Harnessing stability-guided interpretable machine learning for understanding and predicting water quality in freshwater ecosystems

This project uses machine learning and explainable AI (XAI) to predict and understand the phytoplankton dynamics in the Falling Creek and Beaverdam Reservoirs in Virginia, USA. By analyzing environmental variables collected at the sites, we uncover the hidden patterns driving phytoplankton dynamics. Further we introduce XAI-derived feature attribution stability analyses to quantify and demonstrate the reliability of ML modelling workflows for water quality management. Note that this analysis is performed using Python version 3.12.11 and running the full workflow may take an hour or longer, depending on your system.