Identifying Key Taxa for Algal Blooms in a Large Aquatic Ecosystem through Machine Learning.

Area of Science:

• Environmental Science
• Ecology
• Machine Learning

Background:

• Algal blooms (ABs) pose significant ecological challenges, necessitating effective monitoring.
• Current methods for tracking ABs in large water bodies lack taxonomic detail and spatial coverage.
• Identifying key algal species is vital for targeted AB management.

Purpose of the Study:

• To develop an integrated approach for identifying and mapping key algal bloom species.
• To overcome limitations of traditional biomonitoring in large aquatic ecosystems.
• To provide a framework for precise, large-scale simulation of algal dynamics.

Main Methods:

• Supervised machine learning (ML) integrating environmental DNA (eDNA) metabarcoding, remote sensing, and water quality data.
• Gradient boosting tree model for predictive accuracy assessment.
• Principal component regression for quantifying algal taxa contributions to community structure.

Main Results:

• The ML model achieved high predictive accuracy (mean MAPE 11.20%) for 34 algal taxa.
• Spatial distribution maps were generated and validated against morphological data (75% significant correlation).
• Nostocales and Stephanodiscales identified as key drivers of floating algae index (FAI) variations in Poyang Lake.

Conclusions:

• The study presents a novel, accurate framework for species-level algal bloom monitoring and spatial simulation.
• This approach significantly advances the precision and comprehensiveness of AB management strategies.
• The toxic alga Nostocales was identified as a major contributor to FAI in the northern region of Poyang Lake.