deep learning-based solutions

Over the past two decades, I have developed AI tools to support decision-making in biodiversity conservation. Methods such as Markov Decision Processes (MDPs), Partially Observable MDPs (POMDPs), and adaptive management help us plan under uncertainty, incorporate feedback, and learn from dynamic ecological systems. These techniques have supported efforts from invasive species management to threatened species recovery, helping balance short-term actions with long-term outcomes. 

Yet some of the most important insights lie beyond the algorithms themselves. Reversing the biodiversity extinction crisis requires more than technical capability: it demands co-design with ecologists, social scientists, landholders, and policymakers. It calls for a shift from what can we build? to what does biodiversity actually need?

This is the mission of the Environmental Informatics Hub at Monash, a new initiative I lead within the Faculty of IT: to develop AI that supports the future of biodiversity, not just the future of technology. But isolated efforts won’t be enough. If we are serious about halting biodiversity loss, we need a globally coordinated approach, a clear framework, much like the Global Biodiversity Framework (GBF), that defines the AI capabilities required to meet conservation targets.

Learning Objectives:

1. Learn about how AI can help support decision-making for the conservation of biodiversity
2. Learn about the needs and challenges in developing interpretable, non-autonomous and poor data AI systems

Estimating the geographic range of a species from sparse observations is a challenging and important geospatial prediction problem. Given a set of locations where a species has been observed, the goal is to build a model to predict whether the species is present or absent at any location. This problem has a long history in ecology, but traditional methods struggle to take advantage of emerging large-scale crowdsourced datasets which can include tens of millions of observations of hundreds of thousands of species in addition to multi-modal image and text data. In this talk, I will present recent work from my group on deep learning-based solutions for estimating species’ ranges from incomplete data. I will also discuss some of the open challenges that exist in this space.  

Learning Objectives:  

1. Understand the capabilities of current deep learning methods for species range estimation.
2. Recognise the limitations of these models in the context of current open challenges in this space