This thesis presents the results of the fine particulate matter concentration analysis in two sites in Pittsburgh for one year period. The sites represent a downtown commercial area (CMU) and a suburb residential area (NETL). The thesis focuses on variations of PM 2.5 through time and tries to create a model to predict its concentrations. Time-series and meteorology methods were used to investigate association of PM 2.5 with independent variables including fine particulate matter preceding values. The results help to determine the best predictors depositing into the model.
The analysis showed seasonality in PM 2.5 distribution and possible regional allocation of its concentrations. Diurnal variations showed two patterns. The first presents peaks in nights and early mornings which occurred during relatively low 24 hour PM 2.5 levels. The highest 1 hour concentrations in afternoons were observed during the days with high 24 hour average concentrations.
Analysis of PM 2.5 relationship with meteorological variables showed the highest correlation with temperatures and wind characteristics. The other variables impact PM 2.5 variations, but the relationship is probably non-linear.
Investigating the influence of atmospheric stability on PM 2.5 , Lifted Index was used as an independent variable. The contribution of Lifted Index is significant, but its relationship with PM 2.5 appeared to be negative.
However, the most significant contributor to the model predicting current PM 2.5 level is one hour lagged PM 2.5 values. The models containing meteorological and co-pollutants variables were run to predict fine particulate matter concentrations, but the highest prediction was shown by those which contained the lagged variable.