Table 5.

Confusion matrix of the disease classification using BLR models based on spectral indices and the PLSDA model when applied to the validation dataset (n = 758).

Model	Predicted	Actual		Sensitivity (%)	Specificity (%)	Overall accuracy (%)	Kappa
		Healthy 0	Diseased 1
SRRed−edge	0	652	40	16.67	91.83	87.07	0.07
	1	58	8
CIRed−edge	0	660	40	16.67	92.96	88.13	0.09
	1	50	8
NDVI	0	667	40	16.67	93.94	89.05	0.10
	1	43	8
ND705	0	659	40	16.67	92.82	87.99	0.09
	1	51	8
mSR705	0	655	40	16.67	92.25	87.47	0.08
	1	55	8
mND705	0	655	40	16.67	92.25	87.47	0.08
	1	55	8
SIPI	0	673	40	16.67	94.79	89.84	0.12
	1	37	8
PRI	0	604	26	45.83	85.07	82.59	0.17
	1	106	22
WI	0	641	24	50	90.28	87.73	0.28
	1	69	24
NDWI	0	690	20	58.33	97.18	94.72	0.56
	1	20	28
PSRI	0	667	34	29.17	93.94	89.84	0.21
	1	43	14
RRDIRed−edge	0	545	30	37.5	76.76	74.27	0.06
	1	165	18
NPQI	0	415	4	91.67	58.45	60.55	0.13
	1	295	44
CAI	0	678	34	29.17	95.49	91.29	0.25
	1	32	14
HI	0	616	14	70.83	86.76	85.75	0.32
	1	94	34
DSI	0	264	26	45.83	37.18	37.73	−0.03
	1	446	22
PLSDA	0	658	2	95.83	92.68	92.88	0.60
	1	52	46