By: Kathleen Sandusky
29 Nov, 2024
Graham Taylor, Vector Institute: Using AI to track insect biodiversity loss in real time
November 29, 2024
Graham Taylor, Canada CIFAR AI Chair, Vector Institute
The climate crisis poses a severe threat to global insect populations, with a recent NASA study predicting that 65% of the world’s insect species could face extinction over the next century. Such a decline would ripple across ecosystems, affecting plants and animals and threatening human populations through impacts like increased disease spread and loss of crop pollinators.
To counteract these risks, researchers worldwide are collecting extensive data on the range and density of insect species. Ecologists now use tools like “Malaise traps,” or bug tents, that capture and count insect populations, camera systems for visual identification, and audio devices that analyze sound production. This complex, multi-modal data includes images, text, audio and DNA snippets translated into “barcodes” stored in a vast global repository. Each data type requires integration and analysis to provide actionable insights &— a monumental task for human scientists facing the pace of global insect decline.
“…my students are collaborating across teams and messaging through Slack, mentoring and learning from each other. This community of Canada CIFAR AI Chairs has been very valuable for them, too.”
Graham Taylor, Canada CIFAR AI Chair, Vector Institute
Graham Taylor, a Canada CIFAR AI Chair at the Vector Institute and professor at the University of Guelph, is tackling this need through AI. Taylor is Chief Data Officer at the Centre for Biodiversity Genomics, working with the global BIOSCAN initiative to build a comprehensive genetic catalog of planetary biodiversity. This initiative spans multiple disciplines and involves collaborations with policymakers worldwide to address species loss. Taylor is also a primary investigator at the AI and Climate Change Global Biodiversity Center, an international consortium recently funded by the US National Science Foundation and Canada’s NSERC. This project is advancing AI-based methods for integrating and analyzing biodiversity data from diverse sources, including remote sensing from satellites and low-flying aircrafts, ground-based sensors, DNA sequences and citizen science.
The abundance of all these different types of data is both a blessing and a challenge. “Raw, unstructured data from these many sources is essential but it can’t be used by AI without processing,” explains Taylor. “The challenge is to format it all in a way that machine learning can use it to identify specific species, their locations and population abundances. Luckily, this is what deep learning is really good at. We take the images, audio, text, DNA barcodes, and turn it all into more structured and usable information for data analysis.”
Taylor notes that Nanopore sequencing, a technique adapted from healthcare, will be crucial in scaling up DNA barcoding to a global scale. “Previously, to get genetics data for biodiversity you needed very expensive equipment. Someone would need to send a whole dead bug specimen to a central lab for sequencing. But during COVID, new technology was developed that can create the barcoding out in the field, with just a laptop and small DNA sequencer. We can now barcode hundreds of thousands of specimens for less than a cent each.” Collaborators like Canada CIFAR AI Chair Angel Chang at Amii are developing methods to use these DNA barcodes as “weak supervision,” helping machine learning models identify insect photos more accurately and reducing human hours for data labeling.
This work is already making an impact in Canada. In Ontario, Taylor’s team is partnering with the Canadian not-for-profit Alternative Land Use Services to provide farmers and landowners with low-cost methods for assessing the biodiversity potential of fallow or unproductive land. This analysis, called BugShot, supports funding proposals to convert these areas into conservation habitats for threatened species.
Taylor credits his Canada CIFAR AI Chair appointment for connecting him and his students to Canada’s scientific research community. His involvement with the AI and Climate Change Global Biodiversity Center resulted from a connection with Canada CIFAR AI Chair David Rolnick at Mila, who leads the Canadian arm of the project. And Taylor’s DNA barcoding collaboration with Angel Chang at Amii was sparked by a discussion at a CIFAR meeting.
Additionally, Taylor emphasizes the benefits of the Canada CIFAR AI Chair program for his students. “Many of my students have attended the CIFAR Deep Learning + Reinforcement Learning Summer School, which has been really beneficial. Beyond that, my students are collaborating across teams and messaging through Slack, mentoring and learning from each other. This community of Canada CIFAR AI Chairs has been very valuable for them, too.”