2.5.2. Forage Traits

The Kjeldahl method was used for nitrogen content, and the crude protein ratio was calculated using a conversion factor of 6.25. ADF and NDF concentrations were determined according to standard laboratory procedures for forage quality analysis outlined by Ankom Technology. ANKOM F57 filter bags were used for ADF and NDF analysis in this study. Total digestible nutrients (TDN), dry matter intake (DMI), digestible dry matter (DDM), and relative feed value (RFV) were estimated [28] according to the following equations adapted from:

> TDN = (−1.291 × ADF) + 101.35 DMI = 120% NDF% dry matter basis DDM = 88.9 − (0.779 × ADF% dry matter basis) RFV = DDM% × DMI% × 0.775

#### *2.6. Data Analysis*

Qualitative data were analyzed using percentage distribution. Analysis of variance was conducted using SAS 9.1 (Cary, NC, USA) [29]. Augmented randomized complete block design was performed using the "augmentedRCBD" package developed by Aravind et al. [30] in R-Studio (version 2022.02.0) (Boston, MA, USA) [31]. The least significant difference (LSD) test was used for mean comparison in the analysis of variance. Principal component analysis (PCA) was performed with the quantitative traits data using the Minitab 19.1 software (State College, PA, USA) [32]

#### **3. Results**

In this study, the large and diverse grass pea collection grown in highland and lowland conditions was evaluated with three qualitative, 11 quantitative, six forage and two quality traits. The frequency distribution of qualitative traits, flower colors, plant growth habits and seed color of the grass pea genotypes were shown in Figure 2. There was a large variation among grass pea genotypes in terms of flower colors. Nine different flower colors were observed in the grass pea collection as blue, white, blue-white, dark blue, purple, light blue, pink, blue-purple and red with percentages of 28.1, 24.45, 21.53, 13.5, 7.3, 2.55, 1.46, 0.73 and 0.36%. Furthermore, three plant growth habits were determined as erect (50%), semi-erect (40%) and spreading (10%). The seeds of the collection had a gray color with a value of 57%.

**Figure 2.** Frequency distribution of qualitative traits in the whole collection. (**a**) the color of flowers; (**b**) growing type; (**c**) seed color.

According to the analysis of variance, there were significant differences among genotypes for all traits in Antalya (lowland) location (Table S2). The mean number of days of the first flowering varied from 88.8 to 109, and the number of days of 50% flowering ranged from 99.5 to 120.5 in the collection. In the average of two years, the genotypes GP23, GP11, and GP114 had the earliest flowering date. The highest plant height was recorded in GP213 with a value of 90.8, while the shortest plant was GP10 had a plant height of 41.5 cm. The grand mean of number of pods was 23.77 and check cultivars, Corea and Gürbüz showed a higher number of pods than the mean of the germplasm. The maximum and minimum biological yield was detected in GP104 and GP23 with values of 22.5 and 3.3 g, respectively. The seed yield ranged from 1.2 to 7.3 g. The highest seed yield was observed in genotype GP105, followed by GP 104 and GP 249, the mean of collection was 3.01 g. Among the check cultivars, ˙ Ipta¸s gave the highest seed yield, however 40 genotypes of the collection had higher means for seed yield than this check cultivar.

A total of 90 genotypes along with four check cultivars were evaluated at field condition in Isparta (highland) for quantitative agronomic traits using an augmented experimental design (Table S5). The analysis of variance showed that there was a significant (<0.01) difference among the genotypes for all traits except for days of the first flowering. The results revealed no significant differences were observed for blocks except for traits of number of pods, number of branches and stem diameter (Table S6). The number of days to first flowering ranged from 172.1 to 206.3 days, and the number of days to 50% flowering ranged from 181.3 to 213.3 days. The check cultivar, Corea had the earliest 50% flowering date among the check cultivars with a value of 178, only genotypes GP230 and GP247 had higher values compared to this cultivar (Table S5). GP107 and GP105 had the tallest plants, while GP246 was the shortest (15.6 cm). Genotype GP156 and GP145 gave the highest number of pods (62) as well as GP145 produced the highest biological yield being 82.4 g in the highland conditions. There was also a lot of variation in the collection for hundred-seed weight and seed yield traits, which ranged from 4.1 to 82.4 g and 0.9 to 33 g, respectively (Table S5). The genotype GP40 had the highest seed yield followed by GP161, GP18 and GP 19.

Significant differences were observed in the germplasm for quality and forage traits among genotypes grown in Antalya (lowland) (Table S7). The β-ODAP content (%) ranged from 0.25 to 0.49, the average being 0.38. The genotypes GP213, GP49, GP58, GP60 and GP110 had the lowest values for this trait while the genotypes GP248 and GP227 had the highest content. The lowest value was 0.38 among the check cultivars. The maximum amount of protein content was recorded for GP53 in the germplasm, with a notable higher value in the quantity of this trait observed in the genotypes of GP40, GP270 and GP197. Regarding ADF (%), genotypes GP251, GP243, GP248 and GP23 had the highest values over the 9.0 in the mean of two years, the check cultivar, Karada ˘g had the highest check cultivar with value of 8.82. There were four genotypes (GP34, GP156, GP225 and GP149) had an NDF > 17%, the mean of the collection was 13.98. The DMI and DDM ranges in the collection varied from 6.48 to 9.96 and 81.48 to 83.4 with mean values of 8.45 and 82.45, respectively.

The ANOVA analysis of the genotypes showed a highly significant variation in all quality and forage traits in highland conditions (Table S8). There is a non-significant difference among the blocks for these traits in augmented experiment design. Overall, the β-ODAP content (%) corresponding to the genotypes was 0.35 (Table S9). The lowest values were recorded for GP17, GP18 andGP49 with the means of 0.19, 0.21, 0.22, respectively. GP215 had the greatest mean value (0.51) for this trait. The maximum and minimum protein content (%) were detected in the genotypes GP242 and GP225. The genotype, GP248 also had >24% protein content which is higher than the mean of all check cultivars (Table S9). The mean of ADF and NDF values were 8.43 and 15.72, respectively. The highest mean was 10.22 for ADF and 24.23 for NDF. The genotype GP248 was superior for these traits whose means were 10.15 and 19.78, respectively, they were higher than general means and check cultivars. The highest DMI was recorded for GP242, followed by GP43 and GP207. The check cultivar, Corea was the fourth genotype for this trait. When all genotypes were combined, the total mean of total digestible nutrients was 90.47%, with GP213 being the top genotype for this trait. The highest relative feed value was observed in GP242 (668.63) while the lowest value was measured in GP199 (316.72) in the collection.

The grass pea collection examined in this study had 94 genotypes and was evaluated in lowland and highland conditions. The *t*-test of significance for mean values indicated that there were significant differences between the growing areas for all agronomic traits except for the number of branches (Table 2). These genotypes grown in two different environmental area were also compared with quality and forage traits. The mean values for β-ODAP content, raw protein ADF, NDF, DMI and RFV were significantly different between the genotypes grown in lowland and highland (Table 2).

The PCA using the 11 quantitative traits including maturity traits, yield and yield related-trait showed that more than 76.4% and 72.2% variability were accounted for the first four principal components (PCs) with eigenvalues ≥ 1 in the collection grown in Isparta (highland) and Antalya (lowland) (Table 3). The 1st principal component (PC1) had an eigenvalue of 3.89 and explained 35.4% of the total variation in highland. Seed yield and biological yield had the highest positive eigenvectors in PC1, while the pod height had the highest negative eigenvector. The second component (PC2) explained 17.1% of the total variance with an eigenvalue of 1.87 and was mainly correlated with flowering traits, negatively (Table 3). In lowland, the first principal component's (PC1) eigenvalue was 3.36, explaining 30.36% of the total variation, the highest positive eigenvector was biological yield (Table 3). The PC2 explained 16.3% of the total variance and was correlated with days to first flowering, days to 50% flowering, plant height, and the number of pods and stem diameter, positively. The traits of quality and forage were also evaluated with PCA for the grass pea collection grown in lowland and highland (Table 4). Results showed that in the analysis, three components had eigenvalues > 1 for highland andthey explained 93.7% of the variability among the 94 genotypes grown in highland. The PC1 explained 57.2% of the total variance and was positively correlated with all quality and forage traits except for ADF and NDF. The PC1 explained 52.79% of the total variance (75.9%) and was positively

correlated with raw protein, TDN, DMI, DDM and RFV in the collection that was evaluated in lowland.

**Table 2.** Means and standard errors for 11 quantitative and eight quality traits in 94 genotypes produced in Antalya (lowland) and Isparta (highland).


† S.E.: standard error of the mean. # Differences between means of entire and core collection were tested by *t* test; ns is non-significant, *p* = 0.05 and *p* = 0.001, and and --, respectively.

**Table 3.** Eigenvectors for the four principal components (PCs) of traits associated with agronomic performance of 94 grass pea genotypes produced in two different regions.



**Table 4.** Eigenvectors for principal components (PCs) of traits associated with forage and quality value of 94 grass pea genotypes produced in two different regions.
