Skip to main content

Data from: Forecasting animal distribution through individual habitat selection: Insights for population inference and transferable predictions

Cite this dataset

Winter, Veronica et al. (2024). Data from: Forecasting animal distribution through individual habitat selection: Insights for population inference and transferable predictions [Dataset]. Dryad.


Habitat selection models frequently use data collected from a small geographic area over a short window of time to extrapolate patterns of relative abundance to unobserved areas or periods of time. However, these types of models often poorly predict how animals will use habitat beyond the place and time of data collection because space-use behaviors vary between individuals and are context-dependent. Here, we present a modelling workflow to advance predictive distribution performance by explicitly accounting for individual variability in habitat selection behavior and dependence on environmental context. Using global positioning system (GPS) data collected from 238 individual pronghorn, (Antilocapra americana), across 3 years in Utah, we combine individual-year-season-specific exponential habitat-selection models with weighted mixed-effects regressions to both draw inference about the drivers of habitat selection and predict space-use in areas/times where/when pronghorn were not monitored. We found a tremendous amount of variation in both the magnitude and direction of habitat selection behavior across seasons, but also across individuals, geographic regions, and years. We were able to attribute portions of this variation to season, movement strategy, sex, and regional variability in resources, conditions, and risks. We were also able to partition residual variation into inter- and intra-individual components. We then used the results to predict population-level, spatially and temporally dynamic, habitat-selection coefficients across Utah, resulting in a temporally dynamic map of pronghorn distribution at a 30x30m resolution but an extent of 220,000km2. We believe our transferable workflow can provide managers and researchers alike a way to turn limitations of traditional habitat selection models - variability in habitat selection - into a tool to understand and predict species-habitat associations across space and time.

README: Forecasting Animal Distribution through Individual Habitat Selection: Insights for Population Inference and Transferable Predictions

A note on data setup: due to this being a sensitive species, we are unable to share the XY positions. To Implement this analysis, GPS data should be formatted with unique individual IDs, XY, and date-time stamp, as shown here:

         ID               x               y             datetime

 PR17F0001 44xxxx. 42xxxxx. 2017-12-18 20:15:00

 PR17F0001 44xxxx. 42xxxxx. 2017-12-18 22:30:00

 PR17F0001 44xxxx. 42xxxxx. 2017-12-19 01:00:00

 PR17F0001 44xxxx. 42xxxxx. 2017-12-19 03:15:00

 PR17F0001 44xxxx. 42xxxxx. 2017-12-19 05:30:00

This table can include any other information that may pertain to the question at hand for your system/species.

For reproducibility, we have uploaded data files for implementing the analysis post-RSF. 

File structure:

Data preparation and analysis

'xx_' files are functions

'01_' file is to collect data, prepare, and run RSFs

'02_' is for mixed effect model (MEM) preparation and analysis (full, season, and null models)

'03_' file is for plotting MEM outputs

Spatial predictions and Mapping

'04_', '_05', & '_06' are predicting and mapping out of sample data and unconditional mapping


'07_' & '08_' are validating the functional responses, spatial predictions, and unconditional mapping on the full, season, and null models

Example data file information:

For each individual (ID) we have information of

  • ID
  • age class (age_class)
  • WMU (unit)
    • These units are also indicated by ex. 'is.AntelopeIsland' with binary classification
    • Note: Even if an individual dispersed, we did not let it switch WMUs in the data for consistency and predictive purposes
  • Climactic region (clim_reg)

    • based off of where in the state the WMU was located and was generated based off of a climate region map produced for the state
    • These are also present as ex. is.NorthCentral
  • is.mountain was another classification of mountain landscape or not, a more broad classification

  • migration tendency (mig_tend) per spring and fall migration.

    • Tendency is another column of the same information
    • There are corresponding columns of 'is.res' and 'is.mig' with bianary indication
    • 'is.unk.mig' means that either that indiviaul was not being tracked in that season/year or that we did not have enough information to classify 
  • mean availabilities that individual experienced in that season/year ex m_elev

    • These variables also exist as ex: m_SC_elev indicvating that the variable was scaled and centered
  • month and month long are the 'season'

    • Winter = Feb
    • Spring = April
    • Summer = July
    • Fall = November
  • Classifications were made based off of average timing across individuals

  •  beta and stder (standard error) are from the individual RSFs

  • weights are inverse variance weighting from individual RSFs

    • ex: # Tree ---
    • temp <- 1 / (dat$Tree_stder ^ 2) * dat$weight_Tree <- temp / sum(temp, na.rm = TRUE)
  • predictions ex. Elev.mod.full_pred are from the MEM in Winter et al. 2024.

    • Mixed effects models were fit with 2018-2020 data and predictions were made on 2021 data set.
  • PDSI is an acronym for 'Palmer Drought Severity Index' 

Sharing/Access information

Data was collected through partnerships with the Utah Wildlife Migration Initiative:


Code for this manuscript is also available at


Utah Agricultural Experiment Station, Award: UTA01562, Utah Agricultural Experiment Station