In the framework of the Flemish contributions to the LifeWatch infrastructure, a camera trap network was initiated to monitor wildlife (especially mammals) and better understand their distribution, population trends, behaviour and habitat use.
Assessing the size and trends of animal populations is a long-standing challenge in wildlife ecology which requires sufficient and reliable data. However, the collection of data in the field can be daunting and time consuming. Recent technological advances on the hardware have led to the adoption of automated camera traps as research instruments. Camera traps, equipped with infrared sensors that get triggered by motion, can provide a reliable and cost-effective tool for the collection of long-term monitoring data. Besides the fact that these cameras are a time-efficient tool for data collection, it is also an non-invasive technique causing low levels of disturbance as no animals need to be captured or killed.
Camera traps are deployed at a location for a specific duration and generate large quantities of images that may or may not include the species you are interested in. Information that typically can be derived from camera trap images include animal species, number of individuals, their sex, age class and behaviour, and, in case the animals are individually tagged, their name or tag code. Images are often captured as a burst of images, which allows one to better understand their behaviour or measure their walking speed.
Most wildlife researchers are mainly interested in a selected group of animal species. Whereas carnivore specialists meticulously filter out those pictures containing carnivorous species, bird experts could find a valuable source of information in the discarded pictures containing birds. In order to mitigate ecological problems such as a reduced biodiversity and changing species compositions due to climate change and the introduction of invasive alien species, it is of utmost importance that conservation scientists collaborate and share as much information as possible with the scientific community.
As part of the Flemish contribution to the LifeWatch infrastructure, the Research Institute for Nature and Forest (INBO) launched the CATREIN (CAmera Trap REsearch INfrastructure) project in 2017. The aim of the project is two-fold:
The camera trap network was started in 2017 by INBO. Before that time, camera traps were deployed in various short-term projects and monitoring initiatives, resulting in different types of cameras and little attention to data management. This is an impediment to scaling up to long-term and large scale studies. With the installation of CATREIN, a large pool of high quality cameras was purchased.
The first 45 cameras were installed in 2017 in the LTER-site National Park Hoge Kempen by INBO in collaboration with Hasselt University (CMK, centre for environmental sciences). Cameras were put on randomly chosen locations to collect data of the habitat use of wild boar in the framework of a PhD study. Other major research sites include the LTER-sites Forest of Meerdaal and Sonian Forest.
A network of camera traps quickly generates a massive amount of images, which makes a data management platform for aggregating images in projects, project management and the annotation of images indispensable. Among the already existing camera trap management systems, Agouti suited our needs best.
Agouti is an online platform for camera trap data management created by the Wildlife Ecology and Conservation Group at Wageningen University. It lets camera trappers organize collaborative surveys, quickly and easily process and annotate images, obtain standardized summaries of the results, and safely archive images and data.
In 2018 a partnership with INBO was formed as a result of the initiation of a camera trap sensor network (CATREIN) for LifeWatch. With joined forces further development took place. The system was upscaled for handling larger amounts of images, data and users, errors were removed and the processing speed was substantially increased. Furthermore, new modules were added and community (user) management was optimized. New features in development include the automated recognition of people and animals, citizen science through crowdsourcing, a validation module for annotated images, hosting of public images and a standard export format for data including a pipeline to publish camera trap images as open data.
The Agouti infrastructure enhances collaborations between scientists and the reuse of data as all species present on pictures are recorded instead of just focussing on the studied species. The user group keeps growing and includes wildlife managers, policy makers, food and safety authorities, biologists and amateur enthusiasts located in several countries across the globe.
An overview of the INBO projects in Agouti can be found on the organization's landing page.
Camera Trap Data Package (Camtrap DP) was developed to allow the FAIR exchange of camera trap data. It is a community developed data exchange format for camera trap data and can be used for the export and exchange of data from the Agouti platform, but also for data from other data management systems, such as TRAPPER. A Camera Trap Data Package contains data and project metadata, the camera trap deployments, links to the multimedia files captured by the camera traps, and a table with the observations. These data files can be transformed into Darwin Core for the publication of the camera trap data at GBIF.
In 2021, the first camera trap dataset was published as FAIR open data on the GBIF platform. Data and images are archived as a Camera Trap Data Package on Zenodo.