Evaluating multimedia chemical persistence: Classification and regression tree analysis

Deborah H Bennett, Thomas E. McKone, William E. Kastenberg

Research output: Contribution to journalArticlepeer-review

9 Scopus citations


For the thousands of chemicals continuously released into the environment, it is desirable to make prospective assessments of those likely to be persistent. Widely distributed persistent chemicals are impossible to remove from the environment and remediation by natural processes may take decades, which is problematic if adverse health or ecological effects are discovered after prolonged release into the environment. A tiered approach using a classification scheme and a multimedia model for determining persistence is presented. Using specific criteria for persistence, a classification tree is developed to classify a chemical as 'persistent' or 'nonpersistent' based on the chemical properties. In this approach, the classification is derived from the results of a standardized unit world multimedia model. Thus, the classifications are more robust for multimedia pollutants than classifications using a single medium half-life. The method can be readily implemented and provides insight without requiring extensive and often unavailable data. This method can be used to classify chemicals when only a few properties are known and can be used to direct further data collection. Case studies are presented to demonstrate the advantages of the approach.

Original languageEnglish (US)
Pages (from-to)810-819
Number of pages10
JournalEnvironmental Toxicology and Chemistry
Issue number4 I
StatePublished - 2000
Externally publishedYes


  • Characteristic time
  • Classification tree
  • Decay rates
  • Multimedia
  • Persistence

ASJC Scopus subject areas

  • Environmental Science(all)
  • Environmental Chemistry
  • Toxicology
  • Health, Toxicology and Mutagenesis


Dive into the research topics of 'Evaluating multimedia chemical persistence: Classification and regression tree analysis'. Together they form a unique fingerprint.

Cite this