Microcystis abundance is predictable through ambient bacterial communities: A data-oriented approach.

Area of Science:

• Environmental microbiology
• Aquatic ecology
• Machine learning applications in environmental science

Background:

• Increasing prevalence of cyanobacterial harmful algal blooms (cyanoHABs) necessitates improved prediction models.
• Existing cyanoHAB models primarily rely on environmental data, often overlooking the significant influence of bacterial communities.
• Understanding microbial interactions is key to advancing cyanoHAB prediction accuracy.

Purpose of the Study:

• To develop and evaluate a machine learning model for predicting Microcystis dynamics using both bacterial community and environmental data.
• To assess the relative importance of bacterial communities versus environmental factors in cyanoHAB prediction.
• To identify key microbial players and environmental drivers influencing Microcystis blooms.

Main Methods:

• Utilized a multilayer perceptron (MLP) machine learning model for prediction.
• Integrated weekly water quality data with bacterial community data from Daechung Reservoir and Nakdong River.
• Performed post-hoc analysis to determine the contribution of different factors and identified key microbial taxa.

Main Results:

• The MLP model combining bacterial and environmental data achieved a higher R² (0.97) compared to environmental data alone (0.78).
• Nitrogen sources were identified as more critical than phosphorus for Microcystis blooms.
• The bacterial family Microscillaceae showed the strongest association with Microcystis dynamics in both MLP and MLR models.

Conclusions:

• Bacterial community data significantly enhances the predictive power of cyanoHAB models.
• Microbial community composition is a critical factor for understanding and predicting cyanoHABs.
• This approach offers a pathway to more accurate and reliable cyanoHAB prediction by incorporating ambient bacterial data.