Organic-matter loading determines regime shifts and alternative states in an aquatic ecosystem

Abstract

Slow changes in underlying state variables can lead to "tipping points," rapid transitions between alternative states ("regime shifts") in a wide range of complex systems. Tipping points and regime shifts routinely are documented retrospectively in long time series of observational data. Experimental induction of tipping points and regime shifts is rare, but could lead to new methods for detecting impending tipping points and forestalling regime shifts. By using controlled additions of detrital organic matter (dried, ground arthropod prey), we experimentally induced a shift from aerobic to anaerobic states in a miniature aquatic ecosystem: the self-contained pools that form in leaves of the carnivorous northern pitcher plant, Sarracenia purpurea. In unfed controls, the concentration of dissolved oxygen ([O2]) in all replicates exhibited regular diurnal cycles associated with daytime photosynthesis and nocturnal plant respiration. In low prey-addition treatments, the regular diurnal cycles of [O2] were disrupted, but a regime shift was not detected. In high prey-addition treatments, the variance of the [O2] time series increased until the system tipped from an aerobic to an anaerobic state. In these treatments, replicate [O2] time series predictably crossed a tipping point at ~45 h as [O2] was decoupled from diurnal cycles of photosynthesis and respiration. Increasing organic-matter loading led to predictable changes in [O2] dynamics, with high loading consistently driving the system past a well-defined tipping point. The Sarracenia microecosystem functions as a tractable experimental system in which to explore the forecasting and management of tipping points and alternative regimes.

Fig. 1.

Analysis of the time series of [O₂] concentration (in percent; on this scale, atmospheric [O₂] = 20.95 ≡ 1.26 g/L in our greenhouse) in Sarracenia microecosystems as a function of experimental organic-matter loading through prey addition: 0 (control), 0.125, 0.25, 0.5, or 1.0 mg⋅mL⁻¹⋅d⁻¹ dry mass of ground wasps (n = 5,734 observations of [O₂] per replicate within each treatment). (A–E) Mean raw data [O₂] time series (blue line) and 95% CIs (gray region) for each treatment across six replicates. (F–J) Mean decycled and detrended time series (blue line) and 95% CIs (gray region) for each treatment across six replicates. The five vertical cyan lines show the break points for the [O₂] residuals of each replicate (one of the six replicates did not display a significant break point); the vertical red line in each treatment shows the break point for the mean of all of the series within a treatment. (K–O) Relationships between the primary environmental driver (PAR, in μmol⋅m⁻²⋅s⁻¹) and average (residual) [O₂]. Lines of different colors correspond to the 4 d of each trial run: dark blue represents day 1, light blue represents day 2, yellow represents day 3, and brown represents day 4. The red circles in N and O indicate the times of the switches from aerobic to anaerobic states in the two highest prey-addition treatments.

Fig. 2.

Frequency distributions (number of minutes between 0900 and 1500 hours) of [O₂] in the five different prey-addition treatments. The red triangles indicate the locations of four modes in the joint distribution identified with normal mixture modeling and model-based clustering (SI Appendix, Fig. S12). The first identified mode (at 1.682% O₂) corresponds to the mode for the distributions of the two highest prey-addition treatments (0.5 and 1.0 mg⋅mL⁻¹⋅d⁻¹); the second (7.554%) corresponds to the mode of the distribution of the intermediate prey-addition treatment (0.25 mg⋅mL⁻¹⋅d⁻¹); the third (12.146%) corresponds to the mode for the lowest prey-addition treatment (0.125 mg⋅mL⁻¹⋅d⁻¹); and the fourth (16.272%) corresponds to the mode for the distribution of the controls.

Fig. 3.

Box plots of statistical moments and least-squares regression slope coefficients of the time series of [O₂]. Gray boxes represent replicates from treatments with no break points (0, 0.125. 0.25 mg⋅mL⁻¹⋅d⁻¹ added prey). Blue and brown boxes represent the values for these variables in each of the two different states of the system induced by the higher levels of organic-matter loading (0.5 and 1.0 mg⋅mL⁻¹⋅d⁻¹ added prey). Table 1 shows statistical summaries of treatment effects.