Exploring spatiotemporal patterns of algal cell density in lake Dianchi with explainable machine learning

Area of Science:

• Environmental Science
• Ecology
• Water Resource Management

Background:

• Algal blooms are increasing globally, threatening ecosystem services.
• Lake Dianchi faces significant challenges from water quality degradation and algal blooms.

Purpose of the Study:

• To predict daily algal density (AD) in Lake Dianchi using water quality and meteorological data.
• To identify key drivers and their effects on algal blooms across different water quality zones.
• To develop a cost-effective model for sustainable water quality management.

Main Methods:

• Partitioning Lake Dianchi into three clusters based on spatiotemporal water quality heterogeneity.
• Employing ensemble learning and quasi-Monte Carlo methods for predictive modeling.
• Utilizing the Stacking-Elastic-Net regularization model for superior predictive accuracy.

Main Results:

• The Stacking-Elastic-Net model demonstrated high predictive accuracy for algal density (AD) across all clusters.
• The delayed effects of meteorological factors were generally more influential on AD than instantaneous effects.
• A minimum set of drivers for near-optimal accuracy was identified for each cluster, balancing cost and precision.

Conclusions:

• The study provides a scientific basis for understanding and managing algal blooms in Lake Dianchi.
• Findings support the development of targeted, cost-effective regional strategies for water quality improvement.
• Understanding driver-response relationships is crucial for effective ecosystem service protection.