The results of present study revealed that genotypes differed significantly for all the characters except grain yield as indicated by ANOVA for individual traits (Table 1). The observations showed that genotypic variability was high and that prompted and provided us the base to go for further analyses. The estimates of variability parameters for all five traits were worked out and are also presented in Table 1. Mean performance of 62 genotypes for five characters revealed that days to heading varied between 72-86 with an average of 80.14 days having 3.35 critical difference (CD). Similarly, the maturity duration of material ranged between 117-126 days with mean maturity duration of 121 days. The average plant height was 81 cm whereas the average plot yield was recorded as 455g. In general the material under study had a mean of 40 g thousand grain weight over locations.

The variance {genotypic and phenotypic and percent variability along with heritability (broad sense)} estimates were calculated (Table 1). The genotypic coefficient of variation was moderate for most of the traits except thousand grain weight, which otherwise is also considered as a highly heritable character. The highest estimates of heritability were recorded for thousand grain weight followed by days to heading and plant height. Heritability estimates along with the genetic advance are very useful in predicting the gain under selection instead of heritability alone (Singh and Narayana 1993). High heritability coupled with high genetic advance and high coefficient of variability for thousand grain weight and days to heading. Whereas, high estimates of genetic advance and variance for grain yield per plot indicated that due to additive gene effects direct selection maybe effective in this character. However, in case of characters showing high heritability but moderate or low genetic advance, which may be due to non-additive gene action and presence of GxE interaction, simple selection may not be rewarding (Singh et al. 2003, Sharma and Sain 2003). In such cases breeders can go for selections over locations following hybridization and selecting desirable transgressive segregants (Allard 1960; Kumar et al. 2002).

The correlation coefficients (genotypic) among five traits were also worked out to see the association between these traits and are given here in Table 2. Significant and positive correlations were observed between days to heading and maturity duration, thousand grain weight and grain yield, whereas, days to heading and maturity duration showed significant but negative correlations with grain yield. Similarly, thousand grain weight also had significant but negative correlation with heading and maturity duration. The negative association between days to heading, maturity duration with grain yield and thousand grain weight may be due to late heat stress affecting grain development and then ultimately to the yield. This was also evident with moderate mean of thousand grain weight in general. However, the significant and positive association between grain yield and thousand grain weight indicated the direct contribution of grain size to total grain yield. We assume that in durum wheat thousand grain weight has comparatively higher contribution to the yield as compared to bread wheat.

Dendrogram of 62 genotypes was generated on the basis of Euclidean distances created by making use of five parameters and adopting the technique of complete linkage between the genotypes (Statistica software). The critical perusal of the dendrogram reveals that up to the stage 5 the linkage distances between the clusters was appreciable and have little variation thereafter (Fig 1). Plotting of linkage distances across steps following Euclidean distances as presented in Fig. 2 also suggested that up to about 55 steps only 25 % of the total linkage distance has been accounted whereas 75 % of the distance was accounted after step 55 leading to about 5 clusters (Fig. 2). This suggests that the restricting the grouping of genotypes into five clusters will have sufficient distance between the dusters and thus the 62 genotypes were grouped into five clusters. The maximum number of genotypes were in duster V (20 genotypes) while cluster IV had only six genotypes (Table 3). We noticed that out of three checks which were the released varieties for north west and central India, two (PDW 233 and HI 8498) were in cluster III and though the third check (Raj 1555) was in cluster II, but this being the last entry in this cluster could also have been in cluster III, if we slightly increase the Tocher's value. This indicates that in the genotypes under study enough variability exists, which is more than that in already existing released varieties and thus the same can be utilized for hybridization for obtaining the transgressive segregants for desired traits (Deshmukh et al. 1999, Raut et al. 1985). The future yield improvement in durum wheat might be made or achieved by utilizing the information and material developed from the present study.

Acknowledgments

The authors are grateful to CIMMYT for providing germplasm and necessary information like pedigree detais etc for present study. We are thankful to all our colleagues at cooperating centres for conducting the experiments at individual centre successfully.