Applying Time-Series Causal Discovery to Understand Algal Bloom Mechanisms

Harmful algal blooms present a significant ecological and public health challenge, yet their drivers are embedded in complex, time-lagged systems with unobserved confounding variables. Traditional correlational methods often fail to uncover the true causal structure from observational time-series data. In this work, we apply a suite of modern constraint-based causal discovery algorithms to a real-world dataset of daily water quality and weather from Lake Mendota. By explicitly modeling time lags and allowing for the presence of latent confounders using the Time-Series Iterative Causal Discovery (TS-ICD) algorithm, our approach uncovers a scientifically plausible causal graph. We demonstrate the necessity of this approach by comparing its structural output and computational complexity against a baseline algorithm (PC) that assumes causal sufficiency. Our results identify key drivers of chlorophyll-a and reveal evidence of confounding between same-day water temperature and algae levels.