This art icle was downloaded by: [ Oscar Grillo]
On: 11 April 2012, At : 02: 12
Publisher: Taylor & Francis
I nform a Lt d Regist ered in England and Wales Regist ered Num ber: 1072954 Regist ered office: Mort im er House,
37- 41 Mort im er St reet , London W1T 3JH, UK
Systematics and Biodiversity
Publicat ion det ails, including inst ruct ions f or aut hors and subscript ion inf ormat ion:
ht t p: / / www. t andf online. com/ loi/ t sab20
Seed image analysis and taxonomy of Diplotaxis DC.
(Brassicaceae, Brassiceae)
Oscar Grillo
a b
, David Draper
c
, Gianf ranco Venora
a
& Juan Baut ist a Mart ínez-Laborde
d
a
St azione Speriment ale di Granicolt ura per la Sicilia, Via Sirio, 1 - 95041 Sant o Piet ro Calt agirone, It aly
b
Cent ro Conservazione Biodiversit à, Dipart iment o di Scienze della Vit a e dell’ Ambient e,
Universit à di Cagliari – V. le Sant ’ Ignazio da Laconi, 13 – 09123 Cagliari, It aly
c
Inst it ut o de Ecología, Universidad Técnica Part icular de Loj a, San Cayet ano Alt o s/ n, CP
11 01 608, Loj a, Ecuador
d
Depart ament o de Biologia Veget al, Escuela Técnica Superior de Ingenieros Agrónomos,
Universidad Polit écnica de Madrid – Ciudad Universit aria, s/ n – 28040, Madrid, Spain
Available online: 07 Mar 2012
To cite this article: Oscar Grillo, David Draper, Gianf ranco Venora & Juan Baut ist a Mart ínez-Laborde (2012): Seed image
analysis and t axonomy of Diplot axis DC. (Brassicaceae, Brassiceae), Syst emat ics and Biodiversit y, 10: 1, 57-70
To link to this article: ht t p: / / dx. doi. org/ 10. 1080/ 14772000. 2012. 658881
PLEASE SCROLL DOWN FOR ARTI CLE
Full t erm s and condit ions of use: ht t p: / / www.t andfonline.com / page/ t erm s- and- condit ions
This art icle m ay be used for research, t eaching, and privat e st udy purposes. Any subst ant ial or syst em at ic
reproduct ion, redist ribut ion, reselling, loan, sub- licensing, syst em at ic supply, or dist ribut ion in any form t o
anyone is expressly forbidden.
The publisher does not give any warrant y express or im plied or m ake any represent at ion t hat t he cont ent s
will be com plet e or accurat e or up t o dat e. The accuracy of any inst ruct ions, form ulae, and drug doses should
be independent ly verified wit h prim ary sources. The publisher shall not be liable for any loss, act ions, claim s,
proceedings, dem and, or cost s or dam ages what soever or howsoever caused arising direct ly or indirect ly in
connect ion wit h or arising out of t he use of t his m at erial.
Systematics and Biodiversity (2012), 10(1): 57–70
Research Article
Seed image analysis and taxonomy of Diplotaxis DC.
(Brassicaceae, Brassiceae)
OSCAR GRILLO1,2, DAVID DRAPER3, GIANFRANCO VENORA1 & JUAN BAUTISTA MARTÍNEZ-LABORDE4
1
Stazione Sperimentale di Granicoltura per la Sicilia, Via Sirio, 1 - 95041 Santo Pietro - Caltagirone, Italy
Centro Conservazione Biodiversità, Dipartimento di Scienze della Vita e dell’Ambiente, Università di Cagliari – V.le Sant’Ignazio da
Laconi, 13 – 09123 Cagliari, Italy
3
Instituto de Ecologı́a, Universidad Técnica Particular de Loja, San Cayetano Alto s/n, CP 11 01 608, Loja, Ecuador
4
Departamento de Biologia Vegetal, Escuela Técnica Superior de Ingenieros Agrónomos, Universidad Politécnica de Madrid – Ciudad
Universitaria, s/n – 28040, Madrid, Spain
Downloaded by [Oscar Grillo] at 02:12 11 April 2012
2
(Received 29 April 2011; revised 28 December 2011; accepted 16 January 2012)
The genus Diplotaxis, comprising 32 or 34 species, plus several additional infraspecific taxa, displays a considerable degree
of heterogeneity in the morphology, molecular markers, chromosome numbers and geographical amplitude of the species.
The taxonomic relationships within the genus Diplotaxis were investigated by phenetic characterisation of germplasm
belonging to 27 taxa of the genus, because there is an increasing interest in Diplotaxis, since some of its species (D.
tenuifolia, D. muralis) are gathered or cultivated for human consumption, whereas others are frequent arable weeds (D.
erucoides) in many European vineyards. Using a computer-aided vision system, 33 morpho-colorimetric features of seeds
were electronically measured. The data were used to implement a statistical classifier, which is able to discriminate the taxa
within the genus Diplotaxis, in order to compare the resulting species grouping with the current infrageneric systematics of
this genus. Despite the high heterogeneity of the samples, due to the great intra-population variability, the stepwise Linear
Discriminant Analysis method, applied to distinguish the groups, was able to reach over 80% correct identification. The
results obtained allowed us to confirm the current taxonomic position of most taxa and suggested the taxonomic position of
others for reconsideration.
Key words: computer vision, germplasm characterisation, Linear Discriminant Analysis, morpho-colorimetric
measurements, seed identification, statistical classification
Introduction
The genus, Diplotaxis DC. (Brassicaceae, tribe Brassiceae),
currently comprises 32 species (Warwick et al., 2006) or 34
(Table 1), plus several additional infraspecific taxa, native
to Europe, the Mediterranean basin, SW Asia (up to the Himalayas) and Macaronesia. There is an increasing interest
in Diplotaxis, since some of its species (D. tenuifolia, D.
muralis) are gathered or cultivated for human consumption
as rocket salad (Pignone, 1997; Pimpini & Enzo, 1997),
whereas others are frequent arable weeds (D. erucoides) in
many European vineyards (Sans & Masalles, 1994).
The genus displays a considerable degree of heterogeneity in the morphology, molecular markers, chromosome
numbers and geographical amplitude of the species (Table 1). Its highest diversity is found in NW Africa and
Correspondence to: Oscar Grillo. E-mail: oscar.grillo.mail@
gmail.com
ISSN 1477-2000 print / 1478-0933 online
C 2012 The Natural History Museum
http://dx.doi.org/10.1080/14772000.2012.658881
the Iberian Peninsula, where several endemic taxa are restricted to very limited areas (D. ibicensis, D. brachycarpa)
or even small islands (D. siettiana), or occupy more extensive regions (D. assurgens, D. virgata). Other species
display much larger ranges, either across N Africa and SW
Asia (D. harra) or central and southern Europe (D. tenuifolia); a few of these have colonised elsewhere (N and S
America, Australia, etc.).
Chromosome numbers are known for most taxa and range
from n = 7 in D. erucoides, through n = 8, 9, 10 and 11,
to n = 13 in the D. harra and allied species, D. muralis.
The only species with a higher ploidy level has n = 21,
and according to solid evidence (Harberd & McArthur,
1972; Sánchez-Yélamo & Martı́nez-Laborde, 1991; Warwick & Anderson, 1997; Eschmann-Grupe et al., 2003), is
an amphidiploid probably arisen from D. tenuifolia (n =
11) and D. viminea (n = 10). Morphological variation in
this genus includes remarkable differences in habit (from
58
O. Grillo et al.
Table 1. Diplotaxis taxa according to the subgenera (SG) and sections (S) proposed in Gómez-Campo & Martı́nez-Laborde (1998),
diploid chromosome numbers (2n) and geographical areas of distribution.
SGa
Sb
Downloaded by [Oscar Grillo] at 02:12 11 April 2012
D
H
R
Rh
Hc
Hp
aSubgenera:
Species (and subspecies)
2nc
Geographical area
D. tenuifolia (L.) DC.
D. cretacea Kotov
D. muralis (L.) DC.
subsp. muralis
subsp. ceratophylla (Batt.) Mart.-Laborde
D. scaposa DC.
D. simplex (Viv.) Spr.
D. viminea (L.) DC.
D. harra (Forssk.) Boiss.
subsp. harra
subsp. crassifolia (Raf.) Maire
subsp. lagascana (DC.) O. Bolòs & Vigo
D. kohlaanensis A. G. Miller & J. Nyberg
D. villosa Boulos & Jall.
D. pitardiana Maire
D. nepalensis Hara
D. antoniensis Rustan
D. glauca (J. A. Schmidt) O. E. Schulz
D. gorgadensis Rustan
subsp. gorgadensis
subsp. brochmannii Rustan
D. gracilis (Webb) O. E. Schulz
D. hirta (A. Chev.) Rustan & Borgen
D. sundingii Rustan
D. varia Rustan
D. vogelli (Webb) Cout.
D. acris (Forssk.) Boiss.
D. griffithii (Hook.f. & W. Thomps.) Boiss.
D. assurgens (Delile) Grenier
D. berthautii Braun-Blanq. & Maire
D. brachycarpa Godron
D. catholica (L.) DC.
D. ollivieri Maire
D. siifolia Kunze
subsp. siifolia
subsp. bipinnatifida (Coss.) Mart.-Laborde
subsp. vicentina (Sampaio) Mart.-Laborde
D. tenuisiliqua Delile
subsp. tenuisiliqua
subsp. rupestris (J. Ball) Mart.-Laborde
D. virgata (Cav.) DC.
D. ibicensis (Pau) Gómez-Campo
D. brevisiliqua (Coss.) Mart.-Laborde
D. ilorcitana (Sennen) Aedo, Mart.-Laborde & Muñoz Garm.
D. siettiana Maire
D. erucoides (L.) DC.
subsp. erucoides
subsp. longisiliqua (Coss.) Gómez-Campo
22
22
Europe, Middle East
NE Ukraine and adjacent Russia
42
22
20
Europe
NE Algeria
Island of Lampedusa
N Africa
Europe, N Africa, Middle East
26
26
26
26
N Africa, Middle East
Sicily
SE Spain
Yemen
Jordan
NW Africa
Nepal
Cape Verde
Cape Verde
26
26
26
26
22
18
18
18
18
-
Cape Verde
Cape Verde
Cape Verde
Cape Verde
Cape Verde
Cape Verde
Cape Verde
N Africa, Middle East (to Iraq)
Afghanistan, Pakistan
Morocco
S Morocco
NE Algeria
Iberian Peninsula, N Morocco
S Morocco
20
20
Iberian Peninsula, NW Africa
S Morocco
SW Portugal
18
18
16
16
16
16
N Morocco, NW Algeria
S Morocco
Iberian Peninsula, Morocco
E Spain coast, Balearic Islands
NE Morocco, NW Algeria
E Spain
Island of Alboran
14
14
Europe, N Africa, Middle East
NE Algeria
D, Diplotaxis; H, Hesperidium (O. E. Schulz) Nègre; R, Rhynchocarpum (Prantl) Mart.-Laborde.
bSections: Rh, Rhynchocarpum; Hc, Heterocarpum Mart.-Laborde; Hp, Heteropetalum Mart.-Laborde.
cAs reported by Amin (1972), Harberd (1972, 1976), Gómez-Campo (1980), Takahata & Hinata (1978),
Martı́nez-Laborde (1988, 1991), Fernandes &
Queirós (1970–71) and Rustan (1996).
annuals to subshrubby perennials), petal shape (with a distinct claw or a tapering limb), colour (mostly yellow, but
also white or violet) and venation (brochidodromous or
cladodromous to eucamptodromous) and fruit structure. As
in most other Brassiceae, a number of Diplotaxis species
are heteroarthrocarpous (presence of seeds in the stylar
portion of fruit). On the basis of this variation, the
latest infrageneric system, proposed by Gómez-Campo
Downloaded by [Oscar Grillo] at 02:12 11 April 2012
Seed image analysis and taxonomy of Diplotaxis DC.
& Martı́nez-Laborde (1998), recognises three subgenera:
Diplotaxis, Hesperidium and Rhynchocarpum, the latter
including three sections, namely Rhynchocarpum, Heteropetalum and Heterocarpum (Table 1). The system turns
out to be rather consistent with variation in known chromosome numbers and cytodemes (Harberd, 1972, 1976; Takahata & Hinata, 1983; Prakash et al., 1999). The molecular evidence also indicates that Diplotaxis constitutes a
remarkably heterogeneous, even polyphyletic genus. Phylogenetic analyses of chloroplast-DNA restriction site variation of 21 Diplotaxis taxa (Warwick et al., 1992) showed a
clear separation into the two clades previously established
for the whole tribe and designated as the Rapa-Oleracea
and Nigra lineages by Warwick & Black (1991) or the
Brassica and Sinapis lineages by Pradhan et al. (1992).
The Rapa-Oleracea lineage includes the taxa belonging to
subgen. Diplotaxis, plus both subspecies of D. erucoides
(subgen. Rhynchocarpum sect. Heteropetalum), whereas
all the remaining studied taxa belonging to subgen. Rhynchocarpum appear in the Nigra lineage (no taxon of subgen. Hesperidium has been so far examined for molecular
markers). Grouping within each lineage is rather consistent
with known chromosome numbers and cytodemes as recognised by Harberd (1976) and Takahata & Hinata (1983). In
the Rapa-Oleracea lineage, one clade includes two sister
groups, one with D. harra (n = 13) and the other with D.
tenuifolia and related taxa (n = 11), while D. muralis and D.
viminea appear in a third clade. A fourth clade corresponds
to subgen. Rhynchocarpum sect. Heteropetalum. In the Nigra lineage, one clade contains most species of sect. Rhynchocarpum, though D. brachycarpa appears in a second
clade, while a third clade corresponds exactly to sect. Heterocarpum. The dendrogram obtained by Martı́n & SánchezYélamo (2000) on the basis of nuclear DNA microsatellite
Fig. 1. Seeds of D. acris, D. erucoides and D. tenuifolia.
59
markers of 10 Diplotaxis species shows two main branches,
approximately corresponding to the mentioned lineages, although one of the branches combines species of both lineages. The phenogram obtained by Eschmann-Grupe et al.
(2003) with RAPD data of 18 Diplotaxis species is more
ladder-like in structure and therefore the separation of taxa
into the two lineages is less clear-cut, but still, clusters correspond quite well to within-lineage clades in Warwick et al.
(1992) and to known chromosome numbers and cytodemes.
The seed morphology of this genus has received little
attention to date. Bengoechea & Gómez-Campo (1975)
included 15 Diplotaxis taxa in their comprehensive survey
of seed exomorphology and anatomy in the Brassiceae.
Martı́nez-Laborde (1988) examined a few exomorphological seed traits in 29 of the Diplotaxis taxa (species
and subspecies) listed in Table 1. According to these
authors, Diplotaxis seeds are ochre to brown coloured,
small (0.6–1.3 mm long × 0.4–1.0 mm wide), and ovoid
to ellipsoid in shape (Fig. 1). The only, notable exception
is D. siifolia, which has more or less spherical seeds, more
similar to those of Brassica, not only in shape, but also
in the extension of the thickened portion of the radial
(anticlinal) cell walls of the subepidermal, palisade layer
of the testa (Bengoechea & Gómez-Campo, 1975).
In the last two decades, a remarkable increase in image
analysis applications has been applied in the plant biology research field. Until recently, the dimensional measurements as length and width of the seeds were made manually, generally by calipers, while fixed categories officially
recognised, reported by Martin (1946), Stearn (1980) &
Werker (1997), were used to describe contour shapes. With
the same principles, colour evaluation was commonly exeR
Colour Charts (Facuted by comparison with the Munsell
gundez & Izco, 2004). It is evident that there are difficulties
Downloaded by [Oscar Grillo] at 02:12 11 April 2012
60
O. Grillo et al.
making these measurements objective and repeatable, especially when the seed dimension is extremely reduced
(e.g. in Brassicaceae, Cynomoriaceae, Primulaceae, Rubiaceae, Scrophulariaceae, etc.). A technological evolution
occurred that resulted in overcoming some of these limits
(e.g. difficulty to obtain accurate, objective and repeatable
results). The recent literature proves how this innovative
technology improved morphometrics and colour evaluation
(Liao et al., 1994; Granitto et al., 2003; Shahin & Symons,
2003a; Kiliç et al., 2007; Venora et al., 2007, 2009a, 2009b;
Bacchetta et al., 2008, 2011a; Dana & Ivo, 2008; Grillo et
al., 2010, 2011; Smykalova et al., 2011). The importance of
morphologic seed traits such as shape, size and external ornamentation, as diagnostic factors in plant taxonomy, is emphasized by the constantly increasing availability of seeds
collected from wild plants cultivated ex situ (e.g. botanic
gardens) and stored in germplasm banks. The implementation of a workable system, standardised on the basis of a
database of morpho-colorimetric features, could be a valid
tool to support the accession activities of germplasm banks,
during seed cataloguing and identification or for determination and revision of critical taxonomic groups.
The aims of the present study were to use seed morphocolorimetric data obtained by image analysis to implement
a dedicated statistical classifier able to discriminate the taxa
within the genus Diplotaxis and to compare the resulting
species grouping with the current infrageneric systematics of this genus, as well as with groupings more recently
revealed by molecular markers.
Materials and methods
Selected germplasm
Seed samples belonging to the Plant Genebank of the Universidad Politécnica de Madrid (BGV-UPM), and corresponding to most Diplotaxis taxa (27 out of the 34 species
and subspecies listed in Table 1) were analysed. The accession codes of samples in the genebank and the localities of
collection and the amount of the seeds investigated were
recorded (Table 2). All the analysed seed accessions were
stored in the same conditions and for a period longer than
15 years. This allowed us to exclude any possible variation in seed colour, due to the ageing process. In order
to guarantee the representations of accessions and to minimise the intraspecific changes of shape and sizes of the
seeds, due to the seed position inside the fruit and to the
fruit position on the plant (Harper et al., 1970), all seeds of
whole accessions were measured (total of 8918 seeds analysed). Although small differences in seed size and shape
can exist between different populations of the same taxon,
principally due to climatic and edaphic factors as well as to
Table 2. Seed samples of Diplotaxis taxa (species and subspecies) investigated.
Accession code
Species and subspecies
Geographical origin of seeds
Seed amount
BGV-UPM-8860
BGV-UPM-2978
BGV-UPM-7522
BGV-UPM-6467
BGV-UPM-7517
BGV-UPM-2949
BGV-UPM-1445
BGV-UPM-5056
BGV-UPM-6483
BGV-UPM-6472
BGV-UPM-6635
BGV-UPM-9047
BGV-UPM-7032
BGV-UPM-4065
BGV-UPM-4678
BGV-UPM-6486
BGV-UPM-9250
BGV-UPM-3025
BGV-UPM-2964
BGV-UPM-2970
BGV-UPM-7620
BGV-UPM-6453
BGV-UPM-7448
BGV-UPM-7521
BGV-UPM-7527
BGV-UPM-8071
BGV-UPM-3066
D. acris
D. assurgens
D. berthautii
D. brachicarpa
D. brevisiliqua
D. catholica
D. cretacea
D. erucoides subsp. erucoides
D. erucoides subsp. longisiliqua
D. harra subsp. harra
D. harra subsp. crassifolia
D. harra subsp. lagascana
D. ibicensis
D. ilorcitana
D. muralis subsp. muralis
D muralis subsp. ceratophylla
D. ollivieri
D. siettiana
D. siifolia subsp. siifolia
D. siifolia subsp. bipinnatifida
D. siifolia subsp. vicentina
D. simplex
D. tenuifolia
D. tenuisiliqua subsp. tenuisiliqua
D. tenuisiliqua subsp. rupestris
D. viminea
D. virgata
Israel (unknown locality)
Morocco, South of Agadir
Morocco, 120 km North of Marrakech
Algeria, North of Sidi Aı̈ssa
Morocco, Cala Iris
Spain, Toledo, Talaverilla la Nueva
Unknown (Moscow Botanical Garden)
Spain, Lleida, Preixana
Algeria, between El Kantara and Batna
Algeria, 100 km West of Biskra
Italy, Sicily, Caltanissetta
Spain, Alicante, La Albufereta
Spain, Ibiza, Cala Eubarca
Spain, Almerı́a, Tabernas
Tunisia, 11 km East of Gafsa
Aleria, Tazoult-Lambese
Morocco, 10 km South of Goulimine
Spain, Island of Alborán
Morocco, Kenitra
Morocco, Agadir
Portugal, Algarve, Aljezur
Tunisia, North of Gafsa
Turkey, Iznik
Morocco, 75 km North of Ben Guerir
Morocco, Marrakech
Spain, Tarragona
Spain, Madrid
222
329
594
322
784
223
270
444
483
337
416
239
233
241
169
163
69
253
250
357
281
397
298
496
473
110
465
Seed image analysis and taxonomy of Diplotaxis DC.
61
Table 3. List of features measured on seeds.
Downloaded by [Oscar Grillo] at 02:12 11 April 2012
Feature
Morphometric features
A
P
Pconv
PCrof
Pconv /PCrof
Dmax
Dmin
Dmin /Dmax
Sf
Rf
Ecd
EAmax
EAmin
Colourimetric features
Rmean
Rsd
Gmean
Gsd
Bmean
Bsd
Hmean
Hsd
Lmean
Lsd
Smean
Ssd
Densitometric features
Dmean
Dsd
S
K
H
E
Dsum
SqDsum
Description
Area
Perimeter
Convex perimeter
Crofton perimeter
Perimeter ratio
Max diameter
Min diameter
Feret ratio
Shape factor
Roundness factor
Eq. circular diameter
Maximum ellipse axis
Minimum ellipse axis
Seed area (mm2)
Seed perimeter (mm)
Convex perimeter of the seed (mm)
Crofton perimeter of the seed (mm)
Ratio between convex and Crofton’s perimeters
Maximum diameter of the seed (mm)
Minimum diameter of the seed (mm)
Ratio between minimum and maximum diameters
Seed shape descriptor = (4π A)/ P 2 (normalized value)
Seed roundness descriptor = (4A)/(π Dmax 2) (normalized value)
Diameter of a circle with equivalent area (mm)
Maximum axis of an ellipse with equivalent area (mm)
Minimum axis of an ellipse with equivalent area (mm)
Mean red channel
Red std. deviation
Mean green channel
Green std. deviation
Mean blue channel
Blue std. deviation
Mean hue channel
Hue std. deviation
Mean lightness ch.
Lightness std. dev.
Mean saturation ch.
Saturation std. dev.
Red channel mean value of seed pixels (grey levels)
Red channel standard deviation of seed pixels
Green channel mean value of seed pixels (grey levels)
Green channel standard deviation of seed pixels
Blue channel mean value of seed pixels (grey levels)
Blue channel standard deviation of seed pixels
Hue channel mean value of seed pixels (grey levels)
Hue channel standard deviation of seed pixels
Lightness channel mean value of seed pixels (grey levels)
Lightness channel standard deviation of seed pixels
Saturation channel mean value of seed pixels (grey levels)
Saturation channel standard deviation of seed pixels
Mean density
Density std. deviation
Skewness
Kurtosis
Energy
Entropy
Density sum
Square density sum
Density channel mean value of seed pixels (grey levels)
Density channel standard deviation of seed pixels
Asymmetry degree of intensity values distribution (grey levels)
Peakness degree of intensity values distribution (densitometric units)
Measure of the increasing intensity power (densitometric units)
Dispersion power (bit)
Sum of density values of the seed pixels (grey levels)
Sum of the squares of density values (grey levels)
genotype–environment interactions, the differences were
minimal (Bacchetta et al. 2011b); moreover, considering
the specific phenotypic representation of the BGV-UPM
seed accessions, the intra-population differences were not
considered in this study.
Image analysis
The sample images were acquired according to Bacchetta
et al. (2008) by a flatbed scanner (HP Scanjet 4890), with
a resolution of 600 dpi and a scanning area not exceeding
2800 × 2800 pixels. Using a KS-400 V. 3.0 (Carl Zeiss,
Vision, Oberkochen, Germany) image analysis system and
its libraries, a specific macro was developed in the Image Analysis Laboratory of the Stazione Sperimentale di
Granicoltura per la Sicilia (SSG), to obtain measurements
of 33 morpho-colorimetric features of seeds (Table 3). The
scanner was calibrated for colour matching following the
protocol of Shahin and Symons (2003b) before image acquisition, as suggested by Venora et al. (2009b).
Statistical classifier
The data were evaluated statistically by applying the stepwise Linear Discriminant Analysis (LDA) algorithm, using
SPSS software package release 15 (SPSS Inc. 1989–2006).
This approach is commonly used to classify/identify unknown groups characterised by quantitative and qualitative
variables (Fisher, 1936, 1940). The best features for seed
sample identification were detected implementing a stepwise LDA method and a statistical classifier to discriminate
and classify the seeds on the basis of the selected characters.
When several variables are available, the stepwise method
can be useful by automatically selecting the best characters
on the basis of three statistical variables: Tolerance, F-toenter and F-to-remove. The Tolerance value indicates the
proportion of a variable variance not accounted for by other
independent variables in the equation. A variable with very
low Tolerance value provides little information to a model.
F-to-enter and F-to-remove values define the power of each
variable in the model and they are useful to describe what
happens if a variable is inserted or removed, respectively,
62
O. Grillo et al.
from the current model. This method starts with a model
that does not include any of the variables. At each step, the
variable with the largest F-to-enter value that exceeds the
entry criteria chosen (F ≥ 3.84) is added to the model. The
variables left out of the analysis at the last step have F-toenter values smaller than 3.84, hence no more are added.
The process was automatically stopped when no remaining
variables increased the discrimination ability. Afterwards,
the cross-validation procedure was applied to validate the
performance of the developed classifier. This method is
useful to analyse small datasets when a broad group of new
unknown cases is lacking. It tests individual cases and classifies them on the basis of all others (SPSS Application
Guide, 1999).
Downloaded by [Oscar Grillo] at 02:12 11 April 2012
Results and discussion
The morpho-colorimetric analysis of the seed samples of
Diplotaxis allowed us to obtain extremely precise data about
their size, shape and colour (raw data not shown). Seed
length and width mean values were 910.49 ± 162.92 µm
and 633.67 ± 111.51 µm, respectively, with a mean diameter ratio of 0.70 ± 0.08. The shape is ovoid to ellipsoid,
as determined by shape and roundness factors values (0.92
± 0.05 and 0.64 ± 0.08, respectively). The only exception
was D. siifolia, whose seeds are subspherical, with a diameter ratio, shape and roundness factor mean values of
0.87 ± 0.05, 0.97 ± 0.04 and 0.80 ± 0.05, respectively.
These results are in accordance with previous measurements reported by Bengoechea & Gómez-Campo (1975) &
Martı́nez-Laborde (1988).
Key parameters
Evaluating the contribution of the variables using the discrimination algorithm (LDA), it was possible to identify the
features that, more than others, were relevant for the intrageneric separation of the Diplotaxis taxa included in this
study. Considering the number of steps used by the stepwise
method, Tolerance and F-to-remove values, it was possible
to observe that 29 out of the 33 features used were evaluated by the statistical classifier to discriminate among the
taxa (Table 4). The selected parameters were prevalently
related to the colour or more in general to the densitometric features of the seeds, and six of these with highest
F-to-remove values were colorimetric (Table 4). This result
confirms that, apart from a few cases in which the seeds
look morphologically different (D. siettiana, D. siifolia and
D. tenuifolia), the seed shape and size are very similar in
the species investigated.
Species discrimination
A fairly satisfactory level of Diplotaxis taxa (species and
subspecies) discrimination was achieved by image anal-
Table 4. Ranking of the selected features after 29 cycles of
stepwise analysis.
Step
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
Feature
Tolerance
F-to-remove
Gmean
SqDsum
Dsd
Ssd
Gsd
Rsd
A
Dsum
EAmin
Bsd
Ecd
EAmax
Lsd
Smean
Dmax
Rmean
Rf
Bmean
Dmean
E
Hmean
S
Hsd
K
Pconv
H
PCrof
Pconv /PCrof
Dmin /Dmax
0.094
0.017
0.142
0.221
0.197
0.112
0.014
0.014
0.002
0.389
0.001
0.002
0.097
0.003
0.008
0.034
0.032
0.104
0.003
0.246
0.018
0.082
0.213
0.084
0.001
0.719
0.003
0.081
0.100
199.584
158.198
148.171
141.389
131.760
105.189
105.024
93.790
88.927
87.233
76.720
72.771
72.443
72.132
65.834
59.820
59.123
49.907
45.596
36.601
29.588
25.748
24.921
23.867
21.471
20.365
12.535
6.298
4.156
ysis of morpho-colorimetric seed data. An overall crossvalidation percentage of correct classification of 80.7%
was reached for most taxa, with identification performance
ranging from 70.8% (in D. ibicensis) to 97.3% (in D.
viminea), with the only exceptions of D. acris (56.3%), D.
harra subsp. lagascana (65.7%), D. muralis subsp. ceratophylla (66.9%) and D. tenuisiliqua subsp. rupestris (64.5%)
(Table 5).
The rather poor identification of the Saharan D. acris
(subgen. Hesperidium) seeds arose mostly from mistakes
for those of D. brevisiliqua (20.7%) and D. siettiana (4.1%),
both in subgen. Rhynchocarpum and for three subspecies
of D. harra (13.6% altogether). They are all taxa from the
southern part of the geographical range of the genus with
smaller seeds. Even if it was plausible to think of seed
polymorphism phenomena as a cause for the reduced percentage of correct identification of D. acris, the relative
intra-population phenotypic variability is fully considered
by the LDA and included in the seed sample size. Furthermore, seed polymorphism, when it is present, would only
affect a few characters that consequently are automatically
rejected by the stepwise procedure as non-significatively
relevant to the discrimination process. Therefore, the lower
amount of available discriminant features could be the direct cause of the poor performance.
Downloaded by [Oscar Grillo] at 02:12 11 April 2012
Table 5. Cross-validated percentages of correct identification for the Diplotaxis taxa classifier. The number of seeds is in parentheses.
D. acris (1)
1
56,3%
(125)
D. assurgens (2)
D. berthautii (3)
2
3
82,4%
(271)
3,5%
(21)
3,6%
(12)
86,5%
(512)
D. brevisiliqua (5)
73,3%
(236)
4,2%
(33)
D. catholica (6)
5,8%
(13)
0,4%
(1)
D. cretacea (7)
D. erucoides subsp.
erucoides (8)
D. erucoides subsp.
longisiliqua (9)
D. harra subsp. harra
(10)
D. harra subsp.
crassifolia (11)
D. harra subsp.
lagascana (12)
D. ibicensis (13)
D. ilorcitana (14)
D. muralis subsp.
muralis (15)
D. muralis subsp.
ceratophylla (16)
D. ollivieri (17)
D. siettiana (18)
D. siifolia subsp.
siifolia (19)
D. siifolia subsp.
bipinnatifida (20)
D. siifolia subsp.
vicentina (21)
D. simplex (22)
7
4,0%
(13)
0,5%
(3)
0,3%
(1)
0,5%
(2)
13,7%
(66)
0,9%
(4)
0,4%
(2)
1,2%
(2)
3,4%
(14)
5,9%
(14)
10,7%
(25)
4,1%
(10)
5,3%
(9)
D. virgata
(27)
9
22,0%
(71)
4,1%
(18)
83,4%
(403)
0,6%
(1)
2,4%
(12)
7,0%
(33)
0,9%
(1)
1,9%
(9)
0,5%
(2)
96,7%
(326)
2,9%
(12)
1,7%
(4)
0,4%
(1)
0,9%
(7)
13
0,5%
(1)
6,0%
(47)
14
0,5%
(4)
15
1,2%
(4)
0,4%
(3)
3,6%
(16)
0,2%
(1)
0,6%
(2)
89,9%
(274)
9,2%
(22)
1,2%
(3)
5,3%
(9)
5,3%
(9)
0,2%
(1)
0,9%
(3)
1,2%
(5)
65,7%
(157)
1,7%
(4)
1,2%
(3)
0,2%
(1)
0,2%
(1)
2,1%
(5)
70,8%
(165)
11,6%
(28)
5,3%
(9)
0,5%
(2)
0,3%
(1)
16,8%
(50)
0,9%
(1)
18
4,1%
(9)
19
0,9%
(2)
20
0,5%
(1)
21
22
0,9%
(2)
0,3%
(1)
0,7%
(4)
0,3%
(1)
23
0,9%
(2)
3,9%
(9)
75,9%
(183)
3,7%
(29)
25
7,6%
(25)
5,2%
(31)
1,8%
(6)
2,0%
(12)
4,3%
(17)
0,7%
(2)
1,4%
(1)
1,6%
(4)
0,3%
(1)
1,4%
(1)
3,6%
(9)
0,4%
(1)
0,3%
(1)
4,3%
(3)
0,3%
(1)
1,3%
(5)
0,7%
(2)
0,3%
(1)
1,3%
(4)
27
0,2%
(1)
0,3%
(1)
1,3%
(8)
1,2%
(2)
0,6%
(1)
1,3%
(3)
1,3%
(3)
2,6%
(6)
1,2%
(3)
1,3%
(3)
7,7%
(18)
0,2%
(1)
0,8%
(2)
0,6%
(1)
0,2%
(1)
7,0%
(31)
0,6%
(3)
0,9%
(3)
1,2%
(5)
0,4%
(1)
1,35
(3)
0,6%
(1)
0,3%
(1)
0,2%
(1)
26
0,3%
(2)
0,2%
(1)
1,3%
(3)
0,4%
(1)
71,0%
(120)
24
1,6%
(5)
0,3%
(2)
0,4%
(1)
2,5%
(6)
1,8%
(3)
2,4%
(4)
78,3%
(54)
2,0%
(5)
7,2%
(5)
80,6%
(204)
2,0%
(7)
0,4%
(1)
0,3%
(1)
1,4%
(1)
89,2%
(223)
7,6%
(27)
9,3%
(26)
0,3%
(1)
1,4%
(1)
1,2%
(3)
2,8%
(7)
83,8%
(299)
9,6%
(27)
0,5%
(2)
0,3%
(1)
6,4%
(22)
6,2%
(22)
80,8%
(227)
0,4%
(1)
90,7%
(360)
2,0%
(6)
77,5%
(231)
77,8%
(386)
9,5%
(45)
9,7%
(48)
64,5%
(305)
0,9%
(1)
0,9%
(4)
Overall
17
1,8%
(4)
66,9%
(109)
1,4%
(1)
0,2%
(1)
16
0,9%
(4)
1,4%
(7)
20,9%
(34)
1,5%
(3)
2,8%
(14)
0,6%
(2)
0,5%
(4)
12
6,8%
(15)
88,1%
(238)
81,1%
(360)
0,2%
(1)
3,7%
(6)
11
3,2%
(7)
1,3%
(3)
8,5%
(23)
1,4%
(1)
7,5%
(19)
0,8%
(2)
0,8%
(3)
10
3,6%
(8)
0,1%
(1)
D. tenuifolia (23)
D. tenuisiliqua subsp.
tenuisiliqua (24)
D. tenuisiliqua subsp.
rupestris (25)
D. viminea (26)
8
0,3%
(1)
90,1%
(201)
2,2%
(6)
2,5%
(4)
1,4%
(1)
3,6%
(9)
6
0,3%
(1)
83,2%
(652)
1,3%
(3)
0,7%
(2)
0,9%
(3)
1,2%
(5)
10,9%
(26)
1,3%
(3)
0,4%
(1)
0,6%
(1)
4,3%
(7)
5
20,7%
(46)
6,9%
(34)
19,0%
(90)
97,3%
(107)
7,7%
(36)
10,5%
(49)
78,9%
(367)
Total
100,0%
(222)
100,0%
(329)
100,0%
(594)
100,0%
(322)
100,0%
(784)
100,0%
(223)
100,0%
(270)
100,0%
(444)
100,0%
(484)
100,0%
(337)
100,0%
(416)
100,0%
(239)
100,0%
(233)
100,0%
(241)
100,0%
(169)
100,0%
(163)
100,0%
(69)
100,0%
(253)
100,0%
(250)
100,0%
(357
100,0%
(281)
100,0%
(397)
100,0%
(298)
100,0%
(496)
100,0%
(473)
100,0%
(110)
Seed image analysis and taxonomy of Diplotaxis DC.
D. brachycarpa (4)
4
100,0%
(465)
80.7%
(8918)
63
64
O. Grillo et al.
Table 6. Cross-validated percentages of correct identification for D. harra species classifier at subspecies level. The number of seeds is
in parentheses.
Taxa
D. harra subsp. crassifolia
D. harra subsp. lagascana
D. harra subsp. harra
Overall
D. harra subsp. crassifolia
D. harra subsp. lagascana
D. harra subsp. harra
Total
96.7% (402)
9.2% (22)
0.6% (2)
1.4% (6)
88.7% (212)
1.2% (4)
1.9% (8)
2.1% (5)
98.2% (331)
100.0% (416)
100.0% (239)
100.0% (337)
95.3% (992)
since only 66% of identified seeds were correct. Only a fraction of its misidentified seeds were mistaken for those of
other subspecies of D. harra (9.2% as subsp. crassifolia and
1.7% as subsp. harra). In a separate comparison among the
three subspecies of D. harra, an overall percentage of correct classification of 95.3% was achieved, but once again,
the identification of subsp. lagascana (88.7%) was rather
lower than the others (Table 6 and Fig. 2).
Downloaded by [Oscar Grillo] at 02:12 11 April 2012
As expected when subspecies of the same species are
compared, the three taxa belonging to D. harra were mostly
misclassified, while most remaining errors accounted for D.
acris, D. brevisiliqua and D. simplex; three southern species
with generally smaller seeds, which might have similar dispersal strategies. However, whereas D. harra subsp. harra
and subsp. crassifolia did perform satisfactorily (96.7% and
89.9%, respectively), subsp. lagascana was less accurate,
Fig. 2. Graphic representation of the discriminant function scores for Diplotaxis harra.
Seed image analysis and taxonomy of Diplotaxis DC.
65
Table 7. Cross-validated percentages of correct identification for D. muralis species classifier at subspecies level. The number of seeds is
in parentheses.
Taxa
Downloaded by [Oscar Grillo] at 02:12 11 April 2012
D. muralis subsp. ceratophilla
D. muralis subsp. muralis
Overall
D. muralis subsp. ceratophilla
D. muralis subsp. muralis
Total
100.0% (163)
0
0
100.0% (169)
100.0% (163)
100.0% (169)
100.0% (332)
The North African taxon, D. muralis subsp. ceratophylla
(subgen. Diplotaxis), was only correctly identified to the
level of 66.9%. A high proportion of errors were due to
mistakes for the related species D. cretacea (20.9%), and
more than 10% were mistaken for species in subgen. Rhynchocarpum (D. assurgens, 4.3%; D. berthautii, 2.5%; D.
catholica, 3.7%; D. virgata, 0.6%), all of them also growing in northern Africa. The seeds of the type subspecies,
D. muralis subsp. muralis, did not perform much better,
scarcely achieving 71.0% of correct identification. Only
4.2% of the remaining seeds were mistaken for its close
relatives, D. tenuifolia (2.4%) and D. simplex (1.8%), while
most misidentified seeds were wrongly identified as a varied
array of taxa in the same subgenus (D. harra subsp. crassifolia, 5.3%) or in subgen. Rhynchocarpum (D. brevisiliqua,
5.3%; D. ibicensis, 5.3%; D. erucoides subsp. longisiliqua,
5.3%). Such dispersion in misattribution might be related to
the amphidiploid condition of D. muralis (n = 21, the only
known case of polyploidy in the genus), since polymorphism regularly associated with polyploidy might also be
expressed in seed morphology. Surprisingly enough, none
of the misidentified seeds of any of the two subspecies of D.
muralis was mistaken for the other subspecies. Moreover, as
shown in Table 7, comparing both taxa separately, the classifier achieved 100% correct identification. Such clear-cut
separation is more characteristic of distinct species, rather
than of closely related subspecies of the same species and
suggests that the taxonomic relationships between them
should be reconsidered.
The remaining species in subgen. Diplotaxis were more
satisfactorily identified. The seeds of D. simplex reached
90.7% of the correct identification, with 4.3% of the
misidentified seeds attributed to D. harra subsp. harra, in
the same subgenus which is north African. In the case of
D. tenuifolia, 77.5% of its seeds were well discriminated,
with a high proportion of misidentifications being due to
confusion with D. erucoides subsp. erucoides (16.8%), a
species belonging to subgen. Rhynchocarpum, but with a
more similar geographical distribution (mostly central and
southern Europe and the Middle East) and probably adapted
to more similar habitats. In fact, these are the two species
of Diplotaxis most frequently considered as weeds (Sans &
Masalles, 1994; Eschmann-Grupe et al., 2004). A quite satisfactory correct identification (88.1%) was attained with
D. cretacea seeds, and the remaining seeds (8.5%) being
mostly mistaken for those of D. muralis subsp. ceratophylla,
in the same subgenus. Interestingly, no mistakes occurred
between D. cretacea (n = 11) and the morphologically
similar D. tenuifolia (n = 11), which does not support the
subordination of the former to the latter proposed by Sobrino Vesperinas (1996). The seeds of D. viminea, the only
species in the genus that appears to be completely selffertilising, achieved the highest percentage of correct identification (97.3%). This high identification performance is
probably related to the homogeneity regularly associated
with autogamy.
In the case of D. tenuisiliqua subsp. rupestris the low
level of successful identification (64.5%) corresponds to
the partial confusion with subsp. tenuisiliqua (9.5%), but
mostly with D. virgata (19.0%) and also with D. berthautii (7%), both belonging to subgen. Rhynchocarpum sect.
Rhynchocarpum and closely related to D. tenuisiliqua
subsp. rupestris. As for subsp. tenuisiliqua, 77.8% of its
seeds were correctly determined, with all mistakes occurring with closely related taxa in the same section: D.
tenuisiliqua subsp. rupestris (9.7%), D. virgata (6.9%), D.
assurgens (2.8%) and D. berthautii (2.4%). In a separate
comparison between the two subspecies of D. tenuisiliqua,
an identification performance of 92.3% was reached (Table
8 and Fig. 3).
The identification of seeds of all other species in sect.
Rhynchocarpum was well above 70%. The germplasm of
Table 8. Cross-validated percentages of correct identification for D. tenisiliqua species classifier at subspecies level. The number of
seeds is in parentheses.
Taxa
D. tenisiliqua subsp. rupestris
D. tenisiliqua subsp. tenisiliqua
Overall
D. tenisiliqua subsp. rupestris
D. tenisiliqua subsp. tenisiliqua
Total
91.1% (426)
5.6% (28)
9.9% (47)
94.4% (468)
100.0% (473)
100.0% (496)
92.3% (969)
Downloaded by [Oscar Grillo] at 02:12 11 April 2012
66
O. Grillo et al.
Fig. 3. Graphic representation of the discriminant function scores for Diplotaxis siifolia.
D. catholica was quite satisfactorily identified (90.1% of
correctly identified seeds), as were those of D. berthautii
(86.5%), D. assurgens (82.4%) and of D. virgata (78.9%).
Most mistakes occurred among these four species, together
with D. tenuisiliqua. They are quite related in morphological traits, other than those of the seed, have the same
chromosome number (see Table 1) and also share partially similar habitats in the Iberian Peninsula–Morocco
region. The germplasm of D. brachycarpa was also rather
well identified (73.3%). Most errors in this case, however,
were due to confusion with D. erucoides subsp. longisiliqua
(22.0%), of section Heteropetalum, which in turn reached
83.4% of the correct identification, with 13% mistakes for
D. brachycarpa. These two taxa belong to different sections of subgen. Rhynchocarpum, but grow in a rather limited area in north-eastern Algeria. Since size (Harper et
al., 1970) and possibly other seed traits can be affected by
selective pressures, convergence might well have caused
this similarity between these two taxa growing in similar
habitats.
The seeds of D. siifolia were also well discriminated, with
identification performance between 80.8% and 89.2% for
the three subspecies and almost every wrong identification
was due to mistakes for one or another of them. Comparing the three taxa separately, the classifier gave an overall
correct identification of 92.0% (Table 9), with few mistakes
regularly distributed among the subspecies (Fig. 3). These
results are in accordance with previous observations that D.
siifolia is the only Diplotaxis species with globose, nearly
spherical seeds, more close in shape to those of Brassica
and other genera in the tribe Brassiceae, a unique feature already pointed out by Bengoechea & Gómez-Campo (1975).
Seed image analysis and taxonomy of Diplotaxis DC.
67
Table 9. Cross-validated percentages of correct identification for D. siifolia species classifier at subspecies level. The number of seeds is
in parentheses.
Taxa
Downloaded by [Oscar Grillo] at 02:12 11 April 2012
D. siifolia subsp. bipinnatifica
D. siifolia subsp. vicentina
D. siifolia subsp. siifolia
Overall
D. siifolia subsp. bipinnatifica
D. siifolia subsp. vicentina
D. siifolia subsp. siifolia
Total
93.0% (332)
6.4% (18)
2.4% (6)
5.6% (20)
87.5% (246)
2.0% (5)
1.4% (5)
6.0% (17)
95.6% (239)
100.0% (357)
100.0% (281)
100.0% (250)
92.0% (888)
Discrimination of D. ollivieri reached 78.3% of correctly identified seeds, with wrongly identified seeds
corresponding mostly to all four species of sect. Heterocarpum (D. siettiana, 7.2%; D. ilorcitana, 4.3%; D. ibicensis, 1.4%; D. brevisiliqua, 1.4%), and the remaining ones to
five other taxa. Convergence of seed traits due to similarity
of habitats, again, might be the cause for the misidentification among these taxa. However, D. ollivieri has never
been investigated for chromosome number or molecular
markers, and its morphological affinities are poorly understood because descriptions are based on scarce availability
of herbarium specimens. Its taxonomic position in sect.
Rhynchocarpum is therefore weakly sustained and somewhat uncertain, and does not seem to be supported by seed
morpho-colorimetric characters.
Seeds of all four species of section Heterocarpum
achieved percentages of correct identification ranging from
70.8% (D. ibicensis) to 83.2% (D. brevisiliqua). Furthermore, wrongly identified seeds have generally been mistaken for seeds of other species of the same section.
The only two taxa in section Heteropetalum are the
two subspecies of D. erucoides and both turned out to be
well identified in more than 80% of cases. Most errors
were due to misclassification for species from more similar habitats. Seeds of D. erucoides subsp. erucoides were
mainly misidentified for D. tenuifolia (7.0%) and D. harra
subsp. crassifolia (3.6%), both of them with brochidodromously veined petals and a native habitat in central and
southern Europe, as D. erucoides subsp. erucoides; and
only 4.1% of the seeds was mistaken for those of subsp.
longisiliqua, which grows in northern Algeria. On the other
hand, some of the seeds of D. erucoides subsp. longisiliqua were incorrectly identified as D. brachycarpa (13.7%).
Misidentifications between the two subspecies; however,
occurred only in a very small proportion; both were in
fact perfectly discriminated when compared separately
(Table 10).
Infrageneric classification
The distribution of Diplotaxis taxa in the three-dimensional
space, was determined by the first three discriminant functions (DF) derived from morpho-colorimetric parameters
obtained from seed image analysis (Fig. 4). Two major
groups of taxa are recognisable graphically.
This first cloud includes most taxa of sect. Rhynchocarpum, with the exceptions of D. siifolia subspecies,
D. ollivieri (both of questionable taxonomic position) and
D. brachycarpa, as well as a subset constituted by two additional taxa belonging to subgen. Diplotaxis: D. cretacea
and D. muralis subsp. ceratophylla. The latter appears quite
distant from subsp. muralis, which in turn is situated in the
other set of taxa, among the remaining subgen. Diplotaxis.
The fact that the two subspecies of D. muralis appear so
far apart reinforces the idea that, at least on the basis of
morpho-colorimetric traits of seeds, these two taxa might
well be considered as separate species. The striking placement of D. cretacea cannot be explained on the basis of
any affinity other than seed morphology, since according
to plant morphology, isozymes, chromosome numbers and
molecular markers, it is clearly much closer to other taxa
of the subgen. Diplotaxis.
The second major group is located basically in the opposite corner of the graph. It consists of all other taxa belonging to subgen. Diplotaxis, together with those of sect.
Heteropetalum, sect. Heterocarpum and subgen. Hesperidium, plus the previously mentioned D. brachycarpa, D. siifolia subspecies and D. ollivieri of sect. Rhynchocarpum.
The taxa belonging to the type subgenus were all found to
be included in a clade within the Rapa–Oleracea lineage
Table 10. Cross-validated percentages of correct identification for D. erucoides species classifier at subspecies level. The number of
seeds is in parentheses.
Taxa
D. erucoides subsp. longisiliqua
D. erucoides subsp. erucoides
Overall
D. erucoides subsp. longisiliqua
D. erucoides subsp. erucoides
Total
100.0% (483)
0.02% (1)
0
99.8% (443)
100.0% (483)
100.0% (444)
99.9% (927)
Downloaded by [Oscar Grillo] at 02:12 11 April 2012
68
O. Grillo et al.
Fig. 4. Graphic representation of the discriminant function scores for the studied Diplotaxis taxa.
(Warwick et al., 1992) and form a rather compact set, with
negative values of DF2, except for D. viminea, which due
to the self-fertilisation syndrome of D. viminea, makes it
morphologically different from its closest relatives, in many
aspects, and seed morpho-colorimetric characteristics may
well be among them.
Intermixed among taxa of subgen. Diplotaxis D. acris
also appears, whose separate status as subgen. Hesperidium
does not seem to be supported, therefore, by seed characters.
The two subspecies of D. erucoides were plotted close to
each other, but also among taxa of subgen. Diplotaxis, suggesting that sect. Heteropetalum, in terms of seed morphocolorimetric characteristics, is closer to the type subgen.
than to subgen. Rhynchocarpum. Similarly, both taxa in
this section, as well as those of subgen. Diplotaxis, were
found by Warwick et al. (1992) to belong to the same clade,
the Rapa–Oleracea lineage. It should be noted that subgen.
Diplotaxis, subgen. Hesperidium and sect. Heteropetalum
also share the brochidodromous venation of petals, in contrast with the cladodromous to eucamptodromous petals of
the remaining sections of subgen. Rhynchocarpum.
Sect. Heterocarpum is well supported on molecular
grounds (Warwick et al., 1992) and appears to be confirmed
as a taxonomic unit also on the basis of seed morphology, since its four species (D. brevisiliqua, D. ibicensis, D.
ilorcitana and D. siettiana) form a quite concise group positioned rather close to subgen. Diplotaxis. Subgenus Rhynchocarpum is also represented in this second major cloud by
sect. Heteropetalum (discussed above), D. ollivieri (of uncertain taxonomic position), D. brachycarpa and D. siifolia.
The three subspecies of D. siifolia (sect. Rhynchocarpum)
can still be found somewhat above sect. Heterocarpum.
The odd position of this species might well be related to
the spherical shape of its seeds, unique in the genus. Besides, D. siifolia has n = 10 chromosomes (a number only
shared with the not closely related D. viminea) instead of
n = 9 chromosomes, as the rest of sect. Rhynchocarpum.
Several authors (Gómez-Campo & Tortosa, 1974; Takahata & Hinata, 1983, 1986) have pointed out the difficulty
in accommodating D. siifolia, on morphological grounds,
in Diplotaxis as well as its affinities with Brassica or Erucastrum. The placement of D. brachycarpa close to subgen.
Diplotaxis seems to be solely founded on seed morphology,
since according to other morphological traits, chromosome
numbers and molecular markers, it is clearly much closer
to subgen. Rhynchocarpum, though it was found to form
a separate subclade within the Nigra lineage by Warwick
et al. (1992).
Conclusions
This work represents the first approach to investigate taxonomic relationships within the genus Diplotaxis by seed
phenetic characterisation using a seed image analysis system.
The results obtained by the implementation of the general database for this genus and the elaboration of dedicated
seed classifiers for the subspecies groups within this genus,
prove once again how image analysis techniques can be
considered a useful tool in taxonomic studies. In this case,
the application of this innovative kind of identification system allowed us to discriminate among most species and
Seed image analysis and taxonomy of Diplotaxis DC.
Downloaded by [Oscar Grillo] at 02:12 11 April 2012
subspecies with a considerably high percentage of correct
identification, as well as supporting the most recently proposed infrageneric grouping. In particular, the consistency
of section Heterocarpum, the particular position of section
Heteropetalum, and the rather isolated situation of D. siifolia with respect to the rest of the genus are supported by
this analysis.
A certain amount of misidentified seeds was also found,
as should always be expected considering that samples collected in the wild are heterogeneous, and consequently
may have different levels of maturation, may show intrapopulation genetic variation, and probably are subject to any
other endogenous and exogenous interactions or sources of
natural heterogeneity, as well as to plausible phenomena
of convergence among species adapted to similar environmental conditions.
Acknowledgements
The authors thank the “Regione Autonoma Sardegna” for
the support and the realization of this work, on the basis
of the Legge Regionale 7 agosto 2007, n. 7 “Promozione
della ricerca scientifica e dell’innovazione tecnologica in
Sardegna”, pubblica selezione per il conferimento di borse
di ricerca destinate a giovani ricercatori – Bando 2008.
References
AMIN, A. 1972. IOPB Chromosome number reports XXXVIII.
Taxon 21, 679–684.
BACCHETTA, G., FENU, G., GRILLO, O., MATTANA, E. & VENORA,
G. 2011a. Taxonomic identification by image analysis technique of seeds in the Astragalus Sect. Melanocercis Bunge
(Fabaceae) species of Sardinia. Annales Botanici Fennici 48,
449–454.
BACCHETTA, G., GARCÌA, P.E., GRILLO, O., MASCIA, F. & VENORA,
G. 2011b. Seed image analysis provides evidence of taxonomical differentiation within the Lavatera triloba aggregate
(Malvaceae). Flora 206, 468–472.
BACCHETTA, G., GRILLO, O., MATTANA, E. & VENORA, G. 2008.
Morpho-colorimetric characterization by image analysis to
identify diaspores of wild plant species. Flora 203, 669–
682.
BENGOECHEA, G. & GÓMEZ-CAMPO, C. 1975. Algunos caracteres
de la semilla en la tribu Brassiceae. Anales del Instituto Botanico A.J. Cavanilles 32, 793–841.
DANA, W. & IVO, W. 2008. Computer image analysis of seed
shape and seed color for flax cultivar description. Computer
and Electronics in Agriculture 61, 126–135.
ESCHMANN-GRUPE, G., HURKA, H. & NEUFFER, B. 2003. Species
relationships within Diplotaxis (Brassicaceae) and the phylogenetic origin of D. muralis. Plant Systematics and Evolution
243, 13–29.
ESCHMANN-GRUPE, G., NEUFFER, B. & HURKA, H. 2004. Extent
and structure of genetic variation in two colonising Diplotaxis
species (Brassicaceae) with contrasting breeding systems.
Plant Systematics and Evolution 244, 31–43.
FAGUNDEZ, J. & IZCO, J. 2004. Seed morphology of Calluna Salisb.
(Ericaceae). Acta Botanica Malacitana 29, 215–220.
FERNANDES, A. & QUEIRÓS, M. 1970–71. Sur la caryologie de
quelques plantes recoltées pendant la IIIème réunion de
69
botanique peninsulaire. Memória de Sociedade Broteriana
21, 343–385.
FISHER, R.A. 1936. The use of multiple measurements in taxonomic problems. Annals of Eugenics 7(2), 179–188.
FISHER, R.A. 1940. The precision of discriminant functions. Annals of Eugenics 10(4), 422–429.
GÓMEZ-CAMPO, C. 1980. Studies on Cruciferae, V. Chromosome
numbers for twenty-five taxa. Anales del Instituto Botánico
A.J. Cavanilles 35, 177–182.
GÓMEZ-CAMPO, C. & TORTOSA, M.E. 1974. The taxonomic and
evolutionary significance of some juvenile characters in the
tribe Brassiceae. Botanical Journal of the Linnean Society 69,
105–124.
GÓMEZ-CAMPO, C. & MARTÍNEZ-LABORDE, J.B. 1998. Reajustes
taxonómicos y nomenclaturales en la tribu Brassiceae (Cruciferae). Anales del Jardı́n Botánico de Madrid 56, 379–381.
GRANITTO, P.M., GARRALDA, P.A., VERDES, P.F. & CECCATO, H.A.
2003. Boosting classifiers for weed seeds identification. Journal of Computer Science and Technology 3, 34–39.
GRILLO, O., MATTANA, E., VENORA, G. & BACCHETTA, G. 2010.
Statistical seed classifiers of 10 plant families representative
of the Mediterranean vascular flora. Seed Science and Technology 38, 455–476.
GRILLO, O., MICELI, C. & VENORA, G. 2011. Computerised image analysis applied to inspection of vetch seeds for varietal
identification. Seed Science and Technology 39, 490–500.
HARBERD, D.J. 1972. A contribution to the cyto-taxonomy of Brassica (Cruciferae) and its allies. Botanical Journal of the Linnean Society 65, 1–23.
HARBERD, D.J. 1976. Cytotaxonomic studies of Brassica and related genera. In: VAUGHAN, J.G., MACLEOD, A.J. & JONES,
B.M.G., Eds., The Biology and Chemistry of the Cruciferae.
Academic Press, London, pp. 47–68.
HARBERD, D.J. & MCARTHUR, E.D. 1972. The chromosome constitution of Diplotaxis muralis (L.) DC. Watsonia 9, 131–135.
HARPER, J.L., LOVELL, P.H. & MOORE, K.G. 1970. The shapes and
sizes of seeds. Annual Review of Ecology and Systematics 1,
327–356.
KILIÇ, K., BOYACI, I.H., KOKSEL, H. & KUSMENOGLU, I. 2007. A
classification system or beans using computer vision system
and artificial neural networks. Journal of Food Engineering
78, 897–904.
LIAO, K., PAULSEN, M.R. & REID, J.F. 1994. Real-time detection of
colorimetric and surface defects of maize kernels using machine vision. Journal of Agricultural Engineering Research
59, 263–271.
MARTIN, A.C. 1946. The comparative internal morphology of
seeds. American Midland Naturalist Journal 36, 513–660.
MARTÍN, J.P. & SÁNCHEZ-YÉLAMO, M.D. 2000. Genetic relationships among species of the genus Diplotaxis (Brassicaceae)
using inter-simple sequence repeat markers. Theoretical and
Applied Genetics 101, 1234–1241.
MARTÍNEZ-LABORDE, J.B. 1988. Estudio sistemático del género
Diplotaxis DC. (Cruciferae-Brassiceae). PhD dissertation,
Universidad Politécnica de Madrid, Madrid, ES.
MARTÍNEZ-LABORDE, J.B. 1991. Two additional species of
Diplotaxis (Cruciferae, Brassiceae) with n = 8 chromosomes.
Willdenowia 21, 63–68.
PIGNONE, D. 1997. Present status of rocket genetic resources and
conservation activities. In: PADULOSI, S. & PIGNONE, D., Eds.,
Rocket: A Mediterranean Crop for the World. IPGRI, Rome,
pp. 2–12.
PIMPINI, F. & ENZO, M. 1997. Present status and prospects for
rocket cultivation in the Veneto region. In: PADULOSI, S. &
PIGNONE, D., Eds., Rocket: a Mediterranean Crop for the
World. IPGRI, Rome, pp. 51–66.
Downloaded by [Oscar Grillo] at 02:12 11 April 2012
70
O. Grillo et al.
PRADHAN, A.K., PRAKASH, S., MUKHOPADHYAY, A. & PENTAL, D.
1992. Phylogeny of Brassica and allied genera based on variation in chloroplast and mitochondrial DNA patterns: molecular and taxonomic classifications are incongruous. Theoretical
and Applied Genetics 85, 331–340.
PRAKASH, S., TAKAHATA, Y., KIRTI, P. & CHOPRA, V. 1999. Cytogenetics. In: GÓMEZ-CAMPO C., Ed., Biology of Brassica
Coenospecies. Elsevier, Amsterdam, pp. 59–106.
RUSTAN, Ø.H. 1996. Revision of the genus Diplotaxis (Brassicaceae) in the Cape Verde Islands, W Africa. Nordic Journal
of Botany 16, 19–50.
SÁNCHEZ-YÉLAMO, M.D. & MARTÍNEZ-LABORDE, J.B. 1991.
Chemotaxonomic approach to Diplotaxis muralis (Cruciferae:
Brassiceae) and related species. Biochemical Systematics and
Ecology 19, 477–482.
SANS, F.X. & MASALLES, R.M. 1994. Life-history variation in the
annual arable weed Diplotaxis erucoides (Cruciferae). Canadian Journal of Botany 72, 10–19.
SHAHIN, M.A. & SYMONS, S.J. 2003a. Lentil type identification
using machine vision. Canadian Biosystems Engineering 45,
3.5–3.10.
SHAHIN, M.A. & SYMONS, S.J. 2003b. Colour calibration of scanners for scanner independent grain grading. Cereal Chemistry
80, 285–289.
SMYKALOVA, I., GRILLO, O., BJELKOVA, M., HYBL, M. & VENORA,
G. 2011. Morpho-colorimetric traits of Pisum seeds measured
by an image analysis system. Seed Science and Technology
39, 612–626.
SOBRINO VESPERINAS, E. 1996. Posición taxonómica de Diplotaxis
cretacea Kotov (Cruciferae). Anales del Jardı́n Botánico de
Madrid 54, 182–188.
SPSS, 1999. Base 10.0 Application Guide. Prentice Hall, New
Jersey.
STEARN, W.T. 1980. Botanical Latin. David & Charles Publ., London.
TAKAHATA, Y. & HINATA, K. 1978. A description of the genetic
stocks in subtribe Brassicinae by chromosome numbers and
numerical characters. Cruciferae Newsletter 3, 47–51.
TAKAHATA, Y. & HINATA, K. 1983. Studies on cytodemes in subtribe Brassicinae (Cruciferae). Tohoku Journal of Agricultural
Research 33, 111–124.
TAKAHATA, Y. & HINATA, K. 1986. Consideration of the species
relationships in subtribe Brassicinae (Cruciferae) in view of
cluster analysis of morphological characters. Plant Species
Biology 1, 79–88.
VENORA, G., GRILLO, O., SHAHIN, M.A. & SYMONS, S.J. 2007.
Identification of Sicilian landraces and Canadian cultivars of
lentil using image analysis system. Food Research International 40, 161–166.
VENORA, G., GRILLO, O. & SACCONE, R. 2009a. Quality assessment of durum wheat storage centres in Sicily: Evaluation of
vitreous, starchy and shrunken kernels using an image analysis system. Journal of Cereal Science 49, 429–440.
VENORA, G., GRILLO, O., RAVALLI, C. & CREMONINI, R. 2009b.
Identification of Italian landraces of bean (Phaseolus vulgaris
L.) using an image analysis system. Scientia Horticulturae
121, 410–418.
WARWICK, S.L. & ANDERSON, J.K. 1997. Isozyme analysis of
parentage in allopolyploid Diplotaxis muralis (L.) DC. (Brassicaceae). Cruciferae Newsletter 19, 35–36.
WARWICK, S.L. & BLACK, L.D. 1991. Molecular systematics of
Brassica and allied genera (Subtribe Brassicinae, Brassiceae)
chloroplast genome and cytodeme congruence. Theoretical
and Applied Genetics 82, 81–92.
WARWICK, S.L., BLACK, L.D. & AGUINAGALDE, I. 1992. Molecular systematics of Brassica and allied genera (Subtribe
Brassicinae, Brassiceae) chloroplast DNA variation in the
genus Diplotaxis. Theoretical and Applied Genetics 83, 839–
850.
WARWICK, S.L., FRANCIS, A. & AL-SHEHBAZ, I.A. 2006. Brassicaceae: species checklist and database on CD-rom. Plant
Systematics and Evolution 259, 249–258.
WERKER, E. 1997. Seed Anatomy. Encyclopedia of Plant Anatomy,
Vol. 10. Borntraeger, Berlin.
Associate Editor: Charlie Jarvis