Improving G×E Interaction Analysis Through Spatial and Non-Spatial Mixed Models in Multi-Environment Trials of Niger Seed Genotypes

Multi-environment trials (METs) are central to plant breeding programs for evaluating genotype performance and adaptation, yet spatial field variability and genotype × environment interaction (GEI) often reduce the precision of genotype assessment. This study aimed to improve genotype evaluation by integrating spatial linear mixed models, GGE biplot analysis, and parametric and non-parametric stability statistics. Grain yield data from seven environments were analysed using linear mixed models fitted by restricted maximum likelihood. Non-spatial randomized complete block design (RCBD) model was compared with two-dimensional first-order autoregressive spatial model on an environment-specific basis. Spatial model provided a superior fit in three environments, while non-spatial model was adequate in the remaining environments, demonstrating that spatial dependence was not uniform across environments. Genotypic differences for grain yield were detected in most environments, with BLUEs ranging from 0.75 to 0.95 t ha⁻¹ and an overall mean of 0.84 t ha⁻¹. The average-environment coordination view identified Genotypes 3 and 5 as closest to the ideal genotype. Parametric and non-parametric stability analyses supported the GGE results. Overall, the study demonstrates that integrating spatial modelling, GGE biplots, and stability statistics provides a robust framework for accurate genotype evaluation and reliable varietal recommendation in plant breeding programs.