Probabilistic injury assessment and quantification for natural resource damage assessment using bayesian networks: A case-study of the PCB-contaminated Mississinewa River Basin.

The U.S. Federal Natural Resource Damage Assessment and Restoration (NRDAR) program gives tribes and government appointed agencies the authority to assess injury to natural resources and pursue compensatory action for resources injured or lost due to unlawful release of chemicals into the environment. This study was performed to develop and test a Bayesian network (BN) decision support tool to lend quantitative insight into natural resource injury assessment. The BN model represents the causal relationship between the released polychlorinated biphenyls (PCBs) and three common adverse effects of PCB exposure in fish-mortality, growth, and reproductive effects-as well as a combined largest effects model (CLEM) pathway. Each endpoint of a causal pathway is a probabilistic estimation of an injured or uninjured decision based on the PCB concentration in fish tissue and toxicity data. The probability distributions from the Bayesian network's CLEM pathway results were linked to spreadsheets that automate injury quantification in units of discount service acre years (DSAYS). Probabilistic injury determinations and quantifications were performed for individual spatial subregions of the study area and for the entire site. The case study focused on the fish resources of an inactive PCB-contaminated Superfund Site in mid-eastern Indiana-the Little Mississinewa River (LMR) and the larger Mississinewa River into which the LMR drains. Using the BN tool, we determined that there was at least low-level injury to fish resources throughout the Mississinewa River and Reservoir. We found that the likelihood of injury decreased with distance from the original contaminant release site. When quantified, the injury to the entire basin totaled 94,216 lost DSAYs. A secondary analysis determined higher injury to bottom feeding species of fish. This study demonstrated that BNs can be used to characterize and quantify natural resource injury for NRDAR purposes.