Journal of Avian Biology

Transcription

1 Journal of Avian Biology JAV-01 McKnight, A., Blomberg, E. J., Golet, G. H., Irons, D. B., Loftin, C. S. and McKinney, S. T. 01. Experimental evidence of long-term reproductive costs in a colonial nesting seabird. J. Avian Biol. 01: e01 Supplementary material

2 Appendix 1 Table A1. Coefficient values for survival parameters in the top-ranked apparent survival model structure for kittiwakes involved in a long-term cost of reproduction experiment at the Shoup Bay colony in Prince William Sound, Alaska. The top-ranked model did not include the manipulation term, indicating that forced nest failures during -1 had little effect on long-term survival. Beta Estimate SE % CI Significant intercept * state (breeders) trend * state (breeders) X trend *

3 Table A. Coefficient values for transition parameters in the top-ranked transition model structure for kittiwakes involved in a long-term cost of reproduction experiment at the Shoup Bay colony in Prince William Sound, Alaska. The top-ranked model included the manipulation term, indicating that forced nest failures during -1 increased the probability of breeding over the long term. Beta Estimate SE % CI Significant Comments intercept * state (breeders) manip * fixed * * * * * * * confounded state (breeders) X * fixed state (breeders) X * state (breeders) X * state (breeders) X state (breeders) X state (breeders) X state (breeders) X state (breeders) X state (breeders) X state (breeders) X state (breeders) X state (breeders) X state (breeders) X state (breeders) X state (breeders) X 00 confounded Long-term cost of reproduction Supporting Information S

4 Table A. Tag loss sensitivity analysis: Performance of multi-state models estimating the encounter probability (p) for black-legged kittiwakes at the Shoup Bay colony, Prince William Sound, Alaska, during 1-0 with all individuals with degraded marks and reconstructed identities removed. Model structures for apparent survival and transition probability were held constant as state X time. Model weights are denoted by w i, and K represents the number of estimable parameters in each model adjusted for any parameters fixed during analysis. ΔQAIC c values reflect ΔAIC c values adjusted according to a median ĉ estimate of 1.0. Boldface denotes top scoring model. Model ΔQAIC c w i K p (state + year) 0.00 >0. p (state * linear time trend) 1. <0.01 p (state + linear time trend) 1. <0.01 p (state) 1. <0.01 p (state * year). <0.01 p (linear time trend).1 <0.01 p (year) 1.1 <0.01 p (constant). < Long-term cost of reproduction Supporting Information S

5 Table A. Tag loss sensitivity analysis: Performance of multi-state models estimating the probability of apparent survival (φ) for black-legged kittiwakes at the Shoup Bay colony, Prince William Sound, Alaska, during 1-0 with all individuals with degraded marks and reconstructed identities removed. Model structure for transition probability was held constant as state X time, and encounter probability model structure was set to the best competing structure from Table S. Model weights are denoted by w i, and K represents the number of estimable parameters in each model adjusted for any parameters fixed during analysis. ΔQAIC c values reflect ΔAIC c values adjusted according to a median ĉ estimate of 1.0. Boldface denotes top scoring model. Set Model ΔQAIC c w i K Notes General structure φ (state * trend) φ (state + trend) φ (state + period) φ (state * period). <0.01 φ (trend). <0.01 φ (state + year) 1.0 <0.01 φ (state) 1. <0.01 φ (year). < φ (state * year) 0.1 <0.01 φ (constant) 1. <0.01 Temporal covariates φ (state * trend) 0.00 >0. 0 φ (state * prod).0 < φ (state * PDO) 1.0 < Other covariates φ (state * trend) φ (state * trend + manip) Insignificant manipulation effect φ (state * trend + 1 skip) 1. <0.01 Insignificant -1 skipped breeding effect Long-term cost of reproduction Supporting Information S

6 Table A. Tag loss sensitivity analysis: Performance of competing structures for multi-state models estimating the probability of transition (Ψ) between breeding and non-breeding states for Blacklegged Kittiwakes at the Shoup Bay colony, Prince William Sound, Alaska, during 1-0 with all individuals with degraded marks and reconstructed identities removed. Model structure for apparent survival probability was held constant as state X time, and encounter probability model structure was set to the best competing structure from Table A. Model weights are denoted by w i, and K represents the number of estimable parameters in each model adjusted for any parameters fixed during analysis. ΔQAIC c values reflect ΔAIC c values adjusted according to a median ĉ estimate of 1.0. Boldface denotes top scoring model. Set Model ΔQAIC c w i K Notes General structure Ψ (state * year) 0.00 >0. Ψ (state * trend) 0. < Ψ (state + year) 1.1 <0.01 Ψ (state + trend) 1.1 <0.01 Ψ (state + period) 1. <0.0 1 Ψ (state) 1.1 <0.01 Ψ (year).0 < Ψ (trend).1 <0.01 Ψ (period). < Ψ (constant) 0. <0.01 Temporal covariates Ψ (state * year) 0.00 >0. Ψ (state * prod) 1. < Ψ (state * PDO) 0. < Other covariates Ψ (state * year + manip) Significant manipulation effect Ψ (state * year) Ψ (state * year + 1 skip) Insignificant -1 skipped breeding effect Long-term cost of reproduction Supporting Information S

7 Table A. Breeding state designation sensitivity analysis: Performance of multi-state models estimating the encounter probability (p) for black-legged kittiwakes at the Shoup Bay colony, Prince William Sound, Alaska, during 1-0 using a criterion of two sightings on a nest site to designate breeding status, rather than the three sightings used in the primary analysis. Model structures for apparent survival and transition probability were held constant as state X time. Model weights are denoted by w i, and K represents the number of estimable parameters in each model adjusted for any parameters fixed during analysis. ΔQAIC c values reflect ΔAIC c values adjusted according to a median ĉ estimate of 1.0. Boldface denotes top scoring model. Model ΔQAIC c w i K p (state + year) p (state) p (state * linear time trend) p (state + linear time trend) p (state * year) p (year) p (linear time trend) p (constant) Long-term cost of reproduction Supporting Information S

8 Table A. Breeding state designation sensitivity analysis: Performance of multi-state models estimating the probability of apparent survival (φ) for black-legged kittiwakes at the Shoup Bay colony, Prince William Sound, Alaska, during 1-0 using a criterion of two sightings on a nest sight to designate breeding status, rather than the three sightings used in the primary analysis. Model structure for transition probability was held constant as state X time, and encounter probability model structure was set to the best competing structure from Table A. Model weights are denoted by w i, and K represents the number of estimable parameters in each model adjusted for any parameters fixed during analysis. ΔQAIC c values reflect ΔAIC c values adjusted according to a median ĉ estimate of 1.0. Boldface denotes top scoring model. Set Model ΔQAIC c w i K Notes General structure φ (state + trend) φ (state * trend) φ (state + year). 0.1 φ (state + period) φ (state * period) φ (state * year) φ (trend) φ (state) φ (year) φ (constant) Temporal covariates φ (state + trend) φ (state + prod) φ (state + PDO) Other covariates φ (state + trend) φ (state + trend + 1 skip) Insignificant -1 skipped breeding effect φ (state + trend + manipulation) Insignificant manipulation effect Long-term cost of reproduction Supporting Information S

9 Table A. Breeding state designation sensitivity analysis: Performance of multi-state models estimating the probability of transition (Ψ) between breeding and non-breeding states for black-legged kittiwakes at the Shoup Bay colony, Prince William Sound, Alaska, during 1-0 using a criterion of two sightings on a nest sight to designate breeding status, rather than the three sightings used in the primary analysis. Model structure for survival was held constant as state X time, and encounter probability model structure was set to the best competing structure from Table A. Model weights are denoted by w i, and K represents the number of estimable parameters in each model adjusted for any parameters fixed during analysis. Significance of covariate effects noted in comments for competitive models. ΔQAIC c values reflect ΔAIC c values adjusted according to a median ĉ estimate of 1.0. Boldface denotes top scoring model. Set Model ΔQAIC c w i K Notes General structure Ψ (state * year) 0.00 >0. Ψ (state * trend) 1.1 < Ψ (state + year) 1. <0.01 Ψ (state + trend) 1.0 <0.01 Ψ (state + period) 1.0 < Ψ (state) 1. <0.01 Ψ (year).1 < Ψ (trend) 1. <0.01 Ψ (period). < Ψ (constant) 0.1 <0.01 Temporal covariates Ψ (state * year) 0.00 >0. Ψ (state * prod) 1. < Ψ (state * PDO) 0.1 < Other covariates Ψ (state * year + manip) Significant negative manipulation effect Ψ (state * year) Ψ (state * year + 1 skip). <0.01 Insignificant -1 skipped breeding effect Long-term cost of reproduction Supporting Information S

10 Figure A1. Manipulation effect over time: Annual probabilities of transitioning from a breeding to a non-breeding state for Shoup Bay, Alaska, kittiwakes by the number of experimentally forced nest failures during -1, from the multi-state transition model with a three way interaction among breeding state, manipulation category, and year (ΔQAIC c = 1., w i < 0.01). Long-term cost of reproduction Supporting Information S

11 Figure A. Manipulation effect over time: Annual probabilities of remaining in a non-breeding state for Shoup Bay, Alaska, kittiwakes by the number of experimentally forced nest failures during -1, from the multi-state transition model with a three way interaction among breeding state, manipulation category, and year (ΔQAIC c = 1., w i < 0.01). Long-term cost of reproduction Supporting Information S