Stochastic dynamics of Cyanobacteria in long-term high-frequency observations of a eutrophic lake

Abstract

Concentrations of phycocyanin, a pigment of Cyanobacteria, were measured at 1-min intervals during the ice-free seasons of 2008–2018 by automated sensors suspended from a buoy at a central station in Lake Mendota, Wisconsin, U.S.A. In each year, stochastic-dynamic models fitted to time series of log-transformed phycocyanin concentration revealed two alternative stable states and random factors that were much larger than the difference between the alternate stable states. Transitions between low and high states were abrupt and apparently driven by stochasticity. Variation in annual magnitudes of the alternate states and the stochastic factors were not correlated with annual phosphorus input to the lake. At daily time scales, however, phycocyanin concentration was correlated with phosphorus input, precipitation, and wind velocity for time lags of 1–15 d. Multiple years of high-frequency data were needed to discern these patterns in the noise-dominated dynamics of Cyanobacteria.