Stability and Adaptability of Cultivars in Non‐balanced Yield Trials. Comparison of Methods for Selecting ‘High Oleic’ Sunflower hybrids for Grain Yield and Quality

Abstract

AbstractThe best‐yielding and most stable cultivars are identified by growing cultivars in different environments. The stability of grain‐quality traits has been less thoroughly investigated than the stability of grain yield. High‐oleic hybrids of sunflower have been available on the Argentinian seed market for several years. Research on the stability of these genotypes is scarce. The objectives of this work were (i) to compare, using three different methods, the stability and adaptability of high‐oleic hybrids for grain yield and oil and oleic acid contents, and (ii) to explore the advantages and disadvantages of each method in selecting stable or adapted genotypes with high grain yield and high quality. Stability and adaptability analyses were performed on results for grain yields and oil and oleic acid contents of 35 high‐oleic sunflower hybrids from 17 comparative yield trials conducted over 2 years in Argentina. Stability was estimated using two methods: Fisher's protected least significant difference (LSD) test, which compared hybrids with the best‐yielding hybrid in each environment, and the test of relative yield (RY), which uses standard deviation as the measure of stability. Adaptability was estimated using Piepho's method of ‘multiple comparisons with the best’. These three methods can be applied to unbalanced data. Piepho's method made little discrimination amongst the hybrids. The LSD and RY tests coincided in classifying the hybrids as stable and unstable in 85 % of cases for grain yield and 76 % for oil content. It is concluded that the most convenient method depends on characteristics of the experimental design and of variability of the evaluated trait. Results from the LSD test depend on the number of environments in which the cultivar is tested. The RY method is valuable for classifying some cultivars as high‐yielding and stable, avoiding the problem of high‐yielding environments biasing the general average. Use of both methods together could be effective for classifying hybrids when the number of environments is adequate.