AI vs algae

Computer Science PhD candidate Olivier Graffeuille is helping develop solutions to the problem of detecting harmful algal blooms in New Zealand lakes using satellite data and applying machine learning techniques.

Harmful algal blooms frequently affect lakes in New Zealand and other parts of the world, where a rapid buildup of algae mass becomes toxic to human health, ecosystems and aquaculture. Freshwater scientists monitor these algal
blooms by collecting water samples and analysing them in a lab to estimate algal concentration to identify when water becomes unsafe. This process is expensive and time-consuming.

Olivier’s work is part of the Machine Learning to Monitor Harmful Algal Blooms project, based within the Centre of Machine Learning for Social Good. The centre’s primary goal is to advance fundamental knowledge in machine
learning and data analytics to address the most challenging and pressing health, environmental and societal problems for the benefit of our society.

Machine learning is a branch or subfield of artificial intelligence (AI). AI as a field covers a wide range of applications, including recent popular generative AI tools. The process typically uses data and algorithms to imitate how humans will learn, and it can also be used for decision-making.

“Our goal is to create models which take satellite data and then use that information to indicate that an algal bloom is occurring.

“As these environmental data sets are quite small and expensive to collect, our main challenge is to develop machine learning problems that learn from very little data. Most people have heard of tools like ChatGPT, but unlike other machine learning or AI systems where there is a lot of data to train the models on, a key difference here is we have very little data to work with.

“I witnessed a data collection process during the first year of my PhD at Lake Waikare, located between Auckland and Hamilton. The lake is very eutrophic. It’s dirty, shallow and continually full of algae, so it’s an interesting site for that reason. We went out on a boat and collected the water sample from the middle of the lake, which was then sent out to a lab for analysis.

“Every data point is hours and hours of work from environmental scientists, so it was great for me to see that process in person. Now whenever I go past a lake anywhere in New Zealand, I know I’ve seen that lake in my data set.

“There’s a lot of diversity in lake ecosystems, both within New Zealand and around the world. I still have a shallow understanding of lake ecosystems, and when I talk to my supervisors, I’m always impressed by the depth of their knowledge of this space. This diversity challenges us to work in different ways by applying diverse machine learning systems to make predictions.”

Olivier enjoys working as part of a multidisciplinary team, especially learning more about environmental applications and how scientists in other fields work.

“As computer scientists, we’re all a bunch of nerds. We see a bunch of data and want to make models with it. Environmental scientists must understand every detail, every data point and why the information is significant. It’s good to help
build understanding between the different fields and to see how the data is used in real life.”

“My environmental supervisor, Moritz Lehmann from the University of Waikato, has been learning more and more about artificial intelligence while I have been doing my PhD. As I’ve been learning about environmental science and its applications, Moritz has been learning about the machine learning process in return. He has been an applied environmental scientist his whole life and is now able to incorporate these methods in his work, too.”

“We’re also working with Mat Allan from Waikato Regional Council, who has been helpful with supplying data and interesting conversations about the lakes.”

One of Olivier’s supervisors is Professor Yun Sing Koh from the University of Auckland, who is also a co-founder of the Centre of Machine Learning for Social Good.

Yun Sing is excited about the Machine Learning to Monitor Harmful Algal Blooms project, citing it as a good example of a concerted transdisciplinary effort to address a challenging concern, embodying the centre’s foundational principles. She says, “This is the first machine learning for social good centre in Aotearoa. There have been other models out there in other parts of the world, but this is the first one that we have here. Our mission for the centre is to advance fundamental machine learning and data analytics while addressing the most pressing challenges in health.”

“I’m very passionate about what we’re doing for New Zealand here as we potentially tackle problems within our own backyard which can then be applied in a global context.

“Machine learning is a useful tool, but it’s not a silver bullet. Tackling these questions requires a concerted transdisciplinary effort across all sectors of society from the onset of a project and we can see that occurring here.

“I can also see how the learnings from this project can be translated to other questions and also applied to our knowledge of machine learning.”

Olivier is also excited about the local and global possibilities of the project.

“I think there’s many possibilities with the techniques we’re developing here. The nature of satellites means we have global coverage of data all the time and the scalability is exciting.

“Currently, the data is being collected to monitor algal bloom, but we could get to the stage where we’ve got these accurate remote sensing models. Within the next decade this could potentially be used to provide accurate water quality models globally, which would be a huge societal benefit, at little added cost because the satellites are already there.

“Natural hazard monitoring for things like bushfires is another important issue
where this approach could be applied, using global data.”

Olivier would like to pursue a career in industry once he completes his PhD.

“I hope to find a position where I can use my knowledge to help a company in Auckland or overseas. Ideally, where I can make meaningful positive change, using the knowledge I’ve gained here to help solve environmental problems.”