Sex-specific selection of agricultural farmland by a partially migratory ungulate
Data files
Feb 26, 2025 version files 190.59 MB
-
README.md
3.63 KB
-
summer_within_HR.RData
34.15 MB
-
summer_within_landscape.RData
44.08 MB
-
winter_within_HR.RData
51.40 MB
-
winter_within_landscape.RData
60.96 MB
Abstract
Large herbivores at northern latitudes often forage on agricultural farmland. In these populations, presence of both resident and migrant individuals (termed partial migration) is common, but how migrants and residents differ in their selection of farmland is not well understood. Higher access to farmland may provide benefits to residents compensating for not following the ‘green wave’ of emerging vegetation like migrants. According to sexual segregation theory, males and females differ in body-size related nutritional needs and risk-sensitivity associated with farmland. Yet, how the sexes differ in selection of farmland through an annual cycle remains unclear.
We quantified seasonal variation in the selection of farmland by partially migratory red deer (Cervus elaphus) at broad, landscape scale and at fine, within-home range scale using 16 years of data (2005-2020) from 329 females and 115 males in Norway.
We tested predictions related to the partial migration and sexual segregation theories using resource selection functions. We predicted higher selection for farmland by residents than migrants, and higher selection by females than males due to higher nutritional needs, but that higher perceived predation risk would impact their diurnal selection patterns.
The time spent on farmland was higher in winter (14-18%) than summer (8-14%). Residents selected farmland more than migrants mainly at broad, landscape scale, while differences were smaller and less consistent at fine, within-home range scale. Females showed higher broad-scale selection for farmland in winter, and males higher in summer. At fine, within-home range scale, females selected farmland more in summer during darkness, whereas sex-differences were small otherwise. The fine-scale selection of farmland was markedly higher during low-light conditions than during daylight. High population density was correlated with high broad-scale selection of farmland, i.e., high farmland availability in the home ranges, whereas the effect of density was weak at fine, within-home range scale.
Our study emphasises how hypotheses deriving from the theories of partial migration and sexual segregation can improve our understanding of selection of farmland by ungulates. The higher selection by residents during summer highlights the importance of retaining landscape connectivity allowing for migration, reducing pressure on local resources.
https://doi.org/10.5061/dryad.m905qfvc6
Description of the data and file structure
Description of the data variables:
- use_avail: Marks whether the data is a “used” or “available” location, based on the definition following the level of selection of the research selection function. 1 = used, 0 = available.
- hab2: Habitat variable with the habitats counted as available habitat for the red deer. Four levels; forest, marsh, mountain, and aafarmland (= farmland).
- sex: Sex of the red deer individual. m = male, f = female.
- visual.assignment: Movement category, with the levels “migratory” and “stationary”.
- dens: Red deer population density index. Shot red deer per km2.
- log.dens: The log_e of the variable dens (see above).
- elevation: Altitude of the location, given in metres above sea level.
- log.elevation: The log_w og the variable elevation (see above).
- lightlevel: The light level for the given date and time of the location (specific for the used locations, randon within the season for the available locations). Three levels; light, dark, twilight.
- prop.pasture: Proportion available farmland habitat within the individual home range.
- asin.pasture: The prop.pasture variable (see above) arcsine square root transformed.
- region: Region of Norway where the individual is located. Levels: SoF (Sogn & Fjordane), Tr (Trondelag), MoR (More & Romsdal), and SW (Hordaland).
- yr: Year
- id: Individual ID
Data files
File: summer_within_HR.RData
Description: Data used to fit the research selection function (RSF) for the summer season on the within-home range level (third order of selection).
File: summer_within_landscape.RData
Description: Data used to fit the RSF for the summer season on the landscape level (broad-scale, second order of selection).
File: winter_within_HR.RData
Description: Data used to fit the RSF for the winter season on the within-home range level (third order of selection).
File: winter_within_landscape.RData
Description: Data used to fit the RSF for the winter season on the landscape level (broad-scale, second order of selection).
Code/software
R version 4.3.0 was used for the computation.
Separate scripts are available per season and per selection level:
Script: Mod_fit_and_sel_summer_within_HR.R
Description: Script for model fitting and selection for the analysis of the summer season on the within-home range level (fine-scale, third order of selection, data file called summer_within_HR.RData).
Script: Mod_fit_and_sel_summer_within_landscape.R
Description: Script for model fitting and selection for the analysis of the summer season on the landscape level (broad-scale, second order of selection, data file called summer_within_landscape.RData).
Script: Mod_fit_and_sel_winter_within_HR.R
Description: Script for model fitting and selection for the analysis of the winter season on the within-home range level (fine-scale, third order of selection, data file called winter_within_HR.RData).
Script: Mod_fit_and_sel_winter_within_landscape.R
Description: Script for model fitting and selection for the analysis of the winter season on the landscape level (broad-scale, second order of selection, data file called winter_within_landscape.RData).
Versions of the packages used in the scripts:
- lme4: version 1.1-33
- MUMIn: version 1.47.5
- ggplot2: version 3.5.1
The methods comprise fitting research selection functions using use-availability data for red deer (Cervus elaphus) to analyse their selection of farmland on broad landscape scale (second order selection; Johnson, 1980) and on finer, within-home range scale. The data of used locations is collected using GPS collars (Followit, Sweden, and Vectronic, Germany) on male and female adult red deer. The available locations were randomly sampled within each individual's available range (for the second order selection analysis) and seasonal home range (for the third order selection analysis), respectively. The RSF's were fitted using generalised linear mixed-effects models (GLMMs) for the second order selection analysis, and generalised linear models (GLMs) for the third order selection analysis.