Listing Multiple Stressors as Candidate Causes
Effects are often caused by multiple stressors acting together. When developing your list of candidate causes, consider combining stressors that act together. You can reduce the number of causes that must be considered, and more importantly, the combined causes may explain effects better than individual stressors. The following strategies and warnings for combining stressors are discussed further below.
Strategies for Combining Stressors
- Combine stressors that are part of the same causal pathway
- Re-aggregate stressors that have been unnecessarily disaggregated
- Combine similar stressors into one
- Identify independently acting stressors that cause the same effect
- Combine stressors that induce the effect interactively
- Avoid combining causes without an underlying model
- Avoid broad definitions of candidate causes
- Do not lose the independent effects of individual causes
Strategies for Combining Stressors
Combine stressors that are part of the same causal pathway: Sometimes the multiple stressors are not all proximate causes but are related to the same proximate cause. For example, nitrogen, phosphorus, organic matter (BOD), and dissolved oxygen (DO) may all be proposed as candidate causes, but only DO is a proximate cause. Hence, only DO should be listed, and the others should be considered when evaluating the causal pathway as evidence of causation and later when designing remedial actions. Conceptual models are useful tools for making this distinction. Focusing on proximate causes rather than intermediate steps can improve associations as well as reducing the number of listed candidate causes. For example, regional studies of the effects of acid deposition on wood thrush occurrence showed little relationship with soil pH or calcium content as causal variables, but strong associations with the apparent proximate cause, the abundance of calcium-rich invertebrate prey (Hames et al. 2006).
Re-aggregate stressors that have been unnecessarily disaggregated: Multiple stressors that have the same source may more effectively be considered as a group. For example, constituents of an effluent may have each been listed as a candidate cause. In such cases, it may be more appropriate to list the effluent as a candidate cause, perform whole effluent toxicity tests, model or measure dilution and degradation of the effluent, and analyze the evidence that the effluent is the most probable cause. This strategy works well when the exposure and effects of the constituents are measured together as in the effluent toxicity tests, thereby avoiding the need to generate an exposure-response model from measurement of the constituents and tests of their individual effects (Norberg-King et al. 2005). The candidate cause is the effluent and the expression of exposure is the proportional dilution of the effluent.
Combine similar stressors into one: The effects of similarly acting stressors result from their combined exposures, so they should be combined so that an exposure additivity model can be applied to them (U.S. EPA 2000b). This is appropriate for stressors with common modes of action such as organophosphate pesticides or neutral hydrocarbons causing narcosis. For chemicals, this may be done by analyzing the constituents separately, converting their concentrations into a common toxic unit (TU), and then combining them by addition into a measure of combined exposure such as the sum of toxic units (ΣTU). For example, individually measured PAHs have been combined by adding their toxicity-normalized concentrations to estimate their combined toxic effects (Di Toro et al. 2000). Alternatively, multiple regression can generate an additivity model from field data if the independence and linearity assumption is met. Non-chemical stressors may be combined similarly; rocks and large woody debris may be combined as hard substrates. However, this combining of similar nonchemical stressors, like combining chemicals, depends on knowledge of their actual mode of action. For example, deposited sand, silt and clay are often combined as “fines,” which may be appropriate for gravel-spawning fish, but benthic invertebrates may perceive sand as a different substrate from silt and clay. Similarly, suspended mineral particles (sediment) may be combined with suspended algae if the mode of action is reduced light for rooted macrophytes or inhibition of visual predation, but not if it is gill abrasion or interference with filter feeding. For this approach, the expressions of exposure to the candidate causes are the total amount of the similarly acting stressors, either normalized (e.g., ΣTU) or not (e.g., total fines, total habitat structure).
Identify independently acting stressors that cause the same effect: Effects additivity models are used when stressors cause the same effect but they act separately (U.S. EPA 2000b). For example, if entrainment in cooling systems and multiple predators are removing immature fish from a population, those losses may be added. The extent of aggregation depends on the case. In the example, all sources of acute lethality may be combined to be compared to sublethal and chronic candidate causes, or all predators may be combined but distinguished from a cooling system or harvesting. This approach does not actually combine stressors, but it does create a category of stressors that independently cause a discrete effect. The effects of these stressors will be added at the end of the analysis phase to determine whether their total effects account for the magnitude of the impairment.
Combine stressors that induce the effect interactively: Some stressors jointly induce effects through interactions other than additivity. For example, low pH increases the toxicity of many metals, and low flows exacerbate the effects of low oxygen. Combining stressors that interact requires some knowledge of the interaction, preferably a quantitative model. Interactive models can be found in the literature from laboratory tests, generated from knowledge of mechanisms using mathematical simulation models, or generated empirically from regional field data statistical modeling. Examples of interaction models from laboratory tests include interactive toxicities of atrazine and some organophosphate insecticides (Belden and Lydy 2000). Mathematical simulation models of interactive effects are particularly important when the interacting stressors are heterogeneous. For example, the rapid collapse of the Lake Trout population in Lakes Huron and Michigan appeared to have been due to more-than-additive combined effects of harvesting, lamprey parasitism and chlorinated organic chemicals, which could be modeled if assumptions are made about compensatory and depensatory capabilities of the population (Gentile et al. 1999). Some interactions are identified from prior knowledge, but others will be suspected only after the causal analysis cannot account for the impairment.
Avoid combining causes without an underlying model: Sometimes stressors are combined without knowledge of how they interact to produce effects. For example, stressors such as suspended sediment, substrate texture, woody debris, flow velocity, and channel depth are often combined in habitat indices. These indices may be used as a candidate cause, and models relating biological variables to habitat indices may be developed. However, because they are not combined on a mechanistic basis, they are of limited utility for explaining causation or identifying remedial actions. Instead, we recommend disaggregating indices into their constituent metrics and recombining them appropriately based on their modes of action.
Avoid broad definitions of candidate causes: Aggregating candidate causes does not mean broadly defining candidate causes. For example, agricultural land use and suburban development are commonly described as causes of impairment. However, such broadly defined causes cannot be analyzed with any precision or remediated when they are judged to be acting. We will not eliminate agriculture or suburbs, but we can reduce inputs of sediment, nutrients, pesticides or other stressors commonly associated with those land uses. Different stressors usually require different management actions for mitigation. For example, stormwater ponds can reduce flow extremes, but may also increase water temperature. In rare cases, multiple stressors can be eliminated by a single practical remedial action; in those cases, aggregation may make sense (see re-aggregation, above).
Do not lose the independent effects of individual causes: Even when appropriately combining stressors into a single candidate cause, it is important to consider whether any of the stressors may cause an effect independently of the combined effect. For example, elevated temperature inevitably contributes to the effects of low dissolved oxygen by reducing oxygen solubility and increasing oxygen consumption by biota, but elevated temperature may also be sufficient to cause effects independently of dissolved oxygen.