Mammal activity patterns with big data

Observation of a red fox, Vulpes vulpes, captured by a camera trap in a forested habitat in Pickens County, SC.   

Above image and activity below is based on

• Prince, Emily K, Robert T Kelley, and Elizabeth  A. McDonald. “Investigating Mammal Activity Patterns Using Camera Trap Observations.” Teaching Issues and Experiments in Ecology 19 (2023).  CC-BY-NC 4.0.  
  
  

Overview

Many different factors may impact activity patterns and habitat use in organisms.   Students will use camera trap data to design and carry out an ecological experiment focused on related questions.  

Connections to other labs

This lab builds on methods used in earlier labs, including pivot tables ( Adaptations to Introduce Ants , Brood Parasites ), diversity analysis (Spider Diversity , Diversity with Pokémon ), statistical analyis ( Comparing Differences among Groups ), and graphing.

Objectives

Students should be able to

• understand the value of monitoring and big data

• develop hypotheses using pre-recorded data

• test their hypotheses by aggregating and arranging data as needed

• Produce visualizations that match their hypothesis

• Explain their results

Introduction

Although scientists often collect data in the field or lab, they can also make use of existing datasets. NEON, the NSF's LTER program, and community science projects are just examples of large-scale datasets that provide monitoring or occurence data that scientists can use.  These datasets may be  larger than any single scientist can collect and offer unique insight on ecological questions.  Today we will use an existing dataset from camera traps to ask questions about patterns in mammal activity or diversity.  This lab will allow you to design your own set of questions using the scientific method.

Mammals can be extremely difficult to observe directly in the wild because they may inhabit dangerous or inaccessible terrain, be active primarily at night, or actively avoid researchers. Camera traps, remote motion-activated cameras with infrared sensors, can help researchers document mammal activity in a relatively inexpensive and non-intrusive way (Burton et al. 2015). Camera traps are used extensively in ecological research and have been used to answer questions about the distribution, abundance, and behavior of populations as well as to address questions about biodiversity and community structure of mammals (Trolliet et al. 2014).

   A variety of factors influence patterns of mammal abundance and activity. Habitat choice is species-specific. Some species are more common in forested areas, while others select open areas. For example, in a study of mesocarnivores in Illinois, Lesmeister et al. (2015) found that bobcats avoided agricultural areas, coyotes were common in forested areas, and red and gray foxes preferred spatially complex habitats. Within a species, habitat choice depends upon a variety of factors, including foraging or hunting access as well the ability to avoid predators (e.g., Fattebert et al. 2019).

   Human disturbance also influences the spatial distribution of mammals. As a consequence of habitat fragmentation and direct removal, large predators are usually less common in areas with high human footprints (Lesmeister et al. 2015; Nickel et al. 2020; Ritchie & Johnson 2009). Other mammals, like raccoons, are less affected by urban environments (e.g., Mims et al. 2022). In fact, raccoons may benefit from the removal of predators and from using human food waste as a resource (Bozek et al. 2007). The presence of both red foxes and striped skunks has been reported to be positively associated with human disturbance in landscapes as well (Lesmeister et al. 2015).

   Different species of mammals have different diel activity patterns, meaning that they are more active at some times of day than others. Some, like raccoons, are nocturnal, meaning they are mostly active at night. Others, like squirrels, are diurnal, meaning they are primarily active during the day (Ikeda et al. 2016). Bobcats are crepuscular, or active during dawn and dusk (Lesmeister et al. 2015). Red foxes have no clear diel pattern; instead, they tend to be cathemeral, or sporadically active during the day and night (Ikeda et al. 2016). However, even in mammals with clear diel patterns, temporal activity may shift depending on specific conditions. For example, red deer are active during the day in some habitats but during the night in others, presumably in an effort to avoid different predators (Fattebert et al. 2019).

   Human disturbance also alters diel activity patterns. Gaynor et al. (2018) reported a worldwide shift of wildlife towards nocturnality in response to human disturbance. Specifically, red foxes (Lovell et al. 2022) and nine-banded armadillos (DeGregorio et al. 2021) are more nocturnal in urban than rural environments. Even when human infrastructure does not alter the landscape, the temporary presence of humans in the environment has been linked to increased nocturnal activity in predators such as bobcats, coyotes, and pumas (Nickle et al. 2020). This effect is not universal because some mammals, such as the European badger, may not exhibit plasticity in diel activity (Lovell et al. 2022).

Methods

Beginning in late spring 2018, the members of Lander University’s Mammal Ecology Lab set up camera traps in Upstate South Carolina. Each camera is motion activated and records the date and time when each photograph is taken. We began with five camera stations, but we established additional stations over the next few years. Occasionally, we stop monitoring at particular camera stations because cameras that are frequently vandalized or stolen must be relocated. Currently, we have data from 26 cameras in Greenwood, Laurens, and Pickens Counties in South Carolina. The cameras are in six different sites: Grace Street Park in Greenwood County, on the campus of the Greenwood Genetics Center in Greenwood County, at Fellowship Camp and Conference Center in Laurens County, at Lake Greenwood State Park in Greenwood County, in undeveloped land near Table Rock State Park in Pickens County, and in the Jocassee Gorges Wilderness Area in Pickens County (Figure 1).

   Figure 2 shows examples of photographs captured by our camera traps. The locations of the cameras encompass several different types of habitats. Some are in forested areas with substantial tree canopy overhead. Some are in open fields. Other cameras are along the edge of habitats and are positioned in locations where forests transition to fields or to open water. We use the area directly surrounding the camera to classify each habitat type as either forest, edge, or open.

The cameras also differ in the amount of human disturbance that occurs near them. Some sites are in urban areas with extensive landscape-level disturbance in the form of human infrastructure such as roads and bridges. Other sites are in relatively undeveloped land. We used the total length of roads and area of buildings within 1 km of the site to classify the lasting human disturbance (LHD) as high or low. Because urban areas often have lots of artificial light, even at night, we also classified each station as having high or low artificial light at night (ALAN). The stations also differ in the intensity of temporary human activity captured by each camera. Some cameras capture no human activity. Others record the occasional hiker or dog-walker. However, some cameras record frequent human disturbance, which may include people, dogs, farm animals, bicycles, cars, lawn mowers, and industrial vehicles. We used the frequency of images capturing disturbance, weighted by the intensity and duration of the disturbance, to classify each station as having either high, low, or no temporary human disturbance (THD).

   You have received an data file with the records of each photograph taken from May 2018 until August 2022. 

• The sheet labeled “Camera Images” contains observations of mammal activity.Each row represents a single image, and includes the camera site, station, habitat, THD level, LHD level, and ALAN level. The row also includes which species was identified, that species’ trophic level, the number of individuals in the image, and whether the animals were part of a group. Finally, each row includes information about when the image was taken, including the date, the season, the time, and whether the picture was taken during the day or at night. To characterize cathemeral activity, each row also includes information about the time the image was taken divided up into four categories; day, night, dawn, and dusk.

• The sheet labeled "Metadata" provided information on what each column in the "Camera Images" tab measures, including variable levels

•  The sheet labeled “Station Information” provides information about the site, habitat type, deployment time, and human disturbance at each camera station.

• The sheet labeled "Fox example" gives example of questions and analysis (more information below!)

Data review: Exploratory data analysis

First, review the data file. Notice it's big! 

1. How many rows are included in the Camera Images datasheet?

2. Why would analyzing this by hand be difficult?

Note some columns appear to duplicate information. For example, the time of the image is given in columns noting 

• the actual time,

• night or day

• night, day, dusk, or dawn

3. What is the benefit of displaying data in different formats?  

Use pivot tables or other approaches to begin exploring the dataset.  Here are some guiding questions:

4. How many species were recorded among all the traps?  Note since this is long data (Data Summaries in Google Sheets ) you can put any row in for values in a pivot table and use a "count" function to figure out how many rows correspond to each species.

5. Do you see any trends in when species were recorded across seasons? 

6. Do you see any trends in when species were recorded across time of day?

Planning your experiment

Consider questions the dataset may be used to explore.  For example, you might be interested in the effect of habitat type on diel activity or how human disturbance or habitat type impacts diversity.  whether mammals form groups more often in areas with higher human disturbance. You might also be interested in comparing different species or trophic levels to one another. For example, do animals with similar diets, such as raccoons and opossums, live in the same habitat types? You can use these examples, or you can come up with questions of your own. Once you decide on your topic, you need to determine which variables included in the data set will be best to use to test your hypotheses. For example, if you decide that you are interested in the effects of human disturbance on the diel activity patterns of mammals, you must choose which measure of human disturbance to use. You will also need to decide which information provides the best measure of diel activity. Finally, you must consider which observations to include. Are you interested in all mammals? In a single trophic level or species? You could also decide to focus on a single season or habitat type. Discuss your notes with your partner and develop a research question that interests both of you.

Refer to the slides for examples to questions 7-16.

7. What is your research question?

8. What will be your explanatory variables? Make  sure you list  columns in the datasheet.  

9. What are your response variables?  Make  sure you list  columns in the datasheet.  

10. What is your hypothesis?  State it and justify your prediction.

Carry out your (data-based) experiment

To determine if your hypothesis is supported, you will need a way to compare and analyze the number of observations in different categories. We will use a Fisher’s exact test to determine if two variables are associated with one another. If the two variables are associated, it means that the proportions of observations are different between the different categories. Fisher's exact test focuses on tables that display counts of how data fall into various categories. We can produce this type of data summary using pivot tables  (help @ Data Summaries in Google Sheets ).

11. Draw an image of how your pivot table needs to be set up.   What makes the rows and columns? What are the cell values? Do you need to filter the data?

12. Use a pivot table to aggregrate the data so that it matches your hypothesis. 

After you have made your contingency table, you will compute a p-value using a Fisher’s exact test. 

13. Rewrite your hypothesis in statistical terms (null and alternative) if needed.

Put the numbers from your contingency table into the web application available at http://statist.icu/. You can adjust the number of rows and columns for your specific research question. Leave any unused rows and columns blank, and do not include row totals. Press submit to get results

The top of the results (Fisher's exact test for count data) gives you an overall p-value. This tells if you if the proportions are different among  any of your rows (you can transpose the data if that makes  more sense!)

14. What is your overall p-value? 

If there is a difference among your rows, the page also returns pairwise tests so you can determine which rows are different than which others. 

14a. If your overall p-value is significant, report and interpret your pairwise comparisons.

Summarize your findings graphically

15. Produce a  plot that showcases your findings.

Asking a follow-up research question

Answering a question in science often leads to more questions! 

16. What is a related follow-up question you could ask?

17. What will be your explanatory variables? Make  sure you list  columns in the datasheet.  

18. What are your response variables?  Make  sure you list  columns in the datasheet.  

19. What is your hypothesis?  State it and justify your prediction.

20. Draw an image of how your pivot table needs to be set up.   What makes the rows and columns? What are the cell values? Do you need to filter the data?

21. Use a pivot table to aggregrate the data so that it matches your hypothesis. 

22. What is your overall p-value? 

23. Produce a  plot that showcases your findings.

Sharing your results

24. Write a paragraph (or two) summarizing your results and conclusions. Make sure to include your graphs and whether your experimental hypotheses were supported, what conclusions you can draw about mammal activity patterns, and the possible biological explanations for your results.