Phenomic selection perspectives in wheat breeding programs

Wheat breeding in Mediterranean environments requires phenotyping approaches able to increase measurement scale while preserving biological meaning and breeding relevance. This thesis evaluated the use of phenomic information in wheat breeding through three complementary applications: field-based description of early canopy development, temporal genetic interpretation of plant architecture, and prediction of agronomic and quality traits using high-dimensional post-harvest spectra. The work included both durum wheat and bread wheat, providing evidence that phenomic approaches can be adapted to species with different genetic backgrounds and end-use targets but similar developmental constraints under Mediterranean conditions. The first study assessed UAV-derived RGB vegetation indices in durum wheat and showed that low-cost image-based traits can detect genotypic and agronomic differences in early canopy variation. In particular, indices such as NGRDI and TGI provided useful descriptors of early vigour, ground coverage and canopy establishment. The second study extended field phenomics to canopy structure in bread wheat. UAV-derived digital canopy height, modelled across time, allowed the separation of cumulative height from growth rate and supported the detection of stage-specific genetic effects. Growth-rate traits identified genetic associations earlier than whole-height traits and revealed a temporal sequence of developmental control related to phenology and plant architecture. The third study evaluated NIR spectra as static high-dimensional phenomic profiles for the prediction of agronomic, phenological and quality traits in bread wheat. NIR-based phenomic prediction was competitive with genomic prediction, particularly for grain protein content and related indices, while combined genomic-phenomic models showed trait- and scenario-dependent advantages. Optimized reduced training sets also maintained useful prediction accuracy, supporting the practical feasibility of phenomic prediction in breeding pipelines. Overall, the thesis shows that dynamic UAV-derived traits and static NIR-derived profiles have complementary roles. RGB and canopy-height traits are suitable for field screening, developmental characterization and genetic interpretation, whereas NIR spectra can support post-harvest prediction of performance and quality. These results indicate that phenomics can contribute to wheat breeding when sensor-derived data are converted into repeatable, interpretable and operationally useful traits or predictors.