A geospatial and binomial logistic regression model to prioritize sampling for per- and polyfluorinated alkyl substances in public water systems

Sweta Ojha, Ying Li, Nader Rezaei, Ariel Robinson, Anna Hoover, Kelly G. Pennell

Research output: Contribution to journalArticlepeer-review

Abstract

As health-based drinking water standards for per- and polyfluorinated alkyl substances (PFAS) continue to evolve, public health and environmental protection decision-makers must assess exposure risks associated with all public drinking water systems in the United States (US). Unfortunately, current knowledge regarding the presence of PFAS in environmental systems is limited. In this study, a screening approach was established to: (1) identify and direct attention toward potential PFAS hot spots in drinking water sources, (2) prioritize sampling locations, and (3) provide insights regarding the potential PFAS sources that contaminate groundwater and surface water. Our approach incorporates geospatial data from public sources, including the US Environmental Protection Agency's Toxic Release Inventory, to identify locations where PFAS may be present in drinking water sources. An indicator factor (also known as “risk factor”) was developed as a function of distance between potential past and/or present PFAS users (e.g., military bases, industrial sites, and airports) and the public water system, which generates a heat map that visualizes potential exposure risks. A binomial logistic regression model indicates whether PFAS are likely to be detected in public water systems. The results obtained using the developed screening approach aligned well (with a 76% overall model accuracy) with PFAS sampling and chemical analysis data from 81 public drinking water systems in the state of Kentucky. This study proposes this screening model as an effective decision aid to assist key decision-makers in identifying and prioritizing sampling locations for potential PFAS exposure risks in the public drinking water sources in their service areas. Integr Environ Assess Manag 2022;00:1–13.

Original languageEnglish
JournalIntegrated Environmental Assessment and Management
DOIs
StateAccepted/In press - 2022

Bibliographical note

Funding Information:
We are grateful to the staff at Kentucky Department for Environmental Protection (KDEP) for engaging with our team and for sharing their expertise and perspectives throughout this project. The project described is supported by the University of Kentucky Superfund Research Program Grant P42 ES007380, UK‐CARES Grant P30 ES026529 from the National Institute of Environmental Health Sciences, grant number 1452800 from the National Science Foundation, and award number G08LM013185 from the National Library of Medicine. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of Environmental Health Sciences, National Library of Medicine, or National Science Foundation.

Publisher Copyright:
© 2022 SETAC.

Keywords

  • Binomial regression model
  • GIS mapping
  • PFAS
  • Statistics

ASJC Scopus subject areas

  • Geography, Planning and Development
  • Environmental Science (all)

Fingerprint

Dive into the research topics of 'A geospatial and binomial logistic regression model to prioritize sampling for per- and polyfluorinated alkyl substances in public water systems'. Together they form a unique fingerprint.

Cite this