You are here:
Determining Representative National Surface Water Chemical Concentration Ranges for Risk Prioritization in Drinking Water
Citation:
Sayre, R., M. Serre, Woodrow Setzer, AND J. Wambaugh. Determining Representative National Surface Water Chemical Concentration Ranges for Risk Prioritization in Drinking Water. EPA Virtual Environmental Modeling Public Meeting, Virtual, Virtual, August 05, 2020. https://doi.org/10.23645/epacomptox.18822179
Impact/Purpose:
Presentation to the EPA Environmental Modeling Public Meeting August 2020. This work provides concentration ranges in surface waters to allow prioritizing chemicals on the basis of risk through estimation of removal by conventional drinking water treatment.
Description:
Thousands of chemicals can be observed in surface waters. Due to inconsistent data availability on both the hazard and exposure sides, however, it is a complex task to prioritize which of those chemicals may pose a higher relative risk to humans via drinking water. Although many measurements of organic chemical concentrations are taken across the United States, they are collected for different purposes, so many aspects of the samples vary widely. This project evaluates the influence of data curation and statistical conventions on aggregating organic surface water measurements into concentration distributions. Millions of rows of data were available in the National Water Quality Monitoring Council’s Water Quality Portal for over 1700 chemicals sampled in 2114 of 2270 hydrologic subbasins across the United States from 2008 to 2018. Over 50 of these chemicals are on the CCL4 Drinking Water Contaminant Candidate List and are candidates for future drinking water regulation. First, we provide an overview of the variability of surface water chemical concentrations within reported categories such as sampling activity type, time, location, and analytical chemistry method. We explore the effect of grouping sample sets across these categories, striving to include as many samples as possible while maintaining data quality and applicability to drinking water systems. A combination of knowledge-based judgments and statistical tests are used to make grouping decisions. Finally, we evaluate the effect of applying different parametric and nonparametric statistical methods to estimate central tendencies and associated uncertainty intervals per chemical for our chosen samples, accounting for the left-censoring and right-skew of the data. The distributions are used to calibrate a Bayesian metamodel, which creates thousands of predicted chemical distributions for human health relevant, non-monitored compounds. We compare these surface water distributions to distributions of organic chemical analytes reported by public water systems that use surface water. Comparing the distributions allows us to identify which chemicals are most likely to be removed through drinking water treatment, and the chemical features or properties associated with removal. The output of this project is a method to predict the concentration at which compounds not currently monitored by public water systems are likely to be present in drinking water, accounting for existing water treatment processes. The workflow is available as a Python script for reproducibility. The views expressed herein are those of the authors and do not necessarily represent the views or the policies of the U.S. Environmental Protection Agency.
URLs/Downloads:
DOI: Determining Representative National Surface Water Chemical Concentration Ranges for Risk Prioritization in Drinking Water
SAYRE_EMPM_05AUG2020 (2).PDF (PDF, NA pp, 1121.757 KB, about PDF)