NASA Develops AI Tool to Detect Harmful Algal Blooms

NASA Develops AI Tool to Detect Harmful Algal Blooms

NASA scientists have developed an artificial intelligence tool to combat a longstanding challenge in ocean waters: harmful algal blooms (HABs). The new tool, published in AGU Earth and Space Science, successfully fuses data from multiple satellites to detect these severe blooms that occur in western Florida and Southern California. Severe HABs can pose significant health risks and cost coastal economies tens of millions of dollars annually.

Health Risks and Economic Impact

Harmful algal blooms can have devastating effects on marine ecosystems, human health, and local economies. In the Gulf of America, a species called Karenia brevis thrives, causing harmful algal blooms that kill wildlife, foul beaches, and sicken swimmers. On the West Coast, Pseudo-nitzschia blooms have poisoned hundreds of dolphins, California sea lions, and other marine animals in recent years. Toxins from algae can even enter the air, causing respiratory illnesses in humans.

To manage these risks, health agencies regularly test waters and issue warnings or beach closures when necessary. The National Oceanic and Atmospheric Administration (NOAA) collaborates with states and local partners to provide harmful algal bloom forecasts during bloom seasons. On-site testing requires hours of manual water sample collection that must be sent to a lab for analysis, often taking days or longer.

NASA’s Role in Monitoring Algal Blooms

NASA’s Earth-orbiting satellites already track harmful algal blooms with their unique global view. By integrating diverse datasets, the new AI tool could significantly enhance monitoring efforts and serve as a force multiplier to help communities determine where to focus their efforts.

For instance, a hyperspectral sensor aboard NASA’s Plankton, Aerosol, Cloud, ocean Ecosystem (PACE) satellite can identify algal communities by their size, shape, and pigment. Other instruments like TROPOMI (Tropospheric Monitoring Instrument) pick up on the faint red glow emitted by species such as K. brevis during photosynthesis.

Developing the AI Tool

The study team, consisting of Michelle Gierach from NASA’s Jet Propulsion Laboratory in Southern California, Kelly Luis from NASA JPL, and Nick LaHaye from Spatial Informatics Group, combined findings from five space missions or instruments, including PACE and TROPOMI. The challenge they faced was the massive quantity of raw data involved.

The team developed a self-supervised machine learning system designed to learn patterns from multiple kinds of satellite data and compare them with field observations. This approach enables AI to recognize relationships between different data sources without needing any labeling in advance.

The system was trained on satellite data collected in 2018 and 2019, and field and lab measurements were used to add real-world context to the patterns that the system recognized. The scientists evaluated the tool’s performance across later time periods in the same geographic areas. Initial results indicate that it can correctly identify and map harmful blooms, including specific species like K. brevis, even in complex coastal waters swirling with sediment, plants, and runoff.

Potential Benefits of the AI Tool

Applying self-supervised AI to massive streams of satellite data is rapidly becoming a powerful tool for generating actionable ocean intelligence. This AI tool could significantly enhance early detection and monitoring efforts by identifying where and when to collect water samples as an algal bloom starts spreading.

The system can also drive collaboration between specialists, fostering new ways to conduct the science and deliver decision-support products. By providing real-time data on harmful algal blooms, this AI tool can help coastal communities better prepare for and mitigate these environmental challenges.

Conclusion

NASA’s development of an AI tool to detect harmful algal blooms represents a significant step forward in ocean monitoring and management. By integrating diverse satellite datasets, the self-supervised machine learning system can provide actionable intelligence on these harmful events, ultimately helping to protect marine ecosystems, public health, and coastal economies.