A novel machine learning approach for assessing water quality using Daphnia swimming behavior

To bridge the gap between complex water chemistry data and public understanding, we developed a novel Daphnia magna biomonitoring approach combined with deep learning to assess water quality. We recorded Daphnia swimming behavior under different salinity levels (0, 0.01, 0.1, 1, 10%) and extracted key movement features. A Multi-Layer Perceptron (MLP) model was then trained on these features to classify water quality levels. The MLP achieved an overall accuracy of ~ 80%, with class-wise precision, recall, and F1-scores generally ranging from ~ 0.7 to 0.9. Behavioral analysis showed that Daphnia’s average swimming speed and acceleration decreased at higher salinities (p < 0.01), indicating sensitivity to pollutant stress. This study is the first to integrate video-based Daphnia monitoring with machine learning for water quality assessment, demonstrating a simple, low-cost method to empower citizen scientists and enhance environmental education.