Interactions between humans and wildlife are a growing concern associated with increased human presence in wildlife habitats. Collecting reliable geographical data on human-wildlife interactions poses a significant challenge owing to the cryptic nature of wildlife and the fleeting timing of such interactions. In this presentation I will demonstrate a citizen science approach for studying human-wildlife interactions, and how it links with more traditional spatial ecology methods. GPS tracking is used to collect fine-scale spatial-temporal data on the locations of people along a hiking trail. At the same time, hikers were asked to complete a wildlife viewing survey that was linked to the GPS data based on the time attribute. Specifically, I will demonstrate new tools for mapping human-wildlife interactions and studying the environmental context within which these interactions occur.
Human activity in remote and natural areas
is increasing
Collecting robust data on human wildlife
interactions is challenging for a variety of
reasons. First, these interactions are often
fleeting, and may not always be realized by
the human actor. Second, they may be
associated with a distance decay effect, i.e.,
interactions are stronger the closer the two
individuals involved are. Finally,
humanwildlife interactions are generally rare
events, occurring often in more remote
areas. Thus, innovative methods are
required to collect reliable data on such
interactions. Citizen science offers an
opportunity for studying the impacts of
human outdoor activity on local wildlife
Here I demonstrate a study aimed at collecting, analyzing, and mapping humanwildlife interactions. I explore the types of data that can be generated for studying human-wildlife interaction in a citizen science context. The aim of the presentation is to demonstrate new spatial tools for studying these interactions and how these can be used to understand unique spatial events.
The study took place in Glen Lyon in the Perthshire region of Scotland (Figure 1). The site includes a popular 17.5 km hiking trail which includes summits to four prominent munros (defined as peaks above 3000 ft; Carn Gorm, Meall Garbh, Carn Mairg, Creag Mhor). Elevation in the area ranges from 210 m at the trailhead to a maximum of 1042 m (3419 ft; Carn Mairg). The trail is situated on an estate, which also runs several outdoor recreation activities including red deer (Cervus elaphus) stalking, fishing excursions, and has domestic livestock (i.e., sheep) roaming free throughout.
We collected sample data during the summer and autumn months of 2017 and 2018 stratifying our sample days across weekends and weekdays. During sampling days, we asked all hikers entering the trail to carry a GPS device while out on the hill. For each group of hikers (groups defined as individuals from the same party walking together) that agreed to participate we gave them one small portable GPS device (GPSPro 747) to be carried by a single member of each group. The GPS devices were pre-programmed to record position continuously (i.e., one position fix every 5 seconds) prior to being given to a participant. A drop-box was located at the return point (near the car park) where GPS devices could be returned if the team member was no longer present. We did not collect any further information (e.g., age, gender) about hikers during this experiment.
At the same time, we asked participants to carry and fill-out a wildlife viewing survey, which was a piece of card which we provided (along with a pencil). The survey required participants to record the time, species of wildlife, and approximate distance and bearing at which wildlife were viewed while hill-walking (Figure 2). The survey was designed to be simple and easy to fill-out. The cards were then transferred to a digital spreadsheet by a team member.
Based on the time information provided by participants in the wildlife survey the location of the walker at that point in time was cross-referenced based on their GPS tracking data. The locations where walkers viewed wildlife then served as the focal point for estimating the location of the wildlife at that point in time. Where the participant provided an estimate of the distance and bearing of the wildlife encounter, we used this information to map that location using simple geometry. Any wildlife encounter recorded by a participant with a distance estimate of > 500 m was not used in subsequent analysis. When a participant did not provide this information, we simply mapped the encounter based on the location of the participant at the time of the encounter.
Throughout both summers we deployed an array of camera traps situated along transects at various points along the hiking trail and at random locations throughout the study area. The cameras use an infrared sensor to trigger photos and capable of detecting animals in both day and night and across all weather patterns. We focused our study on red deer, but the cameras also captured other animals – mostly sheep). Camera trap photos were manually processed by a team member to codify whether deer were present (and the presence of other animals, i.e., sheep). We also identified various deer behaviors (e.g., running, head-up, head-down) from photos to study how different behavior of red deer were associated with distance to the trail, and the presence of hikers.
To explore contextual factors associated with the hill walker data, we calculated sightlines for all encounters recorded in the wildlife viewing survey and for every camera trap location using GIS-based viewshed analysis.
We collected sample data on 35 (25 in 2017 and 10 in 2018) different days during the summers of 2017 and 2018. Of the 197 people that we approached, 185 agreed to participate, a success rate of 93.9%. From the 185 participants we collected 153 wildlife surveys with useable data. In total, we were able to successfully digitize 323 (2017: 259 and 2018: 64) wildlife encounters.
In order to map wildlife encounters, we needed useable measures of time, distance, and bearing. Where distance or bearing was missing we assumed the encounter occurred in proximity with the hiker. Of the 323 wildlife encounters collected, 143 (2017: 102, 2018: 41) contained both distance and bearing estimates. An example of a single participant’s GPS data and their mapped wildlife encounters is provided in Figure 3.
To test the quality of the participant wildlife encounter data we compare mapped encounters with the observations from the camera traps and visual observations in the field. Here we restrict the analysis to only red deer as they are the focal species for which the camera traps were deployed (e.g., the height and calibration of the cameras), and thus other species are not commonly observed.
Our preliminary results suggest issues with relying on participant generated content to identify human-wildlife interactions. Based on our in-situ visual observations we found that hikers routinely failed to see wildlife, even when they were within sight lines.
We cross-referenced all hiker GPS data with the camera trap data to identify the times when participants were near (within 100m) of the cameras. Over the course of two summers we collected over 100 000 camera trap images, of which approximately 30% contained our focal species (red deer). From this we estimated the error rates associated with the participant survey wildlife encounter data and found that there was a high-error rate of encounters with red deer that were missed by hikers.
Our study explores the feasibility of using a
citizen science approach for collecting
geographical data on human-wildlife
interactions. Specifically, we found an
extremely high level of engagement in our
preliminary study (greater than 90%
participation rate). This rate of participation
at initial glance is high, but other studies
have shown comparatively high willingness
by outdoor recreationalists to participate in
GPS tracking studies
While the data we have collected appears
to be of relatively good quality upon initial
inspection. However, in 2018 we situated a
team member at a viewpoint within the site
and found that hikers routinely did not
identify deer that were within viewing
range. Other problematic aspects of citizen
science studies however need further study,
for example, what might be more important
is the variability between participants in
their capability to observe (or report)
wildlife sightings
The approach we have taken here is highly
labor intensive (i.e., it requires a study
member be present to pass out GPS devices
and the survey). Future work will explore
how to upscale data collection of human
wildlife interactions using mobile-phone
based apps. Other studies have
demonstrated how mobile-phone apps can
be used effectively in collecting similar types
data in ecological field studies
It is generally unknown at what distance
human presence influences different wildlife
species. Previous studies have explored this
in different contexts, for example previous
research has found that rocky mountain elk
(Cervus elaphus L.) respond at large
distances (i.e., up to 2000 m) to all-terrain
vehicles
In summary, collecting reliable and robust geographical data on human-wildlife interactions is a challenge, owing to the cryptic nature of wildlife and the fleeting timing of human-wildlife interactions. We employed a citizen science approach to collect data on wildlife encounters along a popular hill-walking route in the Glen Lyon region of the Scottish Uplands. Specifically, we used voluntary GPS tracking of hikers and a paper-based wildlife viewing survey to map the locations of wildlife encounters along a hiking trail.
Funding and contributions to this work have been provided by The James Hutton Institute, The Carnegie Trust, The British Deer Society, The Association of Deer Management Groups, and the North Chesthill Estate.