A case study: Caution when using food web models early in the ecological risk assessment process

Background/Question/Methods

We compared the results of food web models completed at the SLERA and the BERA stages of the ERA process for the same contaminated site. The SLERA employed the use of hazard indexes (HIs) that were summed across different types of chemicals. Ecological risk predicted in the SLERA involved risks to mallard (from endrin and zinc), heron (from endrin, metals, Aroclors, and zinc), mink (from metals), and shrew (from PAHs). In contrast, the food web models developed in the BERA were based on the results of site-specific ecological observations made during a field sampling reconnaissance. During this site visit, ecological receptors and habitats associated with the site were identified and a list of potential ecological receptors for inclusion in the food web models was developed. Biota tissue (i.e., crayfish, fish, and invertebrates) and abiotic media samples were collected for chemical analysis in support of the BERA and used as inputs into the food web models. Other input parameters such as dietary fraction and site foraging factor were derived for each selected receptor based on site-specific information.

Results/Conclusions

The BERA food web models were limited to chemicals that were site-related, posed ecological risk based on screening evaluations, and had significant bioaccumulation potential (i.e.,  mercury, methyl mercury, and Aroclor 1260). The BERA used the hazard quotient approach where the dose was compared to the no observed adverse and lowest observed adverse effect levels. The primary drivers of ecological risk predicted by the BERA food web models included kingfisher (from methyl and total mercury), mallard (from methyl and total mercury), and heron (from methyl and total mercury), with no predicted risk to raccoon or mink. The BERA results led to the remediation of mercury-contaminated sediments in the creek near the facility. Lessons learned from the comparison of the generic SLERA food web models versus the site specific BERA food web models reveal the need to use caution when considering the use of food web models with specific lessons as follows: (1) Food web models should rarely be used at the SLERA stage of the ERA or whenever site specific input data are lacking because such use will likely lead to inaccurate results; (2) Completing SLERA food web models with default values for chemicals with low bioaccumulation potential will likely overestimate risks; (3) Without adequate site reconnaissance, inappropriate ecological receptors may be selected.