Skip to main content
To KTH's start page

The ocean's smallest creature is mapped

Scientists in lab
Anders Andersson, professor, and Karin Garefelt, PhD student, studying microscopy images of marine plankton analysed and identified with image recognition software (Photo: Magnus Glans)

AI and humans in focus

Published Oct 01, 2024

How will a warmer climate affect the ecosystem? Genetic researchers at KTH Royal Institute of Technology are mapping changes in marine plankton to get closer to the answer. Using AI-based image analysis, the research has taken a big step forward.

Anders Andersson's research group in gene technology aims to improve Swedish marine environmental monitoring, which monitors and protects the sensitive marine environment. The researchers are collaborating with SMHI's oceanographic research unit, which is producing data for assessing the health of marine areas.

The focus is on the billions of microorganisms in the water. Different species of plankton are being mapped and analysed to understand their nutritional needs, relationships and adaptations to the environment.

A new project uses a high-tech camera, the Imaging Flow Cytobot, mounted on the research vessel Svea's expeditions in the Skagerrak, Kattegat and Baltic Sea. The expeditions generate hundreds of thousands of images of plankton that are analysed by a neural network trained to recognise the different species.

“AI is unbeatable when it comes to image analysis. It is much more efficient and we can analyse very large amounts of data, and at a lower cost. Developments are progressing rapidly, and AI methods are also providing more reliable results,” says Anders Andersson , professor of Metagenomics, who leads the research group.

Scientists in laboratory
Researchers expect that the development of AI will provide them with even better tools to map the distribution of plankton in our oceans.

Since all material is stored, it is also easy to reanalyse the images as the AI algorithms improve.

Plankton changes

Researchers have previously shown how the composition of marine plankton has changed over time, probably as a result of climate change. Some species, such as diatoms, have declined while the number of cyanobacteria has increased in many areas.

What the consequences of the shift in the plankton world might be, no one knows for sure. Research results point in different directions. Reports have shown vitamin B1 deficiencies in salmon and eider, among others. This may be related to the decline in diatoms, but is not yet clear. Earlier reports that some small crustaceans would become extinct have also proved to be at least partly incorrect, explains Andersson.

“There are many pieces of the puzzle to be put in place in the future so that we can get a complete picture of how the ecosystem is affected.”

At the same time, there is a lot at stake. Marine plankton creates half of the world's oxygen and produces food for the entire aquatic food chain - from small crustaceans to seals.

“The more and more reliable data we collect, the more we can learn about how the ocean works and take appropriate action when needed.”

Unexpected behaviours

With the help of the AI tool, the research team, together with SMHI, has also been able to gain a more nuanced understanding of phytoplankton behavioural patterns. It shows that plankton are ‘day commuters’ and move from the surface during the day down to the depths where they gather nutrients during the night. This is another valuable piece of the puzzle to get closer to a more complete picture.

“We didn't know that this was such an extensive behaviour and it would have been extremely time-consuming to analyse with traditional microscopy,” says Karin Garefelt , a PhD student working on programming AI networks for plankton analysis.

When will we be able to say something more certain about the impact of changes in marine plankton on the ecosystem?

“Hopefully, within a decade, enough data will have been collected on the ocean food web down to the plankton level. With such mapping as a basis, it will be possible to build models of the ecosystem and calculate the consequences for birds and fish, for example”, Andersson says.

Text: Christer Gummeson

The research in brief

  • The camera instrument, Imaging Flow Cytobot, works by sending an extremely thin jet of water past a microscope lens. When a particle, such as a phytoplankton, passes through the lens, a high-resolution image is automatically taken. The instrument analyses 2 ml of water at a time and can take several thousand plankton images per water sample.
  • The images are analysed using AI, a deep neural network trained to recognise plankton species. For an experienced taxonomist, a person who studies and classifies organisms, it takes about four hours to analyse a plankton sample using microscopy. With the Imaging Flow Cytobot combined with AI, three samples can be analysed every hour, around the clock.
  • The research project ‘AMIME’ is funded by FORMAS and is a collaboration between Anders Andersson, SciLifeLab/KTH, Josephine Sullivan, EECS/KTH, and Bengt Karlson, Oceanography/SMHI.

AI and humans in focus

How can AI contribute to people's lives and the development of society? What are the opportunities, challenges and concerns? A series of articles focusing on artificial intelligence and humans presents exciting research from KTH Royal Institute of Technology.

AI could be your next recruiter

Neuromorphic computer systems challenge AI technology

How to reduce the energy use of buildings

"An enormous potential - but AI is not always right"

Using AI to fight youth gang recruitment

AI virtual human cells: shaping the future of biology and human health

The ocean's smallest creature is mapped

The enzymes that can reduce emissions