Search

Additional file 1: Table S1. A summary of number of tags deployed, device type and attachment method in each year at both colonies. Table S2. Tag and harness mass as a percentage of body mass for all birds tagged at both colonies between 2014 and 2018. Tag and harness mass includes 3g for a colour-ring. For birds some birds tagged with Movetech devices in 2018 (*) the exact tag masses are missing and therefore values represent closest assumed tag masses of 24.5g based on known masses of tags from that year. This value of total tag and harness mass includes 3g for colour ring and 3.5 for harness. Figure S1. Colony area used to define foraging trips for Ribble birds (red polygon). Any GPS fixes outside the colony boundary were defined as a foraging trip. Figure S2. Colony area used to define foraging trips for Walney birds (blue polygon). Any GPS fixes outside the colony boundary were defined as a foraging trip. Table S3. Habitat classifications from the Corine European Landcover database grouped into main foraging habitat types used for habitat selection modelling.abita. Figure S3. Proportion of real gull location fixes vs. randomised pseudoabsences (Pseudo) assigned to each of seven main habitat classes ��� agricultural, coastal, freshwater, landfill, marine, urban and other (scrub woodland and other non-foraging habitats) ��� for gulls breeding at Ribble and Walney in the years before and after landfill closure. Available habitat at Ribble was dominated by agricultural and urban environments whilst marine and freshwater habitats constituted a greater proportion of the pseudoabsence locations for Walney. Table S4. Assessment of habitat selection models containing a habitat variable, site and the site*habitat interaction for all tagged birds (Table 2). All metrics are derived from a confusion matrix based on the original data. AUC (area under the receiver operating curve) ranges from 0 to 1, where 0.5 is random, and higher values indicate better model performance. CC = Correct Classification, PPP = Positive Predictive Power, NPP = Negative Predictive Power, Sen. = Sensitivity, Spec, = Specificity. Table S5. Top five candidate models to explain probability of visiting any landfill at the colony-level ranked by AIC weight. Pseudo-R2 values for the selected model - MR2= 0.0399, CR2 = 0.328. Table S6. Top five candidate models to explain trip duration (hrs) at the colony-level ranked by AIC weight. Pseudo-R2 values for the selected model - MR2 = 0.0833, CR2 = 0.299. Table S7. Top five candidate models to explain trip length (km) at the colony-level ranked by AIC weight. Pseudo-R2 values for the selected model - MR2 = 0.067 CR2= 0.271. Table S8. Top five candidate models to explain distal point (km) at the colony-level ranked by AIC weight. Pseudo-R2 values for the selected model - MR2 = 0.0764, CR2 = 0.269. Table S9. (Generalised) Linear Mixed Model estimates �� standard error for foraging trip duration (hrs), trip length (km) and distal point distance (km) for lesser black-backed gulls in relation to landfill status and colony, with bird ID fitted as a random intercept, Estimates are from the model with the lowest AIC (see Tables S6 ��� S8). S10. Estimates for the effect of an interaction between the habitat variable and landfill status on the probability of a location being a real gull location or a pseudo-absence based on distal foraging trip locations only. Delta (��) AIC refers to the change in AIC caused by removing the interaction. If �� AIC > 2, the interaction is not significant meaning we found no evidence for an effect of breeding habitat on selection for that habitat type. If the interaction effect is significant, habitat selection varied with landfill status. Models were run separately for each site. Stars next to p-values represent significance levels (* < 0.05; ** < 0.01; *** < 0.001). Figure S4. Estimates and 95% confidence intervals from resource selection models for all GPS-tagged lesser black-backed gulls breeding at Ribble and Walney before (gold) and after (purple) closure of the focal landfill site based on distal trip locations. Models estimate the probability of a given location point being a real gull location rather than a pseudo-absence in response to five main foraging habitat categories (agriculture, coastal, landfill, marine, urban). A probability of 0.50 indicates that birds used habitat in proportion to its availability whilst values of > 0.50 indicate selection for that habitat type at the colony-level. Table S11. Assessment of habitat selection models for distal foraging trip locations containing a habitat variable, site and the site*habitat interaction (Table S15). All metrics are derived from a confusion matrix based on the original data. AUC (area under the receiver operating curve) ranges from 0 to 1, where 0.5 is random, and higher values indicate better model performance. CC = Correct Classification, PPP = Positive Predictive Power, NPP = Negative Predictive Power, Sen. = Sensitivity, Spec, = Specificity.