The influence of diet-mediated exposure of avian influenza on adult survival, recruitment and territory occupancy in peregrine falcons

We investigated the influence of waterbird exposure on breeding peregrine falcons within the mid-Atlantic region of North America by comparing inland and coastal subpopulations. We monitored individually marked adults (N = 205) and breeding territories (N = 79) to estimate spatial and temporal patterns of adult survival, recruitment age, and territory occupancy (2016-2025). Three tables are included addressing PEFA survival, site occupancy, and age at recruitment.

Methods
 Survival: We used the package ‘marked’ in Program R [59] to fit POPAN capture–recapture models. These models account for individuals entering the population throughout the study period and provide an estimate of population entry probability (pent) as well as apparent survival. Candidate models included combinations of additive and interactive effects of region (Inland/Coastal) and time (Year) on population entry probability (pent), while detection probability and superpopulation size were held constant. Model selection was based on Akaike’s Information Criterion (AIC). When more than one model was effectively equivalent (ΔAIC ≤ 2.00), and one model was a nested version of the other, we evaluated whether the additional predictor improved model fit and retained the simpler model when it did not. To test for regional differences in estimated survival and population entry probabilities, we conducted pairwise comparisons between inland and outer coast predictions derived from the POPAN survival models.

Recruitment: We provide descriptive statistics on known-age birds recruiting into the breeding population and compare age-at-recruitment for males and females using a two-tailed t-test.  We compiled the frequency of juvenile-plumaged birds recruiting into the breeding population and investigated possible temporal patterns in frequency using a G-test with Yates correction. 

Occupancy: We used the package ‘glmmTMB’  to construct mixed models with territory-level occupancy as a binomial response (occupied/unoccupied). Fixed predictors included year and region. Year was included as a categorical factor to allow non-linear and non-monotonic temporal patterns in occupancy. The model also included the interaction between year and region (inland vs. coastal) to test for differences in region-specific temporal dynamics, and we included territory location as a random intercept to account for repeated observations of the same territory over multiple years. We compared models using all combinations of additive and interactive effects of region and year, and selected the most parsimonious based on AIC, following the same procedure as we did for survival. To test for differences in predicted occupancy between regions while accounting for site-level random effects, we used a parametric bootstrap approach based on our top model. For each bootstrap iteration, we simulated a new response vector from the fitted model using the estimated fixed and random effects and refit the model to the simulated data. Using each refitted model, we generated predicted occupancy probabilities for all observed site × year combinations, including random site effects. We then averaged predicted probabilities across sites within each region and year and calculated the difference in mean occupancy between regions. This procedure was repeated across 10,000 bootstrap replicates, producing an empirical sampling distribution of the regional occupancy difference for each year.