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