Integrating temporal decomposition and data-driven approaches for predicting coastal harmful algal blooms

Area of Science:

• Marine Biology
• Environmental Science
• Data Science

Background:

• Harmful algal blooms (HABs) pose significant threats to coastal ecosystems and human health.
• Biscayne Bay experiences HABs, but the underlying mechanisms require further investigation for effective management.
• Accurate HAB prediction is crucial for environmental protection and public health safety.

Purpose of the Study:

• To develop a robust predictive framework for chlorophyll-a concentrations, an indicator of HABs, in Biscayne Bay.
• To enhance the understanding of HAB dynamics and contributing environmental factors.
• To compare the predictive performance of different machine learning approaches for HAB forecasting.

Main Methods:

• Three predictive scenarios were developed: single nonlinear machine learning (S1), hybrid linear and nonlinear ML (S2), and temporal decomposition combined with ML (TD-ML) (S3).
• Chlorophyll-a concentrations were used as the representative metric for HAB prediction.
• Correlation analysis was employed to identify relationships between environmental variables and chlorophyll-a levels.

Main Results:

• The novel S3 TD-ML hybrid models demonstrated superior predictive accuracy, achieving R² values above 0.9 and Mean Absolute Percentage Error (MAPE) under 30%.
• S3 models significantly outperformed S1 (average R² of 0.16) and S2 (R² of -0.06), effectively capturing complex algal dynamics, including extremes and noise.
• Correlation analysis revealed key environmental drivers of HABs and suggested that climate change may exacerbate bloom intensity in Biscayne Bay.

Conclusions:

• The developed TD-ML framework provides a highly accurate and precise predictive tool for early warning and proactive management of HABs.
• This research offers valuable insights into HAB dynamics and highlights the potential impact of climate change on bloom frequency and intensity.
• The predictive framework has potential for global applicability in addressing HAB challenges and improving coastal environmental management.