Highlights

• Scientists have successfully deployed an Imaging FlowCytobot (IFCB) for continuous, real-time monitoring and identification of algal species in subtropical coastal waters. It is a significant improvement on traditional manual methods, enabling real-time monitoring of toxic algal bloom at species level.  It is the first time IFCB is used in Greater China, and the first time for fisheries management globally.
• This study applies a Random Forest (RF) classifier (a machine learning technique) to identify 15 target harmful algal bloom (HAB) species. With a high accuracy of 94.2%, this classifier (using 25 optimized features) matches the performance of Convolutional Neural Networks (CNNs) developed with transfer learning.
• The integration of the automated classification system with existing HAB early warning systems enhances response capabilities to HAB events, marking a breakthrough in fisheries management and environmental monitoring in Hong Kong’s coastal waters and internationally.

Summary

This research introduces an effective method for managing harmful algal blooms (HABs) in the subtropical coastal waters of Hong Kong (with special reference to marine fish culture zones).

Central to this advancement is the development of an automated real-time system for the identification and concentration measurement of algal bloom species. This system utilizes an Imaging FlowCytobot (IFCB), which is a great improvement over traditional labor-intensive time-consuming manual microscopic methods – enabling real time monitoring and identification of toxic species.

The research is based on the use of machine learning, specifically employing a Random Forest algorithm. This algorithm is optimized for high accuracy and is capable of identifying 15 distinct HAB species. It demonstrates a performance level on par with more complex Convolutional Neural Networks (CNNs).

The integration of AI and advanced real time laser-based high frequency in-situ measurements   represents a significant leap in environmental monitoring techniques. Furthermore, the study enhances existing early warning systems for HAB events, offering a more efficient tool for fisheries management. This integration is crucial in the eutrophic, nutrient-rich waters around Hong Kong. It represents a significant technological advance in monitoring and managing HABs. Other than being vital for protecting marine ecosystems, it also enhances the sustainability of the fisheries industry.

Guo, Y. Ma, and J. H. W. Lee, “Real-time automated identification of algal bloom species for fisheries management in subtropical coastal waters,” Journal of Hydro-environment Research, vol. 36, pp. 1–32, May 2021, doi: .