A geographical analysis of air pollution in the Tucson region

Abstract

This dissertation presents a geographical analysis of air pollution in the Tucson region. Image processing, geographic information system (GIS), climatological, and statistical tools are used to develop and analyze air pollution-related databases. These databases are then used in conjunction with a limited number of spatial measurements of ozone concentrations to create accurate and theoretically sound ground-level ozone maps. High spatial resolution, gridded, multi-temporal, atmospheric emissions inventories (EIs) of ozone precursor chemical (i.e. volatile organic compounds (VOCs) and nitrogen oxides (NOₓ)) emissions are initially developed. GIS-driven "top-down" and "bottom-up" methods are employed to create anthropogenic VOC and NOx emissions inventories while satellite imagery and field surveys are employed to create biogenic VOC (BVOC) emissions inventories. Accounting for approximately 50% of the anthropogenic emissions, on-road vehicles are the dominant anthropogenic source. The forest and desert lands emit nearly all of the BVOCs within the entire Tucson region while exotic trees such as eucalyptus, pine, and palm emit most of the BVOCs within the City of Tucson. Relationships between VOC and NOₓ emissions, atmospheric conditions, and ambient ozone levels are determined by examining spatio-temporal variations in ozone levels, temporal variations in VOC and NOₓ emissions and atmospheric conditions, atmospheric conditions which are conducive to elevated ozone levels. In addition, the likelihood of ozone transport from Phoenix to Tucson is assessed. The highest ozone levels occur at "rural," downwind monitors, occur in August, and occur during the early afternoon hours. Atmospheric conditions conducive to elevated concentrations differ between the months while inter-city ozone transport is most likely to occur in June. Pooled, cross-sectional, times series, regression models are developed with the aid of cluster analysis and principal components analysis to spatially predict daily maximum 1-hr and 8-hr average ozone concentrations. Gridded, multi-temporal estimates of VOCs and NOₓ emissions are the primary predictor variables in the regression models. The pooled models are reasonably accurate with overall R² values from 0.90 to 0.92, 6 to 7% error, and predicted concentrations that are typically within 0.003 to 0.004 ppm of the observed concentrations. The predicted highest ozone concentrations occur in a monitorless area on the eastern edge of the City of Tucson.