A common uncertainty in any aquatic risk assessment is the sensitivity of species and endpoints 
not directly tested. In some cases, there is considerable toxicity data, but certain important data 
are missing, for example, having acute mortality data, but not chronic reproduction data, for a 
critical or sensitive species. Additionally, there are many cases where desired assessments have 
available only a limited number of laboratory toxicity data, if any. Therefore, there is a need for 
models that address extrapolations across endpoints and/or species, and the uncertainties of these 
extrapolations. This relates to the toxicity data needs box in Figure 10 and is a major part of the 
third step in Figure 9. A research effort is needed to improve and test extrapolation methods, and 
to apply these methods to simulated risk assessments with limited data, providing better 
methodology for conducting interspecies extrapolations for nonbioaccumulative toxicants in 
general. 
A final area of major concern is the joint effects of multiple chemical stressors and the effects of 
chemicals in the presence of non-chemical stressors. Much laboratoiy-based research has been 
done on these issues, but has seen little application to EPA criteria, except recently for sediment 
toxicity assessments. A principal need is to determine how this past work should be incorporated 
into assessments, and to better identify what additional work will be worthwhile. 
Population Model 
Toxicity tests can provide information on a diverse set of endpoints, but comparing the relative 
risk of these endpoints and their significance at the population and community level is difficult. 
An important step in better defining the significance of these toxicological effects is to 
incorporate them into population model that will translate these effects into some common 
’’currency” of population dynamics. A critical research need therefore is to develop, test, and 
apply population models for a variety of species relevant to assessments of nonbioaccumulative 
toxicants. This is represented in the second step in Figure 9. These models ^\^ll be used not only 
to describe the significance of observed and predicted organism-level toxicity, but also to 
evaluate the usefulness of toxicity tests and to determine needed changes in the types of tests 
conducted. 
Stage-structured population models can be used to link individual-level effects to the population 
level, such that a set of vital rates defines the dynamics of a population. Such models require 
estimates of vital rates for all life stages, which may or may not show effects from a given level 
of toxic chemical exposure. A set of vital rates also can define the life history strategy of a 
species. Life history strategies vary along a continuum from species with early reproduction, 
high fecundity, and short life expectancy (r-selection model) to species with delayed 
reproduction, low fecundity, and long life expectancy (K-selection model). Population-level 
responses may differ between species because of differences in life history strategies, even 
though individuals of different species may show a similar response to a toxic contaminant. 
Variation in life history strategies, and therefore variation in population-level responses to toxic 
contaminants, also may occur geographically in the same species. 
Population models should be constructed for fish, shellfish, and wildlife species exhibiting 
different life history strategies (i.e., species exhibiting different sets of vital rates). Population 
models generic to groups of species whose life history strategies are similar also can be 
88 
