AI uncovers the history of Australia's iconic trees

Scientists have used new developments in machine learning to look at Australian eucalyptus species and reveal changes over millions of years.

Advances in artificial intelligence are allowing scientists to explore the evolutionary history of Australia's iconic trees, which could help tackle threats such as climate change and biodiversity loss, scientists say.

In the study, scientists from the Sydney Botanic Gardens and the University of New South Wales used AI to analyse the leaves of an unprecedented number of eucalyptus trees (Eucalyptus, Angophora and Corymbia), gaining extraordinary insights into how these native species have evolved with climate.

This paper is Ecology Journal This week, researchers will analyse a dataset of more than 50,000 digital images of eucalyptus specimens – some dating back to 1839 – to determine whether the species' leaves have evolved in response to changing climates.

This follows research last year, in which scientists from the Sydney Botanic Gardens and the University of New South Wales built a machine learning program to examine millions of plant specimens stored in herbariums around the world. The approach introduced a resource that researchers previously couldn't access because herbarium collections are too large for humans to measure.

In this first study, the team used “computer vision” methods to analyze 3,000 samples of Syzygium and Ficus for leaf size. They found that, contrary to frequently observed patterns between species, leaf size within a species does not increase in warmer, wetter climates.

Now a team of researchers is taking this work a step further, focusing on one of Australia's most iconic and beloved trees, the eucalyptus, to understand how its leaf size has changed with the climate over millions of years.

Karina Guo, a scientist at the Sydney Botanic Gardens, said the research shows how AI is transforming plant science.

“Using AI, we can now handle huge amounts of data that were not possible before,” Guo says. “AI is changing the way we find the fine details in plants and helping us paint a very detailed picture of the past.”

“Instead of manually assessing thousands of specimen images, which would take years, machine learning can examine tens of thousands of specimen images in less than four days.”

UNSW research associate Associate Professor Will Cornwell describes this new ability to uncover previously inaccessible data as groundbreaking.

“Digitized specimens will enable us to understand species in greater depth than ever before and ultimately help us address major threats to our flora, such as climate change and biodiversity loss,” Associate Professor Cornwell said.