Kenfack 2011 PDF

Botanical Journal of the Linnean Society, 2011, 165, 186221.
With 17 figures
Resurrection in Carapa (Meliaceae): a reassessment of

morphological variation and species boundaries using
multivariate methods in a phylogenetic context
boj_1104
186..222
DAVID KENFACK*
Center for Tropical Forest Science, Arnold Arboretum, Harvard University, 22 Divinity Avenue,
Cambridge, Massachusetts, 02138, USA
Received 28 January 2010; revised 7 October 2010; accepted for publication 22 October 2010
The taxonomy of the amphi-Atlantic tree genus Carapa (Meliaceae) has long been controversial. Of the three
species currently recognized in the genus, two are known to present substantial morphological variation that has
been used in the past to distinguish several taxa, most of which are currently placed in synonymy. Here, a
combination of field observations, univariate analyses of leaf, floral and seed characters and principal coordinate
analyses of floral characters in the context of a molecular phylogenetic analysis was used to investigate the
patterns of variation and delimit morphological species anew in the genus. These results support the recognition
of 27 species in Carapa, of which 16 are previously described and 11 are new. In general, phylogenetically related
species occurred in the same geographical area, but were morphologically distinct. 2011 The Linnean Society
of London, Botanical Journal of the Linnean Society, 2011, 165, 186221.
ADDITIONAL KEYWORDS: amphi-Atlantic internal transcribed spacer principal coordinate analysis

species delimitation.
INTRODUCTION
Species form the basis of analyses of biogeography,
ecology and conservation biology (Sites & Marshall,
2004), but many remain undescribed. With the
increasing disturbance and destruction of natural ecosystems, there is an urgent need to document and
catalogue new species before they become extinct. For
centuries, systematic biologists, particularly monographers, have relied on morphological characters for
diagnosing and delimiting species. However, speciation is not always accompanied by clear morphological
differentiation. Morphology in some instances is difficult to use in delineating species, especially in
so-called cryptic species, i.e. morphologically similar
populations that are probably, or at least to some
extent, reproductively isolated. Cryptic species are
often found in poorly studied species complexes
or typically widespread morphologically variable
species. Cryptic species are now being uncovered with
the aid of molecular techniques (e.g. Dick, Abdul*Corresponding author. E-mail: kenfackd@si.edu
186
Salim & Bermingham, 2003; Hebert et al., 2004;

Martnez-Ortega et al., 2004; Muellner, Pennington &
Chase, 2009) and, in some instances, are subsequently confirmed with morphological and/or ecological data. This underscores the need for careful
analysis of morphological data, taking into consideration all detectable differences coupled with analysis
of habitat and distributional information, which may
be as important as molecular data in discovering
cryptic species.
An example of cryptic speciation can be found in
Carapa Aubl., a poorly known genus of the mahogany
family (Meliaceae). Species of this genus constitute a
valuable timber resource in many countries and some
also have medicinal uses (Bouquet, 1969; Kerharo &
Adam, 1974; Adjanohoun & Ake Assi, 1979; Styles,
1981; Adjanohoun & Ahyi, 1985; Schultes & Raffauf,
1990; Gueye, Kenfack & Forget, 2010). Carapa comprises small to large trees occurring in a wide range
of habitats from gallery forest in savannas to
humid forest, from sea level to above 2000 m elevation, in Africa, central and northern South America,
and the West Indies (Fig. 1). Within the family,
2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2011, 165, 186221
MORPHOMETRICS OF CARAPA
187
Figure 1. Global distribution of Carapa.
morphological (Pennington & Styles, 1975), anatomical (Kribs, 1930) and molecular data (Muellner et al.,
2003) place Carapa in subfamily Swietenioideae.
Within this subfamily, the septifragal capsule and
unwinged seeds with a woody testa are unique distinctive features of this genus. In the past, questions
have been raised about the generic boundary between
Carapa and the morphologically similar mangrove
genus Xylocarpus Koen. In fact, taxa currently placed
in the two genera were treated under Carapa by
Lamarck (1785) and de Candolle (1878), and it was
only in 1896 that Harms reinstated Xylocarpus
(Harms, 1896). Since then, while there has been
agreement concerning the generic boundaries of
Carapa, there has been substantial disagreement
over the number of species recognized within the
genus and their circumscription.
In his monograph of Meliaceae, de Candolle (1878)
recognized six species of Carapa, of which two were
later transferred to Xylocarpus. By 1917, seven additional species had been described in the genus from
the African rainforest. Harms (1940) in the second
edition of Die natrlichen Pflanzenfamilien, recognized 11 species of Carapa, and Staner (1941), in the
treatment of Meliaceae of the Belgian Congo (DR
Congo), indicated 14 species of which five were found
in tropical America and nine in tropical Africa. The
most recent and comprehensive treatment of the
genus was that of Noamesi (1958). In his revision of
tribe Xylocarpeae, he recognized only seven species

(C. guianensis Aubl., C. surinamensis Miq., C. macrocarpa Ducke and C. nicaraguensis C.DC. in
America and C. batesii C.DC., C. procera DC. and C.
grandiflora Sprague in Africa), reducing several other
names to synonymy. Styles (1981), in his treatment of
Swietenioideae of the Neotropics, reduced C. nicaraguensis and C. macrocarpa to synonymy under C.
guianensis, arguing that the differences in pedicel
length and indumentum used by Noamesi (1958) were
not adequate to discriminate between them. Further,
Styles (1981) treated C. surinamensis as a synonym of
C. procera, thus circumscribing it to have a amphiAtlantic distribution. Styles & White (1991) went
even further, placing all African taxa in synonymy
under C. procera, claiming that, at least on the basis
of herbarium material, there were no clear characters
to differentiate them. Thus, with the description of
C. megistocarpa A.H.Gentry & Dodson (Gentry &
Dodson, 1988) from Ecuador, three species of Carapa
are currently recognized in America and only one in
Africa. However, there is general consensus concerning the extreme variability of C. guianensis and C.
procera, and several authors have emphasized that
further investigations are needed to understand the
taxonomic status of the different morphological variants within these two species better (Staner, 1941;
Hutchinson & Dalziel, 1958; Keay, Onoche & Stanfield, 1964; Keay, 1989; Styles & White, 1991). Varia-
188
D. KENFACK
tion in leaf length, number of pairs of leaflets, petiole

length, leaflet shape and dimensions (Fig. 2), inflorescence and flower indumentum, flower colour (Fig. 3)
and size, pedicel length, number of ovules per locule,
fruit colour, and the development of warty excrescences on the surface of fruits (Fig. 4) have been
mentioned and used in the past to distinguish several
taxa, most of which are currently included within
more broadly defined species.
The main objectives of this study are to use a
combination of field observations, spatial information
and statistical analyses of morphological data within
the framework of a molecular phylogenetic tree to
reassess taxonomic boundaries in Carapa. Three
basic questions are addressed: (1) how many morphological entities can be recognized within the genus?;
(2) what are their distinctive features?; and (3) how is
this morphological variation partitioned within the
distribution range of the genus and the various habitats they grow in? The answers to these three questions are important: (1) to understand the roles that
geography and landscape heterogeneity have played
in morphological divergence; and (2) to delimit robust
and reliably recognizable morphological species in
Carapa.
MATERIAL AND METHODS

PRELIMINARY
HERBARIUM STUDIES AND FIELDWORK
A preliminary morphological study of herbarium

specimens was carried out to assess the variation of
the characters mentioned above and to understand
the geographical distribution of the different character states. Specimens were examined in search of
distinctive combinations of characters that would
enable them to be grouped into cohesive entities (morphospecies). Twenty-seven morphospecies (a priori
groups) were defined based on combinations of characters used in the past to distinguish different
species: number of leaflets per leaf; presence/absence
of acumen on leaflets; presence/absence of indumentum on the rachis, midrib, inflorescence or flowers;
presence/absence of a pedicel; position of inflorescence
(cauliflorous vs. ramiflorous); and flower merosity.
These morphospecies were working hypotheses for
the remainder of the study.
Fieldwork conducted between 2003 and 2006 was
designed to study further most of the morphological
variation mentioned above, in four countries in Africa
(Cameroon, Gabon, Ghana and Senegal) and three in
Central and South America (Panama, Guyana and
Ecuador). Most of these countries include type localities of some published names in Carapa. Fieldwork
was directed towards assessing possible correlations
between the variation of the characters recorded, field
features, such as habit and organ colour, and habitat

requirement for the morphospecies. Collection of
additional herbarium specimens and leaf tissue in
silica gel for morphological and molecular studies was
also carried out. Field trips also provided the opportunity to study herbarium specimens housed at IFAN,
GC, SCA, YA, LBV, BG, STRI, PANAMA, QCA and
QCNA (herbarium abbreviations following Holmgren,
Holmgren & Barnett, 1990).
MOLECULAR
DATA
DNA was extracted from leaf tissue (silica gelpreserved or herbarium material) for one representative of all but three of the morphospecies (Table 1)
and representatives of 15 other genera of Meliaceae.
The amplification of the entire internal transcribed
spacer (ITS) region of the ribosomal DNA (ITS1-5.8SITS2) was carried out by polymerase chain reaction
(PCR), using the primers ITS4 (White et al., 1990)
and LEU1 (Baldwin, 1993). After sequencing in both
forward and reverse directions, phylogenetic analyses
were performed using maximum likelihood (ML) and
all but one of the remaining 13 genera currently
recognized in Swietenioideae and three of Melioideae
as outgroups. For the ML analysis, the GTR+I+G
nucleotide substitution model was chosen using the
Akaike information criterion (AIC) in Modeltest
version 3.06 (Posada & Crandall, 1998, 2001) as the
best fit for our data set. Heuristic searches were
carried out in PAUP* with tree bisection
reconnection (TBR) branch swapping and Multrees
option in effect. Clade support was estimated using a
likelihood bootstrap (LB) of 100 replicates with TBR
branch swapping and Multrees option in effect. The
best tree obtained from this analysis was used as a
guide for partitioning morphological data in the morphometric analyses.
MORPHOLOGICAL
CHARACTERS EXAMINED AND
DATA ANALYSIS
The application of numerical taxonomy to Carapa

posed several problems. First, as with palms (Henderson, 2006), leaves and/or leaflets in several species
of Carapa are quite large (up to 1.8 m long, with eight
to 40 leaflets of up to 65 18 cm), which makes them
difficult to fit onto standard herbarium sheets. Consequently, many herbarium collections are fragmentary and lack potentially useful characters such as
number of leaflets per leaf and length of the petiole,
which preliminary observations suggested might be
useful in discriminating among taxa (Fig. 2). Second,
leaflet size generally increases from the base of the
leaf towards the apex (e.g. see Fig. 2I), precluding
189
Figure 2. Variation in leaf size, number of leaflets and leaflet shape in nine morphospecies of Carapa. Numbers
correspond to the codes of the morphospecies in the multivariate analysis.
Figure 3. Variation in flower colour of 12 morphospecies of Carapa. Numbers correspond to the codes of the morphospecies in the multivariate analysis.
[Photographs: G and L, Pierre-Michel Forget; remaining, David Kenfack].
190
D. KENFACK
Figure 4. Variation in fruit shape, surface ornamentation and colour in Carapa: Numbers corresponds to the codes of the morphospecies in the multivariate
analysis. [Photographs: J, Charles Doumenge; M, Sainge Moses; O, Mathieu Gueye; P and Q, Pierre-Michel Forget; remaining, David Kenfack].
191
192
D. KENFACK
Table 1. List of quantitative and qualitative characters assessed for the study of morphological variation in Carapa
Leaf characters
Floral characters
Number of leaflets (LNUM)

Length of the petiole (PETL)
Length of the rachis (RACL)
Total length of the leaf (LL)
Ratio length of petiole/leaf length (PETL/LL)
Diameter of the rachis (at the insertion of the basal leaflet) (RACD)
Number of glands on the base of the petiole (GNUM)
Length of the basal leaflet petiolule (BLPET)
Length of the basal leaflet lamina (BLL)
Width of the basal leaflet lamina (BLW)
Number of lateral veins of the basal leaflet (BLV)
Width of the basal leaflet lamina at 0.5 mm from the apex (BLAW)
Length of the mid-leaflet petiolule (MLPET)
Length of the mid-leaflet lamina (MLL)
Width of the mid-leaflet lamina (MLW)
Number of lateral veins of the mid-leaflet (MLV)
Width of the mid-leaflet lamina at 0.5 mm from the apex (MLAW)
Terminal leaflet petiolule length (TLPET)
Length of the terminal leaflet lamina (TLL)
Width of the terminal leaflet lamina (TLLW)
Number of lateral veins of the terminal leaflet (TLV)
Width of the terminal leaflet lamina at 0.5 mm from the apex (TLAW)
Length of the petiolule of the longest leaflet
Length of the longest leaflet
Width of the longest leaflet
Number of lateral veins of the longest leaflet
Basal leaflet length/terminal leaflet length ratio (BLL/TLL)
Width/length of the terminal leaflet lamina (TLW/TLL)
Length of the pedicel (PEDL)*

Merosity of the flower (MERO)*
Length of the sepal (SEPL)*
Diameter of the calyx (CALD)*
Length of the petal (PETAL)*
Width of the petal (PETAW)*
Length of the staminal tube (STATL)*
Diameter of the staminal tube (STATD)*
Length of the staminal tube lobe (STATLL)*
Height of the nectary (NECH)*
Diameter of the nectary (NECD)*
Length of the anther (ANTL)*
Width of the anther (ANTW)*
Length of the ovary (OVAL)*
Width of the ovary (OVAW)*
Length of the style (STYL)*
Length of the pistil (PISL)*
Diameter of the stigma (STID)*
Number of ovules per locule (OVUN)*
Longest inflorescence branch
Position of the inflorescence (CAUL)*
Fruit and seed characters

Length of the fruit
Diameter of the fruit
Length of the seed (SEEL)
Width of the seed (SEEW)
Thickness of the seed (SEET)
Length of the hilum (HILL)
Width of the hilum (HILW)
Ratio length of the hilum/length of the seed (HILL/SEEL)
Ratio width of the hilum/length of the seed (HILL/SEEW)
Ratio length of the hilum/width of the seed (HILL/SEEW)
Qualitative data
Leaflet texture
Leaflet acumen (presence absence)
Petiole and rachis indumentum (RACIND)
Leaflet lamina indumentum
Inflorescence length
Indumentum of the inflorescence
Indumentum of the pedicel (PEDIND)*
Indumentum of sepals (CALIND)*
Colour of sepals
Indumentum of petals
Colour of petals
Colour of the nectary
Ornamentation of the surface of the fruit
Colour of the mature fruit
Ornamentation of the seed coat
*Codes are given for characters used in the morphometric analyses.
comparisons among many specimens. In the rare

cases where herbarium specimens comprise complete
leaves, they are frequently folded in a way that does
not allow scoring most characters. Third, fruits in
Carapa vary considerably (Fig. 4), but have been
poorly collected, especially in Africa. Fourth, seed coat
ornamentation and hilum shape are characteristic for
many species (Fig. 5), but again few collections with
seeds exist for most species.
Thus, the flower was the only organ that could
allow inclusive comparative analysis of herbarium
material. Carapa flowers are imperfect, with functional organs of one kind (stamens or ovary) and
developed, but non-functional organs of the other
(Pennington & Styles, 1975). Furthermore, the two
flower types are easily distinguishable. Morphometric
comparisons can be made only between flowers of the
same type, and carpellate flowers represented only 37
and 17%, respectively, of the tetramerous and pentamerous flowers dissected. Because of this paucity of
carpellate flowers, measurements from staminate
flowers were used primarily in the analyses presented
193
Figure 5. Variation in seed coat colour, ornamentation and hilum size in 16 morphospecies of Carapa. Numbers refer to
the denomination of the morphospecies in the multivariate analysis.
below, supplemented with data from other organs

when possible.
Over 1360 herbarium specimens, including my own
collections and loans from BR, F, GH, K, MO, NY, P
and US were examined, but only 292 (Appendix) were
used in the morphometric analysis for reasons enumerated below. Each specimen was scored for as
many of the 72 characters in Table 1 as possible and
assigned to a morphospecies (see above) numbered 1
to 27.
194
D. KENFACK
To display the variation of the characters measured

among morphospecies and depict those that are distinctive for each morphospecies, quantitative characters were represented graphically in the form of boxand-whiskers plots (Tukey, 1977; figures not shown),
using SPSS version 12.0. For each character, and
each morphospecies, the box-and-whiskers plots show
the value of the median, the first and third quartiles,
the minimum and maximum measurements and the
outliers. A one-way ANOVA combined with Tukeys
post hoc test was used to investigate which of the
quantitative characters varied significantly among
the morphospecies and between which morphospecies
they were significantly different.
Of the 72 characters examined, only 18 quantitative and four qualitative floral characters were used
in the multivariate analysis because of the problems
mentioned above. Principal coordinate analysis
(PCoA) was carried out using NTSYSpc version 2.02f
(Rohlf, 1998). In these analyses, quantitative characters were standardized to reduce the effects of different scales of measurement when they comprised
continuous and meristic characters. Qualitative characters were not standardized (Sneath & Sokal, 1973).
PCoA (Gower, 1966) is an exploratory method that
projects in a reduced space the distances among
objects computed using appropriate similarity or dissimilarity measures (Legendre & Legendre, 1998).
Compared with the traditional principal component
analysis (PCA), PCoA performs better in identifying
groups of objects (Rohlf, 1972; Chae & Warde, 2006)
and also allows the use of both quantitative and
qualitative data (Legendre & Legendre, 1998). The
outcome of PCoA does not depend on a priori classification of specimens and was therefore used to investigate independently whether the groups of specimens
initially assigned to particular morphospecies could
be recovered in the analyses.
Multivariate analyses were performed on specimens irrespective of the names they bore, using the
tree topology resulting from the analysis of the ITS
sequence data (Fig. 6) as a guide to partitioning the
data sets. A nested analysis approach was used with
successive PCoAs at global, clades and subclades
level, making a total of 11 PCoAs.
GEOGRAPHICAL
Figure 6. Simplified best tree obtained from the

maximum likelihood analysis of the nuclear internal transcribed spacer (ITS) ribosomal DNA of 24 morphospecies of
Carapa. Only 24 of the 27 morphospecies delimited were
used in this molecular analysis. Open circles on the
branches indicate clades supported by at least one of the
parsimony bootstraps ( 50%), likelihood bootstrap
( 50%) and Bayesian posterior probabilities ( 70%). The
four major clades represented by roman numerals correspond to broad geographical areas.
maps of the morphospecies were produced using

Arcview 3.2.
DISTRIBUTION
During fieldwork, altitude, latitude and longitude were

recorded for each specimen using a hand-held GPS
unit. For herbarium specimens, these data were taken
from labels or were estimated post facto for detailed
localities using the National Geospatial-Intelligence
Agency (NGA) website (http://earth-info.nga.mil/gns/
html/) and/or Google Earth software (available at
http://www.google.com/earth/index.html). Distribution
TAXONOMIC
DECISIONS
The primary goal of reassessing patterns of morphological variation in Carapa was to look for morphological discontinuities among groups of specimens
(morphospecies) that could allow one to distinguish
them from other such groups. By and large, the
decisions were based on the PCoA of staminate
flowers. However, in some instances, results from the
univariate analysis of leaf and seed characters and
other qualitative characters, such as flower colour,
fruit shape and seed coat ornamentation, were used
to supplement the results of the multivariate analysis. In making decisions as to whether these groups of
specimens represented species, as far as possible a
combination of morphological, molecular and spatial
data was used to distinguish separate species. Groups
of specimens that (1) showed strong morphological
discontinuities (e.g. non-overlapping number of leaflets, sessile vs. pedicellate flowers, different number
of ovules per locule, etc.), (2) had overlapping ranges
and/or occur in sympatry and/or (3) belonged to different well-supported clades of the ITS phylogenetic
tree were considered as belonging to separate species.
RESULTS
GLOBAL
MORPHOLOGICAL VARIATION
A one-way ANOVA showed significant differences for

leaf (6.8 < F25,240 < 26, P < 10-14), floral (8.5 < F25,290 <
40, P < 10-20) and seed (51 < F18,697 < 462, P < 10-40)
quantitative characters. Tukeys post hoc analyses
showed on average the most significant differences
(P < 0.05) among the 27 morphospecies for seed characters (64% of pairwise comparison of means) compared with floral (38%) and leaf characters (25%)
(Table 2). The most useful leaf characters for discriminating among morphospecies are the number of pairs
of leaflets (LNUM), the total leaf length (LL) and the
ratio of petiole length to total leaf length (PETL/LL).
Among floral characters, calyx, corolla and staminal
tube dimensions are the most discriminating. The
length and width of the hilum and hilum dimensions
relative to seed dimensions are the most discriminating seed characters (Table 2, Fig. 5).
A PCoA of all 292 specimens spanning the entire
distribution range of the genus in Africa and
America was performed, using 18 quantitative and
four qualitative floral characters (Table 3), to test the
morphological distinctiveness of the four clades of
the ITS tree and the 27 morphospecies. Neither the
clades recovered in the ITS analysis (Fig. 7A) nor the
27 morphospecies (Fig. 7B) could be distinguished in
this analysis. However, it is apparent that there is
only slight overlap of the American (morphospecies
111) and African (morphospecies 1227) specimens,
and the three morphospecies that have biovulate
locules (number 4, 19 and 27), form a cluster separate from the rest. The first three axes explained
68% of the total variance among the specimens. The
variation along the first axis (28% of the total variance) largely agrees with the highest positive loadings for three qualitative characters, merosity,
195
pedicel indumentum and calyx indumentum, and

ovule number; and the highest negative loadings for
petal, pedicel and staminal tube length. The second
axis explained 24.6% of the total variation and had
highest loadings for nectary height and staminal
tube length. Because representatives of the morphospecies from America and Africa formed distinct
clades in the ITS tree topology (Fig. 6), further
analyses were carried out separately for American
and African specimens.
MORPHOLOGICAL VARIATION
AMERICAN SPECIMENS
IN
Specimens studied were collected from throughout

central and northern South America and the West
Indies and were provisionally placed in 11 morphospecies. The analysis of variance of 24 leaf characters measured from 130 leaves of ten of the 11
morphospecies recognized in the Neotropics showed
significant differences (3.5 < F9,130 < 39, P < 10-3)
among the different morphospecies. Leaflet dimensions, secondary veins and leaflet number are the
most variable characters (Table 2). Among floral characters, pedicel length, staminal tube length, sepal and
petal length are the most discriminative characters
(Table 2). The number of ovules per locule is by far
the least variable character within morphospecies.
Seed characters also varied significantly among the
nine morphospecies (13.5 < F8,280 < 39, P < 10-9), with
hilum dimensions and seed width as the best discriminators (Table 2). Morphospecies 3 and 9 have the
smallest seeds with a short hilum and morphospecies
4, 5, 6 and 8 have the largest seeds with the longest
hilum (Fig. 5). However, seed size is not correlated
with hilum length (y = 0.2956x + 8.3115, R2 = 0.0973).
Seeds in morphospecies 1 are smaller compared with
those in morphospecies 4, 5 and 6, but have relatively
longer hilum (Fig. 5). In addition to these quantitative characters, variation was also found in seed coat
colour and ornamentation. The seeds are always
whitish in morphospecies 4 and dark brown in morphospecies 1 and 5. The seed coat is pitted in morphospecies 4, longitudinally fissured in morphospecies
1 and smooth in the other morphospecies (Fig. 5).
A PCoA performed on all 131 specimens from
America using 18 quantitative and four qualitative
characters (Table 4) showed that the two major
molecular clades I and II are not morphologically
distinct (Fig. 8A, B). In the plane of axes 1 and 2, an
overlay of the 11 morphospecies shows two obvious
clouds, a homogeneous one corresponding to morphospecies 4 of clade I and a composite cloud that
includes specimens of the remaining morphospecies
in both clades (Fig. 8C). However, more careful
inspection of this composite cluster shows that there
196
D. KENFACK
Table 2. Number of significant differences (sig. diff.) resulting from multiple comparisons (using Tukeys post hoc
analysis) of means of leaf, floral and seed quantitative characters among 27 morphospecies of Carapa from Africa and
America
All specimens
American specimens
Variables
d.f.
Number of
sig. diff.
Per cent
Leaf
BLAW
BLL
BLL/TLL
BLPET
BLV
BLW
GNUM
LL
LNUM
MLAW
MLL
MLPET
MLV
MLW
PETL
PETL/LL
RACD
RACL
TLAW
TLL
TLL/TLW
TLPET
TLV
TLW
24
24
24
24
24
24
24
24
24
24
24
24
24
24
24
24
24
24
24
24
24
24
24
24
95
97
43
72
103
85
48
99
127
91
88
67
88
100
95
118
41
103
94
71
38
37
84
88
29
30
13
22
32
26
15
31
39
28
27
21
27
31
29
36
13
32
29
22
12
11
26
27
Floral
ANTL
ANTW
CALD
NECD
NECH
OVAL
OVAW
PEDL
PETAL
PETAW
PISL
SEPL
STATD
STATL
STATLL
STID
STYL
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
116
100
165
109
98
40
57
98
161
140
104
149
88
158
171
124
68
Seed
HILL
HILL/SEEL
HILL/SEEW
HILW
HILW/HILW
SEEH
SEEL
SEEW
18
18
18
18
18
18
18
18
128
131
129
105
87
82
105
100
African specimens
Number of
sig. diff.
Per cent
d.f.
Number of
sig. diff.
Per cent
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
24
21
2
14
14
19
4
17
19
18
23
15
19
25
11
2
14
17
23
19
17
15
22
22
53
47
4
31
31
42
9
38
42
40
51
33
42
56
24
4
31
38
51
42
38
33
49
49
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
19
24
28
21
22
14
28
39
52
26
23
19
14
27
38
64
13
39
27
26
2
5
15
34
16
20
23
18
18
12
23
33
43
22
19
16
12
23
32
53
11
33
23
22
2
4
13
28
39
33
55
36
33
13
19
33
54
47
35
50
29
53
57
41
23
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
24
17
28
19
20
11
20
39
29
27
20
28
24
33
32
15
12
53
38
62
42
44
24
44
87
64
60
44
62
53
73
71
33
27
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
36
40
58
49
33
16
11
13
53
50
37
51
21
48
27
56
21
34
38
55
47
31
15
10
12
50
48
35
49
20
46
26
53
20
75
77
75
61
51
48
61
58
8
8
8
8
8
8
8
8
31
28
28
21
15
24
32
29
86
78
78
58
42
67
89
81
9
9
9
9
9
9
9
9
34
39
40
32
22
21
15
19
76
87
89
71
49
47
33
42
d.f.
Table 3. Loadings of the three first axes of the principal
coordinate analysis (PCoA) of all 292 Carapa specimens
from throughout its range, eigenvalues, percentage of variance and cumulative percentage of variance explained by
the first three axes
Variables
PCoA1
PCoA2
PCoA3
MERO
PEDIND
CALIND
OVUN
CAUL
PEDL
NECH
OVAL
STID
STYL
PISL
OVAW
NECD
ANTW
ANTL
STATD
STATLL
STATL
PETW
PETL
SEPL
CALD
Eigenvalue
Percentage of variance
Cumulative percentage
of variance
0.7093
0.6959
0.5584
0.3752
0.3457
0.1919
0.1596
0.1180
0.0781
-0.0949
-0.0980
-0.1074
-0.1414
-0.2120
-0.2649
-0.2765
-0.3009
-0.3138
-0.3171
-0.3218
-0.3631
-0.4202
10.2893
28.9474
28.9474
-0.0584
-0.2815
-0.2482
0.0651
-0.3339
0.2762
0.4921
0.3299
0.3137
0.1613
0.3299
0.0144
0.2190
-0.3123
-0.0364
-0.3452
-0.7813
0.0624
0.0755
0.1569
-0.0336
-0.0655
8.8386
24.8661
53.8135
-0.2825
-0.1053
-0.1003
0.7218
0.0744
0.0319
-0.0390
0.1858
-0.3609
0.2120
0.2378
-0.2365
-0.2893
-0.2757
-0.0574
-0.0146
0.2850
0.2744
-0.0254
0.1203
-0.1494
-0.2074
5.1058
14.3645
68.1781
is no interdigitation between morphospecies 2, 5 and

7 with the rest of the groups. In the scatter plot of
axes 1 and 3, the two clades are again not discernable, but the separation of the morphospecies
becomes more evident (Fig. 8D). Here, specimens of
morphospecies 3 appear well separated from the large
composite cloud. Also the two pentamerous morphospecies (10 and 11) and the cauliflorous morphospecies (8 and 9) are separated from the others.
The first three axes account for 68.3% of the total
variation. The first axis alone accounts for almost
39% of the total variation, with highest loadings contributed by qualitative characters (CAUL, CALIND,
PEDIND, MERO) and the staminal tube diameter
(loaded negatively) and by two staminal tube characters (STATL, STATLL) and petal length (loaded positively) (Table 4). This first axis segregates specimens
with shorter petals and staminal tube and fewer
ovules (morphospecies 4) from those with much larger
flowers and/or more ovules. The second axis explained
17.8% of the total variation contributed by the two
197
qualitative characters PEDIND and CALIND (loaded

positively) and OVUN and OVAL (loaded negatively).
The third axis explained 11.6% of the variation and
had SEPL as a major positive loading; morphospecies
3 is well separated along this axis.
To investigate whether further patterns could be
detected in the variation shown by these specimens,
two subsequent analyses were performed for each of
the two major American clades (I and II). Clade I
comprised four morphospecies (1, 3, 10 and 11).
Except for morphospecies 1, which occurs partly west
of the Andes of Colombia, all specimens of these four
morphospecies were collected east of the Andes, from
Venezuela to French Guiana and southwards to Peru
and Brazil. American morphospecies with pentamerous flowers are restricted to this area, growing on the
Guiana Shield. A PCoA of the 78 specimens from this
area was performed using the same 18 quantitative
and two qualitative floral characters (Table 5). The
first axis of the PCoA (35.5% of the total variance)
separated the two pentamerous morphospecies from
the others, owing to the characters STID, STATD and
MERO (positively correlated) and OVUN, ANTW and
STATL (negatively correlated) (Fig. 9A). The second
axis (19.8% of the total variance) has STATLL and
ANTL as major positive loadings and separates morphospecies 3 from the rest. In the plane of the axes 1
and 2, morphospecies 10 and 11 remain intermixed,
although they can be distinguished based on seed
characters. A PCoA of 69 seeds of the two morphospecies using six quantitative seed characters showed
separate but adjacent groups in the two first axes of
variation (Forget et al., 2009).
Clade II included five morphospecies, all collected
west of the Andes, from Ecuador to Nicaragua. One
additional morphospecies (9) also occurs in this
area, but was not included in the molecular study. A
PCoA of the 53 specimens from this region was performed using 18 quantitative characters and two
qualitative characters. In this analysis, the first
three axes accounted for 80.5% of the total variance.
The first axis explained 65.5% of the total variance,
with OVUN, NECH and PEDL as major positive
loadings and the two qualitative characters
(PEDIND and CALIND) as negative loadings
(Table 6). This axis separated specimens with few
ovules and reduced indumentum (morphospecies 4)
in the negative side from those with more ovules
per locule and no indumentum (morphospecies 8
and 9) in the positive side (Fig. 10). The second axis
accounted for only 11.8% of the variance and had
three characters (STID, ANTW and NECH) positively correlated and two (CALD and SEPL) negatively correlated. Morphospecies 1, 5 and 6 were
well separated from one another along this axis.
Except for morphospecies 8, which clusters with
198
D. KENFACK
Figure 7. Scatter plots of the two first axes of the principal coordinate analysis of specimens of Carapa from Africa and
America. A, all specimens with overlay of the four major clades from the internal transcribed spacer (ITS) topology.
B, with overlay of the 27 morphospecies.
coordinate analysis (PCoA) of all 131 specimens of Carapa
from the New World Tropics, eigenvalues, percentage of
variance and cumulative percentage of variance explained
by the first three axes
Variables
PCoA1
PCoA2
PCoA3
MERO
STID
CALIND
PEDIND
CAUL
OVAL
NECH
OVUN
PEDL
ANTW
OVAW
NECD
STATD
PISL
PETAW
SEPL
ANTL
STYL
CALD
STATLL
PETAL
STATL
Eigenvalue
of variance
0.9342
0.8584
0.7548
0.7423
0.6694
0.4623
0.2049
0.1486
0.0414
0.0169
-0.0496
-0.0841
-0.2747
-0.3593
-0.3665
-0.4513
-0.4785
-0.4963
-0.4973
-0.5231
-0.5826
-0.6698
4.8557
38.8527
38.8527
-0.3260
0.5964
-0.8177
-0.7465
-0.1830
0.6356
0.5750
0.6929
0.3398
0.0466
0.0222
0.2381
0.1656
0.1999
0.1566
-0.4978
-0.2701
-0.3269
-0.4528
-0.1909
0.0849
0.0580
2.22586
17.8101
56.6628
-0.1541
-0.2548
0.0956
0.1096
0.2457
-0.3761
0.0153
0.2729
0.2043
0.7291
-0.4137
-0.1375
-0.0497
-0.1808
0.1423
-0.4977
0.4169
0.0317
-0.5030
0.2519
0.0321
0.0201
1.461783
11.6964
68.3592
morphospecies 9, the others are clearly separated

from each other and occupy different areas in this
morphological space (Fig. 10).
MORPHOLOGICAL
VARIATION IN
AFRICAN
SPECIMENS
The specimens examined in this analysis came from

throughout the range of the genus in Africa. All have
pentamerous flowers, except for specimens from the
Island of So Tom, which are predominantly tetramerous. Fifteen names have been published for
specimens in this region, and 16 morphospecies have
been distinguished.
The analysis of variance of 24 leaf characters measured from 110 African specimens showed significant
differences (3 < F15,110 < 40, P < 10-3) among the 16
morphospecies. Leaf characters revealed more discontinuity between African morphospecies compared
with those from America (Table 2, cf. Fig. 2). Leaves
are also longer in African specimens, particularly in
morphospecies 17, 23 and 25, which also have longer
199
rachises, but not longer petioles. Leaf and rachis

length
are
strongly
positively
correlated
(y = 0.9283x + 21.37, R2 = 0.8725), but there is no correlation between leaf length and petiole length
(y = 0.3813x + 56.23, R2 = 0.0166). The ratio PETL/LL
thus appears to be of taxonomic value for distinguishing these entities (Fig. 2). Long-leaved morphospecies
(17, 23 and 25) have more leaflets and also more
extrafloral nectaries on the base of their petiole than
other morphospecies. Tukeys post hoc analysis
(Table 2) confirms that the ratio PETL/LL is by far
the most useful discriminating leaf character (53% of
significant differences among the 16 morphospecies),
followed by the number of leaflets per leaf (LNUM).
Floral characters also vary significantly among the
16 morphospecies (4.5 < F15,109 < 43.8, P < 10-6).
Dimensions of the tepals, nectary and stigma are the
most discontinuous characters that allow separation
of the different morphospecies (Table 2). The number
of ovules per locule is constant in each morphospecies;
those with four ovules predominate and those with
two or six ovules are exceptional. Although the
number of ovules per locule in these morphospecies
was nearly always even, one specimen from Liberia,
morphologically close to morphospecies 12, which normally has four ovules, had seven ovules in two
flowers.
Seeds were available for only 10 of the 16 African
morphospecies and the two to four seeds included in
the analysis for morphospecies 15, 17, 20 and 25 came
from a single plant. Seeds vary for the eight characters examined (48 < F9,416 < 523, P < < 10-50) and can
be useful in separating the 10 morphospecies. Hilum
size and shape are the most discriminating characters
(Table 2). All West African morphospecies (12, 13 and
14) have a smaller hilum whereas those from high
elevation (4, 8 and 9) have a more developed hilum. In
addition to these quantitative characters, seed coat
ornamentation varies among these species. Morphospecies from high elevation (15, 19 and 20) have a
smooth and shiny seed coat, whereas those from
lower elevation have a more or less rugose seed coat
(Fig. 5).
Specimens of the two African clades III and IV
largely overlapped in the morphospace of the first
three axes of variation in a PCoA of all 160 African
specimens (Fig. 11A, B). Clade III, however, showed
much more discontinuous variation compared with
clade IV, with at least three morphologically distinct
groups in the plane of the two first axes and two
groups in the plane of axes 1 and 3. This analysis
distinguished clearly two of the 16 morphospecies (19
and 27) in the scatter plots of the first three axes
(Fig. 11C, D). Although the rest of the specimens form
a continuum, it is apparent that some morphospecies
(e.g. 14 and 17, 14 and 19, 19 and 25, etc.) do not
200
D. KENFACK
0.30
Principal Coodinate 2
0.20
0.10
0.00
0.10
0.20
0.30
Clade I
Clade II
not sequenced
0.40
0.50 0.40 0.30 0.20 0.10
0.00
0.10
0.20
0.30
0.40
0.20
0.30
0.40
0.25
0.20
0.15
0.10
0.05
0.00
0.05
0.10
0.15
0.20
Clade I
Clade II
not sequenced
0.25
0.30
0.50 0.40 0.30 0.20 0.10
0.00
0.10
Figure 8. Scatter plots of the principal coordinate analysis of specimens of Carapa from America. AB, axes 1 and 2 and
1 and 3, with overlay of the two major clades I and II. CD, axes 1 and 2 and 1 and 3 with overlay of the 11 morphospecies.
The first three axes represent 68.8% of the total variation. See Table 4 for the loadings of the different characters used.
0.30
0.20
0.10
0.00
0.10
0.20
0.30
1
5
9
2
6
10
3
7
11
4
8
0.40
0.50 0.40 0.30 0.20 0.10
0.00
0.10
0.20
0.30
0.40
0.20
0.30
0.40
0.25
0.20
0.15
0.10
0.05
0.00
0.05
0.10
0.15
0.20
1
5
9
0.25
2
6
10
3
7
11
4
8
0.30
0.50 0.40 0.30 0.20 0.10
0.00
0.10
Figure 8. Continued
201
202
D. KENFACK

coordinate analysis (PCoA) of 78 specimens of Carapa
from the Cis-Andes and belonging to clade I of the internal transcribed spacer (ITS) topology, eigenvalues, percentage of variance and cumulative percentage of variance
explained by the first three axes
Variables
PCoA1
PCoA2
PCoA3
STID
STATD
MERO
CALD
PEDL
SEPL
OVAL
ANTL
PEDIND
STATLL
OVAW
PETL
NECH
PETW
NECD
PISL
STYL
STATL
ANTW
OVUN
Eigenvalue
of variance
0.4076
0.3083
0.2966
0.2893
0.2808
0.2532
0.0566
0.0462
-0.0566
-0.0685
-0.0697
-0.0750
-0.0857
-0.1206
-0.1622
-0.1712
-0.2370
-0.2655
-0.3095
-0.3170
1.6840
35.5336
35.5336
-0.1481
-0.0829
-0.1096
0.1902
0.0832
0.2067
-0.4510
0.3039
0.0078
0.3988
-0.4764
0.1927
0.0241
0.1679
-0.1678
-0.2733
-0.0645
0.0778
0.0088
0.1118
0.9423
19.8836
55.4172
-0.0849
-0.1438
0.0792
0.0304
0.3254
-0.2820
0.0713
-0.0507
0.1221
-0.3225
0.1107
0.1230
0.0540
0.4762
0.0570
-0.3015
-0.4359
-0.0298
-0.1118
0.3136
0.5001
10.5525
65.9698
overlap in either of the two scatter plots (Fig. 11C, D).

The three first axes of variation in this analysis
account for 71.5% of the total variance (see Table 7 for
the loadings of the variables). The most important
characters that explain the clustering patterns along
the first axis are ANTW, STID and OVUN (loaded
positively) and PISL, STYL and STATL (loaded
negatively).
A PCoA of 51 specimens belonging to the six morphospecies of clade III and one specimen of morphospecies 18 (from So Tom) was performed using
18 quantitative characters and three qualitative characters (Table 8). The seven morphospecies are largely
morphologically distinct on axes 1 and 2 and on axes
1 and 3 (Fig. 12A, B). Morphospecies 23 and 25
largely overlap in axes 1 and 2, but are well separated
in axes 1 and 3, based on PEDL, STATLL, CALD and
PETAL (Table 8).
Clade IV included nine morphospecies from two
major phytogeographical areas in Africa separated by
the Dahomey gap, the Upper Guinea and the Lower
Guinea and Congolia forest block (White, 1979). A
PCoA was performed on 108 specimens representing

these morphospecies, again with the single specimen
of morphospecies 18. In the scatter plot of the two
first axes (55% of the total variance), some of the
morphospecies occupy particular parts of the morphospace, but they all largely overlap (Fig. 13A,
Table 9). Subsequent analyses were then performed
using samples selected on the basis of phytogeographical considerations, separating specimens of
West African morphospecies (morphospecies 12, 13
and 14), which formed a monophyletic group embedded within clade IV (Fig. 3), from those originating
from Central African forest block (see above).
A PCoA of the 55 specimens from West Africa using
17 quantitative characters and two qualitative indumentum characters (Table 10) separated the three
morphospecies in the plane of the first and second
axes (Fig. 13B). The variables contributing the most
to the variation along the first axis (35.57% of the
total variance) were PEDL, STATLL and STAD (positively correlated) and STYL and PISL (negatively
correlated) (Table 10). Specimens of all three groups
varied considerably along this axis. The separation of
specimens along the second axis, which accounted for
27.76% of the total variation, was attributable to the
contribution of the character PEDL and the qualitative character PEDIND (positive loading) and NECD
and OVAL (negative loading).
The data matrix of the remaining six Central
African specimens groups along with the single specimen of group 18 from So Tom comprised 18 quantitative and two qualitative characters measured
from 54 specimens. The first three axes of the PCoA
accounted for 71% of the total variation (Table 11).
The highest positive loadings on the first axis were for
OVUN and OVAL, and ANTW and STALL loaded
negatively. For the second axis, STATD and STATLL
had heavy positive loading and PETAL and PETAW
had the highest negative loadings (Table 11). The
scatter plot of the first and second axes (Fig. 13C)
clearly depicted only distinct groups, of which one is
a composite comprising morphospecies 18, 21, 22, 24
and 26. However, a separate analysis comprising only
specimens from this composite group showed that
four of them were morphologically distinct (Fig. 13D).
Morphospecies 18 appeared embedded within morphospecies 21. However the two are separate in axes
1 and 3 (figure not shown).
DISCUSSION AND
TAXONOMIC INFERENCES
The ITS sequence data and statistical analyses of
leaf, floral and seed characters showed correlated
molecular and morphological variation in Carapa. In
203
Figure 9. Scatter plots of axes 1 and 2 of the principal coordinate analysis of specimens of Carapa from the Cis-Andes
with overlay of the five morphospecies of the area. See Table 5 for the loadings of the different characters used.
contrast to the taxonomic interpretation found in the

current literature, the variation displayed suggests
that far more than three morphological entities exist
in the genus. The 24 (of 27) morphospecies included in
the phylogenetic analysis formed four molecular
clades, two African and two American, with strong
geographical structuring. When analysed simultaneously, specimens of the four clades did not form
discrete morphological groups, perhaps explaining the
tendency of previous authors to recognize a limited
number of species. Nonetheless, in a PCoA using
floral characters, African and American specimens
largely occupied different regions of morphospace
with little interdigitation among them.
The 27 morphospecies (Table 12) likewise did not
form discrete groups in multivariate analyses performed at both global and continental scales,
although each group occupied a specific area of the
morphospace. When specimens of individual clades
(as identified by molecular analyses), which also corresponded to broad phytogeographical regions, were
analysed separately, patterns of variation among the
morphospecies became more apparent.
Clade I comprised four morphospecies (1, 3, 10 and
11), all from the Cis-Andes, i.e. the region east of the
Andes from the Caribbean to Brazil (Fig. 14). Because
there was geographical structuring in the phylogenetic analysis, specimens of morphospecies 7 from
Venezuela were also included in the morphometric

analysis of this clade. The analysis showed morphological discontinuities among morphospecies 1, 3 and
7, all with tetramerous flowers, which had previously
been considered as part of the C. guianensis complex,
and which clearly form separate clusters in the PCoA.
They also differ in their distributional ranges and
habitats.
Specimens of morphospecies 1 occur almost throughout the range of the genus in the New World, from
Brazil and Peru northwards to the Caribbean, and
from French Guiana to Colombia, and along the Caribbean coast of Central America from Panama to Belize
(Fig. 14), thus overlapping with most of the other
American morphospecies. Throughout most of its
range, morphospecies 1 generally grows in flood plains
and marshy conditions, but probably also on terra
firme in French Guiana, Brazil and Peru. Specimens of
this morphospecies 1 are unique in the genus in having
small leaflets with an acute to acuminate leaflet apex
and sessile (to subsessile) flowers with four-ovulate
locules, relatively small seeds that yet have the longest
hila so far documented in the genus. Material assigned
to this morphospecies matched perfectly the type specimen of C. guianensis Aubl. and are hereafter referred
to as C. guianensis sensu stricto (s.s.).
Morphospecies 3 has a narrower geographical
range (Fig. 14), from the Loreto province of Peru and
204
D. KENFACK

coordinate analysis (PCoA) of 53 specimens of Carapa
from the Trans-Andes region belonging to clade II of the
internal transcribed spacer (ITS) topology, eigenvalues,
percentage of variance and cumulative percentage of variance explained by the first three axes
Variables
PCoA1
PCoA2
PCoA3
OVUN
NECH
PEDL
STATD
STATLL
OVAL
STID
PISL
STATL
PETW
PETL
ANTW
NECD
ANTL
OVAW
STYL
CALD
SEPL
PEDIND
CALIND
Eigenvalue
of variance
0.2132
0.2017
0.1514
0.1275
0.1249
0.1227
0.1226
0.1134
0.1121
0.1050
0.1004
0.0679
0.0310
0.0015
-0.0566
-0.0577
-0.1242
-0.1455
-0.5997
-0.6116
3.97535
62.5358
62.5358
0.1964
0.2778
0.1784
-0.0410
-0.1051
0.2003
0.3916
-0.1489
-0.2052
-0.1239
-0.1500
0.2887
0.0446
-0.1016
-0.1114
-0.2642
-0.3841
-0.3618
0.2067
0.2125
0.7504
11.8048
74.3405
-0.1066
-0.0600
0.0549
-0.0527
0.0673
0.4622
0.1033
-0.0238
-0.0191
-0.0532
-0.0275
-0.4007
0.2081
-0.4309
0.4345
-0.3618
0.1708
0.0783
-0.0066
-0.0367
0.3946
6.2081
80.5486
the adjacent Amazonian forest of Colombia and

Brazil, overlapping with the range of morphospecies
1. Specimens of this morphospecies differ significantly from those of C. guianensis s.s. in having
leaflets with rounded to emarginated apices and
brown rusty indumentum on their midribs, distinctly
pedicellate flowers and six-ovulate locules, and seeds
with a short hilum (Fig. 5). This combination of characters is unique in the genus and suggests that
specimens of this morphospecies belong to an undescribed species.
Specimens of morphospecies 7 are restricted to
northern Venezuela, above 800 m altitude (Fig. 14),
and resemble those of C. guianensis s.s. in having
leaflets with an acuminate apex and ovaries with
four-ovulate locules. However there are notable differences distinguishing them, specifically, the leaflets
are always fewer (only four or five pairs vs. six to nine
pairs in C. guianensis s.s.), the flowers are distinctly
pedicellate and, more importantly, the seeds are
smooth and have much more reduced hila (Fig. 5).
Based on these differences in morphology and habitat,
specimens of morphospecies 7 are considered as

belonging to a separate undescribed species.
Morphospecies 10 and 11 have pentamerous flowers
and were placed by Styles (1981) in his C. procera
complex, making it a widespread species with a transAtlantic distribution. If this broad taxonomic view is
accepted, all the morphospecies occurring in Africa
will also have to be sunk into a single species.
However, because African morphospecies formed wellsupported clades in the ITS tree, and also have distinctive morphological features, it seems best to
consider the two morphospecies 10 and 11 as distinct
from their supposed African congeners. Morphospecies 10 and 11 formed a composite cloud separate
from the remainder of the morphospecies in morphometric analyses and had few differences in floral
characters. Tukeys post hoc analysis showed significant differences for CALD, NECH, SEPL and
STATLL. Quantitative leaf characters do not allow
separation of the two entities, except for MLPET and
the ratio BLL/TLL, although the tertiary venation is
always loose and flat in morphospecies 10, but dense
and prominent in morphospecies 11. The seeds of the
two morphospecies are, however, different. Tukeys
post hoc analysis showed significant differences for all
but one of the seed characters (SEEW) between the
two morphospecies, which are also separated on the
ITS tree, albeit by a rather low maximum likelihood
bootstrap value of 60%. Morphospecies 10 is endemic
to central Guyana, where it is found generally on
terra firme, whereas morphospecies 11 is more widespread, ranging from French Guiana to Guyana and
northern Brazil. The two morphospecies grow in sympatry at Iwokrama (central Guyana) were they can be
distinguished from each other, which suggests that
they can be considered as belonging to separate
species. Morphospecies 10 presents a novel combination of characters and corresponds to C. akuri Poncy,
Forget & Kenfack (Forget et al., 2009). Specimens of
morphospecies 11 matched perfectly the type of C.
surinamensis and, following Noamesi (1958), C. surinamensis is recognized in this study as separate from
C. procera.
The molecular clade II comprised five morphospecies, all restricted to the trans-Andean region, the
area west of the Andes from Nicaragua to Ecuador
(Fig. 15). Morphospecies 9 from the highlands of
Colombia was not sampled for the molecular analysis,
but was included in the morphometric analysis. Morphospecies 2, 4, 5 and 6, all with tetramerous flowers,
were considered by previous authors to belong to C.
guianensis, but they appeared well separated from
one another in the morphospace of the PCoA (Fig. 10).
Furthermore, in the ITS tree, they all belonged to the
well-supported clade II, whereas the rest of the presumed conspecific members (morphospecies 1, 3 and
205
Figure 10. Scatter plot of axes 1 and 2 of the principal coordinate analysis of specimens of Carapa from the Trans-Andes
region with overlay of the six morphospecies of the clade II. See Table 6 for the loadings of the different characters used.
7) were placed in clade I. These morphological, biogeographical and phylogenetic differences suggest
that these morphospecies (2, 4, 5 and 6) can be
considered as distinct species from C. guianensis s.s.
Of the five morphospecies of clade II, morphospecies 2, restricted to Panama, has the narrowest distribution range (Fig. 15). It can be distinguished from
the remaining morphospecies of clade II, based on
their coriaceous leaflets with a rounded to emarginated apex and their distinctly pedicellate flowers with
four-ovulate locules. In the field, individuals belonging to morphospecies 2 are readily distinguished from
those of morphospecies 4 that occur in the same area.
The first is a tall tree branched near its top, the
branches spreading upwards, whereas the second is a
small tree of marshy areas branching low and often
with the branches arching downwards. Specimens of
morphospecies 2 do not match any existing type and
are to be described as a new species.
Morphospecies 4 is the second most widespread of
those in the New World, occurring from Nicaragua
to Panama and along the Pacific coast of Colombia
and Ecuador (Fig. 15). It occurs mostly at low altitudes (< 300 m) and favours marshy conditions. Its
range overlaps with those of morphospecies 1, 5, 6,
and 8. In the herbarium, specimens of morphospecies 4 are readily distinguished on the basis of their
numerous prominent secondary veins, the presence
of a rusty indumentum on the lower surface of
leaflet blades, inflorescence branches and pedicels.

More importantly, this entity is the only representative in the New World to have bi-ovulate locules.
Specimens assigned to morphospecies 4 matched
perfectly the type of C. nicaraguensis and were
assigned to this species. Carapa nicaraguensis
should no longer be considered as a synonym of C.
guianensis s.s.
Morphospecies 5 is limited to Ecuador, where it
occurs on the western slopes of the Andes in both
primary and disturbed forests, generally above 600 m
elevation. Individuals of morphospecies 5 were found
growing a few meters from those of morphospecies 8
in the Reserva Rio Silanche. These specimens of morphospecies 5 are close to those of C. guianensis s.s. in
having subsessile flowers with a whitish to yellowish
disk. However, their leaflets have prominent secondary veins (vs. obscure in C. guianensis s.s.), the fruit
is beaked and the seeds have rounded edges and
shorter hila. This set of characters formed the basis
for a new species being described as C. alticola
Kenfack & A.J.Prez (Kenfack & Prez, 2011).
Morphospecies 6, also only known from Ecuador,
favours lower altitudes. It was found in sympatry
with morphospecies 8 in the Reserva Ethnica Aw.
Specimens of morphospecies 6 differ greatly from the
material of C. guianensis s.s. in having long pedicels
and six-ovulate locules. They also have the longest
petals of all New World species and have being
206
D. KENFACK
0.30
0.20
0.10
0.00
0.10
0.20
0.30
Clade I
Clade II
not sequenced
0.40
0.40 0.30 0.20 0.10 0.00
0.10
0.20
0.30
0.40
0.50
0.30
0.40
0.50
0.30
0.20
0.10
0.00
0.10
0.20
0.30
Clade I
Clade II
not sequenced
0.40
0.40 0.30 0.20 0.10
0.00
0.10
0.20
Figure 11. Scatter plots of the principal coordinate analysis of Carapa specimens from Africa. AB, all specimens with
overlay of the two major clades III and IV. CD, all specimens with overlay of the 16 morphospecies. See Table 7 for the
loadings of the different characters used.
0.40
0.30
12
16
20
24
13
17
21
25
14
18
22
27
15
19
23
27
0.20
0.10
0.00
0.10
0.20
0.30
0.40
0.40 0.30 0.20 0.10
0.00
0.10
0.20
0.30
0.40
0.50
0.30
0.20
0.10
0.00
0.10
0.20
12
15
18
21
24
27
0.30
0.40
0.40 0.30 0.20 0.10
0.00
0.10
0.20
0.30
13
16
19
22
25
14
17
20
23
26
0.40
Figure 11. Continued
0.50
207
208
D. KENFACK

coordinate analysis (PCoA) of all Carapa specimens from
Africa, eigenvalues, percentage of variance and cumulative
percentage of variance explained by the first three axes
Variables
PCoA1
PCoA2
PCoA3
ANTW
STID
OVUN
OVAW
NECH
ANTL
NECD
STATD
OVAL
STATLL
PEDL
PEDIND
PETW
CALIND
PETL
CALD
SEPL
STATL
STYL
PISL
Eigenvalue
of variance
0.3340
0.3171
0.2463
0.2358
0.1977
0.1886
0.1285
0.1184
0.0467
0.0396
0.0373
0.0360
-0.0715
-0.0745
-0.2162
-0.2351
-0.2671
-0.3419
-0.3524
-0.3673
6.19
33.4732
33.4732
-0.0442
-0.1278
0.4234
-0.1179
0.1652
-0.0389
-0.2830
-0.2187
0.2798
-0.0601
0.1902
0.1835
-0.3147
0.1654
-0.2038
-0.3954
-0.1445
0.0227
0.2569
0.2618
4.1249
22.306
55.7793
-0.2695
0.0970
-0.0402
0.2830
-0.1962
-0.3473
0.3155
0.5042
-0.1689
-0.0911
0.2119
0.0930
-0.2917
0.0145
-0.2056
-0.0294
-0.2302
0.0787
0.0722
0.2002
2.92
15.79
71.5692
described as a new species, C. longipetala Kenfack

(Kenfack & Prez, 2011).
The two cauliflorous morphospecies (8 and 9) formed
a composite group separate from the others in the plots
resulting from the PCoA of trans-Andean specimens.
Although barely separated from each other in the
analysis of all specimens of clade II, these two morphospecies formed distinct groups in a separate analysis limited to the two (not shown). Specimens of
morphospecies 8 correspond to the type of C. megistocarpa. Individuals of this species are easily distinguished in the field based on having their flowers borne
on the main trunk. The seeds of C. megistocarpa are
unique in having a whitish and pitted seed coat.
Morphospecies 9 also comprises cauliflorous individuals, which differ from morphospecies 8 in having longer
sepals, stout woolly pedicels and eight-ovulate ovary
locules. In addition, morphospecies 9 occurs at higher
elevations, up to 2300 m. They are considered here as
a new species separate from C. megistocarapa.
The two African clades III and IV revealed in the
ITS tree together comprised 16 morphospecies, all
previously considered as variants of C. procera

coordinate analysis (PCoA) of Carapa specimens from
Africa belonging to clade III of the internal transcribed
spacer (ITS) topology, eigenvalues, percentage of variance
and cumulative percentage of variance explained by the
first three axes
Variables
PCoA1
PCoA2
PCoA3
STATLL
CALD
SEPL
ANTL
STATD
ANTW
OVAW
NECD
PETW
PEDL
STID
PETL
MERO
STATL
STYL
NECH
PISL
OVAL
PEDIND
OVUN
Eigenvalue
of variance
0.2964
0.2763
0.2557
0.2216
0.2029
0.1921
0.1725
0.1211
0.0886
0.0414
-0.0026
-0.0267
-0.0983
-0.0984
-0.1155
-0.1846
-0.1928
-0.2787
-0.3179
-0.5531
4.3945
56.6603
56.6603
0.3120
-0.1201
0.0184
-0.0780
0.3048
-0.2339
0.1019
-0.1376
-0.4507
0.3302
-0.1664
-0.3346
0.1644
0.0484
0.2980
-0.2406
0.1373
-0.1202
0.2056
-0.0389
1.377
17.7547
74.415
0.3164
-0.0070
-0.0680
-0.1139
0.0768
0.0890
-0.0244
-0.0702
0.0689
0.1287
0.1579
-0.1015
0.2778
-0.3096
-0.4443
0.1053
-0.4804
-0.1696
0.2396
0.3285
0.7228
9.3188
83.7337
(Styles & White, 1991). Clade III included two wellsupported subclades and six morphospecies, all from
the Central African forest block (Fig. 16).
Morphospecies 16, 19, 23 and 25 formed a wellsupported clade and also formed discrete clusters in
the PCoA morphospace. With the exception of specimens assigned to morphospecies 19, an entity
endemic to the Eastern Africa montane forest (in
Burundi, Rwanda and Uganda), the ranges of these
morphospecies largely overlap.
Morphospecies 16 has a narrow distribution in
lowland humid forest of southern Cameroon, an area
where morphospecies 23 and 25 also occur. It comprises understory treelets growing to only 3 m tall,
whereas morphospecies 23 and 25 are trees to 20 m
tall. The leaves of morphospecies 16 are only three- or
four-jugate, the leaflets being distinctly cuspidate at
the apex, whereas those of the two other morphospecies have up to 22 pairs of leaflets, with rounded to
acuminate apices. Moreover, specimens of morphospecies 16 have bi-ovulate locules, whereas morphospe-
209
Figure 12. Scatter plots of the principal coordinate analysis of 52 specimens of Carapa from the central African forest
block belonging to clade III of the internal transcribed spacer (ITS) topology. A, axes 1 and 2. B, axes 1 and 3. See Table 8
for character loadings.
cies 23 and 25 have four and six ovules, respectively.

These morphological features of morphospecies 16
also differ drastically from those of C. procera. In fact,
specimens of morphospecies 16 matched perfectly the
type of C. batesii, a species described from southern
Cameroon in 1917 and previously placed by as

synonym under C. procera. Following Noamesi (1958),
C. batesii is considered here as a separate species
from C. procera s.s. (see below) and from morphospecies 23 and 25.
210
D. KENFACK
Figure 13. Scatter plots of the principal coordinate analysis of specimens of Carapa from Africa belonging to the clade
IV of the internal transcribed spacer (ITS) topology. A, all specimens with overlay of the 10 morphospecies. B, subset of
specimens from West Africa. C, subset of specimens from central Africa. D, subset of a composite group of five
morphospecies from Central Africa. See Table 9 for character loadings.
Figure 13. Continued
211
212
D. KENFACK

coordinate analysis (PCoA) 108 specimens of Carapa from
Africa belonging to clade IV, eigenvalues, percentage of
variance and cumulative percentage of variance explained
by the first three axes

coordinate analysis (PCoA) of three specimens groups of
Carapa in West Africa, eigenvalues, percentage of variance
and cumulative percentage of variance explained by the
first three axes
Variables
PCoA1
PCoA2
PCoA3
Variables
PCoA1
PCoA2
PCoA3
PEDL
SEPL
CALD
PETL
PETW
STATD
STID
STATL
STATLL
ANTL
ANTW
NECH
NECD
STYL
OVAL
PISL
OVAW
OVUN
PEDIND
Eigenvalue
of variance
0.60467
-0.76507
0.52578
1.88102
0.58930
0.26560
-0.61538
1.44072
-1.04728
-0.96648
-1.21717
-0.78089
0.25741
-0.55421
-0.83612
0.16540
-0.90895
1.45100
0.51065
2.61446
30.91000
30.91000
0.25132
0.02516
0.16590
0.08867
0.07424
0.18351
-0.02243
0.08128
0.01937
-0.01362
-0.03399
-0.06634
0.09845
-0.03253
-0.07738
-0.03270
0.02852
-0.42149
-0.31593
2.03721
24.08530
54.99530
0.44919
-0.02055
-0.08148
-0.16028
-0.22492
-0.00473
-0.01292
-0.00846
-0.02863
-0.03127
-0.02765
-0.01977
-0.06804
0.02968
-0.02046
0.05027
-0.02125
-0.02957
0.23085
1.16258
13.74480
68.74020
PEDL
STATLL
STATD
OVAW
STID
NECH
NECD
STATL
PETL
ANTL
CALD
SEPL
OVAL
PEDIND
PETW
ANTW
LIND
PISL
STYL
Eigenvalue
of variance
0.3718
0.3543
0.3383
0.2639
0.2140
0.1877
0.0545
-0.0170
-0.0323
-0.0413
-0.0605
-0.0607
-0.0944
-0.1490
-0.1566
-0.1648
-0.1806
-0.4066
-0.4207
1.3525
35.57
35.57
0.3474
0.3363
-0.0013
-0.2390
-0.0312
-0.2576
-0.3307
-0.1485
-0.1778
0.1049
0.1566
0.1822
-0.2829
0.3537
-0.2797
-0.1704
0.1796
0.0026
0.2557
1.0553
27.76
63.33
0.0301
0.1996
-0.1792
-0.0697
-0.0528
0.2833
0.0836
-0.0260
0.0231
-0.3263
-0.4258
-0.4025
0.2047
0.5022
-0.1070
0.0428
0.2680
-0.0032
-0.0448
0.322
8.47
71.80
Specimens of morphospecies 19 are restricted to

east African mountains (Fig. 16). They are trees to
25 m tall, generally with a short fluted trunk. This
distinguishes them from individuals of morphospecies
16, which are understory treelets. Furthermore,
although the two morphospecies have an overlapping
number of leaflets, those of morphospecies 19 have an
apex that is broadly acuminate (vs. cuspidate in morphospecies 16). Their seeds also differ. Those of morphospecies 19 are larger (2.54.5 vs. 1.52.0 cm in
morphospecies 16) and have a smooth seed coat and
well-developed hilum. Based on these differences in
morphology, growth form and habitat preferences,
morphospecies 19 is considered here to be a separate
species from C. batesii. Morphospecies 19 also differs
from morphospecies 23 and 25 in having fewer leaflets (four to seven vs. nine to 22 pairs) and, more
importantly, the two latter have four- and six-ovulate
locules, respectively, and their seeds are morphologically different from those of morphospecies 19 (Fig. 5).
Based on these features, morphospecies 19 is considered as a separate species from morphospecies 23 and
25. Specimens of morphospecies 19 are assigned here
to C. grandiflora based on their habitat, the number

of ovules per locule and their overall similarity with
the type specimen.
Morphospecies 23 and 25 comprised specimens
widely distributed in Central African lowland forest,
where they are often found in close proximity growing
on terra firme. Morphospecies 23 is generally a wellbranched tree to 20 m tall, whereas morphospecies 25
is often monocaulous or few-branched. Both entities
have numerous extrafloral nectaries ( 35) at the
base of their petiole, which clearly distinguish them
from the rest of the genus with no more than seven
nectaries. The most distinctive features separating
these two morphospecies are floral and seed characters. The flowers in morphospecies 23 have a farinose
indumentum on the inflorescence branches and
pedicels, pink petals (Fig. 3) and four-ovulate locules,
whereas, in morphospecies 25, they are glabrous,
have purple to dark red petals (Fig. 3) and six-ovulate
locules. In addition, the seeds of morphospecies 23
have an elongated hilum whereas those of morphospecies 25 have a rounded hilum (Fig. 5). Because
these two morphospecies present such strong morpho-
coordinate analysis (PCoA) of seven specimens groups of
Carapa of clade IV from Central Africa, eigenvalues, percentage of variance and cumulative percentage of variance
explained by the first three axes
Variables
PCoA1
PCoA2
PCoA3
OVUN
OVAL
PEDIND
PISL
STYL
NECH
STATL
MERO
PEDL
PETL
STID
NECD
STATD
OVAW
PETW
ANTL
SEPL
CALD
ANTW
STATLL
Eigenvalue
of variance
0.4821
0.3117
0.2906
0.2596
0.1749
0.1621
0.1601
0.0535
0.0240
-0.0139
-0.0253
-0.1247
-0.1403
-0.1513
-0.1684
-0.2100
-0.2198
-0.2544
-0.2758
-0.3347
1.7837
45.8165
45.8165
0.0885
-0.1920
0.2069
-0.0094
0.1697
-0.2626
-0.0524
0.2761
0.1118
-0.3483
-0.1532
-0.0857
0.4918
0.1432
-0.3627
-0.1359
-0.0004
-0.1151
-0.1292
0.3588
0.5764
14.8054
60.6219
0.0464
-0.4102
-0.1283
-0.0140
-0.0482
0.2612
0.0683
-0.2487
0.2831
-0.0591
0.4151
0.3229
0.3597
-0.0781
-0.1156
-0.2514
-0.0814
-0.0184
0.0072
-0.3105
0.4435
11.3923
72.0143
logical differences, even in sympatry, they are considered here to belong to two separate species. The
comparison of their morphological characteristics to
the protologues of validly published species and type
specimens seen, showed that material of morphospecies 23 more or less matches C. parviflora Harms and
specimens of morphospecies 25 match C. macrantha
Harms, both described from central Africa. The name
C. parviflora was applied to specimens of morphospecies 25 because they have small flowers and despite
one of their main characteristic features, the farinose
indumentum of the inflorescence branches and
pedicel, not mentioned in the protologue of C. parviflora. The name C. macrantha was assigned to specimens of morphospecies 23 which has 1022 pairs of
leaflets, despite the protologue of the C. macrantha
mentioning only four to 11 pairs.
The second subclade of the African clade III comprised morphospecies 15 and 27, both restricted to the
rainforest of Cameroon, Equatorial Guinea and
Gabon (Fig. 16). Specimens assigned to the former all
came from submontane cloud forest (8001200 m
elevation). They are tall trees to 35 m, their leaves
213
Table 12. Summary of taxonomic decisions in the present

study
Morphospecies
#
Molecular
clade
Assigned names in this

study
1
2
3
4
5
American
American
American
American
American
I
II
I
II
II
American II
7
8
N/A
American II
9
10
N/A
American I
11
12
13
14
15
16
17
18
American I
African IV
African IV
African IV
African III
African III
African IV
N/A
19
20
21
22
23
24
25
26
African
African
African
African
African
African
African
African
27
African III
III
IV
IV
IV
III
IV
III
IV
C. guianensis Aubl. s.s.

Carapa sp. nov.
Carapa sp. nov.
C. nicaraguensis C.DC.
C. alticola Kenfack &
A.J.Prez sp. nov.
C. longipetala Kenfack
sp. nov.
Carapa sp. nov.
C. megistocarpa
A.H.Gentry & Dodson
Carapa sp. nov.
C. akuri Poncy, Forget
& Kenfack sp. nov.
C. surinamensis Miq.
C. procera DC.
C. microcarpa A.Chev.
C. velutina C.DC.
C. angustifolia Harms
C. batesii C.DC.
C. dinklagei Harms
C. gogo A.Chev. nomen
nudum
C. grandiflora Sprague
Carapa sp. nov.
C. hygrophila Harms
Carapa sp. nov.
C. macrantha Harms
Carapa sp. nov.
C. parviflora Harms
C. procera var. palustre
G.C.C.Gilbert
Carapa sp. nov.
N/A, not available.
have up to 16 pairs of narrow leaflets with fine

tertiary venation impressed in the lower surface, the
flowers have four-ovulate locules, the seeds are small
compared with those of the other species that occur at
high elevations, and also have a well-developed
hilum. Morphospecies 27 comprises specimens disjunctly distributed in lowland forest from southeastern Nigeria and adjacent south-west Cameroon to
Gabon and probably Equatorial Guinea. These are
small straggling trees to 12 m tall, with branches
arching downwards. They differ from members of
morphospecies 15 in having leaves with only four or
five pairs of obovate leaflets (Fig. 2) and farinose
flowers with only bi-ovulate locules. The two morphospecies can be considered as separate species
214
D. KENFACK
Figure 14. Distribution of the five morphospecies of Carapa from the Cis-Andes region of the New World, four of which
formed clade I of the internal transcribed spacer (ITS) topology.
based on their habitat specialization and the morphological differences. In the transitional zone between
submontane and montane forest, specimens of morphospecies 15 matched the protologue of C. angustifolia Harms, one of the five species described from
Cameroon (Harms, 1917), whereas those of morphospecies 27 had a novel combination of characters
that indicate that they belong to a new species.
The African clade IV is the most species rich of the
four clades of the ITS tree, including nine morphospecies that span almost the entire distribution range of
the genus in Africa (Fig. 17). Specimens of morphospecies 18 from So Tome were included in the
PCoA of this clade as they originate from the same
area. The nine morphospecies formed three wellsupported subclades.
Subclade I included specimens from the two morphospecies 20 and 22. The former is restricted to
montane forest in Central Africa, where it often
occurs in sympatry with individuals of morphospecies
15 in the transitional zone between submontane and
montane forest, whereas the latter occurs in coastal
forest from sea level to only 200 m elevation. The two
members of subclade I also present differences in
growth habit: morphospecies 20 is a small tree to

12 m tall, branching low to form a dense crown,
whereas morphospecies 22 includes large trees to
35 m, branched near their top. Although both have
four-ovulate locules, their leaves and seeds present
major differences. The leaflets are in four to six pairs
in morphospecies 20 vs. seven to ten pairs in morphospecies 22, and the hilum is larger (1418 mm)
in morphospecies 20 than in morphospecies 22
(612 mm). Based on their habitat and morphological
differences, the two morphospecies are considered as
belonging to separate species. Specimens of morphospecies 20 were confused in the past with material of the east African species C. grandiflora,
probably as a result of the apparent convergence in
the morphology of their leaves and seeds. Because
they belong to two different well-supported clades,
and because morphospecies 20 has four-ovulate
locules vs. being bi-ovulate in C. grandiflora, they are
considered here as separate species. Botanists have
long been aware of the distinction between C. grandiflora and the central African morphospecies 20.
Indeed, Keay et al. (1964) already noted that the
criteria used to distinguish the East African C. gran-
215
Figure 15. Distribution of the six morphospecies of Carapa from the Trans-Andes region of the New World, five of which
formed clade II of the internal transcribed spacer (ITS) topology.
diflora (bi-ovulate locules and fruits with < 10 seeds)

from C. procera (s.l.) do not allow distinction of the
high elevation species from the Obudu and Bamenda
(in Nigeria and Cameroon, respectively) from the
latter. The combinations of the features exhibited by
morphospecies 20 and 22 suggest that they represent
new species.
Subclade II comprised the morphospecies 21, 24
and 26, which have overlapping ranges and are morphologically similar to one another. They formed a
composite cloud in the centre of the morphospace
(Fig. 13C). The floral characters used in the PCoA did
not distinguish among these entities. However, they
present other morphological, growth and habitat differences. Morphospecies 21 and 26 grow in wet
valleys in association with species of Raphia P.Beauv.,
whereas morphospecies 24 grows mostly along the
inland limits of mangroves in Gabon, often under
saline conditions. Morphospecies 21 and 24 are small
trees to 12 m tall, whereas morphospecies 26 grows to
30 m tall. Fruits are known for only two of the three
morphospecies. Those of morphospecies 21 are conical
and ribbed, whereas those of morphospecies 26 are
globose with irregular warty excrescences (Fig. 4).
Also, the leaflets of morphospecies 21 are coriaceous
and obovate, those of morphospecies 24 are coriaceous
and oblong and those of morphospecies 26 are

oblong and coriaceous. Based on the protologue, I
assigned the specimens of morphospecies 21 to C.
hygrophila Harms, a species described from southern
Cameroon. Specimens of morphospecies 26 matched
the type of C. procera var. palustre G.Gilbert,
described from Congo in 1958 and so far considered a
synonym of C. procera. A new combination raising
this variety to the rank of species will be published
elsewhere. Morphospecies 24 did not match any previously recognized species and will be therefore be
described as a new species.
The third subclade of the African clade IV comprised morphospecies 12, 13, 14 and 17. The first
three include specimens from West Africa (Senegal to
Togo; Fig. 17) and formed distinctive clusters in the
morphospace of a PCoA comprising only specimens
from this region. These three morphospecies also
formed a well-supported clade in the ITS tree.
Morphospecies 12 and 14 are similar, based on
quantitative leaf and floral characters, and also have
four-ovulate locules. However, they can be easily distinguished based on their indumentum. Morphospecies 12 comprises glabrous individuals, whereas the
leaf rachis, the petiolules, the inflorescence branches
and all parts of the flowers in morphospecies 14 are
216
D. KENFACK
Figure 16. Distribution of the six morphospecies of Carapa from Central Africa, which formed clade III of the internal
transcribed spacer (ITS) topology.
densely pubescent. Their ranges largely overlap and

they were observed growing in mixed populations at
Thionk-Essyl in Senegal. Because these two morphospecies present obvious morphological differences
even when growing in close proximity, they can be
considered as separate species. Specimens of morphospecies 13 were assigned to C. velutina, based on
the similarity with the type specimen collected from
Guinea. Specimens of morphospecies 14 matched the
photograph of the presumed type of C. procera and
are referred to as C. procera s.s., an assignment based
on the lack of indumentum on all parts of the trees,
the shape (oblanceolate to oblong) of the leaflets and
as indicated in the protologue, less than six pairs of
leaflets. A more detailed explanation as to why the
name C. procera should not be applied to the New
World pentamerous species of Carapa is found in
Noamesi (1958).
Morphospecies 13 occurs from Western Ivory
Coast eastwards to Benin (Fig. 17) and, in contrast
to the morphospecies 12 and 14, which grow in
gallery forest in savanna or dry forest, it is found
primarily in periodically inundated terrains and
along streams in wet forests. Morphospecies 13
has densely puberulent rachises, petiolules, leaflet
midribs, inflorescence branches, bracts, pedicels,
sepals and petals, but it can easily be separated

from morphospecies 14 (and 12) in having smaller
fruits and small seeds with tiny hila (Fig. 5). Based
on these differences and the similarity with the type
of C. microcarpa A.Chev., specimens of morphospecies 13 were assigned to this species described from
Ivory Coast in 1909.
Morphospecies 17 is endemic to a coastal forest belt
from south-western Nigeria to Gabon, where it grows
in wet valleys and along streams. It differs from the
other members of its clade (morphospecies 12, 13 and
14) in having six-ovulate locules and fruits with warty
excrescences (Fig. 4). It is found in sympatry with
morphospecies 25 within the 50-ha permanent plot in
Korup National Park in Cameroon (Thomas et al.,
2003; Kenfack et al., 2007) and with morphospecies
23 and 27 elsewhere in the park (D. Kenfack, pers.
observ.). On the basis of these observations, morphospecies 17 is considered here to represent a separate species from all of the others mentioned above.
Based on comparisons with the protologues of existing
names and the type specimens seen, morphospecies
17 corresponds to C. dinklagei Harms.
Morphospecies 18 was embedded within morphospecies 21 in the multivariate analysis of the central
African composite group. Specimens of this entity are
217
Figure 17. Distribution of the nine morphospecies of Carapa from Central and West Africa, which formed clade IV of the
internal transcribed spacer (ITS) topology.
endemic to So Tom e Prncipe. Although also placed

as a synonym under C. procera by previous authors,
they are large trees with a long bole (> 1 m in diameter)
which branches high up (O. Hardy, pers. comm.),
whereas C. procera s.s. comprises small trees with
short trunk and low branches. Also, the material of
morphospecies 18 is unique within Africa in having
predominantly tetramerous flowers and small shiny
seeds, which matched historical specimens identified
as C. gogo A.Chev. (nomen nudum), and is to be
formally described under this name.
In conclusion, the PCoA of staminate flowers
showed 23 more or less discontinuous clusters and
two composite morphospecies (8 + 9 and 10 + 11),
which are further separated from one another based
on leaf or seed characters. Most of the morphospecies
have overlapping distribution ranges and some
instances of strict sympatry involving some entities
have been reported. The variation observed appears
to be sufficient to recognize the 27 morphological
entities as separate species. The distinction of these
species is further supported by the variation in habit,
flower and seed characters that have not been used
previously for species delimitation in Carapa. This
result is consistent with geography, habitat and the
results of the molecular analyses.
By and large, the patterns of morphological variation revealed in this study are consistent with the
delimitation of many species recognized in the past,
but placed in synonymy by recent workers (Table 12).
Indeed, 15 of the 27 morphological entities recognized
here can be assigned to previously described species,
based on comparisons with their protologues and/or
the type specimens (Table 12), whereas the 12 morphospecies remaining exhibit novel combinations of
characters that suggest that they represent undescribed taxa.
IMPORTANT
DISCRIMINATING CHARACTERS
This study shows the importance of staminate flowers

in recognizing morphologically distinct species within
Carapa. Compared with leaf characters, floral features are often less phenotypically plastic (Wilken,
1977; Cresswell, Hagen & Woolnough, 2001; Dorken
& Barrett, 2004) and can be more reliable when
comparing specimens from distant geographical
areas. According to previous authors who have
studied Carapa (Styles, 1981; Styles & White, 1991),
floral characters such as pedicel length, presence or
absence of pubescence and the number of ovules per
locule are too variable to be useful discriminators
218
D. KENFACK
and occur sporadically throughout the range of the

African and New World species (Styles, 1981: 418).
But the present study shows that these traits are
often highly correlated and, when considered in combination, they can be of taxonomic value. The most
important quantitative floral characters appear to be
petal length, pedicel length, staminal tube length and
number of ovules. Other floral features not considered
in the present analyses also separate the morphospecies. Although most entities studied had greenish
petals, four (morphospecies 17, 23, 24 and 25) had
unique petal colours that can be used for their easy
identification (Fig. 3). Trichomes are rare in Carapa
and their presence appears to be of potential taxonomic importance. Only five of the 27 morphospecies
recognized had an indumentum on their leaves,
whereas eight had puberulous to farinose flowers.
Although leaves were not as diagnostic as flowers, the
combination of the number of leaflets and the length
of the petiole is of apparent taxonomic value, especially for African material (Fig. 2). Unfortunately,
because of the large size of the leaves in some morphospecies, these characters are rarely observed in
herbarium specimens. Features of leaflet apex are
important in distinguishing between many pairs of
morphospecies. For example, two groups could be
distinguished based on the presence or absence of an
acute apex, and the longcuspidate apex is unique to
morphospecies 16.
Previous authors have relied heavily on the number
of seeds per locule to justify that the character was
continuous among different variants of Carapa.
Indeed, seed number is unreliable because of the
abortion of some ovules or the compression of some
seeds during fruit development. However, fruit colour
and ornamentation, seed coat ornamentations and
the length of the hilum relative to the length of the
seed are important in discriminating among morphospecies.
Carapa is just one of the many examples of plant
genera with species that present obvious and substantial differences in their field characters (e.g. habit and
flower and fruit colour), but more subtle differences in
herbarium specimens. Previous taxonomic treatments
of the genus (Noamesi, 1958; Styles, 1981) were solely
based on herbarium specimens. Herbarium diagnostic
differences among morphospecies were either not
found by these monographers, for example because of
the fragmentary collections, or they were, but, considered insufficient for recognizing separate entities. As
stated by White (1981), at the same time that it is
worthy undertaking taxonomic revisions at globalscale, it is almost impossible that the monographer will
also have a worldwide field experience in his group to
document such critical [field] differences. In the
present investigation, 23 of the 27 morphospecies of
Carapa recognized in this study were examined and

collected in the field, which was instrumental in discriminating among many of them.
ACKNOWLEDGEMENTS
I am grateful to the Kellogg Lab for laboratory assistance; Pierre-Michel Forget, Sainge Moses, Rosa Ortiz
and Fidy Ratovozon, for collection; David Neil,
Gretchen Walters, Kayombo, Ludovic Ngok Banak,
Mathieu Gueye, Renato Valencia, Richard Condit and
Yves Issembe for their help during fieldwork and for
helping with exportation permits; Pete Lowry, Peter
Stevens and Pierre Michel Forget for their useful
comments on the original manuscript. Financial
support for this study was provided by the National
Geographic Society, the Center for Tropical Forest
Sciences, the Whitney R. Harris World Ecology
Center, the Missouri Botanical Garden and Idea Wild.
REFERENCES
Adjanohoun E, Ahyi MRA. 1985. Contribution aux tudes
ethnobotaniques et floristiques au Mali. Paris: Agence de
Cooperation Culturelle et Technique, (A.C.C.T.).
Adjanohoun E, Ake Assi L. 1979. Contribution au recensement des plantes mdicinales de Cte dIvoire. Abidjan:
Centre National de Floristique. Universit dAbidjan.
Baldwin BG. 1993. Molecular phylogenetics of Calycadenia
(Compositae) based on ITS sequences of nuclear ribosomal
DNA: chromosomal and morphological evolution reexamined. American Journal of Botany 80: 222238.
Bouquet A. 1969. Fticheurs et mdecines traditionnelles du
Congo (Brazzaville). Paris: Mmoire O.R.S.T.O.M No 6.
de Candolle C. 1878. Meliaceae. In De Candolle A & De
Candolle C, eds. Monographiae phanegamarum. Paris:
A Masson. 1: 389758.
Chae SS, Warde WD. 2006. Effect of using principal coordinate analysis and principal components on retrieval of
clusters. Computational Statistics and Data Analysis 50:
14071417.
Cresswell JE, Hagen C, Woolnough JM. 2001. Attributes
of individual flowers of Brassica napus L. are affected by
defoliation but not by intra-specific competition. Annals of
Botany 88: 111117.
Dick CW, Abdul-Salim K, Bermingham E. 2003. Molecular
systematic analysis reveals cryptic tertiary diversification of
a widespread tropical rain forest tree. The American Naturalist 162: 691703.
Dorken ME, Barrett SCH. 2004. Phenotypic plasticity of
vegetative and reproductive traits in monoecious and dioecious populations of Sagittaria latifolia (Alismataceae): a
clonal aquatic plant. Journal of Ecology 92: 3244.
Forget PM, Poncy O, Thomas RS, Hammond DS,
Kenfack D. 2009. A cryptic new species of Carapa Aublet
(Meliaceae) from Central Guyana. Brittonia 64: 366374.
Gentry AH, Dodson CH. 1988. New species and a new
combination for plants from trans-Andean South America.
Annals of the Missouri Botanical Garden 75: 1434
1436.
Gower JC. 1966. Some distance properties of latent root and
vector methods used in multivariate analysis. Biometrika
53: 325338.
Gueye M, Kenfack D, Forget PM. 2010. Importance socioculturelle, potentialits conomiques et thrapeutiques du
Carapa (Meliaceae) au Sngal. In: Van der Burgt X, van
der Maesen J, Onana J-M, eds. Systematics and conservation of African plants, Proceedings of the 18th AETFAT
Congress, Yaounde, Cameroon: 359367.
Harms H. 1896. Meliaceae. In: Engler A, Prantl K, eds. Die
natrlichen Pflanzenfamilien 3. Leipzig: W. Engelmann,
258308.
Harms H. 1917. Meliaceae africanae. Notizblatt des Knigl.
botanischen Gartens und Museums zu Berlin 7: 223
232.
Harms H. 1940. Meliaceae. In: Engler A, Prantl K, eds. Die
natrlichen Pflanzenfamilien. ed. 2, 19b 1. Leipzig: W.
Engelmann, 1172.
Hebert PDN, Penton EH, Burns JM, Janzen DH, Hallwachs W. 2004. Ten species in one: DNA barcoding
reveals cryptic species in the Neotropical skipper butterfly
Astraptes fulgerator. Proceedings of the National Academy
of Science of the United States of America 101: 14812
14817.
Henderson A. 2006. Traditional morphometrics in plant systematic and its role in palm systematics. Botanical Journal
of the Linnean Society 151: 103111.
Holmgren PK, Holmgren NH, Barnett LC. 1990. Index
herbariorum. Part I: the herbaria of the world, 8th edn. New
York: New York Botanical Garden.
Hutchinson J, Dalziel JM. 1958. Flora of West Tropical
Africa (2nd edition, revised by R. W. J. Keay), Vol. 1(2).
London: Crown Agents for Overseas Governments and
Administrations.
Keay RWJ. 1989. Trees of Nigeria. Oxford: Clarendon
Press.
Keay RWJ, Onoche CFA, Stanfield DP. 1964. Nigerian
trees. Ibadan: Federal Department of Forest Research.
Kenfack D, Prez A. 2011. Two new species of Carapa
(Meliaceae) from Western Ecuador. Systematic Botany. In
press.
Kenfack D, Thomas DW, Chuyong GB, Condit R. 2007.
Rarity and abundance in a diverse African forest. Biodiversity Conservation 16: 20452074.
Kerharo J, Adam JG. 1974. La pharmacope sngalaise
traditionnelle. Plantes mdicinales et toxiques. Paris: Editions Vigot Frres.
Kribs DA. 1930. Comparative anatomy of the woods of the
Meliaceae. American Journal of Botany 17: 724738.
Lamarck JBAP. 1785. Encyclopdie mthodique 1: 621.
Legendre P, Legendre L. 1998. Numerical ecology, 2nd edn.
Amsterdam: Elsevier.
Martnez-Ortega MM, Luis Delgado L, Dirk C, Albach
DC, Juana A, Elena-Rossell JA, Rico E. 2004. Species
boundaries and phylogeographic patterns in cryptic taxa
219
inferred from AFLP markers: Veronica subgen. Pentasepalae

(Scrophulariaceae) in the western Mediterranean. Systematic Botany 29: 965986.
Muellner AN, Rosabelle S, Johnson SA, Cheek M, Pennington TD, Chase MW. 2003. Molecular phylogenetics of
Meliaceae (Sapindales) based on nuclear and plastid DNA
Sequences. American Journal of Botany 90: 471480.
Muellner AN, Pennington TD, Chase MW. 2009. Molecular phylogenetics of Neotropical Cedreleae (mahogany
family, Meliaceae) based on nuclear and plastid DNA
sequences reveal multiple origins of Cedrela odorata.
Molecular Phylogenetics and Evolution 52: 461469.
Noamesi GK. 1958. A revision of Xylocarpae (Meliaceae). PhD
Thesis, University of Wisconsin.
Pennington TD, Styles BT. 1975. A generic monograph of
Meliaceae. Blumea 22: 419540.
Posada D, Crandall KA. 1998. Modeltest: testing the model
of DNA substitution. Bioinformatics 14: 817818.
Posada D, Crandall KA. 2001. Selecting the best-fit model of
nucleotide substitution. Systematic Biolology 50: 580601.
Rohlf FG. 1972. An empirical ordination of three ordination
techniques in numerical taxonomy. Systematic Zoology 21:
271280.
Rohlf FJ. 1998. NTSYSpc: numerical taxonomy and multivariate system, ver. 2.02f. Setauket: Exeter Publishing Ltd.
Schultes RE, Raffauf RF. 1990. The healing forest. Portland: Dioscorides Press.
Sites JW Jr, Marshall JC. 2004. Operational criteria for
delimiting species. Annual Review of Ecology, Evolution and
Systematics 35: 199227.
Sneath PHA, Sokal RR. 1973. Numerical taxonomy the
principles and practice of numerical classifications. San
Francisco, CA: W. H. Freeman.
Staner P. 1941. Les Mliaces du Congo Belge. Bulletin du
Jardin Botanique dEtat 16: 114125.
Styles BT. 1981. Swietenioideae. In Pennington TD, Styles
BT, Taylor DAH. Meliaceae. Flora neotropica, Monograph
28. New York: New York Botanical Garden Press, 359
420.
Styles BT, White F. 1991. Flora of Tropical East Africa:
Meliaceae. Polhill RM, ed. Rotterdam/Brookfield: Royal
Botanic Gardens, Kew.
Thomas DW, Kenfack D, Chuyong GB, Moses SN, Losos
E, Songwe NC. 2003. Tree species of southwestern Cameroon: tree distribution maps, diameter tables and species
documentation of the 50-ha Korup Forest Dynamics Plot.
Washington, DC: Center for Tropical Forest Science of the
Smithsonian Tropical Research Institute.
Tukey J. 1977. Exploratory data analysis. Reading: AddisonWesley Publishing Co.
White F. 1979. The Guineo-Congolian Region and its relationships to other phytochoria. Bulletin du Jardin Botanique National de Belgique 49: 1155.
White F. 1981. The history of the Afromontane archipelago
and the scientific need for its conservation. African Journal
of Ecology 19: 3354.
White TJ, Bruns T, Lee S, Taylor JW. 1990. Amplification
and direct sequencing of fungal ribosomal RNA genes for
220
D. KENFACK
phylogenetics. In: Innis MA, Gelfand DH, Sninsky JJ, White

TJ, eds. PCR protocols: a guide to methods and applications.
San Diego, CA: Academic Press, 315322.
Wilken DH. 1977. Local Differentiation for Phenotypic Plasticity in the Annual Collomia linearis (Polemoniaceae). Systematic Botany 2: 99108.
APPENDIX
LIST
OF SPECIMENS USED IN THE
MORPHOMETRIC ANALYSIS
Morphospecies 1: BRAZIL. B. Nelson 803; B.A.

Krukoff 1048; B.A. Krukoff 4776; B.A. Krukoff 8120;
B.V. Rabelo & J. Cardoso 2886; B.V. Rabelo, R.
Nonato & H.P. Belo 1865; B.V. Rabelo, S. Mori, D.
Daly, J.A. Cardos, M.G. da Silva, H. Belo, D. Campbell, C.S. Rosario & M.R. dos Santos 2349; D.C. Daly,
J. Reitsma, J. Cardoso, L.G. Mendez & L.C. Lima
3906; Duke 410; G.T. Pance & T.D. Pennington 1990;
G.T. Prance & T.D. Pennington 1255; G.T. Prance,
T.D. Pennington & N.T. Silva 1378; G.T. Prance, T.D.
Pennington & N.T. Silva 1644; J. Ramos 920; N.T.
Silva & U. Brazo 60674; R. Froes & B.A. Krukoff
11706; T.D. Pennington, G.T. Prance, J. Ramos & O
Monteiro 9964. COLOMBIA. H.P. Fushs & L. Zanella
21879; M.M. Amaya & R. Vasques 188. DOMINICA. J.
Higgins & P. Paris 127A; W.H. & B.H. Hodge 3332.
FRENCH GUIANA. Granville T.1093; Oldeman 1027.
GRENADA. P. Beard 1296; GUATEMALA. H. Johnson
1195; GUYANA. A.S. Hitchcock 17524; J.S. De la Cruz
4153; W. Halm & S. Tiwari 5199. HONDURAS. A.
Clewell & G. Cruz 4038. NICARAGUA. A. Grijalva 6.
PANAMA. C. Galdames, M. Correa, L. Dorr & J. Jimeneez 3764. PERU. A. Gentry, C. Diaz, J. Aronson & N.
Jaramillo 27498; J. Pipoly, R. Vsquez, N. Jaramillo,
C. Grndez, J. Ruz & R. Ortz 13353; R. Vsquez &
N. Jaramillo 2728; Vsquez & N. Jaramillo 12708. A.
Castillo 708; Bernardi 2123; E.L. Little Jr 17661; F.J.
Breteler 4947; F.J. Breteler 5063; J.A. Steyermark
87712; J.J. Wurdack 292.
Morphospecies 2: PANAMA. G. de Nevers, H.
Herrera & S. Charnley 4968; M.D.A. Correa & R.L.
Dressler 1115; R.L. Dressler 3368.
Morphospecies 3: BRAZIL. B. Boom, M. Pacheco, E.
Palheta & S. de Souza 8609; C.A. Cid et. 8536; C.A.
Cid Fereira, A.J. Henderson, A.O. Scariot & J.G. de
Olivera 9994; G.T. Prance & J.F. Ramos 23551;
M.A.D. de Souza & E. Pereira da C., Pinheiro, Z.A.;
Mesquita M.R. 425; P. Kukle 113; R. Vsquez & N.
Jaramillo 9243; T.D. Penington, G.T. Prance, J.
Ramos & O. Monteiro 9931. PERU. R. Vasquez & N.
Jaramillo 963; R. Vasquez & N. Jaramillo 4827.
Morphospecies 4: COLOMBIA. J. Espina, F. Garcia &
S. Pino 2882. COSTA RICA. A. Estrada, O. Valverde. L.
Gutierrez & H. Gomez 989; A.R. Molina, L.O. Will-
iams, W.C. Burger & B. Wallena 17674; B. Hammel,

R. & S. Hammel, I. & P. Rivera 18148; C. Kerman &
P. Phillips 893; Fco. Queseda & M.M. Chavarria 392;
J.M. Orozco 492; P.C. Stanley & J. Valerio 52493; P.H.
Allen 6718; W. Burger & G. Matta U. 4716. ECUADOR.
C. Jativa & C. Epling 1113. NICARAGUA. F.C. Englesing H-F572562; J.C. Sandino 4511; J.C. Sandino
4935; J.C. Sandino 4740; M. Nee 27837; P.P. Moreno
25494. PANAMA. D. Kenfack 2005; E.A. Lao, L.R.
Holdridge 239; Fabio Garca C. & Enzo D. Agualimpia
420; G.P. Cooper & G.M. Slater 59; I.M. Johnston
1824; W.L. Stern, K.L. Chambers, J.D. Dwyer & J.E.
Ebinger 969.
Morphospecies 5: ECUADOR. D. Kenfack 2155; D.
Kenfack 2155; D. Kenfack 2155; D. Neil & QCNE
12736; J. Jaramillo 6970; J. Jaramillo 7039; W. Palacios, G. Tipaz, E. Gudio & B. Cuamacos 9689;
Morphospecies 6: ECUADOR. C. Aulestia, G. Tipaz,
L. Delgado & G. Lao 105; ECUADOR. C. Aulestia, G.
Tipaz, L. Delgado & G. Lao 227; D. Kenfack 2159; H.
van der Werff, C. Dodson, & W. Palacios 9501; M.
Tirado, E. Albuja & M. Chapiro 287.
Morphospecies 7: VENEZUELA. J.A. Steyermark
95095.
Morphospecies 8: ECUADOR. C. Aulestia, G. Tipaz,
L. Delgado & G. Lao 46; C. Dodson, A. Gentry, W.
Palacios & J. Zaruma 14492; C. Dodson, A. Gentry, W.
Palacios & J. Zaruma 14492; C. Dodson, A. Gentry,
W. Palacios & J. Zaruma 14492; C. Dodson, A. Gentry,
W. Palacios & J. Zaruma 14492; H. van der Werff, B.
Gray & G. Tipas 12377; H. van der Werff, B. Gray &
G. Tipas 12377; H. van der Werff, B. Gray & G. Tipas
12377; H. van der Werff, B. Gray & G. Tipas 12377;
H. van der Werff, B. Gray & G. Tipas 12377; H. van
der Werff, C. Dodson, & W. Palacios 9501; H. Vargas,
X. Aguirre, R. Miranda & C. Robles 1370; W. Palacios
& E. Freire 7430;
Morphospecies 9: COLOMBIA. A. Gentry 35056; W.
Palacios, G. Tipaz, E. Gudio & B. Cuamacs 9689.
Morphospecies 10: GUYANA. D. Kenfack, P.M.
Forget & O. Poncy 2108; D. Kenfack, P.M. Forget & O.
Poncy 2109; P. Acedevo 3431; P. Mutchnick & B.
Allicock 383; P.M. Forget 501; P.M. Forget 502; R.C.
Ek, M. Williams & A Williams 624; S. Tiwari 437;
Morphospecies 11: BRAZIL. G.T. Prance, T.D. Pennington, M. Leppard, O.P. Monteiro & J.F. Ramos
23033; T.D. Pennington, G.T. Prance, J. Ramos & O.
Moteiro 9996. FRENCH GUIANA. Cremers 5029; M.
Hoff 7189; R. Benoist 286; S. Mori & A. Bolten 8649;
S. Mori & B. Boom 15123; S. Mori et al. 21528; Sagot
979; Sagot s.n. SURINAME. BBS 407; G. Stahel 105; H.
Jimenez-Saa LBB14307; R. Evans & S. Koemar 3181;
R. Evans & S. Koemar, J. Ramlal & D. Traag 2479;
SC 5566.
Morphospecies 12: GHANA. D. Kenfack & J. Amponsah 2091; H.H. Schmidt, J. Amponsah & A. Welsing
1958; GUINEA. A. Chevalier 14865. A. Chevalier
14869; A. Chevalier 20474; Cordonier 243; F. Malaisse
2541. LIBERA. A.G. Voorhoeve 1288; D.H. Linder 905;
D.H. Linder 1156; J.G. Adam 28702; J.T. Baldwin Jr
10284; J.W.A. Jansen 1360. SENEGAL. A. Chevalier
3164; Anonymous 3166; D. Kenfack, M. Gueye & M.
Sadio 2076; D. Kenfack, M. Gueye & M. Sadio 2077;
D. Kenfack, M. Gueye & M. Sadio 2080; D. Kenfack,
M. Gueye & M. Sadio 2085; Fotius K811; Kaoussou
Sambou, 1; Kaoussou Sambou, T. Sarr sn; Madsen
J.E. 3123. SIERRA LEONE. C.E. Lane-Pook 349; D.
Cledhill 301; G.F. Scott Elliot 4153; M. Heudelot 749;
M. Heudelot 749.
Morphospecies 13: BENIN. C. Barter 3248. GHANA.
A.A. Enti 613; C. Vigne 2525; Hutchinson 146; J.
Deaw 380; J.J. Chipp 262; J.K. Morton 25327; N.H.
Johnson 146; IVORY COAST. A. Chevalier 16233;
IVORY COAST. A. Chevalier B. 22279; IVORY COAST.
J.J.F.E. de Wilde 3120; IVORY COAST. Martineau 303;
NIGERIA. J.M. Dalziel 342.
Morphospecies 14: GUINEA. A. Chevalier 408; A.
Chevalier 461; A. Chevalier 18192; H. Pobguin sn.
GUINEA-BISSAU. E. Santo 1310. IVORY COAST. H.
Pobguin 264; L. Gautier, R. Spichiger & H. Tr
2857. MALI. M. R. Dubois 38. SENEGAL. D. Kenfack
2070; D. Kenfack 2074; D. Kenfack 2084; D. Kenfack,
M. Gueye & M. Sadio 2071; D. Kenfack, M. Gueye &
M. Sadio 2072; D. Kenfack, M. Gueye & M. Sadio
2079; D. Kenfack, M. Gueye & M. Sadio 2081; D.
Kenfack, M. Gueye & M. Sadio 2083; Goudiaby A.
1273; Jacques-George A. 17514.
Morphospecies 15: CAMEROON. Kenfack 2007;
Sainge M 318.
Morphospecies 16: CAMEROON. G.L. Bates 535; P.
Tchouto 2963; R. Letouzey SRFK 1289.
Morphospecies 17: CAMEROON. A.J.M. Leeuwenberg
5279; R. Letouzey 14950; A. Binuyo 45466; A.S. Jhoneuill 220; D. Kenfack 737; D. Kenfack 1024; D.
Kenfack 1169; D. Kenfack 1170; D. Kenfack 1364; D.
Kenfack 1656; D. Kenfack 2118; D.W. Thomas 2352;
D.W. Thomas 8256; D.W. Thomas & J. Nemba 5932; G.
Zenker 145; G. Zenker 3713; G. Zenker sn; G. Zenker
sn; J.J. Bos 4374; M. Akogo 174; Mainoud 430; T.D.
Maitland 431. EQUATORIAL GUINEA. M.F. de Carvalho
2252. GABON. C. Barter 153. NIGERIA. P.A. Talbot
1462. NIGERIA. P.W. Richards 3013; G. Mann 1767.
221
Morphospecies 18: SAO TOME. A. Chevalier 14503.

Morphospecies 19: BURUNDI. Auguier P. 4186; C.M.
Harris 187. RWANDA. Bouxin G. 925; P.M. Forget & A.
Nyiramema 581. TANZANIA. J. Kahuramanga 2565.
UGANDA. B.T. Styles 20.
Morphospecies 20: CAMEROON. D. Maitland 1663;
D.W. Thomas 5489; L. Zapfack 1145; Letouzey R.
14560; M. Etuge 3452; M.N. Sainge 1602; Sainge M.
1261. NIGERIA. J.D. Chapman 4354.
Morphospecies 21: CAMEROON. A. Chevalier 27115;
D. Kenfack & L. Zapfack 2035; R. Letouzey 3901; W.
J.J.O. de Wilde & B.E.E. de Wilde-Duyfjes 2765.
GABON. A. Chevalier 27116.
Morphospecies 22: CAMEROON. D. Kenfack 1365; D.
Kenfack 2106; D. Kenfack & M.N. Sainge 2000; D.
Kenfack & M.N. Sainge 2001; D. Kenfack & M.N.
Sainge 2002; D. Kenfack, M.N. Sainge & O. Hardy
2107.
Morphospecies 23: DR CONGO. A. Masanga 17; E.N.
Ewango 1114; J. Lebrun 5869. GABON. A. Chevalier
27117; D. Kenfack 2065; D. Kenfack 2067. D. Kenfack
2068; D. Kenfack & N. Elias 1356.
Morphospecies 24: GABON. D. Kenfack 2050; D.
Kenfack 2051; D. Kenfack 2052.
Morphospecies 25: CAMEROON. J.J.F.E. de Wilde
7477; D. Kenfack 1087; D. Kenfack 1165; D. Kenfack
1166; D.W. Thomas 4782; F.J. Breteler, J.J.F.E. de
Wilde & A.J.M. Leeuwenberg 2583; J.J. Boss 2989;
M.N. Sainge 378; P. Tchouto 3196; R. Letouzey 10155;
T.R. van Andel, B. Mva & D. Mamia 4116. CENTRAL
AFRICAN REPUBLIC. D.J. Harris & J.M. Fay 1953.
GABON. D. Kenfack 2053; D. Kenfack 2057; D.
Kenfack 2060; D. Kenfack 2066; F.J. Breteler & M.E.
Leal 14256; J.M. & B. Reitsma 1321; J.M. & B.
Reitsma 1321; L. White 1053; L. White 1532; R.P.
Klaine 549.
Morphospecies 26: CAMEROON. G. Le Testu 4395; R.
Letouzey 3901; R. Letouzey 4525; DR CONGO. J.
Lebrun 1245; J. Louis 11935.
Morphospecies 27: D. Kenfack 627; CAMEROON. D.
Kenfack 1368; D. Kenfack 1508; D. Kenfack 627B; D.
Kenfack, M.N. Sainge, N. Elias & P. Ndumbe 2116; J.
Watts 632; J.M. Mbani 379; N. Ndam 1315; P. Fraser
356. NIGERIA. M.G. Latilo 29.

Kenfack 2011 PDF

Загружено:

Сведения о документе

Исходное описание:

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Kenfack 2011 PDF

Загружено:

Авторское право:

Доступные форматы

Botanical Journal of the Linnean Society, 2011, 165, 186221.

Resurrection in Carapa (Meliaceae): a reassessment of

ADDITIONAL KEYWORDS: amphi-Atlantic internal transcribed spacer principal coordinate analysis

Salim & Bermingham, 2003; Hebert et al., 2004;

Figure 1. Global distribution of Carapa.

tribe Xylocarpeae, he recognized only seven species

tion in leaf length, number of pairs of leaflets, petiole

MATERIAL AND METHODS

HERBARIUM STUDIES AND FIELDWORK

A preliminary morphological study of herbarium

features, such as habit and organ colour, and habitat

CHARACTERS EXAMINED AND

The application of numerical taxonomy to Carapa

Number of leaflets (LNUM)

Length of the pedicel (PEDL)*

Fruit and seed characters

*Codes are given for characters used in the morphometric analyses.

comparisons among many specimens. In the rare

below, supplemented with data from other organs

To display the variation of the characters measured

Figure 6. Simplified best tree obtained from the

maps of the morphospecies were produced using

During fieldwork, altitude, latitude and longitude were

A one-way ANOVA showed significant differences for

pedicel indumentum and calyx indumentum, and

Specimens studied were collected from throughout

is no interdigitation between morphospecies 2, 5 and

qualitative characters PEDIND and CALIND (loaded

morphospecies 9, the others are clearly separated

The specimens examined in this analysis came from

rachises, but not longer petioles. Leaf and rachis

0.50 0.40 0.30 0.20 0.10

Table 5. Loadings of the three first axes of the principal

overlap in either of the two scatter plots (Fig. 11C, D).

PCoA was performed on 108 specimens representing

contrast to the taxonomic interpretation found in the

Venezuela were also included in the morphometric

Table 6. Loadings of the three first axes of the principal

the adjacent Amazonian forest of Colombia and

specimens of morphospecies 7 are considered as

leaflet blades, inflorescence branches and pedicels.

0.40 0.30 0.20 0.10 0.00

Figure 11. Continued

Table 7. Loadings of the three first axes of the principal

described as a new species, C. longipetala Kenfack

Table 8. Loadings of the three first axes of the principal

cies 23 and 25 have four and six ovules, respectively.

Cameroon in 1917 and previously placed by as

Figure 13. Continued

Table 9. Loadings of the three first axes of the principal

Table 10. Loadings of the three first axes of the principal

Specimens of morphospecies 19 are restricted to

to C. grandiflora based on their habitat, the number

Table 12. Summary of taxonomic decisions in the present

Assigned names in this

C. guianensis Aubl. s.s.

N/A, not available.

have up to 16 pairs of narrow leaflets with fine

growth habit: morphospecies 20 is a small tree to

diflora (bi-ovulate locules and fruits with < 10 seeds)

and oblong and those of morphospecies 26 are

densely pubescent. Their ranges largely overlap and