STATISTICAL SEED CLASSIFIERS
Grillo, O., Mattana, E., Venora, G. and Bacchetta, G. (2010), Seed Sci. & Technol., 38, 455-476
Statistical seed classifiers of 10 plant families representative of
the Mediterranean vascular flora
O. GRILLO1,2, E. MATTANA1, G. VENORA2 AND G. BACCHETTA1
1
2
Centro Conservazione Biodiversità (CCB). Dipartimento di Scienze Botaniche Università degli Studi di
Cagliari. V.le Sant’Ignazio da Laconi, 13 - 09123 Cagliari, Italy (E-mail: [email protected])
Stazione Sperimentale di Granicoltura per la Sicilia (SSG). Via Buganvillea, 20 - 95041 Caltagirone, Italy
(E-mail: [email protected])
(Accepted April 2010)
Summary
The aim of this study was to implement a seed morphometric and colorimetric database in order to develop
specific statistical classifiers for 10 families representative of the Mediterranean vascular flora. The 501 analysed
accessions consisted of 274 taxa belonging to 161 genera and 52 families. Images were acquired by a flatbed
scanner and then elaborated by a macro software program developed for the morpho-colorimetric measurements.
The mean seed weight was also calculated for each accession. Statistical classifiers were elaborated for genera and
specific and infraspecific taxa discrimination within the following families: Apiaceae, Asteraceae, Boraginaceae,
Brassicaceae, Caryophyllaceae, Cistaceae, Fabaceae, Lamiaceae, Poaceae and Scrophulariaceae. Such classifiers,
based on Linear Discriminant Analysis and checked by cross validation, showed a performance ranging from
90.4% to 98.5% and from 87.8% to 98.3% at genus and taxon levels respectively. A first application of this
technique to gymnosperm seeds was also presented. The performance of correct classification (93.6%) obtained
for the genus Juniperus suggests that this technology is also reliable for gymnosperms. Seed weight improved the
accuracy of the classifier system, confirming that an extensive database of bio-morphological traits constituted
by a large number of seed species may be useful for taxonomic identification.
Introduction
Although morphological seed traits such as shape, size and external ornamentation
represent important diagnostic factors in plant taxonomy studies (e.g. Thellung, 1926;
Valdés, 1970; Viano, 1978; Sutton, 1988), they are often not sufficiently considered
in dichotomous keys of many families and genera. In fact they are observable in the
field only for a short period, between the fruit and seed ripening and dispersal, when
other important characters, such as flowers and leaves, often have already disappeared.
However, the increasing availability of seeds collected from wild plants cultivated ex situ
(e.g. botanic gardens) and, above all, stored in seed banks emphasizes the importance of
seed macro and micro morphology studies in plant taxonomy. For example, Agulló et
al. (1991) carried out a morphological seed analysis in 17 taxa belonging to 5 genera of
the subfamily Papilionidae (Fabaceae) by evaluation of seed size, shape, external colour,
weight and protein content to investigate their taxonomic relationships. González-Andrés
455
O. GRILLO, E. MATTANA, G. VENORA AND G. BACCHETTA
et al. (1999) characterized three species belonging to the Medicago sect. Dendrotelis
(Fabaceae), using 12 seed and 20 seedling morphometric characters, showing that they
were more clearly differentiated on the basis of the seed characters. Kirkbride et al.
(2004) created an interactive key accessible via internet for the seed identification of 683
commercially important legume genera. Fagundez and Izco (2003, 2004a, b, c, 2006, 2009)
analysed seed micromorphology, size and colour to investigate the taxonomic position of
European species of Calluna, Daboecia and Erica genera (Ericaceae). Anderberg et al.
(2008) investigated seed shape, seed coat structure and surface patterns in 34 species
of the genus Lysimachia and in 14 species and 2 subspecies of 6 additional genera
(Anagallis, Ardisiandra, Asterolinon, Glaux, Pelletiera and Trientalis). Mapping the seed
characters onto a recent phylogenetic tree, the authors identified potentially synapomorphic
character states for various subclades of Lysimachia. Dadandi et al. (2009) elaborated an
identification key based on macro and micromorphological properties of intact and mature
seeds of 12 Turkish taxa belonging to the genus Nigella (Ranunculaceae) to assess the
systematic position of these species.
The application of image analysis techniques allows implementing quick, precise
and repeatable measurements of colour, size and shapes of the objects inside a digital
image (Serra, 1982). The cost reductions and the increased power of computer hardware
and software packages for image processing have made image analysis systems more
attractive in the automatic inspection of different aspects of quality evaluation and sorting
of agricultural products (Chen and Sun, 1991). Seeds have a three-dimensional (3D) shape,
while captured images displayed on a monitor or on printed pages are two-dimensional
(2D) (Loomis et al., 1999). Nevertheless, several descriptors of seed size, shape and Red,
Green and Blue (RGB) colour component densities can be easily estimated from 2D digital
images (Dell’Aquila, 2007). In a review on new computer imaging techniques applied to
seed quality testing and sorting, Dell’Aquila (2007) reported 2 main applications in seed
science: (1) to monitor seed germination and radical growth (e.g. Jansen, 1995; McDonald
et al., 2001; Dell’Aquila, 2006) and (2) define markers for seed identification and sorting
(e.g. Liao et al., 1994; Venora et al., 2007, 2009a, b; Dana and Ivo, 2008; Zapotoczny et
al., 2008).
Bacchetta et al. (2008) characterized digital images of seed lots of wild plant species
typical of the Mediterranean basin, stored at the Sardinian Germplasm Bank (BG-SAR)
(Mattana et al., 2005), using an image analysis system at the Stazione Sperimentale di
Granicoltura per la Sicilia (SSG). The analysed accessions referred to 148 taxa belonging
to 102 genera and 47 families. Images of diaspores (fruits and seeds) were acquired
by a flatbed scanner and elaborated by a specifically developed macro, to record 13
morphometric and colorimetric measurements of each seed in the acquired images. This
method allowed the development of a database for the characterization of autochthonous
germplasm upon entry into the seed bank and the realization of statistical classifiers
for the discrimination of genera and species within the following families: Apiaceae,
Boraginaceae, Caryophyllaceae, Cistaceae, Fabaceae and Scrophulariaceae. Such
classifiers, based on the Linear Discriminant Analysis (LDA) technique and checked by
cross-validation method, showed a performance ranging between 74.3% and 96.4%. This
system was later improved for the genus Astragalus (Mattana et al., 2008) by adding
456
STATISTICAL SEED CLASSIFIERS
20 new seed and fruit morphometric and colorimetric features. This new classifier had
a better performance, discriminating with 100% accuracy between two closely related
Astragalus species of Sardinia (A. maritimus and A. verrucosus) belonging to the subgenus
Trimeniaeus.
The aims of this study were: (1) to increase the size of the database developed in
previous works and evaluate the classifier identification performance with a larger dataset
and (2) to develop specific statistical classifiers at family level and test the system on the
genus Juniperus as a first application on gymnosperm seeds.
Materials and methods
Seed material and selection of the plant families
The 501 accessions analyzed in this study belonged to 274 taxa (species or lower ranks),
161 genera and 52 families, making a total of 47,493 fruits or seeds (hereafter referred
to as seeds), mainly collected in Sardinia during the harvesting seasons 2004-2008 and
stored at the BG-SAR (Appendix 1, see page 469). Some material collected elsewhere in
the Mediterranean basin (e.g. Italian and Iberian peninsulas, Corsica, Balearic and Hyères,
Crete and other Aegean islands) was also used, as well as seeds obtained by ex situ
cultivation in the Botanic Gardens of Cagliari or provided by other scientific institutions.
To create dedicated seed statistical classifiers, 10 plant families were selected based on
their representativeness of the Mediterranean vascular flora and the amount of available
seed lots stored at BG-SAR (table 1). These 10 selected families comprised more than 50%
of the genera, taxa, accessions, and seeds in the total database. Taxonomic classifications
were made according to the Angiosperm Phylogeny Group (APG II, 2003).
Table 1. Classifiers developed for 10 plant families representative of the Mediterranean vascular flora. The
representativeness of the ten family classifiers respect to the General classifier is reported in brackets.
Classifiers
General
Apiaceae
Families
Genera
Taxa
Accessions
Seeds
52
161
274
501
47,493
12
14
33
3,230
Boraginaceae
3
7
14
1,071
Caryophyllaceae
5
18
26
2,599
Cistaceae
4
13
25
2,510
Fabaceae
12
26
40
3,698
Scrophulariaceae
Asteraceae
2
7
9
902
24
33
49
4,682
Brassicaceae
7
9
18
1,811
Lamiaceae
8
15
22
2,169
Poaceae
Subtotal
11
88 (54.7%)
14
22
2,064
156 (56.9%)
258 (51.5%)
24,736 (52.1%)
457
O. GRILLO, E. MATTANA, G. VENORA AND G. BACCHETTA
Data acquisition
The images of seed samples were acquired according to Bacchetta et al. (2008) by a
flatbed scanner (Epson GT-15000) with a resolution of 200 dpi and a scanning area not
exceeding 1024×1024 pixel, before being dried at 15°C to 15% of R.H. to avoid any
possible variation in dimension, shape and colour. For the analysis of every accession,
a sub-lot of 100 seeds was randomly prepared; when an accession had fewer than 100
seeds, the analysis was done using the entire lot.
As the system works with 2D images and is therefore unable to distinguish between
globose and flattened seeds, the mean seed weight of each accession was also determined
by weighing air dried seeds on a four decimal place balance, as an estimate of the thickness
(3D) of the seed. A total of 34 seed characters were measured (table 2).
Table 2. List of thirty-four morphometric and colorimetric features measured on seeds. (*) indicates the new
feature (seed weight) not considered in Mattana et al., 2008.
Feature
Description
A
Area
Seed area (mm2)
P
Perimeter
Seed perimeter (mm)
Pconv
Convex Perimeter
Convex perimeter of the seed (mm)
PCrof
Crofton’s Perimeter
Crofton perimeter of the seed (mm)
Pconv /PCrof
Perimeter ratio
Ratio between convex and Crofton’s perimeters
Dmax
Max diameter
Maximum diameter of the seed (mm)
Dmin
Min diameter
Minimum diameter of the seed (mm)
Dmin /Dmax
Feret ratio
Ratio between minimum and maximum diameters
Sf
Shape Factor
Seed shape descriptor = (4 × π × area)/perimeter2 (normalized value)
Rf
Roundness Factor
Seed roundness descriptor = (4 × area)/(π × max diameter2)
(normalized value)
Ecd
Eq. circular diameter
Diameter of a circle with equivalent area (mm)
EAmax
Maximum ellipse axis
Maximum axis of an ellipse with equivalent area (mm)
EAmin
Minimum ellipse axis
Minimum axis of an ellipse with equivalent area (mm)
Rmean
Mean red channel
Red channel mean value of seed pixels (grey levels)
Rsd
Red std. deviation
Red channel standard deviation of seed pixels
Gmean
Mean green channel
Green channel mean value of seed pixels (grey levels)
Gsd
Green std. deviation
Green channel standard deviation of seed pixels
Bmean
Mean blue channel
Blue channel mean value of seed pixels (grey levels)
Bsd
Blue std. deviation
Blue channel standard deviation of seed pixels
Hmean
Mean hue channel
Hue channel mean value of seed pixels (grey levels)
Hsd
Hue std. deviation
Hue channel standard deviation of seed pixels
Lmean
Mean lightness ch.
Lightness channel mean value of seed pixels (grey levels)
Lsd
Lightness std. dev.
Lightness channel standard deviation of seed pixels
458
STATISTICAL SEED CLASSIFIERS
Table 2. continued
Smean
Mean saturation ch.
Saturation channel mean value of seed pixels (grey levels)
Ssd
Saturation std. dev.
Saturation channel standard deviation of seed pixels
Dmean
Mean density
Density channel mean value of seed pixels (grey levels)
Dsd
Density std. deviation
Density channel standard deviation of seed pixels
S
Skewness
Asymmetry degree of intensity values distribution (grey levels)
K
Kurtosis
Peakness degree of intensity values distribution (densit. units)
H
Energy
Measure of the increasing intensity power (densitometric units)
E
Entropy
Dispersion power (bit)
Dsum
Density sum
Sum of density values of the seed pixels (grey levels)
SqDsum
Square density sum
SW
( )
Seed weight *
Sum of the squares of density values (grey levels)
Mean seed weight (g)
Image processing
The KS-400 V. 3.0 (Carl Zeiss, Vision, Oberkochen, Germany) image analysis system
was used. To achieve the relative dimensions, shapes and colour of seeds (RGB – Red,
Green, Blue and HLS – Hue, Lightness, Saturation channels), the macro developed for
wild plant species characterization (Bacchetta et al., 2008) and later modified adding 20
new seed morpho-colorimetric features (Mattana et al., 2008) was used.
Classification statistics
Statistical analyses were performed using the stepwise Linear Discriminant Analysis
method (LDA) to determine the best classification variables. The procedure of cross
validation was applied to verify the performance of the developed classifiers. This method,
is useful for small populations of data without a broad group of new unknown cases. The
LDA method tests individual cases and classifies them on the basis of all the others (SPSS
release 15, SPSS Inc. 1989-2006).
The statistical analysis was carried out in two distinct phases. The whole database was
used to develop a general classifier to discriminate all the studied accessions, according
to families, genera and infraspecific taxa. Specific classifiers were then developed for
generic, specific and infraspecific taxa identifications, within each selected family.
Results and discussion
General classifier
The database of seed characteristics has grown since it was first published in 2008, with
many additional accessions included (table 3). After each intermediate analysis the validity
of the classification system and the representativeness of the database were evaluated
(table 3). Values highlighted that the increase of samples and analysed taxa in the April
2009 database (501 accessions) led to a reduction of the general classifier performance,
with a final percentage of correct identification of 54.2% at family level, 67.0% at genus
459
O. GRILLO, E. MATTANA, G. VENORA AND G. BACCHETTA
Table 3. Trend over time of the database dimensions and performances of the general classifier. With (*) data
from Bacchetta et al., 2008.
Elaborations
Accessions
Families
Genera
Taxa
108
34 - 70.0%
67 - 80.3%
103 - 85.1%
September 2006 *
179
45 - 63.0%
93 - 72.0%
146 - 75.3%
January 2007 (*)
220
47 - 57.4%
102 - 66.5%
148 - 72.5%
May 2006 (*)
( )
May 2008
319
51 - 65.1%
114 - 74.3%
182 - 76.4%
April 2009
501
52 - 54.2%
161 - 67.0%
274 - 69.3%
and 69.3% at taxa levels (table 3). The general classifier presented here achieved a lower
performance than in previous studies because of the huge variability in the data. As
considered in our previous report (Bacchetta et al., 2008), when the database becomes too
large, it is better to divide it into separate databases and classifiers at the family level.
Statistical family classifiers
To overcome the relatively low accuracy of the general classifier, 10 dedicated statistical
classifiers were employed for individual families representative of the Mediterranean
vascular flora (table 4). Performance of each developed classifier at genus and specific or
infraspecific levels, including the previous results achieved by Bacchetta et al., (2008).
Following APG (2003) classification, species belonging to the genera Digitalis, Erinus and
Linaria that in the previous work (Bacchetta et al., 2008) were considered as belonging
to the Scrophulariaceae family, were considered with the Plantaginaceae family in this
study. For the Scrophulariaceae family, an additional statistical analysis was also carried
out without considering the APG classification, to compare the achieved results with the
previous data (table 4).
Despite the number of accessions being more than doubled by increasing the number
of genera, species and infraspecific taxa in the 6 comparable families (Apiaceae,
Boraginaceae, Caryophyllaceae, Cistaceae, Fabaceae and Scropholariaceae), the
classification performance was always higher than in the previous classifiers, except for
the family Boraginaceae at genus level that had about a 1% of decrease in performance
(table 4). The results achieved for this family may be explained by a reduction of its
representativeness due to the addition of a new genus (Myosotis) represented by only one
accession of one taxon. The Scrophulariaceae classifier (sensu APG) showed the largest
improvement in classification performances both at genus and specific or infraspecific
levels, notwithstanding the lower number of samples and analysed taxa (table 4). These
results confirm the taxonomic distance between the genera belonging to Plantaginaceae (and
previously considered as Scrophulariaceae) and the genera Scrophularia and Verbascum.
In fact, the analysis carried out considering all these genera (table 4) achieved lower
results (79.2% and 85.3% at genus and specific or infraspecific levels, respectively).
Statistical classifiers were developed for four new families including Asteraceae,
Brassicaceae, Lamiaceae and Poaceae, which reached performances of correct classification
greater than 90.4% and 87.8% at genus and specific or infraspecific levels, respectively
(table 4). In particular, high percentages of correct classification, both at generic (98.1%)
460
STATISTICAL SEED CLASSIFIERS
Table 4. Performance percentages of family classifiers at genus and specific or infraspecific levels. With (*) data
from Bacchetta et al., 2008. For the Scrophulariaceae family, the performance percentages, without considering
the APG (Angiosperm Phylogeny Group) classification, are also reported in brackets.
Family classifier
Genus
Taxa
Apiaceae
95.6 * - 97.2
93.6 (*) - 94.0
Boraginaceae
96.4 (*) - 95.1
76.8 (*) - 88.3
Caryophyllaceae
89.9 (*) - 92.2
84.3 (*) - 97.0
Cistaceae
84.7 (*) - 93.3
88.8 (*) - 93.1
Fabaceae
Scrophulariaceae
( )
( )
82.6 * - 91.4
78.5 (*) - 91.3
74.3 (*)- (79.2) 91.9
74.4 (*) - (85.3) 96.9
Asteraceae
90.4
87.8
Brassicaceae
98.1
98.3
Lamiaceae
96.7
96.7
Poaceae
98.5
97.8
and specific or infraspecific (98.3%) levels, were achieved for Brassicaceae (tables 4,
5). Seed characters are considered some of the most useful diagnostic characters for
the taxonomy in this family, as previously reported in the literature (e.g. Vaughan and
Whitehouse, 1971; Bengoechea and Gomez Campo, 1975).
Key parameters
Evaluating the contribution of all the variables used by the discrimination algorithm
(LDA), it was possible to identify the features that, more than others, could be considered
peculiar to each family (table 6). When there are many predictors (features), the stepwise
method can be useful by automatically selecting the best variables to use in the model.
This method starts with a model that doesn’t include any of the predictors. At each step,
the predictor with the largest “F to enter” value that exceeds the entry criteria (F ≥ 3.84)
is added to the model. The “Tolerance” value is the proportion of a variable’s variance
not accounted for other independent variables in the equation. A variable with very low
tolerance contributes little information to a model. The “F to remove” value is useful for
describing what happens if a variable is removed from the current model. The features
(predictors) included in the first five steps of the stepwise LDA method, are reported
(table 6) together with the number of steps performed by the algorithm and the Tolerance
and “F to remove” values in brackets.
The feature Seed Weight, not included in the previous classifiers (Bacchetta et al.,
2008; Mattana et al., 2008), showed here its discriminatory power by achieving the highest
“F to remove” values. Indeed, this feature was always present in the first five parameters
considered by the discrimination algorithm, in all families’ classifiers at least at one level
(table 6). In particular, it was the first parameter at both levels for Lamiaceae and Poaceae,
and at specific or infraspecific levels for Asteraceae, Boraginaceae, Caryophyllaceae
and Cistaceae (table 6). Distinguishing between morphometric and colorimetric features
461
462
100.0% (100)
Erysimum L.
100.0% (100)
Cakile Mill.
100.0% (200)
Iberis L.
99.0% (99)
3.5% (32)
Carrichtera DC.
100.0% (300)
Thlaspi L.
100.0% (100)
0.2% (2)
Malcolmia
Spreng.
98.1% (1810)
(100)
(300)
(100)
(200)
(100)
(100)
(910)
Total
Caryophyllaceae
Boraginaceae
Apiaceae
E
(29; 0.474; 213.058)
E
(29; 0.494; 163.297)
A
(24; 0.008; 275.175)
Rmean
(25; 0.018; 12.308)
EAmin
(27; 0.019; 7.131)
Pconv /PCrof
(29; 0.129; 105.528)
Pconv /PCrof
(29; 0.282; 164.116)
E
(24; 0.120; 15.363)
SW
(25; 0.546; 498.418)
Dsum
(27; 0.003; 33.439)
SW
(29; 0.584; 5,331.088)
(genus level)
(species or
infrasp. level)
(genus level)
(species or
infrasp. level)
(genus level)
(species or
infrasp. level)
Pconv
(29; 0.002; 13.171)
2
1
Classifier
Lsd
(29; 0.003; 55.220)
Bmean
(27; 0.031; 37.430)
Hsd
(27; 0.133; 94.565)
Bsd
(29; 0.019; 35.634)
Ssd
(25; 0.086; 32.394)
SW
(24; 0.389; 49.319)
A
(29; 0.010; 45.905)
Rsd
(27; 0.008; 9.546)
A
(25; 0.010; 87.080 )
Ssd
(24; 0.107; 43.896)
Hmean
(29; 0.0083; 30.673)
Sf
(29; 0.076; 80.414)
Hmean
(29; 0.095; 34.037)
Ssd
(29; 0.123; 44.555)
5
4
E
(25; 0.196; 7.596)
Ecd
(24; 0.006; 204.346)
SW
(29; 0.599; 536.105)
Ssd
(29; 0.123; 51.423)
3
Table 6. The best five key parameters (see table 2 for the legend of the features) for the correct classifications are reported. The number of steps, the tolerance
and F to remove values are reported in brackets.
Overall
Malcolmia Spreng.
Thlaspi L.
Carrichtera DC.
Iberis L.
Cakile Mill.
1.0% (1)
96.3% (876)
Brassica L.
Erysimum L.
Brassica L.
Genus
Table 5. Cross validated performances for Brassicaceae. Data in brackets correspond to the number of seeds.
O. GRILLO, E. MATTANA, G. VENORA AND G. BACCHETTA
Poaceae
Lamiaceae
Brassicaceae
Asteraceae
Scrophulariaceae
Fabaceae
Cistaceae
Table 6. continued
Bmean
(26; 0.069; 26.227)
SW
(29; 0.180; 278.205)
Smean
(29; 0.047; 53.319)
Gmean
(12; 0.013; 30.652)
Dsum
(21; 0.017; 27.606)
A
(31; 0.008; 47.212)
EAmax
(31; 0.016; 61.423)
Hsd
(24; 0.193; 36.921)
E
(28; 0.102; 8.674)
E
(28; 0.125; 17.731)
E
(31; 0.270; 21.797)
Hmean
(30; 0.061; 43.812)
E
(31; 0.268; 14.343)
SW
(26; 0.678; 487.142)
Sf
(29; 0.147; 39.272)
Sf
(29; 0.194; 42.348)
Rmean
(12; 0.044; 220.113)
Ssd
(21; 0.468; 15.970)
E
(31; 0.098; 65.570)
SW
(31; 0.520; 416.915)
Sf
(24; 0.171; 88.543)
Dmin /Dmax
(28; 0.207; 72.970)
SW
(28; 0.480; 256.786)
SW
(31; 0.673; 895.191)
SW
(30; 0.345; 854.607)
SW
(31; 0.738; 1,313.390)
(species or
infrasp. level)
(genus level)
(species or
infrasp. level)
(genus level)
(species or
infrasp. level)
(genus level)
(species or
infrasp. level)
(genus level)
(species or
infrasp. level)
(genus level)
(species or
infrasp. level)
(genus level)
(species or
infrasp. level)
Pconv /PCrof
(26; 0.161; 17.836)
Bmean
(26; 0.041; 11.673)
(genus level)
E
(30; 0.103; 19.216)
Hmean
(31; 0.109; 41.242)
Ssd
(28; 0.242; 112.283)
Rf
(31; 0.028; 15.099)
EAmax
(24; 0.004; 53.704)
Bsd
(28; 0.030; 77.162)
Dsum
(31; 0.003; 79.652)
Sf
(31; 0.056; 113.770)
Rmean
(21; 0.003; 22.165)
SW
(12; 0.568; 249.120)
E
(29; 0.235; 133.931)
Smean
(29; 0.032; 51.356)
Bsd
(26; 0.015; 53.695)
Rsd
(26; 0.008; 59.257)
EAmax
(30; 0.006; 13.945)
Rsd
(1; 0.015; 16.336)
Dmean
(28; 0.118; 168.874)
Ssd
31; 0.239; 77.794)
EAmin
(24; 0.002; 33.051)
Rsd
(28; 0.020; 35.265)
Gmean
(31; 0.021; 125.992)
Lmean
(31; 0.015; 52.288)
Hmean
(21; 0.148; 26.030)
Smean
(12; 0.239; 65.963)
Bsd
(29; 0.057; 74.343)
Bsd
(29; 0.034; 102.497)
Rmean
(26; 0.019; 315.968)
Rmean
(26; 0.016; 336.206)
Rsd
(30; 0.014; 16.888 )
EAmax
(31; 0.008; 10.821 )
A
(28; 0.017; 224.755)
Dmean
(31; 0.034; 19.521 )
SW
(24; 0.008; 646.148 )
SW
(28; 0.121; 505.188 )
E
(31; 0.126; 59.059 )
EAmax
(31; 0.010; 83.517 )
SW
(21; 0.191; 811.971 )
PCrof
(12; 0.348; 89.877 )
Bmean
(29; 0.082; 79.749 )
Gsd
(29; 0.061; 216.520 )
Gmean
(26; 0.025; 143.990 )
Bsd
(26; 0.010; 29.778 )
STATISTICAL SEED CLASSIFIERS
463
O. GRILLO, E. MATTANA, G. VENORA AND G. BACCHETTA
(table 2), the latter were the more represented among the first five parameters, even if
morphometric features were the first discriminant parameters at both levels for Apiaceae,
Brassicaceae and Fabaceae (table 6). Cistaceae and Scrophulariacaeae were the only two
families for which there were non-morphological parameters in the first step at specific
or infraspecific levels. These data suggest that both types of measurements should be
considered when trying to achieve high percentages of correct classification, confirming
the findings of Bacchetta et al. (2008) and highlighting as well that the integration of
mean seed weight values improved the performance of the system.
Gymnosperms seeds
The percentage of correct identification of the species belonging to the genus Juniperus is
reported (table 7). Using the images of 1,700 seeds of 8 different taxa, the cross-validated
classification performance was 93.6%. Figure 1 shows the first three function scores used
to distinguish the taxa belonging to the genus, explaining 96.4% of the variability.
The high percentage of correct classification between the two subspecies of J.
phoenicea (98.8%), confirmed the differentiation at subspecies level (Amaral Franco
do, 1986; Conti et al., 2005; Jeanmonod and Gamisans, 2007) or species level (RivasMartínez et al., 1993; Lebreton and Pérez de Paz, 2001) for these two taxa (table 8).
The 89.7% of correct classification of J. communis subsp. communis and J. communis
subsp. alpina (table 9) confirmed the taxonomic distance between them. These two taxa
have been considered by several authors as two different subspecies (Amaral Franco do,
1980, 1986; Jeanmonod and Gamisans, 2007) or species (Pignatti, 1982; Lebreton et al.,
2000), but were recently considered as the same taxon by a few authors (Farjon, 2001;
Adams, 2004).
These findings for the genus Juniperus suggest that this seed classifier technology,
based on quick and cheap acquisitions of morphometric and colorimetric feature
measurements of seed accessions upon their entry to a seed bank, is also reliable for
gymnosperms.
J.
J.
J.
J.
J.
J.
J.
Discriminant Fu
nction 2
8
6
4
2
0
communis subsp. communis
phoenicea subsp. phoenicea
phoenicea subsp. turbinata
oxycedrus subsp. macrocarpa
communis subsp. alpina
sabina
thurifera
-2
-4
-6
-8
-10
0
20
15 n 3
10
tio
5
un
Disc 50 100
0
tF
n
rimi
-5
nan 150
ina
t Fu
ncti 200 250 -10 scrim
on 1
Di
c
Figure 1. Graphic representation of the discriminating function scores for the genus Juniperus.
464
4.0% (4)
J. sabina L. (7)
Overall
J. thurifera L. (8)
0.5% (1)
13.0% (26)
0.4% (2)
84.0% (84)
1
J. oxycedrus L. subsp.
oxycedrus (6)
J. communis subsp.
alpina Čelak (5)
J. oxycedrus L. subsp.
macrocarpa (Sibth.
& Sm.) Neilr. (4)
J. phoenicea L. subsp.
turbinata (Guss.)
Nyman (3)
J. phoenicea L. subsp.
phoenicea (2)
J. communis L. subsp.
communis (1)
Taxa
97.0% (97)
2
8.0% (16)
7.0% (14)
96.6% (483)
3.0% (3)
1.0% (1)
3
100.0% (400)
4
1.5% (3)
80.0% (160)
1.4% (7)
15.0% (15)
5
2.0% (2)
86.5% (173)
1.6% (8)
6
94.0% (94)
3.5% (7)
7
100.0% (100)
8
Table 7. Cross validated performances of the classifier developed for the genus Juniperus. Data in brackets correspond to the number of seeds.
93.6% (1700)
(100)
(100)
(200)
(200)
(400)
(500)
(100)
(100)
Total
STATISTICAL SEED CLASSIFIERS
465
O. GRILLO, E. MATTANA, G. VENORA AND G. BACCHETTA
Table 8. Cross validated performances between the two subspecies of Juniperus phoenicea. Data in brackets
correspond to the number of seeds.
Taxa
Juniperus phoenicea L.
subsp. phoenicea
J. phoenicea L. subsp.
phoenicea
J. phoenicea L. subsp.
turbinata (Guss.) Nyman
Total
93.0% (93)
7.0% (7)
(100)
100.0% (500)
(500)
Juniperus phoenicea L.
subsp. turbinata (Guss.) Nyman
Overall
98.8% (600)
Table 9. Cross validated performances between the two subspecies of Juniperus communis. Data in brackets
correspond to the number of seeds.
Juniperus communis L.
subsp. communis
Juniperus communis L.
subsp. alpina Č elak
Total
Juniperus communis L. subsp.
communis
84.0% (84)
16.0% (16)
(100)
Juniperus communis L. subsp.
alpina Č elak
8.0% (16)
92.0% (184)
(200)
Taxa
Overall
89.7% (300)
Conclusions
This work validated the statistical classification system explored in previous works
(Bacchetta et al., 2008, Mattana et al., 2008) by the implementation of a general database
and the elaboration of a dedicated seed classifier for each of ten families representative of
the Mediterranean vascular flora. These results confirm, as stated by Dell’Aquila (2007),
that an extensive database of bio-morphological traits constituted by a large number of
seed species may be applied also for sorting and taxonomy screening.
A further step of this work will be to improve the classifiers adding other features
related to the exact shape of the seeds with a morphological method based on the elliptic
Fourier descriptors of approximation to closed contours in a two-dimensional plane. In
addition, the availability of morphometric data should be helpful for ecological studies
such as the prediction of seed persistence in the soil (e.g. Thompson et al., 1993;
Cerabolini et al., 2003).
Acknowledgements
The authors wish to thank the BG-SAR staff for field work and help on image acquisition
and all the institutions that kindly provided seeds or their images: CBNM de Porquerolles;
DGMN de Murcia; CIEF Comunidad Valenciana; Jardí Botànic de la Universitat de
València; Institut y Jardí Botànic de Barcelona; Museo di Storia Naturale di Livorno;
466
STATISTICAL SEED CLASSIFIERS
CNBF Peri; BGVA Cordoba; BG-MOL Università degli Studi del Molise; Dipartimento
di Scienze Botaniche, Ecologiche e Geologiche Università degli Studi di Sassari and SCD
Kew Gardens. The CCB is supported by the “Provincia di Cagliari - Assessorato Tutela
Ambiente”. This research has also been supported by the P.R.I.N. 2007 project funds
(2007JNJ7MX-002).
References
Adams, R.P. (2004). Junipers of the World: the Genus Juniperus, Trafford Publishing Co., Vancouver, 282pp.
Agulló, M.A., Tortosa, M.E., Ceresuela, J.L. and Ortiz, J.M. (1991). Seed morphology in several taxa of the
subfamily Papilionidae (Fabaceae). Botanika Chronika, 10, 651-660.
Amaral Franco do, J. (1980). Juniperus L. In Flora Europaea, (eds. T.G. Tutin, V.H. Heywood, N.A. Burges,
D.H. Valentine, S.M. Walters and D.A. Webb), vol. 1, pp. 38-39, Cambridge University Press, Cambridge.
Amaral Franco do, J. (1986). Juniperus L. In Flora Iberica (eds. S. Castroviejo, M. Laínz, G. López González,
P. Montserrat, F. Muños Garmendía, J. Paiva, L. Villar), vol. 1, pp. 181-188, Real Jardín Botánico - CSIC,
Madrid.
Anderberg, A.L., Schonenberger, J. and Anderberg, A.A. (2008). Comparative seed morphology and character
evolution in the genus Lysimachia (Myrsinaceae) and related taxa. Plant Systematics and Evolution, 271,
177-197.
APG (2003). An update of the Angiosperm Phylogeny Group classification for the orders and families of
flowering plants: APG II. Botanical Journal of the Linnean Society, 141, 399-436.
Bacchetta, G., Grillo, O., Mattana, E. and Venora, G. (2008). Morpho-colorimetric characterization by image
analysis to identify diaspores of wild plant species. Flora, 203, 669-682.
Bengoechea, G. and Gomez Campo, C. (1975). Algunos caracteres de la semilla en la tribu Brassiceae. Anales
Instituto Botanico Cavanilles, 32, 793-841.
Cerabolini, B., Ceriani, R.M., Caccianiga, M., De Andreis, R. and Raimondi, B. (2003). Seed size, shape and
persistence in soil: a test on Italian flora from Alps to Mediterranean coasts. Seed Science Research, 13,
75-85.
Chen, P. and Sun, Z. (1991). A review of non-destructive methods for quality evaluation and sorting of
agricultural products. Journal of Agricultural Engineering Research, 49, 85-98.
Conti, F., Abbate, G., Alessandrini, A. and Blasi, C. (eds.) (2005). An Annotated Checklist of the Italian
Vascular Flora. Palombi Editori, Roma, 420pp.
Dadandi, M.Y., Kokdil, G., Ilcim, A. and Ozbilgin, B. (2009). Seed macro and micro morphology of the selected
Nigella (Ranunculaceae) taxa from Turkey and their systematic significance. Biologia, 64, 261-270.
Dana, W. and Ivo, W. (2008). Computer image analysis of seed shape and seed color for flax cultivar description.
Computers and Electronics in Agriculture, 61, 126-135.
Dell’Aquila, A. (2006). Computerised seed imaging: a new tool to evaluate germination quality. Communications
in Biometry and Crop Sciences, 1, 20-31.
Dell’Aquila, A. (2007). Towards new computer imaging techniques applied to seed quality testing and sorting.
Seed Science and Technology, 35, 519-538.
Fagundez, J. and Izco, J. (2003). Seed morphology of Erica L. sect. Chlorocodon Bentham. Acta Botanica
Gallica, 150, 401-410.
Fagundez, J. and Izco, J. (2004a). Seed morphology of Calluna Salisb. (Ericaceae). Acta Botanica Malacitana,
29, 215-220.
Fagundez, J. and Izco, J. (2004b). Seed morphology of Daboecia (Ericaceae). Belgian Journal of Botany, 137,
188-192.
Fagundez, J. and Izco, J. (2004c). Taxonomic value of seed characters in the Erica tetralix L. group (Ericaceae).
Plant Biosystems, 138, 207-213.
Fagundez, J. and Izco, J. (2006). A new European heather: Erica lusitanica subsp. cantabrica subsp. nova
(Ericaceae). Nordic Journal of Botany, 24, 389-394.
Fagundez, J. and Izco, J. (2009). Seed morphology of Erica L. sect. Loxomeria Salisb. ex Benth., sect.
Eremocallis Salisb. ex Benth. and sect. Brachycallis I. Hansen, and its systematic implications. Plant
Biosystems, 143, 238-336.
467
O. GRILLO, E. MATTANA, G. VENORA AND G. BACCHETTA
Farjon, A. (2001). World checklist and bibliography of Conifers, 2nd ed. Royal Botanic Gardens, Kew, 309pp.
González-Andrés, F., Chávez, J., Montáñez, G. and Ceresuela, J.L. (1999). Characterisation of woody Medicago
(sect. Dendrotelis) species, on the basis of seed and seedling morphometry. Genetic Resources and Crop
Evolution, 46, 505-519.
Jansen, P.I. (1995). Seed production quality in Trifolium balansae and T. resupinatum: The role of seed colour.
Seed Science and Techology, 23, 353-364.
Jeanmonod, D. and Gamisans, J. (2007). Flora Corsica. Edisud, Aix en Provence, 921pp.
Kirkbride, J.H., Gunn, C.R., Dallwitz, M.J. (2004). Interactive identification of genus of legume (Fabaceae)
fruits and seeds via the internet. Acta Horticulturae Proceedings, 631, 55-61.
Lebreton, P. and Pérez de Paz, P.L. (2001). Définition du Genévrier de Phénicie (Juniperus aggr. phoenicea)
reconsidéré à ses limites biogéographiques: Méditerranéeorientale (Crète et Chypre) et Atlantique (Iles
Canaries). Bulletin mensuel de la Société Linnéenne de Lyon, 70, 73-92.
Lebreton, P., Mossa, L. and Gallet, C. (2000). A propos d'une entité corso-sarde du genévrier nain. Bulletin
mensuel de la Société Linnéenne de Lyon, 69, 133-141.
Liao, K., Paulsen, M.R. and Reid, J.F. (1994). Real-time detection of colour and surface defects of maize kernels
using machine vision. Journal of Agricultural Engineering Research, 59, 263-271.
Loomis, J.J., Fujimura, K., McDonald, M., James, D. and Bennett, M. (1999). Using computer graphics for
three-dimensional seed cataloging. Seed Science and Technology, 27, 439-446.
Mattana, E., Fenu, G. and Bacchetta, G. (2005). La Banca del Germoplasma della Sardegna (BG-SAR): uno
strumento per la conservazione del germoplasma autoctono sardo. Informatore Botanico Italiano, 37, 145146.
Mattana, E., Grillo, O., Venora, G. and Bacchetta, G. (2008). Germplasm image analysis of Astragalus maritimus
and A. verrucosus of Sardinia (subgen. Trimeniaeus, Fabaceae). Anales Jardin Botanico de Madrid, 65, 149155.
McDonald, M.B., Evans, A.F. and Bennett, M.A. (2001). Using scanners to improve seed and seedling
evaluations. Seed Science and Technology, 29, 683-689.
Pignatti, S. (1982). Flora d’Italia, vol. 1. Edagricole, Bologna, 790pp.
Rivas-Martínez, S., Wildpret, W. and Pérez de Paz, P.L. (1993). Datos sobre Juniperus phoenicea aggr.
(Cupressaceae). Itinera Geobotanica, 7, 509-512.
Serra, J. (1982). Image analysis and mathematical morphology, vol. 1. Academic Press, London, 609pp.
Sutton, D.A. (1988). A revision of the tribe Antirrhineae. Oxford University Press, London, 575pp.
Thellung, A. (1926). Umbelliferae (Morison) Juss. In Illustrierte Flora von Mitteleuropa (ed. G. Hegi), vol. 2,
pp. 926-1537. Pichler, Munich.
Thompson, K., Band, S.R. and Hodgson, J.G. (1993). Seeds size and shape predict persistence in soil. Functional
Ecology, 7, 236-241.
Valdés, B. (1970). Revisión de las especies europeas de Linaria con semillas aladas. Anales de la Universidad
Hispalense, Series: Ciencias, 7. Universidad de Sevilla, Sevilla, 25pp.
Vaughan, J.G. and Whitehouse J.M. (1971). Seed structure and the taxonomy of the Cruciferae. Botanical
Journal of the Linnean Society, 64, 383-409.
Venora, G., Grillo, O. and 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. and Cremonini R. (2009b). Identification of Italian landraces of bean
(Phaseolus vulgaris L.) using an image analysis system. Scientia Horticulturae, 121, 410-418.
Venora, G., Grillo, O., Shahin, M.A. and Symons, S.J. (2007). Identification of Sicilian landraces and Canadian
cultivars of lentil using image analysis system. Food Research International, 40, 161-166.
Viano, J. (1978). Les linaries à graines aptères du bassin méditerranéen occidental, 2. Linaria sect. Elegantes,
Bipunctatae, Diffusae, Speciosae, Repentes. Candollea, 33, 209-267.
Zapotoczny, P., Zielinska, M. and Nita, Z. (2008). Application of image analysis for the varietal classification
of barley: Morphological features. Journal of Cereal Science, 48, 104-110.
468
STATISTICAL SEED CLASSIFIERS
Appendix: List of the analysed taxa grouped by family, following APG (2003).
Taxa
Family (APG II)
Accessions
Achyranthes sicula L. (All.)
Amaranthaceae
2
Seeds (n)
200
Chenopodium bonus-henricus L.
Amaranthaceae
1
100
Halopeplis amplexicaulis (Vahl) Ces., Pass. & Gibelli
Amaranthaceae
1
101
Salicornia dolichostachya Moss
Amaranthaceae
1
101
Salicornia emerici Duval-Jouve
Amaranthaceae
1
100
Salicornia patula Duval-Jouve
Amaranthaceae
1
100
Sarcocornia fruticosa (L.) A.J. Scott
Amaranthaceae
1
100
Suaeda vera J.F. Gmel.
Amaranthaceae
1
100
Pancratium maritimum L.
Amaryllidaceae
4
349
Bunium corydalinum DC. subsp. corydalinum
Apiaceae
1
85
Bupleurum fruticosum L.
Apiaceae
1
100
Crithmum maritimum L.
Apiaceae
3
300
Daucus carota L.
Apiaceae
1
100
Eryngium maritimum L.
Apiaceae
1
100
Ferula arrigonii Bocchieri
Apiaceae
9
900
Ferula communis L. s.l.
Apiaceae
4
387
Ferula fontqueri Jury
Apiaceae
1
58
Laserpitium siler L. subsp. garganicum (Ten.) Arcang.
Apiaceae
1
100
Pastinaca divaricata Desf.
Apiaceae
1
100
Pseudorlaya pumila (L.) Grande
Apiaceae
1
100
Ptychotis sardoa Pignatti & Metlesics
Apiaceae
3
300
Rouya polygama (Desf.) Coincy
Apiaceae
1
100
Seseli praecox (Gamisans) Gamisans
Apiaceae
5
500
Arum pictum L. f. subsp. pictum
Araceae
1
100
Biarum dispar (Schott) Talavera
Araceae
1
100
Helicodiceros muscivorus (L. f.) Engl.
Araceae
3
301
Asphodelus fistulosus L.
Asphodelaceae
1
100
Anthemis maritima L.
Asteraceae
2
200
Artemisia campestris L. subsp. variabilis (Ten.) Greuter
Asteraceae
1
99
Asteriscus maritimus (L.) Less.
Asteraceae
1
100
Bellium crassifolium Moris
Asteraceae
1
100
Buphthalmum inuloides Moris
Asteraceae
1
100
Carduus fasciculiflorus Viv.
Asteraceae
3
300
Carduus nutans L. s.l.
Asteraceae
2
199
Carlina corymbosa L.
Asteraceae
1
100
Centaurea corensis Vals. & Filigh.
Asteraceae
1
100
Centaurea filiformis Viv. subsp. filiformis
Asteraceae
1
94
Centaurea horrida Badarò
Asteraceae
1
95
469
O. GRILLO, E. MATTANA, G. VENORA AND G. BACCHETTA
Appendix continued
Taxa
Family (APG II)
Centaurea magistrorum Arrigoni & Camarda
Asteraceae
Accessions
1
Seeds (n)
Crupina crupinastrum (Moris) Vis.
Asteraceae
1
30
Dittrichia viscosa (L.) Greuter s.l.
Asteraceae
1
100
Eupatorium cannabinum L. subsp. corsicum (Loisel.) P. Fourn.
Asteraceae
1
100
Filago pygmaea L.
Asteraceae
1
108
Helichrysum italicum (Roth) G. Don subsp. italicum
Asteraceae
1
100
Helichrysum microphyllum Willd. subsp. tyrrhenicum Bacch.,
Brullo & Giusso
Asteraceae
1
101
Helichrysum saxatile Moris subsp. morisianum Bacch.,
Brullo & Mossa
Asteraceae
1
100
Lactuca longidentata Moris
Asteraceae
1
100
Lamyropsis microcephala Moris
Asteraceae
5
426
Limbarda crithmoides (L.) Dumort. s.l.
Asteraceae
2
197
Otanthus maritimus (L.) Hoffmanns & Link subsp. maritimus
Asteraceae
4
400
Ptilostemon casabonae (L.) Greuter
Asteraceae
2
200
Ptilostemon chamaepeuce (L.) Less.
Asteraceae
1
87
Pulicaria vulgaris Gaertn. var. sardoa Fiori
Asteraceae
1
100
Rhaponticum coniferum (L.) Greuter
Asteraceae
2
147
Robertia taraxacoides (Loisel.) DC.
Asteraceae
1
99
Santolina corsica Jord. & Fourr.
Asteraceae
1
100
Santolina insularis (Gennari ex Fiori) Arrigoni
Asteraceae
3
300
Senecio cineraria DC.
Asteraceae
1
100
Senecio rodriguezii Willk. ex Rodrig.
Asteraceae
1
100
Tanacetum audibertii (Req.) DC.
Asteraceae
1
100
Anchusa capellii Moris
Boraginaceae
1
100
Anchusa crispa Viv. subsp. maritima (Vals.) Selvi & Bigazzi
Boraginaceae
2
121
Anchusa formosa Selvi, Bigazzi & Bacch.
Boraginaceae
3
220
Anchusa littorea Moris
Boraginaceae
2
173
Anchusa montelinasana Angius, Pontecorvo & Selvi
ex Bacch. et al.
Boraginaceae
2
200
177
100
Borago pygmaea (DC.) Chater & Greuter
Boraginaceae
3
Myosotis soleirolii Godr.
Boraginaceae
1
80
Brassica insularis Moris
Brassicaceae
8
801
Brassica tournefortii Gouan
Brassicaceae
1
110
Cakile maritima Scop. subsp. maritima
Brassicaceae
1
100
Carrichtera annua (L.) DC.
Brassicaceae
1
100
Erysimum majellense Polatschek
Brassicaceae
1
100
Iberis integerrima Moris
Brassicaceae
2
200
470
STATISTICAL SEED CLASSIFIERS
Appendix continued
Taxa
Family (APG II)
Malcolmia ramosissima (Desf.) Gennari
Brassicaceae
Accessions
1
Seeds (n)
100
Thlaspi brevistylum (DC.) Mutel
Brassicaceae
1
100
Thlaspi perfoliatum L. s.l.
Brassicaceae
2
200
Campanula forsythii (Arcang.) Bég.
Campanulaceae
2
201
Arenaria bertolonii Fiori
Caryophyllaceae
1
100
Cerastium boisserianum Greuter & Burdet
Caryophyllaceae
1
100
Cerastium soleirolii Duby
Caryophyllaceae
1
100
Cerastium supramontanum Arrigoni
Caryophyllaceae
1
100
Dianthus insularis Bacch., Brullo, Casti & Giusso
Caryophyllaceae
2
200
Dianthus morisianus Vals.
Caryophyllaceae
2
200
Dianthus sardous Bacch., Brullo & Giusso del Galdo
Caryophyllaceae
1
100
Silene canescens Ten.
Caryophyllaceae
1
100
Silene hifacensis Rouy
Caryophyllaceae
1
100
Silene morisiana Bég. & Ravano
Caryophyllaceae
1
100
Silene rosulata Soy.-Will. & Godr. subsp. sanctae-theresiae
(Jeanm.) Jeanm.
Caryophyllaceae
1
100
Silene salzmanii Badarò
Caryophyllaceae
1
100
Silene succulenta Forssk. subsp. corsica (DC.) Nyman
Caryophyllaceae
2
200
Silene valsecchiae Bocchieri
Caryophyllaceae
2
200
Silene velutina Loisel.
Caryophyllaceae
4
400
Silene vulgaris (Moench) Garcke subsp. vulgaris
Caryophyllaceae
1
100
Silene vulgaris (Moench) subsp. prostrata (Gaudin)
Schinz & Thell
Caryophyllaceae
1
100
Spergularia macrorhiza (Loisel.) Heynh.
Caryophyllaceae
2
199
Cistus albidus L.
Cistaceae
2
200
Cistus creticus L. subsp. corsicus (Loisel.) Greuter & Burdet
Cistaceae
1
100
Cistus creticus L. subsp. creticus
Cistaceae
1
100
Cistus creticus L. subsp. eriocephalus (Viv.) Greuter & Burdet
Cistaceae
4
400
Cistus crispus L.
Cistaceae
1
100
Cistus ladanifer L.
Cistaceae
1
100
Cistus monspeliensis L.
Cistaceae
1
100
Cistus salviifolius L.
Cistaceae
7
700
Halimium halimifolium (L.) Willk. subsp. halimifolium
Cistaceae
2
210
Helianthemum aegyptiacum (L.) Mill.
Cistaceae
2
200
Helianthemum caput-felis Moris
Cistaceae
1
100
Helianthemum croceum (Desf.) Pers.
Cistaceae
1
100
Tuberaria guttata (L.) Fourr.
Cistaceae
1
100
Cornus sanguinea L. s.l.
Cornaceae
1
100
471
O. GRILLO, E. MATTANA, G. VENORA AND G. BACCHETTA
Appendix continued
Taxa
Family (APG II)
Bryonia marmorata Petit
Cucurbitaceae
Accessions
2
Seeds (n)
200
Juniperus communis L. subsp. alpina Čelak.
Cupressaceae
2
200
Juniperus communis L. subsp. communis
Cupressaceae
1
100
Juniperus oxycedrus L. subsp. macrocarpa (Sibth. & Sm.) Neilr.
Cupressaceae
4
400
Juniperus oxycedrus L. subsp. oxycedrus
Cupressaceae
2
200
Juniperus phoenicea L. subsp. phoenicea
Cupressaceae
1
100
Juniperus phoenicea L. subsp. turbinata (Guss.) Nyman
Cupressaceae
5
500
Juniperus sabina L.
Cupressaceae
1
100
Juniperus thurifera L.
Cupressaceae
1
100
Cynomorium coccineum L. subsp. coccineum
Cynomoriaceae
2
200
Carex pendula Huds.
Cyperaceae
1
100
Cyperus capitatus Vand.
Cyperaceae
4
400
Schoenus nigricans L.
Cyperaceae
1
100
Scirpoides holoschoenus (L.) Sojak
Cyperaceae
1
100
Cephalaria squamiflora (Sieber) Greuter subsp.
mediterranea (Viv.) Pignatti
Dipsacaceae
5
498
Scabiosa holosericea Bertol.
Dipsacaceae
1
100
Sixalix atropurpurea (L.) Greuter & Burdet s.l.
Dipsacaceae
1
100
Ephedra distachya L. subsp. distachya
Ephedraceae
2
201
Ephedra nebrodensis Guss. subsp. nebrodensis
Ephedraceae
1
100
Erica terminalis Salisb.
Ericaceae
1
100
Euphorbia amygdaloides L. subsp. semiperfoliata
(Viv.) Radcl.-Sm.
Euphorbiaceae
1
100
Euphorbia gayi Salis
Euphorbiaceae
1
13
Euphorbia paralias L.
Euphorbiaceae
1
100
Euphorbia pithyusa L. subsp. pithyusa
Euphorbiaceae
1
100
Euphorbia pithyusa L. subsp. cupanii (Guss. ex Bertol.)
Radcl.-Sm.
Euphorbiaceae
2
200
Euphorbia spinosa L. s.l.
Euphorbiaceae
1
26
Anthyllis barba-jovis L.
Fabaceae
3
300
Anthyllis hermanniae L. subsp. ichnusae Brullo & Giusso
Fabaceae
1
100
Astragalus cavanillesii Podlech
Fabaceae
1
24
Astragalus genargenteus Moris
Fabaceae
2
200
Astragalus greuteri Bacch. & Brullo
Fabaceae
1
100
Astragalus maritimus Moris
Fabaceae
5
500
Astragalus oxyglottis Steven
Fabaceae
2
200
Astragalus tegulensis Bacch. & Brullo
Fabaceae
1
100
Astragalus terraccianoi Vals.
Fabaceae
3
239
472
STATISTICAL SEED CLASSIFIERS
Appendix continued
Taxa
Family (APG II)
Astragalus thermensis Vals.
Fabaceae
Accessions
2
Seeds (n)
200
Astragalus tremolsianus Pau
Fabaceae
1
100
Astragalus verrucosus Moris
Fabaceae
1
100
Bituminaria morisiana (Pignatti & Metlesics) Greuter
Fabaceae
2
200
Calicotome villosa (Poir.) Link
Fabaceae
2
129
Colutea arborescens L.
Fabaceae
1
100
Cytisus villosus Pourr.
Fabaceae
1
100
Dorycnium hirsutum (L.) Ser.
Fabaceae
1
100
Dorycnium pentaphyllum Scop.
Fabaceae
1
100
Genista sardoa Vals.
Fabaceae
1
44
Genista corsica (Loisel.) DC.
Fabaceae
1
100
Genista linifolia L.
Fabaceae
1
62
Genista valsecchiae Bocchieri
Fabaceae
1
100
Lotus cytisoides L. subsp. conradiae Gamisans
Fabaceae
1
100
Medicago arborea L. s.l.
Fabaceae
2
200
Ononis variegata L.
Fabaceae
1
100
Teline monspessulana (L.) K. Koch
Fabaceae
1
100
Centaurium erythraea Rafn s.l.
Gentianaceae
1
100
Gentiana lutea L. s.l.
Gentianaceae
2
201
Erodium corsicum Léman
Geraniaceae
1
100
Ribes multiflorum Kit. ex Roem. & Schult. subsp. multiflorum
Grossulariaceae
1
100
Ribes multiflorum Kit. ex Roem. & Schult. subsp.
sandalioticum Arrigoni
Grossulariaceae
2
200
Ribes sardoum Martelli
Grossulariaceae
4
42
Hypericum hircinum L. subsp. hircinum
Hypericaceae
2
201
Hypericum scruglii Bacch., Brullo & Salmeri
Hypericaceae
1
100
Crocus minimus DC.
Iridaceae
1
100
Iris planifolia (Mill.) Fiori
Iridaceae
1
100
Iris xiphium L.
Iridaceae
2
200
Juncus maritimus Lam.
Juncaceae
2
200
Juncus subulatus Forssk.
Juncaceae
2
200
Clinopodium sandalioticum (Bacch. & Brullo)
Bacch. & Brullo ex Peruzzi & F. Conti
Lamiaceae
1
100
Micromeria filiformis (Aiton) Benth. subsp. cordata
(Bertol.) Pignatti
Lamiaceae
1
89
Micromeria graeca (L.) Benth. ex Rchb. s.l.
Lamiaceae
1
100
Nepeta foliosa Moris
Lamiaceae
1
100
Salvia aethiopis L.
Lamiaceae
1
100
473
O. GRILLO, E. MATTANA, G. VENORA AND G. BACCHETTA
Appendix continued
Taxa
Family (APG II)
Salvia desoleana Atzei & Picci
Lamiaceae
Accessions
1
Seeds (n)
100
Satureja thymbra L.
Lamiaceae
2
200
Stachys corsica Pers.
Lamiaceae
1
80
Teucrium capitatum L. subsp. capitatum
Lamiaceae
1
100
Teucrium flavum L. subsp. glaucum (Jord. & Fourr.) Ronninger
Lamiaceae
4
400
Teucrium marum L.
Lamiaceae
2
200
Teucrium massiliense L.
Lamiaceae
1
100
Teucrium montanum L.
Lamiaceae
1
100
Teucrium polium L. subsp. polium
Lamiaceae
2
200
Vitex agnus-castus L.
Lamiaceae
2
200
Linum muelleri Moris
Linaceae
3
300
Lavatera arborea L.
Malvaceae
5
437
Lavatera maritima Gouan s.l.
Malvaceae
2
200
Lavatera plazzae Atzei
Malvaceae
2
200
Lavatera triloba L. subsp. triloba
Malvaceae
2
190
Anagallis monelli L. subsp. monelli
Myrsinaceae
1
100
Paeonia corsica Sieber ex Tausch.
Paeoniaceae
3
300
Glaucium corniculatum (L.) Rudolph subsp. corniculatum
Papaveraceae
1
100
Glaucium flavum Crantz
Papaveraceae
2
200
Pinus halepensis Mill.
Pinaceae
4
368
Digitalis purpurea L. var. gyspergerae (Rouy) Fiori
Plantaginaceae
3
300
Erinus alpinus L. s.l.
Plantaginaceae
1
101
Linaria arcusangeli Atzei & Camarda
Plantaginaceae
1
100
Linaria cossonii Barratte
Plantaginaceae
2
199
Linaria flava (Poir.) Desf. subsp. sardoa (Sommier) A. Terracc.
Plantaginaceae
2
200
Plantago subulata L. subsp. insularis (Gren. & Godr.) Nyman
Plantaginaceae
1
100
Armeria pungens (Link) Hoffmanns. & Link
Plumbaginaceae
1
100
Armeria sulcitana Arrigoni
Plumbaginaceae
1
100
Limonium avei (De Not.) Brullo & Erben
Plumbaginaceae
1
100
Limonium morisianum Arrigoni
Plumbaginaceae
1
100
Limonium narbonense Mill.
Plumbaginaceae
1
100
Limonium retirameum Greuter & Burdet
Plumbaginaceae
1
100
Limonium virgatum (Willd.) Fourr.
Plumbaginaceae
1
96
Ammophila arenaria (L.) Link. subsp. australis (Mabille) Lainz
Poaceae
3
251
Ampelodesmos mauritanicus (Poir.) T. Durand & Schinz
Poaceae
1
100
Arrhenatherum murcicum Sennen
Poaceae
1
100
Dactylis glomerata L. subsp. hispanica (Roth) Nyman
Poaceae
1
100
Elymus farctus (Viv.) Runemark ex Melderis subsp. farctus
Poaceae
3
301
474
STATISTICAL SEED CLASSIFIERS
Appendix continued
Taxa
Family (APG II)
Elytrigia corsica (Hackel) J. Holub
Poaceae
Accessions
1
Seeds (n)
100
Phleum sardoum (Hack.) Hack.
Poaceae
1
100
Piptatherum caerulescens (Desf.) Richter
Poaceae
1
85
Piptatherum miliaceum (L.) Coss. s.l.
Poaceae
4
400
Poa balbisii Parl.
Poaceae
1
100
Sesleria insularis Sommier subsp. barbaricina Arrigoni
Poaceae
2
200
Sesleria insularis Sommier subsp. insularis
Poaceae
1
27
Sesleria insularis Sommier subsp. morisiana Arrigoni
Poaceae
1
100
Trisetaria gracilis (Moris) Banfi & Arrigoni
Poaceae
1
100
Polygala sardoa Chodat
Polygalaceae
4
350
Polygala sinisica Arrigoni
Polygalaceae
3
239
Rumex scutatus L. subsp. glaucescens (Guss.) Brullo,
Scelsi & Spamp.
Polygonaceae
2
200
Aquilegia barbaricina Arrigoni & E. Nardi
Ranunculaceae
8
800
Aquilegia bernardii Gren. & Greuter
Ranunculaceae
3
236
Aquilegia dumeticola Jord.
Ranunculaceae
1
100
Aquilegia litardierei Briq.
Ranunculaceae
3
218
Aquilegia nugorensis Arrigoni & E. Nardi
Ranunculaceae
12
1188
Aquilegia vulgaris L.
Ranunculaceae
3
245
Clematis flammula L.
Ranunculaceae
1
100
Delphinium gracile DC.
Ranunculaceae
1
100
Delphinium pictum Willd. s.l.
Ranunculaceae
1
100
Helleborus lividus Aiton subsp. corsicus (Briq.) P. Fourn.
Ranunculaceae
2
176
Nigella damascena L.
Ranunculaceae
1
100
Thalictrum minus L.
Ranunculaceae
1
100
Reseda luteola L.
Resedaceae
1
100
Rhamnus alpina L. s.l.
Rhamnaceae
1
100
Rhamnus lycioides L. subsp. oleoides (L.) Jahald. & Maire
Rhamnaceae
1
100
Rhamnus persicifolia Moris
Rhamnaceae
3
300
Pyracantha coccinea M. Roem.
Rosaceae
1
100
Sarcopoterium spinosum (L.) Spach.
Rosaceae
5
500
Sorbus aria (L.) Crantz s.l.
Rosaceae
2
200
Asperula pumila Moris
Rubiaceae
1
100
Crucianella maritima L.
Rubiaceae
3
300
Galium corsicum Spreng.
Rubiaceae
2
200
Galium minutulum Jord.
Rubiaceae
1
99
Galium schmidii Arrigoni
Rubiaceae
2
200
Rubia peregrina L. subsp. requienii (Duby) Cardona & Sierra
Rubiaceae
1
100
475
O. GRILLO, E. MATTANA, G. VENORA AND G. BACCHETTA
Appendix continued
Taxa
Family (APG II)
Cneorum tricoccon L.
Rutaceae
Accessions
3
Seeds (n)
300
Ruta lamarmorae Bacch. & Brullo
Rutaceae
1
100
Acer monspessulanum L. subsp. monspessulanum
Sapindaceae
1
100
Scrophularia canina L. subsp. bicolor (Sm.) Greuter
Scrophulariaceae
2
200
Scrophularia trifoliata L.
Scrophulariaceae
2
200
Verbascum argenteum Ten.
Scrophulariaceae
1
100
Verbascum boerhavii L.
Scrophulariaceae
1
100
Verbascum conocarpum Moris
Scrophulariaceae
1
102
Verbascum niveum Ten. subsp. garganicum (Ten.) Murb.
Scrophulariaceae
1
100
Verbascum plantagineum Moris
Scrophulariaceae
1
100
Lycium europaeum L.
Solanaceae
1
100
Mandragora autumnalis Bertol.
Solanaceae
2
200
Nicotiana glauca Graham
Solanaceae
1
100
Solanum elaeagnifolium Cab.
Solanaceae
2
200
Taxus baccata L.
Taxaceae
3
300
Daphne oleoides Schreb s.l.
Thymeleaceae
1
100
Centranthus amazonum Fridl. & A. Raynal
Valerianaceae
8
360
Centranthus ruber (L.) DC. subsp. ruber
Valerianaceae
6
522
Centranthus trinervis (Viv.) Bég.
Valerianaceae
11
1098
Vitis vinifera L. subsp. sylvestris (C.C. Gmel) Hegi
Vitaceae
1
100
501
47,493
Totals
476