Table 5. Confusion matrix and classification accuracy applied in machine learning models. Values of the confusion matrix are ratios relative to the total number of pixels. (OA: Overall accuracy, Kappa : Cohen’s Kappa).

Model	MLC	Wildfire severity classes		Reference Data
				Crown Fire(%)	Heat Damage(%)	Surface Fire(%)
		Prediction	Crown Fire	15.0	3.7	2.0
			Heat Damage	4.9	10.0	9.6
			Surface Fire	1.5	4.8	48.4
		OA(%): 73.4		Kappa: 0.54
	RF	Wildfire severity classes		Reference Data
				Crown Fire(%)	Heat Damage(%)	Surface Fire(%)
		Prediction	Crown Fire	16.2	4.9	8.0
			Heat Damage	3.5	9.5	12.6
			Surface Fire	1.8	4.0	39.5
		OA(%): 71.9		Kappa: 0.50
	SVM(Linear)	Wildfire severity classes		Reference Data
				Crown Fire(%)	Heat Damage(%)	Surface Fire(%)
		Prediction	Crown Fire	15.0	3.2	1.0
			Heat Damage	4.4	9.8	16.8
			Surface Fire	2.1	5.5	42.2
		OA(%): 65.3		Kappa: 0.43
	SVM(RBF)	Wildfire severity classes		Reference Data
				Crown Fire(%)	Heat Damage(%)	Surface Fire(%)
		Prediction	Crown Fire	17.7	6.8	4.2
			Heat Damage	2.1	4.2	5.8
			Surface Fire	1.7	7.5	50.0
		OA(%): 67.0		Kappa: 0.45