A statistical technique for analysis of environmental data containing periodic variance components

Abstract

A statistical technique for analyzing data containing periodic variance components when the observations are made at irregular intervals is described. The technique uses a less constrained version of harmonic or periodic regression than that usually employed. The main feature of this method is that a known period is hypothesized, and its component is removed from the data if it is found to be significant. This is in contrast to searching for the presence of unknown periodic components following a classic Fourier analysis. For environmental monitoring programs, biological rhythms may create special problems in estimating average abundances present at one or more stations sampled at different times. Correct interpretations of this type of sampling data become even more difficult when time series data is obtained at irregular intervals, and the objective is trend monitoring for environmental disturbances. The steps in the analysis are described in detail, and the new method of analysis is applied to the abundance of a marine flatfish captured at the intake of a coastal power plant.