Genotype by environment interaction and breeding value prediction for morphological traits in Coffea arabica

Genotype by environment (G by E) interaction is one of the main challenges in plant breeding, as it represents differential phenotypic expression of genotypes across environments. This study aimed to predict breeding values and assess the G by E interaction for morphological traits of Coffea arabica cultivars. Seven cultivars were evaluated in Araponga (20°40'05"S, 42°29'42"W, 1,198 m a.s.l.) and São Miguel do Anta (20°45'22"S, 42°40'30"W, 657 m a.s.l.), Minas Gerais, Brazil. The experiments were established in 2019 in a randomized complete block design with five replications and 15 plants per plot, spaced 2.80 × 0.70 m. In 2023, plant height (PH, m), crown diameter (CD, m), and stem diameter (SD, mm) were measured. Individual and joint Analysis of Deviance were performed to get variance components using the Restricted Maximum Likelihood (REML) method and Best Linear Unbiased Predictions (BLUPs) in a linear mixed models approach. The significance of the variance components was obtained by the Likelihood Ratio Test (LRT). Individual analyses showed significant genotypic variance for all traits except PH. Overall, the plants had a better morphological development in São Miguel do Anta than in Araponga. For example, ‘Catuaí Vermelho IAC 144’ and ‘Oeiras MG 6851’ exhibited the lowest SD and CD predictions, respectively, in São Miguel do Anta. The lower morphological development can be attributed to the environmental specificities, such as higher temperatures and less availability of water. Although CD showed significant genotypic variation in individual analyses, the combined analysis revealed strong G by E interaction for this trait, which reduced the overall genotypic variance. The marginal breeding value predictions for CD confirmed the complexity of the performance of the cultivars across environments. In contrast, SD showed no significant genotypic or G by E variances in the joint analysis, indicating phenotypic stability across environments. The overall means across environments for PH, CD, and SD were 2.01 m, 1.57 m, and 53 mm, respectively. Overall, these results indicate that PH is a relatively stable trait. Due to the unstable performance of CD across environments, it is necessary to perform an environment-specific selection or perform multi-environment testing. In contrast, SD was consistent across environments, suggesting it is a reliable trait for indirect selection in diverse growing conditions.