Table 4.

Coefficient of determination and errors of training and test of nine modelling methods.

Modelling Methods †††	Training		Test		Wavelength of Max. R2
	R2	RMSE (g kg−1)	R2	RMSE (g kg−1)
DVI	0.45	3.77	0.42	4.55	2170 nm, 2160 nm
RVI	0.40	5.98	0.38	6.15	1720 nm, 580 nm
NDVI	0.35	7.06	0.32	7.48	1720 nm, 580 nm
PLSR	0.85	2.07	0.76	3.46	——
SVR	0.94	1.57	0.83	1.78	——
NN	0.95	1.33	0.86	1.66	——
Adaboost-SVR	0.93	1.58	0.85	1.66	——
AdaBoost.RT-BP	0.96	1.03	0.91	1.29	——

^††† DVI, RVI, NDVI, PLSR, SVR and NN represent difference vegetation indexes, ratio vegetation indexes, normalized differential vegetation indexes, partial least squares regression, support vector regression and neural networks, respectively. R² is the coefficient of determination; RMSE is the root mean squared errors.