Environmental Engineering Reference
In-Depth Information
tions were made once, multiple times, or continuously. Concentrations derived with
50% confidence were not protective for continuous or multiple pesticide applica-
tions; a safety factor was therefore required to arrive at no-effect concentrations.
The 5th percentile SSD cutoff, which has been validated against multispecies
NOECs in several cases, is commonly used by current methodologies. It is a level
that balances the desire to select a percentile near zero with the need to avoid the
high uncertainty that exists in the extreme tails of the distributions.
Once a percentile is chosen, a decision on desired level of certainty for the
resulting concentration must be made. The USEPA methodology (1985) does not
provide a means to determine levels of confidence in the derived criteria. All other
SSD methodologies result in a criterion derived from a specified percentile level
and a specified level of confidence. Uncertainty in an extrapolated value results
from a probability that the extrapolated value is wrong (Aldenberg and Slob 1993).
The distribution around the extrapolated value can be used to calculate lower
boundaries for the extrapolated value (Kooijman 1987; Van Straalen and Denneman
1989; Wagner and Løkke 1991; Aldenberg and Slob 1993). By evaluating this
uncertainty, it is possible to state that the true HC 5 falls above (or below) the estimated
value with a 50%, 90%, 95%, or other level of certainty. Among the calculated
confidence levels, the most statistically robust is the 50 th percentile, or median, estimate
(ANZECC and ARMCANZ 2000; EVS Environmental Consultants 1999;
Fox 1999). The variability in the tails of the distribution tends to compound, rather
than clarify, the uncertainties.
The Dutch methodology (RIVM 2001) utilizes the 50% confidence, or median,
HC 5 estimate for derivation of MPCs, but they also report a 90% two-sided confi-
dence interval. Similarly, the Dutch methodology utilizes the median estimate of the
HC 50 for derivation of the SRC ECO , but also report the 90% confidence interval. The
Australia/New Zealand guidelines (ANZECC and ARMCANZ 2000) follow the
Dutch example in using the median estimate of the HC 5 to derive the most probable
estimate of the MTC. The Danish methodology, though, uses the lower 95th percen-
tile estimate of the 5th percentile to derive criteria (although the Danish prefer to use
an AF method; Samsoe-Petersen and Pedersen 1995). The EU risk assessment TGD
utilizes the median PNEC estimate, but also considers the 95th percentile estimate
in determining whether or not an AF should be applied to the derived PNEC. The
OECD guidance (OECD 1995) allows calculation of either median or 95th percen-
tile HC 5 estimates, and leaves it to the user to choose which level to use. Figure 1
depicts the median and lower 95th percentile estimates of the 5th percentile.
Maltby et al. (2005) investigated SSDs for pesticides and determined that the
95th percentile estimate of the SSD 5th percentile derived an HC 5 that was pro-
tective of ecosystems. The median 5th percentile level was protective for a single
pesticide application, but was not protective of continuous or multiple applications.
The authors suggest using a safety factor to address this shortfall. However, in this
study, the SSDs were constructed from acute toxicity data, and it is not expected
that an HC 5 , derived from acute data, would be protective in continuous exposure
scenarios. Multiple exposures would be better addressed by consideration of a
frequency component in the criterion statement.
Search WWH ::




Custom Search