Recurrent Selection for Grain Yield in Durum Wheat

Considerable improvements of durum wheat (Triticum turgidum L. var. durum) genotypes with better yield have been achieved with different breeding methods. Because of several constraints in its application in small cereals, recurrent selection in durum wheat has not been used extensively. The objectives of this research were to evaluate two cycles of recurrent selection for grain yield and to assess effects of selection on other agronomic and yield-component traits. Nineteen semidwarf durum wheat genotypes selected for high yield and broad genetic diversity, were intercrossed to form the base population. Fifty-five randomly selected S₂ lines from three cycles of selection and a set of four checks were planted in two 1989 and three 1990 environments, respectively. Genotype × environment interactions were generally of magnitude rather than of rank order changes for traits evaluated and therefore a combined analysis across environments was performed. Regression analysis indicated an average increase in yield of 0.25 ± 0.03 Mg ha⁻¹ per cycle of selection. Cycle 2 (C2) produced lines with grain yield higher than any line in Cycles 0 (C0) or 1 (Cl). Test weight and plant height increased 8.39 ± 2.01 Kg m⁻³ and 2.62 ± 0.3 cm per cycle, respectively. Recurrent selection for grain yield did not affect spikelets per spike and kernels per spike. Kernel weight increased 2.8 mg from C0 to C2. When calculated as a percentage of the total variance, the genetic variance of most of the evaluated traits decreased slightly. Genotype × environment variance was the most important component of variances of five of the evaluated traits. The efficiency of recurrent selection for yield was demonstrated. Indications are that recurrent selection solely for yield did not have any negative correlated response to unselected traits. Contribution from North Dakota Agric. Exp. Stn. Journal Paper no. 2215. Research, partly, supported by North Dakota State Wheat Commission, National Pasta Association and CIMMYT, was from thesis of senior author in partial fulfillment of Ph.D. degree requirements at North Dakota State Univ.