Multivariate analysis of a 1992 SONTOS data subset

Statistical analysis of observations made between 7 and 31 August 1992 at a rural site in Southern Ontario, Canada, is presented using principal component analysis (PCA), as well as rank and partial correlation analyses. Data for ozone, CO, SO 2 , NO, NO y (NO + NO 2 including NO x , PAN and HNO 3 ), ethane, ethylene, propane, acetylene and isoprene are employed, as well as such meteorological information as temperature, relative humidity, wind speed and direction. The analysis involved several scenarios classifying the data according to daytime, nighttime and wind direction. A normal correlation analysis suggests that isoprene might be important in the build-up of ozone in the S-W sector, whereas it is not important in the N-E wind sector. PCA show that the chemical and the meteorological variables interact differently within the different scenarios. Finally, the relationships between ozone and isoprene for both wind sectors are found to be the same when the data are subjected to a partial correlation analysis including the meteorological parameters and it is concluded that the S-W ozone-isoprene relationship might be biased by temperature.