| The Gulf Coast region of the United States faces some of the most challenging air quality problems in the nation. The region has an unusual mix of large industrial emission sources, extensive transport of emission sources, significant biogenic emissions and a complex coastal meteorology. These sources and the meteorology interact to produce high levels of ozone, hazardous air pollutants and fine particulate matter (PM). In this study, a neural network approach has been utilized to analyze the observed PM and 8-hr ozone concentrations in the Beaumont/Port Arthur area located in the Texas Upper Gulf Coast Region. The objective of the analysis is to provide conceptual information relating PM and 8-hr ozone to many meteorological as well as precursor parameters for the area. This analysis has included thirteen parameters, namely temperature, pressure, solar radiation, relative humidity, time of day, wind speed, local wind direction, regional wind direction, mixing height, rain fall, SO2, NO, and NO2. Current results have indicated that the proposed neural network analysis can provide valuable insights in the observed PM2.5 and 8-hr ozone. The R2 values of the ozone analysis can be as high as 0.89 showing strong meteorological and precursors effects. The R2 values of the PM2.5 analysis vary with the month and can be as high as 0.70 when time-delay is not considered. The observed 0.70 R2 value can be improved to 0.80 with the consideration of time-delay, and it can be further improved to 0.90 if the day and night data are analyzed separately. |
