Unraveling the stability of Panicum maximum in multiple harvest-location trials using a probabilistic model

• Ano de defesa: 2023
• Autor(a) principal: Murakami, Vitória Bizão
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Bayesian models; Cultivar recommendation; Forage breeding; Melhoramento de forrageiras; Modelos Bayesianos; Recomendação de cultivar; Variance-Covariance; Variância-covariância
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/11/11137/tde-06022024-113450/

Evaluating cultivar response under different environmental and annual conditions is a critical phase of perennial crop breeding. These multi-harvest location trials allow access to genotypeby-environment (GEI) and genotype-by-harvest (GHI) interactions, which are the main causes of differential phenotypic response across location and years. In this context, linear mixed models and Bayesian models are useful to capture the expression of genotype diversity across location and harvests. Therefore, the objectives of this study were (i) to evaluate different variance-covariance structures for multiple harvest-location trials, and (ii) to explore the genotype-by-environment (GEI) and genotype-by-harvest (GHI) interactions to assess the adaptability and stability of Panicum Maximum. Dry leaf matter phenotypic data were measured in 23 genotypes in a complete randomized block design with up to seventeen harvests in five locations. The covariance structures of the random effects were modeled and their adequacy was tested by the Akaike and Bayesian information criteria. From the selected model, variance components, genetic parameters and adjusted means were estimated. Models that accounted for heterogeneity in the variance-covariance structures were best fitted. We fitted four Bayesian models with homogeneous (M1, M3) and heterogeneous (M2, M4) residual standard deviations. Based on the model selected by WAIC2 (M2), genotypes PM40, MASS, and PM41 had the highest global and pairwise probability of superior performance for LDM. When analyzing the performance within environments, the genotype PM32 showed an adaptation for the site AC. On the reaction norm plot, we observed that the genotype-by-harvest had a complex significant interaction but could not change more than two positions in the rank, reflecting the homogeneity of the probability of performance along harvests. In terms of stability across locations, genotypes TANZ, PM44 and PM42 were the best. The visual representation of probabilities provided straightforward insights into genotype adaptation patterns across environments and harvests, allowing comparison of genotype performance. Therefore, our results support decision making processes when recommending genotypes and reduce the risk of carrying poor performing genotypes into the next breeding phase.