350 Monitoring Threatened Species and Ecological Communities
larger the effort required to manage it. The level of outreach achieved by Wildlife
Spotter is rarely obtainable by individual research groups. Over 50 000 people have
logged on to the website and categorised almost three million sensor camera
images. Wildlife Spotter’s broad engagement has greatly raised the profile and
understanding of northern bettongs and the work being done to save them. The
Wildlife Spotter portal considerably accelerated image processing: more than
29 000 h of work was carried out by volunteers in 2 months, equivalent to one
person working full-time for 15 years. Dedicated individuals contribute to analysis
in a big way: one user alone classified 19 082 images (ABC pers. comm.).
Teams of specialists with diverse backgrounds were integral to the success of
Wildlife Spotter: a marketing team for social media campaigns, interviews, videos
and media releases; a development and design team for building the website and
providing technical support; and scientists for technical advice. This collaboration
ensured the portal was designed to invite broad participation, be user-friendly and
engaging, and encourage technical accuracy. This last aspect was partly ensured by
requiring users to click through a tutorial before beginning photo analysis, and by
having an easily accessed guide for users who get stuck. Image-based instructions
and recording icons allow a range of volunteers with variable skills to participate.
Ensuring data quality
During cage trapping, volunteers undertake relatively complex tasks and, because
integrity of data is paramount, hands-on face-to-face training is required.
Feedback from volunteers is encouraged, and over time the training protocols have
been refined so fieldwork runs smoothly. Electronic tablets are used instead of
paper datasheets and are set to ensure all required data is collected in a
standardised format. Data and collected samples are reviewed while in-the-field so
any discrepancies or missing information can be corrected.
When training is minimal and/or participants are not supervised, further
safeguards are needed to ensure data integrity. Despite the well-thought-out
development and design of Wildlife Spotter, concerns about the reliability of results
remain with any citizen science project. To tackle this potential issue, WWF
analysed a subset of data to determine the accuracy of Wildlife Spotter results to
see if: (1) bettong presence/absence at a camera site was being correctly determined;
and (2) identifications of individual photos were reliable enough for use in
occupancy/abundance studies.
It was found that users correctly identified bettong presence/absence at 100% of
sites, meaning the results can be used reliably for determining distribution.
Looking at individual photo identification, when a majority result occurred (three
or more users out of five identified the same species), the animal was correctly
identified 100% of the time. Out of the subset, 92.5% of the photos had a majority
result (and hence were correctly identified). Photos without a majority result were