Table 1.

Performance of models predicting traits and diversity metrics

		Amphibians	Bats	Bees	Birds	Carabids	Reptiles
Body size	R2obs-pred	62	44	32	18	40	62
	% inc MSE	17	25	12	62	47	62
Feeding	R2obs-pred	67	9	55	20	19	55
	% inc MSE	16	52	13	57	83	13
Mobility	R2obs-pred	17	31	32	16	46	42
	% inc MSE	14	42	12	62	71	8
Reproduction	R2obs-pred	62	65	57	52	48	33
	% inc MSE	19	40	38	44	57	93
Sp. Richness	R2obs-pred	68	56	80	46	61	70
	% inc MSE	53	29	48	70	39	15
FDis	R2obs-pred	53	54	26	16	18	29
	% inc MSE	31	34	13	50	49	15
FRic	R2obs-pred	53	29	46	20	59	59
	% inc MSE	39	23	36	38	40	18
FEve	R2obs-pred	5	50	10	8	17	11
	% inc MSE	60	48	30	64	71	22

Summary statistics of random forests models of community-weighted means of traits and functional diversity metrics. “R²_obs-pred” is the performance of the model on the independent test dataset where high values indicate that the response variable is well-predicted by urban and forest land cover, climate, and latitude. “R²_obs-pred” was calculated as R-squared of the relationship between the predicted and the observed values of the independent test dataset. “% inc MSE” is the average increase in squared residuals when the variable is permuted. It represents the specific contribution (or importance) of the percentage of urban land cover (within a 500 m radius for all other taxa except birds for which we used a 1000 m radius) to the overall model performance. High values suggest that urban land cover is an important predictor.