Dryland Agricultural Research Institute, Sararoud, Kermanshah, Iran.
The occurrence of genotype × environment (GE) interaction has led to the development of several stability
parameters that can be used to estimate the stability of cultivar performance. Repeatability of 20 parametric and non-parametric stability measures across years and yield subsets as well as their association with mean seed yield and interrelationship among them in safflower (Carthamus tinctorius L.) were studied. Seed yields of 14 safflower advanced breeding lines were evaluated in 46 environments (10 locations in the 2000-2005 growing seasons, with some missing combinations) under rainfed conditions in semi-arid areas of Iran. A wide range of stability statistics, including 12 parametric and 8 non-parametric stability measures, were calculated for seed yield. The repeatability of stability measures, the interrelationships among them and their association with mean yield were estimated using Spearman’s rank correlation over environments. A combined analysis of variance revealed highly significant environmental and G × E effects over all years and yield subsets. The AMMI analysis showed that in most of the years and subsets, three first interaction principal component axes (IPCA1, IPCA2 and IPCA3) explained G × E interaction. Rank correlation between stability measures and mean yield was repeatable for four parametric measures including superiority index (Pi), geometric adaptability index (GAI), regression coefficient (bi) and environmental variance (S2
xi), as well as five non-parametric stability statistics including Nassar and Huehn (1987) stability statistics (Si (3), Si (6)) and Thennarasu (1995) measures (NPi (2), NPi (3) and NPi (4)). Rank correlations among
stability measures showed that non-parametric statistics were more correlated than parametric statistics over years and yield subsets. For example; Si (3) can be used instead of Si (2), Si (6), NPi (1), NPi (2) and NPi (3). Repeatability of the stability measures obtained in consecutive single years was low but moderate for bi and GAI in subsets, and highly repeatable for Pi, GAI, NPi (3) and NPi (2) in year/subsets versus the remaining environments. Superiority measure (Pi) and geometric adaptability index (GAI), along with two Thennarasu non-parametric stability measures (NPi (2) and NPi (3)), displayed strong rank correlations with seed yield, and high repeatability. Therefore, these statistics can be used simultaneously with seed yield to select genotypes with high yield and high yield stability in safflower breeding programs.