Equilibrium-partitioning Sediment Guidelines (ESGs) for metals (EPA 2000a) use acid volatile 
sulfide (AVS) and interstitial water to predict biological effects. These methods have been 
demonstrated to be very useful in predicting biological effects in laboratory experiments and in a 
limited number of field experiments, but few of them have examined the importance of temporal 
and spatial variability in exposures within and near sediments, which limits the usefulness of 
these equilibrium approaches. Research is needed to characterize how AVS and metals 
concentrations at the sediment/water interface vary seasonally and due to hydrological events to 
support better characterization of effects. 
Toxicity Model 
The toxicity model in a risk assessment must integrate exposure information, toxicological data, 
interspecies extrapolations, dose-response nK>dels, and effects of exposure conditions into 
characterizations of risks for various organism-level endpoints, species, and life stages. 
Depending on the requirements of the assessment, including the tier level, various research and 
development needs exist. 
A significant shortcoming in current criteria and many assessments for nonbioaccumulative 
toxicants is consideration of toxicity information for only a single level of effect at a set exposure 
duration (e.g., 96-hr LC50). Critical to better risk characterization are models that more 
comprehensively describe toxic response, including both the magnitude of response and the 
effect of exposure time series. Some capabilities for this have long existed. Models that describe 
the relationship of effects to concentration or time are standard tools in toxicity, but traditionally 
have not combined the effects of both. Furthermore, these models generally assume that 
concentrations are constant. Mancini (1983) described a model that would describe the effects of 
fluctuating concentrations based on standard toxicity tests, but did not incorporate variable levels 
of effects into the model. Subsequent work has tested the utility of this approach with mixed 
results and broadened the model to describe variable levels of effects as a function of both 
concentration and time. Although this model does involve significant uncertainties, it is well 
suited to form the core element in risk assessments and criteria development. It can use toxicity 
information from standard tests and describe the level of effects, with uncertainties, expected 
from any exposure series. This model can thus be combined with expected spatial and temporal 
exposure distributions to produce individual-level risk curves for populations of biological 
receptors. This would provide the individual-level risk characterization in Figure 10 needed for 
any further risk characterization at the population or community level. Other modeling 
approaches, such as proportional hazard and accelerated failure models, also might provide a 
basis for better describing toxicological risks and their dependence on exposure time-series. 
Based on current information, the critical immediate need is to further develop these techniques, 
establish their validity and uncertainty, and describe their application to WQC, which would be a 
major part of the first step in Figure 9. This would establish a core framework for improving 
criteria, but also would identify knowledge gaps that should be addressed by further research. 
Chemical toxicity to aquatic organisms can be influenced markedly by various physicochemical 
exposure factors. For example, ammonia toxicity can vary by orders of magnitude due to the 
combined effects of pH and temperature and by significant amounts due to DO and certain major 
ions. EPA aquatic life criteria for ammonia (EPA 1999) are a function of pH and temperature. 
86 
