Predicting Soil Organic Carbon Content Using Hyperspectral Remote Sensing in a Degraded Mountain Landscape in Lesotho

Table 3

Performance of all RF regression and PLSR models in the calibration and validation laboratory datasets (RMSEC and RMSEP in g 100 g⁻¹).

Models			Calibration set			Validation set
	Model	Pretreatment	RMSEC	Rc^2	AIC	RMSEP	Rp^2	RPD
Lab spectra	PLSR	None	0.60	0.90	126	0.87	0.86	2.27
		FD	0.46	0.94	90.98	0.99	0.82	2.00
		K	0.12	0.99	82	0.95	0.84	2.10
		FD-K	0.63	0.89	131.8	0.99	0.82	2.00
	RF	None	0.65	0.89	—	0.79	0.82	3.03
		K	0.80	0.84	—	0.79	0.86	3.03
		FD	0.64	0.89	—	0.79	0.84	3.03
		FD-K	0.59	0.90	—	0.64	0.87	3.77