Figure 1.

Schematic representation of the concept of phenomic selection, including the two scenarios tested in the present work: (a) S1, where the calibration model is trained with true values (TVs) and NIRS data collected at the same - reference - site and (b) S2, where the calibration model is trained with NIRS data collected at the reference site and TVs from other(s) environment(s). In both scenarios, the outcome of the prediction consists of predicted values (PVs).