Exponential Smoothing of Monitoring Data

Abstract

All monitoring data contain sampling and measurement errors. Even in a well‐managed monitoring program these errors may be quite large. Smoothing methods are presented that enhance the underlying signal and facilitate interpretation. Exponential and double exponential smoothing are compared with respect to ability to follow trends, interpolate missing values, and forecast future levels. Sewer monitoring data from a large metropolitan system are used to illustrate the methods. The methods should be applicable to other data from treatment plant and natural systems.