Seeking alternative stable states in a deep lake

Abstract

Hysteresis linked to alternative stable states may explain delays in water quality recovery despite reduced nutrient loadings in shallow lakes. Because deep lakes are assumed to be less prone to critical transitions, similar delays are attributed to the confounding effects of additional environmental disturbances, such as climate warming. Herein, we hypothesised that the lack of evidence of nutrient‐driven alternative stable states in a deep lake arises from the fact that the nutrient threshold that causes the critical transition is lower than the nutrient threshold in shallow lakes. Thereby, it might have been crossed much earlier in the lake history. To test this hypothesis, we focused on the palaeo‐ecological trajectory of Lake Varese, which is a deep, hypereutrophicated peri‐alpine lake undergoing restoration. Proxies for drivers of ecological state (i.e. total phosphorus—TP—through diatoms and pigments) and ecological responses (Cladocera), as well as a repeatable analysis, were used to identify transitions and to distinguish hysteretic delays from those of the ecosystems responding to additional constraints over the past century. Our results suggest spatial heterogeneity in the ecological response. The littoral habitats changed abruptly and prematurely for a low TP threshold, causing a shift that met many criteria of a flickering‐type critical transition. Soon after the littoral shift, a striking increase in the lake phosphorous concentration was recorded and drove the pelagic assemblage towards a new state. This transition was abrupt, and the pelagic communities exhibited limited evidence of recovery; however, we found no evidence of hysteresis. Therefore, the modern ecological trajectory of the pelagic communities is currently driven by climate warming. This detailed analysis allowed us to go beyond the general pattern that links ecological responses to drivers and suggest that a nonlinear transition following eutrophication can take place in a deep lake synchronously with linear transitions. Instead of triggering a new regime shift, climate warming, to which pelagic habitats are more sensitive than littoral ones, has driven the lake further from its safe operating space.

Potential anthropogenic regime shifts in three freshwater lakes in Tropical East Asia

“…Regime shifts may weaken the resilience of an ecosystem, increasing its vulnerability to subsequent pressures, including abrupt shocks (Hughes, Carpenter, Rockström, Scheffer, & Walker, ). Restoring ecosystems that have undergone a regime shift to their former state may be particularly challenging, because of hysteresis and the confounding effects on remediation efforts of a complex of interacting drivers (Bruel et al., ).…”

“…Restoring ecosystems that have undergone a regime shift to their former state may be particularly challenging, because of hysteresis and the confounding effects on remediation efforts of a complex of interacting drivers (Bruel et al, 2018).…”

Can we detect ecosystem critical transitions and signals of changing resilience from paleo‐ecological records?

“…We calculated eigenvalues from the AR parameters according to Box et al (2008). We calculated eigenvalues from the AR parameters according to Box et al (2008).…”

“…Instead, we noted rising variance with inconsistent AR(1) responses, even in simulations where we knew the underlying dynamics. Bruel et al (2018) recently noted an increased variance and autocorrelation preceding non-critical transitions from paleo-records, but that the values and significance of resilience indicators strongly depend on the rolling window size. Similarly, Ditlevsen and Johnsen (2010) found inconsistent signals from autocorrelation and variance in paleo-climate interpretations, while Thomas (2016) found using a simple bifurcation model to mimic paleo-climate data that autocorrelation displayed a better performance for detecting critical slowing down than variance, possibly due to increased sensitivity to additional processes and mechanisms affecting this parameter (Verbesselt et al 2016).…”

“…However, degrading the data length, data resolution and data density diminished both the autocorrelation and variance signals (Thomas 2016). Bruel et al (2018) recently noted an increased variance and autocorrelation preceding non-critical transitions from paleo-records, but that the values and significance of resilience indicators strongly depend on the rolling window size. Lastly, an analysis of forest-grassland transitions using a simple model as well as observed paleorecords found that rolling window variance, but not autocorrelation, could reliably detect the transition (M. Stegner et al, unpublished manuscript).…”