Table 2 presents descriptive indices for various characteristics of sainfoin, with the coefficient of variation (CV) showcasing notable variability. In the year 2022, CV ranged from 10.5% to 10.9% for traits such as days to seed filling (DSF) and dry leaf to dry stem ratio (LSR), while it reached higher levels of 63.2%, 50.0%, and 51.1% for traits like inflorescence dry weight (IDW), total fresh and dry yield (TFY and TDY), respectively. Similarly, in 2023, CV ranged from 7.9% to 6.8% for days to imitation of flowering (DIF) and DSF, while it peaked at 70.8%, 51.5%, and 50.2% for IDW, TFY, and TDY, respectively. Notably, environmental conditions in the second year exhibited less variability compared to the first year, resulting in relatively lower CV values in the second year across all characteristics. The most variable traits, as highlighted in Table 2, included number of nodes per main stem (NMS), powdery mildew percentage (PMP), dry weight of stems and leaves (SDW and LDW), number of florets per inflorescence (NFI), thousand seed weight (TSW), and seed yield (SY). The wide range of variations observed in these characteristics indicates a high level of diversity among the sainfoin landraces. Moreover, the greater variability observed in the year 2023 compared to the year 2022 underscores the rich diversity present in the available sainfoin germplasm, suggesting its potential utility in future breeding programs. Considering the general heritability of sainfoin characteristics across the two years, relatively high heritability was observed for most traits, particularly in the first year (Table 2). The high levels of general heritability suggest that the influence of the environment on the estimation of this genetic parameter is minimal. This finding aligns with the findings of Majidi (2011), who reported a low effect of the environment on the estimation of genetic parameters in sainfoin under salt stress conditions.
The first four principal component analyses (PCAs) with eigenvalues greater than unity explained 83% of the variation in the year 2022, while the first four factors accounted for 78% of the variation in the year 2023 (Table 3). Additionally, the first two PCAs explained 70% of the variation in 2022 and 65% in 2023. To visually represent the relationships among the characteristics, they were shown in a plot of PCA1 versus PCA2. Since the first two PCAs captured a significant magnitude of the variability, the association between any two characteristics can be guessed by the cosine of their angles. From the plot for the first year (Fig. 1), several prominent associations were observed: a strong positive association between days to imitation of flowering (DIF) and days to imitation of seed filling (DSF); between peduncle length (PL) and number of stems per plant (NSP); between stem dry weight (SDW), LI, and NLS; and among number of nodes per main stem (NMS), inflorescence dry weight (IDW), number of florets per inflorescence (NFI), total dry yield (TDY), and total fresh yield (TFY). Conversely, there was a near-zero correlation between dry leaf to dry stem ratio (LSR) with SDW, LI, and NLS; between seed yield (SY) with DIF and DSF; and between powdery mildew percentage (PMP) with PL and NSP. Furthermore, a negative correlation was observed between LSR and PMP in the first year. It's worth noting that some disagreements between the graphic findings and numerical dataset were seen because the PCA described for less than 100% of the variability. Similarly, in the year 2023 (Fig. 2), notable associations were observed: a positively relation among DIF, DSF, and NSP; between NLS and SDW; between number of seeds (NS) and SY; and among TDY, PL, leaf dry weight (LDW), IDW, TFY, NMS, NFI, and LI. Conversely, there was a no relation between thousand seed weight (TSW) and LSR; between DIF, DSF, and NSP with TSW; and between TDY, PL, LDW, IDW, TFY, NMS, NFI, and LI with PMP. Additionally, a negative correlation was observed between LSR and DIF, DSF, and NSP in the year 2023. Furthermore, Veisipoor et al. (2012) identified stem number and dry biomass as the important characteristics influencing yield performance of sainfoin, and Najafipoor and Majidi (2017) reported stem and seed numbers as the important characteristics influencing seed yield of sainfoin.
To gain a deeper understanding of the relationships among the characteristics, they were graphically displayed as a plot of PCA1 versus PCA2. In the plot for the first year (Fig. 3), three distinct groups were identified: group A consisting of genotypes G14, G19, and G25; group B comprising fifteen genotypes; and group C comprising fourteen genotypes. In the plot for the second year (Fig. 4), similar groups were observed with some changes: group A now included genotypes G14, G16, G19, and G25; group B containing of nineteen genotypes; and group C containing of of nine genotypes. Comparison of genotype grouping across the two years revealed that thirteen genotypes from group B, as well as eight genotypes from group C (G1, G3, G11, G23, G24, G26, G28, and G30), exhibited similar responses across both years, indicating minimal genotype by environment interaction. Notably, three genotypes from the same landraces in both years; G1 (Bonab), G3 (Marand), and G11 (Urmia); were from the northwest of Iran, characterized by semi-arid and cold climates. Similarly, three genotypes; G23 (Khomeyn), G24 (Arak), and G30 (Khansar); were from central Iran, known for its arid and warm climates. Additionally, one genotype from the same landraces in both years, G26 (Asadabad), hailed from western Iran with a Mediterranean climate featuring hot summers, while the other genotype, G28 (Damavand), originated from northern Iran with a humid continental climate characterized by warm summers. These findings underscore the acceptable genetic variability present in current sainfoin landraces, highlighting the diverse origins of the most favorable genotypes.
Table 4 displays the averages of the characteristics for each of the three groups across both years. Genotypes in group A exhibited lower values for most characteristics across both years, making them promising candidates for genetic improvement of early maturity in sainfoin, attributed to their lower amounts of days to imitation of flowering (DIF) and days to imitation of seed filling (DSF). Additionally, these genotypes demonstrated low sensitivity to powdery mildew (PMP), which is a favorable trait in target regions. In contrast, genotypes in group B displayed moderate values for most characteristics across both years. This suggests that they could be further investigated for other traits not measured in this research, potentially contributing to future improvement programs for sainfoin (Table 4). Accessions in group B show significant potential for various traits, particularly in terms of economic characteristics such as total fresh and dry yield (TFY and TDY), seed yield (SY), and thousand seed weight (TSW). Therefore, these genotypes merit further exploration and consideration in breeding programs aimed at enhancing the overall performance and productivity of sainfoin.