Variance of Environmental Water Contaminants: Building Bayesian Priors for Use in Modeling Spatial and Temporal Variability of Emerging Contaminants

• Activity Number: 84
• Type: Contributed
• Date/Time: Sunday, July 29, 2012 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Michael Messner*+ and Song Qian
• Companies: EPA and Tetra Tech
• Keywords: Bayesian ; Occurrence ; Hierarchical Modeling

Abstract

Environmental regulatory agencies frequently assess occurrence of regulated and nonregulated contaminants to support policy decisions. Occurrence models for water contaminants are typically hierarchical in structure, with lognormal distributions of location-specific parameters (means or median concentrations) and variance parameters that describe both between- and within-location variability. In a Bayesian framework, vague, noninformative priors are then used together with monitoring data to estimate these parameters. The vague priors reflect a lack of prior information (ignorance) and ensure that the estimates and inferences are driven primarily by the data. Having modeled numerous inorganic and organic contaminants in water, using vague priors for each, we need not approach emerging contaminants from such a state of ignorance. This presentation summarizes the variability observed for historical contaminants in water and develops informative priors that can be used for emerging contaminants.