Five major groups were obtained from the hierarchical classification at a distance of 6.65 (Fig 1), knowing that the number of clades in the final subdivision was based on the number of subspecies of
V. sativa L. reported by
Quezel and Santa (1962). Some accessions belonging to the same subspecies presented homologous profiles, others identified by the same key displayed ultra-polymorphic profiles. The lowest Euclidian distance (d=1) is observed between the two accessions 16 and 17 (subsp.
obovata), these individuals presented two extra similar profiles with a level of similarity equal to 89.4 and only one variable showing two different modalities, confirming their great rapprochement in the circle of individuals in PCA (Fig 2). On the other side, The greatest Euclidian distance (d=6.65) observed between the two accessions 1 and 19 (subsp.
obovata); these individuals presented two hyper heterogeneous profiles with a level of similarity equal to 29.54 and 22 variables with different modalities. These results are consistent with those obtained by
Bechkri and Khelifi (2016) and highlight the taxonomic debates concerning the group
V. sativa L.
In order to study the similarities between individuals, a principal component analysis was carried out on 71 characters coded with ordinal values and ranked from 2 to 6 states (Table 1). 29 characters presented homologous modalities for all accessions and were disqualified. The results of our PCA showed that the first 12 axes had Eigen values greater than 1 and held more than 90% of the information. In our case only the 2 first axes were maintained in order to build a 2D plan with an inertia imprisonment of around 30%, axis1 (18.33%) and axis2 (11.48%).
Among the most explanatory variables having a good correlation with the first factor we distinguished: ClowTL (
r=0.9024), CupTL (
r=0.9024), CTL (
r=0.7175), CHD (
r=0.7124), SP (
r=-0.5170) and SHD (
r=-0.4915) (Fig 2). These results correspond with those cited by
Van De Wouw et al., (2003) which shows that the characters that separate the most subspecies are nearly quantitative. According to
Mettin and Hanelt (1964),
Vicia angustifolia L. and
Vicia amphicarpa L. generally possess hairier calyx than other subspecies. Our results, in addition to the results of
Bechkri and Khelifi (2016) showed quite the opposite for
V. angustifolia L. Five main variables participated in building the second factor, 3 of them exhibited a positive correlation: ABBO (
r=0.745), Lsur (
r=0.696), HD (
r=0.658). In contrast, WC (
r=-0.694) and SC (
r=-0.619) provided a negative correlation. These traits confirmed the results cited by
Potokina (1997) on the possibility of classifying accessions successfully by their types of coat models. In the light of these results, it was shown that the first factor concerned the length of teeth and pubescence of the calyx depending on the stem pilosity. While, the second factor concerned the patterns of pods and seeds according to their color. These demonstrate the usefulness of these characters for the identification of different
Vicia sativa L. subspecies.
The results of our Kruskal-Wallis test executed on 29 continuous quantitative traits showed a significant effect of the factor (accessions) for all the attributes used in our study (Table 1). These genetically diverse accessions can be used as promising parents for hybridization
(Singh et al., 2021), to develop superior hybrids with desirable combination of traits
(Janghel et al., 2021). Mebarkia et al., (2020) revealed a significant positive relationship between the full flowering stage with grain yield (r=0.76) and aerial biomass (r=0.69) and showed that late varieties are the most productive. The results of our study showed that sample 16 was the late entry for the start of flowering (117 d), full flowering (124 d), pod formation (136 d) and full maturity (155 d) with moderate to high values for almost all variables. Contrastingly, sample 5 was the earliest for the same stages (101 d, 109 d, 114 d and 138 d) with very low values for most of the characters studied. It seems that the late and long period of flowering allows these accessions to escape the late frosts of the region, which negatively affects the fertility of the plant. These results are in line with those obtained by
Keating et al., (1991).
The highest plant at harvest (accession 14) recorded a measurement equal to 82 cm, while the most dwarf (accession 6) exhibited a height equivalent to 36 cm. Mebarkia
et al., (2020) revealed a significant positive correlation (r=0.62) between grain yield and plant height. Our results are in disagreement with these findings and showed that both sample 6 and 14 displayed low grain yield with respectively 1.82 and 5.62 g/plant. This can be explained by the low values recorded by these two accessions concerning seed size (3.19 and 3.58 mm) and the weight of one hundred grain (2.46 and 3.92 g). According to
Moles et al., (2005), small-seeded species are able to produce more seeds for a given amount of energy in comparison with large-seeded species. This finding confirmed in our case as the smallest diameter of seeds (3.32 mm) is attributed to accession 24 (subsp.
consobrina) recorded a larger value of NG/P with an average of 7.26.
In exalbuminated seeds like those of vetch, the early stages of radicle elongation seem to depend on reserves accumulated in their own tissues (
Hopkins, 1999). This indicate that it is preferable to use large seeds to improve the emergence and survival of plants, especially in areas with harsh environmental conditions
(Mao et al., 2015; Dong et al., 2016). The accession 20 (subsp.
obovata) presented the largest diameter of seeds (5.38 mm) with the highest WHS (12.17 g). Therefore, it can help breeders working to improve emergence in areas with difficult environmental conditions.
The accession 9 (subsp.
obovata) proved to be the most efficient by recording high values for the most of variables with agronomic interest especially grain yield (8.29 g/plant), number of pods per plant (17.66), number of seeds per pod (6.13) and weight of hundred grain (7.36 g). Hence, direct selection for these traits will lead to the development of desirable high yielding genotypes
(Sharma et al., 2021).
Spearman’s correlation test showed that soil conductivity followed by elevation and soil organic matter are the characters that most influence the variability of morphological traits. However, the use of a large number of accessions for each site with putting emphasis on these characters is necessary in order to confirm the resulting model.