Evaluating Linear and Non-linear Genotype–Environment Interactions in Pearl Millet

Abstract

The nature of genotype–environment (GE) interactions was investigated for two F1 hybrids and eight open pollinated varieties (OPVs) of pearl millet for grain yield in 12 environments spread over 2 years (2010 and 2011) across the pearl millet growing belt of Sudan. The joint regression analysis showed significant linear and non-linear GE interactions corresponding to heterogeneity and remainder mean squares. However, the GE interactions of all genotypes except PGO9PM1 were explained by the linear regression model. Six OPVs and hybrid PGO9PM3 showed general adaptation with b ≈ 1.0 and non-significant remainder mean squares. Of these genotypes, while hybrid PMO9PM3 was the highest yielding (917 kg ha−1) farmers could adopt any genotype by trading off between their desirable traits such as mean grain yield, earliness, fodder yield and quality criteria etc. Two varieties (ISC-III and MCNELC), were specifically adapted to below average environments with their mean grain yields non-significantly different from the other six OPVs. The highest yielding hybrid PMO9PM1, on the other hand, showed specific adaptation to favourable environments but also had large remainder mean squares. More complex models such as quadratic, 2-intersecting-straight lines, 3-intersecting-straight lines were fitted which, however, could not account for the large remainder mean squares. A 3-lines model with quadratic component in the higher yielding segment of environments was found adequate showing that the upper threshold value for the hybrid was not reached and it would continue responding to higher yielding environments.