On the Statistical Analysis of Air Pollution Data.

Some preliminary results are presented concenrning the distribution of air pollutant concentrations in New Jersey. The data, consist of multiple time series, sampled at 15 minute intervals and collected into hourly averages, for up to ten pollutants at twenty one locations throughout the state. The period covered covers the calendar years 1970 and 1971. The data shows daily and seasonal variation, and residual noise. Standard techniques of data analysis are used, including the fitting of parametric time series models. The question of the distribution of pollutant concentration, often thought to be of the lognormal form, is also examined. It is found that a fourth-root transformation does better than logging for sulfur dioxide and carbon monoxide levels. Author