Experimenting with ecosystem interaction networks in search of threshold potentials in real‐world marine ecosystems

Abstract

Abstract. Thresholds profoundly affect our understanding and management of ecosystem dynamics, but we have yet to develop practical techniques to assess the risk that thresholds will be crossed. Combining ecological knowledge of critical system interdependencies with a largescale experiment, we tested for breaks in the ecosystem interaction network to identify threshold potential in real-world ecosystem dynamics. Our experiment with the bivalves Macomona liliana and Austrovenus stutchburyi on marine sandflats in New Zealand demonstrated that reductions in incident sunlight changed the interaction network between sediment biogeochemical fluxes, productivity, and macrofauna. By demonstrating loss of positive feedbacks and changes in the architecture of the network, we provide mechanistic evidence that stressors lead to break points in dynamics, which theory predicts predispose a system to a critical transition.

Bottom–up and top–down mechanisms indirectly mediate interactions between benthic biotic ecosystem components

“…Hobo data loggers (Onset Corp.) and Thermochron i-buttons were used to quantify differences in light intensity and temperature between treatments. These nine experimentally manipulated plots (2 × 2 m) are a subset of treatments that formed part of a larger experiment designed to examine the interactive effects of light intensity, nutrient loading, and Macomona grazing pressure on sand flat community biodiversity and ecosystem functioning (Thrush et al, 2014). The full experiment consisted of 28 treatments that were randomly positioned in 7 blocks (14 × 56 m, one replicate of each treatment per block).…”

Addressing surprise and uncertain futures in marine science, marine governance, and society

“…Often, regime shifts can be traced back to the interaction of environmental drivers with ecological processes , Thrush et al 2014. The importance of these interactions in driving change implies that management actions that only focus on limiting levels of stress, disturbance, or the removal of organisms without also paying attention to ecosystem responses are likely to generate surprise.…”

“…If the seagrass dies off, fine sediments are readily resuspended, reducing light and inhibiting regrowth or restoration (van der Heide et al 2012). Principles are emerging that would allow feedbacks to be predicted and empirically tested to see how interaction networks may change their typology (Thrush et al 2012(Thrush et al , 2014. The strength of interactions within a network may change with a specific intervention emphasising the need to consider multiple actions or a series of actions to guide a system.…”

“…Increases in the mud content of coastal sediments changes the adaptive capacity of coastal habitats to cope with other stressors, e.g., nutrients from land or ocean acidification. http://www.ecologyandsociety.org/vol21/iss2/art44/ Although these risks are difficult to quantify, targeting critical stressors and identifying changes in ecosystem interaction networks associated with stressors through gradient analysis (Thrush et al 2012) or manipulative experiments (Thrush et al 2014) lead to improved assessment of the risk of a threshold response. Risk assessments also need to incorporate community and landscape features, and nonadditive cumulative assessments (Travis et al 2014).…”

Structural equation modeling of the influence of environmental factors on summer phytoplankton growth in the Ross Sea

“…Structural equation modeling is a technique for testing and estimating causal relations using a combination of statistical data, qualitative causal assumptions, and a theoretical model. The concepts of SEM have been well described (Grace 2006;Pugesek et al 2003;Grace et al 2010;Kline 2015), but their application to marine and pelagic systems remains relatively uncommon (Blake and Duffy 2012;Brewer et al 2012;Thrush et al 2014). A SEM tests causal mechanisms, whereas general linear models (including multiple regression techniques) test statistical significance but do not confirm or reject a theoretical model (Kline 2015).…”