Methods for Deriving Pesticide Aquatic Life Criteria - Reviews of Environmental Contamination and Toxicology - page 95

environmental fate model, it is possible to determine if an ERL, derived for a
primary compartment, has the potential to result in exceedance of an ERL, or
human health limit, in another compartment. It is not clearly explained why this
approach is not used to harmonize water, soil, and sediment ERLs, though it
appears that this type of model could be used to harmonize aquatic life criteria
across environmental compartments.
In the German methodology, the most sensitive asset (e.g., drinking water,
aquatic life) is taken as the basis for deriving the WQO (BMU 2001). For example,
if the drinking water target for a substance is 0.1 µg/L, and the aquatic life target is
0.05 µ/L, then the aquatic life target becomes the objective.
Cross-media coherence of criteria is addressed by only the few methodologies
mentioned here. Lack of attention to this issue is probably because of gaps in knowledge
and paucity of data for development of models to describe cross-media processes.
Benson et al. (2003) noted that models have been successfully used for assessing
possible conflicts between water and sediment criteria for some compounds, but fully
integrated quantitative multimedia models are not available for making full interme-
dia assessments. Although it may not be possible to derive fully integrated criteria, it
is important to use available models to determine if excursions above water quality
criteria might adversely affect other environmental compartments.

7.3.5 Utilization of Available Data and Encouragement of Data Generation

Many methodologies make very poor use of available data, because they use only
the lowest values (Lepper 2002; CCME 1999) or focus on the lowest few values in
a data set (USEPA 1985; Roux et al. 1996). The SSD methodologies, utilized in the
Netherlands (RIVM 2001) and Australia/New Zealand (ANZECC and ARMCANZ
2000), make full use of data. Among their good use of data is the fact that they
utilize variability information to derive confidence limits for criteria. In particular,
the Australia/New Zealand curve-fitting method reduces the need to remove out-
liers or truncate data sets that show a degree of multimodality.
A recurring theme in this chapter is that ecotoxicity data are generally too scarce.
Often, insufficient ecotoxicology data are available to derive adequate criteria
(where “adequate” means there will be high certainty that criteria will neither over- nor
under-protect aquatic ecosystems). Therefore, it would be beneficial if a criteria deri-
vation methodology was designed to encourage data generation by all stakeholders.
Okkerman et al. (1991) found an example in which HC 5 values, based on data for
five species, were lower than those based on nine species. Such examples exist
because the uncertainty in the SSD method decreases with increasing sample size
(which results in lower standard deviations and extrapolation factors).
In contrast, for the USEPA method (1985) which uses only the four values nearest
the 5th percentile (the lowest four values in many cases) to calculate the FAV, additional
data may have different effects on the FAV, depending upon whether the new data fall
within the group of four nearest the 5th percentile. This is illustrated in a report prepared
for the California State Water Resources Control Board by the Great Lakes