Skip to main content
Dryad

Accounting for viewshed area and animal availability when estimating density and recruitment of unmarked white-tailed deer

Cite this dataset

Koeck, Molly (2024). Accounting for viewshed area and animal availability when estimating density and recruitment of unmarked white-tailed deer [Dataset]. Dryad. https://doi.org/10.5061/dryad.76hdr7t4g

Abstract

Quantifying demography of wildlife is vital to population monitoring; however, studies using physical capture methods can prove challenging. Camera traps have gained popularity as a density estimator tool in recent decades due to noninvasive data collection, reduced labor, cost efficiency, and large-scale monitoring capabilities. Many wildlife populations are comprised of individuals with no unique natural markers for individual identification, resulting in the need for unmarked abundance models. The recently developed Space-to-Event (STE) model offers a method for density estimation of unmarked populations using timelapse photography. STE relates detections of animals to camera sampling area (i.e., viewshed), resulting in density estimates that can be extrapolated to abundance over large areas. Consequently, this makes STE sensitive to estimates of viewshed area as small changes in viewshed could significantly affect density estimation. Using STE, we estimated density and recruitment of white-tailed deer (Odocoileus virginianus) in a densely forested landscape using measurements of viewshed per camera. We compared estimates of abundance derived from uniquely measured viewshed to estimates of abundance derived from an assumed viewshed area held constant across all cameras. When using a constant viewshed across all cameras, our point estimates of abundance shifted away from uniquely measured viewshed estimates in predictable ways, depending upon how much area was sampled. Additionally, we demonstrated the need for further exploration of animal availability at fine temporal scales by comparing estimates of density derived from sampling the full diel period to estimates derived from periods of peak activity (i.e., crepuscular periods). Finally, we extended the usefulness of the STE model by using densities of fawns and adult females to derive estimates of fawn recruitment.

README: Accounting for viewshed area and animal availability when estimating density and recruitment of unmarked white-tailed deer

Our dataset is comprised of timelapse and motion-detection photo data collected from two study sites across two sampling seasons (June-December of 2021 and 2022). Our dataset is divided into four R Data Serialization (RDS) files. Files are named by site (JC for James Collins Wildlife Management Area or SB for Sans Bois Wildlife Management Area) and year collected.

Description of the data and file structure

We designed our study to comply with the assumptions of the Space-to-Event (STE) unmarked abundance model. Each row corresponds to a single image. Columns necessary to run STE include "cam", "count", "area", and "datetime". The column corresponding to the taxa of interest should be renamed to "count". File structure is identical for the four RDS files.

  • datetime: Date and time of when the image was taken.
  • comments: Location for any necessary comments regarding an image.

  • opstate: Operating state of the camera (e.g., malfunctioning, burned, tilted...etc.).

  • cam: Camera ID.

  • area: Viewshed area per camera in square meters.

  • otherwhat: Categorical data consisting of a species common name. Corresponds to count data in the same row under the "other" column.

  • humanwhat: Categorical data consisting of types of humans (e.g., researcher, hunter, hiker...etc.). Corresponds to count data in the same row under the "human" column.

The following columns are comprised of count data (0-n) representing the number of individuals of a species detected per image.

  • deer, human, turkey, armadillo, raccoon, coyote, feralhog, squirrel, opossum
  • other: Count data representing the number of individuals of a later specified species detected per image. Specified by "otherwhat" column.

The following columns are comprised of count data (0-n) representing the number of individuals of a sex or stage class detected per image.

  • WTDantlerless: Adult female white-tailed deer.
  • WTDantlered: Adult male white-tailed deer.
  • WTDfawn: Juvenile white-tailed deer.
  • WTDunkn: Deer detections where sex or stage was not able to be determined.
  • Turkeyfemale, turkeymale, turkeypoult, turkeyunkn

Code/Software

We used the package 'spaceNtime' to estimate abundance of white-tailed deer using the STE model.

We used the package 'SunCalc' to gather data on daily sunrise and sunset times within our sampling frame to constrain sampling efforts to crepuscular periods.

Methods

We collected photo data over two years (2021 and 2022) from two study sites (the James Collins Wildlife Management Area and the Sans Bois Wildlife Management Area) in southeast Oklahoma. We deployed 100 cameras (50 per site) in June and retrieved them in December of each sampling year. We outfitted each camera with a 32-gigabyte SD card programmed to take timelapse images at 10-minute intervals as well as motion-triggered images in bursts of three with no time delay between triggers. Once deployed, cameras synchronously took timelapse images to create instantaneous sampling occasions at each 10-minute timestep (i.e., 09:00, 09:10, 09:20, etc.).

We used random sampling in the form of generalized random tessellation stratified sampling (GRTS) to generate 50 camera deployments sites per study site.

We calculated viewshed area per camera using the camera lens angle and measurements of maximum distance of detection. We used a viewshed board to divide the camera lens into 6 sectors and took a maximum distance measurement per sector. Area was calculated per sector and the summation of the 6 sectors resulted in a unique sampling area per camera.