Fusing remote sensing data with spatiotemporal in situ samples for red tide (Karenia brevis) detection

Area of Science:

• Environmental Science
• Oceanography
• Remote Sensing

Background:

• Red tide blooms, particularly those caused by Karenia brevis, pose significant ecological and economic challenges.
• Effective monitoring of these blooms is crucial for mitigation and public safety.

Purpose of the Study:

• To develop and validate a novel, high-resolution method for detecting Karenia brevis blooms.
• To improve the accuracy and efficiency of red tide monitoring systems.

Main Methods:

• A neural network classifier integrating MODIS-Aqua satellite data (2002-2021) and in situ sample data.
• Incorporation of depth normalization for satellite features and a K-nearest neighbor spatiotemporal weighting scheme for in situ data.
• 1-km grid resolution bloom detection.

Main Results:

• The developed model significantly outperforms existing remote detection methods.
• Key innovations include depth normalization and an engineered in situ feature for enhanced accuracy.
• Demonstrated effectiveness in detecting Karenia brevis blooms off the west coast of Florida.

Conclusions:

• The novel classifier shows strong potential for operationalization in red tide monitoring and mitigation.
• This approach can lead to more accurate communication regarding bloom extent and distribution.
• The method is adaptable for detecting other harmful algal blooms in coastal environments.