On the statistical analysis of trend in tropospheric ozone levels.

摘要

This paper is a study of methodology to investigate trends in ozone levels in urban areas. Three methods are studied; two parametric (NHPP, and GPD) and one nonparametric.;The nonhomogeneous Poisson process (NHPP) approach: This method is based on the idea that the number of exceedances over a high threshold follows a Poisson distribution. In this method the detection of trend is approached by estimating the intensity function of the process. The intensity function is estimated using parametric and nonparametric methods. A general parametric function over time for the rate of a NHPP is proposed in order to test for nonexponential patterns.;The Generalized Pareto Distribution approach (GPD): In this method the detection of trend in ozone is approached by considering that the magnitude of the exceedances over a high threshold follows a generalized Pareto distribution.;The nonparametric statistical approach: A test for trend and a nonparametric estimator of a trend parameter were studied. The asymptotic distribution of the estimator is also provided. The nonparametric estimator is compared with the least squares estimator.;The three methods are studied empirically using a Monte Carlo method. Some insight into how well these methods perform is obtained. Also, the use of each method is illustrated by examples with ozone data.;The results of this study show that NHPP performs very well as a method to detect ozone trends. The GPD and the nonparametric approaches had low power for detecting trends in simulation experiments.