Zooplankton provide early warnings of a regime shift in a whole lake manipulation

Regime shifts are massive changes among alternate ecosystem states. Predicting regime shifts is difficult, but statistical indicators such as increasing variance or autocorrelation may provide early warnings of their impending onset. We conducted a 4 yr lake manipulation to test for early warnings prior to a food-web transition by adding largemouth bass (Micropterus salmoides) to a lake dominated by small fishes, and thereby drive a trophic cascade that altered zooplankton biomass, community composition, and body size. Declining catches of small fishes were associated with shifts to larger bodied species of Daphnia. We measured zooplankton biomass daily in a reference and manipulated lake and calculated variance and autocorrelation of the zooplankton time series using 28 d rolling windows. We asked whether the variance in the manipulated lake (relative to the reference lake) increased and whether autocorrelation approached unity (a value suggested by theory) prior to the food-web regime shift. The variance and autocorrelation of zooplankton biomass were similar between the two lakes in the first 2 yr of the manipulation. During the third year, variance was much higher and autocorrelation approached unity for sustained periods in the manipulated but not in the reference lake. Variance and autocorrelation were similar between the two lakes during the fourth year as the food-web transition moved to completion. The joint response of variance and autocorrelation in year 3 provided an early warning of the food-web transition consistent with theoretical expectation. Some aquatic ecosystems undergo massive, abrupt changes shifting to a new and fundamentally different state (Holling 1973; Carpenter 2003). These regime shifts are well-described in some of the smallest as well as largest systems considered by limnologists and oceanographers. For example, small, shallow lakes shift between turbid, algal dominated and clear-water, vegetated states (Scheffer et al. 1993). These shifts are induced by interactions among several factors, including nutrient dynamics, food-web interactions, and sediment resuspension (Scheffer 1998). Large regions of the ocean oscillate between states because of the interactions of climate and hydrographic conditions. These dynamics induce different oceanic regimes that may extend over decadal time scales and lead to distinct differences in primary production, food-web structure, and fisheries (Chavez et al. 2003; deYoung et al. 2008). The theoretical foundation for regime shifts lies in nonlinear models that undergo sharp transitions between stable regions (Scheffer 2009). These stable regions are known as alternate attractors, with intermediate conditions being unstable such that a system in an unstable domain will move toward one or the other attractor in the case of two alternate stable states. Transitions among states typically exhibit hysteresis. In addition to direct evidence for regime shifts noted above, there is a concern that regime shifts may occur more widely because of large-scale environmental forcings such as climate warming, sea-level rise, and loss of biodiversity (Scheffer et al. 2001). Regime shifts are a problem for ecosystem management and affect ecosystem services (Millennium Ecosystem Assessment 2005a). The rapid changes during regime shifts are difficult to predict or anticipate, because they often extend beyond the range of historical experience (Carpenter 2002, 2003) and involve thresholds that are rarely known before they are crossed (Groffman et al. 2006). Regional and global environmental change is likely to cause future regime shifts in inland water and marine ecosystems affecting services such as fish production, pollution mitigation, and storm protection, with consequences for human well-being (Millennium Ecosystem Assessment 2005b; Carpenter et al. 2011). One prospect for anticipating regime shifts comes from recent theory that provides evidence for measurable changes in key ecosystem variables before regime shifts (Carpenter and Brock 2006; van Nes and Scheffer 2007; Scheffer et al. 2009). Specifically, there are early warning indicators (EWI) in time series that foreshadow regime shifts (Scheffer et al. 2009). Most of these indicators are based on familiar statistics as, for example, the variance or autocorrelation of a time-series variable across a sequence of time windows. For variables measured at high frequency relative to the time scale of regime shifts, these indicators provide a means of anticipating and possibly forestalling unwanted changes (Biggs et al. 2009; Contamin and Ellison 2009). Despite rapid theoretical advances (Brock and Carpenter 2010; Dakos et al. 2011; Seekell et al. 2011), there have only been a few empirical tests of EWIs and these studies have mainly used either laboratory systems (Drake and Griffin 2010; Dai et al. 2012;Veraart et al. 2012) or retrospective analyses (Dakos et al. 2009; Bestelmeyer et al. 2011; Lindegren et al. 2012). We tested EWI theory using a whole lake manipulation of a top-predator. We previously reported early warnings from temporal phytoplankton dynamics over the first 3 yr of the study (Carpenter et al. 2011). In this paper, we consider high-frequency observations of zooplankton over * Corresponding author: [email protected] Limnol. Oceanogr., 58(2), 2013, 525–532 E 2013, by the Association for the Sciences of Limnology and Oceanography, Inc. doi:10.4319/lo.2013.58.2.0525