Sea lion behavioural diversity and environmental heterogeneity: Colony-level metrics (15 colonies, 2000–2023) with GLMM code
Data files
May 12, 2026 version files 10.47 KB
-
analysis.R
3.54 KB
-
DataSet.csv
2.33 KB
-
README.md
4.60 KB
Abstract
This dataset provides colony-level behavioural and environmental metrics for five sea lion species across 15 colonies worldwide (2000–2023). For each colony, we summarise population-level behavioural diversity (interquartile range of time-at-depth, IQR_TAD; foraging strategy diversity expressed as the percentage of trips following the most common strategy, strategy_perc) and individual-level specialization/flexibility (repeatability of time-at-depth, R_TAD; Shannon diversity of strategies, SDI). Environmental heterogeneity is represented by spatial complexity (bathymetric_roughness) and by primary productivity and its variability (chlorophyl_concentration, chlorophyl_variability). The dataset enables tests of how spatial and temporal heterogeneity relate to behavioural diversity within and among colonies. Intended uses include reproducing the generalized linear mixed models (GLMMs) provided here, sensitivity analyses with/without an identified high-productivity outlier colony (“Lobos”), and comparative or meta-analytic work on habitat heterogeneity and intra-specific behavioural variation. Data are aggregated at the colony level (15 rows; one per colony); no individual identifiers or sensitive location data are included. The accompanying R script (glmmTMB) fits the models used to assess links between behavioural metrics and environmental covariates. These materials support transparent reanalysis, extension (e.g., alternative link functions, priors, or covariate scaling), and integration with other taxa. Ethical approvals and permitting for the underlying biologging come from the original studies; this package contains only derived, aggregated summaries.
Dataset DOI: 10.5061/dryad.qjq2bvqvp
Description of the data and file structure
This dataset contains colony-level summaries (15 rows, one per colony) linking behavioural diversity in sea lions to environmental heterogeneity. Behavioural metrics capture population-level variation in dive shape and foraging strategy, and individual-level specialization/flexibility. Environmental predictors quantify spatial complexity (bathymetric roughness) and productivity & temporal variability (chlorophyll concentration and interannual variability).
Data are aggregated and contain no individual identifiers or sensitive locations. The accompanying R code reproduces the generalized linear mixed models (GLMMs) used to evaluate relationships between behavioural metrics and environmental covariates, including sensitivity analyses with/without a high-productivity outlier colony (“Lobos”).
Intended uses: reproduction of the reported GLMMs; alternative link functions or covariate scaling; sensitivity/robustness checks; meta-analyses on habitat heterogeneity and behavioural diversity.
Ethics: Animal handling approvals/permits pertain to the original biologging studies cited in the associated manuscript. This package includes only derived, aggregated summaries.
Files and variables
File: DataSet.csv
Description:
- Colony-level table (15 rows × 9 columns), comma-separated, dot as decimal.
Variables
- colony: character, colony name (unique row per colony)
- species: character, sea lion species corresponding to the colony (ASL=Australian sea lion, GSL=Galápagos sea lion, CSL=California sea lion, SSL=Southern sea lion, NZSL=New Zealand sea lion)
- IQR_TAD: numeric, interquartile range of time-at-depth across foraging trips within the colony (population-level dive-shape diversity; higher = more diverse shapes)
- strategy_perc: numeric, percentage of trips following the most common foraging strategy (lower = higher strategy diversity)
- R_TAD: numeric, repeatability of time-at-depth across trips within individuals (individual specialization; higher = stronger specialization)
- SDI: numeric, Shannon diversity index of strategies (individual-level flexibility, aggregated at colony)
- chlorophyl_concentration: numeric, mean chlorophyll concentration around colony (productivity proxy) between 1997-2023
- chlorophyl_variability: numeric, interannual SD of chlorophyll concentration (temporal variability), using mean monthly values between 1997-2023
- bathymetric_roughness: numeric, median seafloor roughness around colony window (SD of depth from 15″ grid, aggregated within 100×100 km window)
Notes on transformations used in models (performed inside the R code, not in the CSV):
- log(bathymetric_roughness) used for linearity.
- Some models use log(chlorophyl_concentration) for sensitivity.
- A small offset SDI + 0.05 is added where needed to avoid exact 0/1 in beta models.
Code/software
File: analysis.R
Description:
- R script fitting the GLMMs (as provided below in the repository/upload).
Software requirements
- R ≥ 4.4
- Packages: glmmTMB (for GLMMs)
How to run
- Place DataSet.csv and analysis.R in the same folder.
- Open R (or RStudio), set working directory to that folder.
- Run analysis.R.
- The script fits:
- Colony-level models for IQR_TAD (beta) and strategy_perc (Gaussian).
- Individual-level models for R_TAD and SDI (beta), with and without the “Lobos” outlier.
- Simplified single-predictor models.
- Model summaries print to console.
Model structure (high-level)
- Responses: IQR_TAD, strategy_perc, R_TAD, SDI
- Predictors: chlorophyl_concentration, chlorophyl_variability, log(bathymetric_roughness), and their key interaction(s) as in the script.
- Random effect: (1 | species)
- Families: beta_family() for bounded responses; gaussian() for strategy_perc.
Access information
Other publicly accessible locations of the data:
- None
Data was derived from the following sources:
- Angelakis et al., 2024; Baylis et al., 2015; Chilvers, 2017; Chilvers et al., 2020; Chilvers & Wilkinson, 2009; Fowler et al., 2006; Jeglinski et al., 2015; McHuron et al., 2016; Páez-Rosas et al., 2017; Riet-Sapriza et al., 2013; Schwarz et al., 2021; Villegas-Amtmann et al., 2008; 2011
- See manuscript for complete references
