AI presents a new tool for remote monitoring of global coral health

A new conservation tool in the field of coral reef ecology has been developed by University of Hawaii (UH) researchers in Mānoa using state-of-the-art artificial intelligence (AI) technology. By developing a new deep learning algorithm, coral ecologists at UH Mānoa School of Ocean and Earth Sciences and Technology (SOEST) can now identify and measure coral halo from outer space.

These features, also known as grazing halos or sand halos, consist of ring-like patterns of bare sand that occur around coral reefs, and are easily visible from satellite imagery.

“Reef halos may be important indicators of coral reef health and vitality, but until now, measuring and tracking them has been a challenging and time-consuming process,” said Simon Franceschinilead author of the study and postdoctoral research fellow at Medina Laboratory on Hawaii Institute of Marine Biology (HIMB) at SOEST. “However, with this new method, we can accurately identify and measure coral halos on a global scale in the short time it takes humans to complete the same task.”

“We aim to develop freely available remote sensing tools to monitor ecological processes at large scale to improve understanding and management of coral reef ecosystems,” said Elizabeth Madin, senior author of the study and associate research professor at HIMB. “Our current research shows that halo reefs may represent an emerging opportunity for monitoring reef ecosystem function at a large scale, including in remote and inaccessible areas.”

In recent years, computer vision techniques have been increasingly used to recognize patterns in medical and biological studies. In ecology, the application of image analysis coupled with advances in satellite imaging technology has enhanced large-scale ecosystem analysis and wildlife conservation.

“This work stems from our team’s understanding of the current state of AI technology and its potential application for conservation research in coral reef ecosystems,” added Madin.

Although AI technologies have demonstrated excellent performance in the field of image analysis, identifying halos – complex ecological patterns with many variations – is a challenge requiring the incorporation of multiple deep learning algorithms.

“Coral halos are sometimes very clear in satellite images, with distinct edges and high contrast to the background vegetation, but sometimes very faint and difficult to distinguish – even by highly trained observers,” says Franceschini. “Ultimately, our team was able to develop a suite of algorithms that are able to consider this diversity of patterns globally and identify and measure halos with surprising accuracy. It is very gratifying for us to now have built something that can accurately identify more than 90% of halos in some parts of the world.”

Coral reefs, one of the most diverse ecosystems on the planet, on which millions of people globally depend, are under threat from overfishing, climate change and many other factors. These ecosystems, and in particular the impact of fisheries and marine reserves on them, are extremely difficult to monitor on a large scale and over time.

“This breakthrough is a key step in improving – both in space and time – our ability to monitor and measure aspects of the health of coral reef ecosystems,” said Madin. “By providing a more efficient and effective way to measure coral reefs and their surrounding halos, this new method paves the way for the development of global-scale coral reef monitoring and conservation tools based on the reef halo phenomenon.”

The team aims to develop, in the near future, a freely available web application that will enable conservation practitioners, scientists, and resource managers to monitor aspects of coral reef health remotely, quickly, and inexpensively using satellite or drone imagery.