Statistical Applications for Environmental Analysis and Risk Assessment

Chapter 10Trends, Autocorrelation, and Temporal Dependence

10.1 Introduction and Overview

In environmental applications, the underlying data population is not always stable or stationary over time or space, as is usually assumed or required for the validity of many statistical tests or procedures. A data population is stationary when the magnitudes of the data values are randomly distributed with no predisposition in one direction or another, and the true or population mean and variance remain invariant. On the contrary, environmental data are frequently subject to sustained or cyclical change due to changing land use activities, increasing pollution levels resulting from up-gradient pollutant discharges, declining contaminant concentrations attributable to active site remediation measures or passive natural attenuation, and so on. The predictor or explanatory variable of interest is usually time, that is, we are often interested in changes in contaminant concentrations that occur over time, in which case, the trend is described as temporal. For example, increases or decreases in contaminant concentration with time constitute temporal trend. Temporal trend can further be described as sustained or directional when the trend is only in one direction, that is, continually trending upward or continually trending downward. One-directional trend is also described as monotonic trend. Continually increasing phosphorus concentrations in a water body as the type of land use in the contributing ...