Influence of individual tree characteristics, spatial structure and logging history on tree-related microhabitat occurrence in North American hardwood forests

Forest Ecosystems

Table 5 Results of the logistic regressions analyzing the probability of occurrence of the different TreM classes. n: number of observations, df: degrees of freedom, χ²: chi-square, AIC: Akaike information criterion, AUC: Area under the ROC curve, Std.Err: standard error, “^”: significance at p < 0.001, “^”: significance at p < 0.01, “^”: significance at p < 0.05

Model	n	df	χ²	p	AIC	AUC	Variables in the final model
Bark loss	365	2	4.03	0.132	331.41	0.62	Tree vigor
Borken branch or top	365	2	38.87	< 0.001^***	176.51	0.84	DBH, TreM-trees aggregation
Crack	365	2	6.00	0.049^*	69.21	0.82	DBH
Crown deadwood	365	4	36.33	< 0.001^***	150.70	0.83	DBH, tree vigor, management
Saproxylic fungi	365	4	6.43	0.168	84.34	0.79	DBH, tree vigor, management
Trunk base rot hole	365	2	3.51	0.172	79.44	0.76	DBH
Trunk rot hole	365	2	25.75	< 0.001^***	279.11	0.74	DBH
Woodpecker lodge	365	4	39.07	< 0.001^***	147.96	0.83	DBH, tree vigor, TreM-trees aggregation