
were lower in old-growth forests than in the other habitats, as
was the case for Cu. Zn was present at a higher concentration in
pastures than in the two types of forests. Finally, the soil was more
acidic in old-growth forests than in the two other habitats.
4. Discussion
In general, the high level of diversity of tropical forests results in
a low density of most plant species (Lieberman and Lieberman,
2007). In the present study, Ceroxylon was found to form up to
12% of the stand BA in TMCF, which has itself been shown to cor-
relate strongly with leaf cover and net throughfall input (Ponette-
Table 2
Relationships between densities of each life stage, habitat, and basal area (stepwise multiple regression: alpha to enter: 0.10; alpha to remove: 0.15). The effects of slope and sub-
adults were not significant and are not shown in the table. The response of adults to other life stages was not taken into account nor was the effect of BA on adults in pastures nor
the effects of older life stages on target stages. LD: low-density area. R
2
was adjusted when two or more variables were included in the analysis. + indicates a positive effect,
indicates a negative effect. Significance: n.s.: not significant; (*): p< 0.10; *: p< 0.05; **: p< 0.01; ***: p< 0.001.
Habitat/life stage Basal area Seedlings Seedlings LD J1 J2 Males Females R
2
Old-growth forest
Seedlings *** 0.43
Seedlings LD *** 0.43
J1 ** 0.27
J2 * 0.20
Sub-adults *** 0.46
Adults 0.13
Disturbed forest
Seedlings *** 0.45
Seedlings LD (*) 0.06
J1 * (*) ** 0.28
J2 *** ** 0.41
Sub-adults () 0.10
Adults n.s
Pasture
Seedlings (*) ** ** 0.61
Seedlings LD *** (*) 0.74
J1 *** 0.87
J2 n.s
Sub-adults n.s
Adults n.s
1062 F. Anthelme et al. / Biological Conservation 144 (2011) 1059–1067
Gonzalez et al., 2010). Therefore, Ceroxylon was shown to be a lo-
cally dominant component of the TMCF, although it is fragmented
on a larger scale (Trenel et al., 2008). Combined with its dominant
treelike structure and its high level of production of fleshy fruits
(Anthelme, pers. obs.), the palm acts as an ecosystem engineer or
keystone species, as reported for many palms in South America
(Scariot, 1999; Galetti et al., 2006), and especially in the Andes
(Svenning, 1998; Borchsenius and Moraes, 2006). However, it is
likely that the arrangement of Ceroxylon populations in clusters
on a landscape scale is due to a failure to achieve long-distance dis-
persal (see Trenel et al., 2008). The result is a high level of fragmen-
tation of relatively dense populations. For these reasons, and taking
into account its socio-economic value (Pintaud and Anthelme,
2008), Ceroxylon is an appropriate target genus for ecosystem con-
servation in TMCF.
4.1. Ceroxylon virtually extinct in deforested pastures?
A strong hypothesis when considering the altered demographic
structure in pastures is that the residual adults in this habitat only
reflected the earlier existence of forests from which Ceroxylon was
spared. Therefore, the dense stands that have been observed in open
areas (Paredes-Ruiz, 1995) would represent the remnants of virtu-
ally extinct populations. However, the higher density found in pas-
tures than in old-growth forests indicates very likely the
occurrence of a preliminary intermediate disturbed phase favour-
able to the development of the palm. This applies to C. echinulatum,
but also to gregarious species such as Ceroxylon quindiuense and C.
alpinum, which display a similar pattern of distribution (Sanín and
Galeano, in press; Vergara-Chaparro, 2002). The relatively high num-
ber of seedlings in pastures indicates a positive response of germina-
tion to light intensity, but is probably negated by higher mortality
through desiccation in this habitat, as suggested for the palmA. erin-
acea, which is found in adjacentold-growth forests (Svenning, 1998).
The presence of very few J1 individuals might indicate that it is
extremely difficult for Ceroxylon to establish itself in pastures. The
spatial association between JI individuals and seedlings, which are
themselves influenced by the presence of residual trees and palms,
indicates that their persistence is due to the presence of protecting
trees that provide local shade (see Table 2).
Data on the African oil palm showed that stomatal conductance
and photosynthetic activity are affected negatively by VPD values
greater than 1.8 kPa, even with a sufficient level of water in the
soil (Dufresne and Saugier, 1993). In comparison with Ceroxylon,
the African oil palm shows a better adaptation to water stress be-
cause it grows in areas adjacent to Ceroxylon where the environ-
ment is slightly more stressful (Anthelme, pers. obs.).
Accordingly, when the VPD reaches more than 2 kPa, as observed
in pastures, seedling desiccation and mortality owing to above-
ground water stress might occur. Together with edaphic water
stress and predation by herbivores, it is likely that this physiolog-
ical limitation is responsible for the failure of Ceroxylon to survive
beyond the seedling stage in open areas, as has been shown for
other mid-and late-successional species in the same region (Gun-
ter et al., 2009). The high level of soil nutrients, especially nitrogen,
in pastures indicates that soil composition, which is a major con-
straint for reforestation in TMCF (Gunter et al., 2009), is not
responsible for the absence of Ceroxylon regeneration in this hab-
itat. Similarly, a high level of iron, which might inhibit photosyn-
thesis in wet anoxic soils but not in the well-drained soils that
were observed, was not responsible (Lucassen et al., 2006). The al-
tered seed dispersal in pastures (Holl, 1999) may reduce even
more the probability of palm recruitment but appears not to be
a central factor when compared with physiological limitations in
our study.
An alternative hypothesis to explain the current distribution
pattern of Ceroxylon in pastures, suggested by Rodrigo Bernal (pers.
comm.), is that the presence of adults would result from J1 and J2
Fig. 2. Relationships between the densities of young trees (20–30 cm diameter, white boxes, solid line) and old trees (>60 cm diameter, black boxes, dashed line) and the
density of Ceroxylon individuals for the life stages that are influenced significantly by total basal area (see Table 2).
F. Anthelme et al. / Biological Conservation 144 (2011) 1059–1067 1063
individuals that survived complete deforestation (including adult
palms). According to Bernal, during their rosette phase (J1, J2), that
may reach as much as 57 years (Vergara-Chaparro, 2002), Ceroxy-
lon individuals are particularly resilient, as their meristem is sub-
terranean or almost so. Other sympatric palms, such as Wettinia
kalbreyeri or Euterpe precatoria, which do not go through this ro-
sette phase, are usually absent in deforested pastures, as their sur-
vival depends on adults being intentionally spared (R. Bernal pers.
com.). Future empirical research may resolve this crucial point,
especially by dating remnants adult palms in pastures (see Verg-
ara-Chaparro (2002) for detailed methods).
Finally, it appears unlikely that Ceroxylon will survive in pas-
tures after deforestation without intervention, as suggested for
many mid- or late-successional tree species in TMCF (Hubbell
et al., 2008; Gunter et al., 2009), including palms (Arroyo-Rodri-
guez et al., 2007).
4.2. Antagonistic responses to an intermediate disturbance
When compared with the effects of complete deforestation, the
effects of intermediate disturbances on biodiversity in the TMCF
remain poorly studied and the results obtained are controversial
(Bongers et al., 2009; Montúfar et al., in press; however, see Sven-
ning, 1998), although TMCF will probably be subjected to greater
disturbances in the future (Wright, 2005). In the case of Ceroxylon,
the high density of individuals and the persistence of nutrient-rich
soils in forests undergoing selective logging (disturbed forests)
suggest what appears to be a positive effect of the intermediate
disturbance, as documented previously for the understory palms
Chamadorea linearis and Chamadorea pinnatifrons (Svenning, 1998).
At the same time, we interpret the high concentration of Cerox-
ylon seedlings at the foot of females and the high rate of mortality
during development from seedlings to juveniles as an indication of
Fig. 3. Variations in the abiotic environment in disturbed forests (DF), pastures with Setaria sphacelata (Pn), and pastures without S. sphacelata (P). Common letters indicate
no difference among pairwise treatments (Mann–Whitney test, p< 0.05). Error bars represent 95% confidence intervals.
Table 3
Variation in the soil nutrient content, pH, and organic matter (MO) among habitats. Significance was determined using the Kruskal–Wallis test for each variable. Common letters
indicate no difference among pairwise habitats (Mann–Whitney test). Sign.: significance, ***: p< 0.001; **: p< 0.01; *: p< 0.05.
pH NH
4
PS K CaMgZnCuFeMnBMO
PF 5.17
a
58.9
a
5.7
a
10.26
a
0.242
ab
8.24
a
2.23
a
2.64
a
3.16
a
91
a
8.86
a
0.52
a
14.03
a
DF 5.61
b
85.3
b
16.0
b
8.66
a
0.186
a
8.27
a
1.43
ab
2.97
a
5.68
b
344
b
8.25
a
0.42
a
11.12
a
P 5.74
b
98.4
c
15.3
b
8.82
a
0.349
b
8.97
a
1.27
b
6.47
b
5.78
b
1105
c
8.55
a
0.48
a
9.35
a
Sign. ** *** *** – * – – *** ** *** – – –
1064 F. Anthelme et al. / Biological Conservation 144 (2011) 1059–1067
an altered structure of Ceroxylon populations (see e.g. Blundell and
Peart, 2004; Silva Matos et al., 1999). Moreover, local aggregation
of stemmed individuals and seedlings – often a specific feature of
rare species within inhospitable habitats (Hubbell, 1979) – indi-
cates deterioration in the fitness of populations of Ceroxylon in dis-
turbed forests. This is about to be confirmed by preliminary results
on the population genetics of Ceroxylon within the same sampling
design, which tend to indicate that intra-specific diversity is higher
in old-growth forests than in disturbed forests (Montúfar, unpub-
lished data).
Accordingly, the effects of the intermediate disturbance appear
to be antagonistic. We raise the hypothesis that our disturbed for-
ests were affected by both selective logging and adjacent fragmen-
tation within a growing matrix of pastures. Although we did not
analyse these two factors separately, our results suggest that the
high density of individuals would be due to the positive effects
of partial clearing by the logging of old live trees. By taking advan-
tage of these clearings Ceroxylon displays an interesting tolerance
to moderate disturbance.
Reversely, Ceroxylon may be negatively affected by concomitant
forest fragmentation through dispersal limitation and dissemina-
tion limitation, patterns that are reported commonly for neotropi-
cal palms (Fleury and Galetti, 2006; Galetti et al., 2006; Aguirre and
Dirzo, 2008). Weak dispersal is suggested to be a potentially criti-
cal factor in explaining the distribution of palms in TMCF (Sven-
ning et al., 2009), especially through the alteration of populations
of seed dispersers (Wright and Duber, 2001; Jorge and Howe,
2009). The populations of birds and mammals that disperse Cerox-
ylon (Mejía Londoño, 1999; Krabbe, 2000) have thus very likely
been altered by fragmentation as these types of animals have been
shown especially sensitive to fragmentation (Lefevre and Rodd,
2009; Chapman et al., 2010). This would explain the genetic bottle-
neck found, as commonly documented in neotropical forests (see
Lowe et al. (2005) for a review).
4.3. Distribution of Ceroxylon and plant–plant interactions
Our data indicated that canopy cover is obligatory for the regen-
eration of Ceroxylon without intervention. The shade provided by
trees drastically reduced the effects of light intensity on evapo-
transpiration and the relative humidity of soil, two well-docu-
mented mechanisms by which the tree canopy promotes plant fit-
ness (Callaway, 2007). The resulting absence of water stress
explains the balanced demographic structure of Ceroxylon in for-
ests. However, the higher density of individuals of all life stages
in disturbed forests as compared with old-growth forests indicated
that the wax palm takes advantage of selective logging of large
trees in disturbed forests. This mid-successional status makes it
very sensitive to deforestation, because its development can be
facilitated by selective logging but inhibited by complete defores-
tation. Future studies should determine more accurately the
threshold between these two states, which might also be relevant
for other mid- and late-successional species in the TMCF.
In pastures, the reduction in water stress above and below
ground that is generated by S. sphacelata and the absence of a neg-
ative effect on nutrient availability indicate that grass – or other
nurse plants – might exert a net positive effect on seedlings of
Ceroxylon under stressful conditions, despite the highly competi-
tive nature of this grass. This hypothesis, which needs to be ex-
plored in detail, is supported by the assertion that the probability
that facilitation will occur increases with distance from the funda-
mental niche optimum of a target species (see Gomez-Aparicio
(2009) for a review). This hypothesis is rarely considered with re-
spect to TMCF, in contrast to its frequent consideration for mar-
ginal alpine or desert ecosystems (Brooker et al., 2008; Maestre
et al., 2009) or even for dry tropical ecosystems (Gomez-Aparicio,
2009). The meta-analysis of Gomez-Aparicio demonstrated that
tropical plant communities are prone to be driven by facilitation
processes, but the author did not observe a difference between
wet and tropical environments. Nonetheless, these environments
might display very different responses to stress and disturbance
(e.g., Bongers et al., 2009). We recommend that studies should be
undertaken to test the stress gradient hypothesis on gradients
caused by deforestation in wet tropical forests, and to identify
more precisely the impact of this ecological process on the conser-
vation of these exceptionally rich ecosystems (see Maestre et al.
(2009) for adapted sampling designs).
4.4. Resilience and conservation
The most appropriate option for the conservation of Ceroxylon
would be to protect existing old-growth forests, because these
are: (1) the primary habitat of Ceroxylon; (2) the habitat where
interactions with other taxa, especially birds and mammals, are
the most intense (e.g. Becker et al., 2008); and (3) the habitat
where dispersal is most efficient, which is likely to generate a high-
er intra-specific diversity. The latter is a key factor when consider-
ing the long-term conservation of populations.
At the same time, disturbed forests might provide a comple-
mentary approach to the protection of old-growth forests to en-
hance the conservation of populations. This is supported by the
observation of a higher density of individuals in disturbed forests
than in old-growth forests, which agrees with the assertion that
tropical secondary forests can serve as habitat refuges for mature
tropical forest tree species (Norden et al., 2009). One concrete
application might be to use disturbed forests as a source of seed-
lings for transplantation to locations where adults are absent;
however, the rare seedlings that are found in low-density areas
should be preserved as beneficiaries of efficient dispersal.
Any attempt to regenerate Ceroxylon in pastures might appear
idealistic at first glance. Although it has been shown experimen-
tally that it is difficult to recover TMCF after decades of pasture
activity in areas dominated by S. sphacelata (Myster, 2007), the
reduction of abiotic stress by the grass provides a suitable habitat
for seedlings. The grass might also protect the seedlings from large
herbivores by providing cover (predator avoidance; Milchunas and
Noy-Meir, 2002). These findings support the hypothesis of Gunter
et al. (2009) that if S. sphacelata is not subjected to manual weed-
ing, it might facilitate reforestation. However, this species is known
to inhibit succession through the limitation of dispersal owing to
its spatial structure (Myster, 2004), as detected in our dataset.
Therefore, to benefit from the protective effects of the grass, it
would be necessary helping seeds or seedlings recruitment by
transplanting them inside the nurse plant, a location that they can-
not reach by themselves. This was not attempted in Myster’s
experiment.
Alternatively, abandonment of pastures can be considered a
passive pathway to promote the survival of Ceroxylon. This type
of intervention is appropriate when it is considered that, world-
wide, 15% of areas that were subjected to deforestation in the
1990s have been rehabilitated through natural secondary succes-
sion (Wright, 2005). Seed dispersers in tropical pastures tend to
avoid open areas (Laborde et al., 2005); therefore, the presence of
female trees in the pastures is a prerequisite for palm resilience,
corroborating the necessity of presence of remnant trees in pas-
tures for promoting forest regeneration (Schlawin and Zahawi,
2008). This approach might be improved by the transplantation
of shrubs in order to mimic classical grass-shrub-tree succession,
which has been shown to be far more efficient for forest restoration
than grass-tree succession (two-phase restoration strategy; Go-
mez-Aparicio, 2009).
F. Anthelme et al. / Biological Conservation 144 (2011) 1059–1067 1065
Facilitation is still considered infrequently during the develop-
ment of approaches to restore species within a habitat (Padilla
and Pugnaire, 2006), despite it being recognised increasingly as a
major ecological process in the structuring of plant communities
(Brooker et al., 2008). A recent example of grass-to-tree facilitation
(Anthelme and Michalet, 2009) can be used as an ecological prec-
edent to expand the use of this type of positive interaction to the
rehabilitation of endangered forest trees and their associated com-
munities in the face of massive deforestation in TMCF.
Acknowledgements
We thank L. Gomez-Aparicio, J.C. Pintaud, R. Bernal and an
anonymous reviewer for constructive comments on the manu-
script. We acknowledge the people of Rio Bravo and Inti-Llacta Re-
serves for their kind reception, and M. Grimaldi for comments on
soil microclimate. This research was funded by the Ecuadorian
government (ECOFONDO grant n°019-ECO7-Inv1), and the project
Palm Harvest Impacts in Tropical Forests-PALMS FP7-ENV-2007-I
(http://www.fp7-palms.org).
Appendix A
Influence of maximal light intensity on minimal soil humidity
and maximal VPD among habitats (DF: disturbed forest; Pn: pas-
ture with Setaria sphacelata; P: pasture without S. sphacelata). Lines
represent linear regressions for each habitat.
References
Aguirre, A., Dirzo, R., 2008. Effects of fragmentation on pollinator abundance and
fruit set of an abundant understory palm in a Mexican tropical forest. Biol.
Conserv. 141, 375–384.
Anthelme, F., Michalet, R., 2009. Grass-to-tree facilitation in an arid grazed
environment (Aïr mountains, Sahara). Basic Appl. Ecol. 10, 437–446.
Anthelme, F., Montúfar-Galárraga, R., Pintaud, J.C., 2010. Caracterización de la
resiliencia ecológica de poblaciones de palmeras. Ecol. Bolivia 45, 23–29.
Arroyo-Rodriguez, V., Aguirre, A., Benitez-Malvido, J., Mandujano, S., 2007. Impact of
rain forest fragmentation on the population size of a structurally important
palm species: Astrocaryum mexicanum at Los Tuxtlas, Mexico. Biol. Conserv. 138,
198–206.
Becker, C.D., Loughin, T.M., Santander, T., 2008. Identifying forest-obligate birds in
tropical moist cloud forest of Andean Ecuador. J. Field Ornithol. 79, 229–
244.
Blundell, A.G., Peart, D.R., 2004. Density-dependent population dynamics of a
dominant rainforest canopy tree. Ecology 85, 704–715.
Bongers, F., Poorter, L., Hawthorne, W.D., Sheil, D., 2009. The intermediate
disturbance hypothesis applies to tropical forests, but disturbance contributes
little to tree diversity. Ecol. Lett. 12, 798–805.
Borchsenius, F., Moraes, M., 2006. Diversidad y usos de palmeras andinas
(Arecaceae). In: Moraes, M.M., Øllgaard, R.B., Kvist, L.P., Borchsenius, F.,
Balslev, H. (Eds.), Botánica Económica de los Andes Centrales. Universidad
Mayor de San Andrés, La Paz, pp. 412–433.
Brooker, R.W., Maestre, F.E., Callaway, R.M., Lortie, C.L., Cavieres, L.K., Kunstler, G.,
Liancourt, P., Tielbörger, K., Travis, J.M.J., Anthelme, F., Armas, C., Coll, L.,
Corcket, M., Delzon, S., Forey, E., Kikvidze, Z., Olofsson, J., Pugnaire, F.I., Quiroz,
C.L., Saccone, P., Schiffers, K., Seifan, M., Touzard, B., Michalet, R., 2008.
Facilitation in plant communities: the past, the present, and the future. J. Ecol.
96, 18–34.
Brummit, N., Lughada, E.I., 2003. Biodiversity. Where’s hot and where’s not.
Conserv. Biol. 17, 1442–1448.
Callaway, R.M., 2007. Positive Interactions and Interdependence in Plant
Communities. Springer, Dordrecht.
Cayuela, L., Golicher, D.J., Rey Benayas, J.M., Gonzalez-espinoza, M., Ramirez-
marcial, N., 2006. Fragmentation, disturbance and tree diversity conservation in
tropical montane forests. J. Appl. Ecol. 43, 1172–1181.
Chapman, C.A., Chapman, L.J., Jacob, A.L., Rothman, J.M., Omeja, P., Reyna-Hurtado,
R., Hartter, J., Lawes, M.J., 2010. Tropical tree community shifts: implications for
wildlife conservation. Biol. Conserv. 143, 366–374.
Cottam, G., Curtis, J.T., 1956. The use of distance measures in phytosociological
sampling. Ecology 37, 451–460.
Debinski, D.M., Holt, R.D., 2000. A survey and overview of habitat fragmentation
experiments. Conserv. Biol. 14, 342–355.
Dufresne, E., Saugier, B., 1993. Gas exchange of oil palm in relation to light, vapour
pressure deficit, temperature and leaf age. Funct. Ecol. 7, 97–104.
Fleury, M., Galetti, M., 2006. Forest fragment size and microhabitat effects on palm
seed predation. Biol. Conserv. 131, 1–13.
Galetti, M., Donatti, C.I., Pires, A.S., Guimaraes, P.R., Jordano, P., 2006. Seed survival
and dispersal of an endemic Atlantic forest palm: the combined effects of
defaunation and forest fragmentation. Bot. J. Linn. Soc. 151, 141–149.
Goerner, A., Gloaguen, R., Makeschin, F., 2007. Monitoring of the Ecuadorian
mountain rainforest with remote sensing. J. Appl. Remote Sens. 1 (article
013527).
Gomez-Aparicio, L., 2009. The role of plant interactions in the restoration of
degraded ecosystems: a meta-analysis across life-forms and ecosystems. J. Ecol.
97, 1202–1214.
Guevara, M., Fuentes Pozo, P., Josse, C., Peñiafiel, M., 2001. Tres decadas de cambios
en el uso de la tierra en el area de Nanegal: 1966–1996. In: Rhoades, R.E. (Ed.),
Tendiendo Puentes entre los Paisajes Humanos y Naturales. Abya Yala, Quito,
pp. 35–54.
Gunter, S., Gonzalez, P., Alvarez, G., Aguirre, N., Palomeque, X., Haubrich, F., Weber,
M., 2009. Determinants for successful reforestation of abandoned pastures in
the Andes: soil conditions and vegetation cover. Forest Ecol. Manage. 258, 81–
91.
Hansen, M.C., Stehman, S.V., Potapov, P.V., 2010. Quantification of global gross
forest cover loss. Proc. Natl. Acad. Sci. Am. 107, 8650–8655.
Hering, K.G., 2003. A scientific formulation of tropical forest management. Ecol.
Model. 166, 211–238.
Holl, K.D., 1999. Factors limiting tropical rain forest regeneration in abandoned
pasture: seed rain, seed germination, microclimate, and soil. Biotropica 31,
229–242.
Hubbell, S.P., 1979. Tree dispersion, Abundance, and diversity in a tropical dry
forest. Science 203, 1299–1309.
Hubbell, S.P., He, F.L., Condit, R., Borda-de-Agua, L., Kellner, J., ter Steege, H., 2008.
How many tree species and how many of them are there in the Amazon will go
extinct? Proc. Natl. Acad. Sci. Am. 105, 11498–11504.
Jorge, M.L.S.P., Howe, H.F., 2009. Can forest fragmentation disrupt a conditional
mutualism? A case from central Amazon. Oecologia 161, 709–718.
Krabbe, N., 2000. Overview of conservation priorities for parrots in the Andean
region with special consideration for Yellow-eared parrot (Ognorhynchus
icterotis). Int. Zoo Yearbook 37, 283–288.
La Torre-Cuadros, M.A., Herrando-Perez, S., Young, K.R., 2007. Diversity and
structural patterns for tropical montane and premontane forests of Central
1066 F. Anthelme et al. / Biological Conservation 144 (2011) 1059–1067
Peru, with an assessment of the use of higher taxon surrogacy. Biodiv. Conserv.
16, 2965–2988.
Laborde, J., Guevara, S., Sanchez-Rios, G., 2005. Tree and shrub seed dispersal in
pastures: the importance of rainforest trees outside forest fragments.
Ecoscience 15, 6–16.
Lefevre, K.L., Rodd, F.H., 2009. How human disturbance of tropical rainforest can
influence avian fruit removal. Oikos 118, 1405–1415.
Lieberman, M., Lieberman, D., 2007. Nearest-neighbor tree species combinations in
tropical forest: the role of chance, and some consequences of high diversity.
Oikos 116, 377–386.
Lowe, A.J., Boshier, D., Ward, M., Bacles, C.F.E., Navarro, C., 2005. Genetic resource
impacts of habitat loss and degradation; reconciling empirical evidence and
predicted theory for neotropical trees. Heredity 95, 255–273.
Lucassen, E.C.H.E.T., Smolders, A.J.P., Boedeltje, G., van den Munckhof, P.J.J., Roelofs,
J.G.M., 2006. Groundwater input affecting plant distribution by controlling
ammonium and iron availability. J. Veg. Sci. 17, 425–434.
Maestre, F.T., Callaway, R.H., Vallarades, F., Lortie, C.J., 2009. Refining the stress-
gradient hypothesis for competition and facilitation in plant communities. J.
Ecol. 97, 199–205.
Mejía Londoño, G.D., 1999. Dispersión de Semillas de la Palma de Cera Ceroxylon
alpinum y estado actual de la población de Aves en un Bosque Montano del
Departamento del Quindío-Colombia. Pregraduate thesis, Universidad de los
Andes, Bogota, Colombia.
Milchunas, D.G., Noy-Meir, I., 2002. Grazing refuges, external avoidance of
herbivory and plant diversity. Oikos 99, 113–130.
Montúfar, R., Anthelme, F., Pintaud, J.C., Balslev, H., in press. A review on the effects
of disturbance on the resilience of palm populations and communities in
Neotropical forests. Bot. Rev.
Myster, R.W., 2004. Regeneration filters in post-agricultural fields of Puerto Rico
and Ecuador. Plant Ecol. 172, 199–209.
Myster, R.W., 2007. Early successional pattern and process after sugarcane, banana,
and pasture cultivation in Ecuador. New Zeal. J. Bot. 45, 101–110.
Norden, N., Chazdon, R.L., Chao, A., Jiang, Y.H., Vilchez-Alvarado, B., 2009. Resilience
of tropical rainforests: tree community reassembly in secondary forests. Ecol.
Lett. 12, 385–394.
OSU, 2009. Soil, Water and Forage Analytical Laboratory. <http://
www.soiltesting.okstate.edu/>.
Padilla, F.M., Pugnaire, F.I., 2006. The role of nurse plants in the restoration of
degraded environments. Front. Ecol. Environ. 4, 196–202.
Paredes-Ruiz, T.K., 1995. Primeros estudios de la Palma de Ramos (Ceroxylon
echinulatum) presente en Cosanga (Provincia del Napo) entre Agosto de 1991 y
Octubre de 1992. Pregraduate thesis, Pontificia Universidad Catolica de
Ecuador, Quito.
Pintaud, J.C., Anthelme, F., 2008. Ceroxylon echinulatum in an agroforestry system of
Northern Peru. Palms 52, 97–103.
Ponette-Gonzalez, A.G., Weatheres, K.C., Curran, L.M., 2010. Water inputs across a
tropical montane landscape in Veracruz, Mexico: synergistic effects of land
cover, rain and fog seasonality, and interannual precipitation variability. Glob.
Change Biol. 16, 946–953.
Sanín, M.J., Galeano, G., in press. Systematics of Ceroxylon Bonpl. ex DC. (Arecaceae:
Ceroxyloideae). Phytotaxa.
Scariot, A., 1999. Forest fragmentation effects on palm diversity in central
Amazonia. J. Ecol. 87, 66–76.
Schlawin, J., Zahawi, R.A., 2008. ‘Nucleating’ succession in recovering neotropical
wet forests: the legacy of remnant trees. J. Veg. Sci. 19, 485–492.
Silva Matos, D.M., Freckleton, R.P., Watkinson, A.R., 1999. The role of density
dependence in the population dynamics of a tropical palm. Ecology 80, 2635–
2650.
Svenning, J.C., 1998. The effect of land-use on the local distribution of palm species
in an Andean rain forest fragment in northwestern Ecuador. Biodiv. Conserv. 7,
1529–1537.
Svenning, J.C., Balslev, H., 1998. The palm flora of the Maquipucuña montane forest
reserve, Ecuador. Principes 42, 218–226.
Svenning, J.C., Harlev, D., Sorensen, M.M., Balslev, H., 2009. Topographic and spatial
controls of palm species distributions in a montane rain forest, southern
Ecuador. Biodiv. Conserv. 18, 219–228.
Trenel, P., Hansen, M.M., Normand, S., Borchsenius, F., 2008. Landscape genetics,
historical isolation and cross-Andean gene flow in the wax palm, Ceroxylon
echinulatum (Arecaceae). Mol. Ecol. 17, 3528–3540.
Valarezo, G.R., 2001. La gente, la tierra y la sociedad de Nanegal desde los tiempos
aborigenes. In: Rhoades, R.E. (Ed.), Tendiendo Puentes entre los Paisajes
Humanos y Naturales. Abya Yala, Quito, pp. 35–54.
Vergara-Chaparro, L.K., 2002. Demografía de Ceroxylon alpinum en bosques
relictuales del valle de Cocora, Salento (Quindio). Pregraduate thesis,
Departamento de Biología, Universidad Nacional de Colombia, Bogota.
Wright, J.S., 2005. Tropical forests in a changing environment. Trends Ecol. Evol. 20,
553–560.
Wright, J.S., Duber, H.C., 2001. Poachers and forest fragmentation alter seed
dispersal, seed survival, and seedling recruitment in the palm Attalea butyracea,
with implications for tropical tree diversity. Biotropica 33, 583–595.
F. Anthelme et al. / Biological Conservation 144 (2011) 1059–1067 1067