ARE OZONE EXCEEDANCE RATES DECREASING? COMMENT ON "EXTREME VALUE ANALYSIS OF ENVIRONwlENTAL TIME-SERIES: AN APPLICATION TO TREND DETECTION IN GROUND-LEVEL OZONE" BY

R.L. Smith (1989) in his Statistical Science discussion paper, proposed new methods fora.Ilalyzingextreme values based on the point process view of highlevelexceedances,andillustratedthem with a detailed analysis·of ozone data frqm Houston, Texas. The methods are powerful and, in particular, the point process ofcluster peaks over a high threshold provides a remarkable condensation of the massive data set that he analyzes. It involves little loss of relevant information and permits fairly simple analyses. Smith's conclusion is that there is no trend in the overall levels of the series, but that there is a marked downward trend in the extreme values. It seems hard to find physical explanations for this, and here the evidence is reassessed in terms of a comparison between competing models for the intensity of a Poisson process. This suggests that there is some evidence for a decreasing trend in exceedance-rates.but that it is rather weak, If there .is-a trend, it seems more likylyto consist of afairly abrupt change thana gradual decrea.se. Thepossibility that such a change is due to an improvement in measurement technology is discussed. The possibility of long-memory dependence is also considered and the clustering method used is discussed. Adrian E. is Associate De]i>artment of of WA 98195. This research was the Office of Naval Research under Contract no. N-00014-88-k-0265. The author is to Richard L. Smith for nrovidine the ozone data on which the are to Peter for very to Michael A. Newton comments. The COn1lputier programs may be obtained from the author