- NASA developed an AI tool to detect harmful algal blooms using multiple satellite data
- The system identified blooms in Florida and Southern California accurately from space
- Harmful blooms threaten marine life, human health, and cause economic losses annually
Scientists at NASA have developed a new artificial intelligence tool that can help detect harmful algal blooms in ocean waters by combining data from multiple satellites. Researchers said the system successfully identified blooms in western Florida and Southern California using satellite observations, reported NASA.
The findings were published in AGU Earth and Space Science. Harmful algal blooms are a major problem in several coastal areas of the United States. Severe blooms can harm marine life, affect public health and cause economic losses worth tens of millions of dollars every year.
In parts of Florida, including Tampa Bay and Sarasota, a species known as Karenia brevis has caused harmful blooms for decades. These blooms can kill wildlife, pollute beaches and make swimmers sick.
On the West Coast, blooms of Pseudo-nitzschia have poisoned dolphins, California sea lions and other marine animals in recent years. Scientists said toxins from algae can also enter the air and lead to respiratory illness in humans.
To reduce risks, health agencies regularly collect water samples and issue warnings or close beaches when necessary. The National Oceanic and Atmospheric Administration works with states and local agencies to release harmful algal bloom forecasts during bloom seasons.
However, collecting water samples requires hours of fieldwork by boat, followed by laboratory testing that can take a day or more. Researchers also said it is difficult to know where to test before blooms begin spreading. NASA's Earth-observing satellites already monitor harmful algal blooms from space. Scientists said the new AI system improves this process by combining information from different satellites and instruments.
Michelle Gierach, a scientist at NASA Jet Propulsion Laboratory and coauthor of the study, said the tool could help scientists decide where and when to collect water samples as a bloom begins. She also said the system could encourage collaboration among specialists and support the development of better scientific and decision-making tools.
The AI system was developed by Gierach, Kelly Luis of NASA JPL, and research data scientist Nick LaHaye of Spatial Informatics Group.
The team combined information from five space missions and instruments, including NASA's PACE satellite and the TROPOMI instrument.
Researchers said one challenge was handling the huge amount of raw satellite data and teaching the AI system to recognise harmful blooms across different environments and data streams.
To solve this, the team created a self-supervised machine learning system that learned patterns from different kinds of satellite observations and compared them with field measurements.
The system was trained using satellite data collected in 2018 and 2019. Scientists later tested the tool in the same regions during later periods and found that it was able to correctly identify and map harmful blooms, including specific species such as Karenia brevis.
Researchers said the tool performed well even in difficult coastal waters filled with sediment, plants and runoff.
Nadya Vinogradova Shiffer, lead program scientist at NASA Headquarters in Washington, said applying self-supervised AI to large streams of satellite data was becoming a powerful method for generating useful ocean intelligence.
