Season-specific genetic variation underlies early-life migration in a partially migratory bird
Data files
Aug 21, 2024 version files 3.15 GB
Abstract
Eco-evolutionary responses to environmentally-induced selection fundamentally depend on magnitudes of genetic variation underlying traits that facilitate population persistence. Additive genetic variances and associated heritabilities can vary across environmental conditions, especially for labile phenotypic traits expressed through early life. However, short-term seasonal dynamics of genetic variances are rarely quantified in wild populations, precluding inference on eco-evolutionary outcomes in seasonally dynamic systems. This limitation applies to seasonal migration versus residence, constituting one key trait where rapid micro-evolution could rescue partially migratory populations from changing seasonal environments. We fitted novel quantitative genetic ‘capture-recapture animal models’ to multi-generational pedigree and year-round resighting data from 11 cohorts of European shags (Gulosus aristotelis), to estimate season-specific additive genetic variances in liabilities to migrate, and in resulting expression of migration, in juveniles’ first autumn and winter. We demonstrate non-negligible genetic variation underlying early-life migration, with twice as large additive genetic variances and heritabilities in autumn than winter. Since early-life survival selection on migration typically occurs in winter, highest genetic variation and strongest selection are seasonally desynchronised. Our results reveal complex within-year and among-year dynamics of early-life genetic and phenotypic variation, demonstrating that adequate inference of eco-evolutionary outcomes requires quantifying micro-evolutionary potential on appropriate scales and seasonal timeframes.
README: Season-specific genetic variation underlies early-life migration in a partially migratory bird
https://doi.org/10.5061/dryad.j6q573nnr
In our study, capture-recapture data and pedigree data were analysed using a Bayesian statistical model coded in Stan language and run in R using package rstan.
The model outputs, i.e. posterior samples of the model parameters and derived parameters, are the primary results of our analyses.
NB: Full understanding of these files requires careful reading of the paper and electronic supplementary materials.
Description of the data and file structure
The following files represent the data used for the analysis:
- Data file with capture-recapture histories for all individuals
crh_clean.txt
** Contains the following columns
# BirdID - identity of the individual
# Column 2 to 7 (X1 to X6) - resighting history of the individual across six defined occasions, with cell values 1 to 4 representing the four observed states (see paper for details).
# HatchYear - year of birth i.e. cohort of the individual
# BreedEventID - identification of the breeding event in which the individual was born (unique for each nest; NA values are for individuals from unknown nests/breeding events, each were attributed a unique brood_id)
# brood_id - dummy values of the brood identity attributed to each individual
# dam_id - dummy values of the mother identity attributed to each individual
# sire_id - dummy values of the father identity attributed to each individual
- Matrix representing relatedness between individuals obtained using a social pedigree
cov_matrix.RData
** Contains the following
# One row and one column for each of the 20,859 individuals contained in the pedigree, with cell values representing their relatedness to each of the other individuals
# It is a square matrix where relatedness values will be repeated in both sides of the diagonal, and diagonal represents relatedness of the individual with itself
# Please be aware this is a heavy file, and can be loaded and prepared for analyses using the "Run_CRAM_RF_clean.R" script.
Code/Software
The following files are Stan code files representing the different models used for analysing the data:
CRAM_RF_v7.9_clean - Stan model code for the primary CRAM
CRAM_RF_v7.9.1_clean - Stan model code for the brood/maternal/paternal variance model
CRAM_RF_v7.2.1_clean - Stan model code for the re-scaled variance model
CRAM_RF_v5.1_clean - Stan model code for the annual variance decomposition model
The following R code reads the data into R, codes it to appropriate input format for the analysis with Stan, runs the analysis of the Stan model, saves posterior samples of model parameters and calculates detailed numerical summaries for all model parameters and derived quantities:
Run_CRAM_RF_clean.R
The following R code builds all the figures in the paper and electronic supplementary materials representing the results from the Stan models:
Plots_CRAM_clean.R
Methods
We utilised a pedigreed partially migratory population of European shags (Gulosus aristotelis) breeding on Isle of May National Nature Reserve (hereafter ‘IoM’), Scotland (56°11′N, 2°33′W).
Data Collection
Each breeding season (April-July) since 1997, all breeding attempts on IoM were systematically monitored (223-1068 per year). Chicks reaching approximately 3 weeks old were individually marked with uniquely coded metal and colour rings readable from ≤150m with a telescope (559-1143 chicks ringed per year, mean 863, recently comprising >95% of fledglings). Ringed breeding adults were identified, or caught and colour-ringed if initially unringed, with sexes inferred through behaviour, vocalizations and/or genotyping.
During 2010-2022, we undertook large-scale non-breeding season (August-March) resighting surveys to locate colour-ringed individuals, and thereby assign juveniles as seasonal migrant or resident. Regular (approximately biweekly) surveys were undertaken on IoM and adjacent day roost sites known to be used by IoM residents, and at focal roost sites spanning the north-east UK coast (ca. 100-500 km from IoM), comprising the relevant autumn-winter range. These core surveys were complemented by ad hoc resightings at other non-focal sites (spanning ca. 800km), including substantial citizen science contributions.
Data processing
Encounter histories
Resightings of 9,359 juvenile shags fledged and ringed in 11 cohorts (fledged 2010–2020) were used to build individual ‘encounter histories’ for the five first-year occasions. A sixth ‘ever after’ occasion, comprising resightings of individuals from age one year onwards, was included to account for surviving individuals that were not observed during their first autumn and winter. From a total of 84,234 post-fledging resightings, multiple resightings within each occasion were condensed to one of four possible events: observed as R (i.e. on or near IoM), observed as M1 (i.e. at a regularly surveyed migratory site), observed as M2 (i.e. at an irregularly surveyed migratory site), or not observed. Here, the fundamental biological distinction is that individuals observed on IoM and adjacent daily foraging sites were categorized as current residents, while individuals observed outside of these locations, and hence that do not roost at night on IoM, were categorized as current migrants at M1 or M2. When an individual was observed in more than one state within an occasion (<1%), its latest observation was used. In the ‘ever after’ occasion 6, individuals were solely classified as observed or not, independent of their location.
(Please see associated manuscript and electronic supplementary materials for details about resighting locations and state attributions)
Pedigree and relatedness
We assembled a social pedigree using breeding data collected on IoM during 1984–2021. The pedigree contained 15,974 ringed chicks from 7,306 breeding events with at least one identified social parent (N=3,749 identified parents, mean identified parents per chick = 1.65), spanning up to six generations and totalling 20,859 individuals. We used this pedigree to compute the A matrix comprising twice the pairwise coefficient of kinship between individuals.
(Please see associated manuscript and electronic supplementary materials for details about pedigree assemblage)
To create the final encounter history and matrix datasets, we included only individuals that were phenotypically and genetically informative, totalling 8,598 individuals (out of 9,359 colour-ringed chicks: 744 excluded due to no pedigree information, 17 excluded due to no known relatedness with other resighted individuals).