Integration of UAV-Based Phenotyping Increases Reliability and Supports Modeling of Genotypic Responses to Environmental Covariates in Eucalyptus

High-resolution unmanned aerial vehicle (UAV) phenotyping enables detailed assessment of crop-growth dynamics, yet spatial heterogeneity and environmental responsiveness remain critical challenges for accurate genetic inference. This study evaluated UAV-derived canopy height and RGB-based vegetation indices across developmental stages in two contrasting environments, integrating spatially adjusted mixed models and genotype-specific regression to environmental covariates. Spatial distribution patterns revealed structured heterogeneity across growth stages, particularly during early development, with persistent gradients across sites. Correlation analyses among BI, GLI, H, NGRDI, and VARI indicated increasing convergence as canopy architecture consolidated. However, VARI exhibited the most stable and consistent associations across models and phases, aligning with structural canopy development. Spatially adjusted models consistently increased heritability, especially in early phases, demonstrating improved partitioning of genetic and residual variance. Differences between spatial-only and spatial plus UAV-covariate models were minor, indicating that fine-scale spatial structure was the primary source of residual heterogeneity. Reaction norms to environmental covariates revealed that in São Paulo, precipitation and temperature differentiated clone performances, whereas in Mato Grosso do Sul, growing degree days were the dominant driver. Finally, integrating spatial correction, UAV-based phenotypes, and environmental regression enhances genetic parameter estimation and enables dynamic characterization of genotype performance, supporting precision-oriented breeding under heterogeneous, climatically variable conditions.