Durum wheat (Triticum turgidum ssp. durum) is the only tetraploid species of wheat grown in the world. Productivity of this crop is very low in Ethiopia due to lack of improved varieties. Estimation of genotype by environment interaction (GEI) can facilitate variety development. Therefore, this study was carried out with the objectives of understanding the relative contribution of GEI, assessing the stability of varieties in grain yield, characterizing the environments, and understanding the associations of traits. The experiment was conducted at Takusa, Dembia, Dabat, Debark, Debre Tabor and Moretina Jiru during 2018 main cropping season. Twelve durum wheat varieties were used using randomized complete block design with three replications. The experimental unit had a plot size of 3 m2. Spacing between plots and rows were 0.5 m and 0.2 m, respectively. A seed rate of 150 kg ha-1 with the recommended fertilizer rate for the respective areas was used. Yield, yield related and quality data were collected and subjected to analysis of variance for individual and combined locations. Stability analysis of grain yield was executed using Ebrehart and Russel’s, Wricks’ ecovalence (Wi), cultivar superiority measure (Pi), Nassar and Huehn’s mean absolute rank difference (Si1) and variance of ranks (Si2), Kang’s yield stability index (YSI), AMMI stability value (ASV), AMMI and GGE biplot models. Analysis of variance for individual locations showed significant difference among the genotypes in most of the traits, indicating the presence of variability among the genotypes. Combined analysis of variance revealed highly significant difference among genotypes, environments and GEI for grain yield and yield-related traits, indicating the presence of adequate variability among the genotypes, environments and the differential response of the genotypes across environments. Cultivar superiority measure and YSI were positively and significantly correlated with grain yield indicating these models were best in identifying stable durum wheat varieties. Based on ‘which won where’ of GGE biplot model, the locations were classified in to two mega-environments. The first mega-environment consists of Moretina Jiru, Debre Tabor, Takusa and Dembia with the best performing variety Fetan, while the second mega-environment consists of Debark and Dabat with the best performing variety Hitossa. Takusa, Dembia and Debark should be included for future testing locations for durum wheat breeding program because of their good discriminative abilities. Grain yield had negative and significant association with days to heading, days to maturity, and grain protein content, providing these traits could be used as the main selection criteria for the breeders with the index selection method to improve several traits at a time. Since, this study was limited to a small number of locations with one year data, it is better to undertake further study including more number of locations across years. Keywords: AMMI, Correlation, Durum Wheat, GEI, GGE, Stability parameter ix TABLE OF CONTENTS Contents Page