Evidence for interactions among environmental stressors in the Laurentian Great Lakes

Highlights

• Interactions among stressors in ecosystems are common but hard to predict.

• We identified interactions by pairing a systematic review with expert elicitation.

• Many potential interactions (esp. synergies) were proposed but poorly studied.

• Nutrient loading and invasive species were often the stressors in these interactions.

• We used a conceptual model to integrate several potential interactions.

Abstract

Co-occurrence of environmental stressors is ubiquitous in ecosystems, but cumulative effects are difficult to predict for effective indicator development. Individual stressors can amplify (synergies) or lessen (antagonisms) each other's impacts or have fully independent effects (additive). Here we use the Laurentian Great Lakes, where a multitude of stressors have been studied for decades, as a case study for considering insights from both a systematic literature review and an expert elicitation (or structured expert judgment) to identify stressor interactions. In our literature search for pairs of stressors and interaction-related keywords, relatively few studies (9%, or 6/65) supported additive interactions with independent stressor effects. Instead, both antagonisms (42%, or 27/65) and synergies (49%, or 32/65) were common. We found substantial evidence for interactions of invasive dreissenid mussels with nutrient loading and between pairs of invasive species (predominantly dreissenids × round goby), yet both sets of records included mixtures of synergies and antagonisms. Complete quantification of individual and joint effects of stressors was rare, but effect sizes for dreissenid mussels × nutrient loading supported an antagonism. Our expert elicitation included discussion in focus groups and a follow-up survey. This process highlighted the potential for synergies of nutrient loading with dreissenid mussels and climate change as seen from the literature review. The elicitation also identified additional potential interactions less explored in the literature, particularly synergies of nutrient loading with hypoxia and wetland loss. To stimulate future research, we built a conceptual model describing interactions among dreissenid mussels, climate change, and nutrient loading. Our case study illustrates the value of considering results from both elicitations and systematic reviews to overcome data limitations. The simultaneous occurrence of synergies and antagonisms in a single ecosystem underscores the challenge of predicting the cumulative effects of stressors to guide indicator development and other management and restoration decisions.

Introduction

The full suite of anthropogenic stressors impacting an ecosystem can greatly affect the development and performance of indicators for environmental management (Niemi and McDonald, 2004). The effects of individual stressors may not accurately predict changes in the condition of species and ecosystems when stressors co-occur, since multiple stressors can amplify (synergies) or lessen (antagonisms) cumulative impact. Such interactions among environmental stressors may be common. For example, one broad meta-analysis revealed interactions among stressors in 77% of the experiments examined (Darling and Côté, 2008). Interpretations of past meta-analyses have emphasized the importance of antagonisms over synergies overall (Côté et al., 2016, Darling and Côté, 2008) and for aquatic ecosystems (Jackson et al., 2016). However, identifying the interactions from pairwise studies and integrating those interactions into accurate predictions of stressor effects in real ecosystems can be challenging.

Few ecosystems can be systematically assessed for stressor interactions (e.g., Brennan and Collins, 2015), and harnessing multiple approaches may improve our ability to integrate available information (Mastrandrea et al., 2010, Memon and Thapa, 2011). Factorial experiments, comparative observational studies, and simulation models can be used to produce estimates of the effects of stressors on response variables occurring individually versus together (Côté et al., 2013). Quantitative meta-analyses synthesizing such studies are ideal to evaluate the overall incidence of interactions, but meta-analyses are not reliable with low sample sizes (Rosenberg et al., 2013). In situations where meta-analysis is not feasible, rigorous synthesis can be derived from other forms of evidence synthesis or structured expert judgment (reviewed by Cook et al., 2017, Martin et al., 2012, respectively). Several forms of data synthesis are available to suit the purpose and question, required level of confidence in output, and available expertise and resources (Cook et al., 2017). One particularly transparent and reliable form of synthesis is a systematic literature review. In systematic reviews, published studies are found and evaluated using explicit criteria to address predefined research questions (Lortie, 2014), allowing an unbiased and replicable survey of the literature for qualitative or quantitative synthesis. A second alternative is expert elicitation (or structured expert judgment), which capitalizes on the broad knowledge and deep abilities of experts to synthesize disparate types of information and to reason through complex relationships (Martin et al., 2012). Many elicitation designs are available to balance different considerations (e.g., respondent bias, group social dynamics, respondent time and effort). Another benefit of expert elicitation in the context of conservation issues is that many designs can be completed rapidly (Martin et al., 2012). As complementary approaches, systematic reviews and expert elicitation together can provide a solid basis for understanding complex environmental problems. Comparing methods may improve data synthesis in policy contexts (Pullin et al., 2016), but even narrative literature reviews and elicitations are rarely formally used together (but see: Memon and Thapa, 2011, Rositano and Ferraro, 2014).

The Laurentian Great Lakes are a globally important system subject to multiple stressors (Allan et al., 2013), and there is a critical need to understand stressor effects, including their interactions. The Lakes are the largest source (18% globally) of surficial freshwater on the planet (Sellinger et al., 2008), support millions of people living and working within the basin, and generate billions (USD) each year economically, particularly in tourism and recreation (Allan et al., 2015, Austin et al., 2007, Vaccaro and Read, 2011). Many anthropogenic stressors co-occur in the basin, including invasive species, climate change effects, urban development, and nonpoint and toxic chemical pollution (Allan et al., 2013). Many ecological indicators for specific stressors and overall condition have been in use (IJC, 2014). The potential for stressor interactions has been identified as a major unmet research need for the Great Lakes (Sterner et al., 2017). Yet, the Great Lakes are reasonably well studied and thus may offer transferable lessons that can be applied to other ecosystems. Expert elicitation has been used successfully to assess current and future stressor impacts in the Great Lakes (Rothlisberger et al., 2010, Smith et al., 2015, Wittmann et al., 2015) but not to assess interactions among stressors.

We expected interactions to be common in the Great Lakes based on the nonlinear changes observed in recent decades, such as food web collapses and algal blooms (Bunnell et al., 2014, Michalak et al., 2013). Bails et al. (2005) hypothesized synergies among various categories of stressors (climate change, nutrient loading, overfishing, toxic chemicals, invasive species, and land use and hydrologic alterations), while others have considered effects of toxic chemicals to be largely additive (e.g., Jackson et al., 2016). Based on previous meta-analyses of interactions (Côté et al., 2016, Crain et al., 2008), we expected antagonisms to be most common. Some stressor combinations have documented mechanistic potential for nonlinear effects, such as invasive mussels moving anthropogenic nutrient inputs from pelagic to benthic food webs (nearshore phosphorus shunt hypothesis: Hecky et al., 2004) and climate change with other stressors promoting regime shifts (Barnosky et al., 2012, Pace et al., 2015).

We paired a systematic literature review and an expert elicitation to explore the incidence of strong and likely stressor interactions in the Laurentian Great Lakes. We aimed to answer the following questions.

• Which pairs of stressors are likely to interact strongly to affect ecosystem condition in the Great Lakes?

• What is the primary direction of each potential interaction (synergy or antagonism)?

• For the pairs of stressors identified as likely to interact based on the elicitation, what is the mechanistic basis of each potential interaction?

Although we found the availability of empirical data on stressor interactions for the Great Lakes to be limited, the systematic review and expert elicitation together allowed us to summarize current evidence, identify potential effects, and highlight the importance of further study of stressor interactions to better understand ecological impacts.