Analysis of variance showed significant differences for all of the measured traits of 23 sumac accessions (Table 2). These results indicate a large variation among accessions, which reflects a good chance for genetic improvement of sumac. The phenotypic coefficient of variation (CV) ranged from 3.6% in seed length to 37.9% in the weight of the bunch (Table 2), which verifies the existence of considerable high variation among sumac accessions.
The mean leaf length was 17.7 cm and varied from 11.5 to 22.8 cm, which accessions YA3, YA5, and YA11 had high values for leaf length (Table 3). The mean leaf width was 8.6 cm and varied from 4.7 to 12.0 cm, whereas accessions YA3, YA5, and YA11 had high values for leaf width (Table 3). All of these superior accessions were belonging to Yasinabad in the west of Iran. Accession YA2 had the highest magnitudes of terminal leaflet length and terminal leaflet width while the mean values were 4.7 and 2.4 mm for TLL and TLW, respectively (Table 3). The mean bunch width was 26.6 mm and varied from 12.9 to 59.9 mm, whereas accession YA7 had high values for leaf length (Table 3). The superior accessions for all of the mentioned traits were from Yasinabad in the west of Iran. Accession YA1, YA3, YA4, YA5, YA9, YA10, KB1, KB4, and KO2 had the highest magnitudes of leaf density while YA7, YA11, and KB4 had the highest values of leaf number (Table 3). Accession YA7 had the highest values for bunch length, the weight of the bunch, fruit density per bunch, fruit length, fruit width and seed length while accession TA following YA6, AB1, and AB2 accessions had the highest values for the weight of ten fruit (Table 4). Thus, accession YA7 showed good performance for bunch and fruit related traits except yield. Generally, accessions YA6, AB1, and AB2 following to YA10, KB1, KB2, KB3, and AB3 had highest or relatively highest values for most of the measured morphological traits.
To determine in the most precise manner the interrelation of traits, correlation coefficients were established (Table 5) and showed there were not significant positive correlations between the weight of ten fruit and all of the measured traits while leaf length, leaf width, terminal leaflet length and terminal leaflet width indicated significant negative correlations with the weight of ten fruit. Similar results for traits associations were reported by Mohammadi-Alaghoz et al. (2021a) in study of morphological traits of sumac in five different populations in Iran. Seed length and fruit width had significant positive correlation with each other and with leaf number, bunch width, the weight of the bunch, fruit density per bunch and fruit length (Table 5). Fruit length indicated negative correlations with leaf length, leaf width and terminal leaflet width and positive correlations with leaf number, bunch width, the weight of the bunch, fruit density per bunch (Table 5). The weight of the bunch and fruit density per bunch had significant positive correlation with each other and with leaf length, leaf width, bunch width, bunch length (Table 5). Bunch length showed significant positive correlations with leaf length, leaf width, terminal leaflet length, terminal leaflet width, leaf density and bunch width (Table 5). Leaf length and leaf width traits had significant positive correlation with each other and with terminal leaflet length and terminal leaflet width (Table 5). Similarly, Fereidoonfar et al. (2018) found relatively similar results in investigation of morphological traits in some sumac accessions in a central province of Iran. Considering the dendrogram (Fig. 1), the 23 sumac accessions are classified into four distinct clusters, which is the cutting point was determined via Wilks’ lambda statistic of multivariate ANOVA (results are not shown). In Cluster-I, six accessions (AB2, AB4, AB5, KO1, TA, and YA8) were grouped while in Cluster-II, three accessions (KB1, YA9, and YA10) were grouped. In Cluster-IV, three accessions (YA1, YA7, and Y11) were grouped and other remined accessions were grouped as Cluster-III with eleven accessions (Fig. 1). The mean values of each cluster for measured traits are given in Table 6. The Cluster-I had lowest values for all traits expect the weight of ten fruit while The Cluster-IV had lowest values for all traits expect the weight of ten fruit (Table 6). Cluster-II and Cluster-III had high values for some traits, low values for other traits and moderate values for some other traits. It is obvious that accessions of Cluster-I and Cluster-IV are the distinct groups and maybe genetically different heterotic groups, which can be used in crossing planning. Also, accessions of Cluster-I can be considered for potential for the weight of ten fruit while the accessions of Cluster-IV can be considered for potential for the weight of the bunch. The accessions of Cluster-II and Cluster-III had high or average values for the most of the characters and can be used for improving these traits of sumac regarding targets of breeder. The accessions of Cluster-I can be regarded from good potential for the weight of ten fruit while the accessions of Cluster-IV can be regarded from good potential for the weight of the bunch. The accessions of Cluster-II and Cluster-III had high or average magnitudes for the most of the traits and can be applied for improving these traits in sumac considering breeders’ goals.
Figure 2 indicated the distribution of the 23 areas of origin of the accessions along the first two axes of the factor analysis. The factor analysis explained 74.81% (39.71% and 35.10% by factor-1 and factor-2, respectively) of the total variation, which this percentage indicates the complexity of the associations among the accessions. The first factor axis separated accessions to two left and right groups and the second factor axis separated them to four groups; left-up group consist on KO1, KO2 and YA8; left-down group consist on AB1, AB2, AB3, AB4, AB5, KB4, and TA; right-down group consist on YA7 and YA11; and right-up group consist on the other remined accessions (Fig. 2). Some of the results can be verified from the cluster analysis (Fig. 1) but some are not consistent with the data and such mentioned discrepancies are expected because the first two factors usually explain less than 100% (in this case, about 75%) rather than 100% of the total variation.
To better grasp the association among the sumac traits they are graphically presented as a plot of Factor 1 versus Factor 2 (Fig. 3). In this figure, for each trait, a vector is dragged from the plot origin to simplify embodiment of the association between and among the traits by the cosine of the angle between vectors. Two traits are positive correlated if the angle between vectors is 0°, independent if the angle is 90° and negative correlated if the angle is 180°. Thus, regarding angle 0° between and among TLW, TLL, LL, and LW, they are positively correlated as well as between and among FDD, WB, and BW traits and between and among FL, FW and SL traits (Fig. 3). The association between FDD, WB, and BW traits with TLW, TLL, LL, and LW traits is independent due to the angle of 90° between their vectors. Also, interrelationship between WTF with TLW, TLL, LL, and LW traits is negative due to the angle of 1800° between their vectors (Fig. 3). Most of the above predictions on traits’ association can be verified from the original data but some are not consistent with the data due to less explanations of two factors than the total variation. In this plot, traits with taller vectors are more reactionary, while traits with shorter vectors are less reactionary to the accessions; and those assessed at the plot origin are not reactionary at all, and most of our sumac traits had taller vectors and so they are more reactionary to the 23 sumac accessions.