Table 1. Candidate models considered in model selection and multimodel inference procedures.

Models were compared based on the second-order Akaike information criterion (AICc). K: number of model parameters; w: Akaike weight. Bold characters represent the best model for each response variable. INT: Proportion of intensive crops; FLO: Proportion of flowering crops; LOC: Local floral richness; JJ: Julian day.

Candidate models*	Weight			Thorax width			Wing length			Head length			Tibia length			Foraging activity			Pollen load richness			Difference pollen vs habitat
	K	ΔAICc	w	K	ΔAICc	w	K	ΔAICc	w	K	ΔAICc	w	K	ΔAICc	w	K	ΔAICc	w	K	ΔAICc	w	K	ΔAICc	w
1 ~ Null	4	0.00	0.76	4	0.00	0.82	4	1.85	0.20	4	0.00	0.80	4	0.00	0.79	5	198.72	0.00	4	1.92	0.17	2	21.75	0.00
2 ~ INT + FLO	6	3.38	0.14	6	3.76	0.12	6	0.00	0.49	6	3.51	0.14	6	3.36	0.15	10	2.38	0.20	6	0.00	0.45	4	14.74	0.00
3 ~ INT + FLO + LOC	7	5.57	0.05	7	5.90	0.04	7	1.94	0.19	7	5.68	0.05	7	5.54	0.05	11	3.86	0.10	7	1.34	0.23	5	16.43	0.00
4 ~ INT + FLO + LOC + INT:FLO	8	5.46	0.05	8	8.09	0.01	8	3.31	0.09	8	7.84	0.02	8	7.75	0.02	12	5.86	0.04	8	3.36	0.08	6	0.00	1
5 ~ INT + FLO + LOC + INT:FLO + INT:LOC + FLO:LOC + INT:JJ	11	8.91	0.01	11	14.71	0.00	11	8.91	0.01	11	11.59	0.00	11	12.59	0.00	14	11.02	0.00	11	4.03	0.06
6 ~ Temperature + Time of the day																8	0.00	0.66

*Temperature and time of day were also included in all models (#1 to #5) for the foraging activity response variable. The interaction INT:JJ was not included in model #5 for the foraging activity response variable.