Multivariate distributions of disinfection by-products in chlorinated drinking water.

Drinking water disinfection by-product (DBP) occurrence research is important in supporting risk assessment and regulatory performance assessment. Recent DBP occurrence surveys have expanded their scope to include non-regulated priority DBPs as well as regulated DBPs. This study applies a Box-Cox transformed multivariate normal model and data augmentation methods for left-censored and missing observations to US EPA Information Collection Rule (ICR) drinking water data to describe the variability in the trihalomethane (THM4), trihaloacetic acid (THAA), dihaloacetic acid (DHAA), and dihaloacetonitrile (DHAN) DBP classes, the relationship between class-sum and the occurrence of individual DBPs within these classes. Inferences about bromine incorporation in these classes are then compared to those made by Obolensky and Singer (2005). Results reported herein show that class-based and individual DBP concentrations are strongly related to bromine substitution, and that speciation and bromine substitution patterns are consistent across DBP classes. In addition, the multiple imputation approach employed reveals that uncertainties related to missing and left-censored DBPs have important implications for understanding bromine substitution in the THAA class. These concerns should be considered through alternative approaches to DBP regulation in subsequent Stage II D/DBP assessment and revisions, where appropriate.