Methods for deriving pesticide aquatic life criteria

Environmental regulators charged with protecting water quality must have scientifically defensible water quality goals. For protection of aquatic life, regulators need to know what levels of contaminants a water body can tolerate, without producing adverse effects. The USEPA has developed water quality criteria for many chemicals, but few are for current-use pesticides. Other countries also derive aquatic life criteria utilizing a variety of methodologies. As a prelude to developing a new criteria derivation methodology, this chapter explores the current state of aquatic life criteria derivation around the world. Rather than discussing each methodology independently, this review is organized according to critical elements that must be part of a scientifically defensible methodology. All of the reviewed methodologies rely on effects data to derive aquatic life criteria. Water quality criteria may be derived from single-species toxicity data by statistical extrapolation procedures (for adequate data sets), or by use of empirically based AFs (for data sets of any size). Assessment factor methods are conservative and have a low probability of underestimating risk, with a concomitant high probability of overestimating risk. Extrapolation methods may also under-, or overestimate risk, but uncertainty is quantifiable and is reduced when larger data sets are used. Although less common, methods are also available for criteria derivation using multispecies toxicity data. Environmental toxicity of chemicals is affected by several factors. Some of these factors can be addressed in criteria derivation, and some cannot. For example, factors such as magnitude, duration and frequency of exposure may be incorporated into criteria, either through use of time-to-event and population models or by derivation of both acute and chronic criteria that have duration and frequency components. Aquatic species may be exposed to hydrophobic organic chemicals by multiple routes. They may take up residues directly from water, or may be exposed dietarily, or combinations of both. Unfortunately, to properly address such multiple routes in criteria derivation, food web models are needed that work for chemicals that have specific modes of action. Similarly, both bioavailability and toxicity parameters may contribute to derivation of criteria, providing sufficient data are available. Ecotoxicological effects and physical-chemical data are needed for criteria derivation. The quality and quantity of required data are clearly stated in existing methodologies; some guidelines provide very specific data quality requirements. The level of detail provided by guidelines varies among methodologies. Most helpful are those that provide lists of acceptable data sources, descriptions of adequate data searches, schemes for rating ecotoxicity data, specifications of required data types (e.g., acute vs chronic), and instructions for data reduction. Many methodologies present procedures for deriving criteria from both large and small data sets. Very small data sets may be supplemented through the use of QSARs for selected pesticides, and through the use of models such as ICE (for prediction of toxicity to under-tested species), and ACE (for estimation of chronic toxicity from acute data). The toxicity of mixtures is addressed by several existing methodologies. In some cases, additional AFs are applied to criteria to account for exposure to mixtures, whereas in others, concentration addition models are used to assess compliance. Multiple stressors and bioaccumulation are also addressed in some methodologies, by providing for application of additional safety factors. Methods are also available for translating dietary exposure limits for humans or other fish-eating animals into water concentrations. Options for addressing the safety of TES exist, and rely heavily on data from surrogate species to derive criteria. Utilizing partition coefficients, criteria may be harmonized across media to ensure that levels set to protect one compartment do not result in unacceptable levels in other compartments. Several methodologies derive criteria from entire data sets through the use of statistical extrapolations; other methods utilize only the lowest (most sensitive) data point or points. Utilization of entire data sets allows derivation of confidence limits for criteria, and encourages data generation. Criteria derivation methodologies have improved over the past two decades as they have incorporated more ecological risk assessment techniques. No single existing methodology is ideal, but elements of several may be combined, and when used with newer risk assessment tools, will produce more usable and flexible criteria derivation procedures that are protective.