Comparative Evaluation of AMMI, GGE, GYT, and Bayesian Models for Risk-Based Selection of Forage Sorghum in multi-environmental trials

<title>Abstract</title> <p> Forage sorghum [ <italic>Sorghum bicolor</italic> (L.) Moench] is a low-input, drought-tolerant, high-biomass multipurpose crop whose deployment is constrained by strong genotype × environment interaction (GEI). We evaluated 30 forage sorghum genotypes across five environments spanning Assam and Hyderabad during the kharif, rabi and summer seasons of 2020–2021. Twelve morphological and yield traits were recorded; green forage yield per plant (GFYP) is the focal selection target and all analysis were done using a dual analytical framework. Classical multivariate stability was characterised through additive main-effects and multiplicative interaction (AMMI), genotype plus G×E interaction (GGE) biplot and genotype-by-yield*trait (GYT) biplot analyses, while a complementary Bayesian multi-environment model delivered posterior genotypic effects, variance-component inference and probability-based superiority and stability metrics. Genotype, environment and GEI effects were highly significant (P &lt; 0.01); the first two interaction principal components captured 88.6% of GEI, validating the AMMI biplots. All AMMI stability parameters and the GGE biplot converged on G24 (348B), G26 (314B), G30 (SSG-59-3) and G19 (327B) as the most stable, high-yielding genotypes, and resolved E3, E2 and E1 as ideal, discriminating test environments within a two mega-environment target region. The Bayesian model showed robust convergence (Bayesian P-values ≈ 0.5; genetic variance ≈ 96% of total; close empirical–sampled density agreement) and a predominantly genetic variance structure; posterior probability rankings independently identified G30 (SSG-59-3), G24 (348B), G21(CSV32F) and G25 (424B) as having the highest probability of superior performance (&gt; 0.75), while joint performance stability probabilities confirmed that high mean yield alone does not guarantee reliability. The GYT biplot revealed strong positive associations of yield with SGT, FDF, NLP, PL, LFW, LAI, LFL and NNP. The strong concordance between frequentist stability statistics and probabilistic risk metrics provides a rigorous, decision-oriented basis for recommending stable, widely adapted forage sorghum and demonstrates the value of integrating classical and Bayesian approaches in multi-environment cultivar evaluation in climate-resilient forage sorghum improvement. </p>