TWS Annual Conference Travel Grant Recipient: Jennifer Feltner
Thanks to a travel grant from WY-TWS, I was able to spend almost a week in Reno, Nevada for The Wildlife Society 2019 Annual Conference and also attend the TWS workshop “Advances in Habitat-Selection Modeling.” As a PhD Candidate at the University of Montana (UM), my research focuses on the impacts of recovering wolf and grizzly bear populations on subordinate carnivores (cougars) and their shared prey (elk) in the southern Greater Yellowstone Ecosystem (SGYE). One of my chapters will examine key factors driving the habitat selection and movements of cougars in the SGYE from 2001-2016 including prey availability, risk of dominant competitor encounter, human activities, and other environmental factors. While I was fortunate to take Dr. Mark Hebblewhite’s intense course on wildlife habitat modeling here at UM in 2017, I was excited for the opportunity to learn more about some of the most recent advances in habitat-selection modeling at the workshop prior to launching my dissertation research analysis this fall. And, I apparently wasn’t the only one! I only got off the waiting list for the workshop after repeated e-mails to conference organizers, who allowed five of us to slip in at the last minute.
The all day workshop was led by Dr. John Fieberg of the University of Minnesota, Dr. Tal Avgar of Utah State University and R-coding master Dr. Johannes Signer of the University of Göttingen, as well as Dr. Stefanie Muff of the Norwegian University of Science and Technology who was unable to attend. These four scientists have co-authored a number of recent publications on methodical advances in habitat selection modeling such as integrated step selection functions (Avgar et al. 2016) and more efficient procedures for accounting for individual-specific variation in habitat and movement models (Muff et al. 2019). They have also developed or introduced R-packages such as amt (Signer et al. 2018), glmmTBM and INLA (Muff et al. 2019) that can allow for much smoother fitting of resource selection function (RSF) and step-selection function (SSF) models of animal location data.
The Sunday workshop began with an introductory presentation on habitat selection and RSF models, and was then followed by presentations on three recent developments covered in their above-mentioned publications: 1) Integrated Step-Selection Function (iSSFs), a simple framework for simultaneous modeling of animal movement and habitat selection processes using conditional logistic regression. These models allow one to relax the assumption that movement characteristics (i.e., step lengths and turn angles) are independent of habitat features; 2) the amt (animal movement tools) package in R, which provides tools for exploratory analysis of animal location data, functions for data development prior to fitting RSFs or SSFs, and a simple tidyverse workflow for seamless fitting of RSF and SSF models to data from individual animals; 3) methods for efficient estimation of mixed-effects RSFs and SSFs using frequentist (glmmTMB) or Bayesian (INLA) methods.
Following these presentations, Dr. Signer lead workshop participants through two coding sessions: 1) an introduction to the amt package; and 2) fitting mixed-effects RSF and SSF models using animal location and covariate data provided by the workshop organizers. Amongst other things, I was amazed at some of the data summaries the amt package can quickly provide such as a summary of each animal’s sampling rates (fix rates). While I will likely focus on RSFs for my dissertation analysis, I was also excited to find a much more efficient way of fitting mixed-effects SSF models, something which had previously been a much less straight forward process.
After our coding sessions, Drs. Fieberg and Avgar gave two more presentations on more advanced topics: 1) interpreting and predicting exponential habitat selection functions and iSSFs; and 2) using calibration plots to validate RSF and SSF models. With my brain overflowing with new information, the workshop concluded right at 5pm, allowing participants to recover for a couple of hours prior to the official kickoff social of the conference at 7pm that evening.
I learned so much from this workshop and will definitely be applying some of the new methods and R packages in my own research. The workshop organizers have generously shared the workshop code, data and presentations on GitHub: https://github.com/jmsigner/workshop_2019_tws.
They plan to create a more comprehensive website detailing these methods soon, so stay tuned. In the meantime, if you are interested in gaining some new skills in habitat modeling or finding more efficient ways to fit RSF and SSF models, I highly recommend you check out the workshop materials and read the three papers discussed here. Additional super helpful code and data from Muff et al. 2019 (including a Bayesian framework for fitting mixed-effects RSFs and SSFs) can also be found at https://conservancy.umn.edu/handle/11299/204737.
Literature cited
Avgar, T., Potts, J. R., Lewis, M. A., & Boyce, M. S. (2016). Integrated step selection analysis: Bridging the gap between resource selection and animal movement. Methods in Ecology and Evolution, 7:619–630.
Muff, S., Signer, J., & Fieberg, J. (2019). Accounting for individual‐specific variation in habitat‐selection studies: Efficient estimation of mixed‐effects models using Bayesian or frequentist computation. Journal of Animal Ecology, 00: 1–13.
Signer, J., Fieberg, J., & Avgar, T. (2019). Animal movement tools (amt): R package for managing tracking data and conducting habitat selection analyses. Ecology and Evolution, 9: 880–890.