A real time data driven algal bloom risk forecast system for mariculture management

Area of Science:

• Marine Ecology
• Oceanography
• Environmental Science

Background:

• Harmful algal blooms (HABs) are a recurring challenge in eutrophic coastal waters, impacting environmental and fisheries management.
• Existing HAB early warning systems lack sufficient field validation, hindering their practical application.

Purpose of the Study:

• To develop and validate a daily algal bloom risk forecast system for eutrophic coastal environments.
• To improve the reliability and applicability of HAB prediction models through rigorous field data validation.

Main Methods:

• Utilized a vertical stability theory, validated against 191 historical HAB events.
• Developed a data-driven artificial neural network (ANN) model to assimilate high-frequency data.
• Predicted sea surface temperature (SST) and vertical temperature/salinity differentials with high accuracy (0.35°C, 0.51°C, 0.58 PSU).

Main Results:

• The ANN model demonstrated accurate predictions for key environmental parameters.
• The forecast system was validated against extensive field data, proving its robustness.
• Operational forecasts were successfully demonstrated for HAB events in Tolo Harbour, Hong Kong.

Conclusions:

• The proposed daily algal bloom risk forecast system offers a reliable tool for environmental and fisheries management.
• The model's independence from chlorophyll measurements and its AI-based approach represent a significant advancement.
• This system complements traditional monitoring and other AI-based methods for HAB management.