The stability of the number of tillers of barley varieties and its relation with consistency of performance under semi-arid conditions

Abstract

In 1981/82 60 barleys varieties were grown at 18 environments (three locations, three seed rates, two soil fertility regimes) and in 1982/83 22 varieties were grown again at 21 environments (seven locations, three seed rates). Grain yield was recorded only in the second year. The estimates of variances sx 2 (untransformed data) and s2 logx (transformed data) for number of tillers and grain yield varied significantly among varieties. When untransformed data were used, high tillering varieties had a higher variance for number of tillers per unit area than low tillering varieties. When the variation was measured by CV of untransformed data or by s2 logx on transformed data a reverse relation occurred, i.e. the correlation coefficient between number of tillers and the variation of number of tillers was negative. High grain yielding varieties had a high mean number of tillers and a low variance for number of tillers per unit area when transformed data were used to compute variances. All interactions between variety, seed rate and location for grain yield were significant. The regression analysis of variety grain yield (Y) on environment mean grain yield (X) gave regression coefficients, b, ranging from 0.51 to 1.69. There was no significant correlation between regression coefficient and grain yield. The highest yielding varieties had b values around 1.0. High grain yielding varieties had low variance of yield over environments and low values for deviations from the regression when transformed data were used. However, the correlations between mean yield and variance for yield or mean yield and deviations from regression were positive when untransformed data were used. From this study it is concluded that high mean number of tillers per unit area and low variance for number of tillers could be used in selecting varieties with consistently high yield at varying environments. Techniques are proposed for application early in the breeding programme, i.e. in segregating population or in nurseries.