Multivariate nonparametric trend assessment with environmental applications

Abstract

A major goal of environmental monitoring is to determine whether the environment is improving or deteriorating. Questions about the health of the environment are usually questions about trends in environmental indicators, typically a number of chemical, physical, and biological variables. Because multiple indicators are required to characterize any but the most simple environment, the problem is statistically a multivariate problem. In this work, methods for analyzing multivariate environmental trends are presented and illustrated on 17 years of approximately monthly observations on 5 water quality variables from southwestern Virginia, USA. Multivariate methods can also be applied to analyze correlated univariate data collected on a seasonal or monthly basis. A variety of methods from the literature are discussed. A unified approach is described based on a general class of correlation measures to construct a general framework for the nonparametric analysis of multivariate trends.