Environment-aware genomic prediction enhances the transferability of polygenic resistance to ash dieback in Fraxinus excelsior

Ash dieback caused by Hymenoscyphus fraxineus threatens European ash (Fraxinus excelsior L.) across its range, yet natural populations retain heritable, polygenic variation in disease response. A major challenge for genomic prediction in long-lived trees is reduced transferability across heterogeneous environments, where genotype-by-environment (GxE) interactions may influence phenotypic expression. Here, we combined nationwide sampling across Poland (320 trees from 107 populations), whole-genome SNP data, and climate-derived predictors to test whether modelling environmental similarity and GxE can improve the prediction of ash dieback severity, quantified using a synthetic tree damage index (Syn). Environmental ordination identified a primary hydroclimatic gradient as a key driver of Syn (PC1env: {beta} = 0.45 {+/-} 0.14, p = 0.0016), although broad-scale environmental predictors explained only a modest proportion of phenotypic variance. Genome-wide association analyses revealed substantial additive genetic signal (SNP-based heritability h{superscript 2}SNP = 0.63; extreme-phenotype h{superscript 2}SNP = 0.81) and identified 414 suggestive loci (p < 1 x 10-), consistent with a broadly polygenic architecture of resistance, but with pronounced local enrichment of association signals in two candidate regions on chromosomes 2 and 4. In genomic prediction, trait-enriched SNP panels consistently outperformed random panels across marker densities. Predictive ability reached r {approx} 0.89 in internal validation for a 500-SNP panel and remained robust (r {approx} 0.80) in an independent external validation set (n = 64). Incorporating GxE in a multi-kernel framework yielded modest but consistent gains over main-effect models, particularly under environmental extrapolation, with REML variance partitioning supported a non-zero interaction component (VGxE {approx} 14.9% and 20.9%). Our results demonstrate that ash dieback resistance is predictably polygenic and that accounting for environmental heterogeneity enhances the robustness and transferability of genomic prediction, supporting environment-aware selection and assisted migration strategies for European ash restoration.