Table 3.

Error matrix, accuracy, and confidence interval for each class (forest and non-forest) at each threshold compared to the PRODES data.

Predictions	Class	Observed		Estimated Area (%)	Estimated Area ± 95% Confidence (km²)	User’s Accuracy (%)	Producer’s Accuracy (%)	Overall Accuracy (%)
Predictions	Class	Non-Forest	Forest	Estimated Area (%)	Estimated Area ± 95% Confidence (km²)	User’s Accuracy (%)	Producer’s Accuracy (%)	Overall Accuracy (%)
10%	Non-forest	544	96	39.3	352,159 ± 12,849	85	64	84
10%	Forest	300	1601	60.7	543,914 ± 12,849	84	94	84
30%	Non-forest	600	118	39.4	353,006 ± 12,606	84	71	86
30%	Forest	244	1579	60.6	543,067 ± 12,606	87	93	86
50%	Non-forest	688	186	38.6	346,277 ± 12,568	79	82	87
50%	Forest	156	1511	61.4	549,796 ± 12,568	91	74	87
80%	Non-forest	787	389	36.0	322,227 ± 13,082	67	93	82
80%	Forest	57	1308	64.0	573,846 ± 13,082	96	77	82
85%	Non-forest	803	461	35.2	315,401 ± 13,256	64	95	80
85%	Forest	41	1236	64.8	580,672 ± 13,256	97	73	80
90%	Non-forest	821	651	34.3	307,281 ± 13,985	56	97	73
90%	Forest	23	1046	65.7	588,792 ± 13,985	98	62	73
95%	Non-forest	832	737	33.6	302,454 ± 14,090	53	99	71
95%	Forest	12	960	66.4	593,618 ± 14,090	99	57	71
PRODES	Non-forest	748	572	36.6	327,194 ± 15,134	57	89	74
PRODES	Forest	96	1125	63.4	567.323 ± 15,134	92	66	74