Measurability (can we get the data we need to do the assessment?);
Correlation (there might be little value in studying an end point that is highly correlated with one already selected);
Policy relevance (can the end point be linked to feasible policy options?);
Tracking and enforcement (can future efforts tell whether the management actions based on risk assessment have been effective?).
Many ecological risk assessments necessarily deal with complex systems that offer an abundance of possible end points for study, and selection of one or a few of them for the intense effort required in a full-scale risk assessment is likely to be time-consuming and expensive—perhaps as long and expensive as the risk assessment itself.
As a strategy for selecting end points, the group consensus favored starting with a broad focus and then narrowing to the appropriate level of detail to define the design of the assessment. Taking an initially broad approach prevents missing the broader implications of hazard and stress. Institutional forms of risk assessment, such as premanufacture reviews, are so routinized that the level of organization (e.g., population) is predetermined. For noninstitutional applications, the ability to quantify will probably dictate the level of organization.
Consideration of Nonlinearities And Discontinuities
Nonlinearities and discontinuities are likely in the response of ecological systems to stress. The group consensus was that the likelihood of observing a threshold or mean-threshold in the stress-response function increases with system complexity. Because thresholds are common in ecological systems, goals of stress-response analysis should include identification of degrees of stress at which thresholds occur and estimation of the upper ends of the threshold ranges.
The slope of the stress-response curves might be steeper as the scale of organization increases—and might approach a step function for communities and ecosystems. Therefore, the assessor needs to be sensitive to the probabilities of catastrophic changes that have few analogues at lower levels of organization and, consequently, use a greater margin of