Clustering stability methods towards selecting best performing and stable durum wheat genotypes

Genotype × environment (GE) interaction is one of the main challenges in identifying stable genotypes with high mean yield performance under different growing conditions. The main objectives were to investigate the efficiency of different stability models for assessing GE interaction and to identify high-yielding and stable genotypes under diverse environments differing in water regime and climatic conditions. In this study, twenty-five durum wheat genotypes (breeding lines and new cultivars) were evaluated at six research stations, representing durum wheat growing regions, for three cropping seasons. After confirming significant GE interaction for grain yield, the phenotypic stability of durum wheat genotypes was analyzed by 34 different stability methods, as well as AMMI, GGE and Y × WAASB biplots methods. Estimation of stability indices, AMMI-based stability parameters, BLUP-based stability parameters (HMGV, RPGV, and HMRPGV), WAASB (Weighted Average of Absolute Scores from the singular value decomposition of the matrix of BLUPs) for the GE interaction effects and WAASBY (a combination of WAASB and yield) scores; and parametric and non-parametric stability methods for identifying high-yielding and stable genotypes were highlighted. The findings of this study suggest that some of the applied methods such as GGE biplot, AMMI2 biplot and Y × WAASB biplot are most advisable for enhancing selection of genotypes with high-yielding and stability performance in diverse environments. In addition, the stability methods of superiority index (Pi), geometric adaptability index (GAI), WASSBY, SIPCY (AMMIF + mean yield) and GGE distance (GGED) were significantly (P < 0.01) correlated with genotypic mean yields and can be successfully recommended to identify high-yielding and stable genotypes. Cluster analysis was applied to analyze the relationships between stability methods and genotypes simultaneously, which allowed to identify six groups of stability methods and four groups of genotypes differing in stability performance and adaptability. The applied methods resulted in identifying superior promising breeding lines i.e., G6, G20, G11, G10, G5 and G15 that out-yielded the national check varieties based on high-performing and stability. In conclusion, this study applied a complete set of univariate and multivariate models with a selection pattern based on each of model for classification and interpretation of GE interaction, which allowed the identification of efficient methods that towards selecting promising durum wheat breeding lines with high mean performance and greater phenotypic stability across the diverse rainfed conditions.