Principal component and canonical correlation analysis for examining air pollution and meteorological data

Five years data on CO, NO, NO2, O-3, smoke and SO2 concentrations recorded at one air-pollution monitoring station in the city of Athens were analysed using principal component analysis (PCA). Separate analyses were undertaken for summer and winter periods. PCA was also applied to meteorological data concerning relative humidity, temperature, sunshine duration, wind velocity and wind direction. It was found that the main principal components extracted from the air pollution data were related to gasoline combustion, oil combustion and ozone interactions. The most prominent principal components from the meteorological data were related to dry conditions (summer period) and high-speed southwestern winds (for both periods). Finally, canonical correlation analysis determined relationships between the two different data sets. The main relationship was between total pollution and high humidity in combination with the low-velocity wind. (C) 1998 Elsevier Science Ltd. All rights reserved.