Let’s apply what we’ve learned about iSSFs. Refer to the lecture slides (05a_lecture.html) and the walkthrough code (05b_walkthrough.R) as a refresher.

We encourage you to use your own data for this exercise. We will be available while you are working to answer any questions.

If you don’t have your own data, you can use some sample cougar data from UT. These data were published by Mahoney et al. (2017) here. These data were originally used in this manuscript:

Mahoney PJ, Young JK (2016) Uncovering behavioural states from animal activity and site fidelity patterns. Methods in Ecology and Evolution 8(2): 174–183. doi:10.1111/2041-210X.12658

We’ve already downloaded and processed the data, which you can find in data/coyote_cougar.csv. We’ve also downloaded and processed some environmental raster data for you to use, which you can find as a multi-band GeoTiff in data/coyote_cougar_habitat.tif.

Habitat layers are:

We will go over how to model multiple animals in a later module, so for now, subset your data to just one animal.

# Load `tidyverse` and `terra`
library(tidyverse)
library(terra)

# Location data
dat <- read_csv("data/coyote_cougar.csv") %>% 
  # Subset to just cougar F53
  filter(id == "F53")

# Set the timezone
tz(dat$t_) <- "US/Mountain"

# Habitat data as stack
hab <- rast("data/coyote_cougar_habitat.tif")
names(hab) <- c("elevation", "trees", "biomass", "dist_to_road")

Instructions

  1. Load your data or the example data. Subset to a single individual.

    • Recall that iSSFs assume a constant step duration. Make sure your data are (within a reasonable tolerance) regularly sampled before you continue.
    • You will want a fairly large sample size as your model grows in parameters by including interaction terms.

  2. Before you get started, consider your habitat variables and your movement variables.

    • Decide a priori whether each variable represents a resource, risk, or condition.
    • Should any of your habitat covariates interact with each other? Why?
    • Which of your covariates might affect the movement process?
    • Should any of your covariates interact with a temporal variable? Might selection change between seasons? Might selection change between times of day?
    • Include your thoughts about each variable as comments in your analysis script.

  3. Fit an iSSF, with at least one interaction.

    • Make sure the form of each variable matches your a priori hypotheses about it from above. Pay careful attention to understand what your interaction terms mean.
    • You might also be interested in fitting more than one model. E.g., maybe you want to compare different seasons or different years.
    • It’s up to you whether to decide if you’d rather capture effects with interactions or separate models.

  4. Make at least one plot showing the RSS (or log-RSS) for one of your habitat variables.

    • What does this plot tell you about the biology of your system?

  5. Make at least one plot showing the step-length or turn-angle distribution for your animal.

    • If you have interactions with step-length or turn-angle, try to show this.