Table 2.

Summary of the generalized linear mixed-effects model results disregarding some of the components used to deal with biases associated with citizen-collected data. Only results for the highest-ranked model (veg + water + temp.range + elevation + roads) are shown. Predictors include the following: veg (all vegetation variables, i.e., koala Eucalyptus – koalaeuc, other Eucalyptus – othereuc, and other vegetation); water (density of watercourses and distance to water bodies), temp.range (temperature range), elevation, and roads (density of sealed and unsealed roads). All models included a spatial random effect (2-km² grid cell). Shown for each model are biased-corrected model probabilities based on weights of Akaike's information criterion corrected for small sample sizes (wAIC_c), the percentage of deviance explained (%De), the cross-validation error (CV_error), and Cohen's κ and both marginal and conditional R²

Disregarding:	wAICc	%De	CVerror	κ	R2(m)	R2(c)	Largest effect
Original	∼1	61.6	0.10 ± 0.01	0.46 ± 0.07	76.4	81.0	Temperature (10.13)
Offset	1	62.7	0.11 ± 0.03	0.48 ± 0.07	75.9	80.6	Temperature (10.20)
Weights	1	49.0	0.25 ± 0.02	0.15 ± 0.03	82.3	83.2	Sealed roads (21.26)
Pseudo-absences1	1	49.0	0.01 ± <0.01	0.14 ± 0.02	82.3	83.2	Sealed roads (21.26)

¹The model disregarding pseudo-absences uses the entire background environmental data within the area covered by the dashed line in Fig. 1.