In a world first, UNSW and Sydney Botanic Garden scientists will use data from millions of botanical specimens held in botanic gardens around the world to study and combat the impacts of climate change on flora. trained an AI to unleash
“The specimen collection is a wonderful time capsule of herbarium,” says associate professor Will Cornwell, lead author of the study. “The National Herbarium of New South Wales alone adds more than 8,000 specimens each year, so it is no longer possible to examine them all by hand.”
Using a new machine-learning algorithm to process more than 3,000 leaf samples, the research team found that within a single species, leaf size increased in temperate climates, in contrast to frequently observed interspecific patterns. I found that it doesn’t increase.
Published in American Journal of Botanythis study not only revealed that non-climatic factors strongly influence leaf size within plant species, but also used AI to transform static specimen collections and explore the effects of climate change. demonstrated how to quickly and effectively document
The Herbarium Collection Moves to the Digital World
Herbaria is a scientific library of botanical specimens that has existed since at least the 16th century.
“Historically, a valuable scientific undertaking was to go out and collect plants and store them in herbaria. Every record has a time and place, a collector, and a putative species ID. says A/Prof. Mr. Cornwell is a Researcher at the School of BEES and a member of the UNSW Data Science His Hub.
A few years ago there was a move to transfer these collections online to facilitate scientific collaboration.
“Herbarium collections used to be confined to small boxes in specific locations, but the world is now very digital. Efforts were made to scan the herbarium to provide a high-resolution digital copy of the specimen.”
The largest herbarium imaging project was carried out at the Sydney Botanic Gardens, transforming over one million herbarium specimens at the National Botanic Gardens of New South Wales into high-resolution digital images.
“The digitization project took over two years, but soon after it was completed, one of the researchers, Dr. Jason Bragg, contacted me from the Sydney Botanic Gardens. We wanted to see how we could incorporate machine learning into some of the digital images of specimens.”
“We were excited to work with Professor Cornwell to develop a model to detect leaves in plant images and use those large datasets to study the relationship between leaf size and climate,” says Dr. Bragg. .
Measure leaf size with “computer vision”
With Dr. Bragg of Sydney Botanic Gardens and UNSW Honors Professor Brendan Wilde. Cornwell has created an algorithm that can be automated to detect and measure leaf size in scanned herbarium samples for two plant genera. Syzygium (commonly known as liripili, blush cherry, or satina) and fig (a genus of about 850 species of trees, shrubs and vines).
“This is a type of AI called convolutional neural networks, also known as computer vision,” says Professor A/. Cornwell. This process essentially teaches AI how to see and identify plant components in the same way humans do.
“We had to build a training data set to tell the computer that this is a leaf, this is a stem, this is a flower,” says Professor A/. Cornwell. “So we basically taught the computer to locate the leaves and measure their size.
“Measuring leaf size is nothing new, as many people do it. But the speed with which we can process these specimens and record individual features is a new development.”
Breaking out of frequently observed patterns
A general rule of thumb in the botanical world is that plants grow larger leaves in humid climates, such as rainforests, than in dry climates, such as deserts.
“And it’s a very consistent pattern seen in interspecies leaves around the world,” says Professor A/. Cornwell. “The first test we did was to see if we could reconstruct that relationship from the machine-learned data, and it was possible. We have a lot of data, so the question was, can we see the same thing, within a species?”
Machine learning algorithms were developed, validated and applied to analyze leaf size–climate relationships within and between species. Syzygium and fig plant.
The results of this test were astonishing. The research team found that although this pattern is seen across different plant species, the same correlation is not seen within a single species worldwide. This is probably due to a different process known as gene flow. Active within species. This process may weaken plant adaptations on a local scale and prevent leaf size–climate relationships from developing within species.
Using AI to predict future climate change responses
The machine learning approach used here for leaf detection and measurement, although not pixel-perfect, provided a good level of accuracy for examining associations between leaf characteristics and climate.
“But the world is changing so fast, and there is so much data, that these kinds of machine learning techniques can be used to effectively document the impacts of climate change,” said A/ says the professor. Cornwell.
Additionally, machine learning algorithms can be trained to identify trends that are not readily apparent to human researchers. This could lead to new insights into plant evolution and adaptation, as well as predictions about how plants will respond to future climate change impacts.
