MICROCYSTIN IN AGRICULTURAL PONDS

Abstract

Microcystin is a common cyanotoxin of concern in freshwater systems due to its impacts on water quality and on animal and human health. However, its dynamics in agricultural irrigation and livestock watering ponds remain poorly characterized. This study aimed to characterize the spatiotemporal variability and predictability of microcystin in small agricultural ponds using in situ measurements, high-throughput imaging flow cytometry, and machine learning approaches. Spatial and temporal patterns of microcystin were assessed in three ponds in Southeastern Georgia, USA over a 17-month period. Microcystin levels were relatively stable and showed strong to moderate correlations to water quality parameters such as photosynthetic pigments and turbidity. Imaging flow cytometry combined with machine learning enabled detailed characterization of a rare winter Microcystis bloom and demonstrated that microcystin concentrations may be reliably predicted from cyanobacteria community composition. Findings also indicated that cyanobacteria blooms can produce elevated toxin concentrations outside of the typical monitoring window of late spring through fall. To further quantify the relationships between microcystin concentrations and environmental variables, five machine learning algorithms were tested using in situ water quality measurements. Among these, the random forest model performed best. Overall, this work highlights the heterogeneous and seasonal nature of microcystin dynamics in small agricultural ponds. Integrating spatial analysis, imaging flow cytometry technologies, and machine learning can provide a practical and adaptable framework for monitoring and predicting cyanotoxin risks in agricultural waters, offering a cost-effective approach for early detection and improved protection of livestock, crops, and human health.