The optical properties of plant tissues and organs are the result of a combination of the biochemical composition (e.g. pigments, water, carbohydrates and proteins), internal structure and physiology (e.g. chlorophyll fluorescence). Measuring the canopy spectral reflectance provides an insight of these attributes in a high throughput and non-destructive way. We have developed a suite of hyperspectral systems for screening biochemical composition of plants at different scales, from single organs to the canopy scale. Hyperspectral data, collected from field and controlled environment trials, and wet chemistry determinations of different metabolites were combined in training machine learning predictive algorithms. This approach allowed for accurate prediction of content for water, nitrogen, carbohydrates and other metabolites. These non-destructive determinations can be used in a dynamic fashion throughout the different phenological stages of the plant development. However the effects of the canopy structure and illumination geometry (sun position) have to be taken into account in order to effectively retrieve these biochemical properties. The combination of hyperspectral imaging (optical properties) and LiDAR (canopy structure) in the context of spectranomics has been proposed as the way to overcome this scale issue in natural ecosystems but its application in agricultural crops has been limited.