Applying early warning indicators to predict critical transitions in a lake undergoing multiple changes

Abstract

Lakes are dynamic ecosystems that can transition among stable states. Since ecosystem-scale transitions can be detrimental and difficult to reverse, being able to predict impending critical transitions in state variables has become a major area of research. However, not all transitions are detrimental, and there is considerable interest in better evaluating the success of management interventions to support adaptive management strategies. Here, we retrospectively evaluated the agreement between time series statistics (i.e., standard deviation, autocorrelation, skewness, and kurtosis-also known as early warning indicators) and breakpoints in state variables in a lake (Lake Simcoe, Ontario, Canada) that has improved from a state of eutrophication. Long-term (1980 to 2019) monitoring data collected fortnightly throughout the ice-free season were used to evaluate historical changes in 15 state variables (e.g., dissolved organic carbon, phosphorus, chlorophyll a) and multivariate-derived time series at three monitoring stations (shallow, middepth, deep) in Lake Simcoe. Time series results from the two deep-water stations indicate that over this period Lake Simcoe transitioned from an algal-dominated state toward a state with increased water clarity (i.e., Secchi disk depth) and silica and lower nutrient and chlorophyll a concentrations, which coincided with both substantial management intervention and the establishment of invasive species (e.g., Dreissenid mussels). Consistent with improvement, Secchi depth at the deep-water stations demonstrated expected trends in statistical indicators prior to identified breakpoints, whereas total phosphorus and chlorophyll a revealed more nuanced patterns. Overall, state variables were largely found to yield inconsistent trends in statistical indicators, so many breakpoints were likely not reflective of traditional bifurcation critical transitions. Nevertheless, statistical indicators of state variable time series may be a valuable tool for the adaptive management and long-term monitoring of lake ecosystems, but we call for more research within the domain of early warning indicators to establish a better understanding of state variable behavior prior to lake changes.

Keywords: critical transitions; early warning indicators; invasive species; management actions; resilience indicators.

FIGURE 1

Map displaying Lake Simcoe stations studied (C1 [LS‐S], C9 [LS‐M], K42 [LS‐D]) and watershed land cover.

FIGURE 2

Principal component (PC) analysis of state variables in Lake Simcoe, where points represent the PC scores of biweekly time steps, and arrows show PC loadings of state variables.

FIGURE 3

(Left) Breakpoint timing, direction of change (▲ = increase; ▼ = decrease), and effect size (white: Cohen's d < 0.3; gray: 0.3 < Cohen's d < 0.5; light red: 0.5 < Cohen's d < 0.8; red: 0.8 < Cohen's d) for principal component (PC) score and state variable time series across three Lake Simcoe monitoring stations (shallow: LS‐S; middepth: LS‐M; deep: LS‐D). (Right) Number of EWIs with expected trends prior to corresponding critical transitions identified through breakpoint analysis (light gray = 0; medium gray = 1; dark gray = 2; dark red = 3; red = 4). EWI trends prior to 1988 (shaded time period) could not be evaluated due to lack of data availability.

FIGURE 4

Flow diagram illustrating connections between introduction of invasive species and selected lake water quality variables. Similar patterns may be predicted with reduced phosphorus inputs.