ODYSSEA contributes to research; will benefit from data produced by machine learning

The EU-funded ODYSSEA project’s marine observation platform has contributed to research on seagrass detection in the Mediterranean, and in the process helped developing a machine learning (ML) model, a form of artificial intelligence, which will in turn provide data to be integrated into the ODYSSEA platform, said ODYSSEA coordinator Prof. Georgios Sylaios of the Democritus University of Thrace (DUTH).

“The ODYSSEA platform is a unique system containing all data, both static – like the bottom topography, habitat types, sediment characteristics and seagrass distributions, and dynamic – like the changes in oceanographic and environmental parameters, such as salinity, temperature, dissolved oxygen, chlorophyll-a and many more,” Sylaios explained.

“This creates an environment of ‘Big Data’ that are easily retrieved, combined and exploited to reach conclusions. Developing the data representing the oceanographic and environmental conditions in which the  seagrass species live in the Mediterranean, utilising machine learning and other advanced tools, to exploit the factors affecting their presence or absence and the species’ overall distribution at the sea bottom is a major breakthrough for scientists. Advanced scenarios may be built to lead us to conclusions related to the impact of climate change on future seagrass species distribution,” he added.

Word of the research and its use of the machine learning model was posted on the website of the European Marine Observation and Data Network (EMODnet).

Noting the DUTH team’s involvement in the ODYSSEA project, EMODnet described how DUTH researchers provided accurate forecasts, assessed wave climate, described extreme conditions and defined wave erosion hotspots, by simultaneously using hydrodynamic model results, EMODnet habitat-type data and wave and water quality data produced by Copernicus satellite image analysis in relation to shoreline changes. The goal of the research was to investigate the influence of environmental conditions on the presence or absence of seagrass species, and their distribution over the Mediterranean Sea.

“To achieve this scope, DUTH leveraged on advanced machine learning techniques and combined a great number of morphodynamic, environmental and human impact variables, […] perform[ing] comparative tests between the ML results when using the random forest and maximum entropy algorithms,” EMODnet said.

EMODnet emphasised researchers’ plans to continue using EMODnet data in the future to examine the abundance and diversity of fish species in relation to environmental variables, “producing a significant impact to the society.”

The DUTH research, carried out by Dimitrios Effrosynidis, Avi Arampatzis and Georgios Sylaios, was titled “Seagrass Detection in the Mediterranean: A Supervised Learning Approach” and published in Ecological Informatics, Volume 48. Their results suggest that the primary environmental factors affecting the distribution of seagrass are winter chlorophyll-α and salinity levels (especially in December), autumn phosphate concentrations and bathymetry, along with the seasonal changes in temperature and light are the main factors controlling the distribution of seagrass species in the benthic environment of the Mediterranean.