429 – Statistical Analysis of Air Pollution Data
Dynamic PCA for Multiple Air Pollutants
Oleg Melnikov
Rice University
Loren Raun
Rice University
Katherine B. Ensor
Rice University
Abstract In this study of air quality in the Houston metropolitan area we apply dynamic (a.k.a. moving window or rolling) principal component analysis (DPCA) to a suitably normalized/standardized (non-local) multivariate time series of daily concentration measurements of five pollutants (O3, CO, NO2, SO2, PM2.5) from 1/1/2009 to 12/31/2011 at each of the 24 hours. The resulting dynamic components are examined in two dimensional time domain (hour by day). Diurnal and seasonal patterns are revealed underlining times when DPCA performs best and principal components (PCs) explain most variability. DPCA is shown to be superior to (static) PCA in discovery of linear relations among transformed pollutants. DPCA captures the time-dependent correlation structure of the underlying (normalized) pollutants recorded at 34 monitoring sites located in the emission traffic regions. In winter mornings the first principal component (PC1) (mainly CO and NO2) explains up to 70% of variability, while cumulatively with PC2 (mainly driven by SO2) the explained variability rises to 90%. In the afternoon, O3, a secondary pollutant with a few-hour formation lag, gains prominence in PC2. The seasonal profile