The U.S. Environmental Protection Agency promulgates regulations on industrial water discharges that often include numerical limitations that restrict the amounts of specific pollutants typically discharged by each industry. In developing these limitations, EPA uses a statistical framework to evaluate the data. Statistical methods are appropriate for dealing with effluent data because the quality of effluent, even in well-operated systems, is subject to a certain amount of fluctuation or uncertainty. The statistical analysis generally assumes the effluent measurements have a lognormal distribution, sometimes with left-censored values (non-detects). Non-detects may be handled using a modified delta-lognormal distribution. EPA also evaluates trends and serial correlation that may be present in the performance data. After adjusting for any trends or other factors, the statistical analysis involves estimating distributional percentiles of individual measurements and average values. These percentiles form the basis of the daily maximum limitation and the monthly or weekly average limitation on pollutant discharges. In conjunction with the statistical methods, EPA performs an engineering review to verify that the limitations are reasonable, based upon the design and expected operation of the control technologies and the facility conditions. This presentation will describe the statistical methodology and its application to performance data from several industries.
Keywords: Effluent guidelines; Lognormal; Serial correlation
Biography: David Marker is a Senior Statistician and Associate Director of Westat, where he has worked for 28 years. He has assisted EPA's Office of Water for almost that entire period. Dr. Marker is an Elected Member of the ISI, a member of TIES and IASS, and chaired the IASS 2010 Nominating Committee. He is a Fellow of the ASA and currently serves on its Board of Directors.