TABLE 4.

Selected components analysis to reduce the dimensionality of the phenomics data. For each trial year (combining both sites), the first three principal components (PCO) are shown as well as the components with the highest loadings in a Partial least squares regression (PLS analysis) with yield as the response variable. For the PLS components, the trait categories with the highest loadings are indicated X). See text for dates and approximate growth stages when measurements were taken.

Trial year	Component	Correlation to Yield	Main composition (PLS loadings)
			Early multispectral	Late multispectral	Hyperspectral	Field scores
2016	PCO 1	0.16
	PCO 2	0.14
	PCO 3	−0.04
	PLS 1	0.47	X		X
	PLS 2	0.35	X	X	X
	PLS 3	0.21		X	X
	PLS 4	0.17	X		X	X
2017	PCO 1	0.17
	PCO 2	0.11
	PCO 3	0.040
	PLS 1	0.63		X	X
	PLS 2	0.35	X	X	X
	PLS 3	0.25			X	X