Research highlight

Scientific Data: Citizen science project reveals penguin breeding dynamics

Scientific Data

June 27, 2018

Data from almost 74,000 images capturing the dynamics of brush-tailed penguin breeding colonies across the Antarctic Peninsula, South Shetland Islands, and South Georgia are reported this week in Scientific Data. Using a citizen science approach, the images (taken from 15 different cameras as part of the Zooniverse project ‘Penguin Watch’) were classified by volunteers.

Owing to its remote, harsh environment, large-scale on-ground monitoring studies in Antarctica are challenging and therefore rare. As such, most studies investigating penguin population dynamics have focused on a specific location, or collectively several locations, with subsequent extrapolation of local data to cover a wider region or regions. However, this approach is insufficient when it comes to understanding numerous widespread populations. A greater understanding of penguin population dynamics, reproductive success, and phenology will allow the impact of threats, such as climate change and over-fishing, to be monitored, thus paving the way for more effective conservation measures

Fiona Jones and colleagues present anonymized volunteer classifications for the 73,802 images, alongside associated metadata, including date, time, and temperature information. Each camera in the ‘Penguin Watch’ network generally captures images once per hour, between 7am and 8pm, year-round. Volunteers classified images by tagging individuals and labelling them as ‘adult’, ‘chick’, or ‘egg’ for penguins, or ‘other’ to indicate the presence of other animals, humans, or ships. This level of annotation allows detection of important phenological stages, such as chick hatching. To date, ‘Penguin Watch’ has processed over six million images that have been classified by nearly 48,000 registered volunteers and a wealth of anonymous participants.

The authors suggest that, in addition to the benefits for ecological monitoring, this type of annotated time-lapse imagery can be employed as a training tool for machine-learning algorithms to automate data extraction. They also suggest that the methodology described in the paper provides validation for the use of citizen science processes.

doi: 10.1038/sdata.2018.124

Return to research highlights

PrivacyMark System