Data from: Daily activity is repeatable but varies across the breeding season in female great tits
Data files
Jan 03, 2025 version files 98.96 KB
-
chickmass_data_for_dryad.csv
4.67 KB
-
Female_activity_for_dryad.csv
91.32 KB
-
README.md
2.97 KB
Abstract
Wild animals typically organize activity around a 24-hour day and daily timing across the year is optimized for both survival and reproductive success. Among-individual variation in chronotype, when individuals begin or end their active day relative to a cue such as photoperiod, often exists within a population. Both intrinsic and extrinsic factors contribute to this variation and activity patterns may change across and within different life history stages as energetic investment changes. Here we describe population level changes in female great tit (Parus major) activity patterns of onset and offset of activity as well as assess variation and repeatability in daily activity both within- and across-breeding stages. We fitted individuals with accelerometers to track activity prior to nest building through chick rearing. Prior to clutch initiation females began their active day before sunrise, however in the days prior to laying their first egg, activity was delayed until after sunrise. Females ended activity prior to sunset across the monitoring period and earliest during egg laying and incubation. Additionally, females exhibited greater among- and within-individual variance in chronotype during parental care. Individual female daily activity was moderately repeatable within breeding stages and strongly covaried across several breeding stages. These findings expand our understanding of individual variation in chronotype during reproduction and the potential fitness implications of chronotype in wild animals.
README: Data from: Daily activity is repeatable but varies across the breeding season in female great tits
https://doi.org/10.5061/dryad.7h44j104j
Description of the data and file structure
Included within this dataset are the daily activity measures and reproductive data used within the publication: “Daily activity is repeatable but varies across the breeding season in female great tits” accepted for publication in Behavioral Ecology.
In this study we fitted individual females with accelerometer data loggers to track activity prior to nest building through chick rearing. We described population level changes in free-living female great tit (Parus major) activity patterns of onset and offset of activity as well as assess variation and repeatability in daily activity both within- and across-breeding stages. We also ran a model to investigate co-variation between onset of female activity and nestling mass.
Files and variables
File: chickmass_data_for_dryad.csv
Description: This data file includes mass of individual chicks collected at 15 days of age. The data is organized by female parent. This data was combined with onset of activity data for the analysis as described in the methods.
Variables
- Year: Year of data collection
- FemaleID: Unique ring number for each individual female parent
- Nestbox: ID of the nestbox in which the female bird made their reproductive attempt
- ChickID: Unique ring number for each individual chick
- D15 Chick Mass: Mass of individual chick collected on day 15 post hatch. Units are grams.
File: Female_activity_for_dryad.csv
Description: This csv file contains all the daily activity data used in the analysis for evaluating variation and repeatability of daily onset, offset, and total length of active day. It is organized by individual bird and the day of behavioral data collection. Any blanks indicate behavioral data was not collected on that day.
Variables
- Year: Year of data collection
- Bird ID: Unique ring number for each individual female
- Nestbox: ID of the nestbox in which the female bird made their reproductive attempt
- Date: The date of the behavioral data collection (i.e. onset would be the initiation of activity on this morning, offset is the ending of activity in the evening of this date)
- Stage: Stage of reproduction. Details on how stage was assigned is included in the methods.
- DiffOnset: Onset of individual female activity relative to sunrise on this particular date. Units are minutes.
- DiffOffset: Offset of individual female activity relative to sunset on this particular date. Units are minutes.
- DurationActivity: Total active day as calculated by subtracting time of onset of activity from time of offset. Units are hours.
- DayLengthActual: Total length of daylight (sunrise to sunset) on this date. Units are hours.
Code/software
All analyses were conducted in R.
Methods
To observe activity patterns across the season, we focused on five different breeding stages: pre-breeding, pre-laying, egg laying, incubation, and chick rearing. The ‘pre-breeding’ stage included a period of 10 days beginning on day 23 and ending on day 14 prior to the first egg. The ‘pre-laying’ stage was the one-week period (day 8 to day 2) prior to the initiation of egg laying for each individual. It is during this period that the majority of ovarian development and commitment to egg laying should occur (Williams, 2012). The data was then assigned for the period of ‘egg laying’ (including any days in which egg laying was ‘skipped’), ‘incubation’ which was determined by visual confirmation either by finding the female sitting on the eggs or by the eggs being cleared of any nest material and warm to the touch, and finally the ‘chick-rearing’ period which began on hatch day and continued until the BitTag data loggers were collected (See Supplemental Table 2 for sample sizes across stages).
Statistical Analysis
To assess differences in mean and variances of timing of daily activity across the five different breeding stages, we used univariate mixed-effects models for onset, offset, and total duration of daily activity (Dingemanse & Dochtermann, 2013). We included stage and year as fixed effects and individual bird ID as a random effect. All statistical models and analyses were specified in the Stan computational framework (http://mc-stan.org/) using the brms R package for Bayesian mixed models (Bürkner, 2017, 2018; Royauté & Dochtermann, 2021). All models were run with default student-t priors and a gaussian distribution for all behavioral measures. 4 chains, each with an iteration of 6000, were fitted for each model. Models were defined as having converged when Rhat values were equal to or less than 1. All effective samples size (ESS) values for models run were greater than 1000 indicating stable estimates (Bürkner, 2017). We also examined density plots of observed and predicted values for evidence of lack of convergence. We first compared the fit of four models structured as follows :
Model 1: Vi = & Vw = A null model where the among-(Vi) and within-individual (Vw) variances is fixed to be constant across breeding stages.
Model 2: Vi ≠ & Vw = A model where the among-individual variance differs across breeding stages, but the within-individual variance is fixed to be constant.
Model 3: Vi = & Vw ≠ The within-individual variance is allowed to differ across breeding stages, but the among-individual variance is fixed to be constant.
Model 4: Vi ≠ & Vw ≠ Both the among-and within-individual variances are allowed to vary across stages.
To determine whether breeding stage affected mean activity, we first ran two models (Models 1 and 4 above) with and without stage as a fixed effect and compared the models using leave-one-out information criterion (LOO; Vehtari et al., 2017). If the average of the behavioral measures differed across stages, the models with stages included would be expected to have lower LOO values. Based on this result, we then ran and compared all models with stage included to determine whether among- and/or within-individual variances differed (Royauté & Dochtermann, 2021). The model with the lowest LOO value was considered the best model and that model was then used to evaluate average behavioral differences across stages using posterior modal estimates for each stage. Variances components were also extracted to compare within- (Vw) and among- (Vi) individual variance for each stage, as well as to calculate repeatability (R = Vi / (Vi + Vw)) for each breeding stage (Nakagawa & Schielzeth, 2010). To assess whether daily timing of activity is integrated across stages, we re-ran Model 4 as above but allowed for covariance across stages. The analyses described above were repeated for onset and offset of daily activity and total length of active day.
We conducted an additional mixed-effects model for both onset and offset of daily activity to determine if onset or offset of daily activity during the different reproductive stages covaried with day 15 nestling mass. Nestling mass is a fitness related trait (Tinbergen & Boerlijst, 1990) and this model allowed us to investigate whether there was a relationship between the size of a nestling and female daily activity during any of the five breeding stages. These models were run as described above for activity but now also included nestling mass measures at day 15 for each individual female as a separate grouping. All analyses were conducted in R 4.3.2 (R Core Team, 2023).