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

Table 4

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

Models			Calibration set			Validation set
	Model	Pretreatment	RMSEC	Rc^2	AIC	RMSEP	Rp^2	RPD
Field spectra	PLSR	None	0.89	0.99	-123	1.05	0.80	1.88
		FD	0.550	0.99	-186	0.88	0.86	2.26
		K	0.67	0.86	63	0.75	0.86	2.97
		FD-K	0.64	0.89	132	1.04	0.80	1.90
	RF	None	0.9	0.79	-	0.79	0.76	3.03
		FD	0.74	0.85	-	0.79	0.89	3.03
		K	0.87	0.80	-	0.87	0.77	3.03
		FD-K	0.60	0.90	-	0.64	0.88	3.77