Prediction of Aureococcus anophageffens using machine learning and deep learning.

Area of Science:

• Marine Ecology and Harmful Algal Blooms

Background:

• The Qinhuangdao sea area faces ecological degradation and economic losses due to recurrent brown tides.
• Aureococcus anophagefferens (A. anophagefferens) is the causative agent of these harmful algal blooms.

Purpose of the Study:

• To predict A. anophagefferens population density using machine learning and deep learning.
• To understand the occurrence mechanisms and influencing factors of brown tide events.

Main Methods:

• Random Forest (RF) algorithm for imputing missing water quality data.
• Machine learning models: RF, Support Vector Regression (SVR), Multilayer Perceptron (MLP).
• Deep learning model: Convolutional Neural Network (CNN) for algal population prediction.

Main Results:

• All tested models (RF, SVR, MLP, CNN) achieved high prediction accuracy (R² > 0.75).
• RF demonstrated exceptional predictive performance (R² > 0.8).
• Identified key factors influencing A. anophagefferens density: ammonia nitrogen, pH, total nitrogen, temperature, and silicate.

Conclusions:

• Machine and deep learning models are effective for predicting A. anophagefferens blooms.
• Water quality parameters like ammonia nitrogen, pH, total nitrogen, temperature, and silicate are critical drivers of brown tide occurrence.