American Journal of Botany 81(9): 1104-1111. 1994.

GENETIC DIVERSITY AND MATING SYSTEM IN A

TROPICAL TREE, CARAPA GUIANENSIS (MELIACEAE) 1

PAMELA HALL,2 LELIA C. ORRELL, AND KAMAUIT S. BAWA3

Department of Biology, University of Massachusetts, Harbor Campus, Boston, Massachusetts 02125

Carapa guianensis Aublet. (Meliaceae) is an abundant, canopy tree species often exported as fine wood products. The
populations in the Atlantic coastal plain ofCosta Rica have undergone a wide variety ofhuman-induced changes in distribution
and abundance, including isolation due to deforestation. Using protein electrophoretic analysis of seeds, we investigated the
genetic variability and population differentiation of nine populations and the outcrossing rate in two populations that differed
in density by a factor of 50%. Average polymorphism (35%) and heterozygosity (0.12) across populations were relatively
low in comparison with other tropical tree species. Only 4.6% of the total genetic variability could be attributed to population
differentiation even though populations were separated by as much as 70 km. However, genetic differentiation did increase
with distance between pairs of compared populations. There were no apparent differences in genetic variation between
logged and unlogged forests. The outcrossing rate for each population was not significantly different from 1.0, but there was
evidence of nonrandom mating in the lower density population. High levels of gene flow are probably maintained by seed
dispersal as seeds travel well throughout flooded forests and rivers. High population density and synchronous flowering also
contribute to high outcrossing rates. These genetic characteristics suggest that Carapa guianensis is a good candidate for
natural forest management where it occurs at high density. However, current management practices need to account for the
rapid rate of deforestation occurring around managed natural forest sites that will result in the elimination of intervening
populations.

Genetic diversity in tropical lowland rain forest trees
is imperiled due to deforestation and logging operations.
Deforestation may reduce population size or altogether
eliminate local populations. Moreover, forest fragmen-
tation may lead to genetic isolation of once continuous
populations, which, following isolation, may lose genetic
diversity as a result of inbreeding and genetic drift. Log-
ging may lower population density and, to the extent
inbreeding is a function of stand density, lead to an in-
creased level of inbreeding (Knowles et aI., 1987; Muraw-
ski et aI., 1990; Murawski, Dayanandan, and Bawa, 1994).
An assessment of the genetic effects of deforestation
and management practices such as selective logging is
hampered by lack of information about the spatial or-
ganization ofgenetic variability, levels ofgene flow among
populations, and the effects ofdensity on inbreeding. This
is particularly true for species that are intensively logged
or occur in areas characterized by high rates of defores-
tation. One such group is the mahoganies (Meliaceae).
This family is an important component oftropical forests
throughout the world. The several tree species that yield
valuable timbers are often selectively harvested and typ-
ically occur in low densities. Here we report the results
I Manuscript received 14 August 1993; revision accepted 25 March
1994.
The authors thank Portico, SA and Leopoldo Torres for use oftheir
forests and assistance in obtaining collections; Rodolfo Peralta for the
great amount of time he spent collecting and discussing many aspects
of this work with them; Victor Robles, Rosa Sandoval, and Leocardio
Centeno for field assistance; Mike Chase for developing the extraction
buffer and providing advice in the lab; and K. Ritland and M. Slatkin
for supplying software for computations of mating systems and pairwise
comparison of populations. This research was supported by grants from
the National Science Foundation, the Pew Charitable Trust's Pew Schol-
ars Program in Conservation and the Environment, and the Conser-
vation, Food and Health Foundation.
2 Current address: Harvard Institute for International Development,
One Eliot Street, Cambridge, MA 02138.
J Author for correspondence.
1104
of a study aimed at determining genetic diversity and
other population genetic parameters of one member of
the Meliaceae, Carapa guianensis Aublet., a large canopy
tree. The species is found in the Atlantic lowland forests
of Central America, an area that is experiencing one of
the highest rates of deforestation. Specifically, our goals
were: 1) to estimate the amount ofgenetic diversity within
populations; 2) to determine the degree of genetic differ-
entiation and gene flow among populations; and 3) to
measure the amount of outcrossing in stands that differ
in density.
MATERIALS AND METHODS
Species description - Carapa guianensis occurs from
Belize to Amazonian Brazil and also in the Antilles
(McHargue and Hartshorn, 1983a). The trees often reach
over 1 m in diameter and over 40 m in height at maturity
(R. Peralta, San Jose, Costa Rica, personal communica-
tion). The species is occasionally found along streams and
usually found in periodically inundated swamp forests of
the Atlantic coastal plains of Costa Rica (McHargue and
Hartshorn, 1983a). Local abundance can be very high,
representing up to 20% of the stand density of trees over
10 cm in diameter, and more than 15 reproductive in-
dividuals per hectare (R. Peralta, personal communica-
tion). flowering occurs synchronously throughout the At-
lantic coastal plain ofCosta Rica between November and
January, although a few individuals may flower in June
or July. Fruits mature 6 to 8 months later during May to
August (Frankie, Baker, and Opler, 1974). The pollination
mechanism is not known. Fruits are large capsules, 10-
14 em in diameter, with four to eight seeds. Capsules
dehisce when ripe fruits fall to the ground (McHargue and
Hartshorn, 1983a). Fresh seeds usually weigh around 25
to 35 g, but we have found seeds over 100 g. Despite their
weight, the seeds are very buoyant and are probably dis-
persed primarily by water and large rodents. Once seeds
September 1994] HALL ET AL.-GENETIC DIVERSITY AND MATING SYSTEM
have been shed, mortality can be quite high due to con-
sumption by various vertebrates and infestation by larvae
of the moth Hypsipyla ferrealis (McHargue and Hart-
shorn, 1983a). However, germination can still occur if
the embryo has not been damaged by the predators (P.
Hall, personal observation).
Field collection - Two types of seed collections were
made: one for analysis of mating systems and a second
for population differentiation. For the mating system anal-
ysis, 20 or more seeds each were collected from 20 adults
within mapped populations at each of the two sites: Ger-
ardo Fallas (GF) and La Selva (LS). The Gerardo Fallas
site is a logged swamp forest owned and managed for
timber production by the private company Portico, SA.
Logging took place in 1988, and approximately four trees
of Carapa per ha greater than 70 em diameter at breast
height (dbh) were removed. Seed collection occurred with-
in a 5-ha marked plot which, after logging, had 15 trees
per ha of reproductive size (~35 em dbh). The area sur-
rounding the plots is composed of logged forest and pas-
tures. The La Selva site is part ofa biological reserve from
which no cutting has taken place. A permanently marked
demographic plot near the site ofseed collection contained
less than ten trees per ha of reproductive size (Lieberman
and Lieberman, 1987). The collection site is located well
within the boundaries of the 1,51O-ha reserve and over
1 km from the nearest pasture. Seed collections at both
sites were done in the middle of the period of fruit fall,
in July 1991 at Gerardo Fallas and in July 1992 at La
Selva. Both of these sites are at elevations between 10
and 60 m and are 42 km apart.
For the analysis ofpopulation differentiation, seeds were
collected randomly from reproductively mature trees from
nine populations within the Atlantic coastal plain ofCosta
Rica (Table 1; Fig. 1). The sites included an undisturbed
forest within a biological reserve (La Selva), unlogged
forests and remanent forests within a patchwork of pas-
tures (Sardinas and Guyacan), and managed logged forests
(Cafio Chiquero, Ferrat, Gerardo Fallas, Guido Madrigal,
Mata Banano, and Parismina). Geographical distances
between sites ranged from 4 to 70 km, and altitudes ranged
from below 10 m (Cafio Chiquero) to 300 m (Guayacan),
Most sites were at elevations of less than 100 m. An
average of 52 seeds per population with three seeds per
tree was collected from each site.
For all collected seeds, the smooth brown seed coat was
removed with a pocket knife until the embryo was located.
The embryo and some endosperm were immediately re-
moved, labeled, wrapped individually in aluminum foil,
and placed directly in liquid nitrogen for storage and ship-
ment.
Electrophoresis-Embryos were stored at our labora-
tory in Boston at - 80 C until enzyme extraction. Embryos
were ground in liquid nitrogen and a buffer solution, elec-
trophoresed, and stained following the procedures of
O'Malley, Wheeler, and Guries (1980). The extraction
buffer consists ofthe following components: 1 mM EDTA,
10 mM KCl, 20 mM MgCI2 , 17 mM sodium metabisulfite,
0.03 mM NADP, 0.03 mM NAD, 3 mM Dithiothreitol,
4% PVP-40, 10% Glycerol, and 0.1 M Na-phosphate.
ll05
Data analysis-Allozymes most anodal were designat-
ed allele 1 with increasing numbers used for increasing
cathodal allozymes. Percent polymorphism (P), hetero-
zygosity (H), and the effective number ofalleles (Ae ) were
calculated (Hartl and Clark, 1989). The expected hetero-
zygosity for populations in Hardy-Weinberg equilibrium
was computed for each locus and for each population as:
He = 1 - 2; Pi2 •
where Pi is the allele frequency. Deviations ofthe observed
distribution of genotypes from the expected distribution
under the assumptions of Hardy-Weinberg equilibrium
were tested using a chi-square statistic, and deviations
were expressed as the fixation index:
F = 1 - (Ho/H e)
where H, is the observed number ofheterozygotes (Wright,
1969). The effective allele number, orthe number ofalleles
needed within a locus to maintain the current level of
heterozygosity, was determined as:
Ae = 1/2; Pi2 •
Population differentiation was investigated using
Wright's (1978) F-statistics and Nei's (1973) genetic di-
versity statistics. F-statistics are related by:
(1 - F is )(1 - F st) = (1 - Fit)
where Fit is the total deviation from expected frequencies
under Hardy Weinberg equilibrium, F st is the deviation
due to population subdivision, and Fi; is the deviation
within population subdivisions. An estimate ofgene flow
among populations (Nm) can be computed from the mul-
tilocus value of F st as:
F st = 1/(4Nma + 1)
where F st is a multilocus value of F st , a = [n/(n - 1)]2
and n is the number ofpopulations (Nei, 1973). Nei (1973)
referred to this statistic as Gst ; however, we shall use F st
for both single and multilocus computations.
Analysis of mating systems was done using the method
of Ritland and Jain (1981) and software provided by
Ritland (Ritland, 1990). Multilocus outcrossing rates were
computed for populations at Gerardo Fallas and La Selva
in order to assess the effect ofdifferences in stand density
on mating patterns. For both the Gerardo Fallas and La
Selva populations, only trees with at least 20 seeds and
for which all polymorphic loci had been scored were used
for outcrossing computations. In order to have a balanced
data set for computation of outcrossing rates, families of
five trees at Gerardo Fallas for which a large number of
seeds had been collected were artificially rarefied by ran-
domly selecting 30 seeds from each family. The mean
number of seeds per family was 24 at La Selva and 26 at
Gerardo Fallas. Bootstrapping 100 times was used to make
estimations ofthe standard errors for all outcrossing rates
(Ritland and Jain, 1981). Pollen and ovule allele fre-
quencies were not constrained to being equal for com-
putation ofoutcrossing rates. However, when outcrossing
rates for individual trees were computed, pollen and ovule
frequencies were set equal. The relatively small size of
sampled families inhibits accurate computations of out-
crossing rates when both pollen frequencies and outcross-
ing rates are estimated for each family. Maternal genotype
1106 AMERICAN JOURNAL OF BOTANY [Vol. 81

TABLE I. Allelic frequencies and summary statistics for genetic variation for Carapa guianensis in Costa Rica. Population labels follow Fig. I.
Allele frequencies are given for all polymorphic loci.'
Populations
CH AF GF GM GY l.S MB PA SA
Forest 18 18 18 ul ul rs Ig Ig ul Mean
ADH I 0.00 0.00 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00
2 0.09 0.11 0.09 0.13 0.23 0.11 0.05 0.08 0.11 0.11
3 0.91 0.89 0.91 0.87 0.76 0.89 0.95 0.92 0.89 0.89
MDH I 0.85 0.93 0.80 0.81 0.71 0.90 0.96 0.85 0.91 0.86
2 0.02 0.02 0.02 0.07 0.00 0.04 0.00 0.03 0.03 0.02
3 0.13 0.05 0.18 0.12 0.29 0.06 0.04 0.12 0.06 0.12
PGI I 0.84 0.81 0.79 0.69 0.91 0.44 0.77 0.83 0.64 0.75
2 0.16 0.19 0.21 0.31 0.09 0.56 0.23 0.17 0.36 0.25
PGM I 0.23 0.28 0.33 0.20 0.22 0.39 0.23 0.09 0.27 0.25
2 0.66 0.71 0.66 0.80 0.78 0.59 0.72 0.88 0.72 0.73
3 0.11 0.01 0.01 0.00 0.00 0.02 0.04 0.03 0.01 0.02
UGPP-I I 0.02 0.01 0.01 0.00 0.00 0.00 0.05 0.04 0.01 0.01
2 0.98 0.99 0.99 1.00 1.00 1.00 0.89 0.96 0.99 0.98
3 0.00 0.00 0.00 0.00 0.00 0.00 0.06 0.00 0.00 0.01
UGPP-2 I 0.18 0.27 0.10 0.24 0.44 0.31 0.28 0.11 0.21 0.24
2 0.14 0.25 0.25 0.17 0.11 0.10 0.21 0.29 0.20 0.19
3 0.61 0.31 0.49 0.52 0.31 0.47 0.43 0.46 0.51 0.46
4 0.07 0.17 0.16 0.07 0.14 0.12 0.08 0.14 0.08 0.11
N(seeds) 52 52 52 52 52 54 52 52 50 468
N(trees) 25 31 31 26 24 20 26 29 19 231
X(sds/tr) 2.1 I.7 I.7 2.0 2.2 2.7 2.0 1.8 2.6 2.1
H(all), 0.11 0.11 0.12 0.12 0.12 0.13 0.12 0.11 0.13 0.12
H, 0.30 0.30 0.33 0.32 0.33 0.34 0.30 0.28 0.32 0.31
P 0.38 0.38 0.38 0.31 0.33 0.31 0.35 0.38 0.38 0.35
Ae 1.14 1.21 1.19 1.17 1.19 1.21 1.18 1.15 1.18 1.18
Test of Hardy-Weinberg Equilibrium
ADH 0.151 0.089 -0.088 0.179 0.192 -0.127 -0.055 0.190 -0.122
MDH 0.060 -0.063 -0.170 0.016 0.249 0.309- -0.055 -0.004 -0.078
PGI -0.055 -0.116 -0.018 0.188 -0.088 -0.167 -0.192 -0.077 0.132
PGM 0.109 -0.065 0.066 -0.255 0.049 0.068 0.026- 0.168 -0.320
UGPP-I -0.027 0.050 0.050 0.000 0.000 0.000 0.611 -0.055 0.000
UGPP-2 -0.005 0.072 0.071 -0.023 -0.102- 0.017 -0.124 -0.088 0.014
a In addition there were ten loci that were monomorphic in all populations. Column labeled Mean gives mean values for populations. Values

given are: Forest, Ig(logged), ul (unlogged forest in patchwork oflogged forest and pasture), rs (large reserve, undisturbed forest); N(seeds), number
of seeds analyzed; N(trees), number of trees sampled; X(sds/tr), mean number of seeds analyzed per tree; Htallj., mean heterozygosity for all loci
(ten monomorphic, six polymorphic); H" mean heterozygosity for polymorphic loci; P, percent polymorphism; and A., the effective number of
alleles calculated as a mean of the effective number of alleles for each locus. F = I - (Hob/Hu p ) is given for each locus, and the significance of
chi-square tests of Hardy Weinberg Equilibrium (HWE) are indicated by - = p < 0.05 or lower; no star = P > 0.05.

was estimated from the genotypic array ofseedlings (Brown
and Allard, 1970). Differences between pollen and ovule
frequencies are tested by computing a chi-square distrib-
uted statistic:
D = NFst(A - 1)
which has A-I degrees of freedom. N is the sum of the
number of pollen and ovule gametes, A is the number of
alleles for the locus, and F st is the genetic difference be-
tween populations-treating pollen and ovules as separate
populations (Murawski and Hamrick, 1992a).
RESULTS
Genetic interpretations-An enzyme survey was con-
ducted to identify polymorphic, Mendelian segregating
markers. We initially screened for 30 enzyme systems
with three electrophoretic buffers. Eleven systems were
variable but, based on the tertiary structure of the en-
zymes, displayed complicated, uninterpretable patterns
of segregation mostly due to gene duplication. An addi-
tional 11 systems either did not have consistent staining
or failed to resolve any presence ofisozymes. Ultimately,
eight systems were found to be reliable for genetic inter-
pretation: malate dehydrogenase (MDH) exhibited five
monomorphic bands, aconitate hydratase (ACO) and
{j-fluorescent esterase (FLE) each exhibited two mono-
morphic bands, and glutamate-pyruvate transaminase
(GPT) yielded a single monomorphic band, which we
classified as a total often monomorphic loci. Five systems
exhibited polymorphic banding patterns that conformed
to Mendelian expectations for segregation: alcohol de-
hydrogenase (ADH), phosphoglucoisomerase (PGI),
phosphoglucomutase (PGM), and an additional locus of
MDH each exhibited one polymorphic banding zone, and
uridine diphosphoglucose pyrophosphorylase (UGPP)
displayed two variable banding regions, designated as
UGPP-l and UGPP-2. We observed three alleles for ADH,
MDH, and PGM, two for PGI and UGPP-l, and four
for UGPP-2. These enzyme systems all exhibited banding
patterns that were consistent with subunit structure as
previously reported (Weeden and Wendel, 1989). Ge-
September 1994] HALL ET AL.-GENETIC DIVERSITY AND MATING SYSTEM 1107

TABLE 2. Heterozygosities (total [HI]' population [H,], and individual

[H,], Wright's F statistics, Nei's D", F", and Nm for each locus,
Caribbean Sea and all loci combined for seeds of Carapa guianensis populations
11 from the Atlantic coastal plain of Costa Rica for six polymorphic
loci.'
Sardina ___
Population differentiation statistics
Guido Madrigal ________
Laselva'/ I
Gerado Falias
Statistic

H,
ADH

0.200
MDH

0.250
PGI

0.378
PGM

0.411
UGPp.(

0.044
UGPP-2

0.685
All

0.328
Parismina H, 0.180 0.227 0.355 0.403 0.028 0.674 0.311
10
• • H, 0.195 0.239 0.343 0.397 0.042 0.662 0.313
San Jose • SEH, 0.093
Guyacan 0.005 0.011 0.035 0.014 0.002 0.024 0.D15
D"
SED" 0.009
F" 0.024 0.046 0.092 0.034 0.040 0.034 0.046
Nm 8.03 4.10 1.95 5.61 4.74 5.61 4.10
a Standard errors (SE) for H, and D" for multiallelic analysis are also
Pacific Ocean 9
given. Columns give values for each locus and All refers to multilocus
values.
N

50 km
(Table l) was highest at La Selva (H, = 0.339), the pop-

86 W
r 85 84
Fig. I. Location of populations of a Carapa guianensis collected on
the Atlantic coastal plain of Costa Rica.
notype frequencies were consistent with expectations of
Hardy-Weinberg equilibrium for all but three of the 54
combinations ofloci and populations (Table I). In sum-
mary, we used six polymorphic loci and ten monomorphic
loci for calculation ofgenetic variation, and restricted our
analysis of population differentiation and mating systems
to the six polymorphic loci.
Genetic diversity- Polymorphism (P) ranged from 0.31
to 0.38 among populations with a mean of 0.35 (Table
I). Overall heterozygosity for Carapa guianensis was 0.12
and varied little among populations, ranging from 0.11
to only 0.13. Most loci (Table I) had a very common
allele which was the most common in all populations.
There was one exception. For the PGI locus, allele 2 had
the highest frequency in La Selva, while allele I had the
highest frequency in all other populations. Of the four
alleles found for UGPP-2, allele 3 tended to be the most
common in each ofthe populations. A rare allele for ADH,
MDH, PGM, UGPP- Nvas sporadically distributed among
populations, frequently absent, or only found in a single
seed. Despite all but one polymorphic system being rep-
resented by three or more alleles, the effective number of
alleles was not much greater than I in all populations.
Population differentiation -Only 4.6% ofthe total mea-
sured genetic variability was due to differences among
populations (Fst = 0.046, Table 2). The amount of the
total genetic variation that can be ascribed to population
differentiation varied among loci, with 9.2% for PGI and
only 2.4% for ADH. Nm, a measure of gene flow among
populations, was much greater than I (Table 2) for all
loci with a multilocus value of 4.10. The most distant
populations, La Selva and Guayacan, were also the most
genetically distinct. Heterozygosity for polymorphic loci
83
8
ulation located within a large reserve, and lowest at Par-
ismina (H, = 0.276). The population at Parismina was
more heavily logged than the other logged sites, is the
most isolated from other forest fragments, and is sur-
rounded by pastures and banana plantations.
The relationship between genetic differentiation and
geographic distribution was investigated by regressing the
log ofNm based on multilocus F st computed for all pairs
of populations on the log of distance between pairs of
populations (Fig. 2) (Slatkin and Barton, 1989; Slatkin,
1993). Even though there was relatively little population
differentiation, there was a negative correlation between
Nm calculated using F st and distance between pairs of
populations (approximation of relationship using Spear-
man rho = -0.76, P < 0.001; Dietz, 1983).
Mating systems- The locus UGPP-I was fixed in the
La Selva population and nearly so in Gerardo Fallas and
was dropped from the mating system analysis leaving five
polymorphic loci for estimation of outcrossing rates. The
multilocus outcrossing rate for the population at Gerardo
Fallas was 0.986 and was not significantly different from
1.0 (Table 3). At La Selva, the multilocus outcrossing rate
was 0.967 and was greater than I standard error from a
value of 1.0. However, the outcrossing rates at Gerardo
Fallas and La Selva were not statistically different from
each other, overlapping within a single standard error.
The mean of single locus estimates of outcrossing was
lower than the multilocus estimate at La Selva, but not
at Gerardo Fallas. In addition, there were no differences
in the genotypes of ovules and the pool of pollen parents
at Gerardo Fallas (Table 4). However, at La Selva, three
out of the five loci showed significantly different allele
frequencies between pollen and ovules. These results are
indicative of nonrandom mating patterns within the pop-
ulation at La Selva.
Outcrossing rates for individual trees varied from 0.32
and 0.34 to over 1.0 at La Selva and Gerardo Fallas,
respectively. At each site, four trees had outcrossing rates
below 0.9, but the individual outcrossing rate was sig-
nificantly below 1.0 for only a single tree at each site.
Detecting deviations from complete outcrossing is diffi-
1108 AMERICAN JOURNAL OF BOTANY [Vol. 81

Relationship between Geographic Distance and Nm for TABLE 4. Estimated ovule and pollen frequencies for two populations
Populations of Carapa guianensis of Carapa guianensis. Standard errors for pollen frequencies are
given in parentheses. Alleles I and 4 are combined for UGPP-2.
100-,....----------------------,
Values for UGPP-I are not given because it was not used in the
mating systems analysis due to rarity of second allele. Test statistic
for difference between ovule and pollen frequency for each locus,
degrees of freedom, and significance level is provided. a
~
0
•
.",
• • Allele Ovules Pollen Statistic- df
;;'"
c,
0
0..
'0
.~ 10
•
•
••
•
• •
• ••
• .• -~
•
ADH I
ADH 2
0.000
0.075
La Selva
0.00 I (0.000)
0.112 (0.015)
3.971"' 2
'"
0.. • •• ADH3 0.925 0.888 (0.015)
.E • ••• • e•
• MDH I 0.875 0.871 (0.016) 13.759*** 2
za • MDH2 0.100 0.021 (0.007)
eo
0
..J •• MDH3 0.025 0.108 (0.014)
PGM I 0.525 0.280 (0.021) 54.202*** 2
PGM 2 0.475 0.700 (0.022)
PGM 3 0.000 0.019 (0.006)
1+-----------r------------.-f
100
POI I 0.500 0.460 (0.028) 0.771"'
1 10 POI 2 0.500 0.540 (0.028)
Log Distance between Pairs of Populations (km)
UGPP-2 I 0.375 0.330 (0.026) 8.318* 2
Fig. 2. Log Nm and log distance between pairs of populations for UGPP-22 0.075 0.191 (0.018)
populations of Carapa guianensis throughout the Atlantic coastal plain UGPP-23 0.550 0.479 (0.027)
of Costa Rica. Nm was computed from multilocus F at for each pair of Gerardo Fallas
populations. Spearman rho and significance of the correlation of log-
transformed N m on log-transformed distance is provided for each group. ADH I 0.000 0.000 0.106"'
Significance indicated by: .*, P < 0.01. ADH 2 0.100 0.109 (0.012)
ADH 3 0.900 0.891 (0.012)
MDHI 0.800 0.876 (0.017) 4.746"' 2
MDH2 0.000 0.010 (0.004)
cult due to the small number ofseeds sampled per family MDH3 0.200 0.114 (0.016)
which results in large standard errors. PGM I 0.300 0.311 (0.022) 0.069"'
PGM2 0.700 0.689 (0.022)
PGM 3 0.000 0.000
DISCUSSION POI I 0.800 0.822 (0.017) 0.380"'
POI 2 0.200 0.178 (0.017)
Although dominance by any given species is uncom- UGPP-2 I 0.200 0.268 (0.020) 4.571"' 2
mon in tropical forests, Carapa guianensis can dominate UGPP-22 0.225 0.146 (0.018)
forests in which it is found. Within some ofthe stands of UGPP-23 0.575 0.587 (0.027)
logged forest, C. guianensis may represent over 20% of a Significance levels as: * P < 0.05, ** P < 0.01, *** P < 0.001, ns P
the stand density and over 80% of the commercial stand- > 0.05. See text for explanation of test statistic.
ing timber volume (R. Peralta, San Jose, Costa Rica,
personal communication). The high density and high
commercial value of Carapa are reminiscent of some
temperate conifers. However, the genetic diversity of this rick and Loveless, 1989; Hamrick and Godt, 1990). Even
species as measured by percent polymorphism (35%) or in comparison with tropical tree species, the total genetic
heterozygosity (0.12) was much lower than mean values diversity of Carapa guianensis is low (H = 0.211, P =
reported for conifers (70.9% and 0.21, respectively; Ham- 60.9), but well within the range of values reported by
Hamrick and Loveless (1989). However, in comparison
with species that also occur throughout the Atlantic coast-
al plain of Costa Rica, the genetic variability of C. gui-
TABLE 3. Outcrossing rates for multilocus estimates (tm) and single anensis is similar to that of less abundant canopy tree
locus estimates (ts), with standard errors in parentheses, are given
for two populations of Carapa guianensis. All families had at least species Pithecellobium pedicel/are and Cordia alliodora,
20 seeds and none had more than 30. Sample sizes are: 481 seeds but much higher than the very abundant Pentaclethra
analyzed for La Selva (mean 24 seeds/tree) and 523 seeds analyzed macroloba (Chase, Boshier, and Bawa, in press; Hall,
for Gerardo Fallas (mean 26 seeds/tree). Chase, and Bawa, 1994; P. Hall, unpublished data). Our
results are not consistent with Hamrick and Murawski's
La Selva Gerardo FalJas
(1991) findings that genetic variation was lower in un-
MuItilocus (trn) 0.967 (0.022) 0.986 (0.028) common than in common tree species.
Average of single locus (ts) 0.932 (0.024) 1.000 (0.030) However, strict interpretation of differences in genetic
Individual loci variability among species is fraught with problems. Our
ADH 1.121 (0.301) 0.860 (0.097) values of polymorphism and heterozygosity are based on
MDH 0.842 (0.069) 0.976 (0.069) a relatively low number ofloci investigated in comparison
POI 1.038 (0.059) 1.121 (0.071) to studies ofa wide variety of species as cited in Hamrick
PGM 0.906 (0.055) 0.974 (0.062) and Godt (1990). Furthermore, our values are biased
UGPP-2 0.898 (0.060) 1.018 (0.038)
downward because of exclusion of loci that are clearly
September 1994] HALL ET AL.-GENETIC DIVERSITY AND MATING SYSTEM
polymorphic but have complex segregating banding pat-
terns, probably due to gene duplication. In addition, stat-
ing the number ofloci present in multibanded but mono-
morphic enzyme systems is largely based on individual
experience with scoring protein electrophoretic gels. With
respect to MDH, we have used the most parsimonious
approach in considering the total number of monomor-
phic loci by viewing each fixed band as a separate locus.
The number of monomorphic loci attributed to such iso-
zyme systems will strongly affect computations of relative
genetic variability such as percent polymorphism and het-
erozygosity. With these caveats in mind, the genetic vari-
ability of Carapa guianensis does seem to be rather low
for a highly outcrossed, long-lived, widespread, tropical,
animal-pollinated tree with wide seed dispersal (Hamrick
and Godt, 1990).
One explanation for this may be that these populations
are near the northern limit of the species distribution.
Populations that occur at the edges of their range are at
the outer limits ofgene flow, whereas central populations
can experience gene flow from all directions. Populations
at distributional limits may also exhibit a fewer number
of alleles due to a recent bottleneck in population size
during colonization of a new region (Ayala, 1982; Barrett
and Shore, 1989; Leberg, 1992; Hall, Chase, and Bawa,
1994), or by natural selection for extreme genotypes at
marginal areas.
Population differentiation - The overall degree of pop-
ulation differentiation was small (Fst = 0.046) for Carapa
guianensis. Such low values have also been reported for
other tropical trees (Hamrick and Loveless, 1989). Pop-
ulation differentiation is also low for many conifer species
(Hamrick, Linhart, and Mitton, 1979; Hamrick, 1983;
Govindaraju, 1988). However, the distances among tree
populations that have been examined are highly variable,
with most conifers often investigated throughout their
natural range, while tropical trees have usually been in-
vestigated over a small portion of their ranges. Penta-
clethra macroloba, studied at the same geographic scales
(approximately 70 km maximum distance between two
populations), provides a better comparison. In this spe-
cies, 21.9% of the genetic variation is due to differences
among populations (Hall, Chase, and Bawa, 1994). Al-
though both C. guianensis and P. macroloba are abundant
species, the spatial organization of genetic variation in
the two species is very different.
The low levels of population differentiation are cor-
roborated by high levels of gene migration among pop-
ulations (Nm = 4.10, Table 2). Geographic distance and
Nm, calculated for pairs of populations, are inversely
related. This indicates that the amount of genetic differ-
entiation that exists between populations increases with
geographic distance, so that in spite of the apparent high
migration rate, populations are not completely panmictic.
It is unlikely that this differentiation is due to selection
for different habitats, as most of the population samples
come from similar environmental conditions in seasonal
inundated swamplands located within the same geograph-
ic region (Vasquez, 1983). The one exception, the Guaya-
can population, was found at a higher elevation (300 m)
than C. guianensis is commonly found, but located within
a narrow river valley clustered along the water's edge.
1109
However, the La Selva and Guayacan populations both
exhibited differences in allele frequencies in comparison
to the other populations as indicated by PGI for La Selva
and UGPP-2 for Guayacan. Interestingly, these two pop-
ulations are the most distant from the group of popula-
tions that constituted the core for this study.
It is probable that a high level of gene flow indicated
by the low degree of population differentiation has oc-
curred historically and may be currently maintained by
pollen and seed dispersal. While we have no information
about the pollination mechanisms of the species, it is
known that the seeds are very buoyant and can survive
in water for substantial periods of time and germinate
even when partially submerged (McHargue and Hart-
shorn, 1983b). As was stated earlier, C. guianensis grows
primarily in periodically inundated tropical lowland
swamps interconnected by watercourses. The movement
and change ofthe river courses over time, which has been
found to occur very rapidly in the Peruvian Amazon,
would aid in the migration of floating seeds to adjacent
and distant populations (Rasanen et aI., 1992).
Our data suggest that isolation of populations by de-
forestation may lead to the loss of genetic diversity. The
La Selva population, the only sample collected within a
large reserve, had the highest heterozygosity of all the
populations. In contrast, the population at Parismina had
the lowest heterozygosity. This population was heavily
logged before Portico controlled the management of this
site. Moreover, this area is now a forest fragment sur-
rounded by pastures and banana plantations. Logging, by
reducing density, may decrease the number of fathers
contributing pollen to each tree, and fragmentation may
reduce long distance gene flow.
Mating systems- The majority of rain forest species
investigated so far appears to have mating systems that
strongly enhance outcrossing (Bawa, Perry, and Beach,
1985). Allozyme analysis has further supported the prev-
alence of outcrossing among neotropical tree species
(O'Malley and Bawa, 1987; O'Malley et aI., 1988; Muraw-
ski et aI., 1990; Murawski and Hamrick, 1991, 1992b).
The multilocus outcrossing rate for Carapa guianensis
was indistinguishable from 100% in both populations we
analyzed (Table 3).
The local abundance of a species and the synchrony
with which a population flowers may affect the rate of
effective outcrossing even in self-incompatible species
(O'Malley and Bawa, 1987; Murawski et aI., 1990; Mu-
rawski and Hamrick, 1991). For the studied populations
of C. guianensis flowering occurred synchronously for the
seed collections, but stand densities varied between the
two sites by a factor of 1.5. The rate of outcrossing was
less in the population with lower density, although the
difference in outcrossing rates between the two popula-
tions was small. At La Selva, there was also an indication
of nonrandom mating as shown by statistically significant
differences in the pollen and ovule allele frequencies for
some loci and lower average single locus outcrossing rate
(Ritland and Jain, 1981). Population density at La Selva
was less than ten trees per ha, but higher than many other
tropical tree species both at La Selva and other sites. Yet
even at this high level oflocal abundance, panmixis does
1110 AMERICAN JOURNAL OF BOTANY
not occur and effective neighborhood sizes vary among
individual trees.
The effects of stand density on outcrossing rates have
been documented for several species (Murawski et al.,
1990; Murawski and Hamrick, 1991, 1992b). In Shorea
congestifolia, the logged populations have a lower rate of
outcrossing than populations in unlogged forests, although
it is not known if the reduction in outcrossing is due to
selective harvests of trees (Murawski, Dayanandan, and
Bawa, 1994). In Carapa guianensis, the reduction in den-
sity as a result of logging could influence the rate of out-
crossing. However, our logged populations had a higher
density of trees than the unlogged populations.
Conservation and management implications - Natural
forest management is being practiced by Portico, SA at
seven ofthe nine sites ofour study. These practices include
extracting only trees greater than 70 cm in diameter, min-
imizing damage to the remaining stand, and monitoring
growth rates to determine rotation cycles (R. Peralta, San
Jose, Costa Rica, personal communication). Because Ca-
rapa guianensis attains reproductive size between 35 and
40 em in diameter, many reproductive trees are left stand-
ing in the logged forests. All of the Portico sites from
where collections were made have C. guianensis densities
equal to or exceeding that of the La Selva reserve because
Portico currently manages only forests that have a high
density of Carapa guianensis. The results of our genetic
analysis indicate that there is little population differen-
tiation, which is probably a reflection of historically per-
vasive gene flow. Furthermore, the outcrossing rate of the
species, even at densities lower than managed forests, is
not significantly different from 100%. Therefore, there is
little threat to increasing rates of inbreeding at these cur-
rent population densities. Ifthe current population struc-
ture and larger landscape pattern ofdistribution and abun-
dance do not change, then the current management
practices of Portico do not seem to threaten the genetic
diversity of the species and will probably maintain the
current level of genetic diversity.
However, forests surrounding the Portico sites and the
La Selva Biological Reserve continue to be rapidly con-
verted to agricultural use or pasture (Sader and Joyce,
1988). Recently, 400 ha offorest were converted to banana
plantation in the Sarapiqui River Valley, the heart of the
remaining populations of Carapa guianensis (Fernandez,
1992). The Atlantic coastal plain was once continuous
forest, but at the present time it is a patchwork ofremanent
forest with two reserves accounting for only a small por-
tion of the once extensive lowland forest. Our current
estimate of low population differentiation reflects histor-
ical patterns of gene flow as only the closest populations
that are now interconnected by waterways may be un-
dergoing current gene exchange. Further deforestation will
no doubt isolate remaining populations, lower current
effective population densities, and destroy the avenues
for seed dispersal, thereby reducing migration and ulti-
mately leading to lower genetic diversity within popula-
tions. The future genetic health of Carapa guianensis in
Costa Rica, as with many other tree species ofthe lowland
Atlantic forest, will depend upon the exact pattern of this
spiraling loss of populations and habitats.
[Vol. 81
LITERATURE CITED
AYALA, F. J. 1982. Population and evolutionary genetics. Benjamin!
Cummings, Menlow Park, CA.
BARREIT, S. C. H., AND T. S. SHORE. 1989. Isozyme variation in
colonizing plants. In D. E. Soltis and P. S. Soltis [eds.), Isozymes
in plant biology, 106-126. Dioscorides Press, Portland, OR.
BAWA, K. S., D. R. PERRY, AND J. H. BEACH. 1985. Reproductive
biology of tropical lowland rain forest trees. I. Sexual systems and
incompatibility mechanisms. American Journal ofBotany 72: 331-
345.
BROWN, A. H. D., AND R. W. ALlARD. 1970. Estimation of the mating
system in open-pollination maize populations using isozyme poly-
morphisms. Genetics 66:133-145.
CHASE, M. R., D. BosHIER, AND K. S. BAWA. In press. Conservation
implications of population genetic structure of Cordia alliodora, a
neotropical forest tree species. American Journal of Botany.
DIETZ, E. J. 1983. Permutation tests for association between two dis-
tance matrices. Systematic Zoology 32: 21-26.
FERNANDEZ, L. 1992. Fuerte aumento en importaciones. San Jose,
Costa Rica: La Nacion July 10: Sect. A, 1-4.
FRANKIE, G. W., H. G. BAKER, AND P. A. OPLER. 1974. Comparative
phenological studies of trees in tropical wet and dry forests in the
lowlands of Costa Rica. Journal ofEcology 62: 881-919.
GOVINDARAJU, D. R. 1988. Relationship between dispersal ability and
levels of gene flow in plants. Oikos 52: 31-35.
HALL, P., M. R. CHASE, AND K. S. BAWA. 1994. Low generic variation
but high population differentiation in a common tropical forest tree
species. Conservation Biology 8: 471-482.
HAMRICK, J. L. 1983. The distribution ofgenetic variation within and
among natural plant populations. In C. M. Schoewald-Cox, S. M.
Chamber, B. MacBryde, and W. L. Thomas [eds.), Genetics and
conservation, 501-508. Benjamin/Cummings, Menlo Park, CA.
- - , AND M. J. GoDT. 1990. Allozyme diversity in plant species.
In A. H. D. Brown, M. T. Clegg, A. L. Kahler, and B. S. Weir [eds.),
Population genetics and germ plasm resources in crop improvement,
43-46. Sinauer, Sunderland, MA.
- - , Y. B. LINHART, ANDJ. B. MITTON. 1979. Relationships between
life history characteristics and electrophoretically detectable genetic
variation in plants. Annual Review ofEcology and Systematics 10:
173-200.
- - , AND M. D. LoVELESS. 1989. The genetic structure of tropical
tree populations: associations with reproductive biology. In J. H.
Bock and Y. B. Linhart [eds.), The evolutionary ecology of plants,
129-146. Westview Press, Boulder, CO.
- - , AND D. A. MURAWSKI. 1991. Levels ofallozyme diversity in
populations ofuncommon neotropical tree species. Journal ofTrop-
ical Ecology 7: 395-399.
HARTL, D. L., AND A. G. CLARK. 1989. Principles of population ge-
netics. Sinauer, Sunderland, MA.
KNOWLES, P., G. R. FURNIER, M. A. ALEKSIUK, AND D. J. PERRY. 1987.
Significant levels of self-fertilization in natural populations of tam-
arack. Canadian Journal of Botany 65: 1087-1091.
LEBERG, P. L. 1992. Effects of population bottlenecks on genetic di-
versity as measured by allozyme electrophoresis. Evolution 46: 477-
494.
LIEBERMAN, D., AND M. LIEBERMAN. 1987. Forest tree growth and
dynamics at La Selva, Costa Rica (1969-1982). Journal ofTropical
Ecology 3: 347-358.
McHARGUE, L. A., AND G. S. HARTSHORN. 1983a. Carapa guianensis.
In D. H. Janzen [ed.), Costa Rican natural history, 206-207. Uni-
versity of Chicago Press, Chicago, IL.
- - , AND - - . 1983b. Seed and seedling ecology of Carapa
guianensis. Turrialba 33: 399-404.
MURAWSKI, D. A., B. DAYANANDAN, AND K. S. BAWA. 1994. Out-
crossing rates of two endemic Shorea sp. from Sri Lankan tropical
rain forests. Biotropica 26: 23-29.
--,ANDJ. L. HAMRICK. 1991. Theeffectofthedensityofflowering
individuals on the mating systems of nine tropical tree species.
Heredity 67: 167-174.
- - , AND--. 1992a. The mating system and phenology ofCeiba
pentandra (Bombacaceae) in Central Panama. Heredity 83: 401-
403.
September 1994] HALL ET AL.-GENETIC DIVERSITY AND MATING SYSTEM
- - - , AND - - - . 1992b. The mating system of Cavanillesia pla-
tanifolia under extremes of flowering-tree density: a test of predic-
tions. Biotropica 24: 99-101.
- - - , - - - , S. P. HUBBELL, AND R. B. FOSTER. 1990. Mating sysems
of two Bombacaceous trees ofa neotropical moist forest. Oecologia
82:501-506.
NEI, M. 1973. Analysis of gene diversity in subdivided populations.
Proceedings of the National Academy of Sciences. USA 70: 3321-
3323.
O'MALLEY, D. M., AND K. S. BAWA. 1987. Mating systems ofa tropical
rain forest tree species. American Journal ofBotany 74: 1143-1149.
- - , D. P. BuCKLEY, G. T. PRANCE, AND K. S. BAWA. 1988. Genetics
of Brazil nut (Bertholetia excelsa Humb. & Bonpl.: Lecythidaceae).
2. Mating system. Theoretical and Applied Genetics 76: 929-932.
- - - , N. C. WHEELER, AND R. P. GURIES. 1980. A manual for starch
gel electrophoresis. University of Wisconsin Press, Madison, WI.
1UsANEN, M., R. NELLER., J. SALO, AND H. JUNGNER. 1992. Recent
and ancient fluvial deposition systems in the Amazonian foreland
basin, Peru. Geological Magazine 129: 293-306.
RITLAND, K. 1990. A series of FORTRAN computer programs for
estimating plant mating systems. Journal ofHeredity 81: 235-237.
1111
- - - , AND S. K. JAIN. 1981. A model for the estimation ofoutcrossing
rate and gene frequencies based on n independent loci. Heredity
47: 35-52.
SADER, S. A., AND A. T. JOYCE. 1988. Deforestation rates and trends
in Costa Rica, 1940 to 1983. Biotropica 20: 11-19.
SLATKIN, M. 1993. Isolation by distance in equilibrium and non-
equilibrium populations. Evolution 47: 264-279.
- - - , AND N. H. BARTON. 1989. A comparison of three indirect
methods for estimating average levels of gene flow. Evolution 43:
1349-1368.
VAsQUEZ M., A. 1983. Soils. In D. H. Janzen [ed.},Costa Rican natural
history, 206-207. University of Chicago Press, Chicago, IL.
WEEDEN, N. F., AND J. F. WENDEL. 1989. Genetics of plant isozymes.
In D. E. Soltis and P. S. Soltis [eds.}, Isozymes in plant biology,
46-72. Dioscorides Press, Portland, OR.
WRIGHT, S. 1969. Evolution and the genetics of populations, vol. 2.
The theory of gene frequencies. University of Chicago Press, Chi-
cago,IL.
- - - . 1978. Evolution and the genetics of populations, vol. 4. Vari-
ability within and among natural populations. University ofChicago
Press, Chicago, IL.