Phenotypic flexibility in the city: A meta-analysis on variation
Data files
May 30, 2026 version files 43.09 KB
-
Meta-analysis-urbanisation.Rproj
218 B
-
R-Script.zip
38.51 KB
-
README.md
4.37 KB
Abstract
Among global changes, urbanisation entangles a variety of human-induced rapid environmental changes, such as habitat fragmentation, temperature change, introduction of human food sources, and pollution. Urban environments, in contrast to non-urban ones, are often assumed to be more heterogeneous and variable in space and time. A key feature of animals coping with high environmental variability ought to be phenotypic flexibility, i.e., the capacity of individuals to express reversible variation in labile traits. However, this “phenotypic flexibility hypothesis” has not been tested rigorously. We compiled available raw data and used a meta-analysis to estimate overall differences in among- and within-individual variation between urban and non-urban population pairs of wild animals. We considered within-individual variation as a proxy of phenotypic flexibility. Across all taxa, among-individual variation did not differ between urban and non-urban populations. Within-individual variation was marginally lower in urban populations compared to non-urban ones. The potential decrease of phenotypic flexibility in urban individuals could result from the multidimensionality and complexity of urban environmental conditions. Overall, the effects of urbanisation on phenotypic variation are not generalisable and depend on the taxa, species, and traits. Future studies should increase efforts to directly link temporal and spatial environmental variability with phenotypic individual variation.
https://doi.org/10.5061/dryad.2rbnzs82d
Preprint
Petit J. & Dammhahn M. (2025). Phenotypic flexibility in the city: A meta-analysis on variation (https://doi.org/10.32942/X2ZW6J)
Data description
Due to licensing restrictions, the folders 'Data', 'Image' and 'Plot' are hosted on Zenodo. Please find these files via the Related Works link for Supplemental Information. The remaining files are hosted on Dryad.
Data.zip/0-Literature review
The folder '0a-Literature processed in ryan' contains the RIS file with all the literature extracted from the final search across all literature database. Those RIS files are stored in the folder 'RIS file imported in Ryan'. The folder '0a-Literature processed in ryan' also contains the compilation of all the RIS file together containing the duplicates ('article (contains duplicates).ris') and all the RIS file with the duplicate removed ('articles(duplicates removed) used for 1st screening.ris').
'article (contains duplicates).ris' was used as a raw input in the R-script '0-Deduplication Rscript.R' and 'articles(duplicates removed) used for 1st screening.ris' was the output of this script. More details about the files written in the R-script.
The folder '0b-Literature manually processed' contains literature (1 csv and 1 word document) that could not be compiled in Ryan due to a bug. Therefore, they were manually processed.
The file 'Summary screening process from 2nd screening.xlsx' contains the explicit reasons of why papers were discarded from the 2nd screening until the final screening providing the final dataset. We recommend reading Figure S1 from the article in supplementary material to see the summary of the screening process.
Data.zip/1-Extracted estimates
Contains all the variance estimates extracted per paper for each trait and for each linear mixed effects model type (FixedFactor is AD-LMM and InterceptOnly is IO-LMM). Excel files are named with the name of the first author and the year of publication if published. More details with specific README files in each author's folder.
Data.zip/2-Tables to build Effectsize IOM_FEM
- Contains the effect sizes of the variance coming from 'intercept-only' mixed effects models (IOM or IO-LMM) and of 'adjusted' mixed effects models (FEM or AD-LMM). In those files, ID_pair is the unique ID of each pair which is consistent with the ID_pair present in the files 'Effect size_FEM' and 'Effectsize_IOM' in Data/3-Tables main analysis.
- d_lnCVR_FEM_BET_N contains the lnCVR estimates and sampling variance (.sv) based on among-individual variance coming from 'adjusted' mixed effects models
- d_lnCVR_FEM_WIT_N contains the lnCVR estimates and sampling variance (.sv) based on within-individual variance coming from 'adjusted' mixed effects models
- d_lnCVR_IOM_BET_N contains the lnCVR estimates and sampling variance (.sv) based on among-individual variance coming from 'intercept-only' mixed effects models
- d_lnCVR_IOM_WIT_N contains the lnCVR estimates and sampling variance (.sv) based on within-individual variance coming from 'intercept-only' mixed effects models
Data.zip/3-Tables main analysis
Contains all the data used in the R scripts. An additional README file is provided to explain the names of the columns.
Image.zip
Contains Figure 1B (animals picture mosaic) and Figure S1 (PRISMA chart with criteria table).
Plot.zip
Contains Figure 1_ABC, Figure 2, Figure 3, Figure S2, Figure S3, Figure S4 and Figure S5. For titles and legends see main manuscript or supplementary materials.
R-Script.zip
Contains the R-script to deduplicate the literature (0-Deduplication Rscript.R), the R-script to extract all the variance estimates for each paper (1-Variance final script extraction.qmd), the R-script to calculate the effect size and make the phylogenetic tree (2-Effect size calculation_Phylotree_Figure 1 assemblage.qmd) as well as the script to perfect the meta-analytic models (3-Meta-analytic models script.qmd).
Meta-analysis-urbanisation.Rproj
This is the R project set up to facilitate the determination of the working environment once you saved all folders (from Dryad and Zenodo) in a specific location on a computer.
We performed the literature search in five databases (Web of Sciences collection, Scopus, ProQuest, EBSCOhost Open Dissertations, and OpenGrey) on the 10th of March 2023. In total, we found 4,322 papers.We performed three screening phases resulting in 113 papers. Raw data was extracted if made open access or authors were contacted to obtain raw data or asked to run a specific R-script. After the final contacting phase and a deep data check, we collected data from 31 studies.
The two main objectives of the meta-analysis were to assess whether within- and between-individual variation change under urbanisation. For all traits, we partitioned the variance using linear mixed effects models (LMMs) to estimate both variance components. The variance partitioning was based on the following steps. First, we ran ‘intercept-only’ LMMs (IO-LMMs) with individual identity as random intercept. Second, we ran ‘adjusted’ LMMs (AD-LMMs) with trial number and sex (if available) as fixed factors as well as other fixed factors when the authors mentioned that they improved the model fit significantly. For all variance estimates that were validated, we calculated the log coefficient of variation ratio (lnCVR) to investigate differences in the variability between urban and non-urban populations.
The data deposed in this repository contains the variance estimation from LMMs and the lnCVR contrasting urban and non-urban populations.