The Wildlife Observatory of Australia (WildObs)

The Challenge

Australia leads the world in deploying wildlife cameras and other autonomous sensors to monitor terrestrial wildlife populations. However, the infrastructure for storing, integrating, analysing, detecting, annotating, and reporting on this vast data stream needs further development. By enhancing these capabilities, Australia can fully leverage machine-based observations to monitor biodiversity, address climate change challenges, and inform national and local biodiversity strategies.

Overcoming key hurdles, such as collating and processing millions of wildlife camera images, creating compatible datasets, and merging data streams, will be essential for establishing a national wildlife camera data commons that unlocks the powerful analyses that are possible when rich data are brought together. Additionally, the shared development between wildlife cameras and acoustic recorders will open up unprecedented integration of machine observation data.

The Response

The Wildlife Observatory of Australia (WildObs) aims to create an efficient pipeline for processing large image datasets of Australian species. The key steps to be worked on before June 2026 include:

1. Creation of a Tagged Image Repository: collect tagged images to serve as a training repository for AI models.
2. Wildlife Image Management Platform Selection: evaluate and choose a long-term Australian data processing solution for AI/ML models capable of identifying fauna in images.
3. Wildlife Camera Detections Database: develop a database that seamlessly imports camera-trap data and exports post-processing detection histories to repositories.

We’ll create APIs to upload and share tabular datasets from contributors. By standardising data formats across wildlife cameras and acoustics, we’ll ensure compatibility between existing datasets. Collaboration with TERN and the Atlas of Living Australia, which both feed into ARDC-supported analytical platforms EcoCommons and Biosecurity Commons, will enhance data storage capacity and encourage contributions from various stakeholders.

The next phase involves iterative development, stakeholder consultation, and feedback. We’ll create a roadmap for a user-friendly machine observation data commons, allowing approved users to submit and query data seamlessly. Regular workshops and training events will engage the wildlife monitoring community, ensuring widespread adoption.

By following this plan, we’ll integrate historical and emerging data streams, leverage AI-powered computer vision, and enhance scalability. These will mark the first steps toward building a fully integrated machine observation ecosystem which would open the door to powerful analyses.WildObs is a partnership led by QCIF and supported by the ARDC’s Planet Research Data Commons, as part of the Machine Observation Data Processing Infrastructure (MODS) program. The MODS program aims to establish national shared infrastructure, services and standards to enable processing and reuse of automated observation and sensor data.

Who Will Benefit

Governments, museums, academics, NGOs and citizen science groups

The Partners

• QCIF (project lead)
• ARDC (project lead)
• TERN
• Atlas of Living Australia (ALA)
• Queensland Government Department of Environment and Science (DESI)
• University of Queensland
• University of Tasmania
• Australian Government Department of Climate Change, Energy, the Environment and Water (DCCEEW)
• Bush Heritage
• University of Sydney
• Australian Museum

Target Outcomes

WildObs will be a purpose-built analysis infrastructure and a wildlife camera data commons that will address the hurdles of bottlenecks, siloing, variation in survey method, and access to advanced analysis and compute.

Our long-term goal, in collaboration with the Planet Research Data Commons’ Machine Observation Data Processing (MODs) program, is to establish a world-leading observatory for monitoring and predicting the distribution and changes in terrestrial vertebrate populations. This initiative will revolutionise collaboration and data utilisation.

Through the MODS program, WildObs will provide the component infrastructures for wildlife camera-trap data, while working with the acoustic community on standardisation.

Key Components of the Future Wildlife Observatory:

1. Shared National Infrastructure: The observatory will leverage national infrastructure to process, manage, aggregate, and report on data from wildlife cameras.
   
2. Tagged Images and Recordings Repository: We’ll create an open repository for AI/ML research, allowing tagged images and recordings to be accessed seamlessly.
   
3. Centralised Species Database: Information from recordings and images (species detections) will feed into a centralised database. This scalable approach integrates historical and ongoing wildlife datasets.
   
4. Advanced Compute and Analytics: Coupled with cutting-edge technology, the observatory will facilitate timely translation from detections to actionable information. This includes image and acoustic detection, identification, annotation, and reporting.
   

Derived Results and Modelling, Analytics and Decision Support Infrastructure (MADSI) Program:

• The observatory will provide derived metrics, maps, and reports on threatened or invasive species. These outputs will inform policy and decision-making.
• Non-specialists will benefit from user-friendly layers and reports, potentially supported by high-performance computing (HPC) resources through the MADSI program.

Key Resources

• Read the QCIF project announcement
• Read about WildObs on the TERN website
• Learn about project partner the Ecological Cascades Lab at The University of Queensland, led by Dr Matthew Luskin