Ecological Risk Assessment of Diffuse and Local Soil Contamination Using Species Sensitivity Distributions - Dealing with Contaminated Sites: From Theory Towards Practical Application - page 641

the cited article). In other words: the uniform methods adopted for the “reverse”
use (like averaging all input data for a species) need not be applicable to problem
solving in the “forward” use (selecting appropriate data).
In general, we think, three rules should be applied when discussing technical
issues and choosing technical approaches in SSD modeling. First, the specific prob-
lem definition of an environmental case should be the guide to appropriate choices
in the Risk Assessment with SSDs. Second, when a statistical model is used, it
should show a sufficient fit to the data. Third, when a problem occurs frequently
(such as in setting quality standards for chemicals), fixation of SSD methods may
be efficient and necessary. The latter is especially the case for legally binding SSD
output. Evidently, all efforts should focus on results that are valid for the problem
and situation.
14.5 Validity of SSD-Based Output in Ecological
Risk Assessment
SSDs are statistical models derived from laboratory toxicity data that should be
applied in combination with appropriate exposure models and extrapolations, to
obtain potentially meaningful and useful output. The big question is: how can these
statistical models predict anything relevant for ecosystems?
This practical question has triggered validation studies. Suter ( 2002 ) has identi-
fied two levels of validation of models in Ecological Risk Assessment:
1. validation by practical use;
2. validation by model-derived values resembling the relevant field phenomena.
To begin with the former, various organizations have formally adopted SSDs
as a method in Ecological Risk Assessment (for an overview see Posthuma et al.
2002b ), and especially for the derivation of soil quality standards. When soil qual-
ity standards are used, they are continuously being challenged in practice. Cases
where impacts are shown when they are unexpected and cases where standards are
unrealistically low will lead to public and policy debate, and eventually to develop-
ment and adoption of more defensible methods. Applications of the SSD-based U.S.
Ambient Water Quality Criteria, for example, have been legally challenged over the
years since they were published in 1985. So far, they have withstood those chal-
lenges, like the soil and water quality standards in Europe (which in part relate to
the use of SSDs). Nonetheless, there are ongoing debates on certain contaminants
(such as zinc) and certain system issues, such as accounting for natural background
concentrations. Such discussions result in technical and conceptual improvements
in practice.
The latter type of validation implies scientific comparison between predicted and
observed impacts. Two tracks can be recognized here: studies that focus on point
estimates (such as HC5) and impacts at such exposures and those that focus on
complete SSD curves in relation to changes in species assemblages in the field.