Automated quality assurance checks on model structure in ecological risk assessments

Publisher Website
Google Scholar
Abstract
A major criticism routinely leveled against probabilistic risk assessments of putative ecological impacts is that there has been insufficient validation of the model employed in the assessment. The only complete solution to this problem would involve concentrated scientific effort to test and validate the model, which is generally beyond the scope of the risk assessment effort. However, there are several computational checks applicable to all models used in risk analysis that can be employed to ensure the absence of several classes of structural and mathematical errors. Several kinds of profound errors which are routinely committed in practice, including violations of can all be detected using currently available software approaches. Additionally, domain knowledge from ecology can also be marshalled to check the validity of the model in its representation of biological reality. Such knowledge includes facts such as Taken together, these two classes of checks on the model constitute elementary methods of quality assurance for model structure.