Moving averages for environmental standards

The use of moving averages is gaining acceptance in environ mental standards and regulations. Indexes based on moving averages are useful because they summarize a fixed number of the immediately preceding observations. If calendar-based monthly averages, for example, are used rather than moving averages of the last 30 days, data from industrial sources would not have a uniform time base. Only four observations would be averaged from a source inspected on the 4th of the month, while 28 would be averaged for a source inspected on the 28th. Many problems occur in implementing a standard based on moving averages: (1) Most environmental data are autocorrelated because they are averages of successive measurements. Autocor relation changes the variance of the statistical distribution and, therefore, the probability of detecting a violation. (2) When continuous monitors are installed, a suitable aver aging time must be determined. (3) Time-averaging affects the variance of the statistical dis tribution. This effect must be incorporated into simulation studies. (4) The relationships between exceedances (probable viola tions), transient changes, and long-term trends should be specified and used as inputs in simulation models. In en vironmental monitoring, a statistical decision about whether a change is permanent may result in substantial penalties. Cumulative sum (CUSUM) charts are a practical approach to establishing standards and determining compliance for auto regressive data. An example of an air-quality monitor illus trates the procedure.