Forest Ecology and Management 258 (2009) 1260–1266

Contents lists available at ScienceDirect

Forest Ecology and Management

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

High levels of pollen dispersal detected through paternity analysis from a

continuous Symphonia globulifera population in the Brazilian Amazon
Francimary da Silva Carneiro a, Bernd Degen b, Milton Kanashiro a, Andre Eduardo Biscaia de Lacerda a,
Alexandre Magno Sebbenn c,*
a
Embrapa Amazônia Oriental, C.P. 48, 66095-100 Belém-PA, Brazil
b
Johann Heinrich von Thünen Institut, Institut für Forstgenetik, Sieker Landstrasse 2, D-22927 Grosshansdorf, Germany
c
Instituto Florestal de São Paulo, Caixa Postal 1322, 01059-970 São Paulo, Brazil

A R T I C L E I N F O A B S T R A C T

Article history: In this study, six highly polymorphic microsatellite loci and a categorical paternity analysis approach
Received 17 February 2009 were used to investigate the contemporary pollen gene flow in the neotropical tree species Symphonia
Received in revised form 8 June 2009 globulifera. Data for this study were taken from a 500 ha experimental plot in a dense terra firme forest in
Accepted 15 June 2009
the Eastern Brazilian Amazon and included the mapping and genotyping of 161 reproductive trees,
representing more than 90% of all adult trees, and the collection of 748 open-pollinated seeds from 56
Keywords: seed-trees over two field seasons (2002 and 2003). High levels of pollen immigration from outside of the
Tropical tree species
study plot were detected in both sampled seed-years (49%) suggesting long distance pollen gene flow.
Microsatellite
Contemporary gene flow
Low levels of self-fertilization were also detected (2%). The analysis showed long distance pollen
Paternity analysis dispersal occurred within the study area in both 2002 (d = 907  652 m SD) and 2003 (d = 963  542 m
SD). Patterns of pollen dispersal distance within the plot were also found to be shorter than the distances
between potential male parents and seed-trees. This result indicates that the distance between trees does not
explain the identified pollen dispersal pattern. Our results support the hypothesis that animal pollinated
species occurring in low-density populations can disperse pollen in long distances, despite the very dense
nature of the forest.
ß 2009 Elsevier B.V. All rights reserved.

1. Introduction Fragmentation and selective logging can therefore affect the

In recent years, tropical forests have experienced high rates of
deforestation through processes of clear cutting and logging which
have reduced many large continuous forests to isolated patches in
highly fragmented landscapes (Sebbenn et al., 2008). The process
of land conversion from forest to agriculture has turned continuous
tree populations into small isolated ones, while selective logging
has reduced the effective population size of continuous popula-
tions by harvesting the largest and reproductively active trees. The
fragmentation of populations and the removal of large numbers of
the reproductive population may affect the viability of many tree
species populations, especially in the species-rich Amazon tropical
forest. In the Amazon, more than 100 different tree species might
be found in a single hectare with most species occurring in very
low-densities, typically with less than 1 tree/ha (Dick et al., 2003a;
Azevedo et al., 2007; Carneiro et al., 2007; Cloutier et al., 2007;
Lemes et al., 2007; Lacerda et al., 2008; Silva et al., 2008).
* Corresponding author. Tel.: +55 19 34351681.
E-mail address: alexandresebbenn@yahoo.com.br (A.M. Sebbenn).
mating system of a species by increasing the distance between
conspecifics.
Studies have shown, however, that many low-density tropical
species have a generally high outcrossing rate, i.e. >0.8 (Ward
et al., 2005), suggesting that most of the pollen dispersal occurs
over long distances. These high outcrossing values can be partially
the result of inbreeding depression which eliminates selfed-seeds
from progeny arrays (Gribel and Gibbs, 2002; Hufford and
Hamrick, 2003; Naito et al., 2005; Del Castillo and Trujillo,
2007). Long distance pollen dispersal has been reported for many
low-density animal pollinated tropical trees in both natural and
fragmented stands (Dawson et al., 1997; White et al., 2002; Dick
et al., 2003a; Kenta et al., 2004; Latouche-Hallé et al., 2004; Lacerda
et al., 2008; Lourmas et al., 2007; Silva et al., 2008). Thus, it is
possible that some of these tropical tree species could be resilient
to various levels of spatial isolation caused by anthropogenic
factors, at both the individual and population levels.
Both pollen and seed dispersal are factors which influence the
gene flow of a population. Gene flow, in turn, determines the genetic
diversity both within a population and among different generations
of a population (Dow and Ashley, 1998). Understanding the extent to

0378-1127/$ – see front matter ß 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.foreco.2009.06.019
 F.S. Carneiro et al. / Forest Ecology and Management 258 (2009) 1260–1266 1261

which both pollen and seed dispersal influence gene flow is essential
in examining the potential of a species to adapt quickly to
environmental and land use changes and it is also imperative in
the creation and implementation of appropriate strategies for
conservation (Dawson et al., 1997; Odduo-Muratorio et al., 2005;
Larsen and Kjaer, 2008), logging (Kanashiro et al., 2002), and seed
collection for reforestation purposes (Bittencourt and Sebbenn,
2007).
Successful pollen dispersal can be directly studied in tree
populations using microsatellite loci, paternity analysis, and
progeny arrays, preferably from known seed-trees (in this study
seed-trees are defined as the individual tree which provided the
seed for germination and genetic study). Microsatellite loci, due to
their typically high polymorphism in terms of the number of low-
frequency alleles, generally present a high exclusion probability
(>0.99) and are therefore an ideal marker for paternity analysis
studies (Dawson et al., 1997; Dow and Ashley, 1998; Smouse and
Sork, 2004). Microsatellite paternity analysis assumes that there
are no mutations between parents and offsprings and that adults
and offsprings with matching genotypes are related (Dow and
Ashley, 1998). In order to study contemporary pollen flow, a well
defined area is required in which the seed-trees and putative
pollen donors are genotyped using microsatellite markers and the
spatial position of each individual is known (Smouse and Sork,
2004). The optimal study area can be a small forest fragment or a
Fig. 1. Spatial position of Symphonia globulifera trees in the 500 ha experimental plot
plot established inside a continuous forest. For low-density tree
at Tapajós National Forest (Brazilian Amazon).
populations (<1 tree/ha) in a continuous forest, the plot must be
sufficiently large to ensure a substantial number of trees for which
the paternity of the seeds can be tested (for example, at least 100 90 cm in diameter at breast height (dbh). In the Central Amazon,
putative pollen donors and therefore an area of more than 100 ha). hummingbirds are the principal pollinators (Bittrich and Amaral,
Thus, with a sufficient number of microsatellite loci evaluated in 1996; Maues, 2001) and the density of the population is less than
progeny arrays and collected within a well defined population or 1 tree/ha (Carneiro et al., 2007).
plot, pollen gene flow can be estimated. This estimate considers the
percentage of pollen donors for which all trees in the plot are 2.2. Site and the sample size
excluded as possible fathers of the seeds. From this, pollen
dispersal distance can be traced by analyzing the relative locations The study population is situated in a 500 ha experimental plot
of the seed-trees and the possible pollen donors (Dow and Ashley, in the Tapajós National Forest (TNF) south of Santarém, Pará State,
1998; Burczyk et al., 2004; Smouse and Sork, 2004). Brazil (28510 S and 548570 W and 175 m above sea level) (Fig. 1). The
In the present study, pollen gene flow was investigated over plot is located on a flat plateau which is covered by a dense terra
two consecutive years in a continuous low-density Symphonia firme forest and the area is not influenced by seasonal river
globulifera population, located within an experimental plot in flooding. The climate is tropical with a mean annual rainfall of
Tapajós National Forest, Brazil. In a previous study within the same 2000 mm (concentrated from February to May) and a dry season of
plot, S. globulifera’s mating system was shown to occur only by two to three months (August–October) with annual temperatures
outcrossing with a small proportion of mating occurring among averaging at 25 8C. The experimental plot was established by the
relatives (Carneiro et al., 2007). Carneiro et al.’s (2007) study used Dendrogene Project (Embrapa Amazônia Oriental/DFID) as an
an indirect method (TWOGENER analysis) to determine pollen Intensive Study Plot for Ecological and Genetic Studies (Kanashiro
dispersal distance which was estimated between 144 and et al., 2002).
444 m. In our study, a direct method of pollen dispersal is used The 500 ha plot was subdivided into six subplots: four 100 ha
based on paternity analysis and six microsatellite loci in S. subplots (subplots 3–6) and two 50 ha subplots each (subplots 1
globulifera. This study seeks to address the following questions: (i) and 2) (Fig. 1). Within each subplot, all S. globulifera trees with a
what is the proportion of pollen immigration from outside of the diameter at breast height (dbh) of 20 cm or greater were identified
plot? (ii) Considering the low-density of reproductive trees in the and mapped, resulting in a density of 0.332 tree/ha. The distance
study site, what is the average distance of pollen dispersal? (iii) Is among trees in all subplots ranged from 14 to 2962 m, with an
the pollen dispersal within the plot a function of the spatial average of 1234 m.
distance between seed-trees and potential pollen donors? Data collection for this study included cambium samples from
90% of all mapped trees (161 sampled out of a total of 179 trees).
2. Materials and methods Seeds were collected from 26 seed-trees in 2002 and from 30 seed-
trees in 2003 (Fig. 1). In order to ensure accuracy in seed sampling,
2.1. Study species collection of seeds was conducted directly in the tree’s canopy.
Seeds from the same trees were collected in both years from 19
S. globulifera L. F. (Clusiaceae) is a hermaphroditic animal individuals because not all seed-trees sampled in 2002 produced
pollinated tropical tree species with a wide natural distribution. S. seeds in 2003. Seeds were germinated as a single tree’s progeny.
globulifera occurs in the neotropics, from Mexico to Rio de Janeiro Microsatellites were amplified for 261 seeds from 2002 (7–15
State (South-eastern Brazil), and is also found in tropical West seeds per tree) and 487 seeds from 2003 (12–30 seeds per tree).
Africa (Dick et al., 2003b). In the Amazon forest, the species is a The distance between sampled seed-trees varied between 26 and
medium sized tree, reaching up to 40 m in height and more than 2951 m with a mean of 1282 m in 2002 and 1295 m in 2003. The
1262 F.S. Carneiro et al. / Forest Ecology and Management 258 (2009) 1260–1266

the critical D. For simulations we used the following parameters:

50,000 repetitions; the proportion of loci at 0.01; confidence levels
of 95% (strict) and 80% (relaxed); and 161 sampled individuals (all
adult trees in the plot) as probable male candidate parents for each
seed-tree.
As the location of a seed-tree could affect the rate of pollen
immigration, seed-trees were classified as central seed-trees
(closer to the center of the plot defined by a minimum distance
to the edge ranging from 245 to 940 m) or as outer seed-trees
(closer to the edge of the plot defined by a minimum distance to the
edge ranging from 0 to 240 m) (Fig. 1). Paternity analysis was
carried out separately for each class. The proportion of candidate
males within the plot was assumed to be 90% for central seed-trees
and 50% for outer seed-trees, because outer seed-trees are likely to
Fig. 2. Frequency distribution of dbh classes in Symphonia globulifera population in
the 500 ha experimental plot at Tapajós (Brazilian Amazon). have a greater number of neighbors outside of the plot. Progenies
with no probable paternal candidate within the plot were assumed
to represent pollination from trees located outside of the plot
average dbh was 29.9 cm (standard deviation of 13.1 cm) and the (pollen immigration). From paternity analysis, the pollen dispersal
largest was 94 cm (Fig. 2). distance was calculated based on the position of the seed-tree and
the assumed parent. In order to investigate if the mating success is
2.3. Microsatellite analysis a function of distance between trees and sampled seed-trees, the
frequency distribution of effective pollen dispersal was compared
DNA was extracted from the cambium of adult trees and from with the frequency distribution of the distances among all
the seedlings according to the protocol described by Colpaert et al. reproductive trees and the seed-trees in the plot; this analysis
(2005). Six primer pairs Sg03, Sg18, Sg06, Sg08, Sg09 and Sg10 was conducted for both years using the Kolmogorov–Smirnov test
were used for microsatellite analysis according to Degen et al. (Sokal and Rohlf, 1981). To determine if the patterns of pollen
(2004) and Vinson et al. (2005). Amplification products were dispersal were different between years, the frequency distribution
visualized by electrophoresis with a LIZ-500 size Standard (Applied of effective pollen dispersal in 2002 was compared with the
Biosystems Incorporated, ABI) on an ABI PRISM1 310 automated frequency distribution of effective pollen dispersal in 2003, also
sequencer. Allele sizes were scored using the program Genotyper1 using the Kolmogorov–Smirnov test. This test was also made only
Software from Applied Biosystems. The observed segregation of for the same seed-trees sampled in both years. We also tested the
alleles within families showed co-dominant inheritance of the hypotheses that center seed-trees received 90% of the pollen from
microsatellite fragments (Carneiro et al., 2007). trees located within the plot and 10% from outside the plot, and
outer seed-trees received 50% of the pollen from trees located
2.4. Analysis of genetic diversity and fixation index within and 50% from outside the plot, using a log-likelihood G test
(Sokal and Rohlf, 1981).
Genetic diversity was analyzed using the total number of alleles
from all loci (k), average number of alleles per locus (A), observed 3. Results
heterozygosity (Ho), and expected heterozygosity of adult trees
under Hardy–Weinberg equilibrium (He). The fixation index (F) 3.1. Genetic diversity and fixation index
was calculated by F̂ ¼ 1  ðĤo =Ĥe Þ. The statistical significance of F
for each locus and average loci were tested using permutation. All A high level of polymorphism and genetic diversity were
cited indexes and permutations were calculated using the FSTAT detected in the loci used in this study. Nineteen to 29 alleles were
program (Goudet, 1995). The paternity exclusion of the second detected among locus in the total sample (759 individuals),
parent [Pr(Ex2)] was estimated using the program CERVUS 3.0 resulting in 146 alleles and an average of all loci of 24.4. Among the
(Marshall et al., 1998; Kalinowski et al., 2007). sampled population, 144 alleles were detected in adult trees
(Table 1), 133 in seeds from 2002, and 136 in seeds from 2003. Of
2.5. Paternity analysis all adult trees (Table 1), the observed heterozygosity was high and
ranged from 0.728 to 0.866, with an average of 0.783. The expected
Paternity analysis was carried out by maximum-likelihood heterozygosity was higher than the observed heterozygosity
assignment (Meagher, 1986) using the CERVUS 3.0 program.
Results from the program were based on multilocus genotypes of Table 1
206 seeds from 2002 and 392 seeds from 2003 and a population of Characteristics of the six microsatellite loci for adult trees of Symphonia globulifera
(n = 161).
161 reproductive trees (dbh  20 cm). The seeds genotyped with
less than four loci were excluded from the paternity analysis Locus k Ho He F PExcl(2)
following the accepted current methodology which requires a
Sg03 21 0.857 0.883 0.029 0.239
higher number of loci analyzed. Therefore of the 261 seeds Sg18 19 0.866 0.888 0.025 0.228
genotyped from 2002, 55 were discarded, while 95 of the 487 Sg06 29 0.762 0.932 0.182** 0.140
genotyped in 2003 were also rejected. The most likely parents and Sg08 24 0.744 0.891 0.164** 0.211
parent-pairs were determined by D statistic (Marshall et al., 1998) Sg09 24 0.740 0.925 0.200** 0.157
Sg10 29 0.728 0.925 0.213** 0.152
using the allele frequencies of the adult population as reference. Means 24 0.783 0.907 0.137**
The significance of D (critical D) was determined through paternity All locus 144 0.99996
tests simulated by the software. From this, the tree with the
For each loci: observed number of alleles (k); observed heterozygosity (Ho);
highest calculated D value was accepted as father of a seed if the expected heterozygosity (He); fixation index (F); second parent exclusion
difference between its LOD score (logarithm of likelihood ratios) probability (PExcl(2)).
and the second most likely candidate’s LOD score was greater than **
P < 0.01.
 F.S. Carneiro et al. / Forest Ecology and Management 258 (2009) 1260–1266 1263

Table 2
Profiles of pollen flow for grouped seed-trees located closer to the center (central) and closer to the edges (outer) of the plot for seeds collected in 2002 and 2003. Sample size
(n); self-fertilization rate (s); pollen flow from within plot, pollen immigration; mean pollination distance within the plot (standard deviation) in Symphonia globulifera.

Sample n s Pollen flow Distance (m)

Within plot Outside plot Including selfing Excluding selfing Min/max

2002
Central 64 0.016 (1) 0.88 (56) 0.13 (8) 777  457 791  457 87/1771
Outer 142 0.021 (3) 0.54 (77) 0.46 (65) 955  770 994  760 46/2832
Total 206 0.019 (4) 0.65 (133) 0.35 (73) 880  660 907  652 46/2832

2003
Central 173 0.023 (4) 0.63 (109) 0.37 (64) 883  463 917  437 26/2061
Outer 219 0.027 (6) 0.53 (116) 0.47 (103) 954  648 1006  625 46/2658
Total 392 0.026 (10) 0.57 (225) 0.43 (167) 919  566 963  542 26/2658

Min/max: minimum and maximum distance of pollen dispersal, excluding selfing.

(ranging from 0.883 to 0.932, with an average 0.907), suggesting
heterozygosity deficiency in relation to the Hardy–Weinberg
principle. The fixation index ranged from 0.025 to 0.213, with an
average of 0.137. The fixation index was statistically significantly
different for four loci among the six analyzed as well as for the
average of all loci (P < 0.01).
3.2. Paternity analysis
The paternity exclusion probability of the second parent for
six loci was very high in this study (0.99996; Table 1). Thus, the
probability of cryptic gene flow (foreign gametes with multi-
locus genotypes that can also be generated by local males) was
very low, resulting in 0.0064 (1  0.99996161; Dow and Ashley,
1996). Of the 206 seeds from 2002 and 392 seeds from 2003
genotyped for at least five loci, pollen donors could be assigned
for 137 (67%) and 235 seeds (60%), respectively (Table 2). This
result indicates high levels of pollen immigration into the study
area in 2002 and 2003. A low number of self-sibs was found for
both years and the average selfing rate was estimated at 1.9% in
2002 and 2.6% in 2003. Excluding selfing, father assignment was
higer for seed-trees located in the center of the plot than seed-
trees near the edges. In the central area of the plot, father
assignment rates were 88% in 2002 and 63% in 2003. At the outer
area, father assignment rates were 54% in 2002 and 53% in 2003.
In testing the hypotheses regarding location of pollen donors the
results showed a statistically significant difference for the center
seed-trees (90% within, 10% outside the plot) only in 2003
(G = 86.9, P < 0.01). In this case, an alternative hypothesis was
tested, assuming that 70% of the pollen was received from trees
located within the plot and 30% from outside the plot. This
hypothesis showed no statistical significance (G = 3.09,
P > 0.05). No significant differences were observed for the
hypothesis of outer seed-trees receiving 50% of the pollen from
trees located within the plot and 50% from outside the plot in
2002 (G = 1.02, P > 0.05) and 2003 (G = 0.77, P > 0.05). Thus, the
location of a seed-tree can affect the rate of pollen immigration.
Of the 161 potential pollen donors, 66 (41%) sired at least one
seed in 2002 (1–7 seeds), while in 2003, 82 (51%) donors sired at
least one seed (1–16 seeds). Therefore, over the two-year study
period 111 potential donors (69%) sired at least one seed.
3.3. Pollen dispersal distance
The results of the analysis showed that pollen dispersal occurs
over long distances within the study area (Table 2, Fig. 3). Including
selfing, the average distance of pollen dispersal (d) for seeds
collected in 2002 was 880 m (660 m, SD) and for seeds sampled in
2003 was 919 m (566 m, SD). Excluding selfing, the distance of
pollen dispersal measured for 2002 ranged from 46 to 2832 m
(d = 907  652 m) and in 2003 from 26 to 2658 m (d = 963  542 m).
In both years, there is a significant negative correlation between
effective pollen dispersal and the distance between seed-trees and
assigned pollen donors (r2002 = 0.93, P < 0.05; r2003 = 0.80,
P < 0.05; r2002–2003 = 0.62, P < 0.05). This negative correlation
suggests that mating tends to occur among trees which are closer

Fig. 3. Frequency distribution of pollen-dispersal distance for seeds from 2002 (a), 2003 (b), 2002 and 2003 (c) and distance between reproductive trees and seed-trees (d) in a
plot of Symphonia globulifera.
1264 F.S. Carneiro et al. / Forest Ecology and Management 258 (2009) 1260–1266
together. Our estimate of the correlation between dbh of possible
pollen donors and the number of produced sibs was not statistically
significant (r = 0.11, P > 0.05).
Excluding selfing, 50% of the pollen was dispersed up to
748 m in 2002 and up to 880 m in 2003 (median values).
Considering each year individually and both years together,
more than 90% of the pollen was dispersed up to 1000 m (Fig. 3).
Significant differences between the frequency distribution of
detected pollen dispersal distance and the frequency distribu-
tion of seed-trees and all other sampled trees in the plot was
detected by the Kolmogorov–Smirnov test for seeds in the three
situations: 2002 (D = 0.236, P < 0.0001), 2003 (D = 0.132,
P = 0.004), and both years combined (D = 0.147, P < 0.0001;
Fig. 3). These results indicate that the distance between
potential male parents and seed-trees cannot explain the
observed mating patterns. This test also shows that the
frequency distribution of pollen dispersal distance in 2002
was significantly different from 2003 (D = 0.256, P < 0.0001).
However, when this test was made only for the same seed-trees
sample in both years, no difference was observed (D = 0.134,
P < 0.149). Pollen dispersal distance was also estimated
exclusively for seed-trees located closer to the edges (outer)
of the plot and those closer to the center of the plot (central).
Excluding selfing, the average pollen dispersal distance for
central seed-trees was higher than outer seed-trees with outer
at 791  457 m in 2002 and 917  437 m in 2003 and central at
994  760 m in 2002 and 1006  625 m in 2003 (Table 2).
4. Discussion
4.1. Outcrossing rate
Despite the low-density of reproductive trees in the studied S.
globulifera population, the results of paternity analysis show the
maintenance of very high outcrossing rates and, consequently,
long pollen dispersal distances. Very low levels of selfing were also
found in both reproductive events with a maximum of 2%. These
results are similar to previous estimates of multilocus outcrossing
rates carried out by Carneiro et al. (2007). Although the study
conducted by Carneiro et al. (2007) used the same data set as this
study, the mixed mating-model (Ritland and Jain, 1981) was used
and resulted in outcrossing rates of 1.0 in 2002 and 0.998 in 2003.
Elsewhere, self-fertilization rates have been reported by Aldrich
and Hamrick (1998) for S. globulifera populations in continuous
forests, fragmented forests and pastures in Costa Rica, with a
minimum outcrossing rate of 0.731. Degen et al. (2004) also found
some levels of self-fertilization and an outcrossing rate of 0.920 in a
high-density S. globulifera population (>10.6 trees/ha) in French
Guiana. The variations in outcrossing rates between sites suggest
that the species is not self-incompatible and some self-fertilization
does occur. Therefore, self-incompatible systems can be excluded
as the possible cause of the observed high outcrossing rate in this
study.
An alternative hypothesis is the inbreeding depression that
occurs between the fertilization event and the moment of the
seedlings’ microsatellite analysis. The inbreeding depression could
eliminate part of the self-fertilized embryos, leaving only the
genotypes of outcrossed seedlings accessible to markers, thus
increasing the outcrossing rate estimate. In this case, the observed
outcrossing rate is not the realized one, but the effective
outcrossing rate. This phenomenon has been observed in various
tropical tree species, such as Pseudobombax munguba (Mart. et
Zucc.) Dugand (Gribel and Gibbs, 2002), Platypodium elegans Vog.
(Hufford and Hamrick, 2003), Neobalanocarpus heimii (King)
Ashton (Naito et al., 2005) and Pinus chiapensis (Martı́nez)
Andresen (Del Castillo and Trujillo, 2007).
4.2. Patterns of pollen dispersal
The cryptic gene flow estimate was very low (0.0064) in this
study, indicating that it did not skew the estimates of pollen
dispersal. Therefore, the seeds from both years that could not be
assigned to any potential pollen donor within the plot are likely to
represent pollen that originated from trees located outside the
study area or from the 18 S. globulifera trees in the plot that were
not genotyped.
Our results show extensive pollen movement in the study plot
with a minimum pollen immigration of 35%. However, according
to Kolmogorov–Smirnov test, patterns of effective pollen
dispersal were not explained by distance between seed-trees
and all other potential male parents within plot (Fig. 3). Pollen
was predominantly dispersed in shorter distances than expected
with 63% of the pollen travelling less than 1000 m. The maximum
average pollen dispersal distance within the plot was 963 m
(seeds from 2003; Table 2) which is shorter than the average
distance between seed-trees and all other reproductive trees
(1234 m), thus indicating a tendency toward pollination between
trees which are closer together. This pattern is likely affected by
the reproductive phenology of trees in the study area. During
data collection we observed that not all reproductive trees in the
plot produce flowers and fruits every year suggesting flowering
of individuals may not be fully synchronised. As a consequence,
only approximately 50% of the trees within the plot fathered
seeds in each year studied. Thus, flowering phenology limits the
mating among reproductive trees and might explain the
differences in pollen dispersal patterns which in this study are
related to the occurrence of pollinization between trees that are
closer together as well as the difference in pollinization distance
between years.
The higher frequency of short-distance pollen dispersal is
similar to that observed in other studies of pollen dispersal in
various tree species. Similar results were reported for the
tropical animal pollinated Dicorynia guianensis Amsh (Latouche-
Hallé et al., 2004), Dipterocarpus tempehes Slooten (Kenta et al.,
2004), Entandrophragma cylindricum (Sprague) Sprague (Sapeli)
(Lourmas et al., 2007), Hymenaea courbaril L. (Lacerda et al.,
2008), and Shorea acuminate Dyer (Naito et al., 2008). The
distances of pollen dispersal detected from paternity analysis in
this study were also higher than the results detected from
previous studies. Carneiro et al.’s (2007) study using the
same data set and TWOGENER analysis (Smouse et al., 2001)
reported pollen dispersal rates of 444 m in 2002 and 154 m in
2003. Similarly, a study from a high density S. globulifera
population in Paracou, French Guiana showed dispersal rates
ranging from 27 to 53 m (Degen et al., 2004). The difference in
pollen dispersal between the Amazon forest population and
Paracou is likely due to the differences in population density
as the Amazon population has a low-density at 0.322 tree/ha,
and Paracou a very high density at more than 10.6 trees/ha,
as discussed by Carneiro et al. (2007). The lower estimates
of pollen dispersal distance by TWOGENER analysis, as reported by
Carneiro et al. (2007), are the results of the type of analysis
used in the study. TWOGENER analysis is an indirect method of
quantification of contemporary pollen dispersal and is based
on assumptions about the shape of the pollen dispersal
curve (normal, exponential, leptokurtic, etc.; Smouse and Sork,
2004). As a consequence, TWOGENER analysis tends to under-
estimate average pollen dispersal distance due to an emphasis
on effective pollen movement (Smouse and Sork, 2004). In
contrast, the paternity analysis used in this study is a direct
method which is not dependent on dispersal functions and is
therefore a more accurate estimate of contemporary pollen
dispersal.
 F.S. Carneiro et al. / Forest Ecology and Management 258 (2009) 1260–1266
4.3. Center versus outer seed-trees
The results confirmed our hypothesis that the location of the
seed-trees can affect the rate of pollen immigration. Seed-trees
located in the center of the plot had a significantly higher number
of pollen donors detected inside of the plot than seed-trees located
close to the edges of the plot. In addition, as hypothesized, the
results showed that outer seed-trees received about 50% of the
pollen from trees located within plot and 50% from outside. These
results are significant and they suggest that studies of the patterns
of pollen dispersal in plots established in continuous forest areas
must consider the location of selected seed-trees in order to
increase father assignment and to better estimate the frequency
distribution of effective pollen dispersal.
Differences were also observed between years for the patters of
pollen dispersal (Fig. 3). Patterns and distance of pollen dispersal,
as well as mating systems, can change between years as a result of
changes in the flowering phenology, environmental changes
affecting the behaviour of the pollinators, or anthropogenic
processes such as forest logging and fragmentation. In this study,
the seed-trees sampled in both years showed no difference in the
patterns of pollen dispersal. The observed differences between
years can be attributed to the variation of trees sampled in each
year. However, differences between years in flowering phenology
are common and studies of pollen gene flow must consider open-
pollinated seeds from multiple reproductive events because a
single pollen dispersal event may not represent the population
pattern. Similar results were observed in D. tempehes (Kenta et al.,
2004).
4.4. Implications for selective logging
S. globulifera is harvested in the Brazilian Amazon for timber
production using selective logging practices (Carneiro et al.,
2007; Sebbenn et al., 2008). The selective logging approach
applied in Brazil is an adaptation of the reduced impact logging
(RIL) system which may remove up to 90% of the largest trees
(dbh  50 cm) during a single harvesting episode. Trees with bad
stem form, hollows or other apparent defects are not logged and
are left as ‘‘seed-trees’’, individuals designated to supply seeds
for regeneration. This type of logging reduces the density of
reproductive trees and increases the distance among reproduc-
tive conspecifics. Changes in reproductive population density
caused by forest logging can affect a species’ mating system, by
increasing self-fertilization and correlated mating levels (Lour-
mas et al., 2007). Such changes in mating systems have been
detected in studies comparing outcrossing rates for trees located
in continuous forests against rates from logged populations
(Murawski et al., 1994; Obayashi et al., 2002; Ledig et al., 2005).
Changes in outcrossing rates due to anthropogenic factors were
also reported for S. globulifera (Aldrich and Hamrick, 1998). Their
results suggest that anthropogenic processes increase the
distance among conspecifics and reduce the density of repro-
ductive populations which can affect the mating system of tree
species.
The results of this study suggest that in an S. globulifera
population located in a continuous, dense forest, the outcrossing
rates are very high. Considering the results of Aldrich and Hamrick
(1998), the reduction in population density and an increase in
distance among conspecifics by logging could increase the self-
fertilization rate in the population. However, S. globulifera is a
middle-size low-density tree species in the Amazon and in our
study area the dbh class distribution showed a positive skewed
shape (Fig. 2). Therefore, logging up to 90% of trees with a dbh of
50 cm would reduce the reproductive population from 161 to
131 trees (18%) and would have a limited effect on the distance
1265
between reproductive trees, increasing it from 1234 m before
logging to 1363 m after logging. However, the largest trees have
also been described in some studies as the main contributors to
pollination (Latouche-Hallé et al., 2004; Lourmas et al., 2007).
Changes in basal area of the population along with a reduction in
the number of the largest trees could, therefore, affect the
quantity of pollen dispersal. However, no association was found
between dbh of possible pollen donors and the number of
produced sibs. Thus, we do not expect that selective logging
practices used in the Brazilian Amazon will substantially affect
the mating system of the present population. These findings are in
agreement with the results of Carneiro et al. (2007) which
suggested that logging will not isolate the remaining trees in the
study area.
Further studies modelling the effects of selective logging on tree
populations also suggest the possible resilience of S. globulifera to
fragmentation and isolation due to logging. Degen et al. (2006) and
Sebbenn et al. (2008) have used Ecogene simulation analysis to
study the long-term effects of selective logging practices on
demographic structure and genetic diversity for four tropical tree
species. Both studies observed that the S. globulifera species was
unique in that it would likely be able to sustain the harvesting
methods actually employed in French Guiana (logging intensity of
90%, minimum cutting dbh of 60 cm, and cutting cycles of 65 years)
in the long term (six cutting cycles). This result can be explained by
the small number of reproductive trees logged in each cutting cycle
due to the high minimum cutting diameter. However, this
hypothesis needs to be empirically confirmed through a study
comparing the mating system and pollen dispersal before and after
logging events.
Acknowledgements
The study was financially supported by grants from DFID, UK
and EMBRAPA Brazil ‘‘Dendrogene Project’’, by a grant for regional
co-operation between French Guiana and Brazil ‘‘Formation
supérieure et diffusion des résultats de la recherche pour la gestion
durable des forêts de Guyane Française et d’Amazonie Brésilienne’’ as
well as by the EU-INCO project ‘‘Distribution of genetic diversity in
tree species from the Neotropics based on DNA fingerprinting assays:
Implications for conservation, sustainable utilization and manage-
ment’’ (contract number: ICA4-CT-2001-10101, http://thoth.
nbu.ac.uk/geneo). We thank Eric Bandou for his excellent technical
assistance in the lab and Evelyn Nimmo for her assistance with
the editing of the text. We are very grateful to three anonymous
referees for helpful comments and constructive suggestions on a
previous version of the manuscript.
References
Aldrich, P.R., Hamrick, J.L., 1998. Reproductive dominance of pasture trees in a
fragmented tropical forest mosaic. Science 281, 103–105, doi:10.1126/science.
281.5373.103.
Azevedo, V.C.R., Kanashiro, M., Ciampi, A.Y., Grattapaglia, D., 2007. Genetic struc-
ture and mating system of Manilkara huberi (Ducke) A. Chev., a heavily logged
Amazonian timber species. J. Hered. 98, 646–654, doi:10.1093/jhered/esm074.
Bittencourt, J.M., Sebbenn, A.M., 2007. Patterns of pollen and seed dispersal in a
small fragmented population of a wind pollinated Araucaria angustifolia in
southern Brazil by paternity analysis. Heredity 99, 580–591 DOI:0018-067X/07.
Bittrich, V., Amaral, M.C.E., 1996. Pollination biology of Symphonia globulifera
(Clusiaceae). Plant. Syst. Evol. 200, 101–110, doi:10.1007/BF00984751.
Burczyk, J., DiFazio, S.P., Adams, W.T., 2004. Gene flow in forest trees: how far do
genes really travel? Forest Genet. 11, 1–14.
Carneiro, F., Sebbenn, A.M., Kanashiro, M., Degen, B., 2007. Low inter-annual
variation of mating system and gene flow of Symphonia globulifera in the
Brazilian Amazon. Biotropica 39, 628–636, doi:10.1111/j.1744-7429.2007.
00314.x.
Cloutier, D., Kanashiro, M., Ciampi, A.Y., Schoen, D.J., 2007. Impacts of selective
logging on inbreeding and gene dispersal in an Amazonian tree population of
Carapa guianensis Aubl. Mol. Ecol. 16, 797–809, doi:10.1111/j.1365-294X.
2006.03193.x.
1266 F.S. Carneiro et al. / Forest Ecology and Management 258 (2009) 1260–1266
Colpaert, N., Cavers, S., Bandou, E., Caron, H., Gheysen, G., Lowe, A.J., 2005. Sampling
tissue for DNA analysis of trees: trunk cambium as an alternative to canopy
leaves. Silvae Genet. 6, 265–269.
Dawson, I.K., Waugh, R., Simons, A.J., Powell, W., 1997. Simple sequence repeats
provide a direct estimate of pollen-mediated gene dispersal in the tropical tree
Gliricidia sepium. Mol. Ecol. 6, 79–183, doi:10.1046/j.1365-294X.1997.00163.x.
Degen, B., Bandou, E., Caron, H., 2004. Limited pollen dispersal and biparental
inbreeding in Symphonia globulifera in French Guiana. Heredity 93, 585–591,
doi:10.1038/sj.hdy.6800560.
Degen, B., Blanc, L., Caron, H., Maggie, L., Kremer, A., Gourlet-Fleury, S., 2006. Impact
of selective logging on genetic composition and demographic structure of four
tropical tree species. Biol. Conserv. 131, 386–401, doi:10.1016/j.biocon.2006.
02.014.
Del Castillo, R.F., Trujillo, S., 2007. Effect of inbreeding depression on outcrossing
rates among populations of a tropical pine. New Phytol. 177, 517–524,
doi:10.1111/j.1469-8137.2007.02260.x.
Dick, C.W., Abdul-Salim, K., Bermingham, E., 2003b. Molecular systematic reveals
cryptic Terceary diversification of a widespread tropical rainforest tree. Am.
Nat. 162, 691–703, doi:10.1086/379795.
Dick, C.W., Etchelecu, G., Austerlitz, F., 2003a. Pollen dispersal of Neotropical trees
(Dinizia excelsa: Fabaceae) by native insects and Africa honeybees in pristine
and fragmented Amazonian rainforest. Mol. Ecol. 12, 753–764, doi:10.1046/
j.1365-294X.2003.01760.x.
Dow, B.D., Ashley, M.V., 1996. Microsatellite analysis of seed dispersal and par-
entage of sampling in bur oak, Quercus macrocarpa. Mol. Ecol. 5, 615–627,
doi:10.1111/j.1365-294X.1996.tb00357.x.
Dow, B.D., Ashley, M.V., 1998. High levels of gene flow in bur oak revealed by
paternity analysis using microsatellites. J. Hered. 89, 62–70, doi:10.1093/
jhered/89.1.62.
Goudet, J., 1995. Fstat (Version 2.9.3.2.): a computer program to calculate F-
statistics. J. Hered. 86, 485–486.
Gribel, R., Gibbs, P.E., 2002. High outbreeding as a consequence of selfed ovule
mortality and single vector bat pollination in the Amazonian tree Pseudobombax
munguba (Bombacaceae). Int. J. Plant Sci. 163, 1035–1043, doi:10.1086/342518.
Hufford, K.M., Hamrick, J.L., 2003. Variability selection at three early life stages of
the tropical tree Platypodium elegans (Fabaceae, Papilonoidae). Evolution 57,
518–526, doi:10.1111/j.0014-3820.2003.tb01543.x.
Kalinowski, S.T., Taper, M.L., Marshall, T.C., 2007. Revising how the computer
program CERVUS accommodates genotyping error increases success in paternity
assignment. Mol. Ecol. 16, 1099–1106, doi:10.1111/j.1365-294x.2007.03089.x.
Kanashiro, M., Thompson, I.S., Yared, J.A.G., Lovelless, M.D., Coventry, P., Martins-
da-Silva, R.C.V., Degen, B., Amaral, W.A., 2002. Improving conservation values of
managed forest: the Dendrogene Project in the Brazilian Amazon. Unasylva 209,
26–33.
Kenta, T., Isagi, Y., Nakagawa, M., Yamashita, M., Nakashizuka, T., 2004. Variation in
pollen dispersal between years with different pollination conditions in a
tropical emergent tree. Mol. Ecol. 13, 3575–3584, doi:10.1111/j.1365-294X.
2004.02345.x.
Lacerda, A.E.B., Sebbenn, A.M., Kanashiro, M., 2008. Long-pollen movement and
deviation of random mating in a low-density continuous population of Hyme-
naea courbaril in the Brazilian Amazon. Biotropica 40, 462–470, doi:10.1111/
j.1744-7429.2008.00.402.x.
Larsen, A.S., Kjaer, E.D., 2008. Pollen mediated gene flow in the native population of
Malus sylvestris and its implications for contemporary gene conservation man-
agement. Conserv. Genet., doi:10.1007/s10592-008-9713-z.
Latouche-Hallé, C., Ramboer, A., Bandou, E., Caron, H., Kremer, A., 2004. Long-
distance pollen flow and tolerance to selfing in a neotropical tree species. Mol.
Ecol. 14, 1055–1064, doi:10.1111/j.1365-294X.2004.02127.x.
Ledig, F.T., Hodgskiss, P.D., Johnson, D.R., 2005. Genetic diversity, genetic structure,
and mating system of Brewer spruce (Pinaceae), a relict of the Arcto-Tertiary
forest. Am. J. Bot. 92 1975-1968.
Lemes, M.R., Grattapaglia, D., Grogan, J., Proctor, J., Gribel, R., 2007. Flexible mating
system in a logged population of Swietenia macrophylla Kind (Meliaceae):
implication for the management of threatened neotropical tree species. Plant
Ecol. 192, 169–179, doi:10.1007/s11258-007-9322-9.
Lourmas, M., Kjellberg, F., Dessard, H., Joly, H.I., Chevallier, M.-H., 2007. Reduced
density due to logging and its consequences on mating system and pollen flow
in the African mahogany Entandrophragma cylindricum. Heredity 99, 151–160,
doi:10.1038/sj.hdy.6800976.
Marshall, T.C., Slate, J., Kruuk, L.E.B., Pemberton, J.M., 1998. Statistical confidence for
likelihood-based paternity inference in natural populations. Mol. Ecol. 7, 639–
655, doi:10.1046/j.1365-294x.1998.00374.x.
Maues, M.M., 2001. Importance of the floral biology and pollinators on the sustain-
ability of forest management. Acta Hortic. 561, 81–85.
Meagher, T.R., 1986. Analysis of paternity within a natural population of Chamae-
lirium luteum. 1. Identification of most-likely male parents. Am. Nat. 128, 199–
215, doi:10.1086/284554.
Murawski, D.A., Gunatilleke, I.A.U.N., Bawas, K.S., 1994. The effects of selective
logging on inbreeding in Shorea megistophylla (Dipterocarpaceae) from Sri
Lanka. Conserv. Biol. 8, 997–1002, doi:10.1046/j.1523-1739.1994.08040997.x.
Naito, Y., Konuma, A., Iwata, H., Suyama, Y., Seiwa, K., Okudo, T., Lee, S.L., Muham-
mad, N., Tsumura, Y., 2005. Selfing and inbreeding depression in seeds and
seedlings of Neobalanocarpus heimii (Dipterocarpaceae). J. Plant Res. 118, 423–
430, doi:10.1007/s10265-005-0245-z.
Naito, Y., Kanzaki, M., Iwata, H., Obayashi, K., Lee, S.L., Muhammad, N., Okuda, T.,
Tsumura, Y., 2008. Density-dependence selfing and its effects on seed perfor-
mance in a tropical canopy tree species, Shore acuminata (Dipterocarpaceae).
Forest Ecol. Manage. 256, 375–383, doi:10.1016/j.foreco.2008.04.031.
Obayashi, K., Tsumara, Y., Ihara-Unino, T., Niyama, K., Tanouchi, H., Syama, Y.,
Washitani, I., Lee, C., Lee, S.L., Muhammad, N., 2002. Genetic diversity and
outcrossing rate between undisturbed and selective logged forest of Shorea
curtissi using microsatellite DNA analysis. Int. J. Plant. Sci. 163, 151–158,
doi:10.1086/324549.
Odduo-Muratorio, S., Klein, E.K., Austerlitz, F., 2005. Pollen flow in the wild service
tree, Sorbus torminalis (L.) Crantz. II. Pollen dispersal and heterogeneity in
mating success inferred from parent-offspring analysis. Mol. Ecol. 14, 4441–
4452, doi:10.1111/j.1365-294X.2005.02720.x.
Silva, M.B., Kanashiro, K., Ciampi, A.Y., Thompson, I., Sebbenn, A.M., 2008. Genetic
effects of selective logging and pollen gene flow in a low-density population of
the dioecious tropical tree Bagassa guianensis in the Brazilian Amazon. Forest
Ecol. Manage. 255, 1548–1558, doi:10.1016/j.foreco.2007.11.012.
Ritland, K., Jain, S., 1981. A model for the estimation of outcrossing rate and gene
frequencies using independent loci. Heredity 47, 35–52, doi:10.1038/
hdy.1981.57.
Sebbenn, A.M., Degen, B., Azevedo, V.C.R., Silva, M., Lacerda, A.E.B., Ciampi, A.Y.,
Kanashiro, M., Carneiro, F.S., Tompson, I., Loveless, M.D., 2008. Modelling the
long-term impact of selective logging on genetic diversity and demographic
structure of four tropical tree species in the Amazon forest. Forest Ecol. Manage.
254, 335–349, doi:10.1016/j.foreco.2007.08.009.
Smouse, P.E., Sork, V.L., 2004. Measuring pollen flow in forest trees: an exposition of
alternative approaches. Forest Ecol. Manage. 197, 21–38, doi:10.1016/j.foreco.
2004.05.049.
Smouse, P.E., Dyer, R.J., Westfall, R.D., Sork, V.L., 2001. Two-generation analysis of
pollen flow across a landscape. I. Male gamete heterogeneity among females.
Evolution 55, 260–271, doi:10.1111/j.0014-3820.2001.tb01291.x.
Sokal, R.R., Rohlf, F.J., 1981. Biometry. W.H. Freeman and Company, New York.
Vinson, C.C., Amaral, A.C., Sampaio, I., Ciampi, A.Y., 2005. Characterization and
isolation of DNA microsatellite primers for the tropical tree, Symphonia globu-
lifera Linn. Forest Mol. Ecol. Notes 5, 202–204, doi:10.1111/j.1471-8286.
2005.00876.x.
Ward, M., Dick, C.W., Gribel, R., Lowe, A.J., 2005. To self, or not to self. . . A review of
outcrossing and pollen-mediated gene flow in neotropical trees. Heredity 95,
246–254, doi:10.1038/sj.hdy.6800712.
White, G.M., Boshier, D.H., Powell, W., 2002. Increased pollen flow counteracts
fragmentation in tropical dry forest: an example from Swietenia humilis Zuc-
carini. Proc. Natl. Acad. Sci. U.S.A. 99, 2038–2042, doi:10.1073/pnas.042649999.