Modeling fire ignition probability and frequency using Hurdle models: a cross-regional study in Southern Europe

Ecological Processes

Table 4 Performances of the Negative Binomial Hurdle models (hurdle and count parts) evaluated for the three study regions and each of the five subsamples

	Hurdle part						Count part
	AUC value			Prediction accuracy (%)			Pearson correlation coefficient			RMSE
	c)	s)	a)	c)	s)	a)	c)	s)	a)	c)	s)	a)
S1	0.72	0.80	0.81	0.67	0.73	0.78	0.23	0.41	0.35	0.73	2.99	0.90
S2	0.71	0.81	0.83	0.66	0.74	0.71	0.13	0.40	0.32	0.82	3.11	0.83
S3	0.71	0.79	0.84	0.66	0.72	0.79	0.15	0.43	0.39	0.77	2.85	0.79
S4	0.72	0.81	0.83	0.70	0.74	0.76	0.22	0.43	0.35	0.84	2.86	0.85
S5	0.72	0.80	0.83	0.62	0.72	0.76	0.20	0.43	0.33	0.74	2.85	0.81
ED	0.71	0.80	0.83	0.66	0.73	0.76	0.23	0.43	0.36	0.76	2.85	0.81

c) Catalonia, s) Sardinia, a) Apulia, AUC area under the curve, S1–S5 subsamples, ED entire dataset, RMSE root mean square error