logo
Pioneers

Machine learning beneath the sea

Understanding the impact of climate change

All around us, it seems like things are changing too fast to keep track of. From social media bombardments, constantly updating news cycles, and daily reminders about the threat of climate change, usually accompanied by images that are hard to look away from. Temperatures are rising. Ice caps are melting. Ecosystems everywhere seem to be threatened by local impacts and global climate change.

Pixabay

Climate change illustrated

It is all so overwhelming that by the time we look away from our screens, we are so consumed by the big picture impacts that we overlook the smaller – yet not less important - parts of our planet that are also affected by climate change. The parts of planet that are both figuratively, and literally, hidden from us.

In many ways, it's these tiny, invisible worlds that keep our ecosystems - and us as humans that rely on them - moving along and functioning. But they're hard to see and even harder to study. It's what keeps these tiny organisms, that form our planet's foundation, largely a mystery to scientists who have spent decades studying them.

Phyto... what?

This is exactly the case for phytoplankton, tiny plant-like organisms that photosynthesize in the ocean and produce most of Earth's newly supplied oxygen. Most of them aren't strong swimmers, and they simply drift along ocean currents. Phytoplankton make up the very bottom of marine food web, meaning that the health of everything else - from fish, octopus, squids, and mammals like dolphins, otters, and whales - all directly or indirectly depend on them.

Pixabay

Phytoplankton

In the past, tracking how these organisms change has been extraordinarily difficult. First the plankton were captured by passing water through small filters. Then tens to thousands of them would be manually counted underneath a microscope by a trained professional who spent years learning to correctly identify them. In the 1930s scientists managed to automatically capture plankton over long oceanic distances by equipment dragging behind volunteer merchant and research vessels. However, even with these technological advances the plankton would still have to be processed in a lab under a microscope.

https://tinyurl.com/yxjk93eo

A Continuous Plankton Recorder

Making the invisible visible

This type of research formed the foundation of incredibly important ocean science, but its cost and labor intensity limited researchers' ability to assess phytoplankton communities in detail. For example how the communities responded to long-term ecological changes like global warming, seasonality, and even day-to-day differences driven by weather, precipitation, or ocean currents.

In other words, their size made most patterns in phytoplankton communities hidden from researchers. All this has started to change with a new wave of ocean technology that has been combined with advances in machine learning. Opening up new possibilities and finally allowing ocean scientists to make what was once invisible to us, visible.

Big data becoming meaningful

Scripps Plankton Camera, built by the Jaffe Lab, anchored to a pier piling at the Scripps Institution of Oceanography collects tens of thousands of images every day. Image courtesy of the Jaffe Lab.

Scientists at the Scripps Institution of Oceanography in La Jolla, California are using new underwater microscopes, that can spend weeks below the surface with little maintenance. The system captures images of particles floating pass by the camera, including phytoplankton. They are imaged at an impressive rate of up to 8 pictures per second.

"Over the last four years, the camera collected nearly 1 billion images of suspended particles drifting onto the microscope. Collection of such a huge amount of data, as exciting as it is to any scientist, comes with a lot of challenges related to data storage and processing," said Dr. Kasia Kenitz, a postdoctoral researcher at the Scripps Institution of Oceanography working on the ecology of phytoplankton communities in the ocean.

"The next challenge is identifying which organisms are depicted in every single image and how can we use this information in a way that will provide meaningful insights into the ecology of these tiny organisms. This is where recent advances in computer vision and machine learning become an integral part in collecting biologically meaningful data."

The Ellen Browning Scripps Memorial Pier hosts an abundance of scientific research equipment to collect data and help solve mysteries of our planet's processes. Image by Rishi Sugla.

Perhaps it is not as bad as looking at a billion samples of phytoplankton slides underneath a microscope, but looking at a billion of images on your computer by hand is still next to impossible. Researchers like Dr. Kenitz have started using machine learning to train image classifiers to automatically identify subgroups of different phytoplankton.

The underwater images of phytoplankton highlight the complex forms of phytoplankton and how beautiful and diverse the invisible can be. This also represents a challenge for automatic object identification since their unique forms are unlike anything else we see in the visible world. Dr. Kenitz and her colleagues have manually inspected tens of thousands of these images to train classifiers to correctly recognize phytoplankton species and calculate their relative abundance through time.

In other words, these methods are revolutionizing our ability to see how these species, which influence every other living creature in the ocean, may change in years to come.

The diversity of phytoplankton. (Images taken by the Scripps Plankton Camera courtesy of the Jaffe Lab).

​So what comes next?

Dr. Kenitz expects that this line of research will become increasingly relevant to humans and how they interact with oceans. "These new underwater technologies have revolutionized the way we observe these tiny-but-mighty organisms. It's especially important to start using these tools for coastal populations that rely on seafood, as uncontrolled growth of some phytoplankton can produce toxic substances, generate huge economic losses and lead to public health issues. We're now working with a community of experts in plankton identification to help us build better, cost-effective ecosystem monitoring."

In the end, this new technology helps us remember that some of the most spectacular parts of our planet are hidden from us. These images, taken in a way previously impossible, help us understand that we can't afford to forget about the people, plants, and tiny organisms living in worlds we aren't able to see.

Even in the midst of what sometimes feel chaotic, we all have a chance to understand the effects of the smallest change in our world just by staying curious.

More about Rishi Sugla you can read here or follow him on Instagram.