Detecting ecological regime shifts from transect data

Abstract

Timely detection of ecological regime shifts is a key problem for ecosystem managers, because changed ecosystem dynamics and function will usually necessitate a change in management strategies. However, currently available methods for detecting regime shifts depend on having multiple long time series data from both before and after the regime shift. This data requirement is prohibitive for many ecosystems. Here, we present a new approach for detecting regime shifts from one‐dimensional spatial (transect) data from just a single time step either side of the transition. Characteristic length scale (CLS) estimation is a method of attractor reconstruction combined with nonlinear prediction that enables identification of the emergent scale at which deterministic behavior of the system is best observed. Importantly, previous studies show that a fundamental change in ecosystem dynamics, from one domain of attraction to another, is reflected in a change in the CLS, i.e., the approach enables distinguishing regime shifts from variability in dynamics around a single attractor. Until now the method required highly resolved two‐dimensional spatial data, but here we adapted the approach so that the CLS can be estimated from one‐dimensional transect data. We demonstrate its successful application to both model and real ecosystem data. In our model test cases, we detected change in the CLS in cases where the shape (topology) of the interaction network had changed, leading to a shift in community composition. In an examination of benthic transect data from four Indonesian coral reefs, changes in the CLS for two of the reefs indicate a regime shift. This new development in estimating CLSs makes it possible to detect regime shifts in systems where data are limited, removing ambiguity in the interpretation of community change.

Inferring critical thresholds of ecosystem transitions from spatial data

“…, Ward et al. ). However, the above methods are limited to a few ecosystems where data are available over large enough spatial or temporal scales, thus limiting their applicability.…”

“…These points can also be estimated from long-term data of ecosystems that have previously undergone transitions (Ratajczak et al 2014). Alternatively, one could construct a complete characterization of ecosystem states as a function of drivers (Hirota et al 2011, Staver et al 2011b, Staal et al 2016, and estimate critical thresholds; these studies implicitly assume a space-for-time substitution approach, i.e., response of ecosystems to temporal evolution of drivers can be inferred by investigating ecosystem states along spatial gradients of drivers (Eby et al 2017, Ward et al 2018). However, the above methods are limited to a few ecosystems where data are available over large enough spatial or temporal scales, thus limiting their applicability.…”

Widespread Mismatch Between Phenology and Climate in Human‐Dominated Landscapes

“…Looking forward, a concerted effort will be needed to further establish the mechanisms and consequences of phenological mismatch. First, the scaling of phenological mismatch at finer resolutions should be investigated, as phenology is highly heterogeneous and mediated by local microclimate variables in complex terrain (Villegas et al., 2010; Ward et al., 2018). Second, insights from land surface phenology in this study can be enhanced by the inclusion of individual‐to species‐level phenology data that are free of confounding factors such as species composition, disturbance, and snow (C. Wang et al., 2017).…”

Long‐term empirical evidence, early warning signals and multiple drivers of regime shifts in a lake ecosystem

“…However, there is little direct empirical evidence to confirm system‐level nonlinear changes in natural ecosystems (Capon et al., 2015), as monitoring programmes are often limited by short time‐series lengths and low sampling resolutions. Although regime shifts can also be tested indirectly using methods of space‐for‐time substitutions in field observations (Su, Wu, et al., 2019; van Nes & Scheffer, 2005; Ward et al., 2018) or remote sensing archives (Staver et al., 2011; Xu et al., 2016), spatial data often have different temporal contexts and cannot provide site‐specific mechanisms of change.…”