Relatedness and inbreeding
The inbreeding coefficient (Fis ) was calculated with GENETIX v4.05.2 (Belkhir et al., 2004) on all genotyped females and one offspring per nest (n=665). Pairwise relatedness coefficients (r ) were calculated with the R package related v1.0. (Pew et al., 2015) for all parent pairs found using COLONY, using sample allele frequency. Simulations were performed based on the observed allele frequency to determine which of the four relatedness estimators implemented in the package performs best on our data (i.e. is the closest to expected relatedness values). A hundred simulated genotypes for each of the four relationships were generated and their rcoefficient was calculated (parent-offspring, expected r = 0.5; full-sibling, expected r = 0.5, half-sibling, expected r = 0.25, unrelated, expected r = 0). The Wang estimator, which yielded the best fit to expected relatedness values, was then used for relatedness calculation of reconstructed parent pairs (Supplementary Data 3). Relatedness of reconstructed vs potential parent pairs was calculated, and a two-tailed Student’s t-test on the mean was performed to test for significant differences between the two relatedness distributions. To explore the effect of relatedness on fitness, clutch size, total number of hatchlings, and number of dead embryos were plotted against the relatedness of parent pairs. Additionally, the relatedness of pairs involving i) a male that mated with multiple partners (within a season or across seasons), ii) a female that mated with multiple partners (within a season or across seasons), iii) the female and dominant male from clutches with multipaternity, and iv) a female that that nested over several years, were compared with the relatedness of all the reconstructed parent pairs to determine whether any of these behaviours led to a deviation of the observed relatedness, and significant differences on the mean relatedness were tested with a two-tailed Student’s t-test.
Results