Treatment of drinking water decreases human health risks by reducing pollutants, but the required materials, chemicals, and energy emit pollutants and increase health risks. We explored human carcinogenic and non-carcinogenic disease tradeoffs of water treatment by comparing pollutant dose-response curves against life cycle burden using USEtox methodology. An illustrative wellhead sorbent groundwater treatment system removing hexavalent chromium or pentavalent arsenic serving 3200 people was studied. Reducing pollutant concentrations in drinking water from 20 μg L−1 to 10 μg L−1 avoided 37 potential cancer cases and 64 potential non-cancer disease cases. Human carcinogenicity embedded in treatment was 0.2–5.3 cases, and non-carcinogenic toxicity was 0.2–14.3 cases, depending on technology and degree of treatment. Embedded toxicity impacts from treating Cr(VI) using strong-base anion exchange were <10% of those from using weak base anion exchange. Acidification and neutralization contributed >90% of the toxicity impacts for treatment options requiring pH control. In scenarios where benefits exceeded burdens, tradeoffs still existed. Benefits are experienced by a local population but burdens are born externally where the materials and energy are produced, thus exporting the health risks. Even when burdens clearly exceeded benefits, cost considerations may still drive selecting a detrimental treatment level or technology.

Original languageEnglish (US)
Pages (from-to)246-254
Number of pages9
JournalWater Research
StatePublished - Jan 1 2018


  • Arsenic
  • Hexavalent chromium
  • Life cycle assessment
  • Water treatment tradeoffs

ASJC Scopus subject areas

  • Water Science and Technology
  • Ecological Modeling
  • Pollution
  • Waste Management and Disposal
  • Environmental Engineering
  • Civil and Structural Engineering


Dive into the research topics of 'Human health tradeoffs in wellhead drinking water treatment: Comparing exposure reduction to embedded life cycle risks'. Together they form a unique fingerprint.

Cite this