## LSU Historical Dissertations and Theses

1989

Dissertation

#### Degree Name

Doctor of Philosophy (PhD)

#### Department

Plant, Enviromental and Soil Sciences

Stephen A. Harrison

#### Abstract

Genotypic stability of performance across environments in general, or across levels of a specific environmental factor in particular, is an important concern of plant breeders. The objectives of this research were to examine the relationships and repeatability of the commonly used stability statistics, the relative merits of these statistics in the stability analyses of wheat (Triticum aestivum L.) yield and test weight, and the relationship of genotypic tolerance to low soil-fertility with yield stability. Yield and test weight data of the Louisiana Agricultural Experiment Station (LAES) winter wheat performance trials from 37 year-location combinations were used in the stability analyses. Three field and two greenhouse experiments were conducted to determine low-fertility tolerance of wheat cultivars. The two-line regression model proposed by Verma et al. (1978) was compared with a linear regression model in characterizing cultivar response to environments. A low-fertility tolerance index (LFTI), similar to the susceptibility index of Fischer and Maurer (1978), was used to rank cultivars for tolerance to low fertility. The stability variance ($\sigma\sbsp{\rm i}{2}$), deviation mean square (Sd$\sbsp{\rm i}{2}$), and the coefficient of determination (r$\sbsp{\rm i}{2}$) were equivalent stability statistics. The variance of genotypic mean (S$\sbsp{\rm i}{2}$) was a function of high b$\sb{\rm i}$ values, which was highly correlated with S$\sbsp{\rm i}{2}$. Genotypes with b$\sb{\rm i}$ values different from 1 had $\sigma\sbsp{\rm i}{2}$ values higher than their Sd$\sbsp{\rm i}{2}$ values. Mean yield (X$\sb{\rm i}$), b$\sb{\rm i}$, and r$\sbsp{\rm i}{2}$ were the most repeatable statistics between subsets of environments. $\sigma\sbsp{\rm i}{2}$ and Sd$\sbsp{\rm i}{2}$ were not repeatable. The two-line regression model was superior to the linear regression model in describing genotypic responses. Low-fertility tolerance, measured by LFTI, was uncorrelated with mean yield and yield stability. Since there was no negative genetic correlation between yield and test weight, and b$\sb{\rm i}$'s of yield and test weight were highly correlated, selection for high yield and high test weight can be effectively performed simultaneously. If heterogeneity of regression is significant, selection for desired adaptability and stability of performance in winter wheat, measured by b$\sb{\rm i}$ and Sd$\sbsp{\rm i}{2}$ or r$\sbsp{\rm i}{2}$, respectively, is likely to be effective. LFTI is a good index for stress tolerance, provided data fit the linear regression model.

215

COinS