TERN Webinar – Towards an Australian Ecosystem Foundation Model

A foundation model is a large-scale AI model trained on a wide variety of data. This year, a Nature paper reported on the Microsoft Foundation Model, Aurora, and its capacity to leverage the latest advances in AI to more accurately predict not just the weather, but a wide range of environmental events in a series of retrospective analyses.

Australia has extensive long-term terrestrial, ocean, and coastal ecosystem monitoring data. They are part of our national research infrastructure, derived from sources such as satellite remote sensing, UAVs, in-situ sensors, and human-collected systematic surveys. To understand complex ecosystems and run multifaceted analyses, these multi-scale, multi-domain datasets need to be used together and independently in models, if we are to successfully span Australia’s diverse environments and answer questions covering time periods from minutes to centuries. Fortunately, with our machine-readable ecosystem datasets, we possess the required unique resources that could train foundation models to understand continental-scale ecological processes.

The webinar speakers will share their thinking about building both domain-specific and cross-domain omnibus foundation models, tailored for Australian landscapes to address ecosystem monitoring, resource management, and disaster response needs—contributing to Australian climate change mitigation efforts.

This webinar isn’t just another technical discussion—it’s a strategic conversation about our environmental future, embracing AI foundation models as the research infrastructure needed for the coming decades for understanding ecosystems. It is also about answering the critical question—if we don’t act now, will we fall behind international efforts while others unlock the potential of our openly accessible, free environmental data?

Speakers

• Dr Andrew Prata, CSIRO
• Dr Roozbeh Valavi, CSIRO
• Siddeswara Guru, TERN Australia
• Stefan Hajkowicz, CSIRO