Towards precision limnology: An explainable AI framework decoding spatiotemporal algal dynamics in Chinese major...

Area of Science:

• Environmental Science
• Limnology
• Machine Learning Applications

Background:

• Lake ecosystems face increasing threats from eutrophication and harmful algal blooms.
• Climate change and human activities exacerbate these challenges, impacting water resource management.

Purpose of the Study:

• To develop an interpretable machine learning framework for analyzing algal cell density (AD) dynamics.
• To identify key drivers of algal blooms in 21 major Chinese lakes.
• To provide a tool for evidence-based lake management.

Main Methods:

• Integrated high-resolution water quality and meteorological data (2021-2024).
• Developed lake-specific predictive models using interpretable machine learning.
• Applied automated feature selection and explainable AI (XAI) with nonlinear modeling.

Main Results:

• Achieved high predictive accuracy (mean R² > 0.8) with lake-specific models.
• Reduced input variables by 50% via feature selection without significant performance loss.
• Uncovered spatially varying driver-response relationships and critical environmental thresholds.

Conclusions:

• Model structures and key drivers of algal blooms vary significantly among lakes, necessitating tailored management strategies.
• The XAI framework aids in optimizing monitoring and targeting algal bloom mitigation efforts.
• Provides practical tools for sustainable lake management in dynamic environments.