Methods for Deriving Pesticide Aquatic Life Criteria - Reviews of Environmental Contamination and Toxicology - page 33

Zabel and Cole (1999) point out that, in the case of algae, if a sensitive species were
eliminated from an ecosystem, the photosynthetic function could be quickly
replaced by another, less sensitive species. Ecosystem structure will have changed,
but function is maintained. In contrast, Daily et al. (1993) note that the disappear-
ance of a single species could lead to the unraveling of community structure as a
result of complex interactions among species. Lawton (1994) explores the impor-
tance of “ecosystem engineers” and states that loss of keystone species, whether
they are engineers or important trophic links, may cause dramatic and sudden eco-
system changes.
It would seem, then, that ecosystems may not be protected if water quality
criteria are derived by a method that does not offer protection to 100% of species.
However, there is no way to ensure such complete protection, because it is not
possible to know the entire composition of an ecosystem. Even if an ecosystem
were fully comprehended, it would not be possible to determine the sensitivity of
all component species of that ecosystem. This chapter presents and evaluates alter-
native methods for estimating ecosystem no-effect concentrations (levels not detri-
mental to aquatic organisms) by extrapolating from available toxicity data, the bulk
of which comprises single-species laboratory studies. To determine if numbers
derived from these methods are adequately protective (i.e., meet policy goals), they
must be validated in field or semi-field studies.

5.3 Probability of Over- or Underprotection

It is very useful to express criteria with confidence limits, because it provides
environmental managers with insight on how likely it is that a criterion will pro-
vide the intended level of protection. Criteria that overprotect may result in
unnecessary expenditures, whereas criteria that under protect may lead to ecosys-
tem damage. Many criteria methodologies (Canada, France, Germany, and UK)
involve compilation of data and then selection of the single most sensitive datum
(often multiplied by an extrapolation factor) to represent the criterion (CCME
1999; Lepper 2002; Zabel and Cole 1999). Criteria so derived do not have confi-
dence limits associated with them. Although the value set may be protective,
there is no way to know to what degree the value over- or underprotects. Criteria
derived by the USEPA methodology (1985) also do not have associated confi-
dence limits, despite the fact that they use a species sensitivity distribution (SSD)
methodology. Australia and New Zealand (ANZECC and ARMCANZ 2000),
The Netherlands (RIVM 2001), and OECD (1995) use SSD techniques that
derive criteria at specified confidence levels. For example, for a criterion derived
at a 50% confidence level, the true no-effect level may be either above or below
the derived criterion with equal probability. If derived at a 95% confidence level,
there is only a 5% chance that the true no-effect level rests below the derived
criterion. This kind of information can provide environmental managers with
some sense as to the reliability of criteria.