Experimental Perturbations Modify the Performance of Early Warning Indicators of Regime Shift

Abstract

Ecosystems may shift abruptly between alternative states in response to environmental perturbations. Early warning indicators have been proposed to anticipate such regime shifts, but experimental field tests of their validity are rare. We exposed rocky intertidal algal canopies to a gradient of press perturbations and recorded the response of associated assemblages over 7 years. Reduced cover and biomass of algal canopies promoted the invasion of algal turfs, driving understory assemblages toward collapse upon total canopy removal. A dynamic model indicated the existence of a critical threshold separating the canopy- and turf-dominated states. We evaluated common indicators of regime shift as the system approached the threshold, including autocorrelation, SD, and skewness. These indicators captured changes in understory cover due to colonization of algal turfs. All indicators increased significantly as the system approached the critical threshold, in agreement with theoretical predictions. The performance of indicators changed when we superimposed a pulse disturbance on the press perturbation that amplified environmental noise. This treatment caused several experimental units to switch repeatedly between the canopy- and the turf-dominated state, resulting in a significant increase in overall variance of understory cover, a negligible effect on skewness and no effect on autocorrelation. Power analysis indicated that autocorrelation and SD were better suited at anticipating a regime shift under mild and strong fluctuations of the state variable, respectively. Our results suggest that regime shifts may be anticipated under a broad range of fluctuating conditions using the appropriate indicator.