Detecting changes in statistical indicators of resilience prior to algal blooms in shallow eutrophic lakes

Algal blooms in lakes and reservoirs can be considered regime shifts from a clear‐water to algae‐dominated state that often occurs abruptly. Under experimental conditions, these regime shifts have been predicted from rises in variance and autocorrelation (generic resilience indicators) of state variables monitored at a high frequency. The goal of this study was to evaluate the behavior of resilience indicators prior to a critical transition in lakes that naturally experience algal blooms. Ambient lake conditions provide several potential hurdles that could inhibit the detection of meaningful changes in resilience indicators prior to a critical transition such as stochastic nutrient loading, spatial complexity, and decreased resilience due to higher baseline nutrient concentrations. We compiled five lake‐years of high‐frequency monitoring of chlorophyll a, phycocyanin, dissolved oxygen, and pH from four hypereutrophic lakes. Despite the factors that might hinder detecting statistical indicators of changing resilience in hypereutrophic ecosystems, we found that a rise in resilience indicators did occur prior to a critical transition in three out of four possible lake‐years, with rise beginning between 5 and 33 d prior. In one lake‐year, a critical transition occurred soon after the monitoring began, preventing detection of rising variance or autocorrelation signals which are calculated using a 21‐d rolling window. These results add to the growing body of evidence that rises in resilience indicators can be detected in ecosystems prior to a regime shift if monitoring programs are properly designed to capture the dynamics; however, continued research is needed to better understand the conditions under which resilience indicators may be useful as an early warning detection tool for lake management.