Algal community structure prediction by machine learning.

Area of Science:

• Ecology
• Environmental Science
• Machine Learning

Background:

• Algal community structure is crucial for aquatic ecosystem management.
• Modeling complex environmental and biological interactions in aquatic systems presents significant challenges.

Purpose of the Study:

• To investigate the efficacy of random forests (RF) in predicting phytoplankton community shifts.
• To identify key environmental drivers influencing algal community dynamics using multi-source data.

Main Methods:

• Utilized random forests (RF) models to predict phytoplankton community structure and total biomass.
• Incorporated multi-source environmental factors: physicochemical, hydrological, and meteorological variables.
• Performed importance analysis and ecological interpretation of RF model outputs.

Main Results:

• RF models demonstrated robust predictions for 13 major algal classes (Bray-Curtis dissimilarity = 9.2 ± 7.0%) and total biomass (validation R² > 0.74).
• Hydro-meteorological variables were identified as the most influential factors regulating phytoplankton communities.
• Ecological interpretation revealed interactive stress-response effects of environmental drivers (temperature, inflow, nutrients) on algal shifts.

Conclusions:

• Machine learning, specifically RF, offers a powerful approach for predicting complex algal community structures.
• Understanding the interplay of environmental drivers is essential for managing and predicting aquatic ecosystem changes.
• The study provides valuable insights into model interpretability for ecological applications.