Evaluation of temporal variations in ambient air quality at Jahra using multivariate techniques

• Multivariate statistical techniques identify important pollution sources.
• Principal component analysis (PCA) discriminates sources of variations in air quality data.
• PCA reduced the number of quality parameters that need to be monitored by 40%.
• Contributions of anthropogenic pollution sources prevailed during winter seasons.

This study demonstrates the effectiveness of multivariate statistical methods in recognizing temporal trends and interdependency of air pollutants from large and complex air quality datasets. Eight years of ambient air quality data for the city of Jahra, Kuwait, were evaluated using various multivariate statistical techniques in order to enhance the understanding of the temporal variations in the dataset. The data are a record of 5-minute measurements of nine air quality variables (sulfur dioxide, SO2, non-methane hydrocarbon, NM-HC, methane, CH4, total nitrogen oxides, NOx; as nitric oxide, NO and nitrogen dioxide, NO2, carbon monoxide, CO, Ozone, O3, particulate matter, PM10 and carbon dioxide, CO2) and four meteorological parameters (wind speed, wind direction, ambient temperature and solar intensity). Exploratory analyses (scatter plots and box plots) and multivariate statistical analyses (principal component analysis, PCA, and correlation analysis, CA) techniques were used to assess and discriminate sources of variations in the dataset. The box plots showed a high variability in the CH4, NM-HC and O3 concentrations. It also showed that O3, PM10, NO, SO2 and CO have significant seasonal patterns. CA analysis revealed significant positive correlations (p<0.01p<0.01) between O3 and temperature and between PM10 and temperature. CA, however, also showed significant inverse correlations (p<0.01p<0.01) between CO2 and temperature, and between NO and temperature. PCA allowed the identification of two different sets of 4 factors that explain 79.4% and 76.5% of the total variations in the winter and summer datasets, respectively. Furthermore, PCA resulted in a 40% reduction in the number of quality parameters. Additionally, it showed that the contributions of anthropogenic sources of air pollution (traffic, power plants and water desalination plants) prevail particularly during the winter. The obtained results are especially valuable for local authorities in planning analytical protocols and in designing effective air pollution control measures.