Scientists mimicked the neural networks of the brain to map phytoplankton types in the Mediterranean Sea. A new study published in the Journal of Geophysical Research: Oceans presented a new method of classifying phytoplankton that relies on artificial intelligence clustering.
Phytoplankton blanket surface waters of the world’s oceans, and pigments in their cells absorb certain wavelengths of light, like the chlorophyll that gives plants their green color. Viewed from space, the color of the ocean’s surface changes depending on the phytoplankton growing there. In the Mediterranean Sea, where the latest study focused its efforts, an array of phytoplankton species bloom throughout the year.
Past research has mined satellite images of ocean color in the Mediterranean for common pigments found in phytoplankton. A combination of pigments can reveal a certain type of dominant phytoplankton in the area, like certain species of diatoms that can be spotted because of their unique orange pigment, fucoxanthin. But connecting the complex relationships between satellite image pixels, pigments, and phytoplankton types can make for a tricky analysis.
The latest study turns to artificial intelligence to parse through the multidimensional data. The process mimics the brain’s ability to take in new information and learn over time, giving the algorithm a chance to identify relationships in the data that may not be readily apparent. The algorithms cluster similar nodes of information near one another, creating a two-dimensional diagram called a “self-organized map.” The scientists trained two algorithms used in the study with 3 million pixels from satellite images and over a thousand measurements taken by boat in the Mediterranean.
The results show six types of phytoplankton and how they come and go by season. In winter, haptophytes and chlorophytes (both algae) are common in the western Mediterranean. In the summer months, the most abundant photosynthetic organism on Earth, the cyanobacteria Prochlorococcus, rules broad swaths of the sea. The new method revealed how the blooms changed over time, giving the scientists a way to ask questions about marine food chains and possible effects of climate change in the future.
The scientists called the new method “very general” in their paper and said that it could be applied elsewhere in the world’s oceans.
—Jenessa Duncombe (@jrdscience), News Writing and Production Fellow
Duncombe, J. (2019), Artificial intelligence can spot plankton from space, Eos, 100, https://doi.org/10.1029/2019EO132391. Published on 06 September 2019.
Text © 2019. AGU. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.