Genetic variability in breeding materials is essential for a successful plant breeding program. Understanding the magnitude of variability in crop species is crucial because it forms the foundation for selection.
Trait Ranges and Analysis of Variance
In the present study, all traits except biomass yield did not exhibit homogeneity of error variances, as indicated by P-values below the significance levels (0.05 and 0.01). Consequently, separate analyses were conducted. Based on data from two seasons, substantial ranges between maximal and minimal mean values were observed for all evaluated traits (Table 4). For instance, in the main season, the mean value ranges for days to heading, days to maturity, and grain filling period were 32-44, 86- 101.8, and 44-62.8 days, respectively. In the off-season, these ranges were 43.0-54.1, 87.0- 97.4, and 43.3-44 days, respectively (Table 4). Similarly, broad ranges were also noted for all studied traits (Table 4).
Across the two seasons of field experiments, the ranges of mean values were 50.6-114.2 cm for plant height, 28-71.6 cm for culm length, 22.2-50.6 cm for panicle length, and 11.0-28.4 cm for peduncle length (Table 4). The minimum and maximum mean values across the two seasons were 0.023 and 0.5 g for thousand seed weight, 350 and 3500 kg/ha for grain yield, 2000 and 24000 kg/ha for biomass yield, 0.01 and 1.53 for grain yield per panicle, 21-88 for lodging index, and 0.042 and 22.1% for harvest index.
The separate analyses of variance for the two seasons revealed that, except for thousand seed weight and fertile tiller number per plant in the main season, and peduncle length and number of fertile tillers in the off-season, most of the genotypes did not show statistically significant differences in terms of variance (P ≤ 0.01 and 0.05) for all evaluated traits (Table 5).Likewise, notable difference concerning origin (regions) were observed in the mean squares of a few traits: biomass yield, grain yield per panicle, and main shoot panicle weight during the off- season, as well as plant height in the main season (Table 5).
The current results from the analyses of variance contrast with previous findings in genetic diversity studies of tef germplasm, where substantial variations were reported for many of the evaluated traits [[9], [10], [11], [24]]. These discrepancies could be due to differences in genotypes and testing environments [25]. However, the range values of traits are in alignment with the report by Assefa et al. [9], except for total and fertile tillers per plant, grain yield, biomass yield, and harvest index. As indicated the fertile tillers per plant ranged from 4.6 to 25 in the main season and 1.8 to 10.6 in the off-season, the mean value for the main season was higher than that reported by Assefa et al.[9] , Jifar et al. [11], and Fikre et al. [13]. Nonetheless, the mean values of tiller number in the off-season were nearly similar with the reports of Assefa et al. [9], Jifar et al. [11], and Fikre et al. [13].
Similarly, mean value ranges for shoot biomass and grain yield in the main season were 5.2 to 24.0 t/ha and 0.35 to 3.50 t/ha, respectively. In the off-season, these ranges were 0.7-2.3 t/ha and 5.0-19.0 t/ha. This indicates that the main season biomass and grain yield per hectare were roughly comparable to those reported by Jifar et al. [26], while off-season yields were relatively lower. Additionally, the range of harvest index in this study was 1.8-22.1 in the main season and 0.042-14.18 in the off-season. These values significantly deviate from the ranges of 5.0- 38.8 and 14.7-24.3 reported by Assefa et al. [9] and Jifar et al. [27], respectively. Such differences can be attributed to seasonal variations in moisture and temperature, as well as differences in test genotypes [25]
Cluster and Distance Analysis
Cluster analysis grouped the genotypes into four clusters based on their similarity. The first cluster (C1, n=31=38.27%) comprised the largest number of core germplasm lines, originating from Jimma (6), Gojam (4), Tigray (4), Wello (11), Wellega (4), and East Shoa (2) (Figure 1 and Table 6). Subsequently, the fourth cluster (n=19=23.46%) included lines from West Shoa (17) and East Shoa (2), followed by the second cluster (n=16=19.75%) encompassing all released varieties (7), as well as germplasm lines from Arsi (5) and West Shoa (4). The smallest cluster was cluster three (C3, n=15=18.51%), consisting of 15 genotypes, all originating from East Shoa (Figure 1 and Table 6).
The clusters displaying the least genetic divergence were clusters 3 and 4, with a D2 value of 11.80, while relatively high divergence was observed between cluster 1 and cluster 2, followed by cluster 1 and cluster 4 (Table 7). Conversely, within-cluster divergence was relatively high for cluster three, followed by cluster two, with the least observed within cluster 1 (Table 7).
When comparing the four clusters formed based on the similarity or differentiation of 17 pheno-morphic and agronomic traits, significant differences among clusters were only observed for three traits: grain filling period, plant height, and grain yield per panicle (Table8). The findings of this study diverge in terms of the number of clusters from those reported by different authors using different sets of tef genotypes. For example, the reported number of clusters was 3 for 18 genotypes[12] , 6 for 28 semi-dwarf genotypes [28], 6 for 188 genotypes [27], and 7 for 49 genotypes [13].
The tested tef genotypes, including germplasm lines from various zones and released varieties from the same origin, clustered into different classes, while those from different origins were grouped together. This confirms the conclusion drawn by Assefa et al.[10] That the genetic diversity level in tef germplasm is comparatively higher within populations (origin) than among populations (origins). Consequently, accessions originating from the same region and altitude were not distinctly separated into distant clusters. Therefore, although this study indicates relatively lower diversity, tef genotypes did not cluster into a small number of groups, as noted in earlier studies [[29],[26]].
In this study, the least genetic divergence was observed between cluster 3 and cluster 4, with a D2 value of 11.8, while high significant divergence was noted between cluster 2 and cluster 1 (D2 = 85.49), followed by the divergence between cluster 1 and cluster 4 (D2 = 45) (Table 7). The notably high inter-cluster distance between cluster 1 and cluster 2 may be attributed to the inclusion of released varieties. Consequently, it is advisable to consider crosses from this cluster for enhanced heterotic expression [27]. Furthermore, within clusters, relatively high distances were observed for cluster 3 and 2 (Table 7), indicating the presence of diverse genotypes within the same cluster, which could hold potential for further breeding.
Principal Component Analysis
The principal component analysis revealed that the first six principal components, each with eigenvalues greater than one, collectively accounted for approximately 70.6% of the total variation among the 74 tef core germplasm lines and 7 released varieties assessed for 17 traits (Table 9). Among these, the first principal component (PC) explained 20% of the overall phenotypic variation among the tef genotypes, primarily attributed to variations in panicle length, number of fertile florets per spikelet, culm length, plant height, days to maturity, number of total and fertile tillers, and grain filling period.
The second principal component, which accounted for 15.20% of the total variation, was primarily influenced by variations in grain yield per plot, biomass yield, harvest index, and peduncle length. The third principal component, contributing to 10.80% of the total variation, was chiefly due to variations in grain yield per panicle, main panicle shoot weight, and peduncle length. The fourth principal component, responsible for 8.90% of the total variation, primarily resulted from high variations in the number of fertile and total tillers per plant.
Similarly, the fifth principal component, explaining 7.80% of the total variation, was mainly influenced by the number of fertile florets per spikelet and thousand seed weight. The sixth principal component, also accounting for 7.8% of the total variation, was chiefly affected by variations in the number of days to heading.
The observed variation in principal components in this study is lower than that reported in studies by Assefa et al. [29], Jifar et al. [26], Jifar et al. [28], and Fikre et al. [13], where the first PC accounted for 40%, 44.7%, 41.3%, and 30.65% of the gross variability, respectively. Furthermore, the proportion of variation explained by the first three principal components in this study (46%) was lower than values previously reported: 64.7% by Assefa et al. [10][10], 68.67% by Assefa et al. [29], 74.66% by Adnew et al. [30], 71.03% by Plaza-Wüthrich et al.[12] , 78.3% by Jifar et al. [26], 69.1% by Jifar et al. [28], and 55.9% by Fikre et al. [13].
This suggests that the phenotypic diversity among the tested tef genotypes cannot be adequately explained solely by a few principal components. This observation persists despite the fact that the analyses of variance did not reveal substantial variations among the genotypes in most of the evaluated traits.