Causal discovery of hydroclimatic drivers influencing water quality in a large lake

Abstract Harmful algal blooms (HABs) are a threat to ecosystem services, with adverse economic and public health impacts. Large-scale climate processes will influence local environmental conditions, potentially favoring HAB formation through complex, nonlinear interactions. This study employs explainable (i.e., SHAP) machine learning to give insight on the predictions and causal analysis (i.e., PCMCI) to identify the relationships between climate indices, physical drivers, and the resultant Chlorophyll-a (CHL) concentrations in western Lake Erie. Our causal analysis revealed that runoff and water temperature directly affect CHL but also act to mediate the impacts of the Arctic Oscillation on influencing CHL. Moreover, our explainable analysis further confirmed this by identifying runoff as the main driving factor, followed by water temperature. The study highlights that water quality in the basin is subject to confounding effects resulting from interactions between global atmospheric circulation patterns and local hydro-meteorological factors, expanding HAB forecasting beyond synoptic scale meteorology.