Supplementary material for "Estimating seasonal survival of migratory birds using continuous-time capture-recapture models"

Abstract

Understanding the population dynamics of migratory species requires consideration of their entire annual cycle. A key requirement is the decomposition of annual survival into seasonal periods. While there is increasing evidence that migrations are the most dangerous phase of the annual cycle, temporal variation in seasonal survival and its effect on population dynamics is still poorly understood. Estimating seasonal survival is challenging when individuals are only encountered during a restricted period of their annual cycle. We developed and tested a continuous-time capture-recapture model applicable to data from marked individuals that can only be encountered during the breeding season. This model allows annual survival to be separated into breeding and non-breeding periods. We applied this model to encounter data of hoopoes (Upupa epops) collected between 2002 and 2024 from a Swiss population during the breeding period to estimate survival during the breeding and the non-breeding periods and their respective annual variations. We then linked seasonal survival with environmental variables from the breeding, stopover, and wintering grounds, which have been documented by previous research on the same population. Mortality hazards were 17 times greater during the non-breeding than during the breeding period, with stronger annual variation observed in the former. Mortality increased during both seasonal periods when a surrogate of prey abundance at the wintering grounds was lower. The effect during the non-breeding period highlighted that environmental conditions outside the breeding season were an important driver of hoopoe population dynamics, whereas the effect during the breeding period was weaker and provided evidence of a carry-over effect. In contrast, the environmental variables considered on the breeding and stopover grounds had little impact on seasonal survival. The study adds to the growing body of evidence suggesting that migration and wintering are critical phases in the life cycle of migratory birds, and that population dynamics are strongly affected by environmental conditions encountered during wintering. The methodological approach developed is applicable to datasets that record multiple encounters of individuals during the breeding season. It enables the analysis of existing data to study temporal variation in survival across periods of the annual cycle.

Readme

The supplementary material contains R code files to analyse the data and to conduct the simulation (6 files), and the hoopoe data (3 files):

1. MainAnalyses.R: Text file with R code for the main analysis. It loads the files Functions.R and Models.R with custom-written R and NIMBLE functions, the file nimbleSummary.R for NIMBLE output processing, and the three data files HoopoeData_2002_2024.csv, NDVI.csv, and BreedingCovariates.csv.

2. Appendix_S1_1.R: Text file with R code for producing results of Appendix S1 (section 1). It loads the files Functions.R and Models.R with custom-written R and NIMBLE functions, the file nimbleSummary.R for NIMBLE output processing.

 3. Appendix_S1_3.R: Text file with code for producing results of Appendix S1 (section 3). It loads the files Functions.R and Models.R with custom-written R and NIMBLE functions, the file nimbleSummary.R for NIMBLE output processing, and the data file HoopoeData_2002_2024.csv.

4. Functions.R: Text file with R and NIMBLE functions needed for the analysis of the data and for the simulation study.

5. Models.R: Text file with the definition of the analysing models in NIMBLE language.

6. nimbleSummary.R: Text file with R functions for NIMBLE output processing.

7. HoopoeData_2002_2024.csv: Data file with Capture dates of adult hoopoes in the Valais study area from 2002-2024. It is a data matrix (4095 x 3), where each line represents a capture, and the 3 columns contain information about the bird identity ('RingNumber'), the capture date ('CaptureDate') and the capture year ('CaptureYear').

 8. NDVI.csv: Data file with scaled NDVI values from wintering, spring and autumn stopover grounds (2001-2023). It is a data matrix (23 x 4), where each line represents a year, and the 4 columns indicate the year ('year'), the winter grounds NDVI ('NDVIw'), the autumn stopover grounds NDVI ('NDVIa'), and the spring stopover grounds NDVI ('NDVIs'). See Appendix S1 (section 2) for more information about these variables.

9. BreedingCovariates.csv: Data file with data on temperature and amount of rainfall in the breeding area (2002-2023). It is a data matrix (22 x 3), where each line represents a year, and the 3 columns indicate the year ('year'), the mean amount of daily rainfall ('rain') [mm], and the mean daily temperature ('temp') [Celsius]. See Appendix S1 (section 2) for more information about these variables