SELECTION OF STATISTICALLY APPROPRIATE
COMPLIANCE SAMPLING INTERVALS
Roy O. Ball, Principal
Clement A. Vath, Principal
Environmental Resources Management-North Central, Inc.
Palatine, Illinois 60067
INTRODUCTION
Environmental discharges are rarely constant or completely predictable. This is especially true in
industrial wastewater discharges either as direct discharges under NPDES or indirect discharges to
POTW's. There are many industries whose wastewater discharges vary over the production cycle.
Industries with batch or semi-continuous manufacturing process cycles are especially prone to such
variabilities. Furthermore, discharges generally occur over 5 days per week and usually only 1 or 2
shifts during the day. The most significant include breweries, fermentation pharmaceuticals, and
some food production processes, among others. The presence of such production cycles significantly
complicates the accurate measurement and prediction of the wastewater discharge quality and increases the difficulty of compliance monitoring.
The use of statistical procedures in the selection of appropriate sampling intervals is important
to ensure that the measured values truly represent the overall discharge characteristics. This is especially significant when a limited number of samples per sampling interval are taken.
The objective of this paper is to analyze the problems presented by batch or semi-continuous
production cycle wastewater dischargers with respect to compliance sampling and to provide a methodology for the selection of statistically appropriate compliance sampling intervals. While the methodology herein can be used for direct discharges, the paper focuses on the use for indirect dischargers
who face a user charge based upon the results of compliance sampling by the industry (self-monitoring) or a second party (such as a sanitary district).
ANALYSIS OF THE PROBLEM
As in all statistical problems, we must first determine the nature of our historical data base that
we will use to predict the reliability of monitoring. In general, our data will be of two types. The
first type is where historical data represents production levels different than the current operations.
We will refer to this as "Case 1 Prediction at Future Production Levels." The second type is where
we have production level vs. effluent quality data taken for production periods with an average
production level similar to present production levels (or production levels for which compliance monitoring is important). We will refer to this type of problem as "Case 2 Prediction at Current Production Levels." These problems are treated in general and then a specific example is provided for
Case 1.
CASE 1-FUTURE PRODUCTION LEVELS
Summary of Approach
The approach involves evaluating the user charge formula to determine the cost sensitivity of the
various effluent parameters. Once the most cost sensitive paramenter is determined, then we analyze
the relationship of the cost sensitive parameter to production for both production days and non-
production days. We develop a linear relationship between the variables and then calculate (at some
confidence level) the error in prediction of the wastewater discharge quality as a function of both
the total number of samples taken and the distribution of the samples between production and non-
production days.
853